Initial revision

src/.cvsignore

@@ -0,0 +1,2 @@
+bin
+gcccheck

src/Config

@@ -0,0 +1,450 @@
+#
+# Config for Etherboot/32
+#
+# Do not delete the tag OptionDescription and /OptionDescription
+# It is used to automatically generate the documentation.
+#
+# @OptionDescription@
+#	User interaction options:
+#
+#	-DASK_BOOT=n
+#			Ask "Boot from (N)etwork ... or (Q)uit? " 
+#			at startup, timeout after n seconds (0 = no timeout).
+#			If unset or negative, don't ask and boot immediately
+#			using the default.
+#	-DBOOT_FIRST
+#	-DBOOT_SECOND
+#	-DBOOT_THIRD
+#			On timeout or Return key from previous
+#			question, selects the order to try to boot from
+#			various devices.
+#			(alternatives: BOOT_NIC, BOOT_DISK,
+#			 BOOT_FLOPPY, BOOT_NOTHING)
+#			See etherboot.h for prompt and answer strings.
+#			BOOT_DISK and BOOT_FLOPPY work only where a driver
+#			exists, e.g. in LinuxBIOS.
+#			They have no effect on PCBIOS.
+#	-DBOOT_INDEX	The device to boot from 0 == any device.
+#			1 == The first nic found.
+#			2 == The second nic found
+#			...
+#			BOOT_INDEX only applies to the BOOT_FIRST.  BOOT_SECOND 
+#			and BOOT_THIRD search through all of the boot devices.
+#	-DBAR_PROGRESS
+#			Use rotating bar instead of sequential dots
+#			to indicate an IP packet transmitted.
+#
+#	Boot order options:
+#
+#	-DBOOT_CLASS_FIRST
+#	-DBOOT_CLASS_SECOND
+#	-DBOOT_CLASS_THIRD
+#			Select the priority of the boot classes
+#			Valid values are:
+#				BOOT_NIC
+#				BOOT_DISK
+#				BOOT_FLOPPY
+#	BOOT_DISK and BOOT_FLOPPY work only where a driver exists,
+#	e.g. in LinuxBIOS.  They have no effect on PCBIOS.
+#
+#	Boot autoconfiguration protocol options:
+#
+#	-DALTERNATE_DHCP_PORTS_1067_1068
+#			Use ports 1067 and 1068 for DHCP instead of 67 and 68.
+#			As these ports are non-standard, you need to configure
+#			your DHCP server to use them. This option gets around
+#			existing DHCP servers which cannot be touched, for
+#			one reason or another, at the cost of non-standard
+#			boot images.
+#	-DNO_DHCP_SUPPORT
+#			Use BOOTP instead of DHCP.
+#	-DRARP_NOT_BOOTP
+#			Use RARP instead of BOOTP/DHCP.
+#	-DREQUIRE_VCI_ETHERBOOT
+#			Require an encapsulated Vendor Class Identifier
+#			of "Etherboot" in the DHCP reply
+#			Requires DHCP support.
+#	-DDHCP_CLIENT_ID=\"Identifier\"
+#	-DDHCP_CLIENT_ID_LEN=<Client ID length in octets>
+#	-DDHCP_CLIENT_ID_TYPE=<Client ID type>
+#			Specify a RFC2132 Client Identifier option, length and type.
+#			Requires DHCP support.
+#	-DDHCP_USER_CLASS=\"UserClass\"
+#	-DDHCP_USER_CLASS_LEN=<User Class length in octets>
+# 			Specify a RFC3004 User Class option and length. Use this
+#			option to set a UC (or multiple UCs) rather than munge the
+#			client Vendor Class ID.
+#			Requires DHCP support.
+#	-DALLOW_ONLY_ENCAPSULATED
+#			Ignore Etherboot-specific options that are not within
+#			the Etherboot encapsulated options field.  This option
+#			should be enabled unless you have a legacy DHCP server
+#			configuration from the bad old days before the use of
+#			encapsulated Etherboot options.
+#	-DDEFAULT_BOOTFILE=\"default_bootfile_name\"
+#			Define a default bootfile for the case where your DHCP
+#			server does not provide the information.  Example:
+#			  -DDEFAULT_BOOTFILE="tftp:///tftpboot/kernel"
+#			If you do not specify this option, then DHCP offers that
+#			do not specify bootfiles will be ignored.
+#
+#	NIC tuning parameters:
+#
+#	-DALLMULTI
+#			Turns on multicast reception in the NICs.
+#
+#	Boot tuning parameters:
+#
+#	-DCONGESTED
+#			Turns on packet retransmission.	 Use it on a
+#			congested network, where the normal operation
+#			can't boot the image.
+#	-DBACKOFF_LIMIT
+#			Sets the maximum RFC951 backoff exponent to n.
+#			Do not set this unreasonably low, because on networks
+#			with many machines they can saturate the link
+#			(the delay corresponding to the exponent is a random
+#			time in the range 0..3.5*2^n seconds).	Use 5 for a
+#			VERY small network (max. 2 minutes delay), 7 for a
+#			medium sized network (max. 7.5 minutes delay) or 10
+#			for a really huge network with many clients, frequent
+#			congestions (max. 1  hour delay).  On average the
+#			delay time will be half the maximum value.  If in
+#			doubt about the consequences, use a larger value.
+#			Also keep in mind that the number of retransmissions
+#			is not changed by this setting, so the default of 20
+#			may no longer be appropriate.  You might need to set
+#			MAX_ARP_RETRIES, MAX_BOOTP_RETRIES, MAX_TFTP_RETRIES
+#			and MAX_RPC_RETRIES to a larger value.
+#	-DTIMEOUT=n
+#			Use with care!! See above.
+#			Sets the base of RFC2131 sleep interval to n.
+#			This can be used with -DBACKOFF_LIMIT=0 to get a small
+#			and constant (predictable) retry interval for embedded
+#			devices. This is to achieve short boot delays if both
+#			the DHCP Server and the embedded device will be powered
+#			on the same time. Otherwise if the DHCP server is ready
+#			the client could sleep the next exponentially timeout,
+#			e.g. 70 seconds or more. This is not what you want.
+#			n should be a multiple of TICKS_PER_SEC (18).
+#
+#	Boot device options:
+#
+#	-DTRY_FLOPPY_FIRST
+#			If > 0, tries that many times to read the boot
+#			sector from a floppy drive before booting from
+#			ROM. If successful, does a local boot.
+#			It assumes the floppy is bootable.
+#	-DEXIT_IF_NO_OFFER
+#			If no IP offer is obtained, exit and
+#			let the BIOS continue.
+#			The accessibility of the TFTP server has no effect,
+#			so configure your DHCP/BOOTP server properly.
+#			You should probably reduce MAX_BOOTP_RETRIES
+#			to a small number like 3.
+#
+#	Boot image options:
+#
+#	-DTAGGED_IMAGE
+#			Add tagged image kernel boot support (recommended).
+#	-DAOUT_IMAGE
+#			Add a.out kernel boot support (generic).
+#	-DELF_IMAGE
+#			Add generic ELF kernel boot support (recommended).
+#	-DEL64F_IMAGE
+#			Add generic ELF64 kernel boot support (useful for > 4GB disks).
+#	-DWINCE_IMAGE
+#			Add the ability to boot WINCE.... now only sis630 OK!
+#	-DPXE_IMAGE
+#			Add the ability to boot PXE NBPs.  Requires
+#			PXE_EXPORT.  Currently not supported on
+#			anything other than i386
+#	-DFREEBSD_PXEEMU
+#			Add the ability to boot PXE images... only FreeBSD supported
+#	-DIMAGE_MULTIBOOT
+#			Add Multiboot image support (currently only
+#			for ELF images).
+#			Without this, generic ELF support is selected.
+#	-DIMAGE_FREEBSD
+#			Add FreeBSD image loading support (requires at least
+#			-DAOUT_IMAGE and/or -DELF_IMAGE).
+#	-DFREEBSD_KERNEL_ENV
+#			Pass in FreeBSD kernel environment
+#	-DAOUT_LYNX_KDI
+#			Add Lynx a.out KDI support
+#	-DMULTICAST_LEVEL1
+#			Support for sending multicast packets
+#	-DMULTICAST_LEVEL2
+#			Support for receiving multicast packets
+#	-DDNS_RESOLVER
+#			Support for resolving hostnames in bootfile name (experimental)
+#	-DDOWNLOAD_PROTO_TFTP
+#			If defined, includes TFTP support (recommended).
+#	-DDOWNLOAD_PROTO_NFS
+#			If defined, includes NFS support.
+#	-DDEFAULT_PROTO_NFS
+#			If defined, makes NFS the default protocol instead
+#			of TFTP. Requires DOWNLOAD_PROTO_NFS.
+#	-DDOWNLOAD_PROTO_SLAM
+#			If defined, includes Scalable Local Area Multicast
+#			support.
+#	-DDOWNLOAD_PROTO_TFTM
+#			If defined, includes TFTP Multicast mode support.
+#	-DDOWNLOAD_PROTO_HTTP
+#			If defined, includes HTTP support.
+#
+#	Console options:
+#
+#	-DCONSOLE_FIRMWARE
+#			Set for firmware/BIOS provided (default if nothing else is set).
+#			Normally this is shows up on your CRT.
+#	-DCONSOLE_SERIAL
+#			Set for serial console.
+#	-DCONSOLE_DUAL
+#			Both of the above
+#	-DCONSOLE_DIRECT_VGA
+#			Set for direct VGA console (only for x86).
+#	-DCOMCONSOLE
+#			Set port, e.g. 0x3F8.
+#	-DCONSPEED
+#			Set speed, e.g. 57600.
+#	-DCOMPARM
+#			Set Line Control Register value for data bits, stop
+#			bits and parity. See a National Semiconditor 8250/
+#			16450/16550 data sheet for bit meanings.
+#			If undefined, defaults to 0x03 = 8N1.
+#	-DCOMPRESERVE
+#			Ignore COMSPEED and COMPARAM and instead preserve
+#			the com port parameters from the previous user
+#			of the com port.  Examples of previous user are a BIOS
+#			that implements console redirection, lilo and LinuxBIOS.
+#			This makes it trivial to keep the serial port
+#			speed setting in sync between multiple users.
+#			You set the speed in the first user and the
+#			rest follow along.
+#
+#	Interface export options:
+#
+#	-DPXE_EXPORT
+#			Export a PXE API interface.  This is work in
+#			progress.  Note that you won't be able to load
+#			PXE NBPs unless you also use -DPXE_IMAGE.
+#	-DPXE_STRICT
+#			Strict(er) compliance with the PXE
+#			specification as published by Intel.  This may
+#                       or may not be a good thing depending on your
+#			view of the spec...
+#	-DPXE_DHCP_STRICT
+#			Strict compliance of the DHCP request packets
+#			with the PXE specification as published by
+#			Intel.  This may or may not be a good thing
+#			depending on your view of whether requesting
+#			vendor options which don't actually exist is
+#			pointless or not. You probably want this
+#			option if you intend to use Windows RIS or
+#			similar.
+#
+#	Obscure options you probably don't need to touch:
+#
+#	-DPOWERSAVE
+#			Halt the processor when waiting for keyboard input
+#			which saves power while waiting for user interaction.
+#			Good for compute clusters and VMware emulation.
+#			But may not work for all CPUs.
+#	-DBUILD_SERIAL
+#			Include an auto-incrementing build number in
+#			the Etherboot welcome message.  Useful when
+#			developing, to be sure that the file you
+#			compiled is the one you're currently testing.
+#	-DBUILD_ID
+#			Include a build ID string in the Etherboot
+#			welcome message.  Useful when developing, if
+#			you have multiple builds with different
+#			configurations and want to check you're
+#			running the one you think you are.  Requires
+#			-DBUILD_SERIAL.
+#
+#	BUS options:
+#	
+#	-DCONFIG_PCI
+#			Include support for devices using the pci bus.
+#	-DCONFIG_ISA
+#			Include support for devices using isa bus.
+#	-DCONFIG_PCMCIA
+#			Include support for PCMCIA in general *development*
+# @/OptionDescription@
+
+# These default settings compile Etherboot with a small number of options.
+# You may wish to enable more of the features if the size of your ROM allows.
+
+
+# Select which buses etherboot should support
+CFLAGS+=	-DCONFIG_PCI -DCONFIG_ISA
+# CFLAGS+=	-DCONFIG_PCMCIA
+
+# For prompting and default on timeout
+CFLAGS+=	-DASK_BOOT=3 -DBOOT_FIRST=BOOT_NIC
+# If you would like to attempt to boot from other devices as well as the network.
+# CFLAGS+=	-DBOOT_SECOND=BOOT_FLOPPY
+# CFLAGS+=	-DBOOT_THIRD=BOOT_DISK
+# CFLAGS+=	-DBOOT_INDEX=0
+
+# If you prefer the old style rotating bar progress display
+# CFLAGS+=	-DBAR_PROGRESS
+
+# Show size indicator
+# CFLAGS+=	-DSIZEINDICATOR
+
+# Enabling this creates non-standard images which use ports 1067 and 1068
+# for DHCP/BOOTP
+# CFLAGS+=	-DALTERNATE_DHCP_PORTS_1067_1068
+
+# Enabling this makes the boot ROM require a Vendor Class Identifier
+# of "Etherboot" in the Vendor Encapsulated Options
+# This can be used to reject replies from servers other than the one
+# we want to give out addresses to us, but it will prevent Etherboot
+# from getting an IP lease until you have configured DHCPD correctly
+# CFLAGS+=	-DREQUIRE_VCI_ETHERBOOT
+
+# EXPERIMENTAL! Set DHCP_CLIENT_ID to create a Client Identifier (DHCP
+# option 61, see RFC2132 section 9.14) when Etherboot sends the DHCP
+# DISCOVER and REQUEST packets.  This ID must UNIQUELY identify each
+# client on your local network.  Set DHCP_CLIENT_ID_TYPE to the
+# appropriate hardware type as described in RFC2132 / RFC1700; this
+# almost certainly means using '1' if the Client ID is an Ethernet MAC
+# address and '0' otherwise. Set DHCP_CLIENT_ID_LEN to the length of
+# the Client ID in octets (this is not a null terminated C string, do
+# NOT add 1 for a terminator and do NOT add an extra 1 for the
+# hardware type octet).  Note that to identify your client using the
+# normal default MAC address of your NIC, you do NOT need to set this
+# option, as the MAC address is automatically used in the
+# hwtype/chaddr field; note also that this field only sets the DHCP
+# option: it does NOT change the MAC address used by the client.
+
+# CFLAGS+=	-DDHCP_CLIENT_ID="'C','L','I','E','N','T','0','0','1'" \
+#		-DDHCP_CLIENT_ID_LEN=9 -DDHCP_CLIENT_ID_TYPE=0
+
+# CFLAGS+=	-DDHCP_CLIENT_ID="0xDE,0xAD,0xBE,0xEF,0xDE,0xAD" \
+#		-DDHCP_CLIENT_ID_LEN=6 -DDHCP_CLIENT_ID_TYPE=1
+
+# EXPERIMENTAL! Set DHCP_USER_CLASS to create a User Class option (see
+# RFC3004) when Etherboot sends the DHCP DISCOVER and REQUEST packets.
+# This can be used for classification of clients, typically so that a
+# DHCP server can send an appropriately tailored reply.  Normally, a
+# string identifies a class of to which this client instance belongs
+# which is useful in your network, such as a department ('FINANCE' or
+# 'MARKETING') or hardware type ('THINCLIENT' or 'KIOSK').  Set
+# DHCP_USER_CLASS_LEN to the length of DHCP_USER_CLASS in octets.
+# This is NOT a null terminated C string, do NOT add 1 for a
+# terminator.  RFC3004 advises how to lay out multiple User Class
+# options by using an octet for the length of each string, as in this
+# example.  It is, of course, up to the server to parse this.
+
+# CFLAGS+=	-DDHCP_USER_CLASS="'T','E','S','T','C','L','A','S','S'"	\
+#		-DDHCP_USER_CLASS_LEN=9
+
+# CFLAGS+=	-DDHCP_USER_CLASS="5,'A','L','P','H','A',4,'B','E','T','A'" \
+#		-DDHCP_USER_CLASS_LEN=11
+
+# for btext console support
+# CFLAGS+=	-DCONSOLE_BTEXT 
+# for direct PC kbd support 
+# CFLAGS+=	-DCONSOLE_PC_KBD
+# Set to enable FILO support
+# for FILO support it will make main call pci_init
+# INCLUDE_FILO=y
+ifdef	INCLUDE_FILO
+CFLAGS+=	-DCONFIG_FILO
+endif
+
+# Enabling this causes Etherboot to ignore Etherboot-specific options
+# that are not within an Etherboot encapsulated options field.
+# This option should be enabled unless you have a legacy DHCP server
+# configuration from the bad old days before the use of
+# encapsulated Etherboot options.
+CFLAGS+=	-DALLOW_ONLY_ENCAPSULATED
+
+# Disable DHCP support
+# CFLAGS+=	-DNO_DHCP_SUPPORT
+
+# Specify a default bootfile to be used if the DHCP server does not
+# provide the information.  If you do not specify this option, then
+# DHCP offers that do not contain bootfiles will be ignored.
+# CFLAGS+=	-DDEFAULT_BOOTFILE=\"tftp:///tftpboot/kernel\"
+
+# Limit the delay on packet loss/congestion to a more bearable value. See
+# description above.  If unset, do not limit the delay between resend.
+CFLAGS+=	-DBACKOFF_LIMIT=7 -DCONGESTED
+
+# More optional features
+# CFLAGS+=	-DTRY_FLOPPY_FIRST=4
+# CFLAGS+=	-DEXIT_IF_NO_OFFER
+
+# For a serial console, which can run in parallel with FIRMWARE console
+# CFLAGS+=	-DCONSOLE_DUAL -DCOMCONSOLE=0x3F8 -DCONSPEED=9600
+
+# Enable tagged image, generic ELF, Multiboot ELF
+# or FreeBSD ELF/a.out boot image support
+CFLAGS+=	-DTAGGED_IMAGE -DELF_IMAGE
+# CFLAGS+=	-DAOUT_IMAGE -DIMAGE_MULTIBOOT -DIMAGE_FREEBSD
+# CFLAGS+=	-DAOUT_IMAGE -DAOUT_LYNX_KDI
+# CFLAGS+=	-DCOFF_IMAGE 
+# CFLAGS+=	-DRAW_IMAGE
+
+# Download files via TFTP
+CFLAGS+=	-DDOWNLOAD_PROTO_TFTP
+# Change download protocol to NFS, default is TFTP
+# CFLAGS+=	-DDOWNLOAD_PROTO_NFS
+# Change download protocol to HTTP, default is TFTP
+# CFLAGS+=	-DDOWNLOAD_PROTO_HTTP
+# Change default protocol to NFS
+# CFLAGS+=	-DDEFAULT_PROTO_NFS
+# Support to resolve hostnames in boot filename
+# CFLAGS+=	-DDNS_RESOLVER
+
+# Multicast Support
+# CFLAGS+=	-DALLMULTI -DMULTICAST_LEVEL1 -DMULTICAST_LEVEL2 -DDOWNLOAD_PROTO_TFTM
+
+# Etherboot as a PXE network protocol ROM
+CFLAGS+=	-DPXE_IMAGE -DPXE_EXPORT
+# Etherboot stricter as a PXE network protocol ROM
+# CFLAGS+=	-DPXE_DHCP_STRICT
+
+# Support for PXE emulation. Works only with FreeBSD to load the kernel
+# via pxeboot, use only with DOWNLOAD_PROTO_NFS
+# CFLAGS+=	-DFREEBSD_PXEEMU
+
+# Include an auto-incrementing build serial number and optional build
+# ID string
+# CFLAGS+=	-DBUILD_SERIAL
+# CFLAGS+=	-DBUILD_SERIAL -DBUILD_ID=\"testing\"
+
+# Do not relocate
+# core/relocate.c should really be moved to an arch specific directory
+# but this is here for archs that don't support relocation
+# CFLAGS+=	-DNORELOCATE
+
+# you should normally not need to change these
+HOST_CC=	gcc
+CPP=		gcc -E -Wp,-Wall
+RM=		rm -f
+TOUCH=		touch
+PERL=		/usr/bin/perl
+CC=		gcc
+AS=		as
+LD=		ld
+SIZE=		size
+AR=		ar
+RANLIB=		ranlib
+OBJCOPY=	objcopy
+
+CFLAGS+=	-Os -ffreestanding
+CFLAGS+=	-Wall -W -Wno-format
+CFLAGS+=	$(EXTRA_CFLAGS)
+ASFLAGS+=	$(EXTRA_ASFLAGS)
+LDFLAGS+=	$(EXTRA_LDFLAGS)
+# For debugging
+# LDFLAGS+=	-Map $@.map
+
+# Location to place generated binaries, and files
+BIN=bin

src/Families

@@ -0,0 +1,133 @@
+# This is the config file for creating Makefile rules for Etherboot ROMs
+#
+# To make a ROM for a supported NIC locate the appropriate family
+# and add a line of the form
+#
+# ROM		PCI-IDs		Comment
+#
+# ROM is the desired output name for both .rom and .lzrom images.
+# PCI IDs are the PCI vendor and device IDs of the PCI NIC
+# For ISA NICs put -
+#
+# All PCI ROMs that share a single driver are only built once (because they
+# only have different PCI-IDs, but identical code).  ISA ROMS are built for
+# each ROM type, because different vendors used a different logic around the
+# basic chip.  The most popular example is the NS8390, which some cards use
+# in PIO mode, some in DMA mode.  Two chips currently don't fit into this nice
+# black-and-white scheme (the Lance and the NS8390).  Their driver deals
+# with both PCI and ISA cards.  These drivers will be treated similarly to
+# ISA only drivers by genrules.pl and are compiled for each ROM type that is
+# ISA, and additionally compiled for the PCI card type.
+#
+# Then do: make clean, make Roms and make
+#
+# Please send additions to this file to <kenUNDERSCOREyap AT users PERIOD sourceforge PERIOD net>
+
+# Start of configuration
+
+family		drivers/net/skel
+
+family		arch/ia64/drivers/net/undi_nii
+undi_nii	-
+
+# 3c59x cards (Vortex) and 3c900 cards
+# If your 3c900 NIC detects but fails to work, e.g. no link light, with
+# the 3c90x driver, try using the 3c595 driver. I have one report that the
+# 3c595 driver handles these NICs properly. (The 595 driver uses the
+# programmed I/O mode of operation, whereas the 90x driver uses the bus
+# mastering mode. These NICs are capable of either mode.) When it comes to
+# making a ROM, as usual, you must choose the correct image, the one that
+# contains the same PCI IDs as your NIC.
+family		drivers/net/3c595
+
+# 3Com 3c90x cards
+family		drivers/net/3c90x
+
+# Intel Etherexpress Pro/100
+family		drivers/net/eepro100
+
+#Intel Etherexpress Pro/1000
+family		drivers/net/e1000
+
+#Broadcom Tigon 3
+family		drivers/net/tg3
+
+family		drivers/net/pcnet32
+
+# National Semiconductor ns83820 (Gigabit) family
+family		drivers/net/ns83820
+
+family		drivers/net/tulip
+
+family		drivers/net/davicom
+
+family		drivers/net/rtl8139
+
+family		drivers/net/r8169
+
+family		drivers/net/via-rhine
+
+family		drivers/net/w89c840
+
+family		drivers/net/sis900
+
+family		drivers/net/natsemi
+
+family		drivers/net/prism2_plx
+
+family		drivers/net/prism2_pci
+# Various Prism2.5 (PCI) devices that manifest themselves as Harris Semiconductor devices
+# (with the actual vendor appearing as the vendor of the first subsystem)
+hwp01170	0x1260,0x3873	ActionTec HWP01170
+dwl520		0x1260,0x3873	DLink DWL-520
+
+family		drivers/net/ns8390
+wd		-		WD8003/8013, SMC8216/8416, SMC 83c790 (EtherEZ)
+ne		-		NE1000/2000 and clones
+3c503		-		3Com503, Etherlink II[/16]
+
+family		drivers/net/epic100
+
+family		drivers/net/3c509
+3c509		-		3c509, ISA/EISA
+3c529		-		3c529 == MCA 3c509
+
+family		drivers/net/3c515
+3c515		-		3c515, Fast EtherLink ISA
+
+family		drivers/net/eepro
+eepro		-		Intel Etherexpress Pro/10
+
+family		drivers/net/cs89x0
+cs89x0		-		Crystal Semiconductor CS89x0
+
+family		drivers/net/depca
+depca		-		Digital DE100 and DE200
+
+family          drivers/net/forcedeth
+
+family		drivers/net/sk_g16
+sk_g16		-		Schneider and Koch G16
+
+family		drivers/net/smc9000
+smc9000		-		SMC9000
+
+family		drivers/net/sundance
+
+family		drivers/net/tlan
+
+family		drivers/disk/ide_disk
+ide_disk	0x0000,0x0000	Generic IDE disk support
+
+family		drivers/disk/pc_floppy
+
+family		arch/i386/drivers/net/undi
+undi		0x0000,0x0000	UNDI driver support
+
+family		drivers/net/pnic
+
+family		arch/armnommu/drivers/net/p2001_eth
+
+family		drivers/net/mtd80x
+
+family		drivers/net/dmfe

src/Makefile

@@ -0,0 +1,15 @@
+# Override ARCH here or on the command line
+# ARCH=i386
+# Additionally you can supply extra compilation arguments, e.g. for x86_64
+# EXTRA_CFLAGS=-m32
+# EXTRA_ASFLAGS=--32
+# EXTRA_LDFLAGS=-m elf_i386
+ifndef ARCH
+ARCH:=$(shell uname -m | sed -e s,i[3456789]86,i386,)
+endif
+MAKEDEPS:=
+SUFFIXES:=
+
+include Config
+include arch/$(ARCH)/Config
+include Makefile.main

src/Makefile-armnommu

@@ -0,0 +1,18 @@
+ARCH:=armnommu
+MAKEDEPS:=
+
+include Config
+include arch/$(ARCH)/Config
+
+CC=		$(CROSS_COMPILE)gcc
+AS=		$(CROSS_COMPILE)as
+LD=		$(CROSS_COMPILE)ld
+SIZE=		$(CROSS_COMPILE)size
+AR=		$(CROSS_COMPILE)ar
+RANLIB=		$(CROSS_COMPILE)ranlib
+OBJCOPY=	$(CROSS_COMPILE)objcopy
+
+MAKEDEPS+=Makefile-armnommu
+BIN=bin
+
+include Makefile.main

src/Makefile-e1

@@ -0,0 +1,18 @@
+ARCH:=e1
+MAKEDEPS:=
+
+include arch/$(ARCH)/Config
+include Config
+
+CC=		e1-coff-gcc
+AS=		e1-coff-as
+LD=		e1-coff-ld
+SIZE=	e1-coff-size
+AR=		e1-coff-ar
+RANLIB=	e1-coff-ranlib
+OBJCOPY=e1-coff-objcopy
+
+MAKEDEPS+=Makefile-e1
+BIN=bin-e1
+
+include Makefile.main

src/Makefile-i386

@@ -0,0 +1,18 @@
+ARCH:=i386
+MAKEDEPS:=
+
+include arch/$(ARCH)/Config
+include Config
+
+CC=		i386-linux-gcc
+AS=		i386-linux-as
+LD=		i386-linux-ld
+SIZE=		i386-linux-size
+AR=		i386-linux-ar
+RANLIB=		i386-linux-ranlib
+OBJCOPY=	i386-linux-objcopy
+
+MAKEDEPS+=Makefile-i386
+BIN=bin-i386
+
+include Makefile.main

src/Makefile-ia64

@@ -0,0 +1,18 @@
+ARCH:=ia64
+MAKEDEPS:=
+
+include arch/$(ARCH)/Config
+include Config
+
+CC=		ia64-linux-gcc
+AS=		ia64-linux-as
+LD=		ia64-linux-ld
+SIZE=		ia64-linux-size
+AR=		ia64-linux-ar
+RANLIB=		ia64-linux-ranlib
+OBJCOPY=	ia64-linux-objcopy
+
+MAKEDEPS+=Makefile-ia64
+BIN=bin-ia64
+
+include Makefile.main

src/Makefile.main

@@ -0,0 +1,450 @@
+#
+# Makefile for Etherboot
+#
+# Most of the time you should edit Config
+#
+# Common options:
+#	VERSION_*=v	- Set the major and minor version numbers
+#
+# NS8390 options:
+#	-DINCLUDE_NE	- Include NE1000/NE2000 support
+#	-DNE_SCAN=list	- Probe for NE base address using list of
+#			  comma separated hex addresses
+#	-DINCLUDE_3C503 - Include 3c503 support
+#	  -DT503_SHMEM	- Use 3c503 shared memory mode (off by default)
+#	-DINCLUDE_WD	- Include Western Digital/SMC support
+#	-DWD_DEFAULT_MEM- Default memory location for WD/SMC cards
+#	-DWD_790_PIO    - Read/write to WD/SMC 790 cards in PIO mode (default
+#			  is to use shared memory) Try this if you get "Bogus
+#			  packet, ignoring" messages, common on ISA/PCI hybrid
+#			  systems.
+#	-DCOMPEX_RL2000_FIX
+#
+#	If you have a Compex RL2000 PCI 32-bit (11F6:1401),
+#	and the bootrom hangs in "Probing...[NE*000/PCI]",
+#	try enabling this fix... it worked for me :).
+#	In the first packet write somehow it somehow doesn't
+#	get back the expected data so it is stuck in a loop.
+#	I didn't bother to investigate what or why because it works
+#	when I interrupt the loop if it takes more then COMPEX_RL2000_TRIES.
+#	The code will notify if it does a abort.
+#	SomniOne - somnione@gmx.net
+#
+# 3C509 option:
+#	-DINCLUDE_3C509	- Include 3c509 support
+#
+# 3C90X options:
+#	-DINCLUDE_3C90X	- Include 3c90x support
+#
+#	Warning Warning Warning
+#	If you use any of the XCVR options below, please do not complain about
+#	the behaviour with Linux drivers to the kernel developers. You are
+#	on your own if you do this. Please read 3c90x.txt to understand
+#	what they do. If you don't understand them, ask for help on the
+#	Etherboot mailing list. And please document what you did to the NIC
+#	on the NIC so that people after you won't get nasty surprises.
+#
+#	-DCFG_3C90X_PRESERVE_XCVR - Reset the transceiver type to the value it
+#			  had initially just before the loaded code is started.
+#	-DCFG_3C90X_XCVR - Hardcode the tranceiver type Etherboot uses.
+#	-DCFG_3C90X_BOOTROM_FIX - If you have a 3c905B with buggy ROM
+#			  interface, setting this option might "fix" it.  Use
+#			  with caution and read the docs in 3c90x.txt!
+#
+#	See the documentation file 3c90x.txt for more details.
+#
+# CS89X0 (optional) options:
+#	-DINCLUDE_CS89X0- Include CS89x0 support
+#	-DCS_SCAN=list	- Probe for CS89x0 base address using list of
+#			  comma separated hex addresses; increasing the
+#			  address by one (0x300 -> 0x301) will force a
+#			  more aggressive probing algorithm. This might
+#			  be neccessary after a soft-reset of the NIC.
+#
+# LANCE options:
+#	-DINCLUDE_NE2100- Include NE2100 support
+#	-DINCLUDE_NI6510- Include NI6510 support
+#
+# SK_G16 options:
+#	-DINCLUDE_SK_G16- Include SK_G16 support
+#
+# I82586 options:
+#	-DINCLUDE_3C507	- Include 3c507 support
+#	-DINCLUDE_NI5210- Include NI5210 support
+#	-DINCLUDE_EXOS205-Include EXOS205 support
+#
+# SMC9000 options:
+#       -DINCLUDE_SMC9000   - Include SMC9000 driver
+#       -DSMC9000_SCAN=list - List of I/O addresses to probe
+#
+# TIARA (Fujitsu Etherstar) options:
+#	-DINCLUDE_TIARA	- Include Tiara support
+#
+# NI5010 options:
+#	-DINCLUDE_NI5010 - Include NI5010 support
+#
+# TULIP options:
+#	-DINCLUDE_TULIP	- Include Tulip support
+#
+# RTL8139 options:
+#	-DINCLUDE_RTL8139 - Include RTL8139 support
+#
+# SIS900 options:
+#	-DINCLUDE_SIS900 - Include SIS900 support
+#
+# NATSEMI options:
+#	-DINCLUDE_NATSEMI - Include NATSEMI support
+#
+
+SRCS:=
+BOBJS:=
+
+MAKEROM=	$(PERL) ./util/makerom.pl
+VERSION_MAJOR=	5
+VERSION_MINOR=	3
+VERSION_PATCH=	14
+EXTRAVERSION=	
+VERSION=	$(VERSION_MAJOR).$(VERSION_MINOR).$(VERSION_PATCH)$(EXTRAVERSION)
+MM_VERSION=	$(VERSION_MAJOR).$(VERSION_MINOR)
+CFLAGS+=	-DVERSION_MAJOR=$(VERSION_MAJOR) \
+		-DVERSION_MINOR=$(VERSION_MINOR) \
+		-DVERSION=\"$(VERSION)\" $(OLDGAS) \
+		-I include -I arch/$(ARCH)/include \
+		-DARCH=$(ARCH)
+FILO=filo
+FILO_PROGRAM_NAME = FILO
+FILO_PROGRAM_VERSION = 0.4.1
+FILO_BUILD_INFO = ($(shell whoami)@$(shell hostname)) $(shell LANG=C date)
+
+GCCINCDIR = $(shell $(CC) -print-search-dirs | head -n 1 | cut -d' ' -f2)include
+CPPFLAGS = -nostdinc -imacros filo/config.h 
+#-Ifilo/include -I$(GCCINCDIR) -MD
+ASFLAGS_X = -D__ASSEMBLY__
+
+IDENT=		'$(@F) $(VERSION) (GPL) etherboot.org'
+
+# Find out if we're using binutils 2.9.1 which uses a different syntax in some
+# places (most prominently in the opcode prefix area).
+OLDGAS:=	$(shell $(AS) --version | grep -q '2\.9\.1' && echo -DGAS291)
+
+BUILD_LIBS=	$(BLIB)
+BUILD_IMGS=	$(IMGS)
+
+3C503FLAGS=	-DINCLUDE_3C503 # -DT503_SHMEM
+# Note that the suffix to MAKEROM_ is the (mixed case) basename of the ROM file
+MAKEROM_3c503=	-3
+3C507FLAGS=	-DINCLUDE_3C507
+3C509FLAGS=	-DINCLUDE_3C509
+3C529FLAGS=	-DINCLUDE_3C529
+3C595FLAGS=	-DINCLUDE_3C595
+3C90XFLAGS=	-DINCLUDE_3C90X
+CS89X0FLAGS=	-DINCLUDE_CS89X0
+EEPROFLAGS=	-DINCLUDE_EEPRO
+EEPRO100FLAGS=	-DINCLUDE_EEPRO100
+E1000FLAGS=	-DINCLUDE_E1000
+EPIC100FLAGS=	-DINCLUDE_EPIC100
+EXOS205FLAGS=	-DINCLUDE_EXOS205
+LANCEFLAGS=	-DINCLUDE_LANCE		# Lance/PCI!
+NE2100FLAGS=	-DINCLUDE_NE2100
+NEFLAGS=	-DINCLUDE_NE -DNE_SCAN=0x300,0x280,0x320,0x340,0x380
+NS8390FLAGS=	-DINCLUDE_NS8390	# NE2000/PCI!
+NI5010FLAGS=	-DINCLUDE_NI5010
+NI5210FLAGS=	-DINCLUDE_NI5210
+NI6510FLAGS=	-DINCLUDE_NI6510
+RTL8139FLAGS=	-DINCLUDE_RTL8139
+SK_G16FLAGS=	-DINCLUDE_SK_G16
+SIS900FLAGS=   	-DINCLUDE_SIS900
+NATSEMIFLAGS=  	-DINCLUDE_NATSEMI
+SMC9000FLAGS=   -DINCLUDE_SMC9000
+SUNDANCEFLAGS=	-DINCLUDE_SUNDANCE
+TLANFLAGS=	-DINCLUDE_TLAN
+TIARAFLAGS=	-DINCLUDE_TIARA
+DEPCAFLAGS=	-DINCLUDE_DEPCA	# -DDEPCA_MODEL=DEPCA -DDEPCA_RAM_BASE=0xd0000
+TULIPFLAGS=	-DINCLUDE_TULIP
+OTULIPFLAGS=	-DINCLUDE_OTULIP
+VIA_RHINEFLAGS=	-DINCLUDE_VIA_RHINE
+WDFLAGS=	-DINCLUDE_WD -DWD_DEFAULT_MEM=0xCC000
+W89C840FLAGS=	-DINCLUDE_W89C840
+
+SRCS+=	core/serial.c
+
+SRCS+=	core/btext.c core/pc_kbd.c
+
+SRCS+=	core/main.c core/pci.c core/osloader.c core/nfs.c
+SRCS+=	core/misc.c core/config.c core/isa_probe.c core/pci_probe.c
+SRCS+=	core/relocate.c core/heap.c
+SRCS+=	drivers/disk/floppy.c core/nic.c core/disk.c core/timer.c
+SRCS+=	core/proto_eth_slow.c
+SRCS+=	core/proto_slam.c core/proto_tftm.c core/proto_http.c
+SRCS+=	core/isapnp.c
+SRCS+=	core/pcmcia.c core/i82365.c
+SRCS+=	core/pxe_export.c core/dns_resolver.c
+
+FILO_SRCS+=	$(FILO)/drivers/ide_x.c  
+FILO_SRCS+=	$(FILO)/fs/blockdev.c $(FILO)/fs/eltorito.c $(FILO)/fs/fsys_ext2fs.c $(FILO)/fs/fsys_fat.c $(FILO)/fs/fsys_iso9660.c
+FILO_SRCS+=	$(FILO)/fs/fsys_reiserfs.c $(FILO)/fs/vfs.c $(FILO)/fs/fsys_jfs.c $(FILO)/fs/fsys_minix.c $(FILO)/fs/fsys_xfs.c  
+FILO_SRCS+=	$(FILO)/main/elfload.c $(FILO)/main/elfnote.c $(FILO)/main/filo_x.c $(FILO)/main/lib.c $(FILO)/main/linuxbios_x.c 
+FILO_SRCS+=	$(FILO)/main/pci_x.c $(FILO)/main/malloc_x.c $(FILO)/main/printf_x.c $(FILO)/main/console_x.c 
+FILO_SRCS+=	$(FILO)/$(ARCH)/context.c $(FILO)/$(ARCH)/linux_load.c $(FILO)/$(ARCH)/segment.c $(FILO)/$(ARCH)/sys_info.c 
+FILO_SRCS+=	$(FILO)/$(ARCH)/switch.S $(FILO)/usb/debug_x.c $(FILO)/usb/scsi_cmds.c $(FILO)/usb/uhci.c $(FILO)/usb/usb.c 
+FILO_SRCS+=	$(FILO)/usb/ohci.c $(FILO)/usb/usb_scsi_low.c  $(FILO)/usb/usb_x.c
+
+
+BOBJS+=		$(BIN)/main.o $(BIN)/osloader.o $(BIN)/nfs.o $(BIN)/misc.o
+BOBJS+=		$(BIN)/proto_slam.o $(BIN)/proto_tftm.o $(BIN)/proto_http.o
+BOBJS+=		$(BIN)/floppy.o
+BOBJS+=		$(BIN)/serial.o $(BIN)/timer.o  $(BIN)/relocate.o $(BIN)/heap.o
+BOBJS+=		$(BIN)/btext.o $(BIN)/pc_kbd.o
+BOBJS+=		$(BIN)/nic.o $(BIN)/disk.o
+BOBJS+=		$(BIN)/isapnp.o
+BOBJS+=		$(BIN)/pci.o $(BIN)/isa_probe.o $(BIN)/pci_probe.o
+BOBJS+=		$(BIN)/vsprintf.o $(BIN)/string.o
+BOBJS+=		$(BIN)/pcmcia.o $(BIN)/i82365.o
+BOBJS+=		$(BIN)/pxe_export.o $(BIN)/dns_resolver.o
+
+FILO_OBJS+=		$(BIN)/ide_x.o $(BIN)/pci_x.o
+FILO_OBJS+=		$(BIN)/blockdev.o $(BIN)/eltorito.o $(BIN)/fsys_ext2fs.o $(BIN)/fsys_fat.o $(BIN)/fsys_iso9660.o $(BIN)/fsys_reiserfs.o $(BIN)/vfs.o
+FILO_OBJS+=		$(BIN)/fsys_jfs.o $(BIN)/fsys_minix.o $(BIN)/fsys_xfs.o  
+FILO_OBJS+=		$(BIN)/elfload.o  $(BIN)/elfnote.o  $(BIN)/filo_x.o $(BIN)/lib.o $(BIN)/linuxbios_x.o $(BIN)/malloc_x.o $(BIN)/printf_x.o $(BIN)/console_x.o   
+FILO_OBJS+=		$(BIN)/context.o  $(BIN)/linux_load.o  $(BIN)/segment.o  $(BIN)/sys_info.o $(BIN)/switch.o
+FILO_OBJS+=		$(BIN)/debug_x.o  $(BIN)/scsi_cmds.o $(BIN)/uhci.o $(BIN)/usb.o $(BIN)/ohci.o $(BIN)/usb_scsi_low.o $(BIN)/usb_x.o
+
+BLIB=		$(BIN)/bootlib.a 
+FILOLIB=	$(BIN)/filolib.a
+LIBS=		$(BLIB)
+ifdef	INCLUDE_FILO
+LIBS+=		$(FILOLIB)
+endif
+UTILS+=		$(BIN)/nrv2b
+STDDEPS=	$(START) $(UTILS)
+# MAKEDEPS is the one target that is depended by all ROMs, so we check gcc here
+# If you are confident that gcc 2.96 works for you, you can remove the lines
+# that check gcc in the toolcheck rule
+MAKEDEPS+=	Makefile Makefile.main Config genrules.pl Families
+MAKEDEPS+=	$(BIN)/toolcheck
+MAKEDEPS+=	arch/$(ARCH)/Makefile arch/$(ARCH)/Config
+
+# Start of targets
+
+.PHONY:	noargs
+noargs:	$(BIN)/toolcheck
+	@echo '===================================================='
+	@echo 'No target specified. To specify a target, do: '
+	@echo
+	@echo '    $(MAKE) bin/<rom-name>.<output-format> '
+	@echo
+	@echo 'where <output-format> is one of {zdsk, zrom, iso, liso, zlilo, zpxe, elf, com}'
+	@echo
+	@echo 'or: '
+	@echo
+	@echo '    $(MAKE) all<output-format>s'
+	@echo
+	@echo 'to generate all possible images of format <output-format>'
+	@echo
+	@echo 'For example, '
+	@echo
+	@echo '    make allzroms '
+	@echo
+	@echo 'will generate all possible .zrom (rom burnable) images, and'
+	@echo
+	@echo '    make allzdsks'
+	@echo
+	@echo 'will generate all possible .zdsk (bootable floppy) images, or'
+	@echo
+	@echo '===================================================='
+	@exit 1
+
+$(BIN)/toolcheck:	Makefile Config
+	@if $(CC) -v 2>&1 | grep -is 'gcc version 2\.96' > /dev/null; \
+	then \
+		echo 'gcc 2.96 is unsuitable for compiling Etherboot'; \
+		echo 'Use gcc 2.95 or gcc 3.x instead'; \
+		exit 1; \
+	else \
+		touch $(BIN)/toolcheck; \
+	fi; \
+	if [ `perl -e 'use bytes; print chr(255)' | wc -c` = 2 ]; \
+	then \
+		echo 'Your Perl version has a Unicode handling bug'; \
+		echo 'To workaround, execute this before compiling Etherboot:'; \
+		echo 'export LANG=$${LANG%.UTF-8}'; \
+		exit 1; \
+	fi
+
+include		arch/$(ARCH)/Makefile
+
+# Common files
+
+$(BLIB):	$(BOBJS)
+	$(AR) r $@ $(BOBJS)
+	$(RANLIB) $@
+
+$(FILOLIB):	$(FILO_OBJS)
+	$(AR) r $@ $(FILO_OBJS)
+	$(RANLIB) $@
+
+# LinuxBIOS support code
+$(BIN)/linuxbios.o:	firmware/linuxbios/linuxbios.c include/etherboot.h include/dev.h firmware/linuxbios/linuxbios_tables.h
+
+# Do not add driver specific dependencies here unless it's something the
+# genrules.pl script *can't* deal with, i.e. if it is not C code.
+
+$(FILO)/config.h: $(FILO)/Config
+	/bin/echo -e '/* GENERATED FILE, DO NOT EDIT */\n' >$@
+	sed -e 's/#.*//' -e '/=/!d' -e 's/\([^[:space:]]*\)[[:space:]]*=[[:space:]]*\(.*\).*/#define \1 \2/' -e 's/^#define \([^ ]*\) 0$$/#undef \1/' $^ >>$@
+
+filo_version: $(FILO)/main/version.h
+
+$(FILO)/main/version.h: FORCE
+	echo '#define PROGRAM_NAME "$(FILO_PROGRAM_NAME)"' > $@
+	echo '#define PROGRAM_VERSION "$(FILO_PROGRAM_VERSION) $(FILO_BUILD_INFO)"' >> $@
+
+FORCE:
+
+
+# Utilities
+
+$(BIN)/lzhuf:	util/lzhuf.c
+	$(HOST_CC) -O2 -DENCODE -DDECODE -DMAIN -DVERBOSE -o $@ $<
+
+# Roms file
+# Builds almost silently because this rule is triggered for just about
+# every modification to sources.
+
+$(BIN)/Roms $(BIN)/NIC:	genrules.pl Families $(SRCS)
+	@mkdir -p $(@D)
+	@echo Scanning for ROMs and dependencies...
+	@$(PERL) ./genrules.pl Families $(BIN)/NIC $(ARCH) $(SRCS) > $(BIN)/Roms
+
+# Pattern Rules
+
+# general rules for compiling/assembling source files
+$(BIN)/%.o:	core/%.c $(MAKEDEPS)
+	$(CC) $(CFLAGS) -o $@ -c $<
+
+$(BIN)/%.s:	core/%.c $(MAKEDEPS)
+	$(CC) $(CFLAGS) -S -o $@ -c $<
+
+$(BIN)/%.o:	drivers/disk/%.c $(MAKEDEPS)
+	$(CC) $(CFLAGS) -o $@ -c $<
+
+$(BIN)/%.o:	drivers/net/%.c $(MAKEDEPS)
+	$(CC) $(CFLAGS) -o $@ -c $<
+
+$(BIN)/%.o:	firmware/linuxbios/%.c $(MAKEDEPS)
+	$(CC) $(CFLAGS) -o $@ -c $<
+
+$(BIN)/%.o:     $(FILO)/drivers/%.c $(MAKEDEPS) $(FILO)/config.h
+	$(CC) $(CFLAGS) -imacros $(FILO)/config.h -o $@ -c $<
+
+$(BIN)/%.o:     $(FILO)/fs/%.c $(MAKEDEPS) $(FILO)/config.h
+	$(CC) $(CFLAGS) -imacros $(FILO)/config.h -o $@ -c $<
+
+$(BIN)/%.o:     $(FILO)/$(ARCH)/%.c $(MAKEDEPS) $(FILO)/config.h
+	$(CC) $(CFLAGS) -imacros $(FILO)/config.h -o $@ -c $<
+
+$(BIN)/%.o:     $(FILO)/$(ARCH)/%.S $(MAKEDEPS) $(FILO)/config.h
+	$(CC) $(ASFLAGS_X) $(CPPFLAGS) -c $< -o $@
+
+$(BIN)/%.o:     $(FILO)/main/%.c $(MAKEDEPS) $(FILO)/config.h filo_version
+	$(CC) $(CFLAGS) -imacros $(FILO)/config.h -o $@ -c $<
+
+$(BIN)/%.o:     $(FILO)/usb/%.c $(MAKEDEPS) $(FILO)/config.h
+	$(CC) $(CFLAGS) -imacros $(FILO)/config.h -o $@ -c $<
+
+# Rule for the super etherboot image.
+$(BIN)/etherboot.o: $(DOBJS)
+	$(LD) $(LDFLAGS) -r $(DOBJS) -o $@
+
+$(BIN)/etherboot-pci.o: $(PCIOBJS)
+	$(LD) $(LDFLAGS) -r $(PCIOBJS) -o $@
+
+# General rules for generating runtime (rt) files
+$(BIN)/%.rt.o:  $(BIN)/%.o $(START) $(BIN)/config.o $(LIBS) $(STDDEPS) $(MAKEDEPS)
+	$(LD) $(LDFLAGS) -r $(START) $(BIN)/config.o $< $(LIBS) -o $@
+
+#   Rule for $(BIN)/%.FORMAT.rt is architecture and target-format specific
+
+$(BIN)/%.rt.bin: $(BIN)/%.rt $(MAKEDEPS)
+	$(OBJCOPY) -O binary -R .prefix $< $@
+
+$(BIN)/%.rt1.bin: $(BIN)/%.rt $(MAKEDEPS)
+	$(OBJCOPY) -O binary -j .text.nocompress $< $@
+
+$(BIN)/%.rt2.bin: $(BIN)/%.rt $(MAKEDEPS)
+	$(OBJCOPY) -O binary -R .prefix -R .text.nocompress $< $@
+
+# Rules for generating prefix binary files
+
+#   Rule for $(BIN)/%.FORMAT.prf is architecture and target-format specific
+$(BIN)/%.prf.bin: $(BIN)/%.prf $(MAKEDEPS)
+	$(OBJCOPY) -j .prefix -O binary $< $@
+
+# general rule for .z (compressed binary code), may be overridden
+$(BIN)/%.zbin: $(BIN)/%.bin $(BIN)/nrv2b $(MAKEDEPS)
+	$(BIN)/nrv2b e $< $@
+
+# Housekeeping
+
+clean:
+	$(RM) $(BIN)/*
+	$(RM) $(FILO)/config.h $(FILO)/main/version.h
+
+../index.html:	../index.xhtml
+	(cd ..; m4 -P -DHOSTSITE=SOURCEFORGE index.xhtml > index.html)
+
+../index-berlios.html:	../index.xhtml
+	(cd ..; m4 -P -DHOSTSITE=BERLIOS index.xhtml > index-berlios.html)
+
+tarball: ../index.html ../index-berlios.html
+	(echo -n $(VERSION) ''; date -u +'%Y-%m-%d') > ../VERSION
+	(cd ../..; tar cf /tmp/etherboot-$(VERSION).tar --exclude CVS --exclude doc etherboot-$(VERSION))
+	bzip2 -9 < /tmp/etherboot-$(VERSION).tar > /tmp/etherboot-$(VERSION).tar.bz2
+	gzip -9 < /tmp/etherboot-$(VERSION).tar > /tmp/etherboot-$(VERSION).tar.gz
+
+# Auto-incrementing build serial number.  Is auto-incremented for each
+# make run that specifies a final image file (e.g. bin/undi.zpxe) as a
+# target, or a target of the form "all*".  Enable via -DBUILD_SERIAL
+# in Config.
+
+ifneq ($(findstring -DBUILD_SERIAL,$(CFLAGS)),)
+
+# If no make goals are specified, it means "make all"
+REALGOALS = $(if $(MAKECMDGOALS),$(MAKECMDGOALS),all)
+
+# Filter to see if there are any targets to trigger an auto-increment
+BUILDGOALS = $(filter all,$(REALGOALS)) $(filter all%,$(REALGOALS)) \
+	     $(foreach SUFFIX,$(SUFFIXES),$(filter %.$(SUFFIX),$(REALGOALS)))
+
+ifneq ($(strip $(BUILDGOALS)),)
+# This is an auto-incrementing build.  Forcibly rebuild .buildserial.h
+# and mark config.o as depending on it to force its rebuilding.
+bin/config.o : include/.buildserial.h
+.PHONY : include/.buildserial.h
+endif # BUILDGOALS
+
+include/.buildserial.h :
+	@if [ ! -f $@ ]; then 	echo '#define BUILD_SERIAL_NUM 0' > $@; fi
+	@perl -pi -e 's/(BUILD_SERIAL_NUM)\s+(\d+)/"$${1} ".($${2}+1)/e' $@
+
+buildserial : include/.buildserial.h
+	@perl -n -e '/BUILD_SERIAL_NUM\s+(\d+)/ && ' \
+		-e 'print "Build serial number is $$1\n";' $<
+
+else # -DBUILD_SERIAL
+
+buildserial : 
+	@echo Build serial number is disabled. Enable -DBUILD_SERIAL in Config.
+
+endif # -DBUILD_SERIAL
+
+bs : buildserial
+
+version:
+	@echo $(VERSION)
+
+romlimit:
+	@echo $(ROMLIMIT)
+
+sources:
+	@echo $(SRCS)

src/README.pixify

@@ -0,0 +1,90 @@
+This file documents the driver changes needed to support use as part
+of a PXE stack.
+
+PROPER WAY
+==========
+
+1. The probe() routine.
+
+There are three additional fields that need to be filled in the nic
+structure: ioaddr, irqno and irq.
+
+  ioaddr is the base I/O address and seems to be for information only;
+  no use will be made of this value other than displaying it on the
+  screen.
+
+  irqno must be the IRQ number for the NIC.  For PCI NICs this can
+  simply be copied from pci->irq.
+
+  irq is a function pointer, like poll and transmit.  It must point to
+  the driver's irq() function.
+
+2. The poll() routine.
+
+This must take an additional parameter: "int retrieve".  Calling
+poll() with retrieve!=0 should function exactly as before.  Calling
+poll() with retrieve==0 indicates that poll() should check for the
+presence of a packet to read, but must *not* read the packet.  The
+packet will be read by a subsequent call to poll() with retrieve!=0.
+
+The easiest way to implement this is to insert the line
+  if ( ! retrieve ) return 1;
+between the "is there a packet ready" and the "fetch packet" parts of
+the existing poll() routine.
+
+Care must be taken that a call to poll() with retrieve==0 does not
+clear the NIC's "packet ready" status indicator, otherwise the
+subsequent call to poll() with retrieve!=0 will fail because it will
+think that there is no packet to read.
+
+poll() should also acknowledge and clear the NIC's "packet received"
+interrupt.  It does not need to worry about enabling/disabling
+interrupts; this is taken care of by calls to the driver's irq()
+routine.
+
+Etherboot will forcibly regenerate an interrupt if a packet remains
+pending after all interrupts have been acknowledged.  You can
+therefore get away with having poll() just acknolwedge and clear all
+NIC interrupts, without particularly worrying about exactly when this
+should be done.
+
+3. The irq() routine.
+
+This is a new routine, with prototype
+  void DRIVER_irq ( struct nic *nic, irq_action_t action );
+"action" takes one of three possible values: ENABLE, DISABLE or FORCE.
+ENABLE and DISABLE mean to enable/disable the NIC's "packet received"
+interrupt.  FORCE means that the NIC should be forced to generate a
+fake "packet received" interrupt.
+
+If you are unable to implement FORCE, your NIC will not work when
+being driven via the UNDI interface under heavy network traffic
+conditions.  Since Etherboot's UNDI driver (make bin/undi.zpxe) is the
+only program known to use this interface, it probably doesn't really
+matter.
+
+
+QUICK AND DIRTY WAY
+===================
+
+It is possible to use the system timer interrupt (IRQ 0) rather than a
+genuine NIC interrupt.  Since there is a constant stream of timer
+interrupts, the net upshot is a whole load of spurious "NIC"
+interrupts that have no effect other than to cause unnecessary PXE API
+calls.  It's inefficient but it works.
+
+To achieve this, simply set nic->irqno=0 in probe() and point nic->irq
+to a dummy routine that does nothing.  Add the line
+  if ( ! retrieve ) return 1;
+at the beginning of poll(), to prevent the packet being read (and
+discarded) when poll() is called with retrieve==0;
+
+
+UNCONVERTED DRIVERS
+===================
+
+Drivers that have not yet been converted should continue to function
+when not used as part of a PXE stack, although there will be a
+harmless compile-time warning about assignment from an incompatible
+pointer type in the probe() function, since the prototype for the
+poll() function is missing the "int retrieve" parameter.

