Merge remote-tracking branch 'upstream/master'

.travis.yml

@@ -0,0 +1,54 @@
+dist: trusty
+
+sudo: false
+
+language: c
+
+cache: ccache
+
+compiler:
+  - gcc
+
+addons:
+  apt:
+    packages:
+      - binutils-dev
+      - liblzma-dev
+      - syslinux
+      - genisoimage
+  coverity_scan:
+    project:
+      name: "ipxe/ipxe"
+      version: $TRAVIS_COMMIT
+    build_command_prepend: "make -C src bin/deps"
+    build_command: "make -C src bin/blib.a"
+    branch_pattern: coverity_scan
+
+env:
+  global:
+    - MAKEFLAGS="-j 4"
+
+script:
+  - make -C src bin/blib.a
+  - make -C src bin/ipxe.pxe
+  - make -C src bin/ipxe.usb
+  - make -C src bin/ipxe.iso
+  - make -C src bin/8086100e.mrom
+  - make -C src bin-x86_64-pcbios/blib.a
+  - make -C src bin-x86_64-pcbios/ipxe.pxe
+  - make -C src bin-x86_64-pcbios/ipxe.usb
+  - make -C src bin-x86_64-pcbios/ipxe.iso
+  - make -C src bin-x86_64-pcbios/8086100e.mrom
+  - make -C src bin-x86_64-efi/blib.a
+  - make -C src bin-x86_64-efi/ipxe.efi
+  - make -C src bin-x86_64-efi/intel.efidrv
+  - make -C src bin-x86_64-efi/intel.efirom
+  - make -C src bin-i386-efi/blib.a
+  - make -C src bin-i386-efi/ipxe.efi
+  - make -C src bin-i386-efi/intel.efidrv
+  - make -C src bin-i386-efi/intel.efirom
+  - make -C src bin-x86_64-linux/blib.a
+  - make -C src bin-x86_64-linux/tap.linux
+  - make -C src bin-x86_64-linux/af_packet.linux
+  - make -C src bin-x86_64-linux/tests.linux
+  - ./src/bin-x86_64-linux/tests.linux

contrib/coverity/model.c

@@ -0,0 +1,21 @@
+/*
+ * Coverity modelling file
+ *
+ */
+
+typedef long off_t;
+typedef void * userptr_t;
+typedef long long time_t;
+struct tm;
+
+/* Inhibit use of built-in models for functions where Coverity's
+ * assumptions about the modelled function are incorrect for iPXE.
+ */
+char * strerror ( int errno ) {
+}
+void copy_from_user ( void *dest, userptr_t src, off_t src_off, size_t len ) {
+}
+time_t mktime ( struct tm *tm ) {
+}
+int getchar ( void ) {
+}

src/Makefile

@@ -9,6 +9,7 @@ ASFLAGS		:=
 LDFLAGS		:=
 HOST_CFLAGS	:=
 MAKEDEPS	:= Makefile
+CROSS_COMPILE	?= $(CROSS)
 
 ###############################################################################
 #
@@ -52,9 +53,7 @@ EINFO		:= ./util/einfo
 GENKEYMAP	:= ./util/genkeymap.pl
 DOXYGEN		:= doxygen
 LCAB		:= lcab
-BINUTILS_DIR	:= /usr
-BFD_DIR		:= $(BINUTILS_DIR)
-ZLIB_DIR	:= /usr
+QEMUIMG		:= qemu-img
 
 ###############################################################################
 #
@@ -63,7 +62,7 @@ ZLIB_DIR	:= /usr
 SRCDIRS		:=
 SRCDIRS		+= libgcc
 SRCDIRS		+= core
-SRCDIRS		+= net net/oncrpc net/tcp net/udp net/infiniband net/80211
+SRCDIRS		+= net net/tcp net/udp net/infiniband
 SRCDIRS		+= image
 SRCDIRS		+= drivers/bus
 SRCDIRS		+= drivers/net
@@ -72,17 +71,23 @@ SRCDIRS		+= drivers/net/e1000e
 SRCDIRS		+= drivers/net/igb
 SRCDIRS		+= drivers/net/igbvf
 SRCDIRS		+= drivers/net/phantom
-SRCDIRS		+= drivers/net/rtl818x
-SRCDIRS		+= drivers/net/ath
-SRCDIRS		+= drivers/net/ath/ath5k
-SRCDIRS		+= drivers/net/ath/ath9k
 SRCDIRS		+= drivers/net/vxge
 SRCDIRS		+= drivers/net/efi
 SRCDIRS		+= drivers/net/tg3
+SRCDIRS		+= drivers/net/sfc
 SRCDIRS		+= drivers/block
 SRCDIRS		+= drivers/nvs
 SRCDIRS		+= drivers/bitbash
 SRCDIRS		+= drivers/infiniband
+SRCDIRS		+= drivers/infiniband/mlx_utils_flexboot/src
+SRCDIRS		+= drivers/infiniband/mlx_utils/src/public
+SRCDIRS		+= drivers/infiniband/mlx_utils/mlx_lib/mlx_reg_access
+SRCDIRS		+= drivers/infiniband/mlx_utils/mlx_lib/mlx_nvconfig
+SRCDIRS		+= drivers/infiniband/mlx_utils/mlx_lib/mlx_vmac
+SRCDIRS		+= drivers/infiniband/mlx_utils/mlx_lib/mlx_blink_leds
+SRCDIRS		+= drivers/infiniband/mlx_utils/mlx_lib/mlx_link_speed
+SRCDIRS		+= drivers/infiniband/mlx_utils/mlx_lib/mlx_mtu
+SRCDIRS		+= drivers/infiniband/mlx_nodnic/src
 SRCDIRS		+= drivers/usb
 SRCDIRS		+= interface/pxe interface/efi interface/smbios
 SRCDIRS		+= interface/bofm
@@ -96,6 +101,16 @@ SRCDIRS		+= hci/keymap
 SRCDIRS		+= usr
 SRCDIRS		+= config
 
+# These directories contain code that is not eligible for UEFI Secure
+# Boot signing.
+#
+SRCDIRS_INSEC	+= net/oncrpc
+SRCDIRS_INSEC	+= net/80211
+SRCDIRS_INSEC	+= drivers/net/rtl818x
+SRCDIRS_INSEC	+= drivers/net/ath
+SRCDIRS_INSEC	+= drivers/net/ath/ath5k
+SRCDIRS_INSEC	+= drivers/net/ath/ath9k
+
 # NON_AUTO_SRCS lists files that are excluded from the normal
 # automatic build system.
 #
@@ -151,6 +166,9 @@ all : $(ALL)
 everything :
 	$(Q)$(MAKE) --no-print-directory $(ALL) \
 		bin/3c509.rom bin/intel.rom bin/intel.mrom \
+		bin-x86_64-pcbios/8086100e.mrom bin-x86_64-pcbios/intel.rom \
+		bin-x86_64-pcbios/ipxe.usb bin-x86_64-pcbios/ipxe.pxe \
+		bin-x86_64-pcbios/undionly.kpxe \
 		bin-i386-efi/ipxe.efi bin-i386-efi/ipxe.efidrv \
 		bin-i386-efi/ipxe.efirom \
 		bin-x86_64-efi/ipxe.efi bin-x86_64-efi/ipxe.efidrv \

src/Makefile.efi

@@ -0,0 +1,46 @@
+# -*- makefile -*- : Force emacs to use Makefile mode
+
+# The EFI linker script
+#
+LDSCRIPT	= scripts/efi.lds
+
+# Retain relocation information for elf2efi
+#
+LDFLAGS		+= -q -S
+
+# Media types.
+#
+NON_AUTO_MEDIA	+= efi
+NON_AUTO_MEDIA	+= efidrv
+NON_AUTO_MEDIA	+= drv.efi
+NON_AUTO_MEDIA	+= efirom
+
+# Include SNP driver in the all-drivers build
+#
+DRIVERS_net += snp
+
+# Rules for building EFI files
+#
+$(BIN)/%.efi : $(BIN)/%.efi.tmp $(ELF2EFI)
+	$(QM)$(ECHO) "  [FINISH] $@"
+	$(Q)$(ELF2EFI) --subsystem=10 $< $@
+
+$(BIN)/%.efidrv : $(BIN)/%.efidrv.tmp $(ELF2EFI)
+	$(QM)$(ECHO) "  [FINISH] $@"
+	$(Q)$(ELF2EFI) --subsystem=11 $< $@
+
+$(BIN)/%.drv.efi : $(BIN)/%.efidrv
+	$(QM)$(ECHO) "  [FINISH] $@"
+	$(Q)$(CP) $< $@
+
+$(BIN)/%.efirom : $(BIN)/%.efidrv $(EFIROM)
+	$(QM)$(ECHO) "  [FINISH] $@"
+	$(Q)$(EFIROM) -v $(TGT_PCI_VENDOR) -d $(TGT_PCI_DEVICE) $< $@
+
+$(BIN)/efidrv.cab : $(BIN)/alldrv.efis # $(ALL_drv.efi) is not yet defined
+	$(QM)$(ECHO) "  [CAB] $@"
+	$(Q)$(LCAB) -n -q $(ALL_drv.efi) $@
+
+$(BIN)/%.usb : $(BIN)/%.efi
+	$(QM)$(ECHO) "  [GENEFIDSK] $@"
+	$(Q)bash util/genefidsk -o $@ -b $(EFI_BOOT_FILE) $<

src/Makefile.housekeeping

@@ -299,7 +299,7 @@ endif
 #
 # Select build architecture and platform based on $(BIN)
 #
-# BIN has the form bin[-[arch-]platform]
+# BIN has the form bin[-[<arch>-]<platform>[-sb]]
 
 ARCHS		:= $(patsubst arch/%,%,$(wildcard arch/*))
 PLATFORMS	:= $(patsubst config/defaults/%.h,%,\
@@ -312,17 +312,18 @@ platforms :
 
 ifdef BIN
 
-# Determine architecture portion of $(BIN), if present
-BIN_ARCH	:= $(strip $(foreach A,$(ARCHS),\
-			     $(patsubst bin-$(A)-%,$(A),\
-			       $(filter bin-$(A)-%,$(BIN)))))
-
-# Determine platform portion of $(BIN), if present
-ifeq ($(BIN_ARCH),)
-BIN_PLATFORM	:= $(patsubst bin-%,%,$(filter bin-%,$(BIN)))
+# Split $(BIN) into architecture, platform, and security flag (where present)
+BIN_ELEMENTS	:= $(subst -,$(SPACE),$(BIN))
+BIN_APS		:= $(wordlist 2,4,$(BIN_ELEMENTS))
+ifeq ($(lastword $(BIN_APS)),sb)
+BIN_AP		:= $(wordlist 2,$(words $(BIN_APS)),discard $(BIN_APS))
+BIN_SECUREBOOT	:= 1
 else
-BIN_PLATFORM	:= $(patsubst bin-$(BIN_ARCH)-%,%,$(BIN))
+BIN_AP		:= $(BIN_APS)
+BIN_SECUREBOOT	:= 0
 endif
+BIN_PLATFORM	:= $(lastword $(BIN_AP))
+BIN_ARCH	:= $(wordlist 2,$(words $(BIN_AP)),discard $(BIN_AP))
 
 # Determine build architecture
 DEFAULT_ARCH	:= i386
@@ -339,6 +340,13 @@ CFLAGS		+= -DPLATFORM=$(PLATFORM)
 platform :
 	@$(ECHO) $(PLATFORM)
 
+# Determine security flag
+DEFAULT_SECUREBOOT := 0
+SECUREBOOT	:= $(firstword $(BIN_SECUREBOOT) $(DEFAULT_SECUREBOOT))
+CFLAGS		+= -DSECUREBOOT=$(SECUREBOOT)
+secureboot :
+	@$(ECHO) $(SECUREBOOT)
+
 endif # defined(BIN)
 
 # Include architecture-specific Makefile
@@ -357,6 +365,11 @@ endif
 #
 # Source file handling
 
+# Exclude known-insecure files from Secure Boot builds
+ifeq ($(SECUREBOOT),0)
+SRCDIRS		+= $(SRCDIRS_INSEC)
+endif
+
 # SRCDIRS lists all directories containing source files.
 srcdirs :
 	@$(ECHO) $(SRCDIRS)
@@ -491,6 +504,11 @@ LDFLAGS		+= -static
 #
 CFLAGS		+= -include include/compiler.h
 
+# The section type character (e.g. "@" in "@progbits") varies by
+# architecture.
+#
+CFLAGS		+= -DASM_TCHAR='$(ASM_TCHAR)' -DASM_TCHAR_OPS='$(ASM_TCHAR_OPS)'
+
 # CFLAGS for specific object types
 #
 CFLAGS_c	+=
@@ -518,19 +536,25 @@ POST_O		:=
 POST_O_DEPS	:=
 endif
 
+# Debug level calculations
+#
+DBGLVL_MAX	= -DDBGLVL_MAX=$(firstword $(subst ., ,$(1)))
+DBGLVL_DFLT	= -DDBGLVL_DFLT=$(lastword $(subst ., ,$(1)))
+DBGLVL		= $(call DBGLVL_MAX,$(1)) $(call DBGLVL_DFLT,$(1))
+
 # Rules for specific object types.
 #
 COMPILE_c	= $(CC) $(CFLAGS) $(CFLAGS_c) $(OBJ_CFLAGS)
 RULE_c		= $(Q)$(COMPILE_c) -c $< -o $@ $(POST_O)
 RULE_c_to_ids.o = $(Q)$(ECHO_E) '$(OBJ_IDS_ASM_NL)' | $(ASSEMBLE_S) -o $@
-RULE_c_to_dbg%.o = $(Q)$(COMPILE_c) -DDBGLVL_MAX=$* -c $< -o $@ $(POST_O)
+RULE_c_to_dbg%.o= $(Q)$(COMPILE_c) $(call DBGLVL,$*) -c $< -o $@ $(POST_O)
 RULE_c_to_c	= $(Q)$(COMPILE_c) -E -c $< > $@
 RULE_c_to_s	= $(Q)$(COMPILE_c) -S -g0 -c $< -o $@
 