src/arch/armnommu/Config

@@ -0,0 +1,23 @@
+# Config for armnommu Etherboot
+#
+
+# For a clean compilation, switch in global Config
+# off: -DCONFIG_PCI, -DTAGGED_IMAGE, -DELF_IMAGE, -DPXE*, -DRELOCATE and INCLUDE_FILO
+# on : -DRAW_IMAGE
+
+# Serial line settings
+CFLAGS+= -DCONSOLE_SERIAL -DCONSPEED=57600
+
+# System Frequency
+CFLAGS+= -DSYSCLK=73728000
+
+# Image Download Address
+CFLAGS+= -DRAWADDR=0x40100000
+
+# NIC Debug Outputs
+#CFLAGS+= -DDEBUG_NIC
+
+# Fixed MAC address
+# p2001_eth has no flash and fixed mac address
+#CFLAGS+=	-DMAC_HW_ADDR_DRV="'H','Y','L','N','X','1'"
+CFLAGS+=	-DMAC_HW_ADDR_DRV="0x00,0x09,0x4F,0x00,0x00,0x02"

src/arch/armnommu/Makefile

@@ -0,0 +1,58 @@
+ARCH_FORMAT=	armnommu
+
+ROMLIMIT=	20480
+CHECKSIZE=	{ read d1; read d1 d2 d3 size d4; [ $$size -gt $(ROMLIMIT) ] &&\
+	{ $(RM) $@; echo "ERROR: code size exceeds limit!"; exit 1; }; exit 0; }
+
+START=	$(BIN)/start.o
+
+SRCS+=	arch/armnommu/core/arm_timer.c
+SRCS+=	arch/armnommu/core/start.S
+SRCS+=	arch/armnommu/core/serial.c
+SRCS+=	arch/armnommu/core/mem.c
+SRCS+=	arch/armnommu/core/setjmp.S
+SRCS+=	arch/armnommu/drivers/net/p2001_eth.c
+
+# not greater than 100kB
+ROMLIMIT:=1024000
+
+include		$(BIN)/Roms
+
+ROMS= $(BIN)/p2001_eth.rom
+IMGS= $(BIN)/p2001_eth.img
+
+
+allfiles:	$(ROMS)
+
+BOBJS+=		$(BIN)/arm_timer.o
+BOBJS+=		$(BIN)/serial.o
+BOBJS+=		$(BIN)/mem.o
+BOBJS+=		$(BIN)/setjmp.o
+BOBJS+=		$(BIN)/lib1funcs.o
+
+# Utilities
+
+$(BIN)/nrv2b:	util/nrv2b.c
+	$(HOST_CC) -O2 -DENCODE -DDECODE -DMAIN -DVERBOSE -DNDEBUG -DBITSIZE=32 -DENDIAN=0 -o $@ $<
+
+# Pattern Rules
+# General for compiling/assembly source files
+$(BIN)/%.o:	arch/armnommu/core/%.c $(MAKEDEPS)
+	$(CC) $(CFLAGS) -o $@ -c $<
+
+$(BIN)/%.o:	arch/armnommu/drivers/net/%.c $(MAKEDEPS)
+	$(CC) $(CFLAGS) -o $@ -c $<
+
+$(BIN)/%.S:	arch/armnommu/core/%.c $(MAKEDEPS)
+	$(CC) $(CFLAGS) -S -o $@ -c $<
+
+$(BIN)/%.o:	arch/armnommu/core/%.S $(MAKEDEPS)
+	$(CPP) $(CFLAGS) -D ASSEMBLY $< | $(AS) $(ASFLAGS) -o $@
+
+# general ruls for generating .img files
+$(BIN)/%.tmp: $(BIN)/%.o $(START) $(BIN)/config.o arch/$(ARCH)/core/etherboot.lds $(LIBS) $(STDDEPS) $(MAKEDEPS)
+	$(LD) $(LDFLAGS) -T arch/$(ARCH)/core/etherboot.lds -o $@ $(START) $(BIN)/config.o $< $(LIBS)
+	@$(SIZE) $@ | $(CHECKSIZE)
+
+$(BIN)/%.img: $(BIN)/%.tmp $(MAKEDEPS)
+	$(OBJCOPY) -O binary $< $@

src/arch/armnommu/core/arm_timer.c

@@ -0,0 +1,77 @@
+/*
+ *  Copyright (C) 2004 Tobias Lorenz
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include "etherboot.h"
+#include "timer.h"
+#include "latch.h"
+#include "hardware.h"
+
+
+/* get timer returns the contents of the timer */
+static unsigned long get_timer(void)
+{
+	return P2001_TIMER->Freerun_Timer;
+}
+
+/* ------ Calibrate the TSC ------- 
+ * Time how long it takes to excute a loop that runs in known time.
+ * And find the convertion needed to get to CLOCK_TICK_RATE
+ */
+
+static unsigned long configure_timer(void)
+{
+	return (1);
+}
+
+static unsigned long clocks_per_tick = 1;
+
+void setup_timers(void)
+{
+	if (!clocks_per_tick) {
+		clocks_per_tick = configure_timer();
+	}
+}
+
+unsigned long currticks(void)
+{
+	return get_timer(); /* /clocks_per_tick */
+}
+
+static unsigned long timer_timeout;
+static int __timer_running(void)
+{
+	return get_timer() < timer_timeout;
+}
+
+void udelay(unsigned int usecs)
+{
+	unsigned long now;
+	now = get_timer();
+	timer_timeout = now + usecs * ((clocks_per_tick * TICKS_PER_SEC)/(1000*1000));
+	while(__timer_running());
+}
+void ndelay(unsigned int nsecs)
+{
+	unsigned long now;
+	now = get_timer();
+	timer_timeout = now + nsecs * ((clocks_per_tick * TICKS_PER_SEC)/(1000*1000*1000));
+	while(__timer_running());
+}
+
+void load_timer2(unsigned int timer2_ticks)
+{
+	unsigned long now;
+	unsigned long clocks;
+	now = get_timer();
+	clocks = timer2_ticks * ((clocks_per_tick * TICKS_PER_SEC)/CLOCK_TICK_RATE);
+	timer_timeout = now + clocks;
+}
+
+int timer2_running(void)
+{
+	return __timer_running();
+}

src/arch/armnommu/core/etherboot.lds

@@ -0,0 +1,55 @@
+/*
+ *  Copyright (C) 2004 Tobias Lorenz
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+OUTPUT_FORMAT("elf32-littlearm", "elf32-littlearm", "elf32-littlearm")
+OUTPUT_ARCH(arm)
+ENTRY(_start)
+SECTIONS
+{
+	/*. = 0x00000000;*/	/* PIC */
+	/*. = 0x00000400;*/	/* ROM Bootloader */
+	. = 0x40000000;		/* SDRAM */
+
+	. = ALIGN(4);
+	_text = . ;
+	.text      :
+	{
+	  _start = .;
+	  _virt_start = .;
+	  bin/start.o	(.text)
+	  *(.text)
+
+	  . = ALIGN(16);
+	  isa_drivers = . ;
+	  *(.drivers.isa);
+	  isa_drivers_end = . ;	  
+	}
+
+	. = ALIGN(4);
+	.rodata : { *(.rodata) }
+
+	. = ALIGN(4);
+	.data : { *(.data) }
+
+	. = ALIGN(4);
+	.got : { *(.got) }
+
+	. = ALIGN(4);
+	_bss = . ;
+	.bss : { *(.bss) }
+
+	. = ALIGN(4);
+	_ebss = .;
+	_end = .;
+
+	. = ALIGN(16);
+	.text :
+	{
+		*(.dma.desc);
+		*(.dma.buffer);
+	}
+}

src/arch/armnommu/core/lib1funcs.c

@@ -0,0 +1,105 @@
+/*
+ *  Copyright (C) 2004 Tobias Lorenz
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+
+/*
+ * from gcc/config/udivmodsi4.c
+ */
+unsigned long
+udivmodsi4(unsigned long num, unsigned long den, int modwanted)
+{
+  unsigned long bit = 1;
+  unsigned long res = 0;
+
+  while (den < num && bit && !(den & (1L<<31)))
+    {
+      den <<=1;
+      bit <<=1;
+    }
+  while (bit)
+    {
+      if (num >= den)
+	{
+	  num -= den;
+	  res |= bit;
+	}
+      bit >>=1;
+      den >>=1;
+    }
+  if (modwanted) return num;
+  return res;
+}
+
+
+/*
+ * from gcc/config/udivmod.c
+ */
+long
+__udivsi3 (long a, long b)
+{
+  return udivmodsi4 (a, b, 0);
+}
+
+long
+__umodsi3 (long a, long b)
+{
+  return udivmodsi4 (a, b, 1);
+}
+
+
+/*
+ * from gcc/config/divmod.c
+ */
+long
+__divsi3 (long a, long b)
+{
+  int neg = 0;
+  long res;
+
+  if (a < 0)
+    {
+      a = -a;
+      neg = !neg;
+    }
+
+  if (b < 0)
+    {
+      b = -b;
+      neg = !neg;
+    }
+
+  res = udivmodsi4 (a, b, 0);
+
+  if (neg)
+    res = -res;
+
+  return res;
+}
+
+long
+__modsi3 (long a, long b)
+{
+  int neg = 0;
+  long res;
+
+  if (a < 0)
+    {
+      a = -a;
+      neg = 1;
+    }
+
+  if (b < 0)
+    b = -b;
+
+  res = udivmodsi4 (a, b, 1);
+
+  if (neg)
+    res = -res;
+
+  return res;
+}

src/arch/armnommu/core/mem.c

@@ -0,0 +1,27 @@
+/*
+ *  Copyright (C) 2004 Tobias Lorenz
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include "hooks.h"
+#include "io.h"
+#include "etherboot.h"
+
+struct meminfo meminfo;
+void get_memsizes(void)
+{
+/* We initialize the meminfo structure 
+ * according to our development board's specs
+ * We do not have a way to automatically probe the 
+ * memspace instead we initialize it manually
+ */
+	meminfo.basememsize = 0x00000000;
+	meminfo.memsize     = 0x00008000;
+	meminfo.map_count   = 1;
+
+	meminfo.map[0].addr = 0x40000000;
+	meminfo.map[0].size = 0x01000000;
+	meminfo.map[0].type = E820_RAM;
+}

src/arch/armnommu/core/raw_loader.c

@@ -0,0 +1,48 @@
+/*
+ *  Copyright (C) 2004 Tobias Lorenz
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#ifdef RAW_IMAGE
+static unsigned long raw_load_addr;
+
+int mach_boot(register unsigned long entry_point)
+{
+	void (*fnc)(void) = (void *) entry_point;
+	// r0 = 0
+	// r1 = 625 (machine nr. MACH_TYPE_P2001)
+	(*fnc)();
+
+	return 0; /* We should never reach this point ! */
+}
+
+static sector_t raw_download(unsigned char *data, unsigned int len, int eof)
+{
+	memcpy(phys_to_virt(raw_load_addr), data, len);
+	raw_load_addr += len;
+	if (!eof)
+		return 0;
+
+	done(1);
+	printf("Starting program.\n");
+	mach_boot(RAWADDR);
+	printf("Bootsector returned?");
+	longjmp(restart_etherboot, -2);
+	return 1;
+}
+
+static os_download_t raw_probe(unsigned char *data __unused, unsigned int len __unused)
+{
+	printf("(RAW");
+	// probe something here...
+	printf(")... \n");
+
+	//raw_load_addr = phys_to_virt(_end);
+	raw_load_addr = RAWADDR;
+	printf("Writing image to 0x%x\n", raw_load_addr);
+	return raw_download;
+}
+
+#endif

src/arch/armnommu/core/serial.c

@@ -0,0 +1,66 @@
+/*
+ *  Copyright (C) 2004 Tobias Lorenz
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include "etherboot.h"
+#include "hardware.h"
+#ifdef	CONSOLE_SERIAL
+
+/*
+ * void serial_putc(int ch);
+ *	Write character `ch' to port UART_BASE.
+ */
+void serial_putc(int ch)
+{
+	/* wait for room in the 32 byte tx FIFO */
+	while ((P2001_UART->r.STATUS & 0x3f) > /* 30 */ 0) ;
+	P2001_UART->w.TX1 = ch & 0xff;
+}
+
+/*
+ * int serial_getc(void);
+ *	Read a character from port UART_BASE.
+ */
+int serial_getc(void)
+{
+	while (((P2001_UART->r.STATUS >> 6) & 0x3f) == 0) ;
+	return P2001_UART->r.RX1 & 0xff;
+}
+
+/*
+ * int serial_ischar(void);
+ *       If there is a character in the input buffer of port UART_BASE,
+ *       return nonzero; otherwise return 0.
+ */
+int serial_ischar(void)
+{
+	return (P2001_UART->r.STATUS >> 6) & 0x3f;
+}
+
+/*
+ * int serial_init(void);
+ *	Initialize port to speed 57.600, line settings 8N1.
+ */
+int serial_init(void)
+{
+	static unsigned int N;
+	// const M=3
+	P2001_UART->w.Clear = 0;		// clear
+	N = ((SYSCLK/8)*3)/CONSPEED;
+	P2001_UART->w.Baudrate = (N<<16)+3;	// set 57.600 BAUD
+	P2001_UART->w.Config = 0xcc100;		// set 8N1, *water = 12
+	return 1;
+}
+
+/*
+ * void serial_fini(void);
+ *	Cleanup our use of the serial port, in particular flush the
+ *	output buffer so we don't accidentially loose characters.
+ */
+void serial_fini(void)
+{
+}
+#endif

src/arch/armnommu/core/setjmp.S

@@ -0,0 +1,33 @@
+/*
+ *  Copyright (C) 2004 Tobias Lorenz
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+.text
+
+.global sigsetjmp;
+.type sigsetjmp,%function
+.align 4;
+sigsetjmp:
+	/* Save registers */
+	stmia	r0, {v1-v6, sl, fp, sp, lr}
+
+	mov	r0, #0
+	bx	lr
+.size sigsetjmp,.-sigsetjmp;
+
+
+
+.global longjmp;
+.type longjmp,%function
+.align 4;
+longjmp:
+	mov	ip, r0		/* save jmp_buf pointer */
+
+	movs	r0, r1		/* get the return value in place */
+	moveq	r0, #1		/* can't let setjmp() return zero! */
+
+	ldmia	ip, {v1-v6, sl, fp, sp, pc}
+.size longjmp,.-longjmp;

src/arch/armnommu/core/start.S

@@ -0,0 +1,184 @@
+/*
+ *  Copyright (C) 2004 Tobias Lorenz
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+.global _start
+
+/* Mode definitions */
+#define Mode_USR	0x10
+#define Mode_FIQ	0x11
+#define Mode_IRQ	0x12
+#define Mode_SVC	0x13
+#define Mode_ABT	0x17
+#define Mode_UNDEF	0x1B
+#define Mode_SYS	0x1F				// only available on ARM Arch. v4
+#define I_Bit		0x80
+#define F_Bit		0x40
+
+/* LPEC register definitions */
+#define Adr_SYS_BASE			0x00100000
+#define REL_Adr_SDRAM_Ctrl		0x10
+#define REL_Adr_ExtMem_Ctrl		0x14
+#define REL_Adr_WaitState_Ext		0x18
+#define REL_Adr_WaitState_Asic		0x1c
+#define Adr_TIMER_BASE			0x00110000
+#define REL_Adr_Timer12_PreDiv		0x0c
+#define REL_Adr_PLL_12000_config	0x30
+#define REL_Adr_PLL_12288_config	0x34
+#define REL_Adr_DIV_12288_config	0x38
+#define REL_Adr_FSC_CONFIG		0x44
+#define Adr_GPIO_BASE			0x00120000
+#define REL_Adr_NRES_OUT		0x2c
+
+
+
+/* Define entry point */
+.arm							// Next instruction will be ARM
+_start:
+
+/*
+ * Initialize memory system
+ */
+
+
+/*
+ * Initialize stack pointer registers
+ */
+
+/* Enter SVC mode and set up the SVC stack pointer */
+	mov	r0, #(Mode_SVC|I_Bit|F_Bit)
+	msr	cpsr_c, r0
+	ldr	sp, SP_SVC
+
+
+/*
+ * Initialize critical IO devices
+ */
+
+	/* watchdog off */
+	mov	r0, #Adr_TIMER_BASE
+	ldr	r1, Timer12_PreDiv
+	str	r1, [r0, #REL_Adr_Timer12_PreDiv]
+	
+	/* NRES=1 */
+	mov	r0, #Adr_GPIO_BASE
+	ldr	r1, NRES_OUT
+	str	r1, [r0, #REL_Adr_NRES_OUT]
+
+	/* ExtMem */
+	mov	r0, #Adr_SYS_BASE
+	ldr	r1, ExtMem_Ctrl
+	str	r1, [r0, #REL_Adr_ExtMem_Ctrl]
+
+	/* SDRAM */
+	mov	r0, #Adr_SYS_BASE
+	ldr	r1, SDRAM_Ctrl
+	str	r1, [r0, #REL_Adr_SDRAM_Ctrl]
+/*
+_wait_sdram_ctrl:
+	ldr	r1, [r0]
+	tst	r1, #0x20000
+	beq	_wait_sdram_ctrl
+*/
+
+	/* WaitState_Ext */
+	ldr	r1, WaitState_Ext
+	str	r1, [r0, #REL_Adr_WaitState_Ext]
+	/* WaitState_Asic */
+	ldr	r1, WaitState_Asic
+	str	r1, [r0, #REL_Adr_WaitState_Asic]
+
+	/* PLL_12288 */
+	mov	r0, #Adr_TIMER_BASE
+	ldr	r1, PLL_12288_config
+	str	r1, [r0, #REL_Adr_PLL_12288_config]
+	/* DIV_12288 */
+	ldr	r1, DIV_12288_config
+	str	r1, [r0, #REL_Adr_DIV_12288_config]
+	/* PLL_12200 */
+	ldr	r1, PLL_12000_config
+	str	r1, [r0, #REL_Adr_PLL_12000_config]
+
+	/* FSC_CONFIG */
+	ldr	r1, [r0, #REL_Adr_FSC_CONFIG]
+	bic	r1, r1, #0x07
+	ldr	r2, FSC_CONFIG
+	orr	r1, r1, r2
+	str	r1, [r0, #REL_Adr_FSC_CONFIG]
+
+
+/*
+ * Initialize interrupt system variables here
+ */
+
+
+/*
+ * Initialize memory required by main C code
+ */
+	
+
+/* jump to main program */
+	mov	r0, #0
+	b	main
+
+
+Timer12_PreDiv:
+	.word	0x40bb0000	/* watchdog off */
+NRES_OUT:
+	.word	0x00000003	/* NRES_OUT_DRV=1, NRES_OUT_DAT=1 */
+
+#if SYSCLK == 73728000
+ExtMem_Ctrl:
+	.word	0x000000e8	/* fuer FPGA 32 Bit konfiguriert */
+SDRAM_Ctrl:
+//	.word	0x28fc0037	/* default */
+	.word	0xaef40027	/* p2001_bit_compact */
+WaitState_Ext:
+	.word	0xa0001245	/* fuer 73 MHz */
+//	.word	0x0000fff3	/* rom bootloader */
+WaitState_Asic:
+	.word	0x00ff8a5f	/* fuer 85 MHz */
+//	.word	0x00000203	/* rom bootloader */
+PLL_12288_config:
+	.word	0x00000004	/* fuer 73 MHz */
+DIV_12288_config:
+	.word	0x00010601	/* fuer 73 MHz */
+PLL_12000_config:
+	.word	0x10004e75	/* fuer 85 MHz */
+FSC_CONFIG:
+	.word	0xc0000005	/* fuer 73 MHz */
+#else
+#error "Please define proper timings and wait states for that sysclk."
+#endif
+
+SP_SVC:
+	.word	0x40fffffc
+
+
+
+#ifndef	NORELOCATE
+/**************************************************************************
+RELOCATE_TO - relocate etherboot to the specified address
+**************************************************************************/
+	.global relocate_to
+
+relocate_to:
+	ldr	r1, =_start
+	ldr	r2, =_end
+
+	/* while (r1 < r2) { *(r0++) = *(r1++) } */
+_relocate_loop:
+	cmp	r1, r2
+	ldrcc	r3, [r1], #4
+	strcc	r3, [r0], #4
+	bcc	_relocate_loop
+	mov	pc, lr
+#endif
+
+
+.global __gccmain
+__gccmain:
+	mov	pc, lr		/* return from subroutine */

src/arch/armnommu/drivers/net/p2001_eth.c

@@ -0,0 +1,525 @@
+/**************************************************************************
+Etherboot -  BOOTP/TFTP Bootstrap Program
+P2001 NIC driver for Etherboot
+***************************************************************************/
+
+/*
+ *  Copyright (C) 2004 Tobias Lorenz
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+/* to get some global routines like printf */
+#include "etherboot.h"
+/* to get the interface to the body of the program */
+#include "nic.h"
+/* to get the ISA support functions, if this is an ISA NIC */
+#include "isa.h"
+
+#include "hardware.h"
+#include "lxt971a.h"
+#include "timer.h"
+
+
+/* NIC specific static variables go here */
+static unsigned char MAC_HW_ADDR[6]={MAC_HW_ADDR_DRV};
+
+/* DMA descriptors and buffers */
+#define NUM_RX_DESC     4	/* Number of Rx descriptor registers. */
+#define DMA_BUF_SIZE	2048	/* Buffer size */
+static DMA_DSC txd              __attribute__ ((__section__(".dma.desc")));
+static DMA_DSC rxd[NUM_RX_DESC] __attribute__ ((__section__(".dma.desc")));
+static unsigned char rxb[NUM_RX_DESC * DMA_BUF_SIZE] __attribute__ ((__section__(".dma.buffer")));
+static unsigned char txb[              DMA_BUF_SIZE] __attribute__ ((__section__(".dma.buffer")));
+static unsigned int cur_rx;
+
+/* Device selectors */
+static unsigned int cur_channel;	// DMA channel    : 0..3
+static unsigned int cur_phy;		// PHY Address    : 0..31
+static P2001_ETH_regs_ptr EU;		// Ethernet Unit  : 0x0018_000 with _=0..3
+static P2001_ETH_regs_ptr MU;		// Management Unit: 0x00180000
+
+#define MDIO_MAXCOUNT 1000			/* mdio abort */
+static unsigned int mdio_error;			/* mdio error */
+
+/* Function prototypes */
+static void         p2001_eth_mdio_init ();
+static void         p2001_eth_mdio_write(unsigned int phyadr, unsigned int regadr, unsigned int data);
+static unsigned int p2001_eth_mdio_read (unsigned int phyadr, unsigned int regadr);
+extern unsigned int p2001_eth_mdio_error;
+
+static int          p2001_eth_poll      (struct nic *nic, int retrieve);
+static void         p2001_eth_transmit  (struct nic *nic, const char *d,
+					unsigned int t, unsigned int s, const char *p);
+
+static void         p2001_eth_irq       (struct nic *nic, irq_action_t action);
+
+static void         p2001_eth_init      ();
+static void         p2001_eth_disable   (struct dev *dev);
+
+static int          p2001_eth_check_link(unsigned int phy);
+static int          p2001_eth_probe     (struct dev *dev, unsigned short *probe_addrs __unused);
+
+
+/**************************************************************************
+PHY MANAGEMENT UNIT - Read/write
+***************************************************************************/
+static void p2001_eth_mdio_init()
+{
+	/* reset ethernet PHYs */
+	printf("Resetting PHYs...\n");
+
+	/* GPIO24/25: TX_ER2/TX_ER0 */
+	/* GPIO26/27: PHY_RESET/TX_ER1 */
+	P2001_GPIO->PIN_MUX |= 0x0018;
+	// 31-16: 0000 1111 0000 0000
+	P2001_GPIO->GPIO2_En |= 0x0400;
+
+	P2001_GPIO->GPIO2_Out |= 0x04000000;
+	P2001_GPIO->GPIO2_Out &= ~0x0400;
+	mdelay(500);
+	P2001_GPIO->GPIO2_Out |= 0x0400;
+
+	/* set management unit clock divisor */
+	// max. MDIO CLK = 2.048 MHz (EU.doc)
+	// max. MDIO CLK = 8.000 MHz (LXT971A)
+	// sysclk/(2*(n+1)) = MDIO CLK <= 2.048 MHz
+	// n >= sysclk/4.096 MHz - 1
+#if SYSCLK == 73728000
+	P2001_MU->MU_DIV = 17;	// 73.728 MHZ =17=> 2.020 MHz
+#else
+	//MU->MU_DIV = (SYSCLK/4.096)-1;
+#error "Please define a proper MDIO CLK divisor for that sysclk."
+#endif
+	asm("nop \n nop");
+}
+
+static void p2001_eth_mdio_write(unsigned int phyadr, unsigned int regadr, unsigned int data)
+{
+	static unsigned int count;
+	count = 0;
+
+	/* Warten bis Hardware inaktiv (MIU = "0") */
+	while ((MU->MU_CNTL & 0x8000) && (count < MDIO_MAXCOUNT))
+		count++;
+
+	/* Schreiben MU_DATA */
+	MU->MU_DATA = data;
+
+	/* Schreiben MU_CNTL */
+	MU->MU_CNTL = regadr + (phyadr<<5) + (1<<10);
+
+	/* Warten bis Hardware aktiv (MIU = "1") */
+	while (((MU->MU_CNTL & 0x8000) == 0) && (count < MDIO_MAXCOUNT))
+		count++;
+	//asm("nop \r\n nop");
+
+	/* Warten bis Hardware inaktiv (MIU = "0") */
+	while ((MU->MU_CNTL & 0x8000) && (count < MDIO_MAXCOUNT))
+		count++;
+
+	mdio_error = (count >= MDIO_MAXCOUNT);
+}
+
+static unsigned int p2001_eth_mdio_read(unsigned int phyadr, unsigned int regadr)
+{
+	static unsigned int count;
+	count = 0;
+
+	do {
+		/* Warten bis Hardware inaktiv (MIU = "0") */
+		while ((MU->MU_CNTL & 0x8000) && (count < MDIO_MAXCOUNT))
+			count++;
+
+		/* Schreiben MU_CNTL */
+		MU->MU_CNTL = regadr + (phyadr<<5) + (2<<10);
+
+		/* Warten bis Hardware aktiv (MIU = "1") */
+		while (((MU->MU_CNTL & 0x8000) == 0) && (count < MDIO_MAXCOUNT))
+			count++;
+		//asm("nop \r\n nop");
+
+		/* Warten bis Hardware inaktiv (MIU = "0") */
+		while ((MU->MU_CNTL & 0x8000) && (count < MDIO_MAXCOUNT))
+			count++;
+
+		/* Fehler, wenn MDIO Read Error (MRE = "1") */
+	} while ((MU->MU_CNTL & 0x4000) && (count < MDIO_MAXCOUNT));
+
+	/* Lesen MU_DATA */
+	mdio_error = (count >= MDIO_MAXCOUNT);
+	return MU->MU_DATA;
+}
+
+
+/**************************************************************************
+POLL - Wait for a frame
+***************************************************************************/
+/* Function: p2001_eth_poll
+ *
+ * Description: checks for a received packet and returns it if found.
+ *
+ * Arguments: struct nic *nic:          NIC data structure
+ *
+ * Returns:   1 if a packet was received.
+ *            0 if no pacet was received.
+ *
+ * Side effects:
+ *            Returns (copies) the packet to the array nic->packet.
+ *            Returns the length of the packet in nic->packetlen.
+ */
+static int p2001_eth_poll(struct nic *nic, int retrieve)
+{
+	/* return true if there's an ethernet packet ready to read */
+	/* nic->packet should contain data on return */
+	/* nic->packetlen should contain length of data */
+
+	int retstat = 0;
+
+	if (rxd[cur_rx].stat & (1<<31))	// OWN
+		return retstat;
+
+	if (!retrieve)
+		return 1;
+
+	nic->packetlen = rxd[cur_rx].cntl & 0xffff;
+
+	if (rxd[cur_rx].stat & ((1<<26)|(1<<25)|(1<<24)|(1<<23)|(1<<22))) {
+		/* corrupted packet received */
+		printf("p2001_eth_poll: Corrupted packet received, stat = %X\n",
+		               rxd[cur_rx].stat);
+		retstat = 0;
+	} else {
+		/* give packet to higher routine */
+		memcpy(nic->packet, (rxb + cur_rx*DMA_BUF_SIZE), nic->packetlen);
+		retstat = 1;
+	}
+
+#ifdef DEBUG_NIC
+	printf("p2001_eth_poll: packet from %! to %! received\n", 
+		(rxb+cur_rx*DMA_BUF_SIZE)+ETH_ALEN,
+		(rxb+cur_rx*DMA_BUF_SIZE));
+#endif
+
+	/* disable receiver */
+	// FIXME: is that ok? it can produce grave errors.
+	EU->RMAC_DMA_EN = 0;				/* clear run bit */
+
+	/* return the descriptor and buffer to receive ring */
+	rxd[cur_rx].stat = (1<<31) | (1<<30) | (1<<29);	// DSC0 OWN|START|END
+	rxd[cur_rx].cntl = (1<<23);			// DSC1 RECEIVE
+	rxd[cur_rx].cntl |= cur_channel << 16;		// DSC1 CHANNEL
+	rxd[cur_rx].cntl |= DMA_BUF_SIZE;		// DSC1 LEN
+
+	if (++cur_rx == NUM_RX_DESC)
+        	cur_rx = 0;
+
+	/* enable receiver */
+	if (!(EU->RMAC_DMA_EN & 0x01))
+		EU->RMAC_DMA_EN = 0x01;			/* set run bit */
+
+#ifdef DEBUG_NIC
+	printf("RMAC_MIB0..5: %d:%d:%d:%d:%d:%d\n",
+		EU->RMAC_MIB0, EU->RMAC_MIB1,
+		EU->RMAC_MIB2, EU->RMAC_MIB3,
+		EU->RMAC_MIB4, EU->RMAC_MIB5);
+#endif
+
+	return retstat;	/* initially as this is called to flush the input */
+}
+
+
+/**************************************************************************
+TRANSMIT - Transmit a frame
+***************************************************************************/
+/* Function: p2001_eth_transmit
+ *
+ * Description: transmits a packet and waits for completion or timeout.
+ *
+ * Arguments: char d[6]:          destination ethernet address.
+ *            unsigned short t:   ethernet protocol type.
+ *            unsigned short s:   size of the data-part of the packet.
+ *            char *p:            the data for the packet.
+ *    
+ * Returns:   void.
+ */
+static void p2001_eth_transmit(
+	struct nic *nic __unused,
+	const char *d,			/* Destination */
+	unsigned int t,			/* Type */
+	unsigned int s,			/* size */
+	const char *p)			/* Packet */
+{
+	unsigned int nstype;
+#ifdef DEBUG_NIC
+	unsigned int status;
+#endif
+
+	/* assemble packet */
+	memcpy(txb, d, ETH_ALEN);			// destination
+	memcpy(txb+ETH_ALEN, nic->node_addr, ETH_ALEN);	// source
+	nstype = htons(t);
+	memcpy(txb+2*ETH_ALEN, (char*)&nstype, 2);	// type
+	memcpy(txb+ETH_HLEN, p, s);			// packet
+	s += ETH_HLEN;
+
+	/* pad to minimum packet size */
+//	while (s<ETH_ZLEN)
+//		txb[s++] = '\0';
+	// TMAC_CNTL.ATP does the same
+
+#ifdef DEBUG_NIC
+	printf("p2001_eth_transmit: packet from %! to %! sent\n", txb+ETH_ALEN, txb);
+#endif
+
+	/* configure descriptor */
+	txd.stat = (1<<31) | (1<<30) | (1<<29);	// DSC0 OWN|START|END
+	txd.cntl = cur_channel << 16;		// DSC1 CHANNEL
+	txd.cntl |= s;				// DSC1 LEN
+
+	/* restart the transmitter */
+	EU->TMAC_DMA_EN = 0x01;		/* set run bit */
+	while(EU->TMAC_DMA_EN & 0x01) ;	/* wait */
+
+#ifdef DEBUG_NIC
+	/* check status */
+	status = EU->TMAC_DMA_STAT;
+	if (status & ~(0x40))
+		printf("p2001_eth_transmit: dma status=0x%hx\n", status);
+
+	printf("TMAC_MIB6..7: %d:%d\n", EU->TMAC_MIB6, EU->TMAC_MIB7);
+#endif
+}
+
+
+/**************************************************************************
+IRQ - Enable, Disable or Force Interrupts
+***************************************************************************/
+/* Function: p2001_eth_irq
+ *
+ * Description: Enable, Disable, or Force, interrupts
+ *    
+ * Arguments: struct nic *nic:          NIC data structure
+ *            irq_action_t action:      Requested action       
+ *
+ * Returns:   void.
+ */
+
+static void
+p2001_eth_irq(struct nic *nic __unused, irq_action_t action __unused)
+{
+	switch ( action ) {
+		case DISABLE :
+			break;
+		case ENABLE :
+			break;
+		case FORCE :
+			break;
+	}
+}
+
+
+/**************************************************************************
+INIT - Initialize device
+***************************************************************************/
+/* Function: p2001_init
+ *
+ * Description: resets the ethernet controller chip and various
+ *    data structures required for sending and receiving packets.
+ *    
+ * returns:   void.
+ */
+static void p2001_eth_init()
+{
+	static int i;
+
+	/* disable transceiver */
+//	EU->TMAC_DMA_EN = 0;		/* clear run bit */
+//	EU->RMAC_DMA_EN = 0;		/* clear run bit */
+
+	/* set rx filter (physical mac addresses) */
+	EU->RMAC_PHYU =
+		(MAC_HW_ADDR[0]<< 8) +
+		(MAC_HW_ADDR[1]<< 0);
+	EU->RMAC_PHYL =
+		(MAC_HW_ADDR[2]<<24) +
+		(MAC_HW_ADDR[3]<<16) +
+		(MAC_HW_ADDR[4]<<8 ) +
+		(MAC_HW_ADDR[5]<<0 );
+
+	/* initialize the tx descriptor ring */
+//	txd.stat = (1<<31) | (1<<30) | (1<<29);			// DSC0 OWN|START|END
+//	txd.cntl = cur_channel << 16;				// DSC1 CHANNEL
+//	txd.cntl |= DMA_BUF_SIZE;				// DSC1 LEN
+	txd.buf = &txb;						// DSC2 BUFFER
+	txd.next = &txd;					// DSC3 NEXTDSC @self
+	EU->TMAC_DMA_DESC = &txd;
+
+	/* initialize the rx descriptor ring */
+	cur_rx = 0;
+	for (i = 0; i < NUM_RX_DESC; i++) {
+		rxd[i].stat = (1<<31) | (1<<30) | (1<<29);	// DSC0 OWN|START|END
+		rxd[i].cntl = (1<<23);				// DSC1 RECEIVE
+		rxd[i].cntl |= cur_channel << 16;		// DSC1 CHANNEL
+		rxd[i].cntl |= DMA_BUF_SIZE;			// DSC1 LEN
+		rxd[i].buf = &rxb[i*DMA_BUF_SIZE];		// DSC2 BUFFER (EU-RX data)
+		rxd[i].next = &rxd[i+1];			// DSC3 NEXTDSC @next
+	}
+	rxd[NUM_RX_DESC-1].next = &rxd[0];			// DSC3 NEXTDSC @first
+	EU->RMAC_DMA_DESC = &rxd[0];
+
+	/* set transmitter mode */
+	EU->TMAC_CNTL = (1<<4) |	/* COI: Collision ignore */
+			//(1<<3) |	/* CSI: Carrier Sense ignore */
+			(1<<2);		/* ATP: Automatic Transmit Padding */
+
+	/* set receive mode */
+	EU->RMAC_CNTL = (1<<3) |	/* BROAD: Broadcast packets */
+			(1<<1);		/* PHY  : Packets to out MAC address */
+
+	/* enable receiver */
+	EU->RMAC_DMA_EN = 1;		/* set run bit */
+}
+
+
+/**************************************************************************
+DISABLE - Turn off ethernet interface
+***************************************************************************/
+static void p2001_eth_disable(struct dev *dev __unused)
+{
+	/* put the card in its initial state */
+	/* This function serves 3 purposes.
+	 * This disables DMA and interrupts so we don't receive
+	 *  unexpected packets or interrupts from the card after
+	 *  etherboot has finished. 
+	 * This frees resources so etherboot may use
+	 *  this driver on another interface
+	 * This allows etherboot to reinitialize the interface
+	 *  if something is something goes wrong.
+	 */
+
+	/* disable transmitter */
+	EU->TMAC_DMA_EN = 0;		/* clear run bit */
+
+	/* disable receiver */
+	EU->RMAC_DMA_EN = 0;		/* clear run bit */
+}
+
+
+/**************************************************************************
+LINK - Check for valid link
+***************************************************************************/
+static int p2001_eth_check_link(unsigned int phy)
+{
+	static int status;
+	static unsigned int count;
+	count = 0;
+
+	/* Use 0x3300 for restarting NWay */
+	printf("Starting auto-negotiation... ");
+	p2001_eth_mdio_write(phy, Adr_LXT971A_Control, 0x3300);
+	if (mdio_error)
+		goto failed;
+
+	/* Bits 1.5 and 17.7 are set to 1 once the Auto-Negotiation process to completed. */
+	do {
+		mdelay(500);
+		status = p2001_eth_mdio_read(phy, Adr_LXT971A_Status1);
+		if (mdio_error || (count++ > 6))	// 6*500ms = 3s timeout
+			goto failed;
+	} while (!(status & 0x20));
+	
+	/* Bits 1.2 and 17.10 are set to 1 once the link is established. */
+	if (p2001_eth_mdio_read(phy, Adr_LXT971A_Status1) & 0x04) {
+		/* Bits 17.14 and 17.9 can be used to determine the link operation conditions (speed and duplex). */
+		printf("Valid link, operating at: %sMb-%s\n",
+			(p2001_eth_mdio_read(phy, Adr_LXT971A_Status2) & 0x4000) ? "100" : "10",
+			(p2001_eth_mdio_read(phy, Adr_LXT971A_Status2) & 0x0200) ? "FD" : "HD");
+			return 1;
+	}
+
+failed:
+	if (mdio_error)
+		printf("Failed\n");
+	else
+		printf("No valid link\n");
+	return 0;
+}
+
+
+/**************************************************************************
+PROBE - Look for an adapter, this routine's visible to the outside
+***************************************************************************/
+static int p2001_eth_probe(struct dev *dev, unsigned short *probe_addrs __unused)
+{
+	struct nic *nic = (struct nic *)dev;
+	/* if probe_addrs is 0, then routine can use a hardwired default */
+	static int board_found;
+	static int valid_link;
+
+	/* reset phys and configure mdio clk */
+	p2001_eth_mdio_init();
+
+	/* find the correct PHY/DMA/MAC combination */
+	MU = P2001_MU;	// MU for all PHYs is only in EU0
+	printf("Searching for P2001 NICs...\n");
+	for (cur_channel=0; cur_channel<4; cur_channel++) {
+		switch(cur_channel) {
+			case 0:
+				EU = P2001_EU0;
+				cur_phy = 0;
+				break;
+			case 1:
+				EU = P2001_EU1;
+				cur_phy = 1;
+				break;
+			case 2:
+				EU = P2001_EU2;
+				cur_phy = 2;
+				break;
+			case 3:
+				EU = P2001_EU3;
+				cur_phy = 3;
+				break;
+		}
+
+		/* first a non destructive test for initial value RMAC_TLEN=1518 */
+		board_found = (EU->RMAC_TLEN == 1518);
+		if (board_found) {
+			printf("Checking EU%d...\n", cur_channel);
+
+			valid_link = p2001_eth_check_link(cur_phy);
+			if (valid_link) {
+				/* initialize device */
+				p2001_eth_init(nic);
+
+				/* set node address */
+				printf("Setting MAC address to %!\n", MAC_HW_ADDR);
+				memcpy(nic->node_addr, MAC_HW_ADDR, 6);
+
+				/* point to NIC specific routines */
+				dev->disable  = p2001_eth_disable;
+				nic->poll     = p2001_eth_poll;
+				nic->transmit = p2001_eth_transmit;
+				nic->irq      = p2001_eth_irq;
+
+				/* Report the ISA pnp id of the board */
+				dev->devid.vendor_id = htons(GENERIC_ISAPNP_VENDOR);
+				dev->devid.vendor_id = htons(0x1234);
+				return 1;
+			}
+		}
+	}
+	/* else */
+	return 0;
+}
+
+ISA_ROM("p2001_eth", "P2001 Ethernet Driver")
+static struct isa_driver p2001_eth_driver __isa_driver = {
+	.type    = NIC_DRIVER,
+	.name    = "P2001 Ethernet Driver",
+	.probe   = p2001_eth_probe,
+	.ioaddrs = 0,
+};

src/arch/armnommu/include/bits/byteswap.h

@@ -0,0 +1,47 @@
+/*
+ *  Copyright (C) 2004 Tobias Lorenz
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#ifndef ETHERBOOT_BITS_BYTESWAP_H
+#define ETHERBOOT_BITS_BYTESWAP_H
+
+/* We do not have byte swap functions ... We are
+ * RISC processor ...
+ */
+
+static inline unsigned short __swap16(volatile unsigned short v)
+{
+    return ((v << 8) | (v >> 8));
+}
+
+static inline unsigned int __swap32(volatile unsigned long v)
+{
+    return ((v << 24) | ((v & 0xff00) << 8) | ((v & 0xff0000) >> 8) | (v >> 24));
+}
+
+#define __bswap_constant_16(x) \
+	((uint16_t)((((uint16_t)(x) & 0x00ff) << 8) | \
+		(((uint16_t)(x) & 0xff00) >> 8)))
+
+#define __bswap_constant_32(x) \
+	((uint32_t)((((uint32_t)(x) & 0x000000ffU) << 24) | \
+		(((uint32_t)(x) & 0x0000ff00U) <<  8) | \
+		(((uint32_t)(x) & 0x00ff0000U) >>  8) | \
+		(((uint32_t)(x) & 0xff000000U) >> 24)))
+
+# define __bswap_16(x) \
+    (__extension__							      \
+     ({ unsigned short int __bsx = (x);					      \
+        ((((__bsx) >> 8) & 0xff) | (((__bsx) & 0xff) << 8)); }))
+
+
+# define __bswap_32(x) \
+    (__extension__							      \
+     ({ unsigned int __bsx = (x);					      \
+        ((((__bsx) & 0xff000000) >> 24) | (((__bsx) & 0x00ff0000) >>  8) |    \
+	 (((__bsx) & 0x0000ff00) <<  8) | (((__bsx) & 0x000000ff) << 24)); }))
+
+#endif /* ETHERBOOT_BITS_BYTESWAP_H */

src/arch/armnommu/include/bits/cpu.h

@@ -0,0 +1,13 @@
+/*
+ *  Copyright (C) 2004 Tobias Lorenz
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#ifndef ARM_BITS_CPU_H
+#define ARM_BITS_CPU_H
+
+#define cpu_setup() do {} while(0)
+
+#endif /* ARM_BITS_CPU_H */

src/arch/armnommu/include/bits/elf.h

@@ -0,0 +1,18 @@
+/*
+ *  Copyright (C) 2004 Tobias Lorenz
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#ifndef ARM_BITS_ELF_H
+#define ARM_BITS_ELF_H
+
+/* ELF Defines for the current architecture */
+#define	EM_CURRENT	EM_ARM
+#define ELFDATA_CURRENT	ELFDATA2LSB
+
+#define ELF_CHECK_ARCH(x) \
+	((x).e_machine == EM_CURRENT)
+
+#endif /* ARM_BITS_ELF_H */

src/arch/armnommu/include/bits/endian.h

@@ -0,0 +1,17 @@
+/*
+ *  Copyright (C) 2004 Tobias Lorenz
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#ifndef ETHERBOOT_BITS_ENDIAN_H
+#define ETHERBOOT_BITS_ENDIAN_H
+
+#ifdef __ARMEB__
+#define __BYTE_ORDER __BIG_ENDIAN
+#else
+#define __BYTE_ORDER __LITTLE_ENDIAN
+#endif
+
+#endif /* ETHERBOOT_BITS_ENDIAN_H */

src/arch/armnommu/include/bits/string.h

@@ -0,0 +1,11 @@
+/*
+ *  Copyright (C) 2004 Tobias Lorenz
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#ifndef ETHERBOOT_BITS_STRING_H
+#define ETHERBOOT_BITS_STRING_H
+
+#endif /* ETHERBOOT_BITS_STRING_H */

src/arch/armnommu/include/callbacks_arch.h

@@ -0,0 +1 @@
+/* empty file */

src/arch/armnommu/include/hardware.h

@@ -0,0 +1,179 @@
+/*
+ *  Copyright (C) 2004 Tobias Lorenz
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+/*
+ * Architecture: ARM9TDMI
+ * Processor   : P2001
+ */
+
+#ifndef ARCH_HARDWARE_H
+#define ARCH_HARDWARE_H
+
+#ifndef __ASSEMBLY__
+
+/* DMA descriptor */
+typedef struct {
+	unsigned int stat;			/* status: own, start, end, offset, status */
+	unsigned int cntl;			/* control: loop, int, type, channel, length */
+	char	*buf;				/* buffer */
+	void	*next;				/* nextdsc */
+} DMA_DSC;
+
+
+/* The address definitions are from asic_bf.h */
+typedef struct {					// 0x00100000U
+	volatile unsigned int reserved1[0x3];
+	volatile unsigned int ArmDmaPri;		// 0x0000000CU
+	volatile unsigned int SDRAM_Ctrl;		// 0x00000010U
+	volatile unsigned int ExtMem_Ctrl;		// 0x00000014U
+	volatile unsigned int WaitState_Ext;		// 0x00000018U
+	volatile unsigned int WaitState_Asic;		// 0x0000001CU
+	volatile unsigned int TOP;			// 0x00000020U
+	volatile unsigned int reserved2[0x3];
+	volatile unsigned int Adr1_EQ_30Bit;		// 0x00000030U
+	volatile unsigned int Adr2_EQ_30Bit;		// 0x00000034U
+	volatile unsigned int Adr3_EQ_30Bit;		// 0x00000038U
+	volatile unsigned int Dat3_EQ_32Bit;		// 0x0000003CU
+	volatile unsigned int Adr4_HE_20Bit;		// 0x00000040U
+	volatile unsigned int Adr4_LT_20Bit;		// 0x00000044U
+	volatile unsigned int Adr5_HE_20Bit;		// 0x00000048U
+	volatile unsigned int Adr5_LT_20Bit;		// 0x0000004CU
+	volatile unsigned int Adr_Control;		// 0x00000050U
+	volatile unsigned int ABORT_IA_32Bit;		// 0x00000054U
+} *P2001_SYS_regs_ptr;
+#define P2001_SYS ((volatile P2001_SYS_regs_ptr) 0x00100000)
+
+typedef struct {					// 0x00110000U
+	volatile unsigned int Timer1;			// 0x00000000U
+	volatile unsigned int Timer2;			// 0x00000004U
+	volatile unsigned int TIMER_PRELOAD;		// 0x00000008U
+	volatile unsigned int Timer12_PreDiv;		// 0x0000000CU
+	volatile unsigned int TIMER_INT;		// 0x00000010U
+	volatile unsigned int Freerun_Timer;		// 0x00000014U
+	volatile unsigned int WatchDog_Timer;		// 0x00000018U
+	volatile unsigned int PWM_CNT;			// 0x00000020U
+	volatile unsigned int PWM_CNT2;			// 0x00000024U
+	volatile unsigned int PLL_12000_config;		// 0x00000030U
+	volatile unsigned int PLL_12288_config;		// 0x00000034U
+	volatile unsigned int DIV_12288_config;		// 0x00000038U
+	volatile unsigned int MOD_CNT_768;		// 0x0000003CU
+	volatile unsigned int FSC_IRQ_STATUS;		// 0x00000040U
+	volatile unsigned int FSC_CONFIG;		// 0x00000044U
+	volatile unsigned int FSC_CONSTRUCT;		// 0x00000048U
+	volatile unsigned int FSC_base_clk_reg;		// 0x0000004CU
+	volatile unsigned int SYSCLK_SHAPE;		// 0x00000050U
+	volatile unsigned int SDRAMCLK_SHAPE;		// 0x00000054U
+	volatile unsigned int RING_OSZI;		// 0x00000058U
+} *P2001_TIMER_regs_ptr;
+#define P2001_TIMER ((volatile P2001_TIMER_regs_ptr) 0x00110000)
+
+typedef struct {					// 0x00120000U
+	volatile unsigned int reserved1[0x5];
+	volatile unsigned int GPIO_Config;		// 0x00000014U
+	volatile unsigned int GPIO_INT;			// 0x00000018U
+	volatile unsigned int GPIO_Out;			// 0x0000001CU
+	volatile unsigned int GPIO_IN;			// 0x00000020U
+	volatile unsigned int GPIO_En;			// 0x00000024U
+	volatile unsigned int PIN_MUX;			// 0x00000028U
+	volatile unsigned int NRES_OUT;			// 0x0000002CU
+	volatile unsigned int GPIO2_Out;		// 0x00000030U
+	volatile unsigned int GPIO2_IN;			// 0x00000034U
+	volatile unsigned int GPIO2_En;			// 0x00000038U
+	volatile unsigned int GPIO_INT_SEL;		// 0x0000003CU
+	volatile unsigned int GPI3_IN;			// 0x00000040U
+	volatile unsigned int GPO4_OUT;			// 0x00000044U
+} *P2001_GPIO_regs_ptr;
+#define P2001_GPIO ((volatile P2001_GPIO_regs_ptr) 0x00120000)
+
+typedef struct {					// 0x00130000U
+	volatile unsigned int Main_NFIQ_Int_Ctrl;	// 0x00000000U
+	volatile unsigned int Main_NIRQ_Int_Ctrl;	// 0x00000004U
+	volatile unsigned int Status_NFIQ;		// 0x00000008U
+	volatile unsigned int Status_NIRQ;		// 0x0000000CU
+} *P2001_INT_CTRL_regs_ptr;
+#define P2001_INT_CTRL ((volatile P2001_INT_CTRL_regs_ptr) 0x00130000)
+
+typedef union {						// 0x00140000U
+	struct {	// write
+		volatile unsigned int TX1;		// 0x00000000U
+		volatile unsigned int TX2;		// 0x00000004U
+		volatile unsigned int TX3;		// 0x00000008U
+		volatile unsigned int TX4;		// 0x0000000CU
+		volatile unsigned int Baudrate;		// 0x00000010U
+		volatile unsigned int reserved1[0x3];
+		volatile unsigned int Config;		// 0x00000020U
+		volatile unsigned int Clear;		// 0x00000024U
+		volatile unsigned int Echo_EN;		// 0x00000028U
+		volatile unsigned int IRQ_Status;	// 0x0000002CU
+	} w;		// write
+	
+	struct {	// read
+		volatile unsigned int RX1;		// 0x00000000U
+		volatile unsigned int RX2;		// 0x00000004U
+		volatile unsigned int RX3;		// 0x00000008U
+		volatile unsigned int RX4;		// 0x0000000CU
+		volatile unsigned int reserved1[0x4];
+		volatile unsigned int PRE_STATUS;	// 0x00000020U
+		volatile unsigned int STATUS;		// 0x00000024U
+		volatile unsigned int reserved2[0x1];
+		volatile unsigned int IRQ_Status;	// 0x0000002CU
+	} r;		// read
+} *P2001_UART_regs_ptr;
+#define P2001_UART ((volatile P2001_UART_regs_ptr) 0x00140000)
+
+typedef struct {				// 0x0018_000U _=0,1,2,3
+	volatile DMA_DSC *    RMAC_DMA_DESC;	// 0x00000000U
+	volatile unsigned int RMAC_DMA_CNTL;	// 0x00000004U
+	volatile unsigned int RMAC_DMA_STAT;	// 0x00000008U
+	volatile unsigned int RMAC_DMA_EN;	// 0x0000000CU
+	volatile unsigned int RMAC_CNTL;	// 0x00000010U
+	volatile unsigned int RMAC_TLEN;	// 0x00000014U
+	volatile unsigned int RMAC_PHYU;	// 0x00000018U
+	volatile unsigned int RMAC_PHYL;	// 0x0000001CU
+	volatile unsigned int RMAC_PFM0;	// 0x00000020U
+	volatile unsigned int RMAC_PFM1;	// 0x00000024U
+	volatile unsigned int RMAC_PFM2;	// 0x00000028U
+	volatile unsigned int RMAC_PFM3;	// 0x0000002CU
+	volatile unsigned int RMAC_PFM4;	// 0x00000030U
+	volatile unsigned int RMAC_PFM5;	// 0x00000034U
+	volatile unsigned int RMAC_PFM6;	// 0x00000038U
+	volatile unsigned int RMAC_PFM7;	// 0x0000003CU
+	volatile unsigned int RMAC_MIB0;	// 0x00000040U
+	volatile unsigned int RMAC_MIB1;	// 0x00000044U
+	volatile unsigned int RMAC_MIB2;	// 0x00000048U
+	volatile unsigned int RMAC_MIB3;	// 0x0000004CU
+	volatile unsigned int RMAC_MIB4;	// 0x00000050U
+	volatile unsigned int RMAC_MIB5;	// 0x00000054U
+	volatile unsigned int reserved1[0x1e8];
+	volatile unsigned int RMAC_DMA_DATA;	// 0x000007F8U
+	volatile unsigned int RMAC_DMA_ADR;	// 0x000007FCU
+	volatile DMA_DSC *    TMAC_DMA_DESC;	// 0x00000800U
+	volatile unsigned int TMAC_DMA_CNTL;	// 0x00000804U
+	volatile unsigned int TMAC_DMA_STAT;	// 0x00000808U
+	volatile unsigned int TMAC_DMA_EN;	// 0x0000080CU
+	volatile unsigned int TMAC_CNTL;	// 0x00000810U
+	volatile unsigned int TMAC_MIB6;	// 0x00000814U
+	volatile unsigned int TMAC_MIB7;	// 0x00000818U
+	volatile unsigned int reserved2[0x1];
+	volatile unsigned int MU_CNTL;		// 0x00000820U
+	volatile unsigned int MU_DATA;		// 0x00000824U
+	volatile unsigned int MU_DIV;		// 0x00000828U
+	volatile unsigned int CONF_RMII;	// 0x0000082CU
+	volatile unsigned int reserved3[0x1f2];
+	volatile unsigned int TMAC_DMA_DATA;	// 0x00000FF8U
+	volatile unsigned int TMAC_DMA_ADR;	// 0x00000FFCU
+} *P2001_ETH_regs_ptr;
+#define P2001_EU0 ((volatile P2001_ETH_regs_ptr) 0x00180000)
+#define P2001_EU1 ((volatile P2001_ETH_regs_ptr) 0x00181000)
+#define P2001_EU2 ((volatile P2001_ETH_regs_ptr) 0x00182000)
+#define P2001_EU3 ((volatile P2001_ETH_regs_ptr) 0x00183000)
+#define P2001_MU  P2001_EU0
+
+#endif
+
+#endif  /* ARCH_HARDWARE_H */

src/arch/armnommu/include/hooks.h

@@ -0,0 +1,25 @@
+/*
+ *  Copyright (C) 2004 Tobias Lorenz
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#ifndef ETHERBOOT_ARM_HOOKS_H
+#define ETHERBOOT_ARM_HOOKS_H
+
+struct Elf_Bhdr;
+
+#define arch_main(data, params) do {} while(0)
+//void arch_main(in_call_data_t *data, va_list params);
+
+#define arch_on_exit(status) do {} while(0)
+//void arch_on_exit(int status);
+
+#define arch_relocate_to(addr) do {} while(0)
+//void arch_relocate_to(unsigned long addr);
+
+#define arch_relocated_from(old_addr) do {} while(0)
+//void arch_relocate_from(unsigned long old_addr);
+
+#endif /* ETHERBOOT_ARM_HOOKS_H */

src/arch/armnommu/include/io.h

@@ -0,0 +1,27 @@
+/*
+ *  Copyright (C) 2004 Tobias Lorenz
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#ifndef ETHERBOOT_IO_H
+#define ETHERBOOT_IO_H
+
+#define virt_to_phys(vaddr)	((unsigned long) (vaddr))
+#define phys_to_virt(vaddr)	((void *) (vaddr))
+
+#define virt_to_bus virt_to_phys
+#define bus_to_virt phys_to_virt
+
+#define iounmap(addr)				((void)0)
+#define ioremap(physaddr, size)			(physaddr)
+
+extern unsigned char inb (unsigned long int port);
+extern unsigned short int inw (unsigned long int port);
+extern unsigned long int inl (unsigned long int port);
+extern void outb (unsigned char value, unsigned long int port);
+extern void outw (unsigned short value, unsigned long int port);
+extern void outl (unsigned long value, unsigned long int port);
+
+#endif /* ETHERBOOT_IO_H */

src/arch/armnommu/include/latch.h

@@ -0,0 +1,15 @@
+/*
+ *  Copyright (C) 2004 Tobias Lorenz
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#ifndef LATCH_H
+#define LATCH_H
+
+// Freerun_Timer is always at 12.288 MHZ
+#define TICKS_PER_SEC		(12288000UL)
+//#define TICKS_PER_SEC		(73728000UL)
+
+#endif /* LATCH_H */

src/arch/armnommu/include/limits.h

@@ -0,0 +1,43 @@
+/*
+ *  Copyright (C) 2004 Tobias Lorenz
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#ifndef __LIMITS_H
+#define __LIMITS_H  1
+
+#define MB_LEN_MAX	16
+
+#define CHAR_BIT	8
+
+#define SCHAR_MIN	(-128)
+#define SCHAR_MAX	127
+
+#define UCHAR_MAX	255
+
+#define CHAR_MIN	SCHAR_MIN
+#define CHAR_MAX	SCHAR_MAX
+
+#define SHRT_MIN	(-32768)
+#define SHRT_MAX	32767
+
+#define USHRT_MAX	65535
+
+#define INT_MIN	(-INT_MAX - 1)
+#define INT_MAX	2147483647
+
+#define UINT_MAX	4294967295U
+
+#define LONG_MAX	2147483647L
+#define LONG_MIN	(-LONG_MAX - 1L)
+
+#define ULONG_MAX	4294967295UL
+
+#define LLONG_MAX	9223372036854775807LL
+#define LLONG_MIN	(-LLONG_MAX - 1LL)
+
+#define ULLONG_MAX	18446744073709551615ULL
+
+#endif

src/arch/armnommu/include/lxt971a.h

@@ -0,0 +1,30 @@
+/*
+ *  Copyright (C) 2004 Tobias Lorenz
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+/*
+ * Intel LXT971ALE (MII-compatible PHY)
+ */
+
+#define Adr_LXT971A_Control		0	/* Control Register */
+#define Adr_LXT971A_Status1		1	/* MII Status Register #1 */
+#define Adr_LXT971A_PHY_ID1		2	/* PHY Identification Register 1 */
+#define Adr_LXT971A_PHY_ID2		3	/* PHY Identification Register 2 */
+#define Adr_LXT971A_AN_Advertise	4	/* Auto Negotiation Advertisement Register */
+#define Adr_LXT971A_AN_Link_Ability	5	/* Auto Negotiation Link Partner Base Page Ability Register */
+#define Adr_LXT971A_AN_Expansion	6	/* Auto Negotiation Expansion */
+#define Adr_LXT971A_AN_Next_Page_Txmit	7	/* Auto Negotiation Next Page Transmit Register */
+#define Adr_LXT971A_AN_Link_Next_Page	8	/* Auto Negotiation Link Partner Next Page Receive Register */
+#define Adr_LXT971A_Fast_Control	9	/* Not Implemented */
+#define Adr_LXT971A_Fast_Status		10	/* Not Implemented */
+#define Adr_LXT971A_Extended_Status	15	/* Not Implemented */
+#define Adr_LXT971A_Port_Config		16	/* Configuration Register */
+#define Adr_LXT971A_Status2		17	/* Status Register #2 */
+#define Adr_LXT971A_Interrupt_Enable	18	/* Interrupt Enable Register */
+#define Adr_LXT971A_Interrupt_Status	19	/* Interrupt Status Register */
+#define Adr_LXT971A_LED_Config		20	/* LED Configuration Register */
+#define Adr_LXT971A_Transmit_Control	30	/* Transmit Control Register */

src/arch/armnommu/include/setjmp.h

@@ -0,0 +1,23 @@
+/*
+ *  Copyright (C) 2004 Tobias Lorenz
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#ifndef ETHERBOOT_SETJMP_H
+#define ETHERBOOT_SETJMP_H
+
+#ifndef __ASSEMBLER__
+/* Jump buffer contains v1-v6, sl, fp, sp and pc.  Other registers are not
+   saved.  */
+//typedef int jmp_buf[22];
+typedef int jmp_buf[10];
+#endif
+
+extern int sigsetjmp(jmp_buf __env, int __savemask);
+extern void longjmp(jmp_buf __env, int __val) __attribute__((__noreturn__));
+
+#define setjmp(env) sigsetjmp(env,0)
+
+#endif /* ETHERBOOT_SETJMP_H */

src/arch/armnommu/include/stdint.h

@@ -0,0 +1,23 @@
+/*
+ *  Copyright (C) 2004 Tobias Lorenz
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#ifndef STDINT_H
+#define STDINT_H
+
+typedef unsigned long      size_t;
+
+typedef unsigned char      uint8_t;
+typedef unsigned short     uint16_t;
+typedef unsigned long      uint32_t;
+typedef unsigned long long uint64_t;
+
+typedef signed char        int8_t;
+typedef signed short       int16_t;
+typedef signed int         int32_t;
+typedef signed long long   int64_t;
+
+#endif /* STDINT_H */

src/arch/e1/Config

@@ -0,0 +1,7 @@
+# Config for e1 Etherboot
+#
+
+CFLAGS+= -mgnu-param -DCONSOLE_SERIAL -DCOMCONSOLE=0x01C00000 -DCONSPEED=28800 
+CFLAGS+= -DSIZEINDICATOR  -DNO_DHCP_SUPPORT -DBAR_PROGRESS
+#CFLAGS+= -DEMBEDDED -DMAC_HW_ADDR_DRV="'H','Y','L','N','X','1'"
+

src/arch/e1/Makefile

@@ -0,0 +1,70 @@
+ARCH_FORMAT=	coff-e1
+
+BUILD_ROMS=	$(ROMS)
+BUILD_COFFS=	$(patsubst %img, %coff, $(IMGS))
+SUFFIXES+=	rom zrom coff
+
+CC=		e1-coff-gcc
+AS=		e1-coff-as
+LD=		e1-coff-ld
+SIZE=	e1-coff-size
+AR=		e1-coff-ar
+RANLIB=	e1-coff-ranlib
+OBJCOPY=e1-coff-objcopy
+
+# DMAC_HW_ADDR_DRV holds the ethernet's MAC address. It is passed as
+# flag to the low level driver instead of reading it from an 
+# external EEPROM, which we do not have!
+EXTRA_CFLAGS = -DEMBEDDED -DMAC_HW_ADDR_DRV="'H','Y','L','N','X','1'"
+
+START=	$(BIN)/start.o
+START16=	$(BIN)/start.o
+
+SRCS+=	arch/e1/core/e132_xs.c
+SRCS+=	arch/e1/core/e1_timer.c
+SRCS+=	arch/e1/core/longjmp.c
+SRCS+=	arch/e1/core/memcmp.S
+SRCS+=	arch/e1/core/memcpy.S
+SRCS+=	arch/e1/core/memset.S
+SRCS+=	arch/e1/core/setjmp.c
+SRCS+=	arch/e1/core/strcmp.S
+SRCS+=	arch/e1/core/start.S
+
+ROMLIMIT:=3276800
+
+include		$(BIN)/Roms
+
+ROMS= $(BIN)/cs89x0.rom
+IMGS= $(BIN)/cs89x0.img
+
+#allfiles:	$(BUILD_ROMS)
+all:	$(BUILD_COFFS)
+
+BOBJS+=		$(BIN)/e1_timer.o
+BOBJS+=		$(BIN)/memcmp.o $(BIN)/memcpy.o $(BIN)/memset.o
+BOBJS+=		$(BIN)/setjmp.o $(BIN)/longjmp.o 
+BOBJS+=		$(BIN)/e132_xs.o
+
+# Utilities
+
+$(BIN)/nrv2b:	util/nrv2b.c
+	$(HOST_CC) -O2 -DENCODE -DDECODE -DMAIN -DVERBOSE -DNDEBUG -DBITSIZE=32 -DENDIAN=0 -o $@ $<
+
+# Pattern Rules
+# General for compiling/assembly source files
+$(BIN)/cs89x0.o:	drivers/net/cs89x0.c $(MAKEDEPS)
+	$(CC) $(CFLAGS) $(EXTRA_CFLAGS) -o $@ -c $<
+# With the current tools we have problem with the compilation 
+# of the vsprintf file when the -O2 is selected. So we compile 
+# the aforemntioned file with -O1 !!!
+$(BIN)/vsprintf.o:	core/vsprintf.c $(MAKEDEPS)
+	$(CC) $(CFLAGS) -O1 -o $@ -c $<
+
+$(BIN)/%.o:	arch/e1/core/%.c $(MAKEDEPS)
+	$(CC) $(CFLAGS) -o $@ -c $<
+
+$(BIN)/%.o:	arch/e1/core/%.S $(MAKEDEPS)
+	$(CPP) $(CFLAGS) -D ASSEMBLY $< | $(AS) $(ASFLAGS) -o $@
+
+$(BIN)/%.coff:	$(BIN)/%.tmp $(MAKEDEPS)
+	mv $< $(BIN)/etherboot.coff

src/arch/e1/Makefile.working

@@ -0,0 +1,67 @@
+ARCH_FORMAT=	coff-e1
+
+CC=		e1-coff-gcc
+AS=		e1-coff-as
+LD=		e1-coff-ld
+SIZE=	e1-coff-size
+AR=		e1-coff-ar
+RANLIB=	e1-coff-ranlib
+OBJCOPY=e1-coff-objcopy
+
+EXTRA_CFLAGS = -DEMBEDDED -DMAC_HW_ADDR_DRV="'H','Y','L','N','X','1'"
+
+BUILD_ROMS=	$(ROMS)
+BUILD_COFFS=	$(BIN)/cs89x0.coff
+#BUILD_COFFS=	$(patsubst %img, %coff, $(IMGS))
+
+START=	$(BIN)/start.o
+START16=	$(BIN)/start.o
+
+#SRCS+=	arch/e1/core/coff_loader.c
+SRCS+=	arch/e1/core/e132_xs.c
+SRCS+=	arch/e1/core/e1_timer.c
+SRCS+=	arch/e1/core/longjmp.c
+SRCS+=	arch/e1/core/memcmp.S
+SRCS+=	arch/e1/core/memcpy.S
+SRCS+=	arch/e1/core/memset.S
+SRCS+=	arch/e1/core/setjmp.c
+SRCS+=	arch/e1/core/strcmp.S
+SRCS+=	arch/e1/core/start.S
+
+ROMLIMIT:=3276800
+
+include		$(BIN)/Roms
+
+hyperstone:		$(BUILD_COFFS)
+coff:		$(BUILD_COFFS)
+
+BOBJS+=		$(BIN)/e1_timer.o
+BOBJS+=		$(BIN)/memcmp.o $(BIN)/memcpy.o $(BIN)/memset.o
+BOBJS+=		$(BIN)/setjmp.o $(BIN)/longjmp.o 
+BOBJS+=		$(BIN)/e132_xs.o
+
+# Utilities
+
+$(BIN)/nrv2b:	util/nrv2b.c
+	$(HOST_CC) -O2 -DENCODE -DDECODE -DMAIN -DVERBOSE -DNDEBUG -DBITSIZE=32 -DENDIAN=0 -o $@ $<
+
+# Pattern Rules
+# General for compiling/assembly source files
+
+$(BIN)/cs89x0.o:	drivers/net/cs89x0.c $(MAKEDEPS)
+	$(CC) $(CFLAGS) $(EXTRA_CFLAGS) -o $@ -c $<
+# With the current tools we have problem with the compilation 
+# of the vsprintf file when the -O2 is selected. So we compile 
+# the aforemntioned file with -O1 !!!
+$(BIN)/vsprintf.o:	core/vsprintf.c $(MAKEDEPS)
+	$(CC) $(CFLAGS) -O1 -o $@ -c $<
+
+$(BIN)/%.o:	arch/e1/core/%.c $(MAKEDEPS)
+	$(CC) $(CFLAGS) -o $@ -c $<
+
+$(BIN)/%.o:	arch/e1/core/%.S $(MAKEDEPS)
+	$(CPP) $(CFLAGS) -D ASSEMBLY $< | $(AS) $(ASFLAGS) -o $@
+
+$(BIN)/%.coff:	$(BIN)/%.tmp $(MAKEDEPS)
+	mv $< $(BIN)/etherboot.coff
+

src/arch/e1/README

@@ -0,0 +1,80 @@
+Introduction
+---------------------
+This README file provides guideliness to compile successfully the 
+Etherboot for Hyperstone.
+This directory (src/arch/e1) contains the files that depend on 
+Hyperstone's architecture. The header files that include the 
+configuration of the system are based on Hyperstone's E132-XS 
+development board. The can be very easily modified to support 
+anyother configuration environment. 
+
+Software Perquisites:
+---------------------
+The build environment requires a C compiler for the E1-32XS
+processor. Normally you can simply install a cross-compiling tool
+chain based on the GNU tools (that is binutils and gcc). If you
+are running a Linux system on a x86 CPU then you can just download
+the toolchain as a binary from Hyperstone's official web-site. The
+binary distribution will untar the tools in the /usr/local/e1-coff
+directory. On any other system you will need to build the tool
+chain from the sources.
+
+To compile successfully the following tools should be available:
+ - GNU toolchain:
+    - GCC ver 2.95.2 20030110 (release)         e1-coff-gcc -v
+    - LD ver 2.12.90 20020726                   e1-coff-ld -V
+
+Hardware Perquisites:
+---------------------
+The etherboot has been successfully tested in the E1-32XS
+development board. A second serial device is initialized
+to act as a console. The standard messages
+are redirected to the console. Nevertheless, if one wants not
+to use the serial console he may omit the corresponding switches
+from the Config file located under "src/arch/e1/" directory.
+
+On the E1-32XS board that was  used, a daughter-board was employed
+to connect a second HyIce to the development board. Since the HyIce
+embeds a standard 16550 device, the Etherboot's standard device 
+driver is used.
+
+The position of the jumpers of the development board in order to
+initialize both the second HyIce and the Ethernet device is
+depicted in the following table:
+
+Jumper: Position
+------:--------------
+J3      1-2 (default)
+J4      1-2 (default)
+J13     5-6
+J5      1-2 (default)
+J6      1-2 & 3-4
+J7      3-4
+J9      1-2 (default)
+J10     1-2
+J11     3-4
+
+Compilation
+---------------------
+In order to compile Etherboot for Hyperstone, the following steps should be followed:
+1) Edit the main Makefile (located under "src" directory") and comment-out
+the ARCH variable (by putting a "#" in front of it). Append the following line:
+ARCH:=e1
+2) Edit the Config file (the one located under "src" directory) and make sure that
+the CFLAGS variable will contain *only* the following swithces:
+CFLAGS+=	-DCONFIG_ISA
+CFLAGS+=	-DBOOT_FIRST=BOOT_NIC
+CFLAGS+=	-DALLOW_ONLY_ENCAPSULATED
+CFLAGS+=	-DBACKOFF_LIMIT=7 -DCONGESTED
+CFLAGS+=	-DCOFF_IMAGE 
+CFLAGS+=	-DDOWNLOAD_PROTO_TFTP
+Please note that extra or any other switches may cause failure of compilation!
+3) type "make hyperstone" or "make coff"
+4) the generated file will be located under the "src/bin" directory and will be called :
+ "etherboot.coff". Now you may download it with usual way using e1-coff-gdb ..
+
+Have Fun
+
+Yannis Mitsos, George Thanos
+{gmitsos,gthanos}@telecom.ntua.gr
+

src/arch/e1/core/coff_loader.c

@@ -0,0 +1,176 @@
+/*
+ * Copyright 2003 Yannis Mitsos and George Thanos 
+ * {gmitsos@gthanos}@telecom.ntua.gr
+ * Released under GPL2, see the file COPYING in the top directory
+ * COFF loader is based on the source code of the ELF loader.
+ *
+ */
+#include "coff.h"
+
+#define COFF_DEBUG 0
+
+typedef struct {
+	COFF_filehdr coff32;
+	COFF_opthdr	opthdr32;
+	union {
+		COFF_scnhdr scnhdr32[1];
+		unsigned char dummy[1024];
+	} p;
+	unsigned long curaddr;
+	signed int segment;		/* current segment number, -1 for none */
+	unsigned int loc;		/* start offset of current block */
+	unsigned int skip;		/* padding to be skipped to current segment */
+	unsigned long toread;	/* remaining data to be read in the segment */
+}coff_state;
+
+coff_state cstate;
+
+static sector_t coff32_download(unsigned char *data, unsigned int len, int eof);
+static inline os_download_t coff_probe(unsigned char *data, unsigned int len)
+{
+	unsigned long phdr_size;
+
+	if (len < (sizeof(cstate.coff32)+ sizeof(cstate.opthdr32))) {
+		return 0;
+	}
+	memcpy(&cstate.coff32, data, (sizeof(cstate.coff32)+sizeof(cstate.opthdr32)));
+
+	if ((cstate.coff32.f_magic != EM_E1) ||
+	  	(cstate.opthdr32.magic != O_MAGIC)){
+		return 0;
+	}
+	printf("(COFF");
+	printf(")... \n");
+
+	if (cstate.coff32.f_opthdr == 0){
+		printf("No optional header in COFF file, cannot find the entry point\n");
+                return dead_download;
+	}
+
+	phdr_size = cstate.coff32.f_nscns * sizeof(cstate.p.scnhdr32);
+	if (sizeof(cstate.coff32) +  cstate.coff32.f_opthdr + phdr_size > len) {
+		printf("COFF header outside first block\n");
+                return dead_download;
+	}
+
+	memcpy(&cstate.p.scnhdr32, data + (sizeof(cstate.coff32) +  cstate.coff32.f_opthdr), phdr_size);
+
+	/* Check for Etherboot related limitations.  Memory
+	 * between _text and _end is not allowed.
+	 * Reasons: the Etherboot code/data area.
+	 */
+	for (cstate.segment = 0; cstate.segment < cstate.coff32.f_nscns; cstate.segment++) {
+		unsigned long start, mid, end, istart, iend;
+
+		if ((cstate.p.scnhdr32[cstate.segment].s_flags != S_TYPE_TEXT) && 
+			(cstate.p.scnhdr32[cstate.segment].s_flags != S_TYPE_DATA) && 
+			(cstate.p.scnhdr32[cstate.segment].s_flags != S_TYPE_BSS)){ /* Do we realy need to check the BSS section ? */
+#ifdef COFF_DEBUG
+				printf("Section <%s> in not a loadable section \n",cstate.p.scnhdr32[cstate.segment].s_name);
+#endif
+			continue;
+		}
+
+	start = cstate.p.scnhdr32[cstate.segment].s_paddr;
+	mid = start + cstate.p.scnhdr32[cstate.segment].s_size;
+	end = start + cstate.p.scnhdr32[cstate.segment].s_size;
+
+	/* Do we need the following variables ? */
+	istart = 0x8000;
+	iend = 0x8000;
+
+		if (!prep_segment(start, mid, end, istart, iend)) {
+                	return dead_download;
+		}
+}
+	cstate.segment = -1;
+	cstate.loc = 0;
+	cstate.skip = 0;
+	cstate.toread = 0;
+	return coff32_download;
+}
+
+extern int mach_boot(unsigned long entry_point);
+static sector_t coff32_download(unsigned char *data, unsigned int len, int eof)
+{
+	unsigned long skip_sectors = 0;
+	unsigned int offset;	/* working offset in the current data block */
+	int i;
+
+	offset = 0;
+	do {
+		if (cstate.segment != -1) {
+			if (cstate.skip) {
+				if (cstate.skip >= len - offset) {
+					cstate.skip -= len - offset;
+					break;
+				}
+				offset += cstate.skip;
+				cstate.skip = 0;
+			}
+			
+			if (cstate.toread) {
+				unsigned int cplen;
+				cplen = len - offset;
+				if (cplen >= cstate.toread) {
+					cplen = cstate.toread;
+				}
+				memcpy(phys_to_virt(cstate.curaddr), data+offset, cplen);
+				cstate.curaddr += cplen;
+				cstate.toread -= cplen;
+				offset += cplen;
+				if (cstate.toread)
+					break;
+			}
+		}
+	
+		/* Data left, but current segment finished - look for the next
+		 * segment (in file offset order) that needs to be loaded. 
+		 * We can only seek forward, so select the program headers,
+		 * in the correct order.
+		 */
+		cstate.segment = -1;
+		for (i = 0; i < cstate.coff32.f_nscns; i++) {
+
+			if ((cstate.p.scnhdr32[i].s_flags != S_TYPE_TEXT) && 
+			(cstate.p.scnhdr32[i].s_flags != S_TYPE_DATA))
+				continue;
+			if (cstate.p.scnhdr32[i].s_size == 0)
+				continue;
+			if (cstate.p.scnhdr32[i].s_scnptr < cstate.loc + offset)
+				continue;	/* can't go backwards */
+			if ((cstate.segment != -1) &&
+				(cstate.p.scnhdr32[i].s_scnptr >= cstate.p.scnhdr32[cstate.segment].s_scnptr))
+				continue;	/* search minimum file offset */
+			cstate.segment = i;
+		}
+
+		if (cstate.segment == -1) {
+			/* No more segments to be loaded, so just start the
+			 * kernel.  This saves a lot of network bandwidth if
+			 * debug info is in the kernel but not loaded.  */
+			goto coff_startkernel;
+			break;
+		}
+		cstate.curaddr = cstate.p.scnhdr32[cstate.segment].s_paddr;
+		cstate.skip	   = cstate.p.scnhdr32[cstate.segment].s_scnptr - (cstate.loc + offset);
+		cstate.toread  = cstate.p.scnhdr32[cstate.segment].s_size;
+#if COFF_DEBUG
+		printf("PHDR %d, size %#lX, curaddr %#lX\n",
+			cstate.segment, cstate.toread, cstate.curaddr);
+#endif
+	} while (offset < len);
+
+	cstate.loc += len + (cstate.skip & ~0x1ff);
+	skip_sectors = cstate.skip >> 9;
+	cstate.skip &= 0x1ff;
+	
+	if (eof) {
+		unsigned long entry;
+coff_startkernel:
+		entry = cstate.opthdr32.entry;
+		done();
+		mach_boot(entry);
+	}
+	return skip_sectors;
+}

src/arch/e1/core/e132_xs.c

@@ -0,0 +1,70 @@
+/*
+ * Copyright 2003 Yannis Mitsos and George Thanos 
+ * {gmitsos@gthanos}@telecom.ntua.gr
+ * Released under GPL2, see the file COPYING in the top directory
+ *
+ */
+#include "hooks.h"
+#include "io.h"
+#include "etherboot.h"
+#include "e132_xs_board.h"
+
+unsigned int io_periph[NR_CS] = {[0 ... NR_CS-1] = 0 };
+
+/*
+void arch_main(struct Elf_Bhdr *ptr __unused)
+{
+
+}
+*/
+
+void init_peripherals(void)
+{
+	int i;
+
+	for(i=0; i< NR_CS; i++){
+		io_periph[i]= (SLOW_IO_ACCESS | i << 22);
+	}
+
+	io_periph[ETHERNET_CS] = (io_periph[ETHERNET_CS] | 1 << IOWait);
+
+	asm volatile("
+			ori SR, 0x20
+			movi FCR, 0x66ffFFFF"
+			:
+			:);
+}
+
+struct meminfo meminfo;
+void get_memsizes(void)
+{
+/* We initialize the meminfo structure 
+ * according to our development board's specs
+ * We do not have a way to automatically probe the 
+ * memspace instead we initialize it manually
+ */
+	meminfo.basememsize = BASEMEM;
+	meminfo.memsize = 	SDRAM_SIZE;
+	meminfo.map_count = NR_MEMORY_REGNS;
+
+	meminfo.map[0].addr = SDRAM_BASEMEM;
+	meminfo.map[0].size = SDRAM_SIZE;
+	meminfo.map[0].type = E820_RAM;
+	meminfo.map[1].addr = SRAM_BASEMEM;
+	meminfo.map[1].size = SRAM_SIZE;
+	meminfo.map[1].type = E820_RAM;
+	meminfo.map[2].addr = IRAM_BASEMEM;
+	meminfo.map[2].size = IRAM_SIZE;
+	meminfo.map[2].type = E820_RAM;
+}
+
+int mach_boot(register unsigned long entry_point)
+{
+		asm volatile(
+					 "mov PC, %0"
+					: /* no outputs */
+					: "l" (entry_point) );
+		return 0; /* We should never reach this point ! */
+
+}
+

src/arch/e1/core/e1_timer.c

@@ -0,0 +1,94 @@
+/*
+ * Copyright 2003 Yannis Mitsos and George Thanos 
+ * {gmitsos@gthanos}@telecom.ntua.gr
+ * Released under GPL2, see the file COPYING in the top directory
+ *
+ */
+#include "etherboot.h"
+#include "timer.h"
+#include "e132_xs_board.h"
+
+/* get timer returns the contents of the timer */
+static inline unsigned long get_timer(void)
+{
+	unsigned long result;
+	__asm__ __volatile__("
+					ORI	SR, 0x20
+					mov	%0, TR" 
+					: "=l"(result));
+	return result;
+}
+
+/* ------ Calibrate the TSC ------- 
+ * Time how long it takes to excute a loop that runs in known time.
+ * And find the convertion needed to get to CLOCK_TICK_RATE
+ */
+
+static unsigned long configure_timer(void)
+{
+	unsigned long TPR_value; /* Timer Prescalar Value */
+
+	TPR_value = 0x000C00000;
+	
+	asm volatile (" 
+				FETCH   4	
+				ORI		SR, 0x20
+				MOV		TPR, %0
+				ORI		SR, 0x20
+				MOVI	TR, 0x0"
+				: /* no outputs */
+				: "l" (TPR_value)
+				); 
+
+	printf("The time prescaler register is set to: <%#x>\n",TPR_value);
+	return (1);
+}
+
+static unsigned long clocks_per_tick;
+
+void setup_timers(void)
+{
+	if (!clocks_per_tick) {
+		clocks_per_tick = configure_timer();
+	}
+}
+
+unsigned long currticks(void)
+{
+	return get_timer()/clocks_per_tick;
+}
+
+static unsigned long timer_timeout;
+static int __timer_running(void)
+{
+	return get_timer() < timer_timeout;
+}
+
+void udelay(unsigned int usecs)
+{
+	unsigned long now;
+	now = get_timer();
+	timer_timeout = now + usecs * ((clocks_per_tick * TICKS_PER_SEC)/(1000*1000));
+	while(__timer_running());
+}
+void ndelay(unsigned int nsecs)
+{
+	unsigned long now;
+	now = get_timer();
+	timer_timeout = now + nsecs * ((clocks_per_tick * TICKS_PER_SEC)/(1000*1000*1000));
+	while(__timer_running());
+}
+
+void load_timer2(unsigned int timer2_ticks)
+{
+	unsigned long now;
+	unsigned long clocks;
+	now = get_timer();
+	clocks = timer2_ticks * ((clocks_per_tick * TICKS_PER_SEC)/CLOCK_TICK_RATE);
+	timer_timeout = now + clocks;
+}
+
+int timer2_running(void)
+{
+	return __timer_running();
+}

src/arch/e1/core/etherboot.lds

@@ -0,0 +1,126 @@
+/* Default linker script, for normal executables */
+OUTPUT_FORMAT("coff-e1-big")
+MEMORY
+	{
+        romvec   : ORIGIN = 0x2000000,  LENGTH = 0x0000400
+        flash    : ORIGIN = 0xE0000000, LENGTH = 0x20000 
+        eflash   : ORIGIN = 0x2200000,  LENGTH = 0x20000
+        ram      : ORIGIN = 0x00000000, LENGTH = 0x1000000
+        eram16MB : ORIGIN = 0x01000000, LENGTH = 0
+        sram     : ORIGIN = 0x40000000, LENGTH = 0x40000
+		iram     : ORIGIN = 0xC0000000, LENGTH = 0x4000
+	}
+
+SEARCH_DIR("/usr/local/e1-coff/lib");
+ENTRY(Main)
+MEM0start   = 0x00000000;
+MEM0size    = 0x40000000;
+MEM1start   = 0x40000000;
+MEM1size    = 0x40000000;
+MEM2start   = 0x80000000;
+MEM2size    = 0x40000000;
+IRAMstart   = 0xC0000000;
+IRAMsize    = 0x20000000;
+MEM3start   = 0xE0000000;
+MEM3size    = 0x20000000;
+Stack1Reserve = 560;
+_Stack1Size = DEFINED(_Stack1Size)? _Stack1Size : 1*1024;
+_Stack2Size = DEFINED(_Stack2Size)? _Stack2Size : 16*1024;
+_Stack1Base = DEFINED(_Stack1Base)? _Stack1Base : __bss_end__;
+_Stack2Base = DEFINED(_Stack2Base)? _Stack2Base : __bss_end__ + _Stack1Size + Stack1Reserve;
+_Mem0HeapBase = DEFINED(_Mem0HeapBase)? _Mem0HeapBase : _Stack2Base + _Stack2Size;
+_Mem1HeapBase = DEFINED(_Mem1HeapBase)? _Mem1HeapBase : 0;
+Priority    = DEFINED(Priority)   ? Priority    : 31;
+TextBase    = DEFINED(TextBase)   ? TextBase    : 0xa00000;
+SECTIONS
+{
+  .G6 (DEFINED(G6Base) ? G6Base : 0x9000) : {
+    *(.G6)
+  }
+  .G7 (DEFINED(G7Base) ? G7Base : 0x40001000) : {
+    *(.G7)
+  }
+  .G8 (DEFINED(G8Base) ? G8Base : 0xC0000000) : {
+    *(.G8)
+  }
+  .G9 (DEFINED(G9Base) ? G9Base : 0) : {
+    *(.G9)
+  }
+  .G10 (DEFINED(G10Base) ? G10Base : 0) : {
+    *(.G10)
+  }
+  .G11 (DEFINED(G11Base) ? G11Base : 0) : {
+    *(.G11)
+  }
+  .G12 (DEFINED(G12Base) ? G12Base : 0) : {
+    *(.G12)
+  }
+  .G13 (DEFINED(G13Base) ? G13Base : 0) : {
+    *(.G13)
+  }
+
+  .text TextBase : {
+	__virt_start = .;
+	__text = . ;
+    *(.text)
+    *(.fini)
+ 		. = ALIGN(16);
+		_isa_drivers = . ;
+		*(.drivisa);
+		_isa_drivers_end = . ;
+		. = ALIGN(16);
+
+   *(.init)
+  _etext  =  . ;
+    /* _init = DEFINED(_init) ? _init : 0; */
+    /* _fini = DEFINED(_fini) ? _fini : 0; */
+    /* _argc = DEFINED(_argc) ? _argc : 0; */
+    /* _argv = DEFINED(_argv) ? _argv : 0; */
+    /* _envp = DEFINED(_envp) ? _envp : 0; */
+    /* _hwinit = DEFINED(_hwinit) ? _hwinit : 0; */
+    /* _atexit = DEFINED(_atexit) ? _atexit : 0; */
+    G6Size = SIZEOF(.G6);
+    G7Size = SIZEOF(.G7);
+    G8Size = SIZEOF(.G8);
+    G9Size = SIZEOF(.G9);
+    G10Size = SIZEOF(.G10);
+    G11Size = SIZEOF(.G11);
+    G12Size = SIZEOF(.G12);
+    G13Size = SIZEOF(.G13);
+
+}
+
+  .data  SIZEOF(.text) + ADDR(.text) : {
+    *(.data)
+     _edata  =  . ;
+}
+
+  .bss  SIZEOF(.data) + ADDR(.data) : 
+  {
+     __bss_start__ = ALIGN( 0x10 ) ;
+     __bss = . ;
+    *(.bss)
+    *(COMMON)
+     __end = . ;
+     __bss_end__ = ALIGN( 0x10 ) ;
+     __ebss = . ;
+  }
+
+.eram16MB :    
+	{
+	___ramend = . - 0x7000;
+	} > eram16MB
+
+  .stab  0 (NOLOAD) :
+  {
+    [ .stab ]
+  }
+  .stabstr  0 (NOLOAD) :
+  {
+    [ .stabstr ]
+  }
+  _GOT_  0 (NOLOAD) :
+  {
+    [ _GOT_ ]
+  }
+}

src/arch/e1/core/longjmp.c

@@ -0,0 +1,35 @@
+/*
+ * Copyright 2003 Yannis Mitsos and George Thanos 
+ * {gmitsos@gthanos}@telecom.ntua.gr
+ * Released under GPL2, see the file COPYING in the top directory
+ *
+ */
+#include "setjmp.h"
+
+unsigned long jmpbuf_ptr;
+
+void longjmp(jmp_buf state, int value )
+{
+	if(!value)
+		state->__jmpbuf->ReturnValue = 1;
+	else
+		state->__jmpbuf->ReturnValue = value;
+
+	jmpbuf_ptr = (unsigned long)state; 
+
+#define _state_ ((struct __jmp_buf_tag*)jmpbuf_ptr)
+	asm volatile("mov L0, %0\n\t"
+		     "mov L1, %1\n\t"
+		     "mov L2, %2\n\t"
+		     "mov G3, %3\n\t"
+		     "mov G4, %4\n\t"
+		     "ret PC, L1\n\t"
+		     :/*no output*/
+		     :"l"(_state_->__jmpbuf->ReturnValue),
+		      "l"(_state_->__jmpbuf->SavedPC),
+		      "l"(_state_->__jmpbuf->SavedSR),
+		      "l"(_state_->__jmpbuf->G3),
+		      "l"(_state_->__jmpbuf->G4)
+		     :"%G3", "%G4", "%L0", "%L1" );
+#undef _state_
+}

src/arch/e1/core/memcmp.S

@@ -0,0 +1,54 @@
+/*
+ * Derived from the Hyperstone's library source code.
+ * Modefied src in order to apply the -mgnu-param compiler option.
+ * Copyright (C) 2002-2003 GDT, Yannis Mitsos <gmitsos@telecom.ntua.gr>
+ *                              George Thanos <gthanos@telecom.ntua.gr>
+ */
+	    .text
+	    .align   2
+	    .global  _memcmp
+
+;ENTRY (_memcmp)
+_memcmp:
+	    FRAME    L9, L3                 # get incoming parameters
+	    CMPBI    L2,3                   # check word alignment
+	    BNZ      byte_compare
+	    CMPBI    L1,3                   # check word alignment
+	    BNZ      byte_compare
+
+double_compare:
+	    ADDI     L0, -8
+	    BLT      is_equal
+	    LDD.P    L1, L5
+	    LDD.P    L2, L7
+	    SUB      L5, L7
+	    DBNZ     corr_8
+	    SUB      L6, L8
+	    BZ       double_compare
+	    ADDI     L0, 4
+	    ADDI     L2, -4
+	    ADDI     L1, -4
+	    BR       byte_compare
+
+corr_8:     ADDI     L0, 8
+	    ADDI     L2, -8
+	    ADDI     L1, -8
+byte_compare:
+	    ADDI     L0, -1
+	    BLT      equal
+	    LDBU.N   L2, L5, 1              # Load and compare bytes
+	    LDBU.N   L1, L6, 1
+	    SUB      L5, L6
+	    BZ       byte_compare
+	    MOV      L2, L5
+	    RET      PC, L3
+
+is_equal:   CMPI     L0, -8
+	    DBNE     byte_compare
+	    ADDI     L0, 8
+equal:
+	    MOVI     L2, 0
+	    RET      PC, L3
+
+	    .END
+

src/arch/e1/core/memcpy.S

@@ -0,0 +1,79 @@
+/*
+ * Derived from the Hyperstone's library source code.
+ * Modefied src in order to apply the -mgnu-param compiler option.
+ * Copyright (C) 2002-2003 GDT, Yannis Mitsos <gmitsos@telecom.ntua.gr>
+ *                              George Thanos <gthanos@telecom.ntua.gr>
+ */    
+		.text
+	    .align   2
+	    .global  _memcpy
+;ENTRY(_memcpy)
+_memcpy:
+	    FRAME    L8, L3
+	    MOV      L7, L2                 # Save for return
+
+#*****************************
+# Perform byte copy if both
+# not on a word alignment
+#*****************************
+	    CMPBI    L2, 3                  # check word alignment
+            BNZ      mem_except
+	    CMPBI    L1, 3                  # check word alignment
+            BNZ      mem_except
+
+#*****************************
+# Copy Double,Word,Halfword,
+# then byte
+#*****************************
+DBL_LOOP:
+	    CMPI     L0, 8                  # Copy Doubles
+            BLT      DO_WORD
+            LDD.P    L1, L5
+            ADDI     L0, -8
+            DBR      DBL_LOOP
+            STD.P    L2, L5
+
+DO_WORD:
+	   CMPI      L0, 4                  # Copy leftover word
+           BLT       DO_HALF
+	   LDW.P     L1, L5
+           ADDI      L0, -4
+	   DBZ       DONE                   # Done if L0 is 0
+           STW.P     L2, L5
+
+DO_HALF:
+	   CMPI      L0, 2                  # Copy leftover byte
+           BLT       DO_BYTE
+           LDHU.N    L1, L5, 2
+           ADDI      L0, -2
+	   DBZ       DONE                   # Done if L0 is 0
+           STHU.N    L2, L5, 2
+
+DO_BYTE:
+	   CMPI      L0, 1                  # Copy leftover byte
+           BLT       DONE
+           LDBU.D    L1, L5, 0
+           STBU.D    L2, L5, 0
+
+DONE:                                       # Copy done
+	   MOV       L2, L7                 # Return pointer
+           RET       PC, L3
+
+#****************************
+# Byte memcpy
+#****************************
+mem_except:
+           DBR       L_5
+           MOVI      L6,0
+L_3:
+	   LDBS.D    L1, L5, 0              # Transfer the byte
+           ADDI      L6, 1
+           STBS.D    L2, L5, 0
+           ADDI      L2, 1
+           ADDI      L1, 1
+L_5:                                        # Loop test
+           CMP       L6, L0
+	   BST       L_3
+	   MOV       L2, L7                 # Return pointer
+	   RET       PC, L3
+	   .END

src/arch/e1/core/memset.S

@@ -0,0 +1,47 @@
+/*
+ * Derived from the Hyperstone's library source code.
+ * Modefied src in order to apply the -mgnu-param compiler option.
+ * Copyright (C) 2002-2003 GDT, Yannis Mitsos <gmitsos@telecom.ntua.gr>
+ *                              George Thanos <gthanos@telecom.ntua.gr>
+ */      
+		.text
+	    .align   2
+		.global _memset
+
+;ENTRY(_memset)
+_memset:    FRAME    L9, L3
+	    MASK     L5, L1, 0xFF
+	    MOV      L8, L2
+	    CMPI     L0, 0                  # if n = 0 then return
+            BE       retour
+
+loop0:      CMPBI    L8, 0x3
+	    BZ       word_bound
+	    ADDI     L0, -1
+	    DBNZ     loop0
+	    STBU.N   L8, L5, 1
+retour:     RET      PC, L3
+
+word_bound:
+	    CMPI     L0, 8
+	    DBLT     loop2
+	    MOV      L7, L5
+	    SHLI     L7, 8
+            OR       L5, L7
+            MOV      L7, L5
+            SHLI     L7, 16
+            OR       L5, L7
+            MOV      L6, L5
+loop1:      ADDI     L0, -8
+            CMPI     L0, 8
+            DBGE     loop1
+	    STD.P    L8, L5
+            CMPI     L0, 0
+            DBNZ     loop2
+	    ANDNI    L5, ~ 0xFF
+	    RET      PC, L3
+
+loop2:      ADDI     L0, -1
+	    DBNZ     loop2
+	    STBU.N   L8, L5, 1
+	    RET      PC, L3

src/arch/e1/core/setjmp.c

@@ -0,0 +1,26 @@
+/*
+ * Copyright 2003 Yannis Mitsos and George Thanos 
+ * {gmitsos@gthanos}@telecom.ntua.gr
+ * Released under GPL2, see the file COPYING in the top directory
+ *
+ */
+#include "setjmp.h"
+
+int setjmp( jmp_buf state)
+{
+	asm volatile(	"mov %0, G3\n\t"           
+			"mov %1, G4\n\t" 
+			:"=l"(state->__jmpbuf->G3), 
+			 "=l"(state->__jmpbuf->G4) 
+			:/*no input*/ 
+			:"%G3", "%G4" );
+
+	asm volatile(   "setadr  %0\n\t"
+			"mov %1, L1\n\t"
+			"mov %2, L2\n\t"
+			:"=l"(state->__jmpbuf->SavedSP),
+			 "=l"(state->__jmpbuf->SavedPC),
+			 "=l"(state->__jmpbuf->SavedSR)
+			:/*no input*/);
+	return 0;
+}

src/arch/e1/core/start.S

@@ -0,0 +1,111 @@
+/*
+ * Derived from the Hyperstone's library source code.
+ * Copyright (C) 2002-2003 GDT, Yannis Mitsos <gmitsos@telecom.ntua.gr>
+ *                              George Thanos <gthanos@telecom.ntua.gr>
+ */    
+	.global	Priority	; Task-Priority
+	.global	_Stack1Size	; Size of hardware stack
+	.global	_Stack2Size	; Size of aggregate stack
+	.global	_Stack1Base	; Base of hardware stack
+	.global	_Stack2Base	; Base of aggregate stack
+	.global	_Mem0HeapBase	; Base of Heap in Mem0
+	.global	_Mem1HeapBase	; Base of Heap in Mem1
+
+	.global _init_peripherals
+	.global	_main
+	.global	Main
+
+	.global	__exit
+	.global	__fmode
+	.global	__MaxArgCount
+
+	.text
+BasePtrs:
+	.weak	G6Base,G7Base,G8Base,G9Base,G10Base,G11Base,G12Base,G13Base
+	.long	G6Base,G7Base,G8Base,G9Base,G10Base,G11Base,G12Base,G13Base
+BasePtrsEnd:
+HeapPtrs:
+	.long	_Mem0HeapBase
+	.long	_Mem1HeapBase
+HeapPtrsEnd:
+
+__MaxArgCount:
+        .long	32
+__StandAloneMode:
+	.long	0		; 0 indicate stand alone mode
+
+;============================================================================;
+; Startup-Code                                                               ;
+;============================================================================;
+	.data
+
+; do not change the order of: __argc,..
+__argc:	
+	.long  0
+__argv:	
+	.long  0
+__IsShell:
+	.long  1
+ErrorLevel:
+	.long  0
+__lab:	
+	.long  0
+__fmode:
+	.long	0x4000		; O_TEXT    attribute
+
+_isa_drivers:
+	.long	0
+
+_isa_drivers_end:
+	.long	0
+
+
+	.text
+Main:
+StartUp:
+	FRAME	L5, L0
+	MOVI	L2, __bss_start__	; clear the .bss segment
+0:	CMPI	L2, __bss_end__
+	BHE	0f
+	DBR	0b
+	STW.P	L2, 0
+0:	SUM     L2, PC, BasePtrs-$	; Load BasePtrs G6-G13
+	LDD.P   L2, G6
+	LDD.P   L2, G8
+;	LDD.P   L2, G10
+	LDD.P   L2, G12
+	MOVI    L2, 1
+	SUM     L3, PC, __StandAloneMode-$
+	STW.R   L3, L2
+
+	;----------------------------------------------------------------;
+	; Call main C function                                           ;
+	;----------------------------------------------------------------;
+2:	LDW.D	PC, L2, __argc - $	; pass count of arguments to main
+	LDW.D   PC, L3, __argv - $	; pass pointer array to main
+	CALL	L4, PC, _init_peripherals - $ 
+	CALL    L4, PC, _main - $	; --> Call Main-Program
+	CHK     PC, PC			; trap if execution arrives here
+
+__exit:
+	FRAME   L5, L1
+	STW.D   PC, L0, ErrorLevel - $	; Store ERRORLEVEL
+
+	CHK     PC, PC
+	RET     PC, L1
+
+	.global ___main
+___main:
+	FRAME   L4, L1
+	MOVI    L3, 2
+	STW.D   PC, L3, __StandAloneMode-$
+	RET     PC, L1		; does not return
+
+	.section _GOT_
+	.long   Main+4		; OnCreate
+	.long   Main+8		; OnError
+	.long   BasePtrs	; G6
+	.long   BasePtrs+4	; G7
+	.long   BasePtrs+8	; G8
+	
+	.END

src/arch/e1/core/strcmp.S

@@ -0,0 +1,76 @@
+/*
+ * Derived from the Hyperstone's library source code.
+ * Modefied src in order to apply the -mgnu-param compiler option.
+ * Copyright (C) 2002-2003 GDT, Yannis Mitsos <gmitsos@telecom.ntua.gr>
+ *                              George Thanos <gthanos@telecom.ntua.gr>
+ */
+        .text
+	    .align   2
+	    .global  _strcmp
+;ENTRY(_strcmp)
+_strcmp:
+	    FRAME    L8,L2
+	    CMPBI    L1, 3        # check for word alignment
+	    BNZ      str_except
+	    CMPBI    L0, 3        # check for word alignment
+	    BNZ      str_except
+
+start:
+	    LDD.P    L1, L4       # post inc mode
+	    LDD.P    L0, L6       # post inc mode
+            CMPBI    L4, ANYBZ
+            BE       correct1
+            CMP      L4, L6
+            BNE      correct1
+            CMP      L5, L7
+            BNE      correct
+            CMPBI    L5, ANYBZ
+            BE       correct
+            CMPBI    L6, ANYBZ
+            BE       correct1
+            CMPBI    L7, ANYBZ
+            BNE      start
+
+correct:    MASK     L4, L5, 0xff000000
+	    MASK     L6, L7, 0xff000000
+            CMP      L4, L6
+            BNE      Exit
+            SHLI     L5, 8
+            CMPI     L4, 0
+            DBNE     correct
+            SHLI     L7, 8
+            MOV      L1, L4
+            RET      PC, L2
+
+Exit:       SUB      L4, L6       # Subtract chars
+            SARI     L4, 24
+            MOV      L1, L4
+            RET      PC, L2
+
+correct1:   MASK     L5, L4, 0xff000000
+	    MASK     L7, L6, 0xff000000
+            CMP      L5, L7
+            BNE      Exit1
+            SHLI     L4, 8
+            CMPI     L5, 0
+            DBNE     correct1
+            SHLI     L6, 8
+            MOV      L1, L5
+            RET      PC, L2
+Exit1:      SUB      L5, L7       # Subtract chars
+            SARI     L5, 24
+            MOV      L1, L5
+            RET      PC, L2
+
+testzero:   CMPI     L4, 0
+            BE       L_5
+str_except:
+	    LDBU.N   L1, L4, 1    # Load *s1, compare bytes
+	    LDBU.N   L0, L5, 1    # Load *s2, compare bytes
+            CMP      L4, L5
+	    BE       testzero
+	    SUB      L4, L5       # Subtract chars
+L_5:        MOV      L1, L4
+	    RET      PC, L2
+	    .END
+

src/arch/e1/include/bits/byteswap.h

@@ -0,0 +1,39 @@
+#ifndef ETHERBOOT_BITS_BYTESWAP_H
+#define ETHERBOOT_BITS_BYTESWAP_H
+
+/* We do not have byte swap functions ... We are
+ * RISC processor ...
+ */
+
+static inline unsigned short __swap16(volatile unsigned short v)
+{
+    return ((v << 8) | (v >> 8));
+}
+
+static inline unsigned int __swap32(volatile unsigned long v)
+{
+    return ((v << 24) | ((v & 0xff00) << 8) | ((v & 0xff0000) >> 8) | (v >> 24));
+}
+
+#define __bswap_constant_16(x) \
+	((uint16_t)((((uint16_t)(x) & 0x00ff) << 8) | \
+		(((uint16_t)(x) & 0xff00) >> 8)))
+
+#define __bswap_constant_32(x) \
+	((uint32_t)((((uint32_t)(x) & 0x000000ffU) << 24) | \
+		(((uint32_t)(x) & 0x0000ff00U) <<  8) | \
+		(((uint32_t)(x) & 0x00ff0000U) >>  8) | \
+		(((uint32_t)(x) & 0xff000000U) >> 24)))
+
+#define __bswap_16(x) \
+	(__builtin_constant_p(x) ? \
+	__bswap_constant_16(x) : \
+	__swap16(x))
+
+
+#define __bswap_32(x) \
+	(__builtin_constant_p(x) ? \
+	__bswap_constant_32(x) : \
+	__swap32(x))
+
+#endif /* ETHERBOOT_BITS_BYTESWAP_H */

src/arch/e1/include/bits/cpu.h

@@ -0,0 +1,6 @@
+#ifndef E1_BITS_CPU_H
+#define E1_BITS_CPU_H
+
+#define cpu_setup() do {} while(0)
+
+#endif /* E1_BITS_CPU_H */

src/arch/e1/include/bits/elf.h

@@ -0,0 +1,6 @@
+#ifndef E1_BITS_ELF_H
+#define E1_BITS_ELF_H
+
+/* dummy file, needed for the compilation of core/nic.c */
+
+#endif /* E1_BITS_ELF_H */

src/arch/e1/include/bits/endian.h

@@ -0,0 +1,6 @@
+#ifndef ETHERBOOT_BITS_ENDIAN_H
+#define ETHERBOOT_BITS_ENDIAN_H
+
+#define __BYTE_ORDER __BIG_ENDIAN
+
+#endif /* ETHERBOOT_BITS_ENDIAN_H */

src/arch/e1/include/bits/string.h

@@ -0,0 +1,35 @@
+#ifndef ETHERBOOT_BITS_STRING_H
+#define ETHERBOOT_BITS_STRING_H
+
+/* define inline optimized string functions here */
+
+#define	__HAVE_ARCH_MEMCPY
+//extern void * memcpy(const void *d, const void *s, size_t count);
+
+#define __HAVE_ARCH_MEMCMP
+//extern int memcmp(const void * s ,const void * d ,size_t );
+
+#define __HAVE_ARCH_MEMSET
+//extern void * memset(const void * s, int c, size_t count);
+
+#define __HAVE_ARCH_MEMMOVE
+static inline void *memmove(void *s1, const void *s2, size_t n) {
+
+   unsigned int i;
+   char *tmp = s1;
+   char *cs2 = (char *) s2;
+
+   if (tmp < cs2) {
+      for(i=0; i<n; ++i, ++tmp, ++cs2)
+           *tmp =  *cs2;
+   }
+   else {
+      tmp += n - 1;
+      cs2 += n - 1;
+      for(i=0; i<n; ++i, --tmp, --cs2)
+              *tmp = *cs2;
+   }
+   return(s1);
+}
+
+#endif /* ETHERBOOT_BITS_STRING_H */

src/arch/e1/include/e132_xs_board.h

@@ -0,0 +1,22 @@
+#ifndef __E132_XS_BOARD_H
+#define __E132_XS_BOARD_H
+
+#define CONFIG_HYPERSTONE_OSC_FREQ_MHZ	15
+
+#define NR_MEMORY_REGNS	3
+#define BASEMEM			0x0
+
+/* SDRAM mapping */
+#define	SDRAM_SIZE		0x01000000
+#define SDRAM_BASEMEM	BASEMEM	
+
+/* SRAM mapping */
+#define SRAM_BASEMEM	0x40000000
+#define SRAM_SIZE		0x0003FFFF
+
+/* IRAM mapping */
+#define IRAM_BASEMEM	0xC0000000
+#define IRAM_SIZE		0x00003FFF
+
+
+#endif /* __E132_XS_BOARD_H */

src/arch/e1/include/hooks.h

@@ -0,0 +1,9 @@
+#ifndef ETHERBOOT_E1_HOOKS_H
+#define ETHERBOOT_E1_HOOKS_H
+
+#define arch_main(data,params) do {} while(0)
+#define arch_on_exit(status) do {} while(0)
+#define arch_relocate_to(addr) do {} while(0)
+#define arch_relocated_from(old_addr) do {} while(0)
+
+#endif /* ETHERBOOT_E1_HOOKS_H */

src/arch/e1/include/io.h

@@ -0,0 +1,210 @@
+#ifndef ETHERBOOT_IO_H
+#define ETHERBOOT_IO_H
+
+/* Don't require identity mapped physical memory,
+ * osloader.c is the only valid user at the moment.
+ */
+#if 0
+static inline unsigned long virt_to_phys(volatile const void *virt_addr)
+{
+	return ((unsigned long)virt_addr);
+}
+#else
+#define virt_to_phys(vaddr)	((unsigned long) (vaddr))
+#endif
+
+#if 0
+static inline void *phys_to_virt(unsigned long phys_addr)
+{
+	return (void *)(phys_addr);
+}
+#else
+#define phys_to_virt(vaddr)	((void *) (vaddr))
+#endif
+
+/* virt_to_bus converts an addresss inside of etherboot [_start, _end]
+ * into a memory address cards can use.
+ */
+#define virt_to_bus virt_to_phys
+
+/* bus_to_virt reverses virt_to_bus, the address must be output
+ * from virt_to_bus to be valid.  This function does not work on
+ * all bus addresses.
+ */
+#define bus_to_virt phys_to_virt
+
+#define iounmap(addr)				((void)0)
+#define ioremap(physaddr, size)			(physaddr)
+
+#define IORegAddress	13
+#define IOWait			11
+#define IOSetupTime		8
+#define IOAccessTime	5
+#define IOHoldTime		3
+
+#define SLOW_IO_ACCESS ( 0x3 << IOSetupTime | 0x0 << IOWait | 7 << IOAccessTime | 3 << IOHoldTime )
+
+/* The development board can generate up to 15 Chip selects */
+#define NR_CS	16
+
+extern unsigned int io_periph[NR_CS];
+#define ETHERNET_CS 4
+
+static inline unsigned short _swapw(volatile unsigned short v)
+{
+    return ((v << 8) | (v >> 8));
+}
+
+static inline unsigned int _swapl(volatile unsigned long v)
+{
+    return ((v << 24) | ((v & 0xff00) << 8) | ((v & 0xff0000) >> 8) | (v >> 24));
+}
+
+#define  hy_inpw(addr)  						\
+	({   register unsigned long dummy, dummy1; 			\
+		 dummy  = addr; 						\
+		 asm volatile  ("LDW.IOD   %1, %0, 0" 	\
+				 		: "=l" (dummy1) 		\
+					   	: "l" (dummy)); dummy1; })
+
+
+#define  hy_outpw(x, addr)						  \
+	({   register unsigned long dummy0,dummy1; \
+	 	 dummy0 = addr; 					  \
+		 dummy1 = x;						  \
+		 asm volatile  ("STW.IOD   %1, %0, 0" \
+						 : "=l" (dummy1) 	  \
+						 : "l"(dummy0), "l" (dummy1)); dummy1; })
+
+#define readb(addr)	({ unsigned char  __v = inregb(addr); __v; })
+#define readw(addr)	({ unsigned short __v = inregw(addr); __v; })
+#define readl(addr)	({ unsigned long  __v = inregl(addr); __v; })
+
+#define writeb(b,addr) (void)(outreg(b, addr))
+#define writew(b,addr) (void)(outreg(b, addr))
+#define writel(b,addr) (void)(outreg(b, addr))
+
+static inline unsigned long common_io_access(unsigned long addr)
+{
+	return io_periph[(addr & 0x03C00000) >> 22];
+}
+
+static inline volatile unsigned char inregb(volatile unsigned long reg)
+{
+	unsigned char val;
+
+	val = hy_inpw(common_io_access(reg) | ((0xf & reg) << IORegAddress)); 
+	return val;
+}
+
+static inline volatile unsigned short inregw(volatile unsigned long reg)
+{
+	unsigned short val;
+
+	val = hy_inpw(common_io_access(reg) | ((0xf & reg) << IORegAddress)); 
+	return val;
+}
+
+static inline volatile unsigned long inregl(volatile unsigned long reg)
+{
+	unsigned long val;
+
+	val = hy_inpw(common_io_access(reg) | ((0xf & reg) << IORegAddress)); 
+	return val;
+}
+
+static inline void outreg(volatile unsigned long val, volatile unsigned long reg)
+{
+		
+	hy_outpw(val, (common_io_access(reg) | ((0xf & reg) << IORegAddress)));
+}
+
+static inline void io_outsb(unsigned int addr, void *buf, int len)
+{
+	unsigned long tmp;
+	unsigned char *bp = (unsigned char *) buf;
+
+	tmp = (common_io_access(addr)) | ((0xf & addr) << IORegAddress);
+
+	while (len--){
+		hy_outpw(_swapw(*bp++), tmp);
+	}
+}
+
+static inline void io_outsw(volatile unsigned int addr, void *buf, int len)
+{
+	unsigned long tmp;
+	unsigned short *bp = (unsigned short *) buf;
+	
+	tmp = (common_io_access(addr)) | ((0xf & addr) << IORegAddress);
+
+	while (len--){
+		hy_outpw(_swapw(*bp++), tmp);
+	}
+}
+
+static inline void io_outsl(volatile unsigned int addr, void *buf, int len)
+{
+	unsigned long tmp;
+	unsigned int *bp = (unsigned int *) buf;
+		
+	tmp = (common_io_access(addr)) | ((0xf & addr) << IORegAddress);
+
+	while (len--){
+		hy_outpw(_swapl(*bp++), tmp);
+	}
+}
+
+static inline void io_insb(volatile unsigned int addr, void *buf, int len)
+{
+	unsigned long tmp;
+	unsigned char *bp = (unsigned char *) buf;
+
+	tmp = (common_io_access(addr)) | ((0xf & addr) << IORegAddress);
+
+	while (len--)
+		*bp++ = hy_inpw((unsigned char) tmp);
+	
+}
+
+static inline void io_insw(unsigned int addr, void *buf, int len)
+{
+	unsigned long tmp;
+	unsigned short *bp = (unsigned short *) buf;
+
+	tmp = (common_io_access(addr)) | ((0xf & addr) << IORegAddress);
+
+	while (len--)
+		*bp++ = _swapw((unsigned short)hy_inpw(tmp));
+
+}
+
+static inline void io_insl(unsigned int addr, void *buf, int len)
+{
+	unsigned long tmp;
+	unsigned int *bp = (unsigned int *) buf;
+
+	tmp = (common_io_access(addr)) | ((0xf & addr) << IORegAddress);
+
+	while (len--)
+		*bp++ = _swapl((unsigned int)hy_inpw(tmp));
+}
+
+#define inb(addr)    readb(addr)
+#define inw(addr)    readw(addr)
+#define inl(addr)    readl(addr)
+#define outb(x,addr) ((void) writeb(x,addr))
+#define outw(x,addr) ((void) writew(x,addr))
+#define outl(x,addr) ((void) writel(x,addr))
+
+#define insb(a,b,l) io_insb(a,b,l)
+#define insw(a,b,l) io_insw(a,b,l)
+#define insl(a,b,l) io_insl(a,b,l)
+#define outsb(a,b,l) io_outsb(a,b,l)
+#define outsw(a,b,l) io_outsw(a,b,l)
+#define outsl(a,b,l) io_outsl(a,b,l)
+	
+#define memcpy_fromio(a,b,c)	memcpy((a),(void *)(b),(c))
+#define memcpy_toio(a,b,c)	memcpy((void *)(a),(b),(c))
+
+#endif /* ETHERBOOT_IO_H */

src/arch/e1/include/latch.h

@@ -0,0 +1,12 @@
+#ifndef LATCH_H
+#define LATCH_H
+
+//#define TICKS_PER_SEC		(1000000UL)
+#define TICKS_PER_SEC		(625000UL)
+
+/* Fixed timer interval used for calibrating a more precise timer */
+//#define LATCHES_PER_SEC		10
+
+void sleep_latch(void);
+
+#endif /* LATCH_H */

src/arch/e1/include/limits.h

@@ -0,0 +1,34 @@
+/*--------------------------------------------------------------------------*/
+/* Project:        ANSI C Standard Header Files                             */
+/* File:           LIMITS.H                                                 */
+/* Edited by:      hyperstone electronics GmbH                              */
+/*                 Am Seerhein 8                                            */
+/*                 D-78467 Konstanz, Germany                                */
+/* Date:           January 30, 1996                                         */
+/*--------------------------------------------------------------------------*/
+/* Purpose:                                                                 */
+/* The header file <limits.h> defines limits of ordinal types               */
+/* (char, short, int, long)                                                 */
+/*--------------------------------------------------------------------------*/
+
+#ifndef __LIMITS_H
+#define __LIMITS_H  1
+
+#define MB_LEN_MAX        1
+#define CHAR_BIT          8
+#define SCHAR_MIN         -128L
+#define SCHAR_MAX         127L
+#define UCHAR_MAX         255
+#define CHAR_MIN          0
+#define CHAR_MAX          UCHAR_MAX
+#define SHRT_MIN          -32768
+#define SHRT_MAX          32767
+#define USHRT_MAX         65535
+#define INT_MIN           0x80000000
+#define INT_MAX           0x7FFFFFFF
+#define UINT_MAX          0xFFFFFFFFL
+#define LONG_MIN          INT_MIN
+#define LONG_MAX          INT_MAX
+#define ULONG_MAX         UINT_MAX
+
+#endif