 PREPROCESS_S	= $(CPP) $(CFLAGS) $(CFLAGS_S) $(OBJ_CFLAGS)
 ASSEMBLE_S	= $(AS) $(ASFLAGS)
 RULE_S		= $(Q)$(PREPROCESS_S) $< | $(ASSEMBLE_S) -o $@
-RULE_S_to_dbg%.o = $(Q)$(PREPROCESS_S) -DDBGLVL_MAX=$* $< | $(ASSEMBLE_S) -o $@
+RULE_S_to_dbg%.o= $(Q)$(PREPROCESS_S) $(call DBGLVL,$*) $< | $(ASSEMBLE_S) -o $@
 RULE_S_to_s	= $(Q)$(PREPROCESS_S) $< > $@
 
 GENERIC_TARGETS	+= ids.o dbg%.o c s
@@ -709,6 +733,60 @@ config/named.h : $(CONFIG_LIST)
 
 .PRECIOUS : config/named.h
 
+# (Single-element) list of assertion configuration
+#
+ASSERT_LIST := $(BIN)/.assert.list
+ifeq ($(wildcard $(ASSERT_LIST)),)
+ASSERT_OLD := <invalid>
+else
+ASSERT_OLD := $(shell cat $(ASSERT_LIST))
+endif
+ifneq ($(ASSERT_OLD),$(ASSERT))
+$(shell $(ECHO) "$(ASSERT)" > $(ASSERT_LIST))
+endif
+
+$(ASSERT_LIST) : $(MAKEDEPS)
+
+VERYCLEANUP += $(ASSERT_LIST)
+
+# Assertion configuration
+#
+ifneq ($(ASSERT),)
+CFLAGS	+= -DASSERTING=$(ASSERT)
+endif
+
+include/assert.h : $(ASSERT_LIST)
+	$(Q)$(TOUCH) $@
+
+.PRECIOUS : include/assert.h
+
+# (Single-element) list of profiling configuration
+#
+PROFILE_LIST := $(BIN)/.profile.list
+ifeq ($(wildcard $(PROFILE_LIST)),)
+PROFILE_OLD := <invalid>
+else
+PROFILE_OLD := $(shell cat $(PROFILE_LIST))
+endif
+ifneq ($(PROFILE_OLD),$(PROFILE))
+$(shell $(ECHO) "$(PROFILE)" > $(PROFILE_LIST))
+endif
+
+$(PROFILE_LIST) : $(MAKEDEPS)
+
+VERYCLEANUP += $(PROFILE_LIST)
+
+# Profiling configuration
+#
+ifneq ($(PROFILE),)
+CFLAGS	+= -DPROFILING=$(PROFILE)
+endif
+
+include/ipxe/profile.h : $(PROFILE_LIST)
+	$(Q)$(TOUCH) $@
+
+.PRECIOUS : include/ipxe/profile.h
+
 # These files use .incbin inline assembly to include a binary file.
 # Unfortunately ccache does not detect this dependency and caches
 # builds even when the binary file has changed.
@@ -734,8 +812,8 @@ $(DBGCOL_LIST) : $(MAKEDEPS)
 VERYCLEANUP += $(DBGCOL_LIST)
 
 DBGCOL_COLOURS := $(subst -, ,$(DBGCOL))
-DBGCOL_MIN := $(word 1,$(DBGCOL_COLOURS))
-DBGCOL_MAX := $(word 2,$(DBGCOL_COLOURS))
+DBGCOL_MIN := $(firstword $(DBGCOL_COLOURS))
+DBGCOL_MAX := $(lastword $(DBGCOL_COLOURS))
 
 debug_DEPS += $(DBGCOL_LIST)
 
@@ -896,7 +974,7 @@ endif
 # Device ID tables (using IDs from ROM definition file)
 #
 define obj_pci_id_asm
-	.section ".pci_devlist.$(1)", "a", @progbits
+	.section ".pci_devlist.$(1)", "a", $(ASM_TCHAR)progbits
 	.globl pci_devlist_$(1)
 pci_devlist_$(1):
 	.short ( 0x$(1) & 0xffff )
@@ -960,13 +1038,13 @@ DRIVERS_ipxe	= $(DRIVERS_net) $(DRIVERS_infiniband) \
 # TGT_DRIVERS  : the driver for each element (e.g. "rtl8139 prism2_pci")
 # TGT_ROM_NAME : the ROM name (e.g. "dfe538")
 #
-CARD_DRIVER	= $(firstword $(DRIVER_$(1)) $(1))
 TGT_ELEMENTS	= $(subst --, ,$(firstword $(subst ., ,$(notdir $@))))
 TGT_ROM_NAME	= $(firstword $(TGT_ELEMENTS))
-TGT_DRIVERS	= $(strip $(if $(DRIVERS_$(TGT_ROM_NAME)), \
-			       $(DRIVERS_$(TGT_ROM_NAME)), \
-			       $(foreach TGT_ELEMENT,$(TGT_ELEMENTS), \
-			         $(call CARD_DRIVER,$(TGT_ELEMENT))) ))
+TGT_DRIVERS	= $(strip $(foreach TGT_ELEMENT,$(TGT_ELEMENTS), \
+				    $(if $(DRIVERS_$(TGT_ELEMENT)), \
+					 $(DRIVERS_$(TGT_ELEMENT)), \
+					 $(firstword $(DRIVER_$(TGT_ELEMENT)) \
+						     $(TGT_ELEMENT)))))
 TGT_PREFIX_NAME	= $(word 2,$(subst ., ,$(notdir $@)))
 TGT_PREFIX	= $(strip $(if $(filter rom,$(TGT_PREFIX_NAME)), \
 			       $(ROM_TYPE_$(TGT_ROM_NAME))rom, \
@@ -1014,10 +1092,12 @@ TGT_LD_FLAGS	= $(foreach SYM,$(TGT_LD_ENTRY) $(TGT_LD_DRIVERS) \
 # the target.
 #
 DEBUG_LIST	= $(subst $(COMMA), ,$(DEBUG))
-DEBUG_OBJ_LEVEL	= $(firstword $(word 2,$(subst :, ,$(1))) 1)
-DEBUG_OBJ_BASE	= $(word 1,$(subst :, ,$(1))).dbg$(call DEBUG_OBJ_LEVEL,$(1))
-DEBUG_OBJ	= $(BIN)/$(call DEBUG_OBJ_BASE,$(1)).o
-DEBUG_ORIG_OBJ	= $(BIN)/$(word 1,$(subst :, ,$(1))).o
+DEBUG_MAX	= $(firstword $(word 2,$(subst :, ,$(1))) 1)
+DEBUG_DFLT	= $(if $(word 3,$(subst :, ,$(1))),.$(word 3,$(subst :, ,$(1))))
+DEBUG_LEVEL	= $(call DEBUG_MAX,$(1))$(call DEBUG_DFLT,$(1))
+DEBUG_BASE	= $(firstword $(subst :, ,$(1))).dbg$(call DEBUG_LEVEL,$(1))
+DEBUG_OBJ	= $(BIN)/$(call DEBUG_BASE,$(1)).o
+DEBUG_ORIG_OBJ	= $(BIN)/$(firstword $(subst :, ,$(1))).o
 DEBUG_OBJS	= $(foreach D,$(DEBUG_LIST),$(call DEBUG_OBJ,$(D)))
 DEBUG_ORIG_OBJS	= $(foreach D,$(DEBUG_LIST),$(call DEBUG_ORIG_OBJ,$(D)))
 BLIB_OBJS	= $(DEBUG_OBJS) $(filter-out $(DEBUG_ORIG_OBJS),$(BOBJS))
@@ -1298,21 +1378,15 @@ CLEANUP += $(ZBIN)
 #
 # The EFI image converter
 #
-ELF2EFI_CFLAGS	:= -I$(BINUTILS_DIR)/include -I$(BFD_DIR)/include \
-		   -I$(ZLIB_DIR)/include -idirafter include
-ELF2EFI_LDFLAGS	:= -L$(BINUTILS_DIR)/lib -L$(BFD_DIR)/lib -L$(ZLIB_DIR)/lib \
-		   -lbfd -ldl -lz -Wl,--no-warn-search-mismatch
 
 $(ELF2EFI32) : util/elf2efi.c $(MAKEDEPS)
 	$(QM)$(ECHO) "  [HOSTCC] $@"
-	$(Q)$(HOST_CC) $(HOST_CFLAGS) $(ELF2EFI_CFLAGS) -DEFI_TARGET_IA32 $< \
-		$(ELF2EFI_LDFLAGS) -o $@
+	$(Q)$(HOST_CC) $(HOST_CFLAGS) -idirafter include -DEFI_TARGET32 $< -o $@
 CLEANUP += $(ELF2EFI32)
 
 $(ELF2EFI64) : util/elf2efi.c $(MAKEDEPS)
 	$(QM)$(ECHO) "  [HOSTCC] $@"
-	$(Q)$(HOST_CC) $(HOST_CFLAGS) $(ELF2EFI_CFLAGS) -DEFI_TARGET_X64 $< \
-		$(ELF2EFI_LDFLAGS) -o $@
+	$(Q)$(HOST_CC) $(HOST_CFLAGS) -idirafter include -DEFI_TARGET64 $< -o $@
 CLEANUP += $(ELF2EFI64)
 
 $(EFIROM) : util/efirom.c $(MAKEDEPS)

src/arch/arm/Makefile

@@ -0,0 +1,12 @@
+# Assembler section type character
+#
+ASM_TCHAR	:= %
+ASM_TCHAR_OPS	:= %%
+
+# Include common ARM headers
+#
+INCDIRS		+= arch/arm/include
+
+# ARM-specific directories containing source files
+#
+SRCDIRS		+= arch/arm/interface/efi

src/arch/arm/Makefile.efi

@@ -0,0 +1,6 @@
+# -*- makefile -*- : Force emacs to use Makefile mode
+
+# Include generic EFI Makefile
+#
+MAKEDEPS	+= Makefile.efi
+include Makefile.efi

src/arch/arm/core/arm_io.c

@@ -0,0 +1,93 @@
+/*
+ * Copyright (C) 2016 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * 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 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <ipxe/io.h>
+#include <ipxe/arm_io.h>
+
+/** @file
+ *
+ * iPXE I/O API for ARM
+ *
+ */
+
+/** An ARM I/O qword */
+union arm32_io_qword {
+	uint64_t qword;
+	uint32_t dword[2];
+};
+
+/**
+ * Read 64-bit qword from memory-mapped device
+ *
+ * @v io_addr		I/O address
+ * @ret data		Value read
+ *
+ * This is not atomic for ARM32.
+ */
+static uint64_t arm32_readq ( volatile uint64_t *io_addr ) {
+	volatile union arm32_io_qword *ptr =
+		container_of ( io_addr, union arm32_io_qword, qword );
+	union arm32_io_qword tmp;
+
+	tmp.dword[0] = readl ( &ptr->dword[0] );
+	tmp.dword[1] = readl ( &ptr->dword[1] );
+	return tmp.qword;
+}
+
+/**
+ * Write 64-bit qword to memory-mapped device
+ *
+ * @v data		Value to write
+ * @v io_addr		I/O address
+ *
+ * This is not atomic for ARM32.
+ */
+static void arm32_writeq ( uint64_t data, volatile uint64_t *io_addr ) {
+	volatile union arm32_io_qword *ptr =
+		container_of ( io_addr, union arm32_io_qword, qword );
+	union arm32_io_qword tmp;
+
+	tmp.qword = data;
+	writel ( tmp.dword[0], &ptr->dword[0] );
+	writel ( tmp.dword[1], &ptr->dword[1] );
+}
+
+PROVIDE_IOAPI_INLINE ( arm, phys_to_bus );
+PROVIDE_IOAPI_INLINE ( arm, bus_to_phys );
+PROVIDE_IOAPI_INLINE ( arm, readb );
+PROVIDE_IOAPI_INLINE ( arm, readw );
+PROVIDE_IOAPI_INLINE ( arm, readl );
+PROVIDE_IOAPI_INLINE ( arm, writeb );
+PROVIDE_IOAPI_INLINE ( arm, writew );
+PROVIDE_IOAPI_INLINE ( arm, writel );
+PROVIDE_IOAPI_INLINE ( arm, iodelay );
+PROVIDE_IOAPI_INLINE ( arm, mb );
+#ifdef __aarch64__
+PROVIDE_IOAPI_INLINE ( arm, readq );
+PROVIDE_IOAPI_INLINE ( arm, writeq );
+#else
+PROVIDE_IOAPI ( arm, readq, arm32_readq );
+PROVIDE_IOAPI ( arm, writeq, arm32_writeq );
+#endif

src/arch/arm/include/bits/acpi.h

@@ -0,0 +1,12 @@
+#ifndef _BITS_ACPI_H
+#define _BITS_ACPI_H
+
+/** @file
+ *
+ * ARM-specific ACPI API implementations
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#endif /* _BITS_ACPI_H */