src/arch/e1/include/setjmp.h

@@ -0,0 +1,23 @@
+#ifndef _SETJMP_H
+#define _SETJMP_H
+
+
+typedef struct {
+	unsigned long G3;
+	unsigned long G4;
+	unsigned long SavedSP;
+	unsigned long SavedPC;
+	unsigned long SavedSR;
+	unsigned long ReturnValue;
+} __jmp_buf[1];
+
+typedef struct __jmp_buf_tag	/* C++ doesn't like tagless structs.  */
+  {
+    __jmp_buf __jmpbuf;		/* Calling environment.  */
+    int __mask_was_saved;	/* Saved the signal mask?  */
+  } jmp_buf[1];
+
+void longjmp(jmp_buf state, int value );
+int setjmp( jmp_buf state);
+
+#endif

src/arch/e1/include/stdint.h

@@ -0,0 +1,16 @@
+#ifndef STDINT_H
+#define STDINT_H
+
+typedef unsigned long      size_t;
+
+typedef unsigned char      uint8_t;
+typedef unsigned short     uint16_t;
+typedef unsigned long      uint32_t;
+typedef unsigned long long uint64_t;
+
+typedef signed char        int8_t;
+typedef signed short       int16_t;
+typedef signed int         int32_t;
+typedef signed long long   int64_t;
+
+#endif /* STDINT_H */

src/arch/i386/Config

@@ -0,0 +1,131 @@
+# Config for i386 Etherboot
+#
+# Do not delete the tag OptionDescription and /OptionDescription
+# It is used to automatically generate the documentation.
+#
+# @OptionDescrition@
+#
+#	BIOS interface options:
+#
+#	-DPCBIOS
+#			Compile in support for the normal pcbios
+#	-DLINUXBIOS
+#			Compile in support for LinuxBIOS
+#	-DBBS_BUT_NOT_PNP_COMPLIANT
+#			Some BIOSes claim to be PNP but they don't conform
+#			to the BBS spec which specifies that ES:DI must
+#			point to the string $PnP on entry. This option
+#			works around those. This option must be added to
+#			LCONFIG.
+#	-DNO_DELAYED_INT
+#			Take control as soon as BIOS detects the ROM.
+#			Normally hooks onto INT18H or INT19H. Use only if you
+#			have a very non-conformant BIOS as it bypasses
+#			BIOS initialisation of devices. This only works for
+#			legacy ROMs, i.e. PCI_PNP_HEADER not defined.
+#			This option was formerly called NOINT19H.
+#	-DBOOT_INT18H
+#			Etherboot normally hooks onto INT19H for legacy ROMs.
+#			You can choose to hook onto INT18H (BASIC interpreter
+#			entry point) instead. This entry point is used when
+#			all boot devices have been exhausted. This option must
+#			be added to LCONFIG.
+#	-DCONFIG_PCI_DIRECT
+#			Define this for PCI BIOSes that do not implement
+#			BIOS32 or not correctly. Normally not needed.
+#			Only works for BIOSes of a certain era.
+#	-DCONFIG_TSC_CURRTICKS
+#			Uses the processor time stamp counter instead of reading
+#			the BIOS time counter.  This allows Etherboot to work
+#			even without a BIOS.  This only works on late model
+#			486s and above.
+#	-DCONFIG_NO_TIMER2
+#			Some systems do not have timer2 implemented.
+#			If you have a RTC this will allow you to roughly calibrate
+#			it using outb instructions.
+#	-DIBM_L40
+#			This option uses the 0x92 method of controlling
+#			A20 instead of the traditional method of using the
+#			keyboard controller. An explanation of A20 is here:
+#			http://www.win.tue.nl/~aeb/linux/kbd/A20.html
+#			This occurs on MCA, EISA and some embedded boards,
+#			and sometimes with the Fast Gate A20 option on some
+#			BIOSes.
+#			Enable this only if you are sure of what you are doing.
+#
+#	Extended cpu options
+
+#	-DCONFIG_X86_64	
+#			Compile in support for booting x86_64 64bit binaries.
+#
+#	PXE loader options:
+#
+#	-DPXELOADER_KEEP_ALL
+#			Prevent PXE loader (prefix) from unloading the
+#			PXE stack.  You will want to use this if, for
+#			example, you are booting via PXE-on-floppy.
+#			You may want to use it under certain
+#			circumstances when using the Etherboot UNDI
+#			driver; these are complex and best practice is
+#			not yet established.
+#			
+#	Obscure options you probably don't need to touch:
+#
+#	-DIGNORE_E820_MAP
+#			Ignore the memory map returned by the E820 BIOS
+#			call.  May be necessary on some buggy BIOSes.
+#	-DT503_AUI
+#			Use AUI by default on 3c503 cards.
+#	-DFLATTEN_REAL_MODE
+# 			Use 4GB segment limits when calling out to or
+#			returning to real-mode code.  This is necessary to
+#			work around some buggy code (e.g. OpenBSD's pxeboot)
+#			that uses flat real-mode without being sufficiently
+#			paranoid about the volatility of its segment limits.
+
+#
+# @/OptionDescription@
+
+# BIOS select don't change unless you know what you are doing
+CFLAGS+=	-DPCBIOS
+
+# Compile in k8/hammer support
+# CFLAGS+=	-DCONFIG_X86_64
+
+# Options to make a version of Etherboot that will work under linuxBIOS.
+# CFLAGS+=	-DLINUXBIOS -DCONFIG_TSC_CURRTICKS  -DCONSOLE_SERIAL -DCOMCONSOLE=0x3f8 -DCOMPRESERVE -DCONFIG_PCI_DIRECT -DELF_IMAGE 
+
+# These options affect the loader that is prepended to the Etherboot image
+# LCONFIG+=	-DBBS_BUT_NOT_PNP_COMPLIANT
+# LCONFIG+=	-DBOOT_INT18H
+
+# Produce code that will work inside the Bochs emulator.  The pnic
+# driver is probably the best one to try.
+# CFLAGS+=	-DCONFIG_PCI_DIRECT
+
+# Produce code that will work with OpenBSD's pxeboot
+# CFLAGS+=	-DFLATTEN_REAL_MODE
+
+CFLAGS+= -fstrength-reduce -fomit-frame-pointer -march=i386
+# Squeeze the code in as little space as possible.
+# gcc3 needs a different syntax to gcc2 if you want to avoid spurious warnings.
+GCC_VERSION	 = $(subst ., ,$(shell $(CC) -dumpversion))
+GCC_MAJORVERSION = $(firstword $(GCC_VERSION))
+ifeq ($(GCC_MAJORVERSION),2)
+CFLAGS+=	-malign-jumps=1 -malign-loops=1 -malign-functions=1
+else
+CFLAGS+=	-falign-jumps=1 -falign-loops=1 -falign-functions=1
+endif
+GCC_MINORVERSION = $(word 2, $(GCC_VERSION))
+ifneq ($(GCC_MINORVERSION),4)
+CFLAGS+=	-mcpu=i386
+endif
+
+LDFLAGS+=	-N
+
+ifeq "$(shell uname -s)" "FreeBSD"
+CFLAGS+=	-DIMAGE_FREEBSD -DELF_IMAGE -DAOUT_IMAGE
+endif
+
+# An alternate location for isolinux.bin can be set here
+# ISOLINUX_BIN=/path/to/isolinux.bin

src/arch/i386/Makefile

@@ -0,0 +1,373 @@
+ARCH_FORMAT=	elf32-i386
+
+# For debugging, don't delete intermediates
+#.SECONDARY:
+
+LDSCRIPT=	arch/i386/core/etherboot.lds
+PLDSCRIPT=	arch/i386/core/etherboot.prefix.lds
+
+LCONFIG+=	-Ui386
+
+ROMLIMIT=	524288
+CHECKSIZE=	{ read d1; read d1 d2 d3 size d4; [ $$size -gt $(ROMLIMIT) ] &&\
+	{ $(RM) $@; echo "ERROR: code size exceeds limit!"; exit 1; }; exit 0; }
+
+START=	$(BIN)/start32.o $(BIN)/linuxbios.o \
+	$(BIN)/bios.o $(BIN)/console.o $(BIN)/memsizes.o $(BIN)/basemem.o \
+	$(BIN)/hidemem.o $(BIN)/e820mangler.o \
+	$(BIN)/realmode.o $(BIN)/realmode_asm.o \
+	$(BIN)/callbacks.o $(BIN)/pxe_callbacks.o
+
+SRCS+=	arch/i386/prefix/floppyprefix.S
+SRCS+=	arch/i386/prefix/unhuf.S
+SRCS+=	arch/i386/prefix/unnrv2b.S
+SRCS+=	arch/i386/firmware/pcbios/bios.c
+SRCS+=	arch/i386/firmware/pcbios/console.c
+SRCS+=	arch/i386/firmware/pcbios/memsizes.c
+SRCS+=	arch/i386/firmware/pcbios/basemem.c
+SRCS+=	arch/i386/firmware/pcbios/hidemem.c
+SRCS+=	arch/i386/firmware/pcbios/e820mangler.S
+SRCS+=	arch/i386/prefix/liloprefix.S
+SRCS+=	arch/i386/prefix/elfprefix.S
+SRCS+=	arch/i386/prefix/lmelf_prefix.S
+SRCS+=	arch/i386/prefix/elf_dprefix.S
+SRCS+=	arch/i386/prefix/lmelf_dprefix.S
+SRCS+=	arch/i386/prefix/comprefix.S
+SRCS+=	arch/i386/prefix/exeprefix.S
+SRCS+=	arch/i386/prefix/pxeprefix.S
+SRCS+=	arch/i386/prefix/romprefix.S
+
+SRCS+=	arch/i386/core/init.S
+SRCS+=	arch/i386/core/start32.S
+SRCS+=	arch/i386/core/pci_io.c
+SRCS+=	arch/i386/core/i386_timer.c
+SRCS+=	arch/i386/core/elf.c
+SRCS+=	arch/i386/core/cpu.c
+SRCS+=	arch/i386/core/video_subr.c
+SRCS+=	arch/i386/core/pic8259.c
+SRCS+=	arch/i386/core/hooks.c
+SRCS+=	arch/i386/core/callbacks.c
+SRCS+=	arch/i386/core/realmode.c
+SRCS+=	arch/i386/core/realmode_asm.S
+SRCS+=	arch/i386/core/pxe_callbacks.c
+
+# ROM loaders: ISA and PCI versions
+ISAPREFIX=	$(BIN)/isaprefix.o
+ISAENTRY=	$(BIN)/isaprefix.entry.o
+ISAEXIT=	$(BIN)/isaprefix.exit.o
+PCIPREFIX=	$(BIN)/pciprefix.o
+PCIENTRY=	$(BIN)/pciprefix.entry.o
+PCIEXIT=	$(BIN)/pciprefix.exit.o
+# Variables xxx_ROMTYPE are defined by genrules.pl.  ROMENTRY and
+# ROMEXIT will evaluate to give the correct objects to use.
+TARGETBASE=$(patsubst $(BIN)/%,%,$(firstword $(subst ., ,$@)))
+ROMCARD=$(firstword $(subst --, ,$(TARGETBASE)))
+ROMTYPE=$(firstword $(ROMTYPE_$(ROMCARD)) ISA)
+romENTRY=$($(ROMTYPE)ENTRY)
+romEXIT=$($(ROMTYPE)EXIT)
+
+# Target type for generic prf rules
+TARGETTYPE=$(patsubst .%,%, $(suffix $(basename $@)))
+TARGETENTRY=$($(TARGETTYPE)ENTRY)
+TARGETEXIT=$($(TARGETTYPE)EXIT)
+
+# Other real-mode entry loaders
+dskPREFIX=	$(BIN)/floppyprefix.o
+dskENTRY=	$(BIN)/floppyprefix.entry.o
+dskEXIT=	$(BIN)/floppyprefix.exit.o
+comPREFIX=	$(BIN)/comprefix.o
+comENTRY=	$(BIN)/comprefix.entry.o
+comEXIT=	$(BIN)/comprefix.exit.o
+exePREFIX=	$(BIN)/exeprefix.o
+exeENTRY=	$(BIN)/exeprefix.entry.o
+exeEXIT=	$(BIN)/exeprefix.exit.o
+liloPREFIX=	$(BIN)/liloprefix.o
+liloENTRY=	$(BIN)/liloprefix.entry.o
+liloEXIT=	$(BIN)/liloprefix.exit.o
+bImagePREFIX=	$(BIN)/bImageprefix.o
+bImageENTRY=	$(BIN)/bImageprefix.entry.o
+bImageEXIT=	$(BIN)/bImageprefix.exit.o
+pxePREFIX=	$(BIN)/pxeprefix.o
+pxeENTRY=	$(BIN)/pxeprefix.entry.o
+pxeEXIT=	$(BIN)/pxeprefix.exit.o
+rawPREFIX=	$(BIN)/nullprefix.o
+rawENTRY=	$(BIN)/nullprefix.entry.o
+rawEXIT=	$(BIN)/nullprefix.exit.o
+
+# Protected mode entry loaders
+elfPREFIX=	$(BIN)/elfprefix.o
+elfENTRY=	$(BIN)/elfprefix.entry.o
+elfEXIT=	$(BIN)/elfprefix.exit.o
+lmelfPREFIX=	$(BIN)/lmelf_prefix.o
+lmelfENTRY=	$(BIN)/lmelf_prefix.entry.o
+lmelfEXIT=	$(BIN)/lmelf_prefix.exit.o
+elfdPREFIX=	$(BIN)/elf_dprefix.o
+elfdENTRY=	$(BIN)/elf_dprefix.entry.o
+elfdEXIT=	$(BIN)/elf_dprefix.exit.o
+lmelfdPREFIX=	$(BIN)/lmelf_dprefix.o
+lmelfdENTRY=	$(BIN)/lmelf_dprefix.entry.o
+lmelfdEXIT=	$(BIN)/lmelf_dprefix.exit.o
+
+include		$(BIN)/Roms
+
+all:		$(ROMS)
+allroms:	$(ROMS)
+allzroms:	$(ROMS)
+alldsks:	$(EB_DSKS)
+allzdsks:	$(EB_ZDSKS)
+alllilos:	$(EB_LILOS)
+allzlilos:	$(EB_ZLILOS)
+allbImages:	$(EB_BIMAGES)
+allbzImages:	$(EB_BZIMAGES)
+allpxes:	$(EB_PXES)
+allzpxes:	$(EB_ZPXES)
+allelfs:	$(EB_ELFS)
+allzelfs:	$(EB_ZELFS)
+alllmelfs:	$(EB_LMELFS)
+allzlmelfs:	$(EB_ZLMELFS)
+allelfds:	$(EB_ELFDS)
+allzelfds:	$(EB_ZELFDS)
+alllmelfds:	$(EB_LMELFDS)
+allzlmelfds:	$(EB_ZLMELFDS)
+allcoms:	$(EB_COMS)
+allexes:	$(EB_EXES)
+allisos:	$(EB_ISOS)
+alllisos:	$(EB_LISOS)
+
+BOBJS+=		$(BIN)/pci_io.o $(BIN)/i386_timer.o
+BOBJS+=		$(BIN)/elf.o $(BIN)/cpu.o $(BIN)/video_subr.o
+BOBJS+=		$(BIN)/pic8259.o $(BIN)/hooks.o
+
+# ROM loaders
+
+$(ISAPREFIX):	arch/i386/prefix/romprefix.S $(MAKEDEPS)
+	$(CPP) $(CFLAGS) $(LCONFIG) -Ui386 -D ASSEMBLY $< \
+		| $(AS) $(ASFLAGS) -o $@ 
+
+$(PCIPREFIX):	arch/i386/prefix/romprefix.S $(MAKEDEPS)
+	$(CPP) -DPCI_PNP_HEADER $(CFLAGS) $(LCONFIG) -Ui386 -D ASSEMBLY $< \
+		| $(AS) $(ASFLAGS) -o $@ 
+
+# Prefix splitters
+$(BIN)/%prefix.entry.o: $(BIN)/%prefix.o $(MAKEDEPS)
+	$(OBJCOPY) -R .text16 $< $@
+
+$(BIN)/%prefix.exit.o: $(BIN)/%prefix.o $(MAKEDEPS)
+	$(OBJCOPY) -R .prefix $< $@
+
+# Generic prefix objects
+PREFIXOBJS = $(BIN)/init.o
+ZPREFIXOBJS = $(BIN)/init.o $(BIN)/unnrv2b.o
+
+# Utilities
+$(BIN)/nrv2b:	util/nrv2b.c
+	$(HOST_CC) -O2 -DENCODE -DDECODE -DMAIN -DVERBOSE -DNDEBUG -DBITSIZE=32 -DENDIAN=0 -o $@ $<
+
+ZFILELEN = perl util/zfilelen.pl
+
+# Pattern Rules
+
+# General for compiling/assembly source files
+
+$(BIN)/%.o:	arch/i386/core/%.c $(MAKEDEPS)
+	$(CC) $(CFLAGS) -o $@ -c $<
+
+$(BIN)/%.o:	arch/i386/core/%.S $(MAKEDEPS)
+	$(CPP) $(CFLAGS) -Ui386 -D ASSEMBLY $< | $(AS) $(ASFLAGS) -o $@
+
+$(BIN)/%.o:	arch/i386/firmware/pcbios/%.c $(MAKEDEPS)
+	$(CC) $(CFLAGS) -o $@ -c $<
+
+$(BIN)/%.o:	arch/i386/firmware/pcbios/%.S $(MAKEDEPS)
+	$(CPP) $(CFLAGS) -Ui386 -D ASSEMBLY $< | $(AS) $(ASFLAGS) -o $@
+
+$(BIN)/%.o:	arch/i386/prefix/%.S $(MAKEDEPS)
+	$(CPP) $(CFLAGS) -Ui386 -D ASSEMBLY $< | $(AS) $(ASFLAGS) -o $@
+
+# general rule for 16bit .o, may be overridden
+$(BIN)/%.o:	$(BIN)/%.s
+	$(AS) $(ASFLAGS) -o $@ $<
+
+# general rule for .bin (plain binary loader code), may be overridden
+$(BIN)/%.bin:	$(BIN)/%.o
+	$(OBJCOPY) -O binary $< $@
+
+# general rule for .z (compressed binary code), may be overridden
+# rule for .z is in top level Makefile
+# Give the directory name, e.g. use $(BIN)/rtl8139.com as the target.
+
+$(BIN)/%.zo:	$(BIN)/%.zbin arch/i386/core/prefixzdata.lds $(MAKEDEPS)
+	$(LD) -T arch/i386/core/prefixzdata.lds -b binary $< -o $@
+
+$(BIN)/%.uo:	$(BIN)/%.bin arch/i386/core/prefixudata.lds $(MAKEDEPS)
+	$(LD) -T arch/i386/core/prefixudata.lds -b binary $< -o $@
+
+# Intermediate prf rules
+
+%.prf:  %.rt $(PREFIXOBJS) %.rt1.uo %.rt2.uo $(MAKEDEPS)
+	$(MAKE) $(TARGETENTRY)
+	$(LD) $(LDFLAGS) -T $(PLDSCRIPT) $(TARGETENTRY) -R $(subst $(MAKEDEPS),,$^)  -o $@ 
+
+%.zprf:  %.rt $(ZPREFIXOBJS) %.rt1.uo %.rt2.zo $(MAKEDEPS)
+	$(MAKE) $(TARGETENTRY)
+	$(LD) $(LDFLAGS) -T $(PLDSCRIPT) $(TARGETENTRY) -R $(subst $(MAKEDEPS),,$^)  -o $@ 
+
+# general rules for normal/compressed ROM images, may be overridden
+SUFFIXES +=	rom zrom
+
+$(BIN)/%.rom.rt: $(BIN)/%.rt.o  $(ISAENTRY) $(PCIENTRY) $(ISAEXIT) $(PCIEXIT) $(LDSCRIPT) $(MAKEDEPS)
+	$(LD) $(LDFLAGS) -T $(LDSCRIPT) -o $@ $(romEXIT) $<
+	@$(SIZE) $@ | $(CHECKSIZE)
+
+$(BIN)/%.rom: $(BIN)/%.rom.prf
+	$(OBJCOPY) -O binary $< $@
+	$(MAKEROM) $(MAKEROM_FLAGS) $(MAKEROM_$(ROMCARD)) $(MAKEROM_ID_$(ROMCARD)) -i$(IDENT) $@
+
+$(BIN)/%.zrom: $(BIN)/%.rom.zprf
+	$(OBJCOPY) -O binary $< $@
+	$(MAKEROM) $(MAKEROM_FLAGS) $(MAKEROM_$(ROMCARD)) $(MAKEROM_ID_$(ROMCARD)) -i$(IDENT) $@
+
+# general rules for ELF images
+SUFFIXES +=	elf zelf
+$(BIN)/%.elf.rt:  $(BIN)/%.rt.o $(elfENTRY) $(elfEXIT) $(LDSCRIPT) $(MAKEDEPS)
+	$(LD) $(LDFLAGS) -T $(LDSCRIPT) -o $@ $(elfEXIT)  $< 
+
+$(BIN)/%.elf: $(BIN)/%.elf.prf
+	$(OBJCOPY) -O binary $< $@
+
+$(BIN)/%.zelf: $(BIN)/%.elf.zprf 
+	$(OBJCOPY) -O binary $< $@
+
+# general rules for Long Mode ELF images
+SUFFIXES +=	lmelf zlmelf
+$(BIN)/%.lmelf.rt: $(BIN)/%.rt.o $(lmelfENTRY) $(lmelfEXIT) $(LDSCRIPT) $(MAKEDEPS)
+	$(LD) $(LDFLAGS) -T $(LDSCRIPT) -o $@ $(lmelfEXIT) $<
+
+$(BIN)/%.lmelf: $(BIN)/%.lmelf.prf
+	$(OBJCOPY) -O binary $< $@
+
+$(BIN)/%.zlmelf: $(BIN)/%.lmelf.zprf
+	$(OBJCOPY) -O binary $< $@
+
+# general rules for ELF dynamic images
+SUFFIXES +=	elfd zelfd
+$(BIN)/%.elfd.rt: $(BIN)/%.rt.o $(elfdENTRY) $(elfdEXIT) $(LDSCRIPT) $(MAKEDEPS)
+	$(LD) $(LDFLAGS) -T $(LDSCRIPT) -o $@ $(elfdEXIT) $<
+
+$(BIN)/%.elfd: $(BIN)/%.elfd.prf
+	$(OBJCOPY) -O binary $< $@
+
+$(BIN)/%.zelfd: $(BIN)/%.elfd.zprf
+	$(OBJCOPY) -O binary $< $@
+
+# general rules for Long Mode ELF dynamic images
+SUFFIXES +=	lmelfd zlmelfd
+$(BIN)/%.lmelfd.rt: $(BIN)/%.rt.o $(lmelfdENTRY) $(lmelfdEXIT) $(LDSCRIPT) $(MAKEDEPS)
+	$(LD) $(LDFLAGS) -T $(LDSCRIPT) -o $@ $(lmelfdEXIT) $<
+
+$(BIN)/%.lmelfd: $(BIN)/%.lmelfd.prf
+	$(OBJCOPY) -O binary $< $@
+
+$(BIN)/%.zlmelfd: $(BIN)/%.lmelfd.zprf
+	$(OBJCOPY) -O binary $< $@
+
+# rules to generate a DOS loadable .com executable
+SUFFIXES +=	com
+$(BIN)/%.com.rt: $(BIN)/%.rt.o $(comENTRY) $(comEXIT) $(LDSCRIPT) $(MAKEDEPS)
+	$(LD) $(LDFLAGS) -T $(LDSCRIPT) -o $@ $< $(comEXIT)
+
+$(BIN)/%.com: $(BIN)/%.com.zprf
+	$(OBJCOPY) -O binary $< $@
+
+# rules to generate a DOS loadable .exe executable
+SUFFIXES +=	exe
+$(BIN)/%.exe.rt: $(BIN)/%.rt.o $(exeENTRY) $(exeEXIT) $(LDSCRIPT) $(MAKEDEPS)
+	$(LD) $(LDFLAGS) -T $(LDSCRIPT) -o $@ $< $(exeEXIT)
+	@$(SIZE) $@ | $(CHECKSIZE)
+
+$(BIN)/%.exe: $(BIN)/%.exe.prf
+	$(OBJCOPY) -O binary $< $@
+
+# rules to make a LILO loadable image
+SUFFIXES +=	lilo zlilo
+
+$(BIN)/%.lilo.rt: $(BIN)/%.rt.o $(liloENTRY) $(liloEXIT) $(LDSCRIPT) $(MAKEDEPS)
+	$(LD) $(LDFLAGS) -T $(LDSCRIPT) -o $@ $< $(liloEXIT)
+	@$(SIZE) $@ | $(CHECKSIZE)
+
+$(BIN)/%.lilo: $(BIN)/%.lilo.prf
+	$(OBJCOPY) -O binary $< $@
+
+$(BIN)/%.zlilo: $(BIN)/%.lilo.zprf
+	$(OBJCOPY) -O binary $< $@
+
+# rules to make big linux boot protocol image
+SUFFIXES +=	bImage bzImage
+
+$(BIN)/%.bImage.rt: $(BIN)/%.rt.o $(bImageENTRY) $(bImageEXIT) $(LDSCRIPT) $(MAKEDEPS)
+	$(LD) $(LDFLAGS) -T $(LDSCRIPT) -o $@ $< $(bImageEXIT)
+
+$(BIN)/%.bImage: $(BIN)/%.bImage.prf
+	$(OBJCOPY) -O binary $< $@
+
+$(BIN)/%.bzImage: $(BIN)/%.bImage.zprf
+	$(OBJCOPY) -O binary $< $@
+
+
+# rules to generate a PXE loadable image
+SUFFIXES +=	pxe zpxe
+
+$(BIN)/%.pxe.rt: $(BIN)/%.rt.o $(pxeENTRY) $(pxeEXIT) $(LDSCRIPT) $(MAKEDEPS)
+	$(LD) $(LDFLAGS) -T $(LDSCRIPT) -o $@ $< $(pxeEXIT)
+	@$(SIZE) $@ | $(CHECKSIZE)
+
+$(BIN)/%.pxe: $(BIN)/%.pxe.prf
+	$(OBJCOPY) -O binary $< $@
+
+$(BIN)/%.zpxe: $(BIN)/%.pxe.zprf
+	$(OBJCOPY) -O binary $< $@
+
+# rules to generate the .dsk/.zdsk floppy images
+SUFFIXES +=	dsk zdsk
+
+$(BIN)/%.dsk.rt: $(BIN)/%.rt.o $(dskENTRY) $(dskEXIT) $(LDSCRIPT) $(MAKEDEPS)
+	$(LD) $(LDFLAGS) -T $(LDSCRIPT) -o $@ $< $(dskEXIT)
+	@$(SIZE) $@ | $(CHECKSIZE)
+
+$(BIN)/%.dsk: $(BIN)/%.dsk.prf
+	$(OBJCOPY) -O binary $< $@
+
+$(BIN)/%.zdsk: $(BIN)/%.dsk.zprf
+	$(OBJCOPY) -O binary $< $@
+
+# rules to write the .dsk/.zdsk image onto a blank floppy
+SUFFIXES +=	fd0 zfd0
+%.fd0:	%.dsk
+	dd if=$< bs=512 conv=sync of=/dev/fd0
+	sync
+
+%.zfd0:	%.zdsk
+	dd if=$< bs=512 conv=sync of=/dev/fd0
+	sync
+
+# rules to create raw executable images
+SUFFIXES +=	raw zraw
+$(BIN)/%.raw.rt: $(BIN)/%.rt.o $(rawENTRY) $(rawEXIT) $(LDSCRIPT) $(MAKEDEPS)
+	$(LD) $(LDFLAGS) -T $(LDSCRIPT) -o $@ $< $(rawEXIT)
+
+$(BIN)/%.raw: $(BIN)/%.raw.prf
+	$(OBJCOPY) -O binary $< $@
+
+$(BIN)/%.zraw: $(BIN)/%.raw.zprf
+	$(OBJCOPY) -O binary $< $@
+
+# rule to make a non-emulation ISO boot image
+SUFFIXES +=	iso
+%.iso:	util/geniso %.zlilo
+	ISOLINUX_BIN=${ISOLINUX_BIN} bash util/geniso $*.iso $*.zlilo
+
+# rule to make a floppy emulation ISO boot image
+SUFFIXES +=	liso
+%.liso:	util/genliso %.zlilo
+	bash util/genliso $*.liso $*.zlilo
+

src/arch/i386/core/aout_loader.c

@@ -0,0 +1,144 @@
+/* a.out */
+struct exec {
+	unsigned long      a_midmag;	/* flags<<26 | mid<<16 | magic */
+	unsigned long      a_text;	/* text segment size */
+	unsigned long      a_data;	/* initialized data size */
+	unsigned long      a_bss;	/* uninitialized data size */
+	unsigned long      a_syms;	/* symbol table size */
+	unsigned long      a_entry;	/* entry point */
+	unsigned long      a_trsize;	/* text relocation size */
+	unsigned long      a_drsize;	/* data relocation size */
+};
+
+struct aout_state {
+	struct exec head;
+	unsigned long curaddr;
+	int segment;			/* current segment number, -1 for none */
+	unsigned long loc;		/* start offset of current block */
+	unsigned long skip;		/* padding to be skipped to current segment */
+	unsigned long toread;		/* remaining data to be read in the segment */
+};
+
+static struct aout_state astate;
+
+static sector_t aout_download(unsigned char *data, unsigned int len, int eof);
+static inline os_download_t aout_probe(unsigned char *data, unsigned int len)
+{
+	unsigned long start, mid, end, istart, iend;
+	if (len < sizeof(astate.head)) {
+		return 0;
+	}
+	memcpy(&astate.head, data, sizeof(astate.head));
+	if ((astate.head.a_midmag & 0xffff) != 0x010BL) {
+		return 0;
+	}
+	
+	printf("(a.out");
+	aout_freebsd_probe();
+	printf(")... ");
+	/* Check the aout image */
+	start  = astate.head.a_entry;
+	mid    = (((start + astate.head.a_text) + 4095) & ~4095) + astate.head.a_data;
+	end    = ((mid + 4095) & ~4095) + astate.head.a_bss;
+	istart = 4096;
+	iend   = istart + (mid - start);
+	if (!prep_segment(start, mid, end, istart, iend))
+		return dead_download;
+	astate.segment = -1;
+	astate.loc = 0;
+	astate.skip = 0;
+	astate.toread = 0;
+	return aout_download;
+}
+
+static sector_t aout_download(unsigned char *data, unsigned int len, int eof)
+{
+	unsigned int offset;	/* working offset in the current data block */
+
+	offset = 0;
+
+#ifdef AOUT_LYNX_KDI
+	astate.segment++;
+	if (astate.segment == 0) {
+		astate.curaddr = 0x100000;
+		astate.head.a_entry = astate.curaddr + 0x20;
+	}
+	memcpy(phys_to_virt(astate.curaddr), data, len);
+	astate.curaddr += len;
+	return 0;
+#endif
+
+	do {
+		if (astate.segment != -1) {
+			if (astate.skip) {
+				if (astate.skip >= len - offset) {
+					astate.skip -= len - offset;
+					break;
+				}
+				offset += astate.skip;
+				astate.skip = 0;
+			}
+
+			if (astate.toread) {
+				if (astate.toread >= len - offset) {
+					memcpy(phys_to_virt(astate.curaddr), data+offset,
+						len - offset);
+					astate.curaddr += len - offset;
+					astate.toread -= len - offset;
+					break;
+				}
+				memcpy(phys_to_virt(astate.curaddr), data+offset, astate.toread);
+				offset += astate.toread;
+				astate.toread = 0;
+			}
+		}
+
+		/* Data left, but current segment finished - look for the next
+		 * segment.  This is quite simple for a.out files.  */
+		astate.segment++;
+		switch (astate.segment) {
+		case 0:
+			/* read text */
+			astate.curaddr = astate.head.a_entry;
+			astate.skip = 4096;
+			astate.toread = astate.head.a_text;
+			break;
+		case 1:
+			/* read data */
+			/* skip and curaddr may be wrong, but I couldn't find
+			 * examples where this failed.  There is no reasonable
+			 * documentation for a.out available.  */
+			astate.skip = ((astate.curaddr + 4095) & ~4095) - astate.curaddr;
+			astate.curaddr = (astate.curaddr + 4095) & ~4095;
+			astate.toread = astate.head.a_data;
+			break;
+		case 2:
+			/* initialize bss and start kernel */
+			astate.curaddr = (astate.curaddr + 4095) & ~4095;
+			astate.skip = 0;
+			astate.toread = 0;
+			memset(phys_to_virt(astate.curaddr), '\0', astate.head.a_bss);
+			goto aout_startkernel;
+		default:
+			break;
+		}
+	} while (offset < len);
+
+	astate.loc += len;
+
+	if (eof) {
+		unsigned long entry;
+
+aout_startkernel:
+		entry = astate.head.a_entry;
+		done(1);
+
+		aout_freebsd_boot();
+#ifdef AOUT_LYNX_KDI
+		xstart32(entry);
+#endif
+		printf("unexpected a.out variant\n");
+		longjmp(restart_etherboot, -2);
+	}
+	return 0;
+}

src/arch/i386/core/callbacks.c

@@ -0,0 +1,107 @@
+/* Callout/callback interface for Etherboot
+ *
+ * This file provides the mechanisms for making calls from Etherboot
+ * to external programs and vice-versa.
+ *
+ * Initial version by Michael Brown <mbrown@fensystems.co.uk>, January 2004.
+ */
+
+#include "etherboot.h"
+#include "callbacks.h"
+#include "realmode.h"
+#include "segoff.h"
+#include <stdarg.h>
+
+/* Maximum amount of stack data that prefix may request to be passed
+ * to its exit routine
+ */
+#define MAX_PREFIX_STACK_DATA 16
+
+/* Prefix exit routine is defined in prefix object */
+extern void prefix_exit ( void );
+extern void prefix_exit_end ( void );
+
+/*****************************************************************************
+ *
+ * IN_CALL INTERFACE
+ *
+ *****************************************************************************
+ */
+
+/* in_call(): entry point for calls in to Etherboot from external code. 
+ *
+ * Parameters: some set up by assembly code _in_call(), others as
+ * passed from external code.
+ */
+uint32_t i386_in_call ( va_list ap, i386_pm_in_call_data_t pm_data,
+			uint32_t opcode ) {
+	uint32_t ret;
+	i386_rm_in_call_data_t rm_data;
+	in_call_data_t in_call_data = { &pm_data, NULL };
+	struct {
+		int data[MAX_PREFIX_STACK_DATA/4];
+	} in_stack;
+
+	/* Fill out rm_data if we were called from real mode */
+	if ( opcode & EB_CALL_FROM_REAL_MODE ) {
+		in_call_data.rm = &rm_data;
+		rm_data = va_arg ( ap, typeof(rm_data) );
+		/* Null return address indicates to use the special
+		 * prefix exit mechanism, and that there are
+		 * parameters on the stack that the prefix wants
+		 * handed to its exit routine.
+		 */
+		if ( rm_data.ret_addr.offset == 0 ) {
+			int n = va_arg ( ap, int ) / 4;
+			int i;
+			for ( i = 0; i < n; i++ ) {
+				in_stack.data[i] = va_arg ( ap, int );
+			}
+		}
+	}
+	
+	/* Hand off to main in_call() routine */
+	ret = in_call ( &in_call_data, opcode, ap );
+
+	/* If real-mode return address is null, it means that we
+	 * should exit via the prefix's exit path, which is part of
+	 * our image.  (This arrangement is necessary since the prefix
+	 * code itself may have been vapourised by the time we want to
+	 * return.)
+	 */
+	if ( ( opcode & EB_CALL_FROM_REAL_MODE ) &&
+	     ( rm_data.ret_addr.offset == 0 ) ) {
+		real_call ( prefix_exit, &in_stack, NULL );
+		/* Should never return */
+	}
+		
+	return ret;
+}
+
+#ifdef	CODE16
+
+/* install_rm_callback_interface(): install real-mode callback
+ * interface at specified address.
+ *
+ * Real-mode code may then call to this address (or lcall to this
+ * address plus RM_IN_CALL_FAR) in order to make an in_call() to
+ * Etherboot.
+ *
+ * Returns the size of the installed code, or 0 if the code could not
+ * be installed.
+ */
+int install_rm_callback_interface ( void *address, size_t available ) {
+	if ( available &&
+	     ( available < rm_callback_interface_size ) ) return 0;
+
+	/* Inform RM code where to find Etherboot */
+	rm_etherboot_location = virt_to_phys(_text);
+
+	/* Install callback interface */
+	memcpy ( address, &rm_callback_interface,
+		 rm_callback_interface_size );
+
+	return rm_callback_interface_size;
+}
+
+#endif	/* CODE16 */

src/arch/i386/core/cpu.c

@@ -0,0 +1,86 @@
+#ifdef CONFIG_X86_64
+#include "stdint.h"
+#include "string.h"
+#include "bits/cpu.h"
+
+
+/* Standard macro to see if a specific flag is changeable */
+static inline int flag_is_changeable_p(uint32_t flag)
+{
+	uint32_t f1, f2;
+
+	asm("pushfl\n\t"
+	    "pushfl\n\t"
+	    "popl %0\n\t"
+	    "movl %0,%1\n\t"
+	    "xorl %2,%0\n\t"
+	    "pushl %0\n\t"
+	    "popfl\n\t"
+	    "pushfl\n\t"
+	    "popl %0\n\t"
+	    "popfl\n\t"
+	    : "=&r" (f1), "=&r" (f2)
+	    : "ir" (flag));
+
+	return ((f1^f2) & flag) != 0;
+}
+
+
+/* Probe for the CPUID instruction */
+static inline int have_cpuid_p(void)
+{
+	return flag_is_changeable_p(X86_EFLAGS_ID);
+}
+
+static void identify_cpu(struct cpuinfo_x86 *c)
+{
+	unsigned xlvl;
+
+	c->cpuid_level = -1;		/* CPUID not detected */
+	c->x86_model = c->x86_mask = 0;	/* So far unknown... */
+	c->x86_vendor_id[0] = '\0';	/* Unset */
+	memset(&c->x86_capability, 0, sizeof c->x86_capability);
+	
+	if (!have_cpuid_p()) {
+		/* CPU doesn'thave CPUID */
+
+		/* If there are any capabilities, they'r vendor-specific */
+		/* enable_cpuid() would have set c->x86 for us. */
+	}
+	else {
+		/* CPU does have CPUID */
+
+		/* Get vendor name */
+		cpuid(0x00000000, &c->cpuid_level,
+		      (int *)&c->x86_vendor_id[0],
+		      (int *)&c->x86_vendor_id[8],
+		      (int *)&c->x86_vendor_id[4]);
+		
+		/* Initialize the standard set of capabilities */
+		/* Note that the vendor-specific code below might override */
+
+		/* Intel-defined flags: level 0x00000001 */
+		if ( c->cpuid_level >= 0x00000001 ) {
+			unsigned tfms, junk;
+			cpuid(0x00000001, &tfms, &junk, &junk,
+			      &c->x86_capability[0]);
+			c->x86 = (tfms >> 8) & 15;
+			c->x86_model = (tfms >> 4) & 15;
+			c->x86_mask = tfms & 15;
+		}
+
+		/* AMD-defined flags: level 0x80000001 */
+		xlvl = cpuid_eax(0x80000000);
+		if ( (xlvl & 0xffff0000) == 0x80000000 ) {
+			if ( xlvl >= 0x80000001 )
+				c->x86_capability[1] = cpuid_edx(0x80000001);
+		}
+	}
+}
+
+struct cpuinfo_x86 cpu_info;
+void cpu_setup(void)
+{
+	identify_cpu(&cpu_info);
+}
+#endif /* CONFIG_X86_64 */

src/arch/i386/core/elf.c

@@ -0,0 +1,135 @@
+#include "etherboot.h"
+#include "elf.h"
+
+
+#define NAME "Etherboot"
+
+#if defined(PCBIOS)
+#define FIRMWARE "PCBIOS"
+#endif
+#if defined(LINUXBIOS)
+#define FIRMWARE "LinuxBIOS"
+#endif
+#if !defined(FIRMWARE)
+#error "No BIOS selected"
+#endif
+
+#define SZ(X) ((sizeof(X)+3) & ~3)
+#define CP(D,S) (memcpy(&(D), &(S), sizeof(S)))
+
+struct elf_notes {
+	/* The note header */
+	struct Elf_Bhdr hdr;
+
+	/* First the Fixed sized entries that must be well aligned */
+
+	/* Pointer to bootp data */
+	Elf_Nhdr nf1;
+	char     nf1_name[SZ(EB_PARAM_NOTE)];
+	uint32_t nf1_bootp_data;
+
+	/* Pointer to ELF header */
+	Elf_Nhdr nf2;
+	char     nf2_name[SZ(EB_PARAM_NOTE)];
+	uint32_t nf2_header;
+
+	/* A copy of the i386 memory map */
+	Elf_Nhdr nf3;
+	char     nf3_name[SZ(EB_PARAM_NOTE)];
+	struct meminfo nf3_meminfo;
+
+	/* Then the variable sized data string data where alignment does not matter */
+
+	/* The bootloader name */
+	Elf_Nhdr nv1;
+	char     nv1_desc[SZ(NAME)];
+	/* The bootloader version */
+	Elf_Nhdr nv2;
+	char     nv2_desc[SZ(VERSION)];
+	/* The firmware type */
+	Elf_Nhdr nv3;
+	char     nv3_desc[SZ(FIRMWARE)];
+	/* Name of the loaded image */
+	Elf_Nhdr nv4;
+	char     nv4_loaded_image[128];
+	/* An empty command line */
+	Elf_Nhdr nv5;
+	char     nv5_cmdline[SZ("")];
+};
+
+#define ELF_NOTE_COUNT  (3 + 5)
+
+static struct elf_notes notes;
+struct Elf_Bhdr *prepare_boot_params(void *header)
+{
+	memset(&notes, 0, sizeof(notes));
+	notes.hdr.b_signature = ELF_BHDR_MAGIC;
+	notes.hdr.b_size      = sizeof(notes);
+	notes.hdr.b_checksum  = 0;
+	notes.hdr.b_records   = ELF_NOTE_COUNT;
+
+	/* Initialize the fixed length entries. */
+	notes.nf1.n_namesz = sizeof(EB_PARAM_NOTE);
+	notes.nf1.n_descsz = sizeof(notes.nf1_bootp_data);
+	notes.nf1.n_type   = EB_BOOTP_DATA;
+	CP(notes.nf1_name,   EB_PARAM_NOTE);
+	notes.nf1_bootp_data = virt_to_phys(BOOTP_DATA_ADDR);
+
+	notes.nf2.n_namesz = sizeof(EB_PARAM_NOTE);
+	notes.nf2.n_descsz = sizeof(notes.nf2_header);
+	notes.nf2.n_type   = EB_HEADER;
+	CP(notes.nf2_name,   EB_PARAM_NOTE);
+	notes.nf2_header   = virt_to_phys(header);
+
+	notes.nf3.n_namesz = sizeof(EB_PARAM_NOTE);
+	notes.nf3.n_descsz = sizeof(notes.nf3_meminfo);
+	notes.nf3.n_type   = EB_I386_MEMMAP;
+	CP(notes.nf3_name,   EB_PARAM_NOTE);
+	memcpy(&notes.nf3_meminfo, &meminfo, sizeof(meminfo));
+
+	/* Initialize the variable length entries */
+	notes.nv1.n_namesz = 0;
+	notes.nv1.n_descsz = sizeof(NAME);
+	notes.nv1.n_type   = EBN_BOOTLOADER_NAME;
+	CP(notes.nv1_desc,   NAME);
+
+	notes.nv2.n_namesz = 0;
+	notes.nv2.n_descsz = sizeof(VERSION);
+	notes.nv2.n_type   = EBN_BOOTLOADER_VERSION;
+	CP(notes.nv2_desc,   VERSION);
+
+	notes.nv3.n_namesz = 0;
+	notes.nv3.n_descsz = sizeof(FIRMWARE);
+	notes.nv3.n_type   = EBN_FIRMWARE_TYPE;
+	CP(notes.nv3_desc,   FIRMWARE);
+
+	/* Attempt to pass the name of the loaded image */
+	notes.nv4.n_namesz = 0;
+	notes.nv4.n_descsz = sizeof(notes.nv4_loaded_image);
+	notes.nv4.n_type   = EBN_LOADED_IMAGE;
+	memcpy(&notes.nv4_loaded_image, KERNEL_BUF, sizeof(notes.nv4_loaded_image));
+
+	/* Pass an empty command line for now */
+	notes.nv5.n_namesz = 0;
+	notes.nv5.n_descsz = sizeof("");
+	notes.nv5.n_type   = EBN_COMMAND_LINE;
+	CP(notes.nv5_cmdline,   "");
+
+
+	notes.hdr.b_checksum = ipchksum(&notes, sizeof(notes));
+	/* Like UDP invert a 0 checksum to show that a checksum is present */
+	if (notes.hdr.b_checksum == 0) {
+		notes.hdr.b_checksum = 0xffff;
+	}
+	return &notes.hdr;
+}
+
+int elf_start(unsigned long machine __unused_i386, unsigned long entry, unsigned long params)
+{
+#if defined(CONFIG_X86_64)
+	if (machine == EM_X86_64) {
+		return xstart_lm(entry, params);
+	}
+#endif
+	return xstart32(entry, params);
+}

src/arch/i386/core/etherboot.lds

@@ -0,0 +1,90 @@
+OUTPUT_FORMAT("elf32-i386", "elf32-i386", "elf32-i386")
+OUTPUT_ARCH(i386)
+
+ENTRY(_text)
+SECTIONS {
+	. = ALIGN(16);
+	/* Start address of Etherboot in the virtual address space */
+	_virt_start = 0;
+	_text = . ;
+	.text.nocompress : {
+		*(.text*.nocompress)
+		. = ALIGN(16);
+	} = 0x9090
+
+	.text16 : {
+		_text16 = .;
+		*(.text16)
+		*(.text16.*)
+		_etext16 = . ;
+	}
+	.text.compress : {
+		*(.text)
+		*(.text.*)
+	} = 0x9090
+	.rodata : {
+		. = ALIGN(4);
+		*(.rodata)
+		*(.rodata.*)
+	}
+	. = ALIGN(4);
+	.drivers.pci : {
+		pci_drivers = . ;
+		*(.drivers.pci);
+		pci_drivers_end = . ;
+	}
+	. = ALIGN(4);
+	.drivers.isa : {
+		isa_drivers = . ;
+		*(.drivers.isa);
+		isa_drivers_end = .;
+	}
+	_etext = . ;
+	_data = . ;
+	.data : {
+		*(.data)
+		*(.data.*)
+	}
+	_edata = . ;
+	_uncompressed_verbatim_end = . ;
+	. = ALIGN(16);
+	.bss.preserve : {
+		*(.bss.preserve)
+		*(.bss.preserve.*)
+	}
+	_bss = . ;
+	.bss : {
+		*(.bss)
+		*(.bss.*)
+	}
+	. = ALIGN(16);
+	_ebss = .;
+	_stack = . ;
+	.stack : {
+		_stack_start = . ;
+		*(.stack)
+		*(.stack.*)
+		_stack_end = . ;
+	}
+	_bss_size = _ebss - _bss;
+	_stack_offset = _stack - _text ;
+	_stack_offset_pgh = _stack_offset / 16 ;
+	_stack_size = _stack_end - _stack_start ;
+	. = ALIGN(16);
+	_end = . ;
+
+	/DISCARD/ : {
+		*(.comment)
+		*(.note)
+	}
+
+	/* PXE-specific symbol calculations.  The results of these are
+	 * needed in romprefix.S, which is why they must be calculated
+	 * here.
+	 */
+	_pxe_stack_size = _pxe_stack_t_size
+		+ _pxe_callback_interface_size
+		+ _rm_callback_interface_size
+		+ _e820mangler_size + 15 ;
+
+}

src/arch/i386/core/etherboot.prefix.lds

@@ -0,0 +1,100 @@
+OUTPUT_FORMAT("elf32-i386", "elf32-i386", "elf32-i386")
+OUTPUT_ARCH(i386)
+
+ENTRY(_prefix_start)
+SECTIONS {
+	/* Prefix */
+	.prefix : {
+		_verbatim_start = . ;
+		_prefix_start = . ;
+		*(.prefix)
+		. = ALIGN(16);
+		_prefix_end = . ;
+	} = 0x9090
+	_prefix_size = _prefix_end - _prefix_start;
+
+	.text.nocompress : {
+		*(.prefix.udata)
+	} = 0x9090
+
+	decompress_to = . ;
+	.prefix.zdata : {
+		_compressed = . ;
+		*(.prefix.zdata)
+		_compressed_end = . ;
+	}
+	_compressed_size = _compressed_end - _compressed;
+
+	. = ALIGN(16);
+	_verbatim_end = . ;
+
+
+	/* Size of the core of etherboot in memory */
+	_base_size = _end - _text;
+
+	/* _prefix_size is the length of the non-core etherboot prefix */
+	_prefix_size = _prefix_end - _prefix_start;
+
+	/* _verbatim_size is the actual amount that has to be copied to base memory */
+	_verbatim_size = _verbatim_end - _verbatim_start;
+
+	/* _image_size is the amount of base memory needed to run */
+	_image_size = _base_size +  _prefix_size;
+
+	/* Standard sizes rounded up to paragraphs */
+	_prefix_size_pgh   = (_prefix_size + 15) / 16;
+	_verbatim_size_pgh = (_verbatim_size + 15) / 16;
+	_image_size_pgh    = (_image_size + 15) / 16 ;
+	
+	/* Standard sizes in sectors */
+	_prefix_size_sct   = (_prefix_size + 511) / 512;
+	_verbatim_size_sct = (_verbatim_size + 511) / 512;
+	_image_size_sct    = (_image_size + 511) / 512;
+
+	/* Symbol offsets and sizes for the exe prefix */
+	_exe_hdr_size   = 32;
+	_exe_size       = _verbatim_size; /* Should this be - 32 to exclude the header? */
+	_exe_size_tail  = (_exe_size) % 512;
+	_exe_size_pages = ((_exe_size) + 511) / 512;
+	_exe_bss_size   = ((_image_size - _verbatim_size) + 15) / 16;
+	_exe_ss_offset  = (_stack_offset + _prefix_size - _exe_hdr_size + 15) / 16 ;
+
+	/* This is where we copy the compressed image before decompression.
+	 * Prepare to decompress in place.  The end mark is about 8.25 bytes long,
+	 * and the worst case symbol is about 16.5 bytes long.  Therefore
+	 * We need to reserve at least 25 bytes of slack here.  
+	 * Currently I reserve 2048 bytes of just slack to be safe :)
+	 * 2048 bytes easily falls within the BSS (the defualt stack is 4096 bytes)
+	 * so we really are decompressing in place.
+	 * 
+	 * Hmm. I missed a trick.  In the very worst case (no compression)
+	 * the encoded data is 9/8 the size as it started out so to be completely
+	 * safe I need to be 1/8 of the uncompressed code size past the end.
+	 * This will still fit compfortably into our bss in any conceivable scenario.
+	 */
+	_compressed_copy = _edata + _prefix_size  - _compressed_size +
+		/* The amount to overflow _edata */
+		MAX( ((_edata - _text + 7) / 8) , 2016 ) + 32; 
+	_assert = ASSERT( ( _compressed_copy - _prefix_size ) < _ebss , "Cannot decompress in place" ) ; 
+
+	decompress = DEFINED(decompress) ? decompress : 0;
+	/DISCARD/ : {
+		*(.comment)
+		*(.note)
+	}
+
+	/* Symbols used by the prefixes whose addresses are inconvinient 
+	 * to compute, at runtime in the code.
+	 */
+	image_basemem_size = DEFINED(image_basemem_size)? image_basemem_size : 65536;
+	image_basemem      = DEFINED(image_basemem)?     image_basemem : 65536;
+	_prefix_real_to_prot          = _real_to_prot         + _prefix_size ;
+	_prefix_prot_to_real          = _prot_to_real         + _prefix_size ;
+	_prefix_image_basemem_size    = image_basemem_size    + _prefix_size ;
+	_prefix_image_basemem         = image_basemem         + _prefix_size ;
+	_prefix_rm_in_call            = _rm_in_call           + _prefix_size ;
+	_prefix_in_call               = _in_call              + _prefix_size ;
+	_prefix_rom                   = rom                   + _prefix_size ;
+	_prefix_rm_etherboot_location = rm_etherboot_location + _prefix_size ;
+	_prefix_stack_end             = _stack_end            + _prefix_size ;
+}

src/arch/i386/core/freebsd_loader.c

@@ -0,0 +1,377 @@
+/* bootinfo */
+#define BOOTINFO_VERSION 1
+#define NODEV           (-1)    /* non-existent device */
+#define PAGE_SHIFT      12              /* LOG2(PAGE_SIZE) */
+#define PAGE_SIZE       (1<<PAGE_SHIFT) /* bytes/page */
+#define PAGE_MASK       (PAGE_SIZE-1)
+#define N_BIOS_GEOM     8
+
+struct bootinfo {
+        unsigned int            bi_version;
+        const unsigned char     *bi_kernelname;
+        struct nfs_diskless     *bi_nfs_diskless;
+                                /* End of fields that are always present. */
+#define bi_endcommon            bi_n_bios_used
+        unsigned int            bi_n_bios_used;
+        unsigned long           bi_bios_geom[N_BIOS_GEOM];
+        unsigned int            bi_size;
+        unsigned char           bi_memsizes_valid;
+        unsigned char           bi_pad[3];
+        unsigned long           bi_basemem;
+        unsigned long           bi_extmem;
+        unsigned long           bi_symtab;
+        unsigned long           bi_esymtab;
+	/* Note that these are in the FreeBSD headers but were not here... */
+	unsigned long           bi_kernend;		/* end of kernel space */
+	unsigned long           bi_envp;		/* environment */
+	unsigned long           bi_modulep;		/* preloaded modules */
+};
+
+static struct bootinfo bsdinfo;
+
+#ifdef ELF_IMAGE
+static Elf32_Shdr *shdr;	/* To support the FreeBSD kludge! */
+static Address symtab_load;
+static Address symstr_load;
+static int symtabindex;
+static int symstrindex;
+#endif
+
+static enum {
+	Unknown, Tagged, Aout, Elf, Aout_FreeBSD, Elf_FreeBSD,
+} image_type = Unknown;
+
+static unsigned int off;
+
+
+#ifdef ELF_IMAGE
+static void elf_freebsd_probe(void)
+{
+	image_type = Elf;
+	if (	(estate.e.elf32.e_entry & 0xf0000000) && 
+		(estate.e.elf32.e_type == ET_EXEC))
+	{
+		image_type = Elf_FreeBSD;
+		printf("/FreeBSD");
+		off = -(estate.e.elf32.e_entry & 0xff000000);
+		estate.e.elf32.e_entry += off;
+	}
+	/* Make sure we have a null to start with... */
+	shdr = 0;
+	
+	/* Clear the symbol index values... */
+	symtabindex = -1;
+	symstrindex = -1;
+	
+	/* ...and the load addresses of the symbols  */
+	symtab_load = 0;
+	symstr_load = 0;
+}
+
+static void elf_freebsd_fixup_segment(void)
+{
+	if (image_type == Elf_FreeBSD) {
+		estate.p.phdr32[estate.segment].p_paddr += off;
+	}
+}
+
+static void elf_freebsd_find_segment_end(void)
+{
+	/* Count the bytes read even for the last block
+	 * as we will need to know where the last block
+	 * ends in order to load the symbols correctly.
+	 * (plus it could be useful elsewhere...)
+	 * Note that we need to count the actual size,
+	 * not just the end of the disk image size.
+	 */
+	estate.curaddr += 
+		(estate.p.phdr32[estate.segment].p_memsz - 
+		estate.p.phdr32[estate.segment].p_filesz);
+}
+
+static int elf_freebsd_debug_loader(unsigned int offset)
+{
+	/* No more segments to be loaded - time to start the
+	 * nasty state machine to support the loading of
+	 * FreeBSD debug symbols due to the fact that FreeBSD
+	 * uses/exports the kernel's debug symbols in order
+	 * to make much of the system work!  Amazing (arg!)
+	 *
+	 * We depend on the fact that for the FreeBSD kernel,
+	 * there is only one section of debug symbols and that
+	 * the section is after all of the loaded sections in
+	 * the file.  This assumes a lot but is somewhat required
+	 * to make this code not be too annoying.  (Where do you
+	 * load symbols when the code has not loaded yet?)
+	 * Since this function is actually just a callback from
+	 * the network data transfer code, we need to be able to
+	 * work with the data as it comes in.  There is no chance
+	 * for doing a seek other than forwards.
+	 *
+	 * The process we use is to first load the section
+	 * headers.  Once they are loaded (shdr != 0) we then
+	 * look for where the symbol table and symbol table
+	 * strings are and setup some state that we found
+	 * them and fall into processing the first one (which
+	 * is the symbol table) and after that has been loaded,
+	 * we try the symbol strings.  Note that the order is
+	 * actually required as the memory image depends on
+	 * the symbol strings being loaded starting at the
+	 * end of the symbol table.  The kernel assumes this
+	 * layout of the image.
+	 *
+	 * At any point, if we get to the end of the load file
+	 * or the section requested is earlier in the file than
+	 * the current file pointer, we just end up falling
+	 * out of this and booting the kernel without this
+	 * information.
+	 */
+
+	/* Make sure that the next address is long aligned... */
+	/* Assumes size of long is a power of 2... */
+	estate.curaddr = (estate.curaddr + sizeof(long) - 1) & ~(sizeof(long) - 1);
+	
+	/* If we have not yet gotten the shdr loaded, try that */
+	if (shdr == 0)
+	{
+		estate.toread = estate.e.elf32.e_shnum * estate.e.elf32.e_shentsize;
+		estate.skip = estate.e.elf32.e_shoff - (estate.loc + offset);
+		if (estate.toread)
+		{
+#if ELF_DEBUG
+			printf("shdr *, size %lX, curaddr %lX\n", 
+				estate.toread, estate.curaddr);
+#endif
+			
+			/* Start reading at the curaddr and make that the shdr */
+			shdr = (Elf32_Shdr *)phys_to_virt(estate.curaddr);
+			
+			/* Start to read... */
+			return 1;
+		}
+	}
+	else
+	{
+		/* We have the shdr loaded, check if we have found
+		 * the indexs where the symbols are supposed to be */
+		if ((symtabindex == -1) && (symstrindex == -1))
+		{
+			int i;
+			/* Make sure that the address is page aligned... */
+			/* Symbols need to start in their own page(s)... */
+			estate.curaddr = (estate.curaddr + 4095) & ~4095;
+			
+			/* Need to make new indexes... */
+			for (i=0; i < estate.e.elf32.e_shnum; i++)
+			{
+				if (shdr[i].sh_type == SHT_SYMTAB)
+				{
+					int j;
+					for (j=0; j < estate.e.elf32.e_phnum; j++)
+					{
+						/* Check only for loaded sections */
+						if ((estate.p.phdr32[j].p_type | 0x80) == (PT_LOAD | 0x80))
+						{
+							/* Only the extra symbols */
+							if ((shdr[i].sh_offset >= estate.p.phdr32[j].p_offset) &&
+								((shdr[i].sh_offset + shdr[i].sh_size) <=
+									(estate.p.phdr32[j].p_offset + estate.p.phdr32[j].p_filesz)))
+							{
+								shdr[i].sh_offset=0;
+								shdr[i].sh_size=0;
+								break;
+							}
+						}
+					}
+					if ((shdr[i].sh_offset != 0) && (shdr[i].sh_size != 0))
+					{
+						symtabindex = i;
+						symstrindex = shdr[i].sh_link;
+					}
+				}
+			}
+		}
+		
+		/* Check if we have a symbol table index and have not loaded it */
+		if ((symtab_load == 0) && (symtabindex >= 0))
+		{
+			/* No symbol table yet?  Load it first... */
+			
+			/* This happens to work out in a strange way.
+			 * If we are past the point in the file already,
+			 * we will skip a *large* number of bytes which
+			 * ends up bringing us to the end of the file and
+			 * an old (default) boot.  Less code and lets
+			 * the state machine work in a cleaner way but this
+			 * is a nasty side-effect trick... */
+			estate.skip = shdr[symtabindex].sh_offset - (estate.loc + offset);
+			
+			/* And we need to read this many bytes... */
+			estate.toread = shdr[symtabindex].sh_size;
+			
+			if (estate.toread)
+			{
+#if ELF_DEBUG
+				printf("db sym, size %lX, curaddr %lX\n", 
+					estate.toread, estate.curaddr);
+#endif
+				/* Save where we are loading this... */
+				symtab_load = phys_to_virt(estate.curaddr);
+				
+				*((long *)phys_to_virt(estate.curaddr)) = estate.toread;
+				estate.curaddr += sizeof(long);
+				
+				/* Start to read... */
+				return 1;
+			}
+		}
+		else if ((symstr_load == 0) && (symstrindex >= 0))
+		{
+			/* We have already loaded the symbol table, so
+			 * now on to the symbol strings... */
+			
+			
+			/* Same nasty trick as above... */
+			estate.skip = shdr[symstrindex].sh_offset - (estate.loc + offset);
+			
+			/* And we need to read this many bytes... */
+			estate.toread = shdr[symstrindex].sh_size;
+			
+			if (estate.toread)
+			{
+#if ELF_DEBUG
+				printf("db str, size %lX, curaddr %lX\n", 
+					estate.toread, estate.curaddr);
+#endif
+				/* Save where we are loading this... */
+				symstr_load = phys_to_virt(estate.curaddr);
+				
+				*((long *)phys_to_virt(estate.curaddr)) = estate.toread;
+				estate.curaddr += sizeof(long);
+				
+				/* Start to read... */
+				return 1;
+			}
+		}
+	}
+	/* all done */
+	return 0;
+}
+
+static void elf_freebsd_boot(unsigned long entry) 
+{
+	if (image_type != Elf_FreeBSD)
+		return;
+
+	memset(&bsdinfo, 0, sizeof(bsdinfo));
+	bsdinfo.bi_basemem = meminfo.basememsize;
+	bsdinfo.bi_extmem = meminfo.memsize;
+	bsdinfo.bi_memsizes_valid = 1;
+	bsdinfo.bi_version = BOOTINFO_VERSION;
+	bsdinfo.bi_kernelname = virt_to_phys(KERNEL_BUF);
+	bsdinfo.bi_nfs_diskless = NULL;
+	bsdinfo.bi_size = sizeof(bsdinfo);
+#define RB_BOOTINFO     0x80000000      /* have `struct bootinfo *' arg */  
+	if(freebsd_kernel_env[0] != '\0'){
+		freebsd_howto |= RB_BOOTINFO;
+		bsdinfo.bi_envp = (unsigned long)freebsd_kernel_env;
+	}
+	
+	/* Check if we have symbols loaded, and if so,
+	 * made the meta_data needed to pass those to
+	 * the kernel. */
+	if ((symtab_load !=0) && (symstr_load != 0))
+	{
+		unsigned long *t;
+		
+		bsdinfo.bi_symtab = symtab_load;
+		
+		/* End of symbols (long aligned...) */
+		/* Assumes size of long is a power of 2... */
+		bsdinfo.bi_esymtab = (symstr_load +
+			sizeof(long) +
+			*((long *)symstr_load) +
+			sizeof(long) - 1) & ~(sizeof(long) - 1);
+		
+		/* Where we will build the meta data... */
+		t = phys_to_virt(bsdinfo.bi_esymtab);
+		
+#if ELF_DEBUG
+		printf("Metadata at %lX\n",t);
+#endif
+		
+		/* Set up the pointer to the memory... */
+		bsdinfo.bi_modulep = virt_to_phys(t);
+		
+		/* The metadata structure is an array of 32-bit
+		 * words where we store some information about the
+		 * system.  This is critical, as FreeBSD now looks
+		 * only for the metadata for the extended symbol
+		 * information rather than in the bootinfo.
+		 */
+		/* First, do the kernel name and the kernel type */
+		/* Note that this assumed x86 byte order... */
+		
+		/* 'kernel\0\0' */
+		*t++=MODINFO_NAME; *t++= 7; *t++=0x6E72656B; *t++=0x00006C65;
+		
+		/* 'elf kernel\0\0' */
+		*t++=MODINFO_TYPE; *t++=11; *t++=0x20666C65; *t++=0x6E72656B; *t++ = 0x00006C65;
+		
+		/* Now the symbol start/end - note that they are
+		 * here in local/physical address - the Kernel
+		 * boot process will relocate the addresses. */
+		*t++=MODINFOMD_SSYM | MODINFO_METADATA; *t++=sizeof(*t); *t++=bsdinfo.bi_symtab;
+		*t++=MODINFOMD_ESYM | MODINFO_METADATA; *t++=sizeof(*t); *t++=bsdinfo.bi_esymtab;
+		
+		*t++=MODINFO_END; *t++=0; /* end of metadata */
+		
+		/* Since we have symbols we need to make
+		 * sure that the kernel knows its own end
+		 * of memory...  It is not _end but after
+		 * the symbols and the metadata... */
+		bsdinfo.bi_kernend = virt_to_phys(t);
+		
+		/* Signal locore.s that we have a valid bootinfo
+		 * structure that was completely filled in. */
+		freebsd_howto |= 0x80000000;
+	}
+	
+	xstart32(entry, freebsd_howto, NODEV, 0, 0, 0, 
+		virt_to_phys(&bsdinfo), 0, 0, 0);
+	longjmp(restart_etherboot, -2);
+}
+#endif
+
+#ifdef AOUT_IMAGE
+static void aout_freebsd_probe(void)
+{
+	image_type = Aout;
+	if (((astate.head.a_midmag >> 16) & 0xffff) == 0) {
+		/* Some other a.out variants have a different
+		 * value, and use other alignments (e.g. 1K),
+		 * not the 4K used by FreeBSD.  */
+		image_type = Aout_FreeBSD;
+		printf("/FreeBSD");
+		off = -(astate.head.a_entry & 0xff000000);
+		astate.head.a_entry += off;
+	}
+}
+
+static void aout_freebsd_boot(void)
+{
+	if (image_type == Aout_FreeBSD) {
+		memset(&bsdinfo, 0, sizeof(bsdinfo));
+		bsdinfo.bi_basemem = meminfo.basememsize;
+		bsdinfo.bi_extmem = meminfo.memsize;
+		bsdinfo.bi_memsizes_valid = 1;
+		bsdinfo.bi_version = BOOTINFO_VERSION;
+		bsdinfo.bi_kernelname = virt_to_phys(KERNEL_BUF);
+		bsdinfo.bi_nfs_diskless = NULL;
+		bsdinfo.bi_size = sizeof(bsdinfo);
+		xstart32(astate.head.a_entry, freebsd_howto, NODEV, 0, 0, 0, 
+			virt_to_phys(&bsdinfo), 0, 0, 0);
+		longjmp(restart_etherboot, -2);
+	}
+}
+#endif

src/arch/i386/core/hooks.c

@@ -0,0 +1,35 @@
+#include "etherboot.h"
+#include "callbacks.h"
+#include <stdarg.h>
+
+void arch_main ( in_call_data_t *data __unused, va_list params __unused )
+{
+#ifdef PCBIOS
+	/* Deallocate base memory used for the prefix, if applicable
+	 */
+	forget_prefix_base_memory();
+#endif
+
+}
+
+void arch_relocated_from (unsigned long old_addr )
+{
+
+#ifdef PCBIOS
+	/* Deallocate base memory used for the Etherboot runtime,
+	 * if applicable
+	 */
+	forget_runtime_base_memory( old_addr );
+#endif
+
+}
+
+void arch_on_exit ( int exit_status __unused ) 
+{
+#ifdef PCBIOS
+	/* Deallocate the real-mode stack now.  We will reallocate
+	 * the stack if are going to use it after this point.
+	 */
+	forget_real_mode_stack();
+#endif
+}

src/arch/i386/core/i386_timer.c

@@ -0,0 +1,191 @@
+/* A couple of routines to implement a low-overhead timer for drivers */
+
+ /*
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2, or (at
+ * your option) any later version.
+ */
+
+#include	"etherboot.h"
+#include	"timer.h"
+#include	"latch.h"
+
+void __load_timer2(unsigned int ticks)
+{
+	/*
+	 * Now let's take care of PPC channel 2
+	 *
+	 * Set the Gate high, program PPC channel 2 for mode 0,
+	 * (interrupt on terminal count mode), binary count,
+	 * load 5 * LATCH count, (LSB and MSB) to begin countdown.
+	 *
+	 * Note some implementations have a bug where the high bits byte
+	 * of channel 2 is ignored.
+	 */
+	/* Set up the timer gate, turn off the speaker */
+	/* Set the Gate high, disable speaker */
+	outb((inb(PPC_PORTB) & ~PPCB_SPKR) | PPCB_T2GATE, PPC_PORTB);
+	/* binary, mode 0, LSB/MSB, Ch 2 */
+	outb(TIMER2_SEL|WORD_ACCESS|MODE0|BINARY_COUNT, TIMER_MODE_PORT);
+	/* LSB of ticks */
+	outb(ticks & 0xFF, TIMER2_PORT);
+	/* MSB of ticks */
+	outb(ticks >> 8, TIMER2_PORT);
+}
+
+static int __timer2_running(void)
+{
+	return ((inb(PPC_PORTB) & PPCB_T2OUT) == 0);
+}
+
+#if !defined(CONFIG_TSC_CURRTICKS)
+void setup_timers(void)
+{
+	return;
+}
+
+void load_timer2(unsigned int ticks)
+{
+	return __load_timer2(ticks);
+}
+
+int timer2_running(void)
+{
+	return __timer2_running();
+}
+
+void ndelay(unsigned int nsecs)
+{
+	waiton_timer2((nsecs * CLOCK_TICK_RATE)/1000000000);
+}
+void udelay(unsigned int usecs)
+{
+	waiton_timer2((usecs * TICKS_PER_MS)/1000);
+}
+#endif /* !defined(CONFIG_TSC_CURRTICKS) */
+
+#if defined(CONFIG_TSC_CURRTICKS)
+
+#define rdtsc(low,high) \
+     __asm__ __volatile__("rdtsc" : "=a" (low), "=d" (high))
+
+#define rdtscll(val) \
+     __asm__ __volatile__ ("rdtsc" : "=A" (val))
+
+
+/* Number of clock ticks to time with the rtc */
+#define LATCH 0xFF
+
+#define LATCHES_PER_SEC ((CLOCK_TICK_RATE + (LATCH/2))/LATCH)
+#define TICKS_PER_LATCH ((LATCHES_PER_SEC + (TICKS_PER_SEC/2))/TICKS_PER_SEC)
+
+static void sleep_latch(void)
+{
+	__load_timer2(LATCH);
+	while(__timer2_running());
+}
+
+/* ------ Calibrate the TSC ------- 
+ * Time how long it takes to excute a loop that runs in known time.
+ * And find the convertion needed to get to CLOCK_TICK_RATE
+ */
+
+
+static unsigned long long calibrate_tsc(void)
+{
+	unsigned long startlow, starthigh;
+	unsigned long endlow, endhigh;
+	
+	rdtsc(startlow,starthigh);
+	sleep_latch();
+	rdtsc(endlow,endhigh);
+
+	/* 64-bit subtract - gcc just messes up with long longs */
+	__asm__("subl %2,%0\n\t"
+		"sbbl %3,%1"
+		:"=a" (endlow), "=d" (endhigh)
+		:"g" (startlow), "g" (starthigh),
+		"0" (endlow), "1" (endhigh));
+	
+	/* Error: ECPUTOOFAST */
+	if (endhigh)
+		goto bad_ctc;
+	
+	endlow *= TICKS_PER_LATCH;
+	return endlow;
+
+	/*
+	 * The CTC wasn't reliable: we got a hit on the very first read,
+	 * or the CPU was so fast/slow that the quotient wouldn't fit in
+	 * 32 bits..
+	 */
+bad_ctc:
+	printf("bad_ctc\n");
+	return 0;
+}
+
+static unsigned long clocks_per_tick;
+void setup_timers(void)
+{
+	if (!clocks_per_tick) {
+		clocks_per_tick = calibrate_tsc();
+		/* Display the CPU Mhz to easily test if the calibration was bad */
+		printf("CPU %ld Mhz\n", (clocks_per_tick/1000 * TICKS_PER_SEC)/1000);
+	}
+}
+
+unsigned long currticks(void)
+{
+	unsigned long clocks_high, clocks_low;
+	unsigned long currticks;
+	/* Read the Time Stamp Counter */
+	rdtsc(clocks_low, clocks_high);
+
+	/* currticks = clocks / clocks_per_tick; */
+	__asm__("divl %1"
+		:"=a" (currticks)
+		:"r" (clocks_per_tick), "0" (clocks_low), "d" (clocks_high));
+
+
+	return currticks;
+}
+
+static unsigned long long timer_timeout;
+static int __timer_running(void)
+{
+	unsigned long long now;
+	rdtscll(now);
+	return now < timer_timeout;
+}
+
+void udelay(unsigned int usecs)
+{
+	unsigned long long now;
+	rdtscll(now);
+	timer_timeout = now + usecs * ((clocks_per_tick * TICKS_PER_SEC)/(1000*1000));
+	while(__timer_running());
+}
+void ndelay(unsigned int nsecs)
+{
+	unsigned long long now;
+	rdtscll(now);
+	timer_timeout = now + nsecs * ((clocks_per_tick * TICKS_PER_SEC)/(1000*1000*1000));
+	while(__timer_running());
+}
+
+void load_timer2(unsigned int timer2_ticks)
+{
+	unsigned long long now;
+	unsigned long clocks;
+	rdtscll(now);
+	clocks = timer2_ticks * ((clocks_per_tick * TICKS_PER_SEC)/CLOCK_TICK_RATE);
+	timer_timeout = now + clocks;
+}
+
+int timer2_running(void)
+{
+	return __timer_running();
+}
+
+#endif /* RTC_CURRTICKS */

src/arch/i386/core/init.S

@@ -0,0 +1,305 @@
+#include "callbacks.h"
+	.equ CR0_PE, 1
+	
+	.text
+	.arch i386
+	.section ".prefix", "ax", @progbits
+
+#undef CODE16
+#if defined(PCBIOS)
+#define	CODE16
+#endif
+
+/* We have two entry points: "conventional" (at the start of the file)
+ * and "callback" (at _entry, 2 bytes in).  The "callback" entry
+ * should be used if the caller wishes to provide a specific opcode.
+ * It is equivalent to a call to in_call.  Using the "conventional"
+ * entry point is equivalent to using the "callback" entry point with
+ * an opcode of EB_OPCODE_MAIN.
+ *
+ * Both entry points can be called in either 16-bit real or 32-bit
+ * protected mode with flat physical addresses.  We detect which mode
+ * the processor is in and call either in_call or rm_in_call as
+ * appropriate.  Note that the mode detection code must therefore be
+ * capable of doing the same thing in either mode, even though the
+ * machine code instructions will be interpreted differently.
+ *
+ * The decompressor will be invoked if necessary to decompress
+ * Etherboot before attempting to jump to it.
+ */
+
+/******************************************************************************
+ * Entry points and mode detection code
+ ******************************************************************************
+ */
+
+	.code32
+/* "Conventional" entry point: caller provides no opcode */
+	.globl	_start
+_start:
+	/* Set flag to indicate conventional entry point used */
+	pushl	$0			/* "pushw $0" in 16-bit code */
+	/* Fall through to "callback" entry point */
+	
+/* "Callback" entry point */
+	.globl	_entry
+_entry:
+	
+#ifdef CODE16
+	/* CPU mode detection code */
+	pushl	%eax			/* "pushw %ax" in 16-bit code */
+	pushw	%ax			/* "pushl %eax" in 16-bit code */
+	movl	%cr0, %eax		/* Test protected mode bit */
+	testb	$CR0_PE, %al
+	popw	%ax			/* "popl %eax" in 16-bit code */
+	popl	%eax			/* "popw %eax" in 16-bit code */
+	jz	rmode
+#endif /* CODE16 */
+
+/******************************************************************************
+ * Entered in protected mode
+ ******************************************************************************
+ */
+		
+	.code32
+pmode:
+	cmpl	$0, 0(%esp)		/* Conventional entry point used? */
+	jne	1f
+	/* Entered via conventional entry point: set up stack */
+	xchgl	%eax, 4(%esp)		/* %eax = return addr, store %eax */
+	movl	%eax, 0(%esp)		/* 0(%esp) = return address */
+	movl	$(EB_OPCODE_MAIN|EB_USE_INTERNAL_STACK|EB_SKIP_OPCODE), %eax
+	xchgl	%eax, 4(%esp)		/* 4(%esp) = opcode, restore %eax */
+1:
+	/* Run decompressor if necessary */
+	pushl	%eax
+	movl	$decompress, %eax
+	testl	%eax, %eax
+	jz	1f
+	call	decompress
+1:	popl	%eax
+
+	/* Make in_call to Etherboot */
+	jmp	_prefix_in_call
+
+/******************************************************************************
+ * Entered in real mode
+ ******************************************************************************
+ */
+	
+#ifdef CODE16
+	.code16
+rmode:
+	pushw	%ax			/* Padding */
+	pushw	%bp
+	movw	%sp, %bp
+	cmpw	$0, 6(%bp)		/* Conventional entry point used? */
+	jne	1f
+	/* Entered via conventional entry point: set up stack */
+	pushw	%ax
+	movw	6(%bp), %ax
+	movw	%ax, 2(%bp)		/* Move return address down */
+	movl	$(EB_OPCODE_MAIN|EB_USE_INTERNAL_STACK|EB_SKIP_OPCODE), 4(%bp)
+	popw	%ax
+	popw	%bp
+	jmp	2f
+1:	/* Entered via callback entry point: do nothing */
+	popw	%bp
+	popw	%ax
+2:
+	/* Preserve registers */
+	pushw	%ds
+	pushl	%eax
+	
+	/* Run decompressor if necessary.  Decompressor is 32-bit
+	 * code, so we must switch to pmode first.  Save and restore
+	 * GDT over transition to pmode.
+	 */
+	movl	$decompress, %eax
+	testl	%eax, %eax
+	jz	1f
+	pushw	%ds
+	pushw	%es
+	pushw	%fs
+	pushw	%gs
+	subw	$8, %sp
+	pushw	%bp
+	movw	%sp, %bp
+	sgdt	2(%bp)
+	pushw	%ss			/* Store params for _prot_to_real */
+	pushw	%cs
+	call	_prefix_real_to_prot
+	.code32
+	call	decompress
+	call	_prefix_prot_to_real
+	.code16
+	popw	%ax			/* skip */
+	popw	%ax			/* skip */
+	lgdt	2(%bp)
+	popw	%bp
+	addw	$8, %sp
+	popw	%gs
+	popw	%fs
+	popw	%es
+	popw	%ds
+1:
+
+	/* Set rm_etherboot_location */
+	xorl	%eax, %eax
+	movw	%cs, %ax
+	movw	%ax, %ds
+	shll	$4, %eax
+	addl	$_prefix_size, %eax
+	movl	%eax, _prefix_rm_etherboot_location
+
+	/* Restore registers */
+	popl	%eax
+	popw	%ds
+
+	/* Make real-mode in_call to Etherboot */
+	jmp	_prefix_rm_in_call
+#endif /* CODE16 */
+
+/******************************************************************************
+ * Utility routines that can be called by the "prefix".
+ ******************************************************************************
+ */
+
+#ifdef CODE16
+
+/* Prelocate code: either to an area at the top of free base memory.
+ * Switch stacks to use the stack within the resulting
+ * Etherboot image.
+ *
+ * On entry, %cs:0000 must be the start of the prefix: this is used to
+ * locate the code to be copied.
+ *
+ * This routine takes a single word parameter: the number of bytes to
+ * be transferred from the old stack to the new stack (excluding the
+ * return address and this parameter itself, which will always be
+ * copied).  If this value is negative, the stacks will not be
+ * switched.
+ *
+ * Control will "return" to the appropriate point in the relocated
+ * image.
+ */
+
+#define PRELOC_PRESERVE ( 20 )
+#define PRELOC_OFFSET_RETADDR ( PRELOC_PRESERVE )
+#define PRELOC_OFFSET_RETADDR_E ( PRELOC_OFFSET_RETADDR + 4 )
+#define PRELOC_OFFSET_COPY ( PRELOC_OFFSET_RETADDR_E )
+#define PRELOC_OFFSET_COPY_E ( PRELOC_OFFSET_COPY + 2 )
+
+#define PRELOC_ALWAYS_COPY ( PRELOC_OFFSET_COPY_E )
+	
+	.code16
+	.globl	prelocate
+prelocate:
+	/* Pad to allow for expansion of return address */
+	pushw	%ax
+	
+	/* Preserve registers */
+	pushaw
+	pushw	%ds
+	pushw	%es
+	
+	/* Claim an area of base memory from the BIOS and put the
+	 * payload there.
+	 */
+	movw	$0x40, %bx
+	movw	%bx, %es
+	movw	%es:(0x13), %bx		/* FBMS in kb to %ax */
+	shlw	$6, %bx			/* ... in paragraphs */
+	subw	$_image_size_pgh, %bx	/* Subtract space for image */
+	shrw	$6, %bx			/* Round down to nearest kb */
+	movw	%bx, %es:(0x13)		/* ...and claim memory from BIOS */
+	shlw	$6, %bx
+
+	/* At this point %bx contains the segment address for the
+	 * start of the image (image = prefix + runtime).
+	 */
+
+	/* Switch stacks */
+	movw	%ss, %ax
+	movw	%ax, %ds
+	movw	%sp, %si		/* %ds:si = current %ss:sp */
+	movw	%ss:PRELOC_OFFSET_COPY(%si), %cx
+	testw	%cx, %cx
+	js	1f
+	leaw	_stack_offset_pgh(%bx), %ax /* %ax = new %ss */
+	movw	%ax, %es
+	movw	$_stack_size, %di
+	addw	$PRELOC_ALWAYS_COPY, %cx
+	subw	%cx, %di		/* %es:di = new %ss:sp */
+	movw	%ax, %ss		/* Set new %ss:sp */
+	movw	%di, %sp
+	cld
+	rep movsb			/* Copy stack contents */
+1:	
+
+	/* Do the image copy backwards, since if there's overlap with
+	 * a forward copy then it means we're going to get trashed
+	 * during the copy anyway...
+	 */
+	pushal				/* Preserve 32-bit registers */
+	movw	%bx, %es		/* Destination base for copy */
+	pushw	%cs
+	popw	%ds			/* Source base for copy */
+	movl	$_verbatim_size-1, %ecx	/* Offset to last byte */
+	movl	%ecx, %esi
+	movl	%ecx, %edi
+	incl	%ecx			/* Length */
+	std				/* Backwards copy of binary */
+	ADDR32 rep movsb
+	cld
+	popal				/* Restore 32-bit registers */
+
+	/* Store (%bx<<4) as image_basemem to be picked up by
+	 * basemem.c.  Also store image_size, since there's no other
+	 * way that we can later know how much memory we allocated.
+	 * (_zfile_size is unavailable when rt2 is linked).
+	 */
+	pushl	%eax
+	xorl	%eax, %eax
+	movw	%bx, %ax
+	shll	$4, %eax
+	movl	%eax, %es:_prefix_image_basemem
+	movl	$_image_size, %es:_prefix_image_basemem_size
+	popl	%eax
+
+	/* Expand original near return address into far return to new
+	 * code location.
+	 */
+	movw	%sp, %bp
+	xchgw	%bx, (PRELOC_OFFSET_RETADDR+2)(%bp)
+	movw	%bx, (PRELOC_OFFSET_RETADDR+0)(%bp)
+
+	/* Restore registers and return */
+	popw	%es
+	popw	%ds
+	popaw
+	lret				/* Jump to relocated code */
+
+	/* Utility routine to free base memory allocated by prelocate.
+	 * Ensure that said memory is not in use (e.g. for the CPU
+	 * stack) before calling this routine.
+	 */
+	.globl deprelocate
+deprelocate:	
+	/* Claim an area of base memory from the BIOS and put the
+	 * payload there.
+	 */
+	pushw	%ax
+	pushw	%es
+	movw	$0x40, %ax
+	movw	%ax, %es
+	movw	%es:(0x13), %ax		/* FBMS in kb to %ax */
+	shlw	$6, %ax			/* ... in paragraphs */
+	addw	$_image_size_pgh+0x40-1, %ax /* Add space for image and... */
+	shrw	$6, %ax			/* ...round up to nearest kb */
+	movw	%ax, %es:(0x13)		/* Give memory back to BIOS */
+	popw	%es
+	popw	%ax
+	ret
+	
+#endif /* CODE16 */

src/arch/i386/core/multiboot_loader.c

@@ -0,0 +1,143 @@
+/* Multiboot support
+ *
+ * 2003-07-02 mmap fix and header probe by SONE Takeshi
+ */
+
+struct multiboot_mods {
+	unsigned mod_start;
+	unsigned mod_end;
+	unsigned char *string;
+	unsigned reserved;
+};
+
+struct multiboot_mmap {
+	unsigned int size;
+	unsigned int base_addr_low;
+	unsigned int base_addr_high;
+	unsigned int length_low;
+	unsigned int length_high;
+	unsigned int type;
+};
+
+/* The structure of a Multiboot 0.6 parameter block.  */
+struct multiboot_info {
+	unsigned int flags;
+#define MULTIBOOT_MEM_VALID       0x01
+#define MULTIBOOT_BOOT_DEV_VALID  0x02
+#define MULTIBOOT_CMDLINE_VALID   0x04
+#define MULTIBOOT_MODS_VALID      0x08
+#define MULTIBOOT_AOUT_SYMS_VALID 0x10
+#define MULTIBOOT_ELF_SYMS_VALID  0x20
+#define MULTIBOOT_MMAP_VALID      0x40
+	unsigned int memlower;
+	unsigned int memupper;
+	unsigned int bootdev;
+	unsigned int cmdline;	/* physical address of the command line */
+	unsigned mods_count;
+	struct multiboot_mods *mods_addr;
+	unsigned syms_num;
+	unsigned syms_size;
+	unsigned syms_addr;
+	unsigned syms_shndx;
+	unsigned mmap_length;
+	unsigned  mmap_addr;
+	/* The structure actually ends here, so I might as well put
+	 * the ugly e820 parameters here...
+	 */
+	struct multiboot_mmap mmap[E820MAX];
+};
+
+/* Multiboot image header (minimal part) */
+struct multiboot_header {
+	unsigned int magic;
+#define MULTIBOOT_HEADER_MAGIC 0x1BADB002
+	unsigned int flags;
+	unsigned int checksum;
+};
+
+static struct multiboot_header *mbheader;
+
+static struct multiboot_info mbinfo;
+
+static void multiboot_probe(unsigned char *data, int len)
+{
+    int offset;
+    struct multiboot_header *h;
+
+    /* Multiboot spec requires the header to be in first 8KB of the image */
+    if (len > 8192)
+	    len = 8192;
+
+    for (offset = 0; offset < len; offset += 4) {
+	    h = (struct multiboot_header *) (data + offset);
+	    if (h->magic == MULTIBOOT_HEADER_MAGIC
+			    && h->magic + h->flags + h->checksum == 0) {
+		    printf("/Multiboot");
+		    mbheader = h;
+		    return;
+	    }
+    }
+    mbheader = 0;
+}
+
+static inline void multiboot_boot(unsigned long entry)
+{
+	unsigned char cmdline[512], *c;
+	int i;
+	if (!mbheader)
+		return;
+	/* Etherboot limits the command line to the kernel name,
+	 * default parameters and user prompted parameters.  All of
+	 * them are shorter than 256 bytes.  As the kernel name and
+	 * the default parameters come from the same BOOTP/DHCP entry
+	 * (or if they don't, the parameters are empty), only two
+	 * strings of the maximum size are possible.  Note this buffer
+	 * can overrun if a stupid file name is chosen.  Oh well.  */
+	c = cmdline;
+	for (i = 0; KERNEL_BUF[i] != 0; i++) {
+		switch (KERNEL_BUF[i]) {
+		case ' ':
+		case '\\':
+		case '"':
+			*c++ = '\\';
+			break;
+		default:
+			break;
+		}
+		*c++ = KERNEL_BUF[i];
+	}
+	(void)sprintf(c, " -retaddr %#lX", virt_to_phys(xend32));
+
+	mbinfo.flags = MULTIBOOT_MMAP_VALID | MULTIBOOT_MEM_VALID |MULTIBOOT_CMDLINE_VALID;
+	mbinfo.memlower = meminfo.basememsize;
+	mbinfo.memupper = meminfo.memsize;
+	mbinfo.bootdev = 0;	/* not booted from disk */
+	mbinfo.cmdline = virt_to_phys(cmdline);
+	for (i = 0; i < (int) meminfo.map_count; i++) {
+		mbinfo.mmap[i].size = sizeof(struct multiboot_mmap)
+		    - sizeof(unsigned int);
+		mbinfo.mmap[i].base_addr_low = 
+		    (unsigned int) meminfo.map[i].addr;
+		mbinfo.mmap[i].base_addr_high = 
+		    (unsigned int) (meminfo.map[i].addr >> 32);
+		mbinfo.mmap[i].length_low = 
+		    (unsigned int) meminfo.map[i].size;
+		mbinfo.mmap[i].length_high = 
+		    (unsigned int) (meminfo.map[i].size >> 32);
+		mbinfo.mmap[i].type = meminfo.map[i].type;
+	}
+	mbinfo.mmap_length = meminfo.map_count * sizeof(struct multiboot_mmap);
+	mbinfo.mmap_addr = virt_to_phys(mbinfo.mmap);
+	
+	/* The Multiboot 0.6 spec requires all segment registers to be
+	 * loaded with an unrestricted, writeable segment.
+	 * xstart32 does this for us.
+	 */
+	
+	/* Start the kernel, passing the Multiboot information record
+	 * and the magic number.  */
+	os_regs.eax = 0x2BADB002;
+	os_regs.ebx = virt_to_phys(&mbinfo);
+	xstart32(entry);
+	longjmp(restart_etherboot, -2);
+}

src/arch/i386/core/pci_io.c

@@ -0,0 +1,352 @@
+/*
+** Support for NE2000 PCI clones added David Monro June 1997
+** Generalised to other NICs by Ken Yap July 1997
+**
+** Most of this is taken from:
+**
+** /usr/src/linux/drivers/pci/pci.c
+** /usr/src/linux/include/linux/pci.h
+** /usr/src/linux/arch/i386/bios32.c
+** /usr/src/linux/include/linux/bios32.h
+** /usr/src/linux/drivers/net/ne.c
+*/
+#ifdef CONFIG_PCI
+#include "etherboot.h"
+#include "pci.h"
+
+#ifdef	CONFIG_PCI_DIRECT
+#define  PCIBIOS_SUCCESSFUL                0x00
+
+#define DEBUG 0
+
+/*
+ * Functions for accessing PCI configuration space with type 1 accesses
+ */
+
+#define CONFIG_CMD(bus, device_fn, where)   (0x80000000 | (bus << 16) | (device_fn << 8) | (where & ~3))
+
+int pcibios_read_config_byte(unsigned int bus, unsigned int device_fn,
+			       unsigned int where, uint8_t *value)
+{
+    outl(CONFIG_CMD(bus,device_fn,where), 0xCF8);
+    *value = inb(0xCFC + (where&3));
+    return PCIBIOS_SUCCESSFUL;
+}
+
+int pcibios_read_config_word (unsigned int bus,
+    unsigned int device_fn, unsigned int where, uint16_t *value)
+{
+    outl(CONFIG_CMD(bus,device_fn,where), 0xCF8);
+    *value = inw(0xCFC + (where&2));
+    return PCIBIOS_SUCCESSFUL;
+}
+
+int pcibios_read_config_dword (unsigned int bus, unsigned int device_fn,
+				 unsigned int where, uint32_t *value)
+{
+    outl(CONFIG_CMD(bus,device_fn,where), 0xCF8);
+    *value = inl(0xCFC);
+    return PCIBIOS_SUCCESSFUL;
+}
+
+int pcibios_write_config_byte (unsigned int bus, unsigned int device_fn,
+				 unsigned int where, uint8_t value)
+{
+    outl(CONFIG_CMD(bus,device_fn,where), 0xCF8);
+    outb(value, 0xCFC + (where&3));
+    return PCIBIOS_SUCCESSFUL;
+}
+
+int pcibios_write_config_word (unsigned int bus, unsigned int device_fn,
+				 unsigned int where, uint16_t value)
+{
+    outl(CONFIG_CMD(bus,device_fn,where), 0xCF8);
+    outw(value, 0xCFC + (where&2));
+    return PCIBIOS_SUCCESSFUL;
+}
+
+int pcibios_write_config_dword (unsigned int bus, unsigned int device_fn, unsigned int where, uint32_t value)
+{
+    outl(CONFIG_CMD(bus,device_fn,where), 0xCF8);
+    outl(value, 0xCFC);
+    return PCIBIOS_SUCCESSFUL;
+}
+
+#undef CONFIG_CMD
+
+#else	 /* CONFIG_PCI_DIRECT  not defined */
+
+#if !defined(PCBIOS)
+#error "The pcibios can only be used when the PCBIOS support is compiled in"
+#endif
+
+/* Macro for calling the BIOS32 service.  This replaces the old
+ * bios32_call function.  Use in a statement such as
+ * __asm__ ( BIOS32_CALL,
+ *	     : <output registers>
+ *	     : "S" ( bios32_entry ), <other input registers> );
+ */
+#define BIOS32_CALL "call _virt_to_phys\n\t" \
+		    "pushl %%cs\n\t" \
+		    "call *%%esi\n\t" \
+		    "cli\n\t" \
+		    "cld\n\t" \
+		    "call _phys_to_virt\n\t"
+
+static unsigned long bios32_entry;
+static unsigned long pcibios_entry;
+
+static unsigned long bios32_service(unsigned long service)
+{
+	unsigned char return_code;	/* %al */
+	unsigned long address;		/* %ebx */
+	unsigned long length;		/* %ecx */
+	unsigned long entry;		/* %edx */
+
+	__asm__(BIOS32_CALL
+		: "=a" (return_code),
+		  "=b" (address),
+		  "=c" (length),
+		  "=d" (entry)
+		: "0" (service),
+		  "1" (0),
+		  "S" (bios32_entry));
+
+	switch (return_code) {
+		case 0:
+			return address + entry;
+		case 0x80:	/* Not present */
+			printf("bios32_service(%d) : not present\n", service);
+			return 0;
+		default: /* Shouldn't happen */
+			printf("bios32_service(%d) : returned %#X????\n",
+				service, return_code);
+			return 0;
+	}
+}
+
+int pcibios_read_config_byte(unsigned int bus,
+        unsigned int device_fn, unsigned int where, uint8_t *value)
+{
+        unsigned long ret;
+        unsigned long bx = (bus << 8) | device_fn;
+
+        __asm__(BIOS32_CALL
+                "jc 1f\n\t"
+                "xor %%ah, %%ah\n"
+                "1:"
+                : "=c" (*value),
+                  "=a" (ret)
+                : "1" (PCIBIOS_READ_CONFIG_BYTE),
+                  "b" (bx),
+                  "D" ((long) where),
+                  "S" (pcibios_entry));
+        return (int) (ret & 0xff00) >> 8;
+}
+
+int pcibios_read_config_word(unsigned int bus,
+        unsigned int device_fn, unsigned int where, uint16_t *value)
+{
+        unsigned long ret;
+        unsigned long bx = (bus << 8) | device_fn;
+
+        __asm__(BIOS32_CALL
+                "jc 1f\n\t"
+                "xor %%ah, %%ah\n"
+                "1:"
+                : "=c" (*value),
+                  "=a" (ret)
+                : "1" (PCIBIOS_READ_CONFIG_WORD),
+                  "b" (bx),
+                  "D" ((long) where),
+                  "S" (pcibios_entry));
+        return (int) (ret & 0xff00) >> 8;
+}
+
+int pcibios_read_config_dword(unsigned int bus,
+        unsigned int device_fn, unsigned int where, uint32_t *value)
+{
+        unsigned long ret;
+        unsigned long bx = (bus << 8) | device_fn;
+
+        __asm__(BIOS32_CALL
+                "jc 1f\n\t"
+                "xor %%ah, %%ah\n"
+                "1:"
+                : "=c" (*value),
+                  "=a" (ret)
+                : "1" (PCIBIOS_READ_CONFIG_DWORD),
+                  "b" (bx),
+                  "D" ((long) where),
+                  "S" (pcibios_entry));
+        return (int) (ret & 0xff00) >> 8;
+}
+
+int pcibios_write_config_byte (unsigned int bus,
+	unsigned int device_fn, unsigned int where, uint8_t value)
+{
+	unsigned long ret;
+	unsigned long bx = (bus << 8) | device_fn;
+
+	__asm__(BIOS32_CALL
+		"jc 1f\n\t"
+		"xor %%ah, %%ah\n"
+		"1:"
+		: "=a" (ret)
+		: "0" (PCIBIOS_WRITE_CONFIG_BYTE),
+		  "c" (value),
+		  "b" (bx),
+		  "D" ((long) where),
+		  "S" (pcibios_entry));
+	return (int) (ret & 0xff00) >> 8;
+}
+
+int pcibios_write_config_word (unsigned int bus,
+	unsigned int device_fn, unsigned int where, uint16_t value)
+{
+	unsigned long ret;
+	unsigned long bx = (bus << 8) | device_fn;
+
+	__asm__(BIOS32_CALL
+		"jc 1f\n\t"
+		"xor %%ah, %%ah\n"
+		"1:"
+		: "=a" (ret)
+		: "0" (PCIBIOS_WRITE_CONFIG_WORD),
+		  "c" (value),
+		  "b" (bx),
+		  "D" ((long) where),
+		  "S" (pcibios_entry));
+	return (int) (ret & 0xff00) >> 8;
+}
+
+int pcibios_write_config_dword (unsigned int bus,
+	unsigned int device_fn, unsigned int where, uint32_t value)
+{
+	unsigned long ret;
+	unsigned long bx = (bus << 8) | device_fn;
+
+	__asm__(BIOS32_CALL
+		"jc 1f\n\t"
+		"xor %%ah, %%ah\n"
+		"1:"
+		: "=a" (ret)
+		: "0" (PCIBIOS_WRITE_CONFIG_DWORD),
+		  "c" (value),
+		  "b" (bx),
+		  "D" ((long) where),
+		  "S" (pcibios_entry));
+	return (int) (ret & 0xff00) >> 8;
+}
+
+static void check_pcibios(void)
+{
+	unsigned long signature;
+	unsigned char present_status;
+	unsigned char major_revision;
+	unsigned char minor_revision;
+	int pack;
+
+	if ((pcibios_entry = bios32_service(PCI_SERVICE))) {
+		__asm__(BIOS32_CALL
+			"jc 1f\n\t"
+			"xor %%ah, %%ah\n"
+			"1:\tshl $8, %%eax\n\t"
+			"movw %%bx, %%ax"
+			: "=d" (signature),
+			  "=a" (pack)
+			: "1" (PCIBIOS_PCI_BIOS_PRESENT),
+			  "S" (pcibios_entry)
+			: "bx", "cx");
+
+		present_status = (pack >> 16) & 0xff;
+		major_revision = (pack >> 8) & 0xff;
+		minor_revision = pack & 0xff;
+		if (present_status || (signature != PCI_SIGNATURE)) {
+			printf("ERROR: BIOS32 says PCI BIOS, but no PCI "
+				"BIOS????\n");
+			pcibios_entry = 0;
+		}
+#if	DEBUG
+		if (pcibios_entry) {
+			printf ("pcibios_init : PCI BIOS revision %hhX.%hhX"
+				" entry at %#X\n", major_revision,
+				minor_revision, pcibios_entry);
+		}
+#endif
+	}
+}
+
+static void pcibios_init(void)
+{
+	union bios32 *check;
+	unsigned char sum;
+	int i, length;
+	bios32_entry = 0;
+
+	/*
+	 * Follow the standard procedure for locating the BIOS32 Service
+	 * directory by scanning the permissible address range from
+	 * 0xe0000 through 0xfffff for a valid BIOS32 structure.
+	 *
+	 */
+
+	for (check = phys_to_virt(0xe0000); (void *)check <= phys_to_virt(0xffff0); ++check) {
+		if (check->fields.signature != BIOS32_SIGNATURE)
+			continue;
+		length = check->fields.length * 16;
+		if (!length)
+			continue;
+		sum = 0;
+		for (i = 0; i < length ; ++i)
+			sum += check->chars[i];
+		if (sum != 0)
+			continue;
+		if (check->fields.revision != 0) {
+			printf("pcibios_init : unsupported revision %d at %#X, mail drew@colorado.edu\n",
+				check->fields.revision, check);
+			continue;
+		}
+#if	DEBUG
+		printf("pcibios_init : BIOS32 Service Directory "
+			"structure at %#X\n", check);
+#endif
+		if (!bios32_entry) {
+			if (check->fields.entry >= 0x100000) {
+				printf("pcibios_init: entry in high "
+					"memory, giving up\n");
+				return;
+			} else {
+				bios32_entry = check->fields.entry;
+#if	DEBUG
+				printf("pcibios_init : BIOS32 Service Directory"
+					" entry at %#X\n", bios32_entry);
+#endif
+			}
+		}
+	}
+	if (bios32_entry)
+		check_pcibios();
+}
+#endif	/* CONFIG_PCI_DIRECT not defined*/
+
+unsigned long pcibios_bus_base(unsigned int bus __unused)
+{
+	/* architecturally this must be 0 */
+	return 0;
+}
+
+void find_pci(int type, struct pci_device *dev)
+{
+#ifndef	CONFIG_PCI_DIRECT
+	if (!pcibios_entry) {
+		pcibios_init();
+	}
+	if (!pcibios_entry) {
+		printf("pci_init: no BIOS32 detected\n");
+		return;
+	}
+#endif
+	return scan_pci_bus(type, dev);
+}
+#endif /* CONFIG_PCI */

src/arch/i386/core/pic8259.c

@@ -0,0 +1,331 @@
+/*
+ * Basic support for controlling the 8259 Programmable Interrupt Controllers.
+ *
+ * Initially written by Michael Brown (mcb30).
+ */
+
+#include <etherboot.h>
+#include "pic8259.h"
+#include "realmode.h"
+
+#ifdef DEBUG_IRQ
+#define DBG(...) printf ( __VA_ARGS__ )
+#else
+#define DBG(...)
+#endif
+
+/* State of trivial IRQ handler */
+irq_t trivial_irq_installed_on = IRQ_NONE;
+static uint16_t trivial_irq_previous_trigger_count = 0;
+
+/* The actual trivial IRQ handler
+ *
+ * Note: we depend on the C compiler not realising that we're putting
+ * variables in the ".text16" section and therefore not forcing them
+ * back to the ".data" section.  I don't see any reason to expect this
+ * behaviour to change.
+ *
+ * These must *not* be the first variables to appear in this file; the
+ * first variable to appear gets the ".data" directive.
+ */
+RM_FRAGMENT(_trivial_irq_handler,
+	"pushw %bx\n\t"
+	"call  1f\n1:\tpopw %bx\n\t"   /* PIC access to variables */
+	"incw  %cs:(_trivial_irq_trigger_count-1b)(%bx)\n\t" 
+	"popw  %bx\n\t" 
+	"iret\n\t" 
+	"\n\t"
+	".globl _trivial_irq_trigger_count\n\t"
+	"_trivial_irq_trigger_count: .short 0\n\t"
+	"\n\t"
+	".globl _trivial_irq_chain_to\n\t"
+	"_trivial_irq_chain_to: .short 0,0\n\t"
+	"\n\t"
+	".globl _trivial_irq_chain\n\t"
+	"_trivial_irq_chain: .byte 0\n\t"
+	);
+extern volatile uint16_t _trivial_irq_trigger_count;
+extern segoff_t _trivial_irq_chain_to;
+extern int8_t _trivial_irq_chain;
+
+/* Current locations of trivial IRQ handler.  These will change at
+ * runtime when relocation is used; the handler needs to be copied to
+ * base memory before being installed.
+ */
+void (*trivial_irq_handler)P((void)) = _trivial_irq_handler;
+uint16_t volatile *trivial_irq_trigger_count = &_trivial_irq_trigger_count;
+segoff_t *trivial_irq_chain_to = &_trivial_irq_chain_to;
+uint8_t *trivial_irq_chain = &_trivial_irq_chain;
+
+/* Install a handler for the specified IRQ.  Address of previous
+ * handler will be stored in previous_handler.  Enabled/disabled state
+ * of IRQ will be preserved across call, therefore if the handler does
+ * chaining, ensure that either (a) IRQ is disabled before call, or
+ * (b) previous_handler points directly to the place that the handler
+ * picks up its chain-to address.
+ */
+
+int install_irq_handler ( irq_t irq, segoff_t *handler,
+			  uint8_t *previously_enabled,
+			  segoff_t *previous_handler ) {
+	segoff_t *irq_vector = IRQ_VECTOR ( irq );
+	*previously_enabled = irq_enabled ( irq );
+
+	if ( irq > IRQ_MAX ) {
+		DBG ( "Invalid IRQ number %d\n" );
+		return 0;
+	}
+
+	previous_handler->segment = irq_vector->segment;
+	previous_handler->offset = irq_vector->offset;
+	if ( *previously_enabled ) disable_irq ( irq );
+	DBG ( "Installing handler at %hx:%hx for IRQ %d (vector 0000:%hx),"
+	      " leaving %s\n",
+	      handler->segment, handler->offset, irq, virt_to_phys(irq_vector),
+	      ( *previously_enabled ? "enabled" : "disabled" ) );
+	DBG ( "...(previous handler at %hx:%hx)\n",
+	      previous_handler->segment, previous_handler->offset );
+	irq_vector->segment = handler->segment;
+	irq_vector->offset = handler->offset;
+	if ( *previously_enabled ) enable_irq ( irq );
+	return 1;
+}
+
+/* Remove handler for the specified IRQ.  Routine checks that another
+ * handler has not been installed that chains to handler before
+ * uninstalling handler.  Enabled/disabled state of the IRQ will be
+ * restored to that specified by previously_enabled.
+ */
+
+int remove_irq_handler ( irq_t irq, segoff_t *handler,
+			 uint8_t *previously_enabled,
+			 segoff_t *previous_handler ) {
+	segoff_t *irq_vector = IRQ_VECTOR ( irq );
+
+	if ( irq > IRQ_MAX ) {
+		DBG ( "Invalid IRQ number %d\n" );
+		return 0;
+	}
+	if ( ( irq_vector->segment != handler->segment ) ||
+	     ( irq_vector->offset != handler->offset ) ) {
+		DBG ( "Cannot remove handler for IRQ %d\n" );
+		return 0;
+	}
+
+	DBG ( "Removing handler for IRQ %d\n", irq );
+	disable_irq ( irq );
+	irq_vector->segment = previous_handler->segment;
+	irq_vector->offset = previous_handler->offset;
+	if ( *previously_enabled ) enable_irq ( irq );
+	return 1;
+}
+
+/* Install the trivial IRQ handler.  This routine installs the
+ * handler, tests it and enables the IRQ.
+ */
+
+int install_trivial_irq_handler ( irq_t irq ) {
+	segoff_t trivial_irq_handler_segoff = SEGOFF(trivial_irq_handler);
+	
+	if ( trivial_irq_installed_on != IRQ_NONE ) {
+		DBG ( "Can install trivial IRQ handler only once\n" );
+		return 0;
+	}
+	if ( SEGMENT(trivial_irq_handler) > 0xffff ) {
+		DBG ( "Trivial IRQ handler not in base memory\n" );
+		return 0;
+	}
+
+	DBG ( "Installing trivial IRQ handler on IRQ %d\n", irq );
+	if ( ! install_irq_handler ( irq, &trivial_irq_handler_segoff,
+				     trivial_irq_chain,
+				     trivial_irq_chain_to ) )
+		return 0;
+	trivial_irq_installed_on = irq;
+
+	DBG ( "Testing trivial IRQ handler\n" );
+	disable_irq ( irq );
+	*trivial_irq_trigger_count = 0;
+	trivial_irq_previous_trigger_count = 0;
+	fake_irq ( irq );
+	if ( ! trivial_irq_triggered ( irq ) ) {
+		DBG ( "Installation of trivial IRQ handler failed\n" );
+		remove_trivial_irq_handler ( irq );
+		return 0;
+	}
+	/* Send EOI just in case there was a leftover interrupt */
+	send_specific_eoi ( irq );
+	DBG ( "Trivial IRQ handler installed successfully\n" );
+	enable_irq ( irq );
+	return 1;
+}
+
+/* Remove the trivial IRQ handler.
+ */
+
+int remove_trivial_irq_handler ( irq_t irq ) {
+	segoff_t trivial_irq_handler_segoff = SEGOFF(trivial_irq_handler);
+
+	if ( trivial_irq_installed_on == IRQ_NONE ) return 1;
+	if ( irq != trivial_irq_installed_on ) {
+		DBG ( "Cannot uninstall trivial IRQ handler from IRQ %d; "
+		      "is installed on IRQ %d\n", irq,
+		      trivial_irq_installed_on );
+		return 0;
+	}
+
+	if ( ! remove_irq_handler ( irq, &trivial_irq_handler_segoff,
+				    trivial_irq_chain,
+				    trivial_irq_chain_to ) )
+		return 0;
+
+	if ( trivial_irq_triggered ( trivial_irq_installed_on ) ) {
+		DBG ( "Sending EOI for unwanted trivial IRQ\n" );
+		send_specific_eoi ( trivial_irq_installed_on );
+	}
+
+	trivial_irq_installed_on = IRQ_NONE;
+	return 1;
+}
+
+/* Safe method to detect whether or not trivial IRQ has been
+ * triggered.  Using this call avoids potential race conditions.  This
+ * call will return success only once per trigger.
+ */
+
+int trivial_irq_triggered ( irq_t irq ) {
+	uint16_t trivial_irq_this_trigger_count = *trivial_irq_trigger_count;
+	int triggered = ( trivial_irq_this_trigger_count -
+			  trivial_irq_previous_trigger_count );
+	
+	/* irq is not used at present, but we have it in the API for
+	 * future-proofing; in case we want the facility to have
+	 * multiple trivial IRQ handlers installed simultaneously.
+	 *
+	 * Avoid compiler warning about unused variable.
+	 */
+	if ( irq == IRQ_NONE ) {};
+	
+	trivial_irq_previous_trigger_count = trivial_irq_this_trigger_count;
+	return triggered ? 1 : 0;
+}
+
+/* Copy trivial IRQ handler to a new location.  Typically used to copy
+ * the handler into base memory; when relocation is being used we need
+ * to do this before installing the handler.
+ *
+ * Call with target=NULL in order to restore the handler to its
+ * original location.
+ */
+
+int copy_trivial_irq_handler ( void *target, size_t target_size ) {
+	irq_t currently_installed_on = trivial_irq_installed_on;
+	uint32_t offset = ( target == NULL ? 0 :
+			    target - (void*)_trivial_irq_handler );
+
+	if (( target != NULL ) && ( target_size < TRIVIAL_IRQ_HANDLER_SIZE )) {
+		DBG ( "Insufficient space to copy trivial IRQ handler\n" );
+		return 0;
+	}
+
+	if ( currently_installed_on != IRQ_NONE ) {
+		DBG ("WARNING: relocating trivial IRQ handler while in use\n");
+		if ( ! remove_trivial_irq_handler ( currently_installed_on ) )
+			return 0;
+	}
+
+	/* Do the actual copy */
+	if ( target != NULL ) {
+		DBG ( "Copying trivial IRQ handler to %hx:%hx\n",
+		      SEGMENT(target), OFFSET(target) );
+		memcpy ( target, _trivial_irq_handler,
+			 TRIVIAL_IRQ_HANDLER_SIZE );
+	} else {
+		DBG ( "Restoring trivial IRQ handler to original location\n" );
+	}
+	/* Update all the pointers to structures within the handler */
+	trivial_irq_handler = ( void (*)P((void)) )
+		( (void*)_trivial_irq_handler + offset );
+	trivial_irq_trigger_count = (uint16_t*)
+		( (void*)&_trivial_irq_trigger_count + offset );
+	trivial_irq_chain_to = (segoff_t*)
+		( (void*)&_trivial_irq_chain_to + offset );
+	trivial_irq_chain = (uint8_t*)
+		( (void*)&_trivial_irq_chain + offset );
+
+	if ( currently_installed_on != IRQ_NONE ) {
+		if ( ! install_trivial_irq_handler ( currently_installed_on ) )
+			return 0;
+	}
+	return 1;
+}
+
+/* Send non-specific EOI(s).  This seems to be inherently unsafe.
+ */
+
+void send_nonspecific_eoi ( irq_t irq ) {
+	DBG ( "Sending non-specific EOI for IRQ %d\n", irq );
+	if ( irq >= IRQ_PIC_CUTOFF ) {
+		outb ( ICR_EOI_NON_SPECIFIC, PIC2_ICR );
+	}		
+	outb ( ICR_EOI_NON_SPECIFIC, PIC1_ICR );
+}
+
+/* Send specific EOI(s).
+ */
+
+void send_specific_eoi ( irq_t irq ) {
+	DBG ( "Sending specific EOI for IRQ %d\n", irq );
+	outb ( ICR_EOI_SPECIFIC | ICR_VALUE(irq), ICR_REG(irq) );
+	if ( irq >= IRQ_PIC_CUTOFF ) {
+		outb ( ICR_EOI_SPECIFIC | ICR_VALUE(CHAINED_IRQ),
+		       ICR_REG(CHAINED_IRQ) );
+	}
+}
+
+/* Fake an IRQ
+ */
+
+void fake_irq ( irq_t irq ) {
+	struct {
+		uint16_t int_number;
+	} PACKED in_stack;
+
+	/* Convert IRQ to INT number:
+	 *
+	 * subb	$0x08,%cl	Invert bit 3, set bits 4-7 iff irq < 8
+	 * xorb	$0x70,%cl	Invert bits 4-6
+	 * andb	$0x7f,%cl	Clear bit 7
+	 *
+	 * No, it's not the most intuitive method, but I was proud to
+	 * get it down to three lines of assembler when this routine
+	 * was originally implemented in pcbios.S.
+	 */
+	in_stack.int_number = ( ( irq - 8 ) ^ 0x70 ) & 0x7f;
+
+	RM_FRAGMENT(rm_fake_irq,
+		"popw %ax\n\t"		/* %ax = INT number */
+		"call 1f\n1:\tpop %bx\n\t"
+		"movb %al, %cs:(2f-1b+1)(%bx)\n\t" /* Overwrite INT number..*/
+		"\n2:\tint $0x00\n\t"		    /* ..in this instruction */
+	);
+
+	real_call ( rm_fake_irq, &in_stack, NULL );
+}
+
+/* Dump current 8259 status: enabled IRQs and handler addresses.
+ */
+
+#ifdef DEBUG_IRQ
+void dump_irq_status ( void ) {
+	int irq = 0;
+	
+	for ( irq = 0; irq < 16; irq++ ) {
+		if ( irq_enabled ( irq ) ) {
+			printf ( "IRQ%d enabled, ISR at %hx:%hx\n", irq,
+				 IRQ_VECTOR(irq)->segment,
+				 IRQ_VECTOR(irq)->offset );
+		}
+	}
+}
+#endif