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

@@ -0,0 +1,13 @@
+#ifndef _BITS_ENDIAN_H
+#define _BITS_ENDIAN_H
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/* ARM may be either little-endian or big-endian */
+#ifdef __ARM_BIG_ENDIAN
+#define __BYTE_ORDER __BIG_ENDIAN
+#else
+#define __BYTE_ORDER __LITTLE_ENDIAN
+#endif
+
+#endif /* _BITS_ENDIAN_H */

src/arch/arm/include/bits/entropy.h

@@ -3,7 +3,7 @@
 
 /** @file
  *
- * x86_64-specific entropy API implementations
+ * ARM-specific entropy API implementations
  *
  */
 

src/arch/arm/include/bits/errfile.h

@@ -0,0 +1,19 @@
+#ifndef _BITS_ERRFILE_H
+#define _BITS_ERRFILE_H
+
+/** @file
+ *
+ * ARM-specific error file identifiers
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/**
+ * @addtogroup errfile Error file identifiers
+ * @{
+ */
+
+/** @} */
+
+#endif /* _BITS_ERRFILE_H */

src/arch/arm/include/bits/hyperv.h

@@ -0,0 +1,12 @@
+#ifndef _BITS_HYPERV_H
+#define _BITS_HYPERV_H
+
+/** @file
+ *
+ * Hyper-V interface
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#endif /* _BITS_HYPERV_H */

src/arch/arm/include/bits/io.h

@@ -0,0 +1,14 @@
+#ifndef _BITS_IO_H
+#define _BITS_IO_H
+
+/** @file
+ *
+ * ARM-specific I/O API implementations
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <ipxe/arm_io.h>
+
+#endif /* _BITS_IO_H */

src/arch/arm/include/bits/iomap.h

@@ -0,0 +1,12 @@
+#ifndef _BITS_IOMAP_H
+#define _BITS_IOMAP_H
+
+/** @file
+ *
+ * ARM-specific I/O mapping API implementations
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#endif /* _BITS_IOMAP_H */

src/arch/arm/include/bits/nap.h

@@ -0,0 +1,14 @@
+#ifndef _BITS_NAP_H
+#define _BITS_NAP_H
+
+/** @file
+ *
+ * ARM-specific CPU sleeping API implementations
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <ipxe/efi/efiarm_nap.h>
+
+#endif /* _BITS_MAP_H */

src/arch/arm/include/bits/pci_io.h

@@ -0,0 +1,14 @@
+#ifndef _BITS_PCI_IO_H
+#define _BITS_PCI_IO_H
+
+/** @file
+ *
+ * ARM PCI I/O API implementations
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <ipxe/io.h>
+
+#endif /* _BITS_PCI_IO_H */

src/arch/arm/include/bits/reboot.h

@@ -3,7 +3,7 @@
 
 /** @file
  *
- * x86_64-specific reboot API implementations
+ * ARM-specific reboot API implementations
  *
  */
 

src/arch/arm/include/bits/sanboot.h

@@ -3,7 +3,7 @@
 
 /** @file
  *
- * x86_64-specific sanboot API implementations
+ * ARM-specific sanboot API implementations
  *
  */
 

src/arch/arm/include/bits/smbios.h

@@ -3,8 +3,10 @@
 
 /** @file
  *
- * i386-specific SMBIOS API implementations
+ * ARM-specific SMBIOS API implementations
  *
  */
 
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
 #endif /* _BITS_SMBIOS_H */

src/arch/arm/include/bits/time.h

@@ -3,7 +3,7 @@
 
 /** @file
  *
- * x86_64-specific time API implementations
+ * ARM-specific time API implementations
  *
  */
 

src/arch/arm/include/bits/uaccess.h

@@ -3,8 +3,10 @@
 
 /** @file
  *
- * x86_64-specific user access API implementations
+ * ARM-specific user access API implementations
  *
  */
 
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
 #endif /* _BITS_UACCESS_H */

src/arch/arm/include/bits/uart.h

@@ -0,0 +1,12 @@
+#ifndef _BITS_UART_H
+#define _BITS_UART_H
+
+/** @file
+ *
+ * 16550-compatible UART
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#endif /* _BITS_UART_H */

src/arch/arm/include/bits/umalloc.h

@@ -0,0 +1,12 @@
+#ifndef _BITS_UMALLOC_H
+#define _BITS_UMALLOC_H
+
+/** @file
+ *
+ * ARM-specific user memory allocation API implementations
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#endif /* _BITS_UMALLOC_H */

src/arch/arm/include/bits/xen.h

@@ -0,0 +1,158 @@
+#ifndef _BITS_XEN_H
+#define _BITS_XEN_H
+
+/** @file
+ *
+ * Xen interface
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/* Hypercall registers */
+#ifdef __aarch64__
+#define XEN_HC "x16"
+#define XEN_REG1 "x0"
+#define XEN_REG2 "x1"
+#define XEN_REG3 "x2"
+#define XEN_REG4 "x3"
+#define XEN_REG5 "x4"
+#else
+#define XEN_HC "r12"
+#define XEN_REG1 "r0"
+#define XEN_REG2 "r1"
+#define XEN_REG3 "r2"
+#define XEN_REG4 "r3"
+#define XEN_REG5 "r4"
+#endif
+
+/**
+ * Issue hypercall with one argument
+ *
+ * @v xen		Xen hypervisor
+ * @v hypercall		Hypercall number
+ * @v arg1		First argument
+ * @ret retval		Return value
+ */
+static inline __attribute__ (( always_inline )) unsigned long
+xen_hypercall_1 ( struct xen_hypervisor *xen __unused, unsigned int hypercall,
+		  unsigned long arg1 ) {
+	register unsigned long hc asm ( XEN_HC ) = hypercall;
+	register unsigned long reg1 asm ( XEN_REG1 ) = arg1;
+
+	__asm__ __volatile__ ( "hvc %1"
+			       : "+r" ( reg1 )
+			       : "i" ( XEN_HYPERCALL_TAG ), "r" ( hc )
+			       : "memory", "cc" );
+	return reg1;
+}
+
+/**
+ * Issue hypercall with two arguments
+ *
+ * @v xen		Xen hypervisor
+ * @v hypercall		Hypercall number
+ * @v arg1		First argument
+ * @v arg2		Second argument
+ * @ret retval		Return value
+ */
+static inline __attribute__ (( always_inline )) unsigned long
+xen_hypercall_2 ( struct xen_hypervisor *xen __unused, unsigned int hypercall,
+		  unsigned long arg1, unsigned long arg2 ) {
+	register unsigned long hc asm ( XEN_HC ) = hypercall;
+	register unsigned long reg1 asm ( XEN_REG1 ) = arg1;
+	register unsigned long reg2 asm ( XEN_REG2 ) = arg2;
+
+	__asm__ __volatile__ ( "hvc %2"
+			       : "+r" ( reg1 ), "+r" ( reg2 )
+			       : "i" ( XEN_HYPERCALL_TAG ), "r" ( hc )
+			       : "memory", "cc" );
+	return reg1;
+}
+
+/**
+ * Issue hypercall with three arguments
+ *
+ * @v xen		Xen hypervisor
+ * @v hypercall		Hypercall number
+ * @v arg1		First argument
+ * @v arg2		Second argument
+ * @v arg3		Third argument
+ * @ret retval		Return value
+ */
+static inline __attribute__ (( always_inline )) unsigned long
+xen_hypercall_3 ( struct xen_hypervisor *xen __unused, unsigned int hypercall,
+		  unsigned long arg1, unsigned long arg2, unsigned long arg3 ) {
+	register unsigned long hc asm ( XEN_HC ) = hypercall;
+	register unsigned long reg1 asm ( XEN_REG1 ) = arg1;
+	register unsigned long reg2 asm ( XEN_REG2 ) = arg2;
+	register unsigned long reg3 asm ( XEN_REG3 ) = arg3;
+
+	__asm__ __volatile__ ( "hvc %3"
+			       : "+r" ( reg1 ), "+r" ( reg2 ), "+r" ( reg3 )
+			       : "i" ( XEN_HYPERCALL_TAG ), "r" ( hc )
+			       : "memory", "cc" );
+	return reg1;
+}
+
+/**
+ * Issue hypercall with four arguments
+ *
+ * @v xen		Xen hypervisor
+ * @v hypercall		Hypercall number
+ * @v arg1		First argument
+ * @v arg2		Second argument
+ * @v arg3		Third argument
+ * @v arg4		Fourth argument
+ * @ret retval		Return value
+ */
+static inline __attribute__ (( always_inline )) unsigned long
+xen_hypercall_4 ( struct xen_hypervisor *xen __unused, unsigned int hypercall,
+		  unsigned long arg1, unsigned long arg2, unsigned long arg3,
+		  unsigned long arg4 ) {
+	register unsigned long hc asm ( XEN_HC ) = hypercall;
+	register unsigned long reg1 asm ( XEN_REG1 ) = arg1;
+	register unsigned long reg2 asm ( XEN_REG2 ) = arg2;
+	register unsigned long reg3 asm ( XEN_REG3 ) = arg3;
+	register unsigned long reg4 asm ( XEN_REG4 ) = arg4;
+
+	__asm__ __volatile__ ( "hvc %4"
+			       : "+r" ( reg1 ), "+r" ( reg2 ), "+r" ( reg3 ),
+				 "+r" ( reg4 )
+			       : "i" ( XEN_HYPERCALL_TAG ), "r" ( hc )
+			       : "memory", "cc" );
+	return reg1;
+}
+
+/**
+ * Issue hypercall with five arguments
+ *
+ * @v xen		Xen hypervisor
+ * @v hypercall		Hypercall number
+ * @v arg1		First argument
+ * @v arg2		Second argument
+ * @v arg3		Third argument
+ * @v arg4		Fourth argument
+ * @v arg5		Fifth argument
+ * @ret retval		Return value
+ */
+static inline __attribute__ (( always_inline )) unsigned long
+xen_hypercall_5 ( struct xen_hypervisor *xen __unused, unsigned int hypercall,
+		  unsigned long arg1, unsigned long arg2, unsigned long arg3,
+		  unsigned long arg4, unsigned long arg5 ) {
+	register unsigned long hc asm ( XEN_HC ) = hypercall;
+	register unsigned long reg1 asm ( XEN_REG1 ) = arg1;
+	register unsigned long reg2 asm ( XEN_REG2 ) = arg2;
+	register unsigned long reg3 asm ( XEN_REG3 ) = arg3;
+	register unsigned long reg4 asm ( XEN_REG4 ) = arg4;
+	register unsigned long reg5 asm ( XEN_REG5 ) = arg5;
+
+	__asm__ __volatile__ ( "hvc %5"
+			       : "+r" ( reg1 ), "+r" ( reg2 ), "+r" ( reg3 ),
+				 "+r" ( reg4 ), "+r" ( reg5 )
+			       : "i" ( XEN_HYPERCALL_TAG ), "r" ( hc )
+			       : "memory", "cc" );
+	return reg1;
+}
+
+#endif /* _BITS_XEN_H */

src/arch/arm/include/ipxe/arm_io.h

@@ -0,0 +1,105 @@
+#ifndef _IPXE_ARM_IO_H
+#define _IPXE_ARM_IO_H
+
+/** @file
+ *
+ * iPXE I/O API for ARM
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#ifdef IOAPI_ARM
+#define IOAPI_PREFIX_arm
+#else
+#define IOAPI_PREFIX_arm __arm_
+#endif
+
+/*
+ * Memory space mappings
+ *
+ */
+
+/** Page shift */
+#define PAGE_SHIFT 12
+
+/*
+ * Physical<->Bus address mappings
+ *
+ */
+
+static inline __always_inline unsigned long
+IOAPI_INLINE ( arm, phys_to_bus ) ( unsigned long phys_addr ) {
+	return phys_addr;
+}
+
+static inline __always_inline unsigned long
+IOAPI_INLINE ( arm, bus_to_phys ) ( unsigned long bus_addr ) {
+	return bus_addr;
+}
+
+/*
+ * MMIO reads and writes up to native word size
+ *
+ */
+
+#define ARM_READX( _api_func, _type, _insn_suffix, _reg_prefix )	      \
+static inline __always_inline _type					      \
+IOAPI_INLINE ( arm, _api_func ) ( volatile _type *io_addr ) {		      \
+	_type data;							      \
+	__asm__ __volatile__ ( "ldr" _insn_suffix " %" _reg_prefix "0, %1"    \
+			       : "=r" ( data ) : "Qo" ( *io_addr ) );	      \
+	return data;							      \
+}
+#ifdef __aarch64__
+ARM_READX ( readb, uint8_t, "b", "w" );
+ARM_READX ( readw, uint16_t, "h", "w" );
+ARM_READX ( readl, uint32_t, "", "w" );
+ARM_READX ( readq, uint64_t, "", "" );
+#else
+ARM_READX ( readb, uint8_t, "b", "" );
+ARM_READX ( readw, uint16_t, "h", "" );
+ARM_READX ( readl, uint32_t, "", "" );
+#endif
+
+#define ARM_WRITEX( _api_func, _type, _insn_suffix, _reg_prefix )			\
+static inline __always_inline void					      \
+IOAPI_INLINE ( arm, _api_func ) ( _type data, volatile _type *io_addr ) {     \
+	__asm__ __volatile__ ( "str" _insn_suffix " %" _reg_prefix "0, %1"    \
+			       : : "r" ( data ), "Qo" ( *io_addr ) );	      \
+}
+#ifdef __aarch64__
+ARM_WRITEX ( writeb, uint8_t, "b", "w" );
+ARM_WRITEX ( writew, uint16_t, "h", "w" );
+ARM_WRITEX ( writel, uint32_t, "", "w" );
+ARM_WRITEX ( writeq, uint64_t, "", "" );
+#else
+ARM_WRITEX ( writeb, uint8_t, "b", "" );
+ARM_WRITEX ( writew, uint16_t, "h", "" );
+ARM_WRITEX ( writel, uint32_t, "", "" );
+#endif
+
+/*
+ * Slow down I/O
+ *
+ */
+static inline __always_inline void
+IOAPI_INLINE ( arm, iodelay ) ( void ) {
+	/* Nothing to do */
+}
+
+/*
+ * Memory barrier
+ *
+ */
+static inline __always_inline void
+IOAPI_INLINE ( arm, mb ) ( void ) {
+
+#ifdef __aarch64__
+	__asm__ __volatile__ ( "dmb sy" );
+#else
+	__asm__ __volatile__ ( "dmb" );
+#endif
+}
+
+#endif /* _IPXE_ARM_IO_H */

src/arch/arm/include/ipxe/efi/efiarm_nap.h

@@ -0,0 +1,18 @@
+#ifndef _IPXE_EFIARM_NAP_H
+#define _IPXE_EFIARM_NAP_H
+
+/** @file
+ *
+ * EFI CPU sleeping
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#ifdef NAP_EFIARM
+#define NAP_PREFIX_efiarm
+#else
+#define NAP_PREFIX_efiarm __efiarm_
+#endif
+
+#endif /* _IPXE_EFIARM_NAP_H */

src/arch/arm/interface/efi/efiarm_nap.c

@@ -0,0 +1,53 @@
+/*
+ * Copyright (C) 2016 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * 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 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <ipxe/nap.h>
+#include <ipxe/efi/efi.h>
+
+/** @file
+ *
+ * iPXE CPU sleeping API for EFI
+ *
+ */
+
+/**
+ * Sleep until next interrupt
+ *
+ */
+static void efiarm_cpu_nap ( void ) {
+	/*
+	 * I can't find any EFI API that allows us to put the CPU to
+	 * sleep.  The CpuSleep() function is defined in CpuLib.h, but
+	 * isn't part of any exposed protocol so we have no way to
+	 * call it.
+	 *
+	 * The EFI shell doesn't seem to bother sleeping the CPU; it
+	 * just sits there idly burning power.
+	 *
+	 */
+	__asm__ __volatile__ ( "wfi" );
+}
+
+PROVIDE_NAP ( efiarm, cpu_nap, efiarm_cpu_nap );

src/arch/arm32/Makefile

@@ -0,0 +1,23 @@
+# ARM32-specific directories containing source files
+#
+SRCDIRS		+= arch/arm32/core
+SRCDIRS		+= arch/arm32/libgcc
+
+# ARM32-specific flags
+#
+CFLAGS		+= -mthumb -mcpu=cortex-a15 -mabi=aapcs -mfloat-abi=soft
+CFLAGS		+= -mword-relocations
+ASFLAGS		+= -mthumb -mcpu=cortex-a15
+
+# EFI requires -fshort-wchar, and nothing else currently uses wchar_t
+#
+CFLAGS		+= -fshort-wchar
+
+# Include common ARM Makefile
+MAKEDEPS	+= arch/arm/Makefile
+include arch/arm/Makefile
+
+# Include platform-specific Makefile
+#
+MAKEDEPS	+= arch/arm32/Makefile.$(PLATFORM)
+include arch/arm32/Makefile.$(PLATFORM)

src/arch/arm32/Makefile.efi

@@ -0,0 +1,18 @@
+# -*- makefile -*- : Force emacs to use Makefile mode
+
+# UEFI requires that enums are always 32 bits
+#
+CFLAGS		+= -fno-short-enums
+
+# Specify EFI image builder
+#
+ELF2EFI		= $(ELF2EFI32)
+
+# Specify EFI boot file
+#
+EFI_BOOT_FILE	= bootarm.efi
+
+# Include generic EFI Makefile
+#
+MAKEDEPS	+= arch/arm/Makefile.efi
+include arch/arm/Makefile.efi

src/arch/arm32/core/arm32_bigint.c

@@ -0,0 +1,102 @@
+/*
+ * Copyright (C) 2016 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * 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 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+#include <string.h>
+#include <ipxe/bigint.h>
+
+/** @file
+ *
+ * Big integer support
+ */
+
+/**
+ * Multiply big integers
+ *
+ * @v multiplicand0	Element 0 of big integer to be multiplied
+ * @v multiplier0	Element 0 of big integer to be multiplied
+ * @v result0		Element 0 of big integer to hold result
+ * @v size		Number of elements
+ */
+void bigint_multiply_raw ( const uint32_t *multiplicand0,
+			   const uint32_t *multiplier0,
+			   uint32_t *result0, unsigned int size ) {
+	const bigint_t ( size ) __attribute__ (( may_alias )) *multiplicand =
+		( ( const void * ) multiplicand0 );
+	const bigint_t ( size ) __attribute__ (( may_alias )) *multiplier =
+		( ( const void * ) multiplier0 );
+	bigint_t ( size * 2 ) __attribute__ (( may_alias )) *result =
+		( ( void * ) result0 );
+	unsigned int i;
+	unsigned int j;
+	uint32_t multiplicand_element;
+	uint32_t multiplier_element;
+	uint32_t *result_elements;
+	uint32_t discard_low;
+	uint32_t discard_high;
+	uint32_t discard_temp;
+
+	/* Zero result */
+	memset ( result, 0, sizeof ( *result ) );
+
+	/* Multiply integers one element at a time */
+	for ( i = 0 ; i < size ; i++ ) {
+		multiplicand_element = multiplicand->element[i];
+		for ( j = 0 ; j < size ; j++ ) {
+			multiplier_element = multiplier->element[j];
+			result_elements = &result->element[ i + j ];
+			/* Perform a single multiply, and add the
+			 * resulting double-element into the result,
+			 * carrying as necessary.  The carry can
+			 * never overflow beyond the end of the
+			 * result, since:
+			 *
+			 *     a < 2^{n}, b < 2^{n} => ab < 2^{2n}
+			 */
+			__asm__ __volatile__ ( "umull %1, %2, %5, %6\n\t"
+					       "ldr %3, [%0]\n\t"
+					       "adds %3, %1\n\t"
+					       "stmia %0!, {%3}\n\t"
+					       "ldr %3, [%0]\n\t"
+					       "adcs %3, %2\n\t"
+					       "stmia %0!, {%3}\n\t"
+					       "bcc 2f\n\t"
+					       "\n1:\n\t"
+					       "ldr %3, [%0]\n\t"
+					       "adcs %3, #0\n\t"
+					       "stmia %0!, {%3}\n\t"
+					       "bcs 1b\n\t"
+					       "\n2:\n\t"
+					       : "+l" ( result_elements ),
+						 "=l" ( discard_low ),
+						 "=l" ( discard_high ),
+						 "=l" ( discard_temp ),
+						 "+m" ( *result )
+					       : "l" ( multiplicand_element ),
+						 "l" ( multiplier_element )
+					       : "cc" );
+		}
+	}
+}

src/arch/arm32/core/setjmp.S

@@ -0,0 +1,32 @@
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL )
+
+	.text
+	.arm
+
+/*
+ * Save stack context for non-local goto
+ */
+	.globl	setjmp
+	.type	setjmp, %function
+setjmp:
+	/* Store registers */
+	stmia	r0, { r4, r5, r6, r7, r8, r9, r10, fp, sp, lr }
+	/* Return 0 when returning as setjmp() */
+	mov	r0, #0
+	bx	lr
+	.size	setjmp, . - setjmp
+
+/*
+ * Non-local jump to a saved stack context
+ */
+	.globl	longjmp
+	.type	longjmp, %function
+longjmp:
+	/* Restore registers */
+	ldmia	r0, { r4, r5, r6, r7, r8, r9, r10, fp, sp, lr }
+	/* Force result to non-zero */
+	movs	r0, r1
+	moveq	r0, #1
+	/* Return to setjmp() caller */
+	bx	lr
+	.size	longjmp, . - longjmp

src/arch/arm32/include/bits/bigint.h

@@ -0,0 +1,316 @@
+#ifndef _BITS_BIGINT_H
+#define _BITS_BIGINT_H
+
+/** @file
+ *
+ * Big integer support
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+#include <string.h>
+#include <strings.h>
+
+/** Element of a big integer */
+typedef uint32_t bigint_element_t;
+
+/**
+ * Initialise big integer
+ *
+ * @v value0		Element 0 of big integer to initialise
+ * @v size		Number of elements
+ * @v data		Raw data
+ * @v len		Length of raw data
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_init_raw ( uint32_t *value0, unsigned int size,
+		  const void *data, size_t len ) {
+	size_t pad_len = ( sizeof ( bigint_t ( size ) ) - len );
+	uint8_t *value_byte = ( ( void * ) value0 );
+	const uint8_t *data_byte = ( data + len );
+
+	/* Copy raw data in reverse order, padding with zeros */
+	while ( len-- )
+		*(value_byte++) = *(--data_byte);
+	while ( pad_len-- )
+		*(value_byte++) = 0;
+}
+
+/**
+ * Add big integers
+ *
+ * @v addend0		Element 0 of big integer to add
+ * @v value0		Element 0 of big integer to be added to
+ * @v size		Number of elements
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_add_raw ( const uint32_t *addend0, uint32_t *value0,
+		 unsigned int size ) {
+	bigint_t ( size ) __attribute__ (( may_alias )) *value =
+		( ( void * ) value0 );
+	uint32_t *discard_addend;
+	uint32_t *discard_value;
+	uint32_t *discard_end;
+	uint32_t discard_addend_i;
+	uint32_t discard_value_i;
+
+	__asm__ __volatile__ ( "adds %2, %0, %8, lsl #2\n\t" /* clear CF */
+			       "\n1:\n\t"
+			       "ldmia %0!, {%3}\n\t"
+			       "ldr %4, [%1]\n\t"
+			       "adcs %4, %3\n\t"
+			       "stmia %1!, {%4}\n\t"
+			       "teq %0, %2\n\t"
+			       "bne 1b\n\t"
+			       : "=l" ( discard_addend ),
+				 "=l" ( discard_value ),
+				 "=l" ( discard_end ),
+				 "=l" ( discard_addend_i ),
+				 "=l" ( discard_value_i ),
+				 "+m" ( *value )
+			       : "0" ( addend0 ), "1" ( value0 ), "l" ( size )
+			       : "cc" );
+}
+
+/**
+ * Subtract big integers
+ *
+ * @v subtrahend0	Element 0 of big integer to subtract
+ * @v value0		Element 0 of big integer to be subtracted from
+ * @v size		Number of elements
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_subtract_raw ( const uint32_t *subtrahend0, uint32_t *value0,
+		      unsigned int size ) {
+	bigint_t ( size ) __attribute__ (( may_alias )) *value =
+		( ( void * ) value0 );
+	uint32_t *discard_subtrahend;
+	uint32_t *discard_value;
+	uint32_t *discard_end;
+	uint32_t discard_subtrahend_i;
+	uint32_t discard_value_i;
+
+	__asm__ __volatile__ ( "add %2, %0, %8, lsl #2\n\t"
+			       "cmp %2, %0\n\t" /* set CF */
+			       "\n1:\n\t"
+			       "ldmia %0!, {%3}\n\t"
+			       "ldr %4, [%1]\n\t"
+			       "sbcs %4, %3\n\t"
+			       "stmia %1!, {%4}\n\t"
+			       "teq %0, %2\n\t"
+			       "bne 1b\n\t"
+			       : "=l" ( discard_subtrahend ),
+				 "=l" ( discard_value ),
+				 "=l" ( discard_end ),
+				 "=l" ( discard_subtrahend_i ),
+				 "=l" ( discard_value_i ),
+				 "+m" ( *value )
+			       : "0" ( subtrahend0 ), "1" ( value0 ),
+				 "l" ( size )
+			       : "cc" );
+}
+
+/**
+ * Rotate big integer left
+ *
+ * @v value0		Element 0 of big integer
+ * @v size		Number of elements
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_rol_raw ( uint32_t *value0, unsigned int size ) {
+	bigint_t ( size ) __attribute__ (( may_alias )) *value =
+		( ( void * ) value0 );
+	uint32_t *discard_value;
+	uint32_t *discard_end;
+	uint32_t discard_value_i;
+
+	__asm__ __volatile__ ( "adds %1, %0, %5, lsl #2\n\t" /* clear CF */
+			       "\n1:\n\t"
+			       "ldr %2, [%0]\n\t"
+			       "adcs %2, %2\n\t"
+			       "stmia %0!, {%2}\n\t"
+			       "teq %0, %1\n\t"
+			       "bne 1b\n\t"
+			       : "=l" ( discard_value ),
+				 "=l" ( discard_end ),
+				 "=l" ( discard_value_i ),
+				 "+m" ( *value )
+			       : "0" ( value0 ), "1" ( size )
+			       : "cc" );
+}
+
+/**
+ * Rotate big integer right
+ *
+ * @v value0		Element 0 of big integer
+ * @v size		Number of elements
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_ror_raw ( uint32_t *value0, unsigned int size ) {
+	bigint_t ( size ) __attribute__ (( may_alias )) *value =
+		( ( void * ) value0 );
+	uint32_t *discard_value;
+	uint32_t *discard_end;
+	uint32_t discard_value_i;
+
+	__asm__ __volatile__ ( "adds %1, %0, %5, lsl #2\n\t" /* clear CF */
+			       "\n1:\n\t"
+			       "ldmdb %1!, {%2}\n\t"
+			       "rrxs %2, %2\n\t"
+			       "str %2, [%1]\n\t"
+			       "teq %0, %1\n\t"
+			       "bne 1b\n\t"
+			       : "=l" ( discard_value ),
+				 "=l" ( discard_end ),
+				 "=l" ( discard_value_i ),
+				 "+m" ( *value )
+			       : "0" ( value0 ), "1" ( size )
+			       : "cc" );
+}
+
+/**
+ * Test if big integer is equal to zero
+ *
+ * @v value0		Element 0 of big integer
+ * @v size		Number of elements
+ * @ret is_zero		Big integer is equal to zero
+ */
+static inline __attribute__ (( always_inline, pure )) int
+bigint_is_zero_raw ( const uint32_t *value0, unsigned int size ) {
+	const uint32_t *value = value0;
+	uint32_t value_i;
+
+	do {
+		value_i = *(value++);
+		if ( value_i )
+			break;
+	} while ( --size );
+
+	return ( value_i == 0 );
+}
+
+/**
+ * Compare big integers
+ *
+ * @v value0		Element 0 of big integer
+ * @v reference0	Element 0 of reference big integer
+ * @v size		Number of elements
+ * @ret geq		Big integer is greater than or equal to the reference
+ */
+static inline __attribute__ (( always_inline, pure )) int
+bigint_is_geq_raw ( const uint32_t *value0, const uint32_t *reference0,
+		    unsigned int size ) {
+	const uint32_t *value = ( value0 + size );
+	const uint32_t *reference = ( reference0 + size );
+	uint32_t value_i;
+	uint32_t reference_i;
+
+	do {
+		value_i = *(--value);
+		reference_i = *(--reference);
+		if ( value_i != reference_i )
+			break;
+	} while ( --size );
+
+	return ( value_i >= reference_i );
+}
+
+/**
+ * Test if bit is set in big integer
+ *
+ * @v value0		Element 0 of big integer
+ * @v size		Number of elements
+ * @v bit		Bit to test
+ * @ret is_set		Bit is set
+ */
+static inline __attribute__ (( always_inline )) int
+bigint_bit_is_set_raw ( const uint32_t *value0, unsigned int size,
+			unsigned int bit ) {
+	const bigint_t ( size ) __attribute__ (( may_alias )) *value =
+		( ( const void * ) value0 );
+	unsigned int index = ( bit / ( 8 * sizeof ( value->element[0] ) ) );
+	unsigned int subindex = ( bit % ( 8 * sizeof ( value->element[0] ) ) );
+
+	return ( value->element[index] & ( 1 << subindex ) );
+}
+
+/**
+ * Find highest bit set in big integer
+ *
+ * @v value0		Element 0 of big integer
+ * @v size		Number of elements
+ * @ret max_bit		Highest bit set + 1 (or 0 if no bits set)
+ */
+static inline __attribute__ (( always_inline )) int
+bigint_max_set_bit_raw ( const uint32_t *value0, unsigned int size ) {
+	const uint32_t *value = ( value0 + size );
+	int max_bit = ( 8 * sizeof ( bigint_t ( size ) ) );
+	uint32_t value_i;
+
+	do {
+		value_i = *(--value);
+		max_bit -= ( 32 - fls ( value_i ) );
+		if ( value_i )
+			break;
+	} while ( --size );
+
+	return max_bit;
+}
+
+/**
+ * Grow big integer
+ *
+ * @v source0		Element 0 of source big integer
+ * @v source_size	Number of elements in source big integer
+ * @v dest0		Element 0 of destination big integer
+ * @v dest_size		Number of elements in destination big integer
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_grow_raw ( const uint32_t *source0, unsigned int source_size,
+		  uint32_t *dest0, unsigned int dest_size ) {
+	unsigned int pad_size = ( dest_size - source_size );
+
+	memcpy ( dest0, source0, sizeof ( bigint_t ( source_size ) ) );
+	memset ( ( dest0 + source_size ), 0, sizeof ( bigint_t ( pad_size ) ) );
+}
+
+/**
+ * Shrink big integer
+ *
+ * @v source0		Element 0 of source big integer
+ * @v source_size	Number of elements in source big integer
+ * @v dest0		Element 0 of destination big integer
+ * @v dest_size		Number of elements in destination big integer
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_shrink_raw ( const uint32_t *source0, unsigned int source_size __unused,
+		    uint32_t *dest0, unsigned int dest_size ) {
+
+	memcpy ( dest0, source0, sizeof ( bigint_t ( dest_size ) ) );
+}
+
+/**
+ * Finalise big integer
+ *
+ * @v value0		Element 0 of big integer to finalise
+ * @v size		Number of elements
+ * @v out		Output buffer
+ * @v len		Length of output buffer
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_done_raw ( const uint32_t *value0, unsigned int size __unused,
+		  void *out, size_t len ) {
+	const uint8_t *value_byte = ( ( const void * ) value0 );
+	uint8_t *out_byte = ( out + len );
+
+	/* Copy raw data in reverse order */
+	while ( len-- )
+		*(--out_byte) = *(value_byte++);
+}
+
+extern void bigint_multiply_raw ( const uint32_t *multiplicand0,
+				  const uint32_t *multiplier0,
+				  uint32_t *value0, unsigned int size );
+
+#endif /* _BITS_BIGINT_H */