src/arch/i386/core/prefixudata.lds

@@ -0,0 +1,8 @@
+OUTPUT_FORMAT("elf32-i386", "elf32-i386", "elf32-i386")
+OUTPUT_ARCH(i386)
+
+SECTIONS {
+	 .prefix.udata : {
+		      *(*)
+	 }
+}

src/arch/i386/core/prefixzdata.lds

@@ -0,0 +1,8 @@
+OUTPUT_FORMAT("elf32-i386", "elf32-i386", "elf32-i386")
+OUTPUT_ARCH(i386)
+
+SECTIONS {
+	 .prefix.zdata : {
+		      *(*)
+	 }
+}

src/arch/i386/core/pxe_callbacks.c

@@ -0,0 +1,364 @@
+/* PXE callback mechanisms.  This file contains only the portions
+ * specific to i386: i.e. the low-level mechanisms for calling in from
+ * an NBP to the PXE stack and for starting an NBP from the PXE stack.
+ */
+
+#ifdef PXE_EXPORT
+
+#include "etherboot.h"
+#include "callbacks.h"
+#include "realmode.h"
+#include "pxe.h"
+#include "pxe_callbacks.h"
+#include "pxe_export.h"
+#include "hidemem.h"
+#include <stdarg.h>
+
+#define INSTALLED(x) ( (typeof(&x)) ( (void*)(&x) \
+				      - &pxe_callback_interface \
+				      + (void*)&pxe_stack->arch_data ) )
+#define pxe_intercept_int1a	INSTALLED(_pxe_intercept_int1a)
+#define pxe_intercepted_int1a	INSTALLED(_pxe_intercepted_int1a)
+#define pxe_pxenv_location	INSTALLED(_pxe_pxenv_location)
+#define INT1A_VECTOR ( (segoff_t*) ( phys_to_virt( 4 * 0x1a ) ) )
+
+/* The overall size of the PXE stack is ( sizeof(pxe_stack_t) +
+ * pxe_callback_interface_size + rm_callback_interface_size ).
+ * Unfortunately, this isn't a compile-time constant, since
+ * {pxe,rm}_callback_interface_size depend on the length of the
+ * assembly code in these interfaces.
+ *
+ * We used to have a function pxe_stack_size() which returned this
+ * value.  However, it actually needs to be a link-time constant, so
+ * that it can appear in the UNDIROMID structure in romprefix.S.  We
+ * therefore export the three component sizes as absolute linker
+ * symbols, get the linker to add them together and generate a new
+ * absolute symbol _pxe_stack_size.  We then import this value into a
+ * C variable pxe_stack_size, for access from C code.
+ */
+
+/* gcc won't let us use extended asm outside a function (compiler
+ * bug), ao we have to put these asm statements inside a dummy
+ * function.
+ */
+static void work_around_gcc_bug ( void ) __attribute__ ((used));
+static void work_around_gcc_bug ( void ) {
+	/* Export sizeof(pxe_stack_t) as absolute linker symbol */
+	__asm__ ( ".globl _pxe_stack_t_size" );
+	__asm__ ( ".equ _pxe_stack_t_size, %c0"
+		  : : "i" (sizeof(pxe_stack_t)) );
+}
+/* Import _pxe_stack_size absolute linker symbol into C variable */
+extern int pxe_stack_size;
+__asm__ ( "pxe_stack_size: .long _pxe_stack_size" );
+
+/* Utility routine: byte checksum
+ */
+uint8_t byte_checksum ( void *address, size_t size ) {
+	unsigned int i, sum = 0;
+
+	for ( i = 0; i < size; i++ ) {
+		sum += ((uint8_t*)address)[i];
+	}
+	return (uint8_t)sum;
+}
+
+/* install_pxe_stack(): install PXE stack.
+ * 
+ * Use base = NULL for auto-allocation of base memory
+ *
+ * IMPORTANT: no further allocation of base memory should take place
+ * before the PXE stack is removed.  This is to work around a small
+ * but important deficiency in the PXE specification.
+ */
+pxe_stack_t * install_pxe_stack ( void *base ) {
+	pxe_t *pxe;
+	pxenv_t *pxenv;
+	void *pxe_callback_code;
+	void (*pxe_in_call_far)(void);
+	void (*pxenv_in_call_far)(void);
+	void *rm_callback_code;
+	void *e820mangler_code;
+	void *end;
+
+	/* If already installed, just return */
+	if ( pxe_stack != NULL ) return pxe_stack;
+
+	/* Allocate base memory if requested to do so
+	 */
+	if ( base == NULL ) {
+		base = allot_base_memory ( pxe_stack_size );
+		if ( base == NULL ) return NULL;
+	}
+
+	/* Round address up to 16-byte physical alignment */
+	pxe_stack = (pxe_stack_t *)
+		( phys_to_virt ( ( virt_to_phys(base) + 0xf ) & ~0xf ) );
+	/* Zero out allocated stack */
+	memset ( pxe_stack, 0, sizeof(*pxe_stack) );
+	
+	/* Calculate addresses for portions of the stack */
+	pxe = &(pxe_stack->pxe);
+	pxenv = &(pxe_stack->pxenv);
+	pxe_callback_code = &(pxe_stack->arch_data);
+	pxe_in_call_far = _pxe_in_call_far +  
+		( pxe_callback_code - &pxe_callback_interface );
+	pxenv_in_call_far = _pxenv_in_call_far +
+		( pxe_callback_code - &pxe_callback_interface );
+	rm_callback_code = pxe_callback_code + pxe_callback_interface_size;
+	
+	e820mangler_code = (void*)(((int)rm_callback_code +
+				    rm_callback_interface_size + 0xf ) & ~0xf);
+	end = e820mangler_code + e820mangler_size;
+
+	/* Initialise !PXE data structures */
+	memcpy ( pxe->Signature, "!PXE", 4 );
+	pxe->StructLength = sizeof(*pxe);
+	pxe->StructRev = 0;
+	pxe->reserved_1 = 0;
+	/* We don't yet have an UNDI ROM ID structure */
+	pxe->UNDIROMID.segment = 0;
+	pxe->UNDIROMID.offset = 0;
+	/* or a BC ROM ID structure */
+	pxe->BaseROMID.segment = 0;
+	pxe->BaseROMID.offset = 0;
+	pxe->EntryPointSP.segment = SEGMENT(pxe_stack);
+	pxe->EntryPointSP.offset = (void*)pxe_in_call_far - (void*)pxe_stack;
+	/* No %esp-compatible entry point yet */
+	pxe->EntryPointESP.segment = 0;
+	pxe->EntryPointESP.offset = 0;
+	pxe->StatusCallout.segment = -1;
+	pxe->StatusCallout.offset = -1;
+	pxe->reserved_2 = 0;
+	pxe->SegDescCn = 7;
+	pxe->FirstSelector = 0;
+	/* PXE specification doesn't say anything about when the stack
+	 * space should get freed.  We work around this by claiming it
+	 * as our data segment as well.
+	 */
+	pxe->Stack.Seg_Addr = pxe->UNDIData.Seg_Addr = real_mode_stack >> 4;
+	pxe->Stack.Phy_Addr = pxe->UNDIData.Phy_Addr = real_mode_stack;
+	pxe->Stack.Seg_Size = pxe->UNDIData.Seg_Size = real_mode_stack_size;
+	/* Code segment has to be the one containing the data structures... */
+	pxe->UNDICode.Seg_Addr = SEGMENT(pxe_stack);
+	pxe->UNDICode.Phy_Addr = virt_to_phys(pxe_stack);
+	pxe->UNDICode.Seg_Size = end - (void*)pxe_stack;
+	/* No base code loaded */
+	pxe->BC_Data.Seg_Addr = 0;
+	pxe->BC_Data.Phy_Addr = 0;
+	pxe->BC_Data.Seg_Size = 0;
+	pxe->BC_Code.Seg_Addr = 0;
+	pxe->BC_Code.Phy_Addr = 0;
+	pxe->BC_Code.Seg_Size = 0;
+	pxe->BC_CodeWrite.Seg_Addr = 0;
+	pxe->BC_CodeWrite.Phy_Addr = 0;
+	pxe->BC_CodeWrite.Seg_Size = 0;
+	pxe->StructCksum -= byte_checksum ( pxe, sizeof(*pxe) );
+
+	/* Initialise PXENV+ data structures */
+	memcpy ( pxenv->Signature, "PXENV+", 6 );
+	pxenv->Version = 0x201;
+	pxenv->Length = sizeof(*pxenv);
+	pxenv->RMEntry.segment = SEGMENT(pxe_stack);
+	pxenv->RMEntry.offset = (void*)pxenv_in_call_far - (void*)pxe_stack;
+	pxenv->PMOffset = 0; /* "Do not use" says the PXE spec */
+	pxenv->PMSelector = 0; /* "Do not use" says the PXE spec */
+	pxenv->StackSeg = pxenv->UNDIDataSeg = real_mode_stack >> 4;
+	pxenv->StackSize = pxenv->UNDIDataSize = real_mode_stack_size;
+	pxenv->BC_CodeSeg = 0;
+	pxenv->BC_CodeSize = 0;
+	pxenv->BC_DataSeg = 0;
+	pxenv->BC_DataSize = 0;
+	/* UNDIData{Seg,Size} set above */
+	pxenv->UNDICodeSeg = SEGMENT(pxe_stack);
+	pxenv->UNDICodeSize = end - (void*)pxe_stack;
+	pxenv->PXEPtr.segment = SEGMENT(pxe);
+	pxenv->PXEPtr.offset = OFFSET(pxe);
+	pxenv->Checksum -= byte_checksum ( pxenv, sizeof(*pxenv) );
+
+	/* Mark stack as inactive */
+	pxe_stack->state = CAN_UNLOAD;
+
+	/* Install PXE and RM callback code and E820 mangler */
+	memcpy ( pxe_callback_code, &pxe_callback_interface,
+		 pxe_callback_interface_size );
+	install_rm_callback_interface ( rm_callback_code, 0 );
+	install_e820mangler ( e820mangler_code );
+
+	return pxe_stack;
+}
+
+/* Use the UNDI data segment as our real-mode stack.  This is for when
+ * we have been loaded via the UNDI loader
+ */
+void use_undi_ds_for_rm_stack ( uint16_t ds ) {
+	forget_real_mode_stack();
+	real_mode_stack = virt_to_phys ( VIRTUAL ( ds, 0 ) );
+	lock_real_mode_stack = 1;
+}
+
+/* Activate PXE stack (i.e. hook interrupt vectors).  The PXE stack
+ * *can* be used before it is activated, but it really shoudln't.
+ */
+int hook_pxe_stack ( void ) {
+	if ( pxe_stack == NULL ) return 0;
+	if ( pxe_stack->state >= MIDWAY ) return 1;
+
+	/* Hook INT15 handler */
+	hide_etherboot();
+
+	/* Hook INT1A handler */
+	*pxe_intercepted_int1a = *INT1A_VECTOR;
+	pxe_pxenv_location->segment = SEGMENT(pxe_stack);
+	pxe_pxenv_location->offset = (void*)&pxe_stack->pxenv
+		- (void*)pxe_stack;
+	INT1A_VECTOR->segment = SEGMENT(&pxe_stack->arch_data);
+	INT1A_VECTOR->offset = (void*)pxe_intercept_int1a
+		- (void*)&pxe_stack->arch_data;
+
+	/* Mark stack as active */
+	pxe_stack->state = MIDWAY;
+	return 1;
+}
+
+/* Deactivate the PXE stack (i.e. unhook interrupt vectors).
+ */
+int unhook_pxe_stack ( void ) {
+	if ( pxe_stack == NULL ) return 0;
+	if ( pxe_stack->state <= CAN_UNLOAD ) return 1;
+
+	/* Restore original INT15 and INT1A handlers */
+	*INT1A_VECTOR = *pxe_intercepted_int1a;
+	if ( !unhide_etherboot() ) {
+		/* Cannot unhook INT15.  We're up the creek without
+		 * even a suitable log out of which to fashion a
+		 * paddle.  There are some very badly behaved NBPs
+		 * that will ignore plaintive pleas such as
+		 * PXENV_KEEP_UNDI and just zero out our code anyway.
+		 * This means they end up vapourising an active INT15
+		 * handler, which is generally not a good thing to do.
+		 */
+		return 0;
+	}
+
+	/* Mark stack as inactive */
+	pxe_stack->state = CAN_UNLOAD;
+	return 1;
+}
+
+/* remove_pxe_stack(): remove PXE stack installed by install_pxe_stack()
+ */
+void remove_pxe_stack ( void ) {
+	/* Ensure stack is deactivated, then free up the memory */
+	if ( ensure_pxe_state ( CAN_UNLOAD ) ) {
+		forget_base_memory ( pxe_stack, pxe_stack_size );
+		pxe_stack = NULL;
+	} else {
+		printf ( "Cannot remove PXE stack!\n" );
+	}
+}
+
+/* xstartpxe(): start up a PXE image
+ */
+int xstartpxe ( void ) {
+	int nbp_exit;
+	struct {
+		reg16_t bx;
+		reg16_t es;
+		segoff_t pxe;
+	} PACKED in_stack;
+	
+	/* Set up registers and stack parameters to pass to PXE NBP */
+	in_stack.es.word = SEGMENT(&(pxe_stack->pxenv));
+	in_stack.bx.word = OFFSET(&(pxe_stack->pxenv));
+	in_stack.pxe.segment = SEGMENT(&(pxe_stack->pxe));
+	in_stack.pxe.offset = OFFSET(&(pxe_stack->pxe));
+
+	/* Real-mode trampoline fragment used to jump to PXE NBP
+	 */
+	RM_FRAGMENT(jump_to_pxe_nbp, 
+		"popw %bx\n\t"
+		"popw %es\n\t"
+		"lcall $" RM_STR(PXE_LOAD_SEGMENT) ", $" RM_STR(PXE_LOAD_OFFSET) "\n\t"
+	);
+
+	/* Call to PXE image */
+	gateA20_unset();
+	nbp_exit = real_call ( jump_to_pxe_nbp, &in_stack, NULL );
+	gateA20_set();
+
+	return nbp_exit;
+}
+
+int pxe_in_call ( in_call_data_t *in_call_data, va_list params ) {
+	/* i386 calling conventions; the only two defined by Intel's
+	 * PXE spec.
+	 *
+	 * Assembly code must pass a long containing the PXE version
+	 * code (i.e. 0x201 for !PXE, 0x200 for PXENV+) as the first
+	 * parameter after the in_call opcode.  This is used to decide
+	 * whether to take parameters from the stack (!PXE) or from
+	 * registers (PXENV+).
+	 */
+	uint32_t api_version = va_arg ( params, typeof(api_version) );
+	uint16_t opcode;
+	segoff_t segoff;
+	t_PXENV_ANY *structure;
+		
+	if ( api_version >= 0x201 ) {
+		/* !PXE calling convention */
+		pxe_call_params_t pxe_params
+			= va_arg ( params, typeof(pxe_params) );
+		opcode = pxe_params.opcode;
+		segoff = pxe_params.segoff;
+	} else {
+		/* PXENV+ calling convention */
+		opcode = in_call_data->pm->regs.bx;
+		segoff.segment = in_call_data->rm->seg_regs.es;
+		segoff.offset = in_call_data->pm->regs.di;
+	}
+	structure = VIRTUAL ( segoff.segment, segoff.offset );
+	return pxe_api_call ( opcode, structure );
+}
+
+#ifdef TEST_EXCLUDE_ALGORITHM
+/* This code retained because it's a difficult algorithm to tweak with
+ * confidence
+ */
+int ___test_exclude ( int start, int len, int estart, int elen, int fixbase );
+void __test_exclude ( int start, int len, int estart, int elen, int fixbase ) {
+	int newrange = ___test_exclude ( start, len, estart, elen, fixbase );
+	int newstart = ( newrange >> 16 ) & 0xffff;
+	int newlen = ( newrange & 0xffff );
+
+	printf ( "[%x,%x): excluding [%x,%x) %s gives [%x,%x)\n",
+		 start, start + len,
+		 estart, estart + elen,
+		 ( fixbase == 0 ) ? "  " : "fb",
+		 newstart, newstart + newlen );
+}
+void _test_exclude ( int start, int len, int estart, int elen ) {
+	__test_exclude ( start, len, estart, elen, 0 );
+	__test_exclude ( start, len, estart, elen, 1 );
+}
+void test_exclude ( void ) {
+	_test_exclude ( 0x8000, 0x1000, 0x0400, 0x200 ); /* before */
+	_test_exclude ( 0x8000, 0x1000, 0x9000, 0x200 ); /* after */
+	_test_exclude ( 0x8000, 0x1000, 0x7f00, 0x200 ); /* before overlap */
+	_test_exclude ( 0x8000, 0x1000, 0x8f00, 0x200 ); /* after overlap */
+	_test_exclude ( 0x8000, 0x1000, 0x8000, 0x200 ); /* align start */
+	_test_exclude ( 0x8000, 0x1000, 0x8e00, 0x200 ); /* align end */
+	_test_exclude ( 0x8000, 0x1000, 0x8100, 0x200 ); /* early overlap */
+	_test_exclude ( 0x8000, 0x1000, 0x8d00, 0x200 ); /* late overlap */
+	_test_exclude ( 0x8000, 0x1000, 0x7000, 0x3000 ); /* total overlap */
+	_test_exclude ( 0x8000, 0x1000, 0x8000, 0x1000 ); /* exact overlap */
+}
+#endif /* TEST_EXCLUDE_ALGORITHM */
+
+#else /* PXE_EXPORT */
+
+/* Define symbols used by the linker scripts, to prevent link errors */
+__asm__ ( ".globl _pxe_stack_t_size" );
+__asm__ ( ".equ _pxe_stack_t_size, 0" );
+
+#endif /* PXE_EXPORT */

src/arch/i386/core/pxe_loader.c

@@ -0,0 +1,94 @@
+/*
+ * PXE image loader for Etherboot.
+ * 
+ * Note: There is no signature check for PXE images because there is
+ * no signature.  Well done, Intel!  Consequently, pxe_probe() must be
+ * called last of all the image_probe() routines, because it will
+ * *always* claim the image.
+ */
+
+#ifndef PXE_EXPORT
+#error PXE_IMAGE requires PXE_EXPORT
+#endif
+
+#include "etherboot.h"
+#include "pxe_callbacks.h"
+#include "pxe_export.h"
+#include "pxe.h"
+
+unsigned long pxe_load_offset;
+
+static sector_t pxe_download ( unsigned char *data,
+			       unsigned int len, int eof );
+
+static inline os_download_t pxe_probe ( unsigned char *data __unused,
+					unsigned int len __unused ) {
+	printf("(PXE)");
+	pxe_load_offset = 0;
+	return pxe_download;
+}
+
+static sector_t pxe_download ( unsigned char *data,
+			       unsigned int len, int eof ) {
+	unsigned long block_address = PXE_LOAD_ADDRESS + pxe_load_offset;
+	PXENV_STATUS_t nbp_exit;
+
+	/* Check segment will fit.  We can't do this in probe()
+	 * because there's nothing in the non-existent header to tell
+	 * us how long the image is.
+	 */
+	if ( ! prep_segment ( block_address, block_address + len, 
+			      block_address + len,
+			      pxe_load_offset, pxe_load_offset + len ) ) {
+		longjmp ( restart_etherboot, -2 );
+	}
+
+	/* Load block into memory, continue loading until eof */
+	memcpy ( phys_to_virt ( block_address ), data, len );
+	pxe_load_offset += len;
+	if ( ! eof ) { 
+		return 0;
+	}
+
+	/* Start up PXE NBP */
+	done ( 0 );
+
+	/* Install and activate a PXE stack */
+	pxe_stack = install_pxe_stack ( NULL );
+	if ( ensure_pxe_state ( READY ) ) {
+		/* Invoke the NBP */
+		nbp_exit = xstartpxe();
+	} else {
+		/* Fake success so we tear down the stack */
+		nbp_exit = PXENV_STATUS_SUCCESS;
+	}
+
+	/* NBP has three exit codes:
+	 *   PXENV_STATUS_KEEP_UNDI : keep UNDI and boot next device
+	 *   PXENV_STATUS_KEEP_ALL  : keep all and boot next device
+	 *   anything else : remove all and boot next device
+	 * 
+	 * Strictly, we're meant to hand back to the BIOS, but this
+	 * would prevent the useful combination of "PXE NBP fails, so
+	 * let Etherboot try to boot its next device".  We therefore
+	 * take liberties.
+	 */
+	if ( nbp_exit != PXENV_STATUS_KEEP_UNDI &&
+	     nbp_exit != PXENV_STATUS_KEEP_ALL ) {
+		/* Tear down PXE stack */
+		remove_pxe_stack();
+	}
+
+	/* Boot next device.  Under strict PXE compliance, exit back
+	 * to the BIOS, otherwise let Etherboot move to the next
+	 * device.
+	 */
+#ifdef PXE_STRICT
+	longjmp ( restart_etherboot, 255 );
+#else
+	longjmp ( restart_etherboot, 4 );
+#endif
+	
+	/* Never reached; avoid compiler warning */
+	return ( 0 );
+}

src/arch/i386/core/realmode.c

@@ -0,0 +1,148 @@
+/* Real-mode interface: C portions.
+ *
+ * Initial version by Michael Brown <mbrown@fensystems.co.uk>, January 2004.
+ */
+
+#include "etherboot.h"
+#include "realmode.h"
+#include "segoff.h"
+
+#define RM_STACK_SIZE ( 0x1000 )
+
+/* gcc won't let us use extended asm outside a function (compiler
+ * bug), ao we have to put these asm statements inside a dummy
+ * function.
+ */
+static void work_around_gcc_bug ( void ) __attribute__ ((used));
+static void work_around_gcc_bug ( void ) {
+	/* Export _real_mode_stack_size as absolute linker symbol */
+	__asm__ ( ".globl _real_mode_stack_size" );
+	__asm__ ( ".equ _real_mode_stack_size, %c0" : : "i" (RM_STACK_SIZE) );
+}
+
+/* While Etherboot remains in base memory the real-mode stack is
+ * placed in the Etherboot main stack.  The first allocation or
+ * deallocation of base memory will cause a 'proper' real-mode stack
+ * to be allocated.  This will happen before Etherboot is relocated to
+ * high memory.
+ */
+uint32_t real_mode_stack = 0;
+size_t real_mode_stack_size = RM_STACK_SIZE;
+int lock_real_mode_stack = 0;	/* Set to make stack immobile */
+
+/* Make a call to a real-mode code block.
+ */
+
+/* These is the structure that exists on the stack as the paramters
+ * passed in to _real_call.  We pass a pointer to this struct to
+ * prepare_real_call(), to save stack space.
+ */
+typedef struct {
+	void *fragment;
+	int fragment_len;
+	void *in_stack;
+	int in_stack_len;
+	void *out_stack;
+	int out_stack_len;
+} real_call_params_t;
+
+uint32_t prepare_real_call ( real_call_params_t *p,
+			     int local_stack_len, char *local_stack ) {
+	char *stack_base;
+	char *stack_end;
+	char *stack;
+	char *s;
+	prot_to_real_params_t *p2r_params;
+	real_to_prot_params_t *r2p_params;
+
+	/* Work out where we're putting the stack */
+	if ( virt_to_phys(local_stack) < 0xa0000 ) {
+		/* Current stack is in base memory.  We can therefore
+		 * use it directly, with a constraint on the size that
+		 * we don't know; assume that we can use up to
+		 * real_mode_stack_size.  (Not a valid assumption, but
+		 * it will do).
+		 */
+		stack_end = local_stack + local_stack_len;
+		stack_base = stack_end - real_mode_stack_size;
+	} else {
+		if (!real_mode_stack) {
+			allot_real_mode_stack();
+		}
+		/* Use the allocated real-mode stack in base memory.
+		 * This has fixed start and end points.
+		 */
+		stack_base = phys_to_virt(real_mode_stack);
+		stack_end = stack_base + real_mode_stack_size;
+	}
+	s = stack = stack_end - local_stack_len;
+
+	/* Compile input stack and trampoline code to stack */
+	if ( p->in_stack_len ) {
+		memcpy ( s, p->in_stack, p->in_stack_len );
+		s += p->in_stack_len;
+	}
+	memcpy ( s, _prot_to_real_prefix, prot_to_real_prefix_size );
+	s += prot_to_real_prefix_size;
+	p2r_params = (prot_to_real_params_t*) ( s - sizeof(*p2r_params) );
+	memcpy ( s, p->fragment, p->fragment_len );
+	s += p->fragment_len;
+	memcpy ( s, _real_to_prot_suffix, real_to_prot_suffix_size );
+	s += real_to_prot_suffix_size;
+	r2p_params = (real_to_prot_params_t*) ( s - sizeof(*r2p_params) );
+
+	/* Set parameters within compiled stack */
+	p2r_params->ss = p2r_params->cs = SEGMENT ( stack_base );
+	p2r_params->esp = virt_to_phys ( stack );
+	p2r_params->r2p_params = virt_to_phys ( r2p_params );
+	r2p_params->out_stack = ( p->out_stack == NULL ) ?
+		0 : virt_to_phys ( p->out_stack );
+	r2p_params->out_stack_len = p->out_stack_len;
+
+	return virt_to_phys ( stack + p->in_stack_len );
+}
+
+
+/* Parameters are not genuinely unused; they are passed to
+ * prepare_real_call() as part of a real_call_params_t struct.
+ */
+uint16_t _real_call ( void *fragment, int fragment_len,
+		      void *in_stack __unused, int in_stack_len,
+		      void *out_stack __unused, int out_stack_len __unused ) {
+	uint16_t retval;
+
+	/* This code is basically equivalent to
+	 *
+	 *	uint32_t trampoline;
+	 *	char local_stack[ in_stack_len + prot_to_real_prefix_size +
+	 *			  fragment_len + real_to_prot_suffix_size ];
+	 *	trampoline = prepare_real_call ( &fragment, local_stack );
+	 *	__asm__ ( "call _virt_to_phys\n\t"
+	 *		  "call %%eax\n\t"
+	 *		  "call _phys_to_virt\n\t"
+	 *		  : "=a" (retval) : "0" (trampoline) );
+	 *
+	 * but implemented in assembly to avoid problems with not
+	 * being certain exactly how gcc handles %esp.
+	 */
+
+	__asm__ ( "pushl %%ebp\n\t"
+		  "movl  %%esp, %%ebp\n\t"	/* %esp preserved via %ebp */
+		  "subl  %%ecx, %%esp\n\t"	/* space for inline RM stack */
+		  "pushl %%esp\n\t"		/* set up RM stack */
+		  "pushl %%ecx\n\t"
+		  "pushl %%eax\n\t"
+		  "call  prepare_real_call\n\t"	/* %eax = trampoline addr */
+		  "addl  $12, %%esp\n\t"
+		  "call  _virt_to_phys\n\t"	/* switch to phys addr */
+		  "call  *%%eax\n\t"		/* call to trampoline */
+		  "call  _phys_to_virt\n\t"	/* switch to virt addr */
+		  "movl  %%ebp, %%esp\n\t"	/* restore %esp & %ebp */
+		  "popl  %%ebp\n\t"
+		  : "=a" ( retval )
+		  : "0" ( &fragment )
+		  , "c" ( ( ( in_stack_len + prot_to_real_prefix_size +
+			      fragment_len + real_to_prot_suffix_size )
+			    + 0x3 ) & ~0x3 ) );
+	return retval;
+}

src/arch/i386/core/realmode_asm.S

@@ -0,0 +1,695 @@
+/* Real-mode interface: assembly-language portions.
+ *
+ * Initial version by Michael Brown <mbrown@fensystems.co.uk>, January 2004.
+ */
+
+#include "realmode.h"
+#include "callbacks.h"
+
+#if	1	/* CODE16 */
+	
+#define BOCHSBP xchgw %bx,%bx
+
+#define NUM_PUSHA_REGS (8)
+#define NUM_SEG_REGS (6)
+
+	.text
+	.arch i386
+	.section ".text16.nocompress", "ax", @progbits
+	.code16
+	
+	.equ	CR0_PE,1
+
+#ifdef	GAS291
+#define DATA32 data32;
+#define ADDR32 addr32;
+#define	LJMPI(x)	ljmp	x
+#else
+#define DATA32 data32
+#define ADDR32 addr32
+/* newer GAS295 require #define	LJMPI(x)	ljmp	*x */
+#define	LJMPI(x)	ljmp	x
+#endif
+
+/****************************************************************************
+ * REAL-MODE CALLBACK INTERFACE
+ *
+ * This must be copied down to base memory in order for external
+ * programs to be able to make calls in to Etherboot.  Store the
+ * current physical address of Etherboot (i.e. virt_to_phys(_text)) in
+ * (uint32_t)rm_etherboot_location, then copy
+ * (uint16_t)rm_callback_interface_size bytes starting at
+ * &((void)rm_callback_interface).
+ *
+ * There are two defined entry points:
+ *   Offset RM_IN_CALL     = 0		Near call entry point
+ *   Offset RM_IN_CALL_FAR = 2		Far call entry point
+ *
+ * Note that the routines _prot_to_real and _real_to_prot double as
+ * trampoline fragments for external calls (calls from Etherboot to
+ * real-mode code).  _prot_to_real does not automatically re-enable
+ * interrupts; this is to allow for the potential of using Etherboot
+ * code as an ISR.  _real_to_prot does automatically disable
+ * interrupts, since we don't have a protected-mode IDT.
+ ****************************************************************************
+ */
+
+	.globl	rm_callback_interface
+	.code16
+rm_callback_interface:
+	.globl	_rm_in_call
+_rm_in_call:
+	jmp	_real_in_call
+	.globl	_rm_in_call_far
+_rm_in_call_far:
+	jmp	_real_in_call_far
+
+/****************************************************************************
+ * _real_in_call
+ *
+ * Parameters:
+ *   16-bit real-mode near/far return address (implicit from [l]call
+ *   to routine) Other parameters as for _in_call_far().
+ *
+ * This routine will convert the 16-bit real-mode far return address
+ * to a 32-bit real-mode far return address, switch to protected mode
+ * using _real_to_prot and call in to _in_call_far.
+ ****************************************************************************
+ */
+
+#define RIC_PRESERVE ( 8 )
+#define RIC_OFFSET_CALLADDR ( RIC_PRESERVE )
+#define RIC_OFFSET_CALLADDR_E ( RIC_OFFSET_CALLADDR + 4 )
+#define RIC_OFFSET_CONTADDR ( RIC_OFFSET_CALLADDR_E )
+#define RIC_OFFSET_CONTADDR_E ( RIC_OFFSET_CONTADDR + 4 )
+#define RIC_OFFSET_OPCODE ( RIC_OFFSET_CONTADDR_E )
+#define RIC_OFFSET_OPCODE_E ( RIC_OFFSET_OPCODE + 4 )
+#define RIC_OFFSET_SEG_REGS ( RIC_OFFSET_OPCODE_E )
+#define RIC_OFFSET_SEG_REGS_E ( RIC_OFFSET_SEG_REGS + ( NUM_SEG_REGS * 2 ) )
+#define RIC_OFFSET_PAD ( RIC_OFFSET_SEG_REGS_E )
+#define RIC_OFFSET_PAD_E ( RIC_OFFSET_PAD + 2 )
+#define RIC_OFFSET_FLAGS ( RIC_OFFSET_PAD_E )
+#define RIC_OFFSET_FLAGS_E ( RIC_OFFSET_FLAGS + 2 )
+#define RIC_OFFSET_RETADDR ( RIC_OFFSET_FLAGS_E )
+#define RIC_OFFSET_RETADDR_E ( RIC_OFFSET_RETADDR + 4 )
+#define RIC_OFFSET_ORIG_OPCODE ( RIC_OFFSET_RETADDR_E )
+#define RIC_INSERT_LENGTH ( RIC_OFFSET_OPCODE_E - RIC_OFFSET_CALLADDR )
+	
+	.code16
+_real_in_call:
+	/* Expand near return address to far return address
+	 */
+	pushw	%ax		/* Extend stack, store %ax */
+	pushfw
+	pushw	%bp
+	movw	%sp, %bp
+	movw	%cs, %ax
+	xchgw	%ax, 6(%bp)
+	xchgw	%ax, 4(%bp)	/* also restores %ax */
+	popw	%bp
+	popfw
+	/* Fall through to _real_in_call_far */
+	
+_real_in_call_far:
+	/* Store flags and pad */
+	pushfw
+	pushw	%ax
+
+	/* Store segment registers.  Order matches that of seg_regs_t */
+	pushw	%gs
+	pushw	%fs
+	pushw	%es
+	pushw	%ds
+	pushw	%ss
+	pushw	%cs
+
+	/* Switch to protected mode */
+	call _real_to_prot
+	.code32
+
+	/* Allow space for expanded stack */
+	subl	$RIC_INSERT_LENGTH, %esp
+	
+	/* Store temporary registers */
+	pushl	%ebp
+	pushl	%eax
+
+	/* Copy opcode, set EB_CALL_FROM_REAL_MODE and EP_SKIP_OPCODE.
+	 * Copy it because _in_call() and i386_in_call() expect it at
+	 * a fixed position, not as part of the va_list.
+	 */
+	movl	RIC_OFFSET_ORIG_OPCODE(%esp), %eax
+	orl	$(EB_CALL_FROM_REAL_MODE|EB_SKIP_OPCODE), %eax
+	movl	%eax, RIC_OFFSET_OPCODE(%esp)
+	
+	/* Set up call and return addresses */
+	call	1f
+1:	popl	%ebp
+	subl	$1b, %ebp			/* %ebp = offset */
+	movl	rm_etherboot_location(%ebp), %eax  /* Etherboot phys addr */
+	subl	$_text, %eax
+	addl	$_in_call, %eax			/* _in_call phys addr */
+	movl	%eax, RIC_OFFSET_CALLADDR(%esp)
+	leal	2f(%ebp), %eax			/* continuation address */
+	movl	%eax, RIC_OFFSET_CONTADDR(%esp)
+	
+	/* Restore temporary registers */
+	popl	%eax
+	popl	%ebp
+
+	/* Call to _in_call */
+	ret
+	/* opcode will be popped automatically thanks to EB_SKIP_OPCODE */
+
+2:	/* Continuation point */
+	call	_prot_to_real			/* Return to real mode */
+	/* Note: the first two words of our segment register store
+	 * happens to be exactly what we need to pass as parameters to
+	 * _prot_to_real.
+	 */
+	.code16
+	popw	%ds				/* Restore segment registers */
+	popw	%ds				/* (skip cs&ss since these   */
+	popw	%ds				/* have already been set by  */
+	popw	%es				/* _prot_to_real	     */
+	popw	%fs
+	popw	%gs
+	addw	$2, %sp				/* skip pad */
+
+	/* Check for EB_SKIP_OPCODE */
+	pushw	%bp
+	movw	%sp, %bp
+	testl	$EB_SKIP_OPCODE, 6(%bp)
+	popw	%bp
+	jnz	1f
+	/* Normal return */
+	popfw					/* Restore interrupt status */
+	lret					/* Back to caller */
+1:	/* Return and skip opcode */
+	popfw
+	lret	$4
+
+/****************************************************************************
+ * rm_etherboot_location: the current physical location of Etherboot.
+ * Needed so that real-mode callback routines can locate Etherboot.
+ ****************************************************************************
+ */
+	.globl rm_etherboot_location
+rm_etherboot_location:	.long 0
+		
+/****************************************************************************
+ * _prot_to_real_prefix
+ *
+ * Trampoline fragment.  Switch from 32-bit protected mode with flat
+ * physical addresses to 16-bit real mode.  Store registers in the
+ * trampoline for restoration by _real_to_prot_suffix.  Switch to
+ * stack in base memory.
+ ****************************************************************************
+ */
+	
+	.globl _prot_to_real_prefix
+	.code32
+_prot_to_real_prefix:
+	/* Registers to preserve */
+	pushl	%ebx
+	pushl	%esi
+	pushl	%edi
+	pushl	%ebp
+
+	/* Calculate offset */
+	call	1f
+1:	popl	%ebp
+	subl	$1b, %ebp		/* %ebp = offset for labels in p2r*/
+
+	/* Preserve registers and return address in r2p_params */
+	movl	p2r_r2p_params(%ebp), %ebx
+	subl	$r2p_params, %ebx	/* %ebx = offset for labels in r2p */
+	popl	r2p_ebp(%ebx)
+	popl	r2p_edi(%ebx)
+	popl	r2p_esi(%ebx)
+	popl	r2p_ebx(%ebx)
+	popl	r2p_ret_addr(%ebx)
+	movl	%esp, r2p_esp(%ebx)
+
+	/* Switch stacks */
+	movl	p2r_esp(%ebp), %esp
+
+	/* Switch to real mode */
+	pushl	p2r_segments(%ebp)
+	call	_prot_to_real
+	.code16
+	addw	$4, %sp
+		
+	/* Fall through to next trampoline fragment */
+	jmp	_prot_to_real_prefix_end
+	
+/****************************************************************************
+ * _prot_to_real
+ *
+ * Switch from 32-bit protected mode with flat physical addresses to
+ * 16-bit real mode.  Stack and code must be in base memory when
+ * called.  %cs, %ss, %eip, %esp are changed to real-mode values,
+ * other segment registers are destroyed, all other registers are
+ * preserved.  Interrupts are *not* enabled.
+ *
+ * Parameters:
+ *   %cs		Real-mode code segment (word)
+ *   %ss		Real-mode stack segment (word)
+ ****************************************************************************
+ */
+
+#define P2R_PRESERVE ( 12 )
+#define P2R_OFFSET_RETADDR ( P2R_PRESERVE )
+#define P2R_OFFSET_RETADDR_E ( P2R_OFFSET_RETADDR + 4 )
+#define P2R_OFFSET_CS ( P2R_OFFSET_RETADDR_E )
+#define P2R_OFFSET_CS_E ( P2R_OFFSET_CS + 2 )
+#define P2R_OFFSET_SS ( P2R_OFFSET_CS_E )
+#define P2R_OFFSET_SS_E ( P2R_OFFSET_SS + 2 )
+
+	.globl _prot_to_real
+	.code32
+_prot_to_real:
+	/* Preserve registers */
+	pushl	%ebp
+	pushl	%ebx
+	pushl	%eax
+	
+	/* Calculate offset */
+	call	1f
+1:	popl	%ebp
+	subl	$1b, %ebp		/* %ebp = offset for labels in p2r*/
+
+	/* Set up GDT with real-mode limits and appropriate bases for
+	 * real-mode %cs and %ss.  Set up protected-mode continuation
+	 * point on stack.
+	 */
+	/* Fixup GDT */
+	leal	p2r_gdt(%ebp), %eax
+	movl	%eax, p2r_gdt_addr(%ebp)
+
+	/* Calculate CS base address: set GDT code segment, adjust
+	 * return address, set up continuation address on stack.
+	 */
+	movzwl	P2R_OFFSET_CS(%esp), %eax
+	shll	$4, %eax
+	/* Change return address to real-mode far address */
+	subl	%eax, P2R_OFFSET_RETADDR(%esp)
+	movl	%eax, %ebx
+	shrl	$4, %ebx
+	movw	%bx, (P2R_OFFSET_RETADDR+2)(%esp)
+	/* First real mode address */
+	movl	%eax, %ebx
+	shrl	$4, %ebx
+	pushw	%bx
+	leal	3f(%ebp), %ebx
+	subl	%eax, %ebx
+	pushw	%bx
+	/* Continuation address */
+	pushl	$(p2r_rmcs - p2r_gdt)
+	leal	2f(%ebp), %ebx
+	subl	%eax, %ebx
+	pushl	%ebx
+	/* Code segment in GDT */
+	movw	%ax, (p2r_rmcs+2)(%ebp)
+	shrl	$16, %eax			/* Remainder of cs base addr */
+	movb	%al, (p2r_rmcs+4)(%ebp)
+	movb	%ah, (p2r_rmcs+7)(%ebp)
+
+	/* Calculate SS base address: set GDT data segment, retain to
+	 * use for adjusting %esp.
+	 */
+	movzwl	(12+P2R_OFFSET_SS)(%esp), %eax	/* Set ss base address */
+	shll	$4, %eax
+	movw	%ax, (p2r_rmds+2)(%ebp)
+	movl	%eax, %ebx
+	shrl	$16, %ebx
+	movb	%bl, (p2r_rmds+4)(%ebp)
+	movb	%bh, (p2r_rmds+7)(%ebp)
+
+	/* Load GDT */
+	lgdt	p2r_gdt(%ebp)
+	/* Reload all segment registers and adjust %esp */
+	movw	$(p2r_rmds - p2r_gdt), %bx /* Pmode DS */
+	movw	%bx, %ss
+	subl	%eax, %esp		/* %esp now less than 0x10000 */
+	movw	%bx, %ds
+	movw	%bx, %es
+	movw	%bx, %fs
+	movw	%bx, %gs
+	lret				/* %cs:eip */
+2:	/* Segment registers now have 16-bit limits. */
+	.code16
+
+	/* Switch to real mode */
+	movl	%cr0, %ebx
+	andb	$0!CR0_PE, %bl
+	movl	%ebx, %cr0
+
+	/* Make intersegment jmp to flush the processor pipeline
+	 * and reload %cs:%eip (to clear upper 16 bits of %eip).
+	 */
+	lret
+3:		
+	
+	/* Load real-mode segment value to %ss.  %sp already OK */
+	shrl	$4, %eax
+	movw	%ax, %ss
+
+	/* Restore registers */
+	popl	%eax
+	popl	%ebx
+	popl	%ebp
+
+	/* Return to caller in real-mode */
+	lret
+
+#ifdef FLATTEN_REAL_MODE
+#define RM_LIMIT_16_19__AVL__SIZE__GRANULARITY 0x8f
+#else
+#define RM_LIMIT_16_19__AVL__SIZE__GRANULARITY 0x00
+#endif
+
+p2r_gdt:
+p2r_gdtarg:
+p2r_gdt_limit:		.word p2r_gdt_end - p2r_gdt - 1
+p2r_gdt_addr:		.long 0
+p2r_gdt_padding:	.word 0
+p2r_rmcs:
+	/* 16 bit real mode code segment */
+	.word	0xffff,(0&0xffff)
+	.byte	(0>>16),0x9b,RM_LIMIT_16_19__AVL__SIZE__GRANULARITY,(0>>24)
+p2r_rmds:
+	/* 16 bit real mode data segment */
+	.word	0xffff,(0&0xffff)
+	.byte	(0>>16),0x93,RM_LIMIT_16_19__AVL__SIZE__GRANULARITY,(0>>24)
+p2r_gdt_end:
+
+	/* This is the end of the trampoline prefix code.  When used
+	 * as a prefix, fall through to the following code in the
+	 * trampoline.
+	 */
+p2r_params: /* Structure must match prot_to_real_params_t in realmode.h */
+p2r_esp:	.long 0
+p2r_segments:
+p2r_cs:		.word 0
+p2r_ss:		.word 0
+p2r_r2p_params:	.long 0
+	.globl	_prot_to_real_prefix_end
+_prot_to_real_prefix_end:
+	
+	.globl	_prot_to_real_prefix_size
+	.equ	_prot_to_real_prefix_size, _prot_to_real_prefix_end - _prot_to_real_prefix
+	.globl	prot_to_real_prefix_size
+prot_to_real_prefix_size:	
+	.word	_prot_to_real_prefix_size
+	
+/****************************************************************************
+ * _real_to_prot_suffix
+ *
+ * Trampoline fragment.  Switch from 16-bit real-mode to 32-bit
+ * protected mode with flat physical addresses.  Copy returned stack
+ * parameters to output_stack.  Restore registers preserved by
+ * _prot_to_real_prefix.  Restore stack to previous location.
+ ****************************************************************************
+ */
+
+	.globl _real_to_prot_suffix
+	.code16
+_real_to_prot_suffix:
+
+	/* Switch to protected mode */
+	call	_real_to_prot
+	.code32
+
+	/* Calculate offset */
+	call	1f
+1:	popl	%ebp
+	subl	$1b, %ebp		/* %ebp = offset for labels in r2p */
+
+	/* Copy stack to out_stack */
+	movl	r2p_out_stack(%ebp), %edi
+	movl	r2p_out_stack_len(%ebp), %ecx
+	movl	%esp, %esi
+	cld
+	rep movsb
+
+	/* Switch back to original stack */
+	movl	r2p_esp(%ebp), %esp
+
+	/* Restore registers and return */
+	pushl	r2p_ret_addr(%ebp)	/* Set up return address on stack */
+	movl	r2p_ebx(%ebp), %ebx
+	movl	r2p_esi(%ebp), %esi
+	movl	r2p_edi(%ebp), %edi
+	movl	r2p_ebp(%ebp), %ebp
+	ret
+
+/****************************************************************************
+ * _real_to_prot
+ *
+ * Switch from 16-bit real-mode to 32-bit protected mode with flat
+ * physical addresses.  All segment registers are destroyed, %eip and
+ * %esp are changed to flat physical values, all other registers are
+ * preserved.  Interrupts are disabled.
+ *
+ * Parameters: none
+ ****************************************************************************
+ */
+
+#define R2P_PRESERVE ( 12 )
+#define R2P_OFFSET_RETADDR ( R2P_PRESERVE )
+#define R2P_OFFSET_ORIG_RETADDR ( R2P_OFFSET_RETADDR + 2 )
+
+	.globl _real_to_prot
+	.code16		
+_real_to_prot:
+	/* Disable interrupts */
+	cli
+	/* zero extend the return address */
+	pushw	$0
+
+	/* Preserve registers */
+	pushl	%ebp
+	pushl	%ebx
+	pushl	%eax
+
+	/* Convert 16-bit real-mode near return address to
+	 * 32-bit pmode physical near return address
+	 */
+	movw	%sp, %bp
+	xorl	%ebx, %ebx
+	push	%cs
+	popw	%bx
+	movw	%bx, %ds
+	shll	$4, %ebx
+	movzwl	%ss:R2P_OFFSET_ORIG_RETADDR(%bp), %eax
+	addl	%ebx, %eax
+	movl	%eax, %ss:(R2P_OFFSET_RETADDR)(%bp)
+
+	/* Store the code segment physical base address in %ebp */
+	movl	%ebx, %ebp
+
+	/* Find the offset within the code segment that I am running at */
+	xorl	%ebx, %ebx
+	call	1f
+1:	popw	%bx
+
+	/* Set up GDT */
+	leal	(r2p_gdt-1b)(%bx), %eax	/* %ds:ebx = %ds:bx = &(r2p_gdt) */
+	addl	%ebp, %eax		/* %eax = &r2p_gdt (physical) */
+	movl	%eax, %ds:(r2p_gdt-1b+2)(%bx) /* Set phys. addr. in r2p_gdt */
+
+	/* Compute the first protected mode physical address */
+	leal	(2f-1b)(%bx), %eax
+	addl	%ebp, %eax
+	movl	%eax, %ds:(r2p_paddr-1b)(%bx)
+
+	/* Calculate new %esp */
+	xorl	%eax, %eax
+	push	%ss
+	popw	%ax
+	shll	$4, %eax
+	movzwl	%sp, %ebp
+	addl	%eax, %ebp		/* %ebp = new %esp */
+	
+	/* Load GDT */
+	DATA32 lgdt %ds:(r2p_gdt-1b)(%bx)	/* Load GDT */
+
+	/* Switch to protected mode */
+	movl	%cr0, %eax
+	orb	$CR0_PE, %al
+	movl	%eax, %cr0
+
+	/* flush prefetch queue, and reload %cs:%eip */
+	DATA32 ljmp %ds:(r2p_paddr-1b)(%bx)
+	.code32
+2:
+	
+	/* Load segment registers, adjust %esp */
+	movw	$(r2p_pmds-r2p_gdt), %ax
+	movw	%ax, %ss
+	movl	%ebp, %esp
+	movw	%ax, %ds
+	movw	%ax, %es
+	movw	%ax, %fs
+	movw	%ax, %gs
+
+	/* Restore registers */
+	popl	%eax
+	popl	%ebx
+	popl	%ebp
+
+	/* return to caller */
+	ret
+
+r2p_gdt:
+	.word	r2p_gdt_end - r2p_gdt - 1	/* limit */
+	.long 0					/* addr */
+	.word 0
+r2p_pmcs:
+	/* 32 bit protected mode code segment, physical addresses */
+	.word	0xffff, 0
+	.byte	0, 0x9f, 0xcf, 0
+r2p_pmds:
+	/* 32 bit protected mode data segment, physical addresses */
+	.word	0xffff,0
+	.byte	0,0x93,0xcf,0
+r2p_gdt_end:
+
+r2p_paddr:
+	.long 2b
+	.word r2p_pmcs - r2p_gdt, 0
+
+
+	/* This is the end of the trampoline suffix code.
+	 */
+r2p_params: /* Structure must match real_to_prot_params_t in realmode.h */
+r2p_ret_addr:		.long 0
+r2p_esp:		.long 0
+r2p_ebx:		.long 0
+r2p_esi:		.long 0
+r2p_edi:		.long 0
+r2p_ebp:		.long 0
+r2p_out_stack:		.long 0
+r2p_out_stack_len:	.long 0
+	.globl	_real_to_prot_suffix_end
+_real_to_prot_suffix_end:
+
+	.globl	_real_to_prot_suffix_size
+	.equ	_real_to_prot_suffix_size, _real_to_prot_suffix_end - _real_to_prot_suffix
+	.globl	real_to_prot_suffix_size
+real_to_prot_suffix_size:
+	.word	_real_to_prot_suffix_size
+
+rm_callback_interface_end:
+
+	.globl	_rm_callback_interface_size
+	.equ	_rm_callback_interface_size, rm_callback_interface_end - rm_callback_interface
+	.globl	rm_callback_interface_size
+rm_callback_interface_size:
+	.word	_rm_callback_interface_size
+
+/****************************************************************************
+ * END OF REAL-MODE CALLBACK INTERFACE
+ ****************************************************************************
+ */
+
+
+#ifdef PXE_EXPORT
+/****************************************************************************
+ * PXE CALLBACK INTERFACE
+ *
+ * Prepend this to rm_callback_interface to create a real-mode PXE
+ * callback interface.
+ ****************************************************************************
+ */
+	.section ".text16", "ax", @progbits
+	.globl	pxe_callback_interface
+	.code16
+pxe_callback_interface:
+
+/* Macro to calculate offset of labels within code segment in
+ * installed copy of code.
+ */
+#define INSTALLED(x) ( (x) - pxe_callback_interface )
+
+/****************************************************************************
+ * PXE entry points (!PXE and PXENV+ APIs)
+ ****************************************************************************
+ */
+	/* in_call mechanism for !PXE API calls */
+	.globl	_pxe_in_call_far
+_pxe_in_call_far:
+	/* Prepend "PXE API call" and "API version 0x201" to stack */
+	pushl	$0x201
+	jmp	1f
+	/* in_call mechanism for PXENV+ API calls */
+	.globl	_pxenv_in_call_far
+_pxenv_in_call_far:
+	/* Prepend "PXE API call" and "API version 0x200" to stack */
+	pushl	$0x200
+1:	pushl	$EB_OPCODE_PXE
+	/* Perform real-mode in_call */
+	call	pxe_rm_in_call
+	/* Return */
+	addw	$8, %sp
+	lret
+
+/****************************************************************************
+ * PXE installation check (INT 1A) code
+ ****************************************************************************
+ */
+	.globl	_pxe_intercept_int1a
+_pxe_intercept_int1a:
+	pushfw
+	cmpw	$0x5650, %ax
+	jne	2f
+1:	/* INT 1A,5650 - Intercept */
+	popfw
+	/* Set up return values according to PXE spec: */
+	movw	$0x564e, %ax		/* AX := 564Eh (VN) */
+	pushw	%cs:INSTALLED(_pxe_pxenv_segment)
+	popw	%es			/* ES:BX := &(PXENV+ structure) */
+	movw	%cs:INSTALLED(_pxe_pxenv_offset), %bx
+	clc				/* CF is cleared */
+	lret	$2			/* 'iret' without reloading flags */
+2:	/* INT 1A,other - Do not intercept */
+	popfw
+	ljmp	%cs:*INSTALLED(_pxe_intercepted_int1a)
+
+	.globl	_pxe_intercepted_int1a
+_pxe_intercepted_int1a:	.word 0,0
+	.globl	_pxe_pxenv_location
+_pxe_pxenv_location:
+_pxe_pxenv_offset:	.word 0
+_pxe_pxenv_segment:	.word 0
+	
+pxe_rm_in_call:	
+pxe_attach_rm:
+	/* rm_callback_interface must be appended here */
+
+pxe_callback_interface_end:
+
+	.globl	_pxe_callback_interface_size
+	.equ	_pxe_callback_interface_size, pxe_callback_interface_end - pxe_callback_interface
+	.globl	pxe_callback_interface_size
+pxe_callback_interface_size:
+	.word	_pxe_callback_interface_size
+
+#else	/* PXE_EXPORT */
+	
+/* Define symbols used by the linker scripts, to prevent link errors */
+	.globl	_pxe_callback_interface_size
+	.equ	_pxe_callback_interface_size, 0	
+	
+#endif	/* PXE_EXPORT */
+
+#else	/* CODE16 */
+
+/* Define symbols used by the linker scripts, to prevent link errors */
+	.globl	_rm_callback_interface_size
+	.equ	_rm_callback_interface_size, 0	
+	.globl	_pxe_callback_interface_size
+	.equ	_pxe_callback_interface_size, 0	
+		
+#endif	/* CODE16 */