src/arch/arm32/include/bits/bitops.h

@@ -0,0 +1,100 @@
+#ifndef _BITS_BITOPS_H
+#define _BITS_BITOPS_H
+
+/** @file
+ *
+ * ARM bit operations
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+
+/**
+ * Test and set bit atomically
+ *
+ * @v bit		Bit to set
+ * @v bits		Bit field
+ * @ret old		Old value of bit (zero or non-zero)
+ */
+static inline __attribute__ (( always_inline )) int
+test_and_set_bit ( unsigned int bit, volatile void *bits ) {
+	unsigned int index = ( bit / 32 );
+	unsigned int offset = ( bit % 32 );
+	volatile uint32_t *dword = ( ( ( volatile uint32_t * ) bits ) + index );
+	uint32_t mask = ( 1UL << offset );
+	uint32_t old;
+	uint32_t new;
+	uint32_t flag;
+
+	__asm__ __volatile__ ( "\n1:\n\t"
+			       "ldrex %0, %3\n\t"
+			       "orr %1, %0, %4\n\t"
+			       "strex %2, %1, %3\n\t"
+			       "tst %2, %2\n\t"
+			       "bne 1b\n\t"
+			       : "=&r" ( old ), "=&r" ( new ), "=&l" ( flag ),
+				 "+Q" ( *dword )
+			       : "r" ( mask )
+			       : "cc" );
+
+	return ( old & mask );
+}
+
+/**
+ * Test and clear bit atomically
+ *
+ * @v bit		Bit to set
+ * @v bits		Bit field
+ * @ret old		Old value of bit (zero or non-zero)
+ */
+static inline __attribute__ (( always_inline )) int
+test_and_clear_bit ( unsigned int bit, volatile void *bits ) {
+	unsigned int index = ( bit / 32 );
+	unsigned int offset = ( bit % 32 );
+	volatile uint32_t *dword = ( ( ( volatile uint32_t * ) bits ) + index );
+	uint32_t mask = ( 1UL << offset );
+	uint32_t old;
+	uint32_t new;
+	uint32_t flag;
+
+	__asm__ __volatile__ ( "\n1:\n\t"
+			       "ldrex %0, %3\n\t"
+			       "bic %1, %0, %4\n\t"
+			       "strex %2, %1, %3\n\t"
+			       "tst %2, %2\n\t"
+			       "bne 1b\n\t"
+			       : "=&r" ( old ), "=&r" ( new ), "=&l" ( flag ),
+				 "+Q" ( *dword )
+			       : "r" ( mask )
+			       : "cc" );
+
+	return ( old & mask );
+}
+
+/**
+ * Set bit atomically
+ *
+ * @v bit		Bit to set
+ * @v bits		Bit field
+ */
+static inline __attribute__ (( always_inline )) void
+set_bit ( unsigned int bit, volatile void *bits ) {
+
+	test_and_set_bit ( bit, bits );
+}
+
+/**
+ * Clear bit atomically
+ *
+ * @v bit		Bit to set
+ * @v bits		Bit field
+ */
+static inline __attribute__ (( always_inline )) void
+clear_bit ( unsigned int bit, volatile void *bits ) {
+
+	test_and_clear_bit ( bit, bits );
+}
+
+#endif /* _BITS_BITOPS_H */

src/arch/arm32/include/bits/byteswap.h

@@ -0,0 +1,52 @@
+#ifndef _BITS_BYTESWAP_H
+#define _BITS_BYTESWAP_H
+
+/** @file
+ *
+ * Byte-order swapping functions
+ *
+ */
+
+#include <stdint.h>
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+static inline __attribute__ (( always_inline, const )) uint16_t
+__bswap_variable_16 ( uint16_t x ) {
+	__asm__ ( "rev16 %0, %1" : "=l" ( x ) : "l" ( x ) );
+	return x;
+}
+
+static inline __attribute__ (( always_inline )) void
+__bswap_16s ( uint16_t *x ) {
+	*x = __bswap_variable_16 ( *x );
+}
+
+static inline __attribute__ (( always_inline, const )) uint32_t
+__bswap_variable_32 ( uint32_t x ) {
+	__asm__ ( "rev %0, %1" : "=l" ( x ) : "l" ( x ) );
+	return x;
+}
+
+static inline __attribute__ (( always_inline )) void
+__bswap_32s ( uint32_t *x ) {
+	*x = __bswap_variable_32 ( *x );
+}
+
+static inline __attribute__ (( always_inline, const )) uint64_t
+__bswap_variable_64 ( uint64_t x ) {
+	uint32_t in_high = ( x >> 32 );
+	uint32_t in_low = ( x & 0xffffffffUL );
+	uint32_t out_high = __bswap_variable_32 ( in_low );
+	uint32_t out_low = __bswap_variable_32 ( in_high );
+
+	return ( ( ( ( uint64_t ) out_high ) << 32 ) |
+		 ( ( uint64_t ) out_low ) );
+}
+
+static inline __attribute__ (( always_inline )) void
+__bswap_64s ( uint64_t *x ) {
+	*x = __bswap_variable_64 ( *x );
+}
+
+#endif

src/arch/arm32/include/bits/compiler.h

@@ -0,0 +1,16 @@
+#ifndef _BITS_COMPILER_H
+#define _BITS_COMPILER_H
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/** Dummy relocation type */
+#define RELOC_TYPE_NONE R_ARM_NONE
+
+#ifndef ASSEMBLY
+
+#define __asmcall
+#define __libgcc
+
+#endif /* ASSEMBLY */
+
+#endif /*_BITS_COMPILER_H */

src/arch/arm32/include/bits/profile.h

@@ -0,0 +1,30 @@
+#ifndef _BITS_PROFILE_H
+#define _BITS_PROFILE_H
+
+/** @file
+ *
+ * Profiling
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+
+/**
+ * Get profiling timestamp
+ *
+ * @ret timestamp	Timestamp
+ */
+static inline __attribute__ (( always_inline )) uint64_t
+profile_timestamp ( void ) {
+	uint32_t cycles;
+
+	/* Read cycle counter */
+	__asm__ __volatile__ ( "mcr p15, 0, %1, c9, c12, 0\n\t"
+			       "mrc p15, 0, %0, c9, c13, 0\n\t"
+			       : "=r" ( cycles ) : "r" ( 1 ) );
+	return cycles;
+}
+
+#endif /* _BITS_PROFILE_H */

src/arch/arm32/include/bits/stdint.h

@@ -0,0 +1,23 @@
+#ifndef _BITS_STDINT_H
+#define _BITS_STDINT_H
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+typedef __SIZE_TYPE__		size_t;
+typedef signed long		ssize_t;
+typedef signed long		off_t;
+
+typedef unsigned char		uint8_t;
+typedef unsigned short		uint16_t;
+typedef unsigned int		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;
+
+typedef unsigned long		physaddr_t;
+typedef unsigned long		intptr_t;
+
+#endif /* _BITS_STDINT_H */

src/arch/arm32/include/bits/string.h

@@ -0,0 +1,60 @@
+#ifndef BITS_STRING_H
+#define BITS_STRING_H
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/** @file
+ *
+ * String functions
+ *
+ */
+
+/**
+ * Fill memory region
+ *
+ * @v dest		Destination region
+ * @v character		Fill character
+ * @v len		Length
+ * @ret dest		Destination region
+ */
+static inline __attribute__ (( always_inline )) void *
+memset ( void *dest, int character, size_t len ) {
+
+	/* Not yet optimised */
+	generic_memset ( dest, character, len );
+	return dest;
+}
+
+/**
+ * Copy memory region
+ *
+ * @v dest		Destination region
+ * @v src		Source region
+ * @v len		Length
+ * @ret dest		Destination region
+ */
+static inline __attribute__ (( always_inline )) void *
+memcpy ( void *dest, const void *src, size_t len ) {
+
+	/* Not yet optimised */
+	generic_memcpy ( dest, src, len );
+	return dest;
+}
+
+/**
+ * Copy (possibly overlapping) memory region
+ *
+ * @v dest		Destination region
+ * @v src		Source region
+ * @v len		Length
+ * @ret dest		Destination region
+ */
+static inline __attribute__ (( always_inline )) void *
+memmove ( void *dest, const void *src, size_t len ) {
+
+	/* Not yet optimised */
+	generic_memmove ( dest, src, len );
+	return dest;
+}
+
+#endif /* BITS_STRING_H */

src/arch/arm32/include/bits/strings.h

@@ -0,0 +1,85 @@
+#ifndef _BITS_STRINGS_H
+#define _BITS_STRINGS_H
+
+/** @file
+ *
+ * String functions
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/**
+ * Find first (i.e. least significant) set bit
+ *
+ * @v value		Value
+ * @ret lsb		Least significant bit set in value (LSB=1), or zero
+ */
+static inline __attribute__ (( always_inline )) int __ffsl ( long value ) {
+	unsigned long bits = value;
+	unsigned long lsb;
+	unsigned int lz;
+
+	/* Extract least significant set bit */
+	lsb = ( bits & -bits );
+
+	/* Count number of leading zeroes before LSB */
+	__asm__ ( "clz %0, %1" : "=r" ( lz ) : "r" ( lsb ) );
+
+	return ( 32 - lz );
+}
+
+/**
+ * Find first (i.e. least significant) set bit
+ *
+ * @v value		Value
+ * @ret lsb		Least significant bit set in value (LSB=1), or zero
+ */
+static inline __attribute__ (( always_inline )) int __ffsll ( long long value ){
+	unsigned long high = ( value >> 32 );
+	unsigned long low = ( value >> 0 );
+
+	if ( low ) {
+		return ( __ffsl ( low ) );
+	} else if ( high ) {
+		return ( 32 + __ffsl ( high ) );
+	} else {
+		return 0;
+	}
+}
+
+/**
+ * Find last (i.e. most significant) set bit
+ *
+ * @v value		Value
+ * @ret msb		Most significant bit set in value (LSB=1), or zero
+ */
+static inline __attribute__ (( always_inline )) int __flsl ( long value ) {
+	unsigned int lz;
+
+	/* Count number of leading zeroes */
+	__asm__ ( "clz %0, %1" : "=r" ( lz ) : "r" ( value ) );
+
+	return ( 32 - lz );
+}
+
+/**
+ * Find last (i.e. most significant) set bit
+ *
+ * @v value		Value
+ * @ret msb		Most significant bit set in value (LSB=1), or zero
+ */
+static inline __attribute__ (( always_inline )) int __flsll ( long long value ){
+	unsigned long high = ( value >> 32 );
+	unsigned long low = ( value >> 0 );
+
+	if ( high ) {
+		return ( 32 + __flsl ( high ) );
+	} else if ( low ) {
+		return ( __flsl ( low ) );
+	} else {
+		return 0;
+	}
+}
+
+#endif /* _BITS_STRINGS_H */

src/arch/arm32/include/bits/tcpip.h

@@ -0,0 +1,19 @@
+#ifndef _BITS_TCPIP_H
+#define _BITS_TCPIP_H
+
+/** @file
+ *
+ * Transport-network layer interface
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+static inline __attribute__ (( always_inline )) uint16_t
+tcpip_continue_chksum ( uint16_t partial, const void *data, size_t len ) {
+
+	/* Not yet optimised */
+	return generic_tcpip_continue_chksum ( partial, data, len );
+}
+
+#endif /* _BITS_TCPIP_H */

src/arch/arm32/include/efi/ipxe/dhcp_arch.h

@@ -0,0 +1,40 @@
+/*
+ * Copyright (C) 2015 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * 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 of the
+ * License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+#ifndef _DHCP_ARCH_H
+#define _DHCP_ARCH_H
+
+/** @file
+ *
+ * Architecture-specific DHCP options
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <ipxe/dhcp.h>
+
+#define DHCP_ARCH_CLIENT_ARCHITECTURE DHCP_CLIENT_ARCHITECTURE_ARM32
+
+#define DHCP_ARCH_CLIENT_NDI 1 /* UNDI */ , 3, 10 /* v3.10 */
+
+#endif

src/arch/arm32/include/gdbmach.h

@@ -0,0 +1,45 @@
+#ifndef GDBMACH_H
+#define GDBMACH_H
+
+/** @file
+ *
+ * GDB architecture specifics
+ *
+ * This file declares functions for manipulating the machine state and
+ * debugging context.
+ *
+ */
+
+#include <stdint.h>
+
+typedef unsigned long gdbreg_t;
+
+/* Register snapshot */
+enum {
+	/* Not yet implemented */
+	GDBMACH_NREGS,
+};
+
+#define GDBMACH_SIZEOF_REGS ( GDBMACH_NREGS * sizeof ( gdbreg_t ) )
+
+static inline void gdbmach_set_pc ( gdbreg_t *regs, gdbreg_t pc ) {
+	/* Not yet implemented */
+	( void ) regs;
+	( void ) pc;
+}
+
+static inline void gdbmach_set_single_step ( gdbreg_t *regs, int step ) {
+	/* Not yet implemented */
+	( void ) regs;
+	( void ) step;
+}
+
+static inline void gdbmach_breakpoint ( void ) {
+	/* Not yet implemented */
+}
+
+extern int gdbmach_set_breakpoint ( int type, unsigned long addr, size_t len,
+				    int enable );
+extern void gdbmach_init ( void );
+
+#endif /* GDBMACH_H */

src/arch/arm32/include/limits.h

@@ -0,0 +1,61 @@
+#ifndef LIMITS_H
+#define LIMITS_H	1
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/* Number of bits in a `char' */
+#define CHAR_BIT	8
+
+/* Minimum and maximum values a `signed char' can hold */
+#define SCHAR_MIN	(-128)
+#define SCHAR_MAX	127
+
+/* Maximum value an `unsigned char' can hold. (Minimum is 0.) */
+#define UCHAR_MAX	255
+
+/* Minimum and maximum values a `char' can hold */
+#define CHAR_MIN	SCHAR_MIN
+#define CHAR_MAX	SCHAR_MAX
+
+/* Minimum and maximum values a `signed short int' can hold */
+#define SHRT_MIN	(-32768)
+#define SHRT_MAX	32767
+
+/* Maximum value an `unsigned short' can hold. (Minimum is 0.) */
+#define USHRT_MAX	65535
+
+
+/* Minimum and maximum values a `signed int' can hold */
+#define INT_MIN		(-INT_MAX - 1)
+#define INT_MAX		2147483647
+
+/* Maximum value an `unsigned int' can hold. (Minimum is 0.) */
+#define UINT_MAX	4294967295U
+
+
+/* Minimum and maximum values a `signed int' can hold */
+#define INT_MAX		2147483647
+#define INT_MIN		(-INT_MAX - 1)
+
+
+/* Maximum value an `unsigned int' can hold. (Minimum is 0.) */
+#define UINT_MAX	4294967295U
+
+
+/* Minimum and maximum values a `signed long' can hold */
+#define LONG_MAX	2147483647
+#define LONG_MIN	(-LONG_MAX - 1L)
+
+/* Maximum value an `unsigned long' can hold. (Minimum is 0.) */
+#define ULONG_MAX	4294967295UL
+
+/* Minimum and maximum values a `signed long long' can hold */
+#define LLONG_MAX	9223372036854775807LL
+#define LLONG_MIN	(-LONG_MAX - 1LL)
+
+
+/* Maximum value an `unsigned long long' can hold. (Minimum is 0.) */
+#define ULLONG_MAX	18446744073709551615ULL
+
+
+#endif /* LIMITS_H */

src/arch/arm32/include/setjmp.h

@@ -0,0 +1,38 @@
+#ifndef _SETJMP_H
+#define _SETJMP_H
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+
+/** A jump buffer */
+typedef struct {
+	/** Saved r4 */
+	uint32_t r4;
+	/** Saved r5 */
+	uint32_t r5;
+	/** Saved r6 */
+	uint32_t r6;
+	/** Saved r7 */
+	uint32_t r7;
+	/** Saved r8 */
+	uint32_t r8;
+	/** Saved r9 */
+	uint32_t r9;
+	/** Saved r10 */
+	uint32_t r10;
+	/** Saved frame pointer (r11) */
+	uint32_t fp;
+	/** Saved stack pointer (r13) */
+	uint32_t sp;
+	/** Saved link register (r14) */
+	uint32_t lr;
+} jmp_buf[1];
+
+extern int __asmcall __attribute__ (( returns_twice ))
+setjmp ( jmp_buf env );
+
+extern void __asmcall __attribute__ (( noreturn ))
+longjmp ( jmp_buf env, int val );
+
+#endif /* _SETJMP_H */

src/arch/arm32/libgcc/lldivmod.S

@@ -0,0 +1,50 @@
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL )
+
+	.text
+	.thumb
+
+/**
+ * Unsigned long long division
+ *
+ * @v r1:r0		Dividend
+ * @v r3:r2		Divisor
+ * @ret r1:r0		Quotient
+ * @ret r3:r2		Remainder
+ */
+	.section ".text.__aeabi_uldivmod", "ax", %progbits
+	.globl	__aeabi_uldivmod
+	.type	__aeabi_uldivmod, %function
+__aeabi_uldivmod:
+	/* Allocate stack space for remainder and pointer to remainder */
+	push	{r0, r1, r2, r3, r4, lr}
+	/* Call __udivmoddi4() */
+	add	r4, sp, #8
+	str	r4, [sp]
+	bl	__udivmoddi4
+	/* Retrieve remainder and return */
+	add	sp, sp, #8
+	pop	{r2, r3, r4, pc}
+	.size	__aeabi_uldivmod, . - __aeabi_uldivmod
+
+/**
+ * Signed long long division
+ *
+ * @v r1:r0		Dividend
+ * @v r3:r2		Divisor
+ * @ret r1:r0		Quotient
+ * @ret r3:r2		Remainder
+ */
+	.section ".text.__aeabi_ldivmod", "ax", %progbits
+	.globl	__aeabi_ldivmod
+	.type	__aeabi_ldivmod, %function
+__aeabi_ldivmod:
+	/* Allocate stack space for remainder and pointer to remainder */
+	push	{r0, r1, r2, r3, r4, lr}
+	/* Call __divmoddi4() */
+	add	r4, sp, #8
+	str	r4, [sp]
+	bl	__divmoddi4
+	/* Retrieve remainder and return */
+	add	sp, sp, #8
+	pop	{r2, r3, r4, pc}
+	.size	__aeabi_ldivmod, . - __aeabi_ldivmod

src/arch/arm32/libgcc/llshift.S

@@ -0,0 +1,88 @@
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL )
+
+	.text
+	.arm
+
+/**
+ * Logical shift left
+ *
+ * @v r1:r0		Value to shift
+ * @v r2		Shift amount
+ * @ret r1:r0		Shifted value
+ */
+	.section ".text.__aeabi_llsl", "ax", %progbits
+	.globl	__aeabi_llsl
+	.type	__aeabi_llsl, %function
+__aeabi_llsl:
+	/* r3 = ( shift - 32 ) */
+	subs	r3, r2, #32
+	/* If shift >= 32, then
+	 *   high = ( low << ( shift - 32 ) )
+	 */
+	movpl	r1, r0, lsl r3
+	/* If shift < 32, then
+	 *   high = ( ( high << shift ) | ( low >> ( 32 - shift ) ) )
+	 */
+	movmi	r1, r1, lsl r2
+	rsbmi	r3, r2, #32
+	orrmi	r1, r1, r0, lsr r3
+	/* low = ( low << shift ) */
+	mov	r0, r0, lsl r2
+	bx	lr
+	.size	__aeabi_llsl, . - __aeabi_llsl
+
+/**
+ * Logical shift right
+ *
+ * @v r1:r0		Value to shift
+ * @v r2		Shift amount
+ * @ret r1:r0		Shifted value
+ */
+	.section ".text.__aeabi_llsr", "ax", %progbits
+	.globl	__aeabi_llsr
+	.type	__aeabi_llsr, %function
+__aeabi_llsr:
+	/* r3 = ( shift - 32 ) */
+	subs	r3, r2, #32
+	/* If shift >= 32, then
+	 *   low = ( high >> ( shift - 32 ) )
+	 */
+	movpl	r0, r1, lsr r3
+	/* If shift < 32, then
+	 *   low = ( ( low >> shift ) | ( high << ( 32 - shift ) ) )
+	 */
+	movmi	r0, r0, lsr r2
+	rsbmi	r3, r2, #32
+	orrmi	r0, r0, r1, lsl r3
+	/* high = ( high >> shift ) */
+	mov	r1, r1, lsr r2
+	bx	lr
+	.size	__aeabi_llsr, . - __aeabi_llsr
+
+/**
+ * Arithmetic shift right
+ *
+ * @v r1:r0		Value to shift
+ * @v r2		Shift amount
+ * @ret r1:r0		Shifted value
+ */
+	.section ".text.__aeabi_lasr", "ax", %progbits
+	.globl	__aeabi_lasr
+	.type	__aeabi_lasr, %function
+__aeabi_lasr:
+	/* r3 = ( shift - 32 ) */
+	subs	r3, r2, #32
+	/* If shift >= 32, then
+	 *   low = ( high >> ( shift - 32 ) )
+	 */
+	movpl	r0, r1, asr r3
+	/* If shift < 32, then
+	 *   low = ( ( low >> shift ) | ( high << ( 32 - shift ) ) )
+	 */
+	movmi	r0, r0, lsr r2
+	rsbmi	r3, r2, #32
+	orrmi	r0, r0, r1, lsl r3
+	/* high = ( high >> shift ) */
+	mov	r1, r1, asr r2
+	bx	lr
+	.size	__aeabi_lasr, . - __aeabi_lasr

src/arch/arm64/Makefile

@@ -0,0 +1,33 @@
+# ARM64-specific directories containing source files
+#
+SRCDIRS		+= arch/arm64/core
+
+# ARM64-specific flags
+#
+CFLAGS		+= -mlittle-endian -mcmodel=small
+CFLAGS		+= -fomit-frame-pointer
+ASFLAGS		+= -mabi=lp64 -EL
+
+# We want to specify the LP64 model.  There is an explicit -mabi=lp64
+# on GCC 4.9 and later, and no guarantee as to which is the default
+# model.  In earlier versions of GCC, there is no -mabi option and the
+# default appears to be LP64 anyway.
+#
+ifeq ($(CCTYPE),gcc)
+LP64_TEST = $(CC) -mabi=lp64 -x c -c /dev/null -o /dev/null >/dev/null 2>&1
+LP64_FLAGS := $(shell $(LP64_TEST) && $(ECHO) '-mabi=lp64')
+WORKAROUND_CFLAGS += $(LP64_FLAGS)
+endif
+
+# EFI requires -fshort-wchar, and nothing else currently uses wchar_t
+#
+CFLAGS		+= -fshort-wchar
+
+# Include common ARM Makefile
+MAKEDEPS	+= arch/arm/Makefile
+include arch/arm/Makefile
+
+# Include platform-specific Makefile
+#
+MAKEDEPS	+= arch/arm64/Makefile.$(PLATFORM)
+include arch/arm64/Makefile.$(PLATFORM)

src/arch/arm64/Makefile.efi

@@ -0,0 +1,14 @@
+# -*- makefile -*- : Force emacs to use Makefile mode
+
+# Specify EFI image builder
+#
+ELF2EFI		= $(ELF2EFI64)
+
+# Specify EFI boot file
+#
+EFI_BOOT_FILE	= bootaa64.efi
+
+# Include generic EFI Makefile
+#
+MAKEDEPS	+= arch/arm/Makefile.efi
+include arch/arm/Makefile.efi