src/arch/i386/core/start16.S

@@ -0,0 +1,285 @@
+/*****************************************************************************
+ *
+ * THIS FILE IS NOW OBSOLETE.
+ *
+ * The functions of this file are now placed in init.S.
+ *
+ *****************************************************************************
+ */
+	
+#ifndef PCBIOS
+#error	"16bit code is only supported with the PCBIOS"
+#endif
+
+#define	CODE_SEG 0x08
+#define DATA_SEG 0x10
+
+#define EXEC_IN_SITU_MAGIC	0x45524548	/* 'HERE' */
+
+	.equ CR0_PE, 1
+
+#ifdef	GAS291
+#define DATA32 data32;
+#define ADDR32 addr32;
+#define	LJMPI(x)	ljmp	x
+#else
+#define DATA32 data32
+#define ADDR32 addr32
+/* newer GAS295 require #define	LJMPI(x)	ljmp	*x */
+#define	LJMPI(x)	ljmp	x
+#endif
+
+/*****************************************************************************
+ *
+ * start16 : move payload to desired area of memory, set up for exit
+ *	     back to prefix, set up for 32-bit code.
+ *
+ * Enter (from prefix) with es:di = 0x4552:0x4548 if you want to
+ * prevent start16 from moving the payload.  There are three
+ * motivations for moving the payload:
+ * 
+ * 1. It may be in ROM, in which case we need to move it to RAM.
+ * 2. Whatever loaded us probably didn't know about our memory usage
+ *    beyond the end of the image file.  We should claim this memory
+ *    before using it.
+ *
+ * Unless the prefix instructs us otherwise we will move the payload to:
+ *
+ *    An area of memory claimed from the BIOS via 40:13.
+ *
+ * We use the main Etherboot stack (within the image target) as our
+ * stack; we don't rely on the prefix passing us a stack usable for
+ * anything other than the prefix's return address.  The (first 512
+ * bytes of the) prefix code segment is copied to a safe archive
+ * location.
+ *
+ * When we return to the prefix (from start32), we copy this code back
+ * to a new area of memory, restore the prefix's ss:sp and ljmp back
+ * to the copy of the prefix.  The prefix will see a return from
+ * start16 *but* may be executing at a new location.  Code following
+ * the lcall to start16 must therefore be position-independent and
+ * must also be within [cs:0000,cs:01ff].  We make absolutely no
+ * guarantees about the stack contents when the prefix regains
+ * control.
+ * 
+ * Trashes just about all registers, including all the segment
+ * registers.
+ *
+ *****************************************************************************
+ */
+
+	.text
+	.code16
+	.arch i386
+	.org 0
+	.globl _start16
+_start16:
+
+/*****************************************************************************
+ * Work out where we are going to place our image (image = optional
+ * decompressor + runtime).  Exit this stage with %ax containing the
+ * runtime target address divided by 16 (i.e. a real-mode segment
+ * address).
+ *****************************************************************************
+ */	
+	movw	%es, %ax
+	cmpw	$(EXEC_IN_SITU_MAGIC >> 16), %ax
+	jne	exec_moved
+	cmpw	$(EXEC_IN_SITU_MAGIC & 0xffff), %di
+	jne	exec_moved
+exec_in_situ: 
+	/* Prefix has warned us not to move the payload.  Simply
+	 * calculate where the image is going to end up, so we can
+	 * work out where to put our stack.
+	 */
+	movw	%cs, %ax
+	addw	$((payload-_start16)/16), %ax
+	jmp	99f
+exec_moved:
+	/* Claim an area of base memory from the BIOS and put the
+	 * payload there.  arch_relocated_to() will deal with freeing
+	 * up this memory once we've relocated to high memory.
+	 */
+	movw	$0x40, %ax
+	movw	%ax, %es
+	movw	%es:(0x13), %ax		/* FBMS in kb to %ax */
+	shlw	$6, %ax			/* ... in paragraphs */
+	subw	$__image_size_pgh, %ax	/* Subtract space for image */
+	shrw	$6, %ax			/* Round down to nearest kb */
+	movw	%ax, %es:(0x13)		/* ...and claim memory from BIOS */
+	shlw	$6, %ax
+99:
+	/* At this point %ax contains the segment address for the
+	 * start of the image (image = optional decompressor + runtime).
+	 */
+
+/*****************************************************************************
+ * Set up stack in start32's stack space within the place we're going
+ * to copy Etherboot to, reserve space for GDT, copy return address
+ * from prefix stack, store prefix stack address
+ *****************************************************************************
+ */
+	popl	%esi			/* Return address */
+	mov	%ss, %bx		/* %es:di = prefix stack address   */
+	mov	%bx, %es		/* (*after* pop of return address) */
+	movw	%sp, %di
+	movw	$__offset_stack_pgh, %bx	/* Set up Etherboot stack */
+	addw	%ax, %bx
+	movw	%bx, %ss
+	movw	$__stack_size, %sp
+	subw	$(_gdt_end - _gdt), %sp	/* Reserve space for GDT */
+	movw	%sp, %bp		/* Record GDT location */	
+	/* Set up i386_rm_in_call_data_t structure on stack.  This is
+	 * the same structure as is set up by rm_in_call.
+	 */
+	pushl	$0			/* Dummy opcode */
+	pushl	%esi			/* Prefix return address */
+	pushfw				/* Flags */
+	pushw	%di			/* Prefix %sp */
+	pushw	%gs			/* Segment registers */
+	pushw	%fs
+	pushw	%es
+	pushw	%ds
+	pushw	%es			/* Prefix %ss */
+	pushw	%cs
+	/* Stack is now 32-bit aligned */
+	
+	/* %ax still contains image target segment address */
+
+/*****************************************************************************
+ * Calculate image target and prefix code physical addresses, store on stack
+ * for use in copy routine.
+ *****************************************************************************
+ */
+	movzwl	%es:-2(%di), %ebx	/* Prefix code segment */
+	shll	$4, %ebx
+	pushl	%ebx			/* Prefix code physical address */
+	movzwl	%ax, %edi		/* Image target segment */
+	shll	$4, %edi
+	pushl	%edi			/* Image target physical address */
+
+/*****************************************************************************
+ * Transition to 32-bit protected mode.  Set up all segment
+ * descriptors to use flat physical addresses.
+ *****************************************************************************
+ */
+	/* Copy gdt to area reserved on stack
+	 */
+	push	%cs			/* GDT source location -> %ds:%si */
+	pop	%ds
+	mov	$(_gdt - _start16), %si
+	push	%ss			/* GDT target location -> %es:%di */
+	pop	%es
+	mov	%bp, %di
+	mov	$(_gdt_end - _gdt), %cx
+	cld
+	rep movsb			/* Copy GDT to stack */
+	movl	%ss, %eax
+	shll	$4, %eax
+	movzwl	%bp, %ebx
+	addl	%eax, %ebx		/* Physical addr of GDT copy -> %ebx */
+	movl	%ebx, 2(%bp)		/* Fill in addr field in GDT */
+		
+	/* Compute the offset I am running at.
+	 */
+	movl	%cs, %ebx 
+	shll	$4, %ebx		/* %ebx = offset for start16 symbols */
+
+	/* Switch to 32bit protected mode.
+	 */
+	cli				/* Disable interrupts */
+	lgdt	(%bp)			/* Load GDT from stack */
+	movl	%cr0, %eax		/* Set protected mode bit */
+	orb	$CR0_PE, %al
+	movl	%eax, %cr0
+	movl	%ss, %eax		/* Convert stack pointer to 32bit */
+	shll	$4, %eax
+	movzwl	%sp, %esp
+	addl	%eax, %esp
+	movl	$DATA_SEG, %eax		/* Reload the segment registers */
+	movl	%eax, %ds
+	movl	%eax, %es
+	movl	%eax, %ss
+	movl	%eax, %fs
+	movl	%eax, %gs
+	/* Flush prefetch queue, and reload %cs:%eip by effectively ljmping
+	 * to code32_start.  Do the jump via pushl and lret because the text
+	 * may not be writable/
+	 */
+	pushl	$CODE_SEG
+	ADDR32 leal (code32_start-_start16)(%ebx), %eax
+	pushl	%eax	
+	DATA32 lret /* DATA32 needed, because we're still in 16-bit mode */
+
+_gdt:
+gdtarg:
+	.word	_gdt_end - _gdt - 1	/* limit */
+	.long 0				/* addr */
+	.word 0
+_pmcs:
+	/* 32 bit protected mode code segment */
+	.word	0xffff, 0
+	.byte	0, 0x9f, 0xcf, 0
+_pmds:
+	/* 32 bit protected mode data segment */
+	.word	0xffff,0
+	.byte	0,0x93,0xcf,0
+_gdt_end:
+
+	.code32
+code32_start:	
+	
+/*****************************************************************************
+ * Copy payload to target location.  Do the copy backwards, since if
+ * there's overlap with a forward copy then it means start16 is going
+ * to get trashed during the copy anyway...
+ *****************************************************************************
+ */
+	popl	%edi		/* Image target physical address */
+	pushl	%edi
+	leal	(payload-_start16)(%ebx), %esi /* Image source physical addr */
+	movl	$__payload_size, %ecx /* Payload size (not image size) */
+	addl	%ecx, %edi	/* Start at last byte (length - 1) */
+	decl	%edi
+	addl	%ecx, %esi
+	decl	%esi
+	std			/* Backward copy of image */
+	rep movsb
+	cld
+	popl	%edi		/* Restore image target physical address */
+	leal	__decompressor_uncompressed(%edi), %ebx
+	subl	$_text, %ebx /* %ebx = offset for runtime symbols */
+
+/*****************************************************************************
+ * Copy prefix to storage area within Etherboot image.
+ *****************************************************************************
+ */
+	popl	%esi		/* Prefix source physical address */
+	pushl	%edi
+	leal	_prefix_copy(%ebx), %edi /* Prefix copy phys. addr. */
+	leal	_eprefix_copy(%ebx), %ecx
+	subl	%edi, %ecx	/* Prefix copy size */
+	rep movsb		/* Forward copy of prefix */
+	popl	%edi		/* Restore image target physical address */
+
+/*****************************************************************************
+ * Record base memory used by Etherboot image
+ *****************************************************************************
+ */
+	movl	%edi, _prefix_image_basemem (%ebx)
+
+/*****************************************************************************
+ * Jump to start of the image (i.e. the decompressor, or start32 if
+ * non-compressed).
+ *****************************************************************************
+ */
+	pushl	$0		/* Inform start32 that exit path is 16-bit */
+	jmpl	*%edi		/* Jump to image */
+	
+	.balign 16
+	/* Etherboot needs to be 16byte aligned or data that
+	 * is virtually aligned is no longer physically aligned
+	 * which is just nasty in general.  16byte alignment 
+	 * should be sufficient though.
+	 */
+payload:

src/arch/i386/core/start16.lds

@@ -0,0 +1,8 @@
+/* When linking with an uncompressed image, these symbols are not
+ * defined so we provide them here.
+ */
+
+__decompressor_uncompressed = 0 ;
+__decompressor__start = 0 ;
+
+INCLUDE arch/i386/core/start16z.lds

src/arch/i386/core/start16z.lds

@@ -0,0 +1,65 @@
+OUTPUT_FORMAT("elf32-i386", "elf32-i386", "elf32-i386")
+OUTPUT_ARCH(i386)
+
+/* Linker-generated symbols are prefixed with a double underscore.
+ * Decompressor symbols are prefixed with __decompressor_.  All other
+ * symbols are the same as in the original object file, i.e. the
+ * runtime addresses.
+ */
+
+ENTRY(_start16)
+
+SECTIONS {
+	.text : {
+		*(.text)
+	}
+	.payload : {
+		__payload_start = .;
+		*(.data)
+		__payload_end = .;
+	}
+
+	/* _payload_size is the size of the binary image appended to
+	 * start16, in bytes.
+	 */
+	__payload_size = __payload_end - __payload_start ;
+
+	/* _size is the size of the runtime image
+	 * (start32 + the C code), in bytes.
+	 */
+	__size = _end - _start ;
+
+	/* _decompressor_size is the size of the decompressor, in
+	 * bytes.  For a non-compressed image, start16.lds sets
+	 * _decompressor_uncompressed = _decompressor__start = 0.
+	 */
+	__decompressor_size = __decompressor_uncompressed - __decompressor__start ;
+
+	/* image__size is the total size of the image, after
+	 * decompression and including the decompressor if applicable.
+	 * It is therefore the amount of memory that start16's payload
+	 * needs in order to execute, in bytes.
+	 */
+	__image_size = __size + __decompressor_size ;
+
+	/* Amount to add to runtime symbols to obtain the offset of
+	 * that symbol within the image.
+	 */
+	__offset_adjust = __decompressor_size - _start ;
+
+	/* Calculations for the stack
+	 */
+	__stack_size = _estack - _stack ;
+	__offset_stack = _stack + __offset_adjust ;
+
+	/* Some symbols will be larger than 16 bits but guaranteed to
+	 * be multiples of 16.  We calculate them in paragraphs and
+	 * export these symbols which can be used in 16-bit code
+	 * without risk of overflow.
+	 */
+	__image_size_pgh = ( __image_size / 16 );
+	__start_pgh = ( _start / 16 );
+	__decompressor_size_pgh = ( __decompressor_size / 16 );
+	__offset_stack_pgh = ( __offset_stack / 16 );
+}
+

src/arch/i386/core/start32.S

@@ -0,0 +1,767 @@
+/* #defines because ljmp wants a number, probably gas bug */
+/*	.equ	KERN_CODE_SEG,_pmcs-_gdt	*/
+#define	KERN_CODE_SEG	0x08
+	.equ	KERN_DATA_SEG,_pmds-_gdt
+/*	.equ	REAL_CODE_SEG,_rmcs-_gdt	*/
+#define	REAL_CODE_SEG	0x18
+	.equ	REAL_DATA_SEG,_rmds-_gdt
+	.equ	FLAT_CODE_SEG,_pmcs2-_gdt
+	.equ	FLAT_DATA_SEG,_pmds2-_gdt
+	.equ	CR0_PE,1
+#ifdef CONFIG_X86_64
+	.equ	LM_CODE_SEG,  _lmcs-_gdt
+	.equ	LM_DATA_SEG,  _lmds-_gdt
+#endif
+
+	.equ	MSR_K6_EFER,   0xC0000080
+	.equ	EFER_LME,      0x00000100
+	.equ	X86_CR4_PAE,   0x00000020
+	.equ	CR0_PG,        0x80000000
+
+#ifdef	GAS291
+#define DATA32 data32;
+#define ADDR32 addr32;
+#define	LJMPI(x)	ljmp	x
+#else
+#define DATA32 data32
+#define ADDR32 addr32
+/* newer GAS295 require #define	LJMPI(x)	ljmp	*x */
+#define	LJMPI(x)	ljmp	x
+#endif
+
+#define BOCHSBP xchgw %bx, %bx
+
+#include "callbacks.h"
+#define NUM_PUSHA_REGS (8)
+#define NUM_SEG_REGS (6)
+	
+/*
+ * NOTE: if you write a subroutine that is called from C code (gcc/egcs),
+ * then you only have to take care of %ebx, %esi, %edi and %ebp.  These
+ * registers must not be altered under any circumstance.  All other registers
+ * may be clobbered without any negative side effects.  If you don't follow
+ * this rule then you'll run into strange effects that only occur on some
+ * gcc versions (because the register allocator may use different registers).
+ *
+ * All the data32 prefixes for the ljmp instructions are necessary, because
+ * the assembler emits code with a relocation address of 0.  This means that
+ * all destinations are initially negative, which the assembler doesn't grok,
+ * because for some reason negative numbers don't fit into 16 bits. The addr32
+ * prefixes are there for the same reasons, because otherwise the memory
+ * references are only 16 bit wide.  Theoretically they are all superfluous.
+ * One last note about prefixes: the data32 prefixes on all call _real_to_prot
+ * instructions could be removed if the _real_to_prot function is changed to
+ * deal correctly with 16 bit return addresses.  I tried it, but failed.
+ */
+
+/**************************************************************************
+ * START
+ *
+ * This file is no longer enterered from the top.  init.S will jump to
+ * either _in_call or _rm_in_call, depending on the processor mode
+ * when init.S was entered.
+ **************************************************************************/
+	.text
+	.arch i386
+	.code32
+	
+/**************************************************************************
+_IN_CALL - make a call in to Etherboot.
+**************************************************************************/
+
+/* There are two 32-bit entry points: _in_call and _in_call_far, for
+ * near calls and far calls respectively.  Both should be called with
+ * flat physical addresses.  They will result in a call to the C
+ * routine in_call(); see there for API details.
+ *
+ * Note that this routine makes fairly heavy use of the stack and no
+ * use of fixed data areas.  This is because it must be re-entrant;
+ * there may be more than one concurrent call in to Etherboot.
+ */
+
+#define IC_OFFSET_VA_LIST_PTR ( 0 )
+#define IC_OFFSET_VA_LIST_PTR_E ( IC_OFFSET_VA_LIST_PTR + 4 )
+#define IC_OFFSET_REGISTERS ( IC_OFFSET_VA_LIST_PTR_E )
+#define IC_OFFSET_REGISTERS_E ( IC_OFFSET_REGISTERS + ( NUM_PUSHA_REGS * 4 ) )
+#define IC_OFFSET_SEG_REGS ( IC_OFFSET_REGISTERS_E )
+#define IC_OFFSET_SEG_REGS_E ( IC_OFFSET_SEG_REGS + ( NUM_SEG_REGS * 2 ) )
+#define IC_OFFSET_GDT ( IC_OFFSET_SEG_REGS_E )
+#define IC_OFFSET_GDT_E ( IC_OFFSET_GDT + 8 )
+#define IC_OFFSET_FLAGS ( IC_OFFSET_GDT_E )
+#define IC_OFFSET_FLAGS_E ( IC_OFFSET_FLAGS + 4 )
+#define IC_OFFSET_RETADDR ( IC_OFFSET_FLAGS_E )
+#define IC_OFFSET_RETADDR_E ( IC_OFFSET_RETADDR + 8 )
+#define IC_OFFSET_ORIG_STACK ( IC_OFFSET_RETADDR )
+#define IC_OFFSET_OPCODE ( IC_OFFSET_ORIG_STACK + 8 )
+#define IC_OFFSET_OPCODE_E ( IC_OFFSET_OPCODE + 4 )
+#define IC_OFFSET_VA_LIST ( IC_OFFSET_OPCODE_E )
+	
+	.code32
+	.globl _in_call
+	.globl _in_call_far
+_in_call:
+	/* Expand to far return address */
+	pushl	%eax			/* Store %eax */
+	xorl	%eax, %eax
+	movw	%cs, %ax
+	xchgl	%eax, 4(%esp)		/* 4(%esp) = %cs, %eax = ret addr */
+	xchgl	%eax, 0(%esp)		/* 0(%esp) = ret addr, restore %eax */
+_in_call_far:
+	/* Store flags */
+	pushfl
+	/* Store the GDT */
+	subl	$8, %esp
+	sgdt	0(%esp)
+	/* Store segment register values */
+	pushw	%gs
+	pushw	%fs
+	pushw	%es
+	pushw	%ds
+	pushw	%ss
+	pushw	%cs
+	/* Store general-purpose register values */
+	pushal
+	/* Replace %esp in store with physical %esp value on entry */
+	leal	(IC_OFFSET_ORIG_STACK - IC_OFFSET_REGISTERS)(%esp), %eax
+	movl	%eax, (IC_OFFSET_REGISTERS - IC_OFFSET_REGISTERS + 12)(%esp)
+	/* Store va_list pointer (physical address) */
+	leal	(IC_OFFSET_VA_LIST - IC_OFFSET_VA_LIST_PTR_E)(%esp), %eax
+	pushl	%eax
+	/* IC_OFFSET_*(%esp) are now valid */
+
+	/* Switch to virtual addresses */
+	call	_phys_to_virt
+
+	/* Fixup the va_list pointer */
+	movl	virt_offset, %ebp
+	subl	%ebp, IC_OFFSET_VA_LIST_PTR(%esp)
+
+	/* Check opcode for EB_USE_INTERNAL_STACK flag */
+	movl	IC_OFFSET_OPCODE(%esp), %eax
+	testl	$EB_USE_INTERNAL_STACK, %eax
+	je	2f
+	/* Use internal stack flag set */
+	/* Check %esp is not already in internal stack range */
+	leal	_stack, %esi		/* %esi = bottom of internal stack */
+	leal	_estack, %edi		/* %edi = top of internal stack */
+	cmpl	%esi, %esp
+	jb	1f
+	cmpl	%edi, %esp
+	jbe	2f
+1:	/* %esp not currently in internal stack range */
+	movl	%esp, %esi		/* %esi = original stack */
+	movl	$IC_OFFSET_OPCODE_E, %ecx /* %ecx = length to transfer */
+	subl	%ecx, %edi		/* %edi = internal stack pos */
+	movl	%edi, %esp		/*  = new %esp */
+	rep movsb			/* Copy data to internal stack */
+2:
+
+	/* Call to C code */
+	call	i386_in_call
+
+	/* Set %eax (return code from C) in registers structure on
+	 * stack, so that we return it to the caller.
+	 */
+	movl	%eax, (IC_OFFSET_REGISTERS + 28)(%esp)
+
+	/* Calculate physical continuation address */
+	movl	virt_offset, %ebp
+	movzwl	(IC_OFFSET_SEG_REGS + 0)(%esp), %eax	/* %cs */
+	movzwl	(IC_OFFSET_SEG_REGS + 2)(%esp), %ebx	/* %ss */
+	pushl	%eax			/* Continuation segment */
+	leal	1f(%ebp), %eax
+	pushl	%eax			/* Continuation offset */
+	
+	/* Restore caller's GDT */
+	cli				/* Temporarily disable interrupts */
+	lgdt	(8+IC_OFFSET_GDT)(%esp)
+	/* Reset %ss and adjust %esp */
+	movw	%bx, %ss
+	addl	%ebp, %esp
+	lret				/* Reload %cs:eip, flush prefetch */
+1:
+
+	/* Skip va_list ptr */
+	popl	%eax
+	/* Reload general-purpose registers to be returned */
+	popal
+	/* Reload segment registers as passed in from caller */
+	popw	%gs
+	popw	%fs
+	popw	%es
+	popw	%ds
+	addl	$(4+8), %esp	/* Skip %cs, %ss and GDT (already reloaded) */
+	/* Restore flags (including revert of interrupt status) */
+	popfl
+
+	/* Restore physical %esp from entry.  It will only be
+	 * different if EB_USE_INTERNAL_STACK was specified.
+	 */
+	movl	( 12 + IC_OFFSET_REGISTERS - IC_OFFSET_RETADDR )(%esp), %esp
+		
+	/* Check for EB_SKIP_OPCODE */
+	pushfl
+	testl	$EB_SKIP_OPCODE, 12(%esp)
+	jnz	1f
+	/* Normal return */
+	popfl
+	lret
+1:	/* Return and skip opcode */
+	popfl
+	lret	$4
+	
+/**************************************************************************
+RELOCATE_TO - relocate etherboot to the specified address
+**************************************************************************/
+	.globl relocate_to
+relocate_to:
+	/* Save the callee save registers */
+	pushl	%ebp
+	pushl	%esi
+	pushl	%edi
+
+	/* Compute the virtual destination address */
+	movl	16(%esp), %edi	# dest
+	subl	virt_offset, %edi
+	
+
+	/* Compute the new value of virt_offset */
+	movl	16(%esp), %ebp	# virt_offset
+	subl	$_text, %ebp
+
+	/* Fixup the gdt */
+	pushl	$_pmcs
+	pushl	%ebp		# virt_offset
+	call	set_seg_base
+	addl	$8, %esp
+
+	/* Fixup gdtarg */
+	leal	_gdt(%ebp), %eax
+	movl	%eax, gdtarg +2
+
+	/* Fixup virt_offset */
+	movl	%ebp, virt_offset
+
+	/* Load the move parameters */
+	movl	$_text, %esi
+	movl	$_end, %ecx
+	subl	%esi, %ecx
+
+	/* Move etherboot uses %esi, %edi, %ecx */
+	rep 
+	movsb
+
+	/* Reload the gdt */
+	cs
+	lgdt	gdtarg
+
+	/* Reload %cs */
+	ljmp	$KERN_CODE_SEG, $1f
+1:
+	/* reload other segment registers */
+	movl	$KERN_DATA_SEG, %eax
+	movl	%eax,%ds
+	movl	%eax,%es
+	movl	%eax,%ss
+	movl	%eax,%fs
+	movl	%eax,%gs
+
+	/* Restore the callee save registers */
+	popl	%edi
+	popl	%esi
+	popl	%ebp
+
+	/* return */
+	ret
+
+/**************************************************************************
+XSTART32 - Transfer control to the kernel just loaded
+**************************************************************************/
+	.globl xstart32
+xstart32:
+	/* Save the callee save registers */
+	movl	%ebp, os_regs + 32
+	movl	%esi, os_regs + 36
+	movl	%edi, os_regs + 40
+	movl	%ebx, os_regs + 44
+
+	/* save the return address */
+	popl	%eax
+	movl	%eax, os_regs + 48
+
+	/* save the stack pointer */
+	movl	%esp, os_regs + 52
+
+	/* Get the new destination address */
+	popl	%ecx
+
+	/* Store the physical address of xend on the stack */
+	movl	$xend32, %ebx
+	addl	virt_offset, %ebx
+	pushl	%ebx
+
+	/* Store the destination address on the stack */
+	pushl	$FLAT_CODE_SEG
+	pushl	%ecx
+
+	/* Cache virt_offset */
+	movl	virt_offset, %ebp
+	
+	/* Switch to using physical addresses */
+	call	_virt_to_phys
+
+	/* Save the target stack pointer */
+	movl	%esp, os_regs + 12(%ebp)
+	leal	os_regs(%ebp), %esp
+
+	/* Store the pointer to os_regs */
+	movl	%esp, os_regs_ptr(%ebp)
+
+	/* Load my new registers */
+	popal
+	movl	(-32 + 12)(%esp), %esp
+
+	/* Jump to the new kernel
+	 * The lret switches to a flat code segment
+	 */
+	lret
+
+	.balign 4
+	.globl xend32
+xend32:
+	/* Fixup %eflags */
+	nop
+	cli
+	cld
+	
+	/* Load %esp with &os_regs + virt_offset */
+	.byte	0xbc /* movl $0, %esp */
+os_regs_ptr:
+	.long	0
+
+	/* Save the result registers */
+	addl	$32, %esp
+	pushal
+
+	/* Compute virt_offset */
+	movl	%esp, %ebp
+	subl	$os_regs, %ebp
+	
+	/* Load the stack pointer */
+	movl	52(%esp), %esp
+
+	/* Enable the virtual addresses */
+	leal	_phys_to_virt(%ebp), %eax
+	call	*%eax
+
+	/* Restore the callee save registers */
+	movl	os_regs + 32, %ebp
+	movl	os_regs + 36, %esi
+	movl	os_regs + 40, %edi
+	movl	os_regs + 44, %ebx
+	movl	os_regs + 48, %edx
+	movl	os_regs + 52, %esp
+
+	/* Get the C return value */
+	movl	os_regs + 28, %eax
+
+	jmpl	*%edx
+
+#ifdef CONFIG_X86_64
+	.arch	sledgehammer
+/**************************************************************************
+XSTART_lm - Transfer control to the kernel just loaded in long mode
+**************************************************************************/
+	.globl xstart_lm
+xstart_lm:
+	/* Save the callee save registers */
+	pushl	%ebp
+	pushl	%esi
+	pushl	%edi
+	pushl	%ebx
+
+	/* Cache virt_offset && (virt_offset & 0xfffff000) */
+	movl	virt_offset, %ebp
+	movl	%ebp, %ebx
+	andl	$0xfffff000, %ebx
+
+	/* Switch to using physical addresses */
+	call	_virt_to_phys
+
+	/* Initialize the page tables */
+	/* Level 4 */
+	leal	0x23 + pgt_level3(%ebx), %eax
+	leal	pgt_level4(%ebx), %edi
+	movl	%eax, (%edi)
+
+	/* Level 3 */
+	leal	0x23 + pgt_level2(%ebx), %eax
+	leal	pgt_level3(%ebx), %edi
+	movl	%eax, 0x00(%edi)
+	addl	$4096, %eax
+	movl	%eax, 0x08(%edi)
+	addl	$4096, %eax
+	movl	%eax, 0x10(%edi)
+	addl	$4096, %eax
+	movl	%eax, 0x18(%edi)
+
+	/* Level 2 */
+	movl	$0xe3, %eax
+	leal	pgt_level2(%ebx), %edi
+	leal	16384(%edi), %esi
+pgt_level2_loop:
+	movl	%eax, (%edi)
+	addl	$8, %edi
+	addl	$0x200000, %eax
+	cmp	%esi, %edi
+	jne	pgt_level2_loop
+
+	/* Point at the x86_64 page tables */
+	leal	pgt_level4(%ebx), %edi
+	movl	%edi, %cr3
+
+
+	/* Setup for the return from 64bit mode */
+	/* 64bit align the stack */
+	movl	%esp, %ebx		/* original stack pointer + 16 */
+	andl	$0xfffffff8, %esp
+
+	/* Save original stack pointer + 16 */
+	pushl	%ebx
+
+	/* Save virt_offset */
+	pushl	%ebp
+
+	/* Setup for the jmp to 64bit long mode */
+	leal	start_lm(%ebp), %eax
+	movl	%eax, 0x00 + start_lm_addr(%ebp)
+	movl	$LM_CODE_SEG, %eax
+	movl	%eax, 0x04 + start_lm_addr(%ebp)
+
+	/* Setup for the jump out of 64bit long mode */
+	leal	end_lm(%ebp), %eax
+	movl	%eax, 0x00 + end_lm_addr(%ebp)
+	movl	$FLAT_CODE_SEG, %eax
+	movl	%eax, 0x04 + end_lm_addr(%ebp)
+
+	/* Enable PAE mode */
+	movl	%cr4, %eax
+	orl	$X86_CR4_PAE, %eax
+	movl	%eax, %cr4
+
+	/* Enable long mode */
+	movl	$MSR_K6_EFER, %ecx
+	rdmsr
+	orl	$EFER_LME, %eax
+	wrmsr
+
+	/* Start paging, entering 32bit compatiblity mode */
+	movl	%cr0, %eax
+	orl	$CR0_PG, %eax
+	movl	%eax, %cr0
+
+	/* Enter 64bit long mode */
+	ljmp	*start_lm_addr(%ebp)
+	.code64
+start_lm:
+	/* Load 64bit data segments */
+	movl	$LM_DATA_SEG, %eax
+	movl	%eax, %ds
+	movl	%eax, %es
+	movl	%eax, %ss
+
+	andq	$0xffffffff, %rbx
+	/* Get the address to jump to */
+	movl	20(%rbx), %edx
+	andq	$0xffffffff, %rdx
+	
+	/* Get the argument pointer */
+	movl	24(%rbx), %ebx
+	andq	$0xffffffff, %rbx
+
+	/* Jump to the 64bit code */
+	call	*%rdx
+
+	/* Preserve the result */
+	movl	%eax, %edx
+
+	/* Fixup %eflags */
+	cli
+	cld
+
+	/* Switch to 32bit compatibility mode */
+	ljmp	*end_lm_addr(%rip)
+
+	.code32
+end_lm:
+	/* Disable paging */
+	movl	%cr0, %eax
+	andl	$~CR0_PG, %eax
+	movl	%eax, %cr0
+
+	/* Disable long mode */
+	movl	$MSR_K6_EFER, %ecx
+	rdmsr
+	andl	$~EFER_LME, %eax
+	wrmsr
+
+	/* Disable PAE */
+	movl	%cr4, %eax
+	andl	$~X86_CR4_PAE, %eax
+	movl	%eax, %cr4
+	
+	/* Compute virt_offset */
+	popl	%ebp
+
+	/* Compute the original stack pointer + 16 */
+	popl	%ebx
+	movl	%ebx, %esp
+
+	/* Enable the virtual addresses */
+	leal	_phys_to_virt(%ebp), %eax
+	call	*%eax
+
+	/* Restore the callee save registers */
+	popl	%ebx
+	popl	%esi
+	popl	%edi
+	popl	%ebp
+
+	/* Get the C return value */
+	movl	%edx, %eax
+
+	/* Return */
+	ret
+
+	.arch i386
+#endif /* CONFIG_X86_64 */
+
+/**************************************************************************
+SETJMP - Save stack context for non-local goto
+**************************************************************************/
+	.globl	setjmp
+setjmp:
+	movl	4(%esp),%ecx		/* jmpbuf */
+	movl	0(%esp),%edx		/* return address */
+	movl	%edx,0(%ecx)
+	movl	%ebx,4(%ecx)
+	movl	%esp,8(%ecx)
+	movl	%ebp,12(%ecx)
+	movl	%esi,16(%ecx)
+	movl	%edi,20(%ecx)
+	movl	$0,%eax
+	ret
+
+/**************************************************************************
+LONGJMP - Non-local jump to a saved stack context
+**************************************************************************/
+	.globl	longjmp
+longjmp:
+	movl	4(%esp),%edx		/* jumpbuf */
+	movl	8(%esp),%eax		/* result */
+	movl	0(%edx),%ecx
+	movl	4(%edx),%ebx
+	movl	8(%edx),%esp
+	movl	12(%edx),%ebp
+	movl	16(%edx),%esi
+	movl	20(%edx),%edi
+	cmpl	$0,%eax
+	jne	1f
+	movl	$1,%eax
+1:	movl	%ecx,0(%esp)
+	ret
+
+/**************************************************************************
+_VIRT_TO_PHYS - Transition from virtual to physical addresses
+		Preserves all preservable registers and flags
+**************************************************************************/
+	.globl _virt_to_phys
+_virt_to_phys:
+	pushfl
+	pushl	%ebp
+	pushl	%eax
+	movl	virt_offset, %ebp	/* Load virt_offset */
+	addl	%ebp, 12(%esp)		/* Adjust the return address */
+
+	/* reload the code segment */
+	pushl	$FLAT_CODE_SEG
+	leal	1f(%ebp), %eax
+	pushl	%eax
+	lret
+
+1:
+	/* reload other segment registers */
+	movl	$FLAT_DATA_SEG, %eax
+	movl	%eax, %ds
+	movl	%eax, %es	
+	movl	%eax, %ss	
+	addl	%ebp, %esp		/* Adjust the stack pointer */
+	movl	%eax, %fs	
+	movl	%eax, %gs
+
+	popl	%eax
+	popl	%ebp
+	popfl
+	ret
+
+
+/**************************************************************************
+_PHYS_TO_VIRT - Transition from using physical to virtual addresses
+		Preserves all preservable registers and flags
+**************************************************************************/
+	.globl _phys_to_virt
+_phys_to_virt:
+	pushfl
+	pushl	%ebp
+	pushl	%eax
+
+	call	1f
+1:	popl	%ebp
+	subl	$1b, %ebp
+	movl	%ebp, virt_offset(%ebp)
+
+	/* Fixup the gdt */
+	leal	_pmcs(%ebp), %eax
+	pushl	%eax
+	pushl	%ebp
+	call	set_seg_base
+	addl	$8, %esp
+
+	/* Fixup gdtarg */
+	leal	_gdt(%ebp), %eax
+	movl	%eax, (gdtarg+2)(%ebp)
+
+	/* Load the global descriptor table */
+	cli
+	lgdt	%cs:gdtarg(%ebp)
+	ljmp	$KERN_CODE_SEG, $1f
+1:
+	/* reload other segment regsters */
+	movl	$KERN_DATA_SEG, %eax
+	movl	%eax, %ds
+	movl	%eax, %es	
+	movl	%eax, %ss	
+	subl	%ebp, %esp	/* Adjust the stack pointer */
+	movl	%eax, %fs	
+	movl	%eax, %gs
+
+	subl	%ebp, 12(%esp)	/* Adjust the return address */
+	popl	%eax
+	popl	%ebp
+	popfl
+	ret
+	
+
+/**************************************************************************
+SET_SEG_BASE - Set the base address of a segment register
+**************************************************************************/
+	.globl set_seg_base
+set_seg_base:
+	pushl	%eax
+	pushl	%ebx
+	movl	12(%esp), %eax		/* %eax = base address */
+	movl	16(%esp), %ebx		/* %ebx = &code_descriptor */
+	movw	%ax, (0+2)(%ebx)	/* CS base bits 0-15 */
+	movw	%ax, (8+2)(%ebx)	/* DS base bits 0-15 */
+	shrl	$16, %eax
+	movb	%al, (0+4)(%ebx)	/* CS base bits 16-23 */
+	movb	%al, (8+4)(%ebx)	/* DS base bits 16-23 */
+	movb	%ah, (0+7)(%ebx)	/* CS base bits 24-31 */
+	movb	%ah, (8+7)(%ebx)	/* DS base bits 24-31 */
+	popl	%ebx
+	popl	%eax
+	ret
+	
+/**************************************************************************
+GLOBAL DESCRIPTOR TABLE
+**************************************************************************/
+	.data
+	.align	4
+
+	.globl	_gdt
+	.globl	gdtarg
+_gdt:
+gdtarg:
+	.word	_gdt_end - _gdt - 1	/* limit */
+	.long	_gdt			/* addr */
+	.word	0
+
+	.globl	_pmcs
+_pmcs:
+	/* 32 bit protected mode code segment */
+	.word	0xffff,0
+	.byte	0,0x9f,0xcf,0
+
+_pmds:
+	/* 32 bit protected mode data segment */
+	.word	0xffff,0
+	.byte	0,0x93,0xcf,0
+
+_rmcs:
+	/* 16 bit real mode code segment */
+	.word	0xffff,(0&0xffff)
+	.byte	(0>>16),0x9b,0x00,(0>>24)
+
+_rmds:
+	/* 16 bit real mode data segment */
+	.word	0xffff,(0&0xffff)
+	.byte	(0>>16),0x93,0x00,(0>>24)
+
+_pmcs2:
+	/* 32 bit protected mode code segment, base 0 */
+	.word	0xffff,0
+	.byte	0,0x9f,0xcf,0
+
+_pmds2:
+	/* 32 bit protected mode data segment, base 0 */
+	.word	0xffff,0
+	.byte	0,0x93,0xcf,0
+
+#ifdef CONFIG_X86_64
+_lmcs:
+	/* 64bit long mode code segment, base 0 */
+	.word	0xffff, 0
+	.byte	0x00, 0x9f, 0xaf , 0x00
+_lmds:
+	/* 64bit long mode data segment, base 0 */
+	.word	0xffff, 0
+	.byte	0x00, 0x93, 0xcf, 0x00
+#endif
+_gdt_end:
+
+	/* The initial register contents */
+	.balign 4
+	.globl initial_regs
+initial_regs:
+	.fill 8, 4, 0
+
+	/* The virtual address offset  */	
+	.globl virt_offset
+virt_offset:
+	.long  	0
+
+	.section ".stack"
+	.p2align 3
+	/* allocate a 4K stack in the stack segment */
+	.globl	_stack
+_stack:
+	.space 4096
+	.globl	_estack
+_estack:
+#ifdef CONFIG_X86_64
+	.section ".bss"
+	.p2align 12
+	/* Include a dummy space in case we are loaded badly aligned */
+	.space 4096
+	/* Reserve enough space for a page table convering 4GB with 2MB pages */
+pgt_level4:	
+	.space 4096
+pgt_level3:	
+	.space 4096
+pgt_level2:	
+	.space 16384
+start_lm_addr:
+	.space	8
+end_lm_addr:
+	.space	8
+#endif

src/arch/i386/core/tagged_loader.c

@@ -0,0 +1,201 @@
+#include "realmode.h"
+#include "segoff.h"
+
+struct segheader
+{
+	unsigned char length;
+	unsigned char vendortag;
+	unsigned char reserved;
+	unsigned char flags;
+	unsigned long loadaddr;
+	unsigned long imglength;
+	unsigned long memlength;
+};
+
+struct imgheader
+{
+	unsigned long magic;
+	unsigned long length;			/* and flags */
+	union
+	{
+		segoff_t segoff;
+		unsigned long location;
+	} u;
+	unsigned long execaddr;
+};
+
+/* Keep all context about loaded image in one place */
+static struct tagged_context
+{
+	struct imgheader	img;			/* copy of header */
+	unsigned long		linlocation;		/* addr of header */
+	unsigned long		last0, last1;		/* of prev segment */
+	unsigned long		segaddr, seglen;	/* of current segment */
+	unsigned char		segflags;
+	unsigned char		first;
+	unsigned long		curaddr;
+} tctx;
+
+#define	TAGGED_PROGRAM_RETURNS	(tctx.img.length & 0x00000100)	/* bit 8 */
+#define	LINEAR_EXEC_ADDR	(tctx.img.length & 0x80000000)	/* bit 31 */
+
+static sector_t tagged_download(unsigned char *data, unsigned int len, int eof);
+void xstart16 (unsigned long execaddr, segoff_t location,
+	       void *bootp);
+
+static inline os_download_t tagged_probe(unsigned char *data, unsigned int len)
+{
+	struct segheader	*sh;
+	unsigned long loc;
+	if (*((uint32_t *)data) != 0x1B031336L) {
+		return 0;
+	}
+	printf("(NBI)");
+	/* If we don't have enough data give up */
+	if (len < 512)
+		return dead_download;
+	/* Zero all context info */
+	memset(&tctx, 0, sizeof(tctx));
+	/* Copy first 4 longwords */
+	memcpy(&tctx.img, data, sizeof(tctx.img));
+	/* Memory location where we are supposed to save it */
+	tctx.segaddr = tctx.linlocation = 
+		((tctx.img.u.segoff.segment) << 4) + tctx.img.u.segoff.offset;
+	if (!prep_segment(tctx.segaddr, tctx.segaddr + 512, tctx.segaddr + 512,
+			  0, 512)) {
+		return dead_download;
+	}
+	/* Now verify the segments we are about to load */
+	loc = 512;
+	for(sh = (struct segheader *)(data
+				      + ((tctx.img.length & 0x0F) << 2)
+				      + ((tctx.img.length & 0xF0) >> 2) ); 
+		(sh->length > 0) && ((unsigned char *)sh < data + 512); 
+		sh = (struct segheader *)((unsigned char *)sh
+					  + ((sh->length & 0x0f) << 2) + ((sh->length & 0xf0) >> 2)) ) {
+		if (!prep_segment(
+			sh->loadaddr,
+			sh->loadaddr + sh->imglength,
+			sh->loadaddr + sh->imglength,
+			loc, loc + sh->imglength)) {
+			return dead_download;
+		}
+		loc = loc + sh->imglength;
+		if (sh->flags & 0x04) 
+			break;
+	}
+	if (!(sh->flags & 0x04))
+		return dead_download;
+	/* Grab a copy */
+	memcpy(phys_to_virt(tctx.segaddr), data, 512);
+	/* Advance to first segment descriptor */
+	tctx.segaddr += ((tctx.img.length & 0x0F) << 2)
+		+ ((tctx.img.length & 0xF0) >> 2);
+	/* Remember to skip the first 512 data bytes */
+	tctx.first = 1;
+	
+	return tagged_download;
+
+}
+static sector_t tagged_download(unsigned char *data, unsigned int len, int eof)
+{
+	int	i;
+
+	if (tctx.first) {
+		tctx.first = 0;
+		if (len > 512) {
+			len -= 512;
+			data += 512;
+			/* and fall through to deal with rest of block */
+		} else 
+			return 0;
+	}
+	for (;;) {
+		if (len == 0) /* Detect truncated files */
+			eof = 0;
+		while (tctx.seglen == 0) {
+			struct segheader	sh;
+			if (tctx.segflags & 0x04) {
+				done(1);
+				if (LINEAR_EXEC_ADDR) {
+					int result;
+					/* no gateA20_unset for PM call */
+					result = xstart32(tctx.img.execaddr,
+						virt_to_phys(&loaderinfo),
+						tctx.linlocation,
+						virt_to_phys(BOOTP_DATA_ADDR));
+					printf("Secondary program returned %d\n",
+						result);
+					if (!TAGGED_PROGRAM_RETURNS) {
+						/* We shouldn't have returned */
+						result = -2;
+					}
+					if (result == 0)
+						result = -2;
+					longjmp(restart_etherboot, result);
+						
+				} else {
+					gateA20_unset();
+					xstart16(tctx.img.execaddr, 
+						 tctx.img.u.segoff,
+						 BOOTP_DATA_ADDR);
+					longjmp(restart_etherboot, -2);
+				}
+			}
+			sh = *((struct segheader *)phys_to_virt(tctx.segaddr));
+			tctx.seglen = sh.imglength;
+			if ((tctx.segflags = sh.flags & 0x03) == 0)
+				tctx.curaddr = sh.loadaddr;
+			else if (tctx.segflags == 0x01)
+				tctx.curaddr = tctx.last1 + sh.loadaddr;
+			else if (tctx.segflags == 0x02)
+				tctx.curaddr = (Address)(meminfo.memsize * 1024L
+						    + 0x100000L)
+					- sh.loadaddr;
+			else
+				tctx.curaddr = tctx.last0 - sh.loadaddr;
+			tctx.last1 = (tctx.last0 = tctx.curaddr) + sh.memlength;
+			tctx.segflags = sh.flags;
+			tctx.segaddr += ((sh.length & 0x0F) << 2)
+				+ ((sh.length & 0xF0) >> 2);
+			/* Avoid lock-up */
+			if ( sh.length == 0 ) longjmp(restart_etherboot, -2); 
+		}
+		if ((len <= 0) && !eof)
+			break;
+		i = (tctx.seglen > len) ? len : tctx.seglen;
+		memcpy(phys_to_virt(tctx.curaddr), data, i);
+		tctx.seglen -= i;
+		tctx.curaddr += i;
+		len -= i;
+		data += i;
+	} 
+	return 0;
+}
+
+void xstart16 (unsigned long execaddr, segoff_t location,
+	       void *bootp) {
+	struct {
+		segoff_t execaddr;
+		segoff_t location;
+		segoff_t bootp;
+	} PACKED in_stack;
+
+	/* AFAICT, execaddr is actually already a segment:offset */
+	*((unsigned long *)&in_stack.execaddr) = execaddr;
+	in_stack.location = location;
+	in_stack.bootp.segment = SEGMENT(bootp);
+	in_stack.bootp.offset = OFFSET(bootp);
+
+	RM_FRAGMENT(rm_xstart16,
+		"popl %eax\n\t"	/* Calculated lcall */
+		"pushw %cs\n\t" 
+		"call 1f\n1:\tpopw %bp\n\t" 
+		"leaw (2f-1b)(%bp), %bx\n\t" 
+		"pushw %bx\n\t" 
+		"pushl %eax\n\t"
+		"lret\n2:\n\t"
+	);
+       
+	real_call ( rm_xstart16, &in_stack, NULL );
+}

src/arch/i386/core/video_subr.c

@@ -0,0 +1,94 @@
+/*
+ *
+ * modified from linuxbios code
+ * by Cai Qiang <rimy2000@hotmail.com>
+ *
+ */
+
+#ifdef CONSOLE_DIRECT_VGA
+
+#include <etherboot.h>
+#include <vga.h>
+
+static char *vidmem;		/* The video buffer */
+static int video_line, video_col;
+
+#define VIDBUFFER 0xB8000	
+
+static void memsetw(void *s, int c, unsigned int n)
+{
+	int i;
+	u16 *ss = (u16 *) s;
+
+	for (i = 0; i < n; i++) {
+		ss[i] = ( u16 ) c;
+	}
+}
+
+void video_init(void)
+{
+	static int inited=0;
+
+	vidmem = (unsigned char *)phys_to_virt(VIDBUFFER);
+
+	if (!inited) {
+		video_line = 0;
+		video_col = 0;
+	
+	 	memsetw(vidmem, VGA_ATTR_CLR_WHT, 2*1024); //
+
+		inited=1;
+	}
+}
+
+static void video_scroll(void)
+{
+	int i;
+
+	memcpy(vidmem, vidmem + COLS * 2, (LINES - 1) * COLS * 2);
+	for (i = (LINES - 1) * COLS * 2; i < LINES * COLS * 2; i += 2)
+		vidmem[i] = ' ';
+}
+
+void vga_putc(unsigned char byte)
+{
+	if (byte == '\n') {
+		video_line++;
+		video_col = 0;
+
+	} else if (byte == '\r') {
+		video_col = 0;
+
+	} else if (byte == '\b') {
+		video_col--;
+
+	} else if (byte == '\t') {
+		video_col += 4;
+
+	} else if (byte == '\a') {
+		//beep
+		//beep(500);
+
+	} else {
+		vidmem[((video_col + (video_line *COLS)) * 2)] = byte;
+		vidmem[((video_col + (video_line *COLS)) * 2) +1] = VGA_ATTR_CLR_WHT;
+		video_col++;
+	}
+	if (video_col < 0) {
+		video_col = 0;
+	}
+	if (video_col >= COLS) {
+		video_line++;
+		video_col = 0;
+	}
+	if (video_line >= LINES) {
+		video_scroll();
+		video_line--;
+	}
+	// move the cursor
+	write_crtc((video_col + (video_line *COLS)) >> 8, CRTC_CURSOR_HI);
+	write_crtc((video_col + (video_line *COLS)) & 0x0ff, CRTC_CURSOR_LO);
+}
+
+#endif
+

src/arch/i386/core/wince_loader.c

@@ -0,0 +1,273 @@
+#define LOAD_DEBUG	0
+
+static int get_x_header(unsigned char *data, unsigned long now);
+static void jump_2ep();
+static unsigned char ce_signature[] = {'B', '0', '0', '0', 'F', 'F', '\n',};
+static char ** ep;
+
+#define BOOT_ARG_PTR_LOCATION 0x001FFFFC
+
+typedef struct _BOOT_ARGS{
+	unsigned char ucVideoMode;
+	unsigned char ucComPort;
+	unsigned char ucBaudDivisor;
+	unsigned char ucPCIConfigType;
+	
+	unsigned long dwSig;
+	#define BOOTARG_SIG 0x544F4F42
+	unsigned long dwLen;
+	
+	unsigned char ucLoaderFlags;
+	unsigned char ucEshellFlags;
+	unsigned char ucEdbgAdapterType;
+	unsigned char ucEdbgIRQ;
+	
+	unsigned long dwEdbgBaseAddr;
+	unsigned long dwEdbgDebugZone;	
+	unsigned long dwDHCPLeaseTime;
+	unsigned long dwEdbgFlags;
+	
+	unsigned long dwEBootFlag;
+	unsigned long dwEBootAddr;
+	unsigned long dwLaunchAddr;
+	
+	unsigned long pvFlatFrameBuffer;
+	unsigned short vesaMode;
+	unsigned short cxDisplayScreen;
+	unsigned short cyDisplayScreen;
+	unsigned short cxPhysicalScreen;
+	unsigned short cyPhysicalScreen;
+	unsigned short cbScanLineLength;
+	unsigned short bppScreen;
+	
+	unsigned char RedMaskSize;
+	unsigned char REdMaskPosition;
+	unsigned char GreenMaskSize;
+	unsigned char GreenMaskPosition;
+	unsigned char BlueMaskSize;
+	unsigned char BlueMaskPosition;
+} BOOT_ARGS;
+
+BOOT_ARGS BootArgs;
+
+static struct segment_info{
+	unsigned long addr;		// Section Address
+	unsigned long size;		// Section Size
+	unsigned long checksum;		// Section CheckSum
+} X;
+
+#define PSIZE	(1500)			//Max Packet Size
+#define DSIZE  (PSIZE+12)
+static unsigned long dbuffer_available =0;
+static unsigned long not_loadin =0;
+static unsigned long d_now =0;
+
+unsigned long entry;
+static unsigned long ce_curaddr;
+
+
+static sector_t ce_loader(unsigned char *data, unsigned int len, int eof);
+static os_download_t wince_probe(unsigned char *data, unsigned int len)
+{
+	if (strncmp(ce_signature, data, sizeof(ce_signature)) != 0) {
+		return 0;
+	}
+	printf("(WINCE)");
+	return ce_loader;
+}
+
+static sector_t ce_loader(unsigned char *data, unsigned int len, int eof)
+{
+	static unsigned char dbuffer[DSIZE];
+	int this_write = 0;
+	static int firsttime = 1;
+
+	/*
+	 *	new packet in, we have to 
+	 *	[1] copy data to dbuffer,
+	 *
+	 *	update...
+	 *	[2]  dbuffer_available
+	 */
+	memcpy( (dbuffer+dbuffer_available), data, len);	//[1]
+	dbuffer_available += len;	// [2]
+	len = 0;
+
+	d_now = 0;
+	
+#if 0
+	printf("dbuffer_available =%ld \n", dbuffer_available);
+#endif 
+	
+	if (firsttime) 
+	{
+		d_now = sizeof(ce_signature);
+		printf("String Physical Address = %lx \n", 
+			*(unsigned long *)(dbuffer+d_now));
+		
+		d_now += sizeof(unsigned long);
+		printf("Image Size = %ld [%lx]\n", 
+			*(unsigned long *)(dbuffer+d_now), 
+			*(unsigned long *)(dbuffer+d_now));
+		
+		d_now += sizeof(unsigned long);
+		dbuffer_available -= d_now;			
+		
+		d_now = (unsigned long)get_x_header(dbuffer, d_now);
+		firsttime = 0;
+	}
+	
+	if (not_loadin == 0)
+	{
+		d_now = get_x_header(dbuffer, d_now);
+	}
+	
+	while ( not_loadin > 0 )
+	{
+		/* dbuffer do not have enough data to loading, copy all */
+#if LOAD_DEBUG
+		printf("[0] not_loadin = [%ld], dbuffer_available = [%ld] \n", 
+			not_loadin, dbuffer_available);
+		printf("[0] d_now = [%ld] \n", d_now);
+#endif
+		
+		if( dbuffer_available <= not_loadin)
+		{
+			this_write = dbuffer_available ;
+			memcpy(phys_to_virt(ce_curaddr), (dbuffer+d_now), this_write );
+			ce_curaddr += this_write;
+			not_loadin -= this_write;
+			
+			/* reset index and available in the dbuffer */
+			dbuffer_available = 0;
+			d_now = 0;
+#if LOAD_DEBUG
+			printf("[1] not_loadin = [%ld], dbuffer_available = [%ld] \n", 
+				not_loadin, dbuffer_available);
+			printf("[1] d_now = [%ld], this_write = [%d] \n", 
+				d_now, this_write);
+#endif
+				
+			// get the next packet...
+			return (0);
+		}
+			
+		/* dbuffer have more data then loading ... , copy partital.... */
+		else
+		{
+			this_write = not_loadin;
+			memcpy(phys_to_virt(ce_curaddr), (dbuffer+d_now), this_write);
+			ce_curaddr += this_write;
+			not_loadin = 0;
+			
+			/* reset index and available in the dbuffer */
+			dbuffer_available -= this_write;
+			d_now += this_write;
+#if LOAD_DEBUG
+			printf("[2] not_loadin = [%ld], dbuffer_available = [%ld] \n", 
+				not_loadin, dbuffer_available);
+			printf("[2] d_now = [%ld], this_write = [%d] \n\n", 
+				d_now, this_write);
+#endif
+			
+			/* dbuffer not empty, proceed processing... */
+			
+			// don't have enough data to get_x_header..
+			if ( dbuffer_available < (sizeof(unsigned long) * 3) )
+			{
+//				printf("we don't have enough data remaining to call get_x. \n");
+				memcpy( (dbuffer+0), (dbuffer+d_now), dbuffer_available);
+				return (0);
+			}
+			else
+			{
+#if LOAD_DEBUG				
+				printf("with remaining data to call get_x \n");
+				printf("dbuffer available = %ld , d_now = %ld\n", 
+					dbuffer_available, d_now);
+#endif					
+				d_now = get_x_header(dbuffer, d_now);
+			}
+		}
+	}
+	return (0);
+}
+
+static int get_x_header(unsigned char *dbuffer, unsigned long now)
+{
+	X.addr = *(unsigned long *)(dbuffer + now);
+	X.size = *(unsigned long *)(dbuffer + now + sizeof(unsigned long));
+	X.checksum = *(unsigned long *)(dbuffer + now + sizeof(unsigned long)*2);
+
+	if (X.addr == 0)
+	{
+		entry = X.size;
+		done(1);
+		printf("Entry Point Address = [%lx] \n", entry);
+		jump_2ep();		
+	}
+
+	if (!prep_segment(X.addr, X.addr + X.size, X.addr + X.size, 0, 0)) {
+		longjmp(restart_etherboot, -2);
+	}
+
+	ce_curaddr = X.addr;
+	now += sizeof(unsigned long)*3;
+
+	/* re-calculate dbuffer available... */
+	dbuffer_available -= sizeof(unsigned long)*3;
+
+	/* reset index of this section */
+	not_loadin = X.size;
+	
+#if 1
+	printf("\n");
+	printf("\t Section Address = [%lx] \n", X.addr);
+	printf("\t Size = %d [%lx]\n", X.size, X.size);
+	printf("\t Checksum = %ld [%lx]\n", X.checksum, X.checksum);
+#endif
+#if LOAD_DEBUG
+	printf("____________________________________________\n");
+	printf("\t dbuffer_now = %ld \n", now);
+	printf("\t dbuffer available = %ld \n", dbuffer_available);
+	printf("\t not_loadin = %ld \n", not_loadin);
+#endif
+
+	return now;
+}
+
+static void jump_2ep()
+{
+	BootArgs.ucVideoMode = 1;
+	BootArgs.ucComPort = 1;
+	BootArgs.ucBaudDivisor = 1;
+	BootArgs.ucPCIConfigType = 1;	// do not fill with 0
+	
+	BootArgs.dwSig = BOOTARG_SIG;
+	BootArgs.dwLen = sizeof(BootArgs);
+	
+	if(BootArgs.ucVideoMode == 0)
+	{
+		BootArgs.cxDisplayScreen = 640;
+		BootArgs.cyDisplayScreen = 480;
+		BootArgs.cxPhysicalScreen = 640;
+		BootArgs.cyPhysicalScreen = 480;
+		BootArgs.bppScreen = 16;
+		BootArgs.cbScanLineLength  = 1024;
+		BootArgs.pvFlatFrameBuffer = 0x800a0000;	// ollie say 0x98000000
+	}	
+	else if(BootArgs.ucVideoMode != 0xFF)
+	{
+		BootArgs.cxDisplayScreen = 0;
+		BootArgs.cyDisplayScreen = 0;
+		BootArgs.cxPhysicalScreen = 0;
+		BootArgs.cyPhysicalScreen = 0;
+		BootArgs.bppScreen = 0;
+		BootArgs.cbScanLineLength  = 0;
+		BootArgs.pvFlatFrameBuffer = 0;	
+	}
+
+	ep = phys_to_virt(BOOT_ARG_PTR_LOCATION);
+	*ep= virt_to_phys(&BootArgs);
+	xstart32(entry);
+}