src/arch/arm64/core/arm64_bigint.c

@@ -0,0 +1,103 @@
+/*
+ * Copyright (C) 2016 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * 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 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+#include <string.h>
+#include <ipxe/bigint.h>
+
+/** @file
+ *
+ * Big integer support
+ */
+
+/**
+ * Multiply big integers
+ *
+ * @v multiplicand0	Element 0 of big integer to be multiplied
+ * @v multiplier0	Element 0 of big integer to be multiplied
+ * @v result0		Element 0 of big integer to hold result
+ * @v size		Number of elements
+ */
+void bigint_multiply_raw ( const uint64_t *multiplicand0,
+			   const uint64_t *multiplier0,
+			   uint64_t *result0, unsigned int size ) {
+	const bigint_t ( size ) __attribute__ (( may_alias )) *multiplicand =
+		( ( const void * ) multiplicand0 );
+	const bigint_t ( size ) __attribute__ (( may_alias )) *multiplier =
+		( ( const void * ) multiplier0 );
+	bigint_t ( size * 2 ) __attribute__ (( may_alias )) *result =
+		( ( void * ) result0 );
+	unsigned int i;
+	unsigned int j;
+	uint64_t multiplicand_element;
+	uint64_t multiplier_element;
+	uint64_t *result_elements;
+	uint64_t discard_low;
+	uint64_t discard_high;
+	uint64_t discard_temp_low;
+	uint64_t discard_temp_high;
+
+	/* Zero result */
+	memset ( result, 0, sizeof ( *result ) );
+
+	/* Multiply integers one element at a time */
+	for ( i = 0 ; i < size ; i++ ) {
+		multiplicand_element = multiplicand->element[i];
+		for ( j = 0 ; j < size ; j++ ) {
+			multiplier_element = multiplier->element[j];
+			result_elements = &result->element[ i + j ];
+			/* Perform a single multiply, and add the
+			 * resulting double-element into the result,
+			 * carrying as necessary.  The carry can
+			 * never overflow beyond the end of the
+			 * result, since:
+			 *
+			 *     a < 2^{n}, b < 2^{n} => ab < 2^{2n}
+			 */
+			__asm__ __volatile__ ( "mul %1, %6, %7\n\t"
+					       "umulh %2, %6, %7\n\t"
+					       "ldp %3, %4, [%0]\n\t"
+					       "adds %3, %3, %1\n\t"
+					       "adcs %4, %4, %2\n\t"
+					       "stp %3, %4, [%0], #16\n\t"
+					       "bcc 2f\n\t"
+					       "\n1:\n\t"
+					       "ldr %3, [%0]\n\t"
+					       "adcs %3, %3, xzr\n\t"
+					       "str %3, [%0], #8\n\t"
+					       "bcs 1b\n\t"
+					       "\n2:\n\t"
+					       : "+r" ( result_elements ),
+						 "=&r" ( discard_low ),
+						 "=&r" ( discard_high ),
+						 "=r" ( discard_temp_low ),
+						 "=r" ( discard_temp_high ),
+						 "+m" ( *result )
+					       : "r" ( multiplicand_element ),
+						 "r" ( multiplier_element )
+					       : "cc" );
+		}
+	}
+}

src/arch/arm64/core/arm64_string.c

@@ -0,0 +1,249 @@
+/*
+ * Copyright (C) 2016 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * 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 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+/** @file
+ *
+ * Optimised string operations
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <string.h>
+
+/**
+ * Copy memory area
+ *
+ * @v dest		Destination address
+ * @v src		Source address
+ * @v len		Length
+ * @ret dest		Destination address
+ */
+void arm64_memcpy ( void *dest, const void *src, size_t len ) {
+	void *discard_dest;
+	void *discard_end;
+	const void *discard_src;
+	size_t discard_offset;
+	unsigned long discard_data;
+	unsigned long discard_low;
+	unsigned long discard_high;
+
+	/* If length is too short for an "ldp"/"stp" instruction pair,
+	 * then just copy individual bytes.
+	 */
+	if ( len < 16 ) {
+		__asm__ __volatile__ ( "cbz %0, 2f\n\t"
+				       "\n1:\n\t"
+				       "sub %0, %0, #1\n\t"
+				       "ldrb %w1, [%3, %0]\n\t"
+				       "strb %w1, [%2, %0]\n\t"
+				       "cbnz %0, 1b\n\t"
+				       "\n2:\n\t"
+				       : "=&r" ( discard_offset ),
+					 "=&r" ( discard_data )
+				       : "r" ( dest ), "r" ( src ), "0" ( len )
+				       : "memory" );
+		return;
+	}
+
+	/* Use "ldp"/"stp" to copy 16 bytes at a time: one initial
+	 * potentially unaligned access, multiple destination-aligned
+	 * accesses, one final potentially unaligned access.
+	 */
+	__asm__ __volatile__ ( "ldp %3, %4, [%1], #16\n\t"
+			       "stp %3, %4, [%0], #16\n\t"
+			       "and %3, %0, #15\n\t"
+			       "sub %0, %0, %3\n\t"
+			       "sub %1, %1, %3\n\t"
+			       "bic %2, %5, #15\n\t"
+			       "b 2f\n\t"
+			       "\n1:\n\t"
+			       "ldp %3, %4, [%1], #16\n\t"
+			       "stp %3, %4, [%0], #16\n\t"
+			       "\n2:\n\t"
+			       "cmp %0, %2\n\t"
+			       "bne 1b\n\t"
+			       "ldp %3, %4, [%6, #-16]\n\t"
+			       "stp %3, %4, [%5, #-16]\n\t"
+			       : "=&r" ( discard_dest ),
+				 "=&r" ( discard_src ),
+				 "=&r" ( discard_end ),
+				 "=&r" ( discard_low ),
+				 "=&r" ( discard_high )
+			       : "r" ( dest + len ), "r" ( src + len ),
+				 "0" ( dest ), "1" ( src )
+			       : "memory", "cc" );
+}
+
+/**
+ * Zero memory region
+ *
+ * @v dest		Destination region
+ * @v len		Length
+ */
+void arm64_bzero ( void *dest, size_t len ) {
+	size_t discard_offset;
+	void *discard_dest;
+	void *discard_end;
+
+	/* If length is too short for an "stp" instruction, then just
+	 * zero individual bytes.
+	 */
+	if ( len < 16 ) {
+		__asm__ __volatile__ ( "cbz %0, 2f\n\t"
+				       "\n1:\n\t"
+				       "sub %0, %0, #1\n\t"
+				       "strb wzr, [%1, %0]\n\t"
+				       "cbnz %0, 1b\n\t"
+				       "\n2:\n\t"
+				       : "=&r" ( discard_offset )
+				       : "r" ( dest ), "0" ( len )
+				       : "memory" );
+		return;
+	}
+
+	/* Use "stp" to zero 16 bytes at a time: one initial
+	 * potentially unaligned access, multiple aligned accesses,
+	 * one final potentially unaligned access.
+	 */
+	__asm__ __volatile__ ( "stp xzr, xzr, [%0], #16\n\t"
+			       "bic %0, %0, #15\n\t"
+			       "bic %1, %2, #15\n\t"
+			       "b 2f\n\t"
+			       "\n1:\n\t"
+			       "stp xzr, xzr, [%0], #16\n\t"
+			       "\n2:\n\t"
+			       "cmp %0, %1\n\t"
+			       "bne 1b\n\t"
+			       "stp xzr, xzr, [%2, #-16]\n\t"
+			       : "=&r" ( discard_dest ),
+				 "=&r" ( discard_end )
+			       : "r" ( dest + len ), "0" ( dest )
+			       : "memory", "cc" );
+}
+
+/**
+ * Fill memory region
+ *
+ * @v dest		Destination region
+ * @v len		Length
+ * @v character		Fill character
+ *
+ * The unusual parameter order is to allow for more efficient
+ * tail-calling to arm64_memset() when zeroing a region.
+ */
+void arm64_memset ( void *dest, size_t len, int character ) {
+	size_t discard_offset;
+
+	/* Use optimised zeroing code if applicable */
+	if ( character == 0 ) {
+		arm64_bzero ( dest, len );
+		return;
+	}
+
+	/* Fill one byte at a time.  Calling memset() with a non-zero
+	 * value is relatively rare and unlikely to be
+	 * performance-critical.
+	 */
+	__asm__ __volatile__ ( "cbz %0, 2f\n\t"
+			       "\n1:\n\t"
+			       "sub %0, %0, #1\n\t"
+			       "strb %w2, [%1, %0]\n\t"
+			       "cbnz %0, 1b\n\t"
+			       "\n2:\n\t"
+			       : "=&r" ( discard_offset )
+			       : "r" ( dest ), "r" ( character ), "0" ( len )
+			       : "memory" );
+}
+
+/**
+ * Copy (possibly overlapping) memory region forwards
+ *
+ * @v dest		Destination region
+ * @v src		Source region
+ * @v len		Length
+ */
+void arm64_memmove_forwards ( void *dest, const void *src, size_t len ) {
+	void *discard_dest;
+	const void *discard_src;
+	unsigned long discard_data;
+
+	/* Assume memmove() is not performance-critical, and perform a
+	 * bytewise copy for simplicity.
+	 */
+	__asm__ __volatile__ ( "b 2f\n\t"
+			       "\n1:\n\t"
+			       "ldrb %w2, [%1], #1\n\t"
+			       "strb %w2, [%0], #1\n\t"
+			       "\n2:\n\t"
+			       "cmp %0, %3\n\t"
+			       "bne 1b\n\t"
+			       : "=&r" ( discard_dest ),
+				 "=&r" ( discard_src ),
+				 "=&r" ( discard_data )
+			       : "r" ( dest + len ), "0" ( dest ), "1" ( src )
+			       : "memory" );
+}
+
+/**
+ * Copy (possibly overlapping) memory region backwards
+ *
+ * @v dest		Destination region
+ * @v src		Source region
+ * @v len		Length
+ */
+void arm64_memmove_backwards ( void *dest, const void *src, size_t len ) {
+	size_t discard_offset;
+	unsigned long discard_data;
+
+	/* Assume memmove() is not performance-critical, and perform a
+	 * bytewise copy for simplicity.
+	 */
+	__asm__ __volatile__ ( "cbz %0, 2f\n\t"
+			       "\n1:\n\t"
+			       "sub %0, %0, #1\n\t"
+			       "ldrb %w1, [%3, %0]\n\t"
+			       "strb %w1, [%2, %0]\n\t"
+			       "cbnz %0, 1b\n\t"
+			       "\n2:\n\t"
+			       : "=&r" ( discard_offset ),
+				 "=&r" ( discard_data )
+			       : "r" ( dest ), "r" ( src ), "0" ( len )
+			       : "memory" );
+}
+
+/**
+ * Copy (possibly overlapping) memory region
+ *
+ * @v dest		Destination region
+ * @v src		Source region
+ * @v len		Length
+ */
+void arm64_memmove ( void *dest, const void *src, size_t len ) {
+
+	if ( dest <= src ) {
+		arm64_memmove_forwards ( dest, src, len );
+	} else {
+		arm64_memmove_backwards ( dest, src, len );
+	}
+}