src/arch/i386/drivers/net/undi.h

@@ -0,0 +1,178 @@
+/**************************************************************************
+Etherboot -  BOOTP/TFTP Bootstrap Program
+UNDI NIC driver for Etherboot - header file
+
+This file Copyright (C) 2003 Michael Brown <mbrown@fensystems.co.uk>
+of Fen Systems Ltd. (http://www.fensystems.co.uk/).  All rights
+reserved.
+
+$Id$
+***************************************************************************/
+
+/*
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2, or (at
+ * your option) any later version.
+ */
+
+#include "pxe.h"
+#include "pic8259.h"
+
+/* A union that can function as the parameter block for any UNDI API call.
+ */
+typedef t_PXENV_ANY pxenv_structure_t;
+
+/* BIOS PnP parameter block.  We scan for this so that we can pass it
+ * to the UNDI driver.
+ */
+
+#define PNP_BIOS_SIGNATURE ( ('$'<<0) + ('P'<<8) + ('n'<<16) + ('P'<<24) )
+typedef struct pnp_bios {
+	uint32_t	signature;
+	uint8_t		version;
+	uint8_t		length;
+	uint16_t	control;
+	uint8_t		checksum;
+	uint8_t		dontcare[24];
+} PACKED pnp_bios_t;
+
+/* Structures within the PXE ROM.
+ */
+
+#define ROM_SIGNATURE 0xaa55
+typedef struct rom {
+	uint16_t	signature;
+	uint8_t		unused[0x14];
+	uint16_t	undi_rom_id_off;
+	uint16_t	pcir_off;
+	uint16_t	pnp_off;
+} PACKED rom_t;	
+
+#define PCIR_SIGNATURE ( ('P'<<0) + ('C'<<8) + ('I'<<16) + ('R'<<24) )
+typedef struct pcir_header {
+	uint32_t	signature;
+	uint16_t	vendor_id;
+	uint16_t	device_id;
+} PACKED pcir_header_t;
+
+#define PNP_SIGNATURE ( ('$'<<0) + ('P'<<8) + ('n'<<16) + ('P'<<24) )
+typedef struct pnp_header {
+	uint32_t	signature;
+	uint8_t		struct_revision;
+	uint8_t		length;
+	uint16_t	next;
+	uint8_t		reserved;
+	uint8_t		checksum;
+	uint16_t	id[2];
+	uint16_t	manuf_str_off;
+	uint16_t	product_str_off;
+	uint8_t		base_type;
+	uint8_t		sub_type;
+	uint8_t		interface_type;
+	uint8_t		indicator;
+	uint16_t	boot_connect_off;
+	uint16_t	disconnect_off;
+	uint16_t	initialise_off;
+	uint16_t	reserved2;
+	uint16_t	info;
+} PACKED pnp_header_t;
+
+#define UNDI_SIGNATURE ( ('U'<<0) + ('N'<<8) + ('D'<<16) + ('I'<<24) )
+typedef struct undi_rom_id {
+	uint32_t	signature;
+	uint8_t		struct_length;
+	uint8_t		struct_cksum;
+	uint8_t		struct_rev;
+	uint8_t		undi_rev[3];
+	uint16_t	undi_loader_off;
+	uint16_t	stack_size;
+	uint16_t	data_size;
+	uint16_t	code_size;
+} PACKED undi_rom_id_t;
+
+/* Nontrivial IRQ handler structure */
+typedef struct {
+	segoff_t		chain_to;
+	uint8_t			irq_chain, pad1, pad2, pad3;
+	segoff_t		entry;
+	uint16_t		count_all;
+	uint16_t		count_ours;
+	t_PXENV_UNDI_ISR	undi_isr;
+	char			code[0];
+} PACKED undi_irq_handler_t ;
+
+/* Storage buffers that we need in base memory.  We collect these into
+ * a single structure to make allocation simpler.
+ */
+
+typedef struct undi_base_mem_xmit_data {
+	MAC_ADDR		destaddr;
+	t_PXENV_UNDI_TBD	tbd;
+} undi_base_mem_xmit_data_t;
+
+typedef struct undi_base_mem_data {
+	pxenv_structure_t	pxs;
+	undi_base_mem_xmit_data_t xmit_data;
+	char			xmit_buffer[ETH_FRAME_LEN];
+	/* Must be last in structure and paragraph-aligned */
+	union {
+		char			e820mangler[0];
+		char			irq_handler[0];
+		undi_irq_handler_t	nontrivial_irq_handler;
+	}  __attribute__ ((aligned(16)));
+} undi_base_mem_data_t;
+
+/* Macros and data structures used when freeing bits of base memory
+ * used by the UNDI driver.
+ */
+
+#define FIRING_SQUAD_TARGET_SIZE 8
+#define FIRING_SQUAD_TARGET_INDEX(x) ( (x) / FIRING_SQUAD_TARGET_SIZE )
+#define FIRING_SQUAD_TARGET_BIT(x) ( (x) % FIRING_SQUAD_TARGET_SIZE )
+typedef struct firing_squad_lineup {
+	uint8_t targets[ 640 / FIRING_SQUAD_TARGET_SIZE ];
+} firing_squad_lineup_t;
+typedef enum firing_squad_shoot {
+	DONTSHOOT = 0,
+	SHOOT = 1
+} firing_squad_shoot_t;
+
+/* Driver private data structure.
+ */
+
+typedef struct undi {
+	/* Pointers to various data structures */
+	pnp_bios_t		*pnp_bios;
+	rom_t			*rom;
+	undi_rom_id_t		*undi_rom_id;
+	pxe_t			*pxe;
+	pxenv_structure_t	*pxs;
+	undi_base_mem_xmit_data_t *xmit_data;
+	/* Pointers and sizes to keep track of allocated base memory */
+	undi_base_mem_data_t	*base_mem_data;
+	void			*driver_code;
+	size_t			driver_code_size;
+	void			*driver_data;
+	size_t			driver_data_size;
+	char			*xmit_buffer;
+	/* Flags.  We keep our own instead of trusting the UNDI driver
+	 * to have implemented PXENV_UNDI_GET_STATE correctly.  Plus
+	 * there's the small issue of PXENV_UNDI_GET_STATE being the
+	 * same API call as PXENV_STOP_UNDI...
+	 */
+	uint8_t prestarted;	/* pxenv_start_undi() has been called */
+	uint8_t started;	/* pxenv_undi_startup() has been called */
+	uint8_t	initialized;	/* pxenv_undi_initialize() has been called */
+	uint8_t opened;		/* pxenv_undi_open() has been called */
+	/* Parameters that we need to store for future reference
+	 */
+	struct pci_device	pci;
+	irq_t			irq;
+} undi_t;
+
+/* Constants
+ */
+
+#define HUNT_FOR_PIXIES 0
+#define HUNT_FOR_UNDI_ROMS 1

src/arch/i386/firmware/pcbios/basemem.c

@@ -0,0 +1,317 @@
+#ifdef PCBIOS
+
+#include "etherboot.h"
+#include "realmode.h" /* for real_mode_stack */
+
+/* Routines to allocate base memory in a BIOS-compatible way, by
+ * updating the Free Base Memory Size counter at 40:13h.
+ *
+ * Michael Brown <mbrown@fensystems.co.uk> (mcb30)
+ * $Id$
+ */
+
+#define fbms ( ( uint16_t * ) phys_to_virt ( 0x413 ) )
+#define BASE_MEMORY_MAX ( 640 )
+#define FREE_BLOCK_MAGIC ( ('!'<<0) + ('F'<<8) + ('R'<<16) + ('E'<<24) )
+#define FREE_BASE_MEMORY ( (uint32_t) ( *fbms << 10 ) )
+
+/* Prototypes */
+void * _allot_base_memory ( size_t size );
+void _forget_base_memory ( void *ptr, size_t size );
+
+typedef struct free_base_memory_block {
+	uint32_t	magic;
+	uint16_t	size_kb;
+} free_base_memory_block_t;
+
+/* Return amount of free base memory in bytes
+ */
+
+uint32_t get_free_base_memory ( void ) {
+	return FREE_BASE_MEMORY;
+}
+
+/* Start of our image in base memory.
+ */
+#define __text16_nocompress __attribute__ ((section (".text16.nocompress")))
+uint32_t image_basemem __text16_nocompress = 0;
+uint32_t image_basemem_size __text16_nocompress = 0;
+
+/* Allot/free the real-mode stack
+ */
+
+void allot_real_mode_stack ( void )
+{
+	void *new_real_mode_stack;
+
+	if ( lock_real_mode_stack ) 
+		return;
+
+	/* This is evil hack. 
+	 * Until we have a real_mode stack use 0x7c00.
+	 * Except for 0 - 0x600 membory below 0x7c00 is hardly every used.
+	 * This stack should never be used unless the stack allocation fails,
+	 * or if someone has placed a print statement in a dangerous location.
+	 */
+	if (!real_mode_stack) {
+		real_mode_stack = 0x7c00;
+	}
+	new_real_mode_stack = _allot_base_memory ( real_mode_stack_size );
+	if ( ! new_real_mode_stack ) {
+		printf ( "FATAL: No real-mode stack\n" );
+		while ( 1 ) {};
+	}
+	real_mode_stack = virt_to_phys ( new_real_mode_stack );
+	get_memsizes();
+}
+
+void forget_real_mode_stack ( void )
+{
+	if ( lock_real_mode_stack ) 
+		return;
+
+	if ( real_mode_stack) {
+		_forget_base_memory ( phys_to_virt(real_mode_stack),
+				      real_mode_stack_size );
+		/* get_memsizes() uses the real_mode stack we just freed
+		 * for it's BIOS calls.
+		 */
+		get_memsizes();
+		real_mode_stack = 0;
+	}
+}
+
+/* Allocate N bytes of base memory.  Amount allocated will be rounded
+ * up to the nearest kB, since that's the granularity of the BIOS FBMS
+ * counter.  Returns NULL if memory cannot be allocated.
+ */
+
+static void * _allot_base_memory ( size_t size ) 
+{
+	uint16_t size_kb = ( size + 1023 ) >> 10;
+	void *ptr = NULL;
+
+#ifdef DEBUG_BASEMEM
+	printf ( "Trying to allocate %d kB of base memory from %d kB free\n",
+		 size_kb, *fbms );
+#endif
+
+	/* Free up any unused memory before we start */
+	free_unused_base_memory();
+
+	/* Check available base memory */
+	if ( size_kb > *fbms ) { return NULL; }
+
+	/* Reduce available base memory */
+	*fbms -= size_kb;
+
+	/* Calculate address of memory allocated */
+	ptr = phys_to_virt ( FREE_BASE_MEMORY );
+
+	/* Zero out memory.  We do this so that allocation of
+	 * already-used space will show up in the form of a crash as
+	 * soon as possible.
+	 *
+	 * Update: there's another reason for doing this.  If we don't
+	 * zero the contents, then they could still retain our "free
+	 * block" markers and be liable to being freed whenever a
+	 * base-memory allocation routine is next called.
+	 */
+	memset ( ptr, 0, size_kb << 10 );
+
+#ifdef DEBUG_BASEMEM
+	printf ( "Allocated %d kB at [%x,%x)\n", size_kb,
+		 virt_to_phys ( ptr ),
+		 virt_to_phys ( ptr ) + size_kb * 1024 );
+#endif
+
+	return ptr;
+}
+
+void * allot_base_memory ( size_t size )
+{
+	void *ptr;
+
+	/* Free real-mode stack, allocate memory, reallocate real-mode
+	 * stack.
+	 */
+	forget_real_mode_stack();
+	ptr = _allot_base_memory ( size );
+	get_memsizes();
+	return ptr;
+}
+
+/* Free base memory allocated by allot_base_memory.  The BIOS provides
+ * nothing better than a LIFO mechanism for freeing memory (i.e. it
+ * just has the single "total free memory" counter), but we improve
+ * upon this slightly; as long as you free all the allotted blocks, it
+ * doesn't matter what order you free them in.  (This will only work
+ * for blocks that are freed via forget_base_memory()).
+ *
+ * Yes, it's annoying that you have to remember the size of the blocks
+ * you've allotted.  However, since our granularity of allocation is
+ * 1K, the alternative is to risk wasting the occasional kB of base
+ * memory, which is a Bad Thing.  Really, you should be using as
+ * little base memory as possible, so consider the awkwardness of the
+ * API to be a feature! :-)
+ */
+
+static void _forget_base_memory ( void *ptr, size_t size ) 
+{
+	uint16_t remainder = virt_to_phys(ptr) & 1023;
+	uint16_t size_kb = ( size + remainder + 1023 ) >> 10;
+	free_base_memory_block_t *free_block =
+		( free_base_memory_block_t * ) ( ptr - remainder );
+	
+	if ( ( ptr == NULL ) || ( size == 0 ) ) { 
+		return; 
+	}
+
+#ifdef DEBUG_BASEMEM
+	printf ( "Trying to free %d bytes base memory at 0x%x\n",
+		 size, virt_to_phys ( ptr ) );
+	if ( remainder > 0 ) {
+		printf ( "WARNING: destructively expanding free block "
+			 "downwards to 0x%x\n",
+			 virt_to_phys ( ptr - remainder ) );
+	}
+#endif
+
+	/* Mark every kilobyte within this block as free.  This is
+	 * overkill for normal purposes, but helps when something has
+	 * allocated base memory with a granularity finer than the
+	 * BIOS granularity of 1kB.  PXE ROMs tend to do this when
+	 * they allocate their own memory.  This method allows us to
+	 * free their blocks (admittedly in a rather dangerous,
+	 * tread-on-anything-either-side sort of way, but there's no
+	 * other way to do it).
+	 *
+	 * Since we're marking every kB as free, there's actually no
+	 * need for recording the size of the blocks.  However, we
+	 * keep this in so that debug messages are friendlier.  It
+	 * probably adds around 8 bytes to the overall code size.
+	 */
+	while ( size_kb > 0 ) {
+		/* Mark this block as unused */
+		free_block->magic = FREE_BLOCK_MAGIC;
+		free_block->size_kb = size_kb;
+		/* Move up by 1 kB */
+		free_block = (void *)(((char *)free_block) + (1 << 10));
+		size_kb--;
+	}
+
+	/* Free up unused base memory */
+	free_unused_base_memory();
+}
+
+void forget_base_memory ( void *ptr, size_t size )
+{
+	/* Free memory, free real-mode stack, re-allocate real-mode
+	 * stack.  Do this so that we don't end up wasting a huge
+	 * block of memory trapped behind the real-mode stack.
+	 */
+	_forget_base_memory ( ptr, size );
+	forget_real_mode_stack();
+	get_memsizes();
+}
+
+/* Do the actual freeing of memory.  This is split out from
+ * forget_base_memory() so that it may be called separately.  It
+ * should be called whenever base memory is deallocated by an external
+ * entity (if we can detect that it has done so) so that we get the
+ * chance to free up our own blocks.
+ */
+static void free_unused_base_memory ( void ) {
+	free_base_memory_block_t *free_block = NULL;
+
+	/* Try to release memory back to the BIOS.  Free all
+	 * consecutive blocks marked as free.
+	 */
+	while ( 1 ) {
+		/* Calculate address of next potential free block */
+		free_block = ( free_base_memory_block_t * )
+			phys_to_virt ( FREE_BASE_MEMORY );
+		
+		/* Stop processing if we're all the way up to 640K or
+		 * if this is not a free block
+		 */
+		if ( ( *fbms == BASE_MEMORY_MAX ) ||
+		     ( free_block->magic != FREE_BLOCK_MAGIC ) ) {
+			break;
+		}
+
+		/* Return memory to BIOS */
+		*fbms += free_block->size_kb;
+
+#ifdef DEBUG_BASEMEM
+		printf ( "Freed %d kB base memory, %d kB now free\n",
+			 free_block->size_kb, *fbms );
+#endif
+		
+		/* Zero out freed block.  We do this in case
+		 * the block contained any structures that
+		 * might be located by scanning through
+		 * memory.
+		 */
+		memset ( free_block, 0, free_block->size_kb << 10 );
+	}
+}
+
+/* Free base memory used by the prefix.  Called once at start of
+ * Etherboot by arch_main().
+ */
+void forget_prefix_base_memory ( void )
+{
+	/* runtime_start_kb is _text rounded down to a physical kB boundary */
+	uint32_t runtime_start_kb = virt_to_phys(_text) & ~0x3ff;
+	/* prefix_size_kb is the prefix size excluding any portion
+	 * that overlaps into the first kB used by the runtime image
+	 */
+	uint32_t prefix_size_kb = runtime_start_kb - image_basemem;
+
+#ifdef DEBUG_BASEMEM
+	printf ( "Attempting to free base memory used by prefix\n" );
+#endif
+
+	/* If the decompressor is in allocated base memory
+	 * *and* the Etherboot text is in base
+	 * memory, then free the decompressor.
+	 */
+	if ( ( image_basemem >= FREE_BASE_MEMORY ) &&
+	     ( runtime_start_kb >= FREE_BASE_MEMORY ) &&
+	     ( runtime_start_kb <= ( BASE_MEMORY_MAX << 10 ) ) ) 
+	{
+		forget_base_memory ( phys_to_virt ( image_basemem ),
+				     prefix_size_kb );
+		/* Update image_basemem and image_basemem_size to
+		 * indicate that our allocation now starts with _text
+		 */
+		image_basemem = runtime_start_kb;
+		image_basemem_size -= prefix_size_kb;
+	}
+}
+
+/* Free base memory used by the runtime image.  Called after
+ * relocation by arch_relocated_from().
+ */
+void forget_runtime_base_memory ( unsigned long old_addr )
+{
+	/* text_start_kb is old _text rounded down to a physical KB boundary */
+	uint32_t old_text_start_kb = old_addr & ~0x3ff;
+
+#ifdef DEBUG_BASEMEM
+	printf ( "Attempting to free base memory used by runtime image\n" );
+#endif
+
+	if ( ( image_basemem >= FREE_BASE_MEMORY ) &&
+	     ( image_basemem == old_text_start_kb ) ) 
+	{
+		forget_base_memory ( phys_to_virt ( image_basemem ),
+				     image_basemem_size );
+		/* Update image_basemem to show no longer in use */
+		image_basemem = 0;
+		image_basemem_size = 0;
+	}
+}
+
+#endif /* PCBIOS */

src/arch/i386/firmware/pcbios/bios.c

@@ -0,0 +1,155 @@
+/* Etherboot routines for PCBIOS firmware.
+ *
+ * Body of routines taken from old pcbios.S
+ */
+
+#ifdef PCBIOS
+
+#include "etherboot.h"
+#include "realmode.h"
+#include "segoff.h"
+
+#define CF ( 1 << 0 )
+
+/**************************************************************************
+CURRTICKS - Get Time
+Use direct memory access to BIOS variables, longword 0040:006C (ticks
+today) and byte 0040:0070 (midnight crossover flag) instead of calling
+timeofday BIOS interrupt.
+**************************************************************************/
+#if defined(CONFIG_TSC_CURRTICKS)
+#undef CONFIG_BIOS_CURRTICKS
+#else
+#define CONFIG_BIOS_CURRTICKS 1
+#endif
+#if defined(CONFIG_BIOS_CURRTICKS)
+unsigned long currticks (void) 
+{
+	static uint32_t days = 0;
+	uint32_t *ticks = VIRTUAL(0x0040,0x006c);
+	uint8_t *midnight = VIRTUAL(0x0040,0x0070);
+
+	/* Re-enable interrupts so that the timer interrupt can occur
+	 */
+	RM_FRAGMENT(rm_currticks,
+		"sti\n\t"
+		"nop\n\t"
+		"nop\n\t"
+		"cli\n\t"
+	);	
+
+	real_call ( rm_currticks, NULL, NULL );
+
+	if ( *midnight ) {
+		*midnight = 0;
+		days += 0x1800b0;
+	}
+	return ( days + *ticks );
+}
+#endif	/* CONFIG_BIOS_CURRTICKS */
+
+/**************************************************************************
+INT15 - Call Interrupt 0x15
+**************************************************************************/
+int int15 ( int ax )
+{
+	struct {
+		reg16_t ax;
+	} PACKED in_stack;
+	struct {
+		reg16_t flags;
+	} PACKED out_stack;
+	reg16_t ret_ax;
+
+	RM_FRAGMENT(rm_int15,
+		"sti\n\t"
+		"popw %ax\n\t"
+		"stc\n\t"
+		"int $0x15\n\t"
+		"pushf\n\t"
+		"cli\n\t"
+	);
+
+	in_stack.ax.word = ax;
+	ret_ax.word = real_call ( rm_int15, &in_stack, &out_stack );
+
+	/* Carry flag clear indicates function not supported */
+	if ( ! ( out_stack.flags.word & CF ) ) return 0;
+	return ret_ax.h;
+}
+
+#ifdef	POWERSAVE
+/**************************************************************************
+CPU_NAP - Save power by halting the CPU until the next interrupt
+**************************************************************************/
+void cpu_nap ( void )
+{
+	RM_FRAGMENT(rm_cpu_nap,
+		"sti\n\t"
+		"hlt\n\t"
+		"cli\n\t"
+	);	
+
+	real_call ( rm_cpu_nap, NULL, NULL );
+}
+#endif	/* POWERSAVE */
+
+#if	(TRY_FLOPPY_FIRST > 0)
+/**************************************************************************
+DISK_INIT - Initialize the disk system
+**************************************************************************/
+void disk_init ( void )
+{
+	RM_FRAGMENT(rm_disk_init,
+		"sti\n\t"
+		"xorw %ax,%ax\n\t"
+		"movb $0x80,%dl\n\t"
+		"int $0x13\n\t"
+		"cli\n\t"
+	);
+	
+	real_call ( rm_disk_init, NULL, NULL );
+}
+
+/**************************************************************************
+DISK_READ - Read a sector from disk
+**************************************************************************/
+unsigned int pcbios_disk_read ( int drive, int cylinder, int head, int sector,
+				char *buf ) {
+	struct {
+		reg16_t ax;
+		reg16_t cx;
+		reg16_t dx;
+		segoff_t buffer;
+	} PACKED in_stack;
+	struct {
+		reg16_t flags;
+	} PACKED out_stack;
+	reg16_t ret_ax;
+
+	RM_FRAGMENT(rm_pcbios_disk_read,
+		"sti\n\t"
+		"popw %ax\n\t"
+		"popw %cx\n\t"
+		"popw %dx\n\t"
+		"popw %bx\n\t"
+		"popw %es\n\t"
+		"int $0x13\n\t"
+		"pushfw\n\t"
+		"cli\n\t"
+	);
+
+	in_stack.ax.h = 2; /* INT 13,2 - Read disk sector */
+	in_stack.ax.l = 1; /* Read one sector */
+	in_stack.cx.h = cylinder & 0xff;
+	in_stack.cx.l = ( ( cylinder >> 8 ) & 0x3 ) | sector;
+	in_stack.dx.h = head;
+	in_stack.dx.l = drive;
+	in_stack.buffer.segment = SEGMENT ( buf );
+	in_stack.buffer.offset = OFFSET ( buf );
+	ret_ax.word = real_call ( rm_pcbios_disk_read, &in_stack, &out_stack );
+	return ( out_stack.flags.word & CF ) ? ret_ax.word : 0;
+}
+#endif /* TRY_FLOPPY_FIRST */
+
+#endif /* PCBIOS */

src/arch/i386/firmware/pcbios/console.c

@@ -0,0 +1,85 @@
+/* Etherboot routines for PCBIOS firmware.
+ *
+ * Body of routines taken from old pcbios.S
+ */
+
+#ifdef PCBIOS
+
+#include "etherboot.h"
+#include "realmode.h"
+#include "segoff.h"
+
+#define ZF ( 1 << 6 )
+
+/**************************************************************************
+CONSOLE_PUTC - Print a character on console
+**************************************************************************/
+void console_putc ( int character )
+{
+	struct {
+		reg16_t ax;
+	} PACKED in_stack;
+
+	RM_FRAGMENT(rm_console_putc,
+		"sti\n\t"
+		"popw %ax\n\t"
+		"movb $0x0e, %ah\n\t"
+		"movl $1, %ebx\n\t"
+		"int $0x10\n\t"
+		"cli\n\t"
+	);
+
+	in_stack.ax.l = character;
+	real_call ( rm_console_putc, &in_stack, NULL );
+}
+
+/**************************************************************************
+CONSOLE_GETC - Get a character from console
+**************************************************************************/
+int console_getc ( void )
+{
+	RM_FRAGMENT(rm_console_getc,
+		"sti\n\t"
+		"xorw %ax, %ax\n\t"
+		"int $0x16\n\t"
+		"xorb %ah, %ah\n\t"
+		"cli\n\t"
+	);	
+
+	return real_call ( rm_console_getc, NULL, NULL );
+}
+
+/**************************************************************************
+CONSOLE_ISCHAR - Check for keyboard interrupt
+**************************************************************************/
+int console_ischar ( void )
+{
+	RM_FRAGMENT(rm_console_ischar,
+		"sti\n\t"
+		"movb $1, %ah\n\t"
+		"int $0x16\n\t"
+		"pushfw\n\t"
+		"popw %ax\n\t"
+		"cli\n\t"
+	);
+
+	return ( ( real_call ( rm_console_ischar, NULL, NULL ) & ZF ) == 0 );
+}
+
+/**************************************************************************
+GETSHIFT - Get keyboard shift state
+**************************************************************************/
+int getshift ( void )
+{
+	RM_FRAGMENT(rm_getshift,
+		"sti\n\t"
+		"movb $2, %ah\n\t"
+		"int $0x16\n\t"
+		"andw $0x3, %ax\n\t" 
+		"cli\n\t"
+	);
+
+	return real_call ( rm_getshift, NULL, NULL );
+}
+
+#endif /* PCBIOS */

src/arch/i386/firmware/pcbios/e820mangler.S

@@ -0,0 +1,296 @@
+#undef CODE16
+#if defined(PCBIOS)
+#define	CODE16
+#endif
+
+#ifdef CODE16
+	
+#define BOCHSBP xchgw %bx,%bx
+	
+	.text
+	.arch i386
+	.section ".text16", "ax", @progbits
+	.code16
+
+/****************************************************************************
+ * Memory map mangling code
+ ****************************************************************************
+ */
+
+	.globl	e820mangler
+e820mangler:
+
+/* Macro to calculate offset of labels within code segment in
+ * installed copy of code.
+ */
+#define INSTALLED(x) ( (x) - e820mangler )
+	
+/****************************************************************************
+ * Intercept INT 15 memory calls and remove the hidden memory ranges
+ * from the resulting memory map.
+ ****************************************************************************
+ */
+	.globl	_intercept_int15
+_intercept_int15:
+	/* Preserve registers */
+	pushw	%bp
+	/* Store %ax for future reference */
+	pushw	%ax
+	/* Make INT-style call to old INT15 routine */
+	pushfw
+	lcall	%cs:*INSTALLED(_intercepted_int15)
+	/* Preserve flags returned by original E820 routine */
+	pushfw
+	/* Check for valid INT15 routine */
+	jc	intercept_int15_exit
+	/* Check for a routine we want to intercept */
+	movw	%sp, %bp
+	cmpw	$0xe820, 2(%bp)
+	je	intercept_e820
+	cmpw	$0xe801, 2(%bp)
+	je	intercept_e801
+	cmpb	$0x88, 3(%bp)
+	je	intercept_88
+intercept_int15_exit:
+	/* Restore registers and return */
+	popfw
+	popw	%bp			/* discard original %ax */
+	popw	%bp
+	lret	$2			/* 'iret' - flags already loaded */
+
+	.globl	_intercepted_int15
+_intercepted_int15:	.word 0,0
+		
+/****************************************************************************
+ * Exclude an address range from a potentially overlapping address range
+ *
+ * Note: this *can* be called even if the range doesn't overlap; it
+ * will simply return the range unaltered.  It copes with all the
+ * possible cases of overlap, including total overlap (which will
+ * modify the range to length zero).  If the to-be-excluded range is
+ * in the middle of the target range, then the larger remaining
+ * portion will be returned.  If %di is nonzero on entry then the
+ * range will only be truncated from the high end, i.e. the base
+ * address will never be altered.  All this in less than 30
+ * instructions.  :)
+ *
+ * Parameters:
+ *  %eax	Base address of memory range
+ *  %ecx	Length of memory range
+ *  %ebx	Base address of memory range to exclude
+ *  %edx	Length of memory range to exclude
+ *  %di		0 => truncate either end, 1 => truncate high end only
+ * Returns:
+ *  %eax	Updated base address of range
+ *  %ecx	Updated length of range
+ *  %ebx,%edx	Undefined
+ *		All other registers (including %di) preserved
+ *
+ * Note: "ja" is used rather than "jg" because we are comparing
+ * unsigned ints
+ ****************************************************************************
+ */
+#ifdef TEST_EXCLUDE_ALGORITHM
+	.code32
+#endif /* TEST_EXCLUDE_ALGORITHM */
+exclude_memory_range:
+	/* Convert (start,length) to (start,end) */
+	addl	%eax, %ecx
+	addl	%ebx, %edx
+	/* Calculate "prefix" length */
+	subl	%eax, %ebx		/* %ebx = "prefix" length */
+	ja	1f
+	xorl	%ebx, %ebx		/* Truncate to zero if negative */
+1:	/* %di == 0 => truncate either end
+	 * %di != 0 => truncate only high end
+	 */
+	testw	%di, %di
+	je	use_either
+	cmpl	%eax, %edx
+	jbe	99f			/* excl. range is below target range */
+use_prefix:	/* Use prefix, discard suffix */
+	addl	%eax, %ebx		/* %ebx = candidate end address */
+	cmpl	%ecx, %ebx		/* %ecx = min ( %ebx, %ecx ) */
+	ja	1f
+	movl	%ebx, %ecx
+1:	jmp	99f
+use_either:		
+	/* Calculate "suffix" length */
+	subl	%ecx, %edx		/* %edx = -( "suffix" length ) */
+	jb	1f
+	xorl	%edx, %edx		/* Truncate to zero if negative */
+1:	negl	%edx			/* %edx = "suffix" length */
+	/* Use whichever is longest of "prefix" and "suffix" */
+	cmpl	%ebx, %edx
+	jbe	use_prefix
+use_suffix:	/* Use suffix, discard prefix */
+	negl	%edx
+	addl	%ecx, %edx		/* %edx = candidate start address */
+	cmpl	%eax, %edx		/* %eax = max ( %eax, %edx ) */
+	jb	1f
+	movl	%edx, %eax
+1:	
+99:	subl	%eax, %ecx		/* Convert back to (start,length) */
+	ret
+
+#ifdef TEST_EXCLUDE_ALGORITHM
+	.globl	__test_exclude
+__test_exclude:
+	pushl   %ebx
+	pushl	%edi
+	movl    12(%esp), %eax
+	movl    16(%esp), %ecx
+	movl    20(%esp), %ebx
+	movl    24(%esp), %edx
+	movl	28(%esp), %edi
+	call    exclude_memory_range
+	shll    $16, %eax
+	orl     %ecx, %eax
+	popl	%edi
+	popl	%ebx
+	ret
+	.code16
+#endif /* TEST_EXCLUDE_ALGORITHM */
+	
+/****************************************************************************
+ * Exclude Etherboot-reserved address ranges from a potentially
+ * overlapping address range
+ *
+ * Parameters:
+ *  %eax	Base address of memory range
+ *  %ecx	Length of memory range
+ *  %di		0 => truncate either end, 1 => truncate high end only
+ * Returns:
+ *  %eax	Updated base address of range
+ *  %ecx	Updated length of range
+ *		All other registers (including %di) preserved
+ ****************************************************************************
+ */
+exclude_hidden_memory_ranges:
+	pushw	%si
+	pushl	%ebx
+	pushl	%edx
+	movw	$INSTALLED(_hide_memory), %si
+2:	movl	%cs:0(%si), %ebx
+	movl	%cs:4(%si), %edx
+	call	exclude_memory_range
+	addw	$8, %si
+	cmpw	$INSTALLED(_hide_memory_end), %si
+	jl	2b
+	popl	%edx
+	popl	%ebx
+	popw	%si
+	ret
+	
+	.globl	_hide_memory	
+_hide_memory:
+	.long	0,0			/* Etherboot text (base,length) */
+	.long	0,0			/* Heap (base,length) */
+_hide_memory_end:
+
+/****************************************************************************
+ * Intercept INT 15,E820 calls and remove the hidden memory ranges
+ * from the resulting memory map.
+ ****************************************************************************
+ */
+#define SMAP ( 0x534d4150 )
+intercept_e820:
+	/* Check for valid E820 routine */
+	cmpl	$SMAP, %eax
+	jne	intercept_int15_exit
+	/* If base address isn't in the low 4GB, return unaltered
+	 * (since we never claim memory above 4GB).  WARNING: we cheat
+	 * by assuming that no E820 region will straddle the 4GB
+	 * boundary: if this is not a valid assumption then things
+	 * will probably break.
+	 */
+	cmpl	$0, %es:4(%di)
+	jne	intercept_int15_exit
+	/* Preserve registers */
+	pushl	%eax
+	pushl	%ecx
+	/* Update returned memory range */
+	movl	%es:0(%di), %eax	/* Base */
+	movl	%es:8(%di), %ecx	/* Length */
+	pushw	%di
+	xorw	%di, %di		/* "truncate either end" flag */
+	call	exclude_hidden_memory_ranges
+	popw	%di
+	movl	%eax, %es:0(%di)	/* Store updated base */
+	movl	%ecx, %es:8(%di)	/* Store updated length */
+	/* Restore registers and return */
+	popl	%ecx
+	popl	%eax
+	jmp	intercept_int15_exit
+
+/****************************************************************************
+ * Intercept INT 15,E801 calls and remove the hidden memory ranges
+ * from the resulting memory map.
+ ****************************************************************************
+ */
+intercept_e801:
+	/* Adjust return values */
+	call	e801_adjust
+	xchgw	%ax, %cx
+	xchgw	%bx, %dx
+	call	e801_adjust
+	xchgw	%ax, %cx
+	xchgw	%bx, %dx
+	jmp	intercept_int15_exit
+	
+	/* %ax = #KB from 1MB+, %bx = #64KB from 16MB+
+	 * Return with modified values in %ax, %bx.  Preserver other regs.
+	 */
+e801_adjust:
+	pushw	%di
+	pushl	%ecx
+	pushl	%eax
+	movw	$1, %di			/* "truncate only high end" flag */
+
+	/* Truncate #64KB from 16MB+ as appropriate */
+	movw	%bx, %cx		/* (no need to zero high word) */
+	shll	$16, %ecx		/* %ecx = length in bytes */
+	movl	$(1<<24), %eax		/* 16MB start address */
+	call	exclude_hidden_memory_ranges
+	shrl	$16, %ecx		/* %cx = updated length in 64KB */
+	movw	%cx, %bx		/* Return in %bx */
+	
+	/* Truncate #KB from 1MB+ as appropriate */
+	popw	%cx			/* Orig. %ax (high word already 0) */
+	shll	$10, %ecx		/* %ecx = length in bytes */
+	shrl	$4, %eax		/* 1MB start address */
+	call	exclude_hidden_memory_ranges
+	shrl	$10, %ecx		/* %cx = updated length in KB */
+	pushw	%cx			/* Will be picked up in %eax */
+	
+	popl	%eax
+	popl	%ecx
+	popw	%di
+	ret
+
+/****************************************************************************
+ * Intercept INT 15,88 calls and remove the hidden memory ranges
+ * from the resulting memory map.
+ ****************************************************************************
+ */
+intercept_88:
+	pushw	%bx			/* E801 adjust, ignore %bx */
+	call	e801_adjust
+	popw	%bx
+	jmp	intercept_int15_exit
+
+	.globl	e820mangler_end
+e820mangler_end:
+
+	.globl	_e820mangler_size
+	.equ	_e820mangler_size, e820mangler_end - e820mangler
+	.globl	e820mangler_size
+e820mangler_size:
+	.word	_e820mangler_size
+
+#else
+
+	.globl	_e820mangler_size
+	.equ	_e820mangler_size, 0
+	
+#endif /* CODE16 */

src/arch/i386/firmware/pcbios/hidemem.c

@@ -0,0 +1,94 @@
+/* Utility functions to hide Etherboot by manipulating the E820 memory
+ * map.  These could go in memsizes.c, but are placed here because not
+ * all images will need them.
+ */
+
+#include "etherboot.h"
+#include "hidemem.h"
+
+#ifdef	CODE16
+
+static int mangling = 0;
+static void *mangler = NULL;
+
+#define INSTALLED(x)	( (typeof(&x)) ( (void*)(&x) - (void*)e820mangler \
+					 + mangler ) )
+#define intercept_int15		INSTALLED(_intercept_int15)
+#define intercepted_int15	INSTALLED(_intercepted_int15)
+#define hide_memory		INSTALLED(_hide_memory)
+#define INT15_VECTOR ( (segoff_t*) ( phys_to_virt( 4 * 0x15 ) ) )
+
+int install_e820mangler ( void *new_mangler ) {
+	if ( mangling ) return 0;
+	memcpy ( new_mangler, &e820mangler, e820mangler_size );
+	mangler = new_mangler;
+	return 1;
+}
+
+/* Intercept INT15 calls and pass them through the mangler.  The
+ * mangler must have been copied to base memory before making this
+ * call, and "mangler" must point to the base memory copy, which must
+ * be 16-byte aligned.
+ */
+int hide_etherboot ( void ) {
+	if ( mangling ) return 1;
+	if ( !mangler ) return 0;
+
+	/* Hook INT15 handler */
+	*intercepted_int15 = *INT15_VECTOR;
+	(*hide_memory)[0].start = virt_to_phys(_text);
+	(*hide_memory)[0].length = _end - _text;
+	/* IMPORTANT, possibly even FIXME:
+	 *
+	 * Etherboot has a tendency to claim a very large area of
+	 * memory as possible heap; enough to make it impossible to
+	 * load an OS if we hide all of it.  We hide only the portion
+	 * that's currently in use.  This means that we MUST NOT
+	 * perform further allocations from the heap while the mangler
+	 * is active.
+	 */
+	(*hide_memory)[1].start = heap_ptr;
+	(*hide_memory)[1].length = heap_bot - heap_ptr;
+	INT15_VECTOR->segment = SEGMENT(mangler);
+	INT15_VECTOR->offset = 0;
+
+	mangling = 1;
+	return 1;
+}
+
+int unhide_etherboot ( void ) {
+	if ( !mangling ) return 1;
+
+	/* Restore original INT15 handler
+	 */
+	if ( VIRTUAL(INT15_VECTOR->segment,INT15_VECTOR->offset) != mangler ) {
+		/* Oh dear... */
+
+#ifdef WORK_AROUND_BPBATCH_BUG
+		/* BpBatch intercepts INT15, so can't unhook it, and
+		 * then proceeds to ignore our PXENV_KEEP_UNDI return
+		 * status, which means that it ends up zeroing out the
+		 * INT15 handler routine.
+		 *
+		 * This rather ugly hack involves poking into
+		 * BpBatch's code and changing it's stored value for
+		 * the "next handler" in the INT15 chain.
+		 */
+		segoff_t *bp_chain = VIRTUAL ( 0x0060, 0x8254 );
+
+		if ( ( bp_chain->segment == SEGMENT(mangler) ) &&
+		     ( bp_chain->offset == 0 ) ) {
+			printf ( "\nBPBATCH bug workaround enabled\n" );
+			*bp_chain = *intercepted_int15;
+		}
+#endif /* WORK_AROUND_BPBATCH_BUG */
+
+		return 0;
+	}
+	*INT15_VECTOR = *intercepted_int15;
+
+	mangling = 0;
+	return 1;
+}
+
+#endif	/* CODE16 */

src/arch/i386/firmware/pcbios/memsizes.c

@@ -0,0 +1,201 @@
+#ifdef PCBIOS
+
+#include "etherboot.h"
+#include "realmode.h"
+
+#define CF ( 1 << 0 )
+
+#ifndef MEMSIZES_DEBUG 
+#define MEMSIZES_DEBUG 0
+#endif
+
+/* by Eric Biederman */
+
+struct meminfo meminfo;
+
+/**************************************************************************
+BASEMEMSIZE - Get size of the conventional (base) memory
+**************************************************************************/
+unsigned short basememsize ( void )
+{
+	RM_FRAGMENT(rm_basememsize,
+		"int $0x12\n\t"
+	);	
+	return real_call ( rm_basememsize, NULL, NULL );
+}
+
+/**************************************************************************
+MEMSIZE - Determine size of extended memory
+**************************************************************************/
+unsigned int memsize ( void )
+{
+	struct {
+		reg16_t ax;
+	} PACKED in_stack;
+	struct {
+		reg16_t ax;
+		reg16_t bx;
+		reg16_t cx;
+		reg16_t dx;
+		reg16_t flags;
+	} PACKED out_stack;
+	int memsize;
+
+	RM_FRAGMENT(rm_memsize,
+	/* Some buggy BIOSes don't clear/set carry on pass/error of
+	 * e801h memory size call or merely pass cx,dx through without
+	 * changing them, so we set carry and zero cx,dx before call.
+	 */
+		"stc\n\t"
+		"xorw %cx,%cx\n\t"
+		"xorw %dx,%dx\n\t"
+		"popw %ax\n\t"
+		"int $0x15\n\t"
+		"pushfw\n\t"
+		"pushw %dx\n\t"
+		"pushw %cx\n\t"
+		"pushw %bx\n\t"
+		"pushw %ax\n\t"
+	);
+
+	/* Try INT 15,e801 first */
+	in_stack.ax.word = 0xe801;
+	real_call ( rm_memsize, &in_stack, &out_stack );
+	if ( out_stack.flags.word & CF ) {
+		/* INT 15,e801 not supported: try INT 15,88 */
+		in_stack.ax.word = 0x8800;
+		memsize = real_call ( rm_memsize, &in_stack, &out_stack );
+	} else {
+		/* Some BIOSes report extended memory via ax,bx rather
+		 * than cx,dx
+		 */
+		if ( (out_stack.cx.word==0) && (out_stack.dx.word==0) ) {
+			/* Use ax,bx */
+			memsize = ( out_stack.bx.word<<6 ) + out_stack.ax.word;
+		} else {
+			/* Use cx,dx */
+			memsize = ( out_stack.dx.word<<6 ) + out_stack.cx.word;
+		}
+	}
+	return memsize;
+}
+
+#define SMAP ( 0x534d4150 )
+int meme820 ( struct e820entry *buf, int count )
+{
+	struct {
+		reg16_t flags;
+		reg32_t eax;
+		reg32_t ebx;
+		struct e820entry entry;
+	} PACKED stack;
+	int index = 0;
+
+	RM_FRAGMENT(rm_meme820,
+		"addw $6, %sp\n\t"	/* skip flags, eax */
+		"popl %ebx\n\t"
+		"pushw %ss\n\t"	/* es:di = ss:sp */
+		"popw %es\n\t"
+		"movw %sp, %di\n\t"
+		"movl $0xe820, %eax\n\t"
+		"movl $" RM_STR(SMAP) ", %edx\n\t"
+		"movl $" RM_STR(E820ENTRY_SIZE) ", %ecx\n\t"
+		"int $0x15\n\t"
+		"pushl %ebx\n\t"
+		"pushl %eax\n\t"
+		"pushfw\n\t"
+	);
+
+	stack.ebx.dword = 0; /* 'EOF' marker */
+	while ( ( index < count ) &&
+		( ( index == 0 ) || ( stack.ebx.dword != 0 ) ) ) {
+		real_call ( rm_meme820, &stack, &stack );
+		if ( stack.eax.dword != SMAP ) return 0;
+		if ( stack.flags.word & CF ) return 0;
+		buf[index++] = stack.entry;
+	}
+	return index;
+}
+
+void get_memsizes(void)
+{
+	/* Ensure we don't stomp bios data structutres.
+	 * the interrupt table: 0x000 - 0x3ff
+	 * the bios data area:  0x400 - 0x502
+	 * Dos variables:       0x502 - 0x5ff
+	 */
+	static const unsigned min_addr = 0x600;
+	unsigned i;
+	unsigned basemem;
+	basemem = get_free_base_memory();
+	meminfo.basememsize = basememsize();
+	meminfo.memsize = memsize();
+#ifndef IGNORE_E820_MAP
+	meminfo.map_count = meme820(meminfo.map, E820MAX);
+#else
+	meminfo.map_count = 0;
+#endif
+	if (meminfo.map_count == 0) {
+		/* If we don't have an e820 memory map fake it */
+		meminfo.map_count = 2;
+		meminfo.map[0].addr = 0;
+		meminfo.map[0].size = meminfo.basememsize << 10;
+		meminfo.map[0].type = E820_RAM;
+		meminfo.map[1].addr = 1024*1024;
+		meminfo.map[1].size = meminfo.memsize << 10;
+		meminfo.map[1].type = E820_RAM;
+	}
+	/* Scrub the e820 map */
+	for(i = 0; i < meminfo.map_count; i++) {
+		if (meminfo.map[i].type != E820_RAM) {
+			continue;
+		}
+		/* Reserve the bios data structures */
+		if (meminfo.map[i].addr < min_addr) {
+			unsigned long delta;
+			delta = min_addr - meminfo.map[i].addr;
+			if (delta > meminfo.map[i].size) {
+				delta = meminfo.map[i].size;
+			}
+			meminfo.map[i].addr = min_addr;
+			meminfo.map[i].size -= delta;
+		}
+		/* Ensure the returned e820 map is in sync 
+		 * with the actual memory state 
+		 */
+		if ((meminfo.map[i].addr < 0xa0000) && 
+			((meminfo.map[i].addr + meminfo.map[i].size) > basemem))
+		{
+			if (meminfo.map[i].addr <= basemem) {
+				meminfo.map[i].size = basemem - meminfo.map[i].addr;
+			} else {
+				meminfo.map[i].addr = basemem;
+				meminfo.map[i].size = 0;
+			}
+		}
+	}
+#if MEMSIZES_DEBUG
+{
+	int i;
+	printf("basememsize %d\n", meminfo.basememsize);
+	printf("memsize %d\n",     meminfo.memsize);
+	printf("Memory regions(%d):\n", meminfo.map_count);
+	for(i = 0; i < meminfo.map_count; i++) {
+		unsigned long long r_start, r_end;
+		r_start = meminfo.map[i].addr;
+		r_end = r_start + meminfo.map[i].size;
+		printf("[%X%X, %X%X) type %d\n", 
+			(unsigned long)(r_start >> 32),
+			(unsigned long)r_start,
+			(unsigned long)(r_end >> 32),
+			(unsigned long)r_end,
+			meminfo.map[i].type);
+#if defined(CONSOLE_FIRMWARE)
+		sleep(1); /* No way to see 32 entries on a standard 80x25 screen... */
+#endif
+	}
+}
+#endif
+}
+
+#endif /* PCBIOS */

src/arch/i386/include/bits/byteswap.h

@@ -0,0 +1,45 @@
+#ifndef ETHERBOOT_BITS_BYTESWAP_H
+#define ETHERBOOT_BITS_BYTESWAP_H
+
+static inline uint16_t __i386_bswap_16(uint16_t x)
+{
+	__asm__("xchgb %b0,%h0\n\t"
+		: "=q" (x)
+		: "0" (x));
+	return x;
+}
+
+static inline uint32_t __i386_bswap_32(uint32_t x)
+{
+	__asm__("xchgb %b0,%h0\n\t"
+		"rorl $16,%0\n\t"
+		"xchgb %b0,%h0"
+		: "=q" (x)
+		: "0" (x));
+	return x;
+}
+
+
+#define __bswap_constant_16(x) \
+	((uint16_t)((((uint16_t)(x) & 0x00ff) << 8) | \
+		(((uint16_t)(x) & 0xff00) >> 8)))
+
+#define __bswap_constant_32(x) \
+	((uint32_t)((((uint32_t)(x) & 0x000000ffU) << 24) | \
+		(((uint32_t)(x) & 0x0000ff00U) <<  8) | \
+		(((uint32_t)(x) & 0x00ff0000U) >>  8) | \
+		(((uint32_t)(x) & 0xff000000U) >> 24)))
+
+#define __bswap_16(x) \
+	((uint16_t)(__builtin_constant_p(x) ? \
+	__bswap_constant_16(x) : \
+	__i386_bswap_16(x)))
+
+
+#define __bswap_32(x) \
+	((uint32_t)(__builtin_constant_p(x) ? \
+	__bswap_constant_32(x) : \
+	__i386_bswap_32(x)))
+
+
+#endif /* ETHERBOOT_BITS_BYTESWAP_H */

src/arch/i386/include/bits/cpu.h

@@ -0,0 +1,243 @@
+#ifndef I386_BITS_CPU_H
+#define I386_BITS_CPU_H
+
+
+/* Sample usage: CPU_FEATURE_P(cpu.x86_capability, FPU) */
+#define CPU_FEATURE_P(CAP, FEATURE) \
+	(!!(CAP[(X86_FEATURE_##FEATURE)/32] & ((X86_FEATURE_##FEATURE) & 0x1f)))
+
+#define NCAPINTS	4	/* Currently we have 4 32-bit words worth of info */
+
+/* Intel-defined CPU features, CPUID level 0x00000001, word 0 */
+#define X86_FEATURE_FPU		(0*32+ 0) /* Onboard FPU */
+#define X86_FEATURE_VME		(0*32+ 1) /* Virtual Mode Extensions */
+#define X86_FEATURE_DE		(0*32+ 2) /* Debugging Extensions */
+#define X86_FEATURE_PSE 	(0*32+ 3) /* Page Size Extensions */
+#define X86_FEATURE_TSC		(0*32+ 4) /* Time Stamp Counter */
+#define X86_FEATURE_MSR		(0*32+ 5) /* Model-Specific Registers, RDMSR, WRMSR */
+#define X86_FEATURE_PAE		(0*32+ 6) /* Physical Address Extensions */
+#define X86_FEATURE_MCE		(0*32+ 7) /* Machine Check Architecture */
+#define X86_FEATURE_CX8		(0*32+ 8) /* CMPXCHG8 instruction */
+#define X86_FEATURE_APIC	(0*32+ 9) /* Onboard APIC */
+#define X86_FEATURE_SEP		(0*32+11) /* SYSENTER/SYSEXIT */
+#define X86_FEATURE_MTRR	(0*32+12) /* Memory Type Range Registers */
+#define X86_FEATURE_PGE		(0*32+13) /* Page Global Enable */
+#define X86_FEATURE_MCA		(0*32+14) /* Machine Check Architecture */
+#define X86_FEATURE_CMOV	(0*32+15) /* CMOV instruction (FCMOVCC and FCOMI too if FPU present) */
+#define X86_FEATURE_PAT		(0*32+16) /* Page Attribute Table */
+#define X86_FEATURE_PSE36	(0*32+17) /* 36-bit PSEs */
+#define X86_FEATURE_PN		(0*32+18) /* Processor serial number */
+#define X86_FEATURE_CLFLSH	(0*32+19) /* Supports the CLFLUSH instruction */
+#define X86_FEATURE_DTES	(0*32+21) /* Debug Trace Store */
+#define X86_FEATURE_ACPI	(0*32+22) /* ACPI via MSR */
+#define X86_FEATURE_MMX		(0*32+23) /* Multimedia Extensions */
+#define X86_FEATURE_FXSR	(0*32+24) /* FXSAVE and FXRSTOR instructions (fast save and restore */
+				          /* of FPU context), and CR4.OSFXSR available */
+#define X86_FEATURE_XMM		(0*32+25) /* Streaming SIMD Extensions */
+#define X86_FEATURE_XMM2	(0*32+26) /* Streaming SIMD Extensions-2 */
+#define X86_FEATURE_SELFSNOOP	(0*32+27) /* CPU self snoop */
+#define X86_FEATURE_HT		(0*32+28) /* Hyper-Threading */
+#define X86_FEATURE_ACC		(0*32+29) /* Automatic clock control */
+#define X86_FEATURE_IA64	(0*32+30) /* IA-64 processor */
+
+/* AMD-defined CPU features, CPUID level 0x80000001, word 1 */
+/* Don't duplicate feature flags which are redundant with Intel! */
+#define X86_FEATURE_SYSCALL	(1*32+11) /* SYSCALL/SYSRET */
+#define X86_FEATURE_MMXEXT	(1*32+22) /* AMD MMX extensions */
+#define X86_FEATURE_LM		(1*32+29) /* Long Mode (x86-64) */
+#define X86_FEATURE_3DNOWEXT	(1*32+30) /* AMD 3DNow! extensions */
+#define X86_FEATURE_3DNOW	(1*32+31) /* 3DNow! */
+
+/* Transmeta-defined CPU features, CPUID level 0x80860001, word 2 */
+#define X86_FEATURE_RECOVERY	(2*32+ 0) /* CPU in recovery mode */
+#define X86_FEATURE_LONGRUN	(2*32+ 1) /* Longrun power control */
+#define X86_FEATURE_LRTI	(2*32+ 3) /* LongRun table interface */
+
+/* Other features, Linux-defined mapping, word 3 */
+/* This range is used for feature bits which conflict or are synthesized */
+#define X86_FEATURE_CXMMX	(3*32+ 0) /* Cyrix MMX extensions */
+#define X86_FEATURE_K6_MTRR	(3*32+ 1) /* AMD K6 nonstandard MTRRs */
+#define X86_FEATURE_CYRIX_ARR	(3*32+ 2) /* Cyrix ARRs (= MTRRs) */
+#define X86_FEATURE_CENTAUR_MCR	(3*32+ 3) /* Centaur MCRs (= MTRRs) */
+
+#define MAX_X86_VENDOR_ID 16
+struct cpuinfo_x86 {
+	uint8_t	 x86;		/* CPU family */
+	uint8_t	 x86_model;
+	uint8_t	 x86_mask;
+
+       	int	 cpuid_level;	/* Maximum supported CPUID level, -1=no CPUID */
+	unsigned x86_capability[NCAPINTS];
+	char	 x86_vendor_id[MAX_X86_VENDOR_ID];
+};
+
+
+#define X86_VENDOR_INTEL 0
+#define X86_VENDOR_CYRIX 1
+#define X86_VENDOR_AMD 2
+#define X86_VENDOR_UMC 3
+#define X86_VENDOR_NEXGEN 4
+#define X86_VENDOR_CENTAUR 5
+#define X86_VENDOR_RISE 6
+#define X86_VENDOR_TRANSMETA 7
+#define X86_VENDOR_NSC 8
+#define X86_VENDOR_UNKNOWN 0xff
+
+/*
+ * EFLAGS bits
+ */
+#define X86_EFLAGS_CF	0x00000001 /* Carry Flag */
+#define X86_EFLAGS_PF	0x00000004 /* Parity Flag */
+#define X86_EFLAGS_AF	0x00000010 /* Auxillary carry Flag */
+#define X86_EFLAGS_ZF	0x00000040 /* Zero Flag */
+#define X86_EFLAGS_SF	0x00000080 /* Sign Flag */
+#define X86_EFLAGS_TF	0x00000100 /* Trap Flag */
+#define X86_EFLAGS_IF	0x00000200 /* Interrupt Flag */
+#define X86_EFLAGS_DF	0x00000400 /* Direction Flag */
+#define X86_EFLAGS_OF	0x00000800 /* Overflow Flag */
+#define X86_EFLAGS_IOPL	0x00003000 /* IOPL mask */
+#define X86_EFLAGS_NT	0x00004000 /* Nested Task */
+#define X86_EFLAGS_RF	0x00010000 /* Resume Flag */
+#define X86_EFLAGS_VM	0x00020000 /* Virtual Mode */
+#define X86_EFLAGS_AC	0x00040000 /* Alignment Check */
+#define X86_EFLAGS_VIF	0x00080000 /* Virtual Interrupt Flag */
+#define X86_EFLAGS_VIP	0x00100000 /* Virtual Interrupt Pending */
+#define X86_EFLAGS_ID	0x00200000 /* CPUID detection flag */
+
+/*
+ * Generic CPUID function
+ */
+static inline void cpuid(int op, 
+	unsigned int *eax, unsigned int *ebx, unsigned int *ecx, unsigned int *edx)
+{
+	__asm__("cpuid"
+		: "=a" (*eax),
+		  "=b" (*ebx),
+		  "=c" (*ecx),
+		  "=d" (*edx)
+		: "0" (op));
+}
+
+/*
+ * CPUID functions returning a single datum
+ */
+static inline unsigned int cpuid_eax(unsigned int op)
+{
+	unsigned int eax;
+
+	__asm__("cpuid"
+		: "=a" (eax)
+		: "0" (op)
+		: "bx", "cx", "dx");
+	return eax;
+}
+static inline unsigned int cpuid_ebx(unsigned int op)
+{
+	unsigned int eax, ebx;
+
+	__asm__("cpuid"
+		: "=a" (eax), "=b" (ebx)
+		: "0" (op)
+		: "cx", "dx" );
+	return ebx;
+}
+static inline unsigned int cpuid_ecx(unsigned int op)
+{
+	unsigned int eax, ecx;
+
+	__asm__("cpuid"
+		: "=a" (eax), "=c" (ecx)
+		: "0" (op)
+		: "bx", "dx" );
+	return ecx;
+}
+static inline unsigned int cpuid_edx(unsigned int op)
+{
+	unsigned int eax, edx;
+
+	__asm__("cpuid"
+		: "=a" (eax), "=d" (edx)
+		: "0" (op)
+		: "bx", "cx");
+	return edx;
+}
+
+/*
+ * Intel CPU features in CR4
+ */
+#define X86_CR4_VME		0x0001	/* enable vm86 extensions */
+#define X86_CR4_PVI		0x0002	/* virtual interrupts flag enable */
+#define X86_CR4_TSD		0x0004	/* disable time stamp at ipl 3 */
+#define X86_CR4_DE		0x0008	/* enable debugging extensions */
+#define X86_CR4_PSE		0x0010	/* enable page size extensions */
+#define X86_CR4_PAE		0x0020	/* enable physical address extensions */
+#define X86_CR4_MCE		0x0040	/* Machine check enable */
+#define X86_CR4_PGE		0x0080	/* enable global pages */
+#define X86_CR4_PCE		0x0100	/* enable performance counters at ipl 3 */
+#define X86_CR4_OSFXSR		0x0200	/* enable fast FPU save and restore */
+#define X86_CR4_OSXMMEXCPT	0x0400	/* enable unmasked SSE exceptions */
+
+
+#define MSR_K6_EFER			0xC0000080
+/* EFER bits: */ 
+#define _EFER_SCE 0  /* SYSCALL/SYSRET */
+#define _EFER_LME 8  /* Long mode enable */
+#define _EFER_LMA 10 /* Long mode active (read-only) */
+#define _EFER_NX 11  /* No execute enable */
+
+#define EFER_SCE (1<<_EFER_SCE)
+#define EFER_LME (1<<EFER_LME)
+#define EFER_LMA (1<<EFER_LMA)
+#define EFER_NX (1<<_EFER_NX)
+
+#define rdmsr(msr,val1,val2) \
+     __asm__ __volatile__("rdmsr" \
+			  : "=a" (val1), "=d" (val2) \
+			  : "c" (msr))
+
+#define wrmsr(msr,val1,val2) \
+     __asm__ __volatile__("wrmsr" \
+			  : /* no outputs */ \
+			  : "c" (msr), "a" (val1), "d" (val2))
+
+
+#define read_cr0()	({ \
+	unsigned int __dummy; \
+	__asm__( \
+		"movl %%cr0, %0\n\t" \
+		:"=r" (__dummy)); \
+	__dummy; \
+})
+#define write_cr0(x) \
+	__asm__("movl %0,%%cr0": :"r" (x));
+
+#define read_cr3()	({ \
+	unsigned int __dummy; \
+	__asm__( \
+		"movl %%cr3, %0\n\t" \
+		:"=r" (__dummy)); \
+	__dummy; \
+})
+#define write_cr3x(x) \
+	__asm__("movl %0,%%cr3": :"r" (x));
+
+
+#define read_cr4()	({ \
+	unsigned int __dummy; \
+	__asm__( \
+		"movl %%cr4, %0\n\t" \
+		:"=r" (__dummy)); \
+	__dummy; \
+})
+#define write_cr4x(x) \
+	__asm__("movl %0,%%cr4": :"r" (x));
+
+
+extern struct cpuinfo_x86 cpu_info;
+#ifdef CONFIG_X86_64
+extern void cpu_setup(void);
+#else
+#define cpu_setup() do {} while(0)
+#endif
+
+#endif /* I386_BITS_CPU_H */