src/arch/arm64/core/arm64_tcpip.c

@@ -0,0 +1,175 @@
+/*
+ * Copyright (C) 2016 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * 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 of the
+ * License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/** @file
+ *
+ * TCP/IP checksum
+ *
+ */
+
+#include <strings.h>
+#include <ipxe/tcpip.h>
+
+/** Alignment used by main checksumming loop */
+#define TCPIP_CHKSUM_ALIGN 16
+
+/** Number of steps in each iteration of the unrolled main checksumming loop */
+#define TCPIP_CHKSUM_UNROLL 4
+
+/**
+ * Calculate continued TCP/IP checkum
+ *
+ * @v sum		Checksum of already-summed data, in network byte order
+ * @v data		Data buffer
+ * @v len		Length of data buffer
+ * @ret sum		Updated checksum, in network byte order
+ */
+uint16_t tcpip_continue_chksum ( uint16_t sum, const void *data,
+				 size_t len ) {
+	intptr_t start;
+	intptr_t end;
+	intptr_t mid;
+	unsigned int pre;
+	unsigned int post;
+	unsigned int first;
+	uint64_t discard_low;
+	uint64_t discard_high;
+
+	/* Avoid potentially undefined shift operation */
+	if ( len == 0 )
+		return sum;
+
+	/* Find maximally-aligned midpoint.  For short blocks of data,
+	 * this may be aligned to fewer than 16 bytes.
+	 */
+	start = ( ( intptr_t ) data );
+	end = ( start + len );
+	mid = ( end &
+		~( ( ~( 1UL << 63 ) ) >> ( 64 - flsl ( start ^ end ) ) ) );
+
+	/* Calculate pre- and post-alignment lengths */
+	pre = ( ( mid - start ) & ( TCPIP_CHKSUM_ALIGN - 1 ) );
+	post = ( ( end - mid ) & ( TCPIP_CHKSUM_ALIGN - 1 ) );
+
+	/* Calculate number of steps in first iteration of unrolled loop */
+	first = ( ( ( len - pre - post ) / TCPIP_CHKSUM_ALIGN ) &
+		  ( TCPIP_CHKSUM_UNROLL - 1 ) );
+
+	/* Calculate checksum */
+	__asm__ ( /* Invert sum */
+		  "eor %w0, %w0, #0xffff\n\t"
+		  /* Clear carry flag */
+		  "cmn xzr, xzr\n\t"
+		  /* Byteswap and sum pre-alignment byte, if applicable */
+		  "tbz %w4, #0, 1f\n\t"
+		  "ldrb %w2, [%1], #1\n\t"
+		  "rev16 %w0, %w0\n\t"
+		  "rev16 %w2, %w2\n\t"
+		  "adcs %0, %0, %2\n\t"
+		  "\n1:\n\t"
+		  /* Sum pre-alignment halfword, if applicable */
+		  "tbz %w4, #1, 1f\n\t"
+		  "ldrh %w2, [%1], #2\n\t"
+		  "adcs %0, %0, %2\n\t"
+		  "\n1:\n\t"
+		  /* Sum pre-alignment word, if applicable */
+		  "tbz %w4, #2, 1f\n\t"
+		  "ldr %w2, [%1], #4\n\t"
+		  "adcs %0, %0, %2\n\t"
+		  "\n1:\n\t"
+		  /* Sum pre-alignment doubleword, if applicable */
+		  "tbz %w4, #3, 1f\n\t"
+		  "ldr %2, [%1], #8\n\t"
+		  "adcs %0, %0, %2\n\t"
+		  "\n1:\n\t"
+		  /* Jump into unrolled (x4) main loop */
+		  "adr %2, 2f\n\t"
+		  "sub %2, %2, %5, lsl #3\n\t"
+		  "sub %2, %2, %5, lsl #2\n\t"
+		  "br %2\n\t"
+		  "\n1:\n\t"
+		  "ldp %2, %3, [%1], #16\n\t"
+		  "adcs %0, %0, %2\n\t"
+		  "adcs %0, %0, %3\n\t"
+		  "ldp %2, %3, [%1], #16\n\t"
+		  "adcs %0, %0, %2\n\t"
+		  "adcs %0, %0, %3\n\t"
+		  "ldp %2, %3, [%1], #16\n\t"
+		  "adcs %0, %0, %2\n\t"
+		  "adcs %0, %0, %3\n\t"
+		  "ldp %2, %3, [%1], #16\n\t"
+		  "adcs %0, %0, %2\n\t"
+		  "adcs %0, %0, %3\n\t"
+		  "\n2:\n\t"
+		  "sub %2, %1, %6\n\t"
+		  "cbnz %2, 1b\n\t"
+		  /* Sum post-alignment doubleword, if applicable */
+		  "tbz %w7, #3, 1f\n\t"
+		  "ldr %2, [%1], #8\n\t"
+		  "adcs %0, %0, %2\n\t"
+		  "\n1:\n\t"
+		  /* Sum post-alignment word, if applicable */
+		  "tbz %w7, #2, 1f\n\t"
+		  "ldr %w2, [%1], #4\n\t"
+		  "adcs %0, %0, %2\n\t"
+		  "\n1:\n\t"
+		  /* Sum post-alignment halfword, if applicable */
+		  "tbz %w7, #1, 1f\n\t"
+		  "ldrh %w2, [%1], #2\n\t"
+		  "adcs %0, %0, %2\n\t"
+		  "\n1:\n\t"
+		  /* Sum post-alignment byte, if applicable */
+		  "tbz %w7, #0, 1f\n\t"
+		  "ldrb %w2, [%1], #1\n\t"
+		  "adcs %0, %0, %2\n\t"
+		  "\n1:\n\t"
+		  /* Fold down to a uint32_t plus carry flag */
+		  "lsr %2, %0, #32\n\t"
+		  "adcs %w0, %w0, %w2\n\t"
+		  /* Fold down to a uint16_t plus carry in bit 16 */
+		  "ubfm %2, %0, #0, #15\n\t"
+		  "ubfm %3, %0, #16, #31\n\t"
+		  "adc %w0, %w2, %w3\n\t"
+		  /* Fold down to a uint16_t */
+		  "tbz %w0, #16, 1f\n\t"
+		  "mov %w2, #0xffff\n\t"
+		  "sub %w0, %w0, %w2\n\t"
+		  "tbz %w0, #16, 1f\n\t"
+		  "sub %w0, %w0, %w2\n\t"
+		  "\n1:\n\t"
+		  /* Byteswap back, if applicable */
+		  "tbz %w4, #0, 1f\n\t"
+		  "rev16 %w0, %w0\n\t"
+		  "\n1:\n\t"
+		  /* Invert sum */
+		  "eor %w0, %w0, #0xffff\n\t"
+		  : "+r" ( sum ), "+r" ( data ), "=&r" ( discard_low ),
+		    "=&r" ( discard_high )
+		  : "r" ( pre ), "r" ( first ), "r" ( end - post ),
+		    "r" ( post )
+		  : "cc" );
+
+	return sum;
+}

src/arch/arm64/core/setjmp.S

@@ -0,0 +1,56 @@
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL )
+
+	.text
+
+	/* Must match jmp_buf structure layout */
+	.struct	0
+env_x19_x20:	.quad	0, 0
+env_x21_x22:	.quad	0, 0
+env_x23_x24:	.quad	0, 0
+env_x25_x26:	.quad	0, 0
+env_x27_x28:	.quad	0, 0
+env_x29_x30:	.quad	0, 0
+env_sp:		.quad	0
+	.previous
+
+/*
+ * Save stack context for non-local goto
+ */
+	.globl	setjmp
+	.type	setjmp, %function
+setjmp:
+	/* Store registers */
+	stp	x19, x20, [x0, #env_x19_x20]
+	stp	x21, x22, [x0, #env_x21_x22]
+	stp	x23, x24, [x0, #env_x23_x24]
+	stp	x25, x26, [x0, #env_x25_x26]
+	stp	x27, x28, [x0, #env_x27_x28]
+	stp	x29, x30, [x0, #env_x29_x30]
+	mov	x16, sp
+	str	x16, [x0, #env_sp]
+	/* Return 0 when returning as setjmp() */
+	mov	x0, #0
+	ret
+	.size	setjmp, . - setjmp
+
+/*
+ * Non-local jump to a saved stack context
+ */
+	.globl	longjmp
+	.type	longjmp, %function
+longjmp:
+	/* Restore registers */
+	ldp	x19, x20, [x0, #env_x19_x20]
+	ldp	x21, x22, [x0, #env_x21_x22]
+	ldp	x23, x24, [x0, #env_x23_x24]
+	ldp	x25, x26, [x0, #env_x25_x26]
+	ldp	x27, x28, [x0, #env_x27_x28]
+	ldp	x29, x30, [x0, #env_x29_x30]
+	ldr	x16, [x0, #env_sp]
+	mov	sp, x16
+	/* Force result to non-zero */
+	cmp	w1, #0
+	csinc	w0, w1, w1, ne
+	/* Return to setjmp() caller */
+	br	x30
+	.size	longjmp, . - longjmp

src/arch/arm64/include/bits/bigint.h

@@ -0,0 +1,317 @@
+#ifndef _BITS_BIGINT_H
+#define _BITS_BIGINT_H
+
+/** @file
+ *
+ * Big integer support
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+#include <string.h>
+#include <strings.h>
+
+/** Element of a big integer */
+typedef uint64_t bigint_element_t;
+
+/**
+ * Initialise big integer
+ *
+ * @v value0		Element 0 of big integer to initialise
+ * @v size		Number of elements
+ * @v data		Raw data
+ * @v len		Length of raw data
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_init_raw ( uint64_t *value0, unsigned int size,
+		  const void *data, size_t len ) {
+	size_t pad_len = ( sizeof ( bigint_t ( size ) ) - len );
+	uint8_t *value_byte = ( ( void * ) value0 );
+	const uint8_t *data_byte = ( data + len );
+
+	/* Copy raw data in reverse order, padding with zeros */
+	while ( len-- )
+		*(value_byte++) = *(--data_byte);
+	while ( pad_len-- )
+		*(value_byte++) = 0;
+}
+
+/**
+ * Add big integers
+ *
+ * @v addend0		Element 0 of big integer to add
+ * @v value0		Element 0 of big integer to be added to
+ * @v size		Number of elements
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_add_raw ( const uint64_t *addend0, uint64_t *value0,
+		 unsigned int size ) {
+	bigint_t ( size ) __attribute__ (( may_alias )) *value =
+		( ( void * ) value0 );
+	uint64_t *discard_addend;
+	uint64_t *discard_value;
+	uint64_t discard_addend_i;
+	uint64_t discard_value_i;
+	unsigned int discard_size;
+
+	__asm__ __volatile__ ( "cmn xzr, xzr\n\t" /* clear CF */
+			       "\n1:\n\t"
+			       "ldr %3, [%0], #8\n\t"
+			       "ldr %4, [%1]\n\t"
+			       "adcs %4, %4, %3\n\t"
+			       "str %4, [%1], #8\n\t"
+			       "sub %w2, %w2, #1\n\t"
+			       "cbnz %w2, 1b\n\t"
+			       : "=r" ( discard_addend ),
+				 "=r" ( discard_value ),
+				 "=r" ( discard_size ),
+				 "=r" ( discard_addend_i ),
+				 "=r" ( discard_value_i ),
+				 "+m" ( *value )
+			       : "0" ( addend0 ), "1" ( value0 ), "2" ( size )
+			       : "cc" );
+}
+
+/**
+ * Subtract big integers
+ *
+ * @v subtrahend0	Element 0 of big integer to subtract
+ * @v value0		Element 0 of big integer to be subtracted from
+ * @v size		Number of elements
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_subtract_raw ( const uint64_t *subtrahend0, uint64_t *value0,
+		      unsigned int size ) {
+	bigint_t ( size ) __attribute__ (( may_alias )) *value =
+		( ( void * ) value0 );
+	uint64_t *discard_subtrahend;
+	uint64_t *discard_value;
+	uint64_t discard_subtrahend_i;
+	uint64_t discard_value_i;
+	unsigned int discard_size;
+
+	__asm__ __volatile__ ( "cmp xzr, xzr\n\t" /* set CF */
+			       "\n1:\n\t"
+			       "ldr %3, [%0], #8\n\t"
+			       "ldr %4, [%1]\n\t"
+			       "sbcs %4, %4, %3\n\t"
+			       "str %4, [%1], #8\n\t"
+			       "sub %w2, %w2, #1\n\t"
+			       "cbnz %w2, 1b\n\t"
+			       : "=r" ( discard_subtrahend ),
+				 "=r" ( discard_value ),
+				 "=r" ( discard_size ),
+				 "=r" ( discard_subtrahend_i ),
+				 "=r" ( discard_value_i ),
+				 "+m" ( *value )
+			       : "0" ( subtrahend0 ), "1" ( value0 ),
+				 "2" ( size )
+			       : "cc" );
+}
+
+/**
+ * Rotate big integer left
+ *
+ * @v value0		Element 0 of big integer
+ * @v size		Number of elements
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_rol_raw ( uint64_t *value0, unsigned int size ) {
+	bigint_t ( size ) __attribute__ (( may_alias )) *value =
+		( ( void * ) value0 );
+	uint64_t *discard_value;
+	uint64_t discard_value_i;
+	unsigned int discard_size;
+
+	__asm__ __volatile__ ( "cmn xzr, xzr\n\t" /* clear CF */
+			       "\n1:\n\t"
+			       "ldr %2, [%0]\n\t"
+			       "adcs %2, %2, %2\n\t"
+			       "str %2, [%0], #8\n\t"
+			       "sub %w1, %w1, #1\n\t"
+			       "cbnz %w1, 1b\n\t"
+			       : "=r" ( discard_value ),
+				 "=r" ( discard_size ),
+				 "=r" ( discard_value_i ),
+				 "+m" ( *value )
+			       : "0" ( value0 ), "1" ( size )
+			       : "cc" );
+}
+
+/**
+ * Rotate big integer right
+ *
+ * @v value0		Element 0 of big integer
+ * @v size		Number of elements
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_ror_raw ( uint64_t *value0, unsigned int size ) {
+	bigint_t ( size ) __attribute__ (( may_alias )) *value =
+		( ( void * ) value0 );
+	uint64_t *discard_value;
+	uint64_t discard_value_i;
+	uint64_t discard_value_j;
+	unsigned int discard_size;
+
+	__asm__ __volatile__ ( "mov %3, #0\n\t"
+			       "\n1:\n\t"
+			       "sub %w1, %w1, #1\n\t"
+			       "ldr %2, [%0, %1, lsl #3]\n\t"
+			       "extr %3, %3, %2, #1\n\t"
+			       "str %3, [%0, %1, lsl #3]\n\t"
+			       "mov %3, %2\n\t"
+			       "cbnz %w1, 1b\n\t"
+			       : "=r" ( discard_value ),
+				 "=r" ( discard_size ),
+				 "=r" ( discard_value_i ),
+				 "=r" ( discard_value_j ),
+				 "+m" ( *value )
+			       : "0" ( value0 ), "1" ( size ) );
+}
+
+/**
+ * Test if big integer is equal to zero
+ *
+ * @v value0		Element 0 of big integer
+ * @v size		Number of elements
+ * @ret is_zero		Big integer is equal to zero
+ */
+static inline __attribute__ (( always_inline, pure )) int
+bigint_is_zero_raw ( const uint64_t *value0, unsigned int size ) {
+	const uint64_t *value = value0;
+	uint64_t value_i;
+
+	do {
+		value_i = *(value++);
+		if ( value_i )
+			break;
+	} while ( --size );
+
+	return ( value_i == 0 );
+}
+
+/**
+ * Compare big integers
+ *
+ * @v value0		Element 0 of big integer
+ * @v reference0	Element 0 of reference big integer
+ * @v size		Number of elements
+ * @ret geq		Big integer is greater than or equal to the reference
+ */
+static inline __attribute__ (( always_inline, pure )) int
+bigint_is_geq_raw ( const uint64_t *value0, const uint64_t *reference0,
+		    unsigned int size ) {
+	const uint64_t *value = ( value0 + size );
+	const uint64_t *reference = ( reference0 + size );
+	uint64_t value_i;
+	uint64_t reference_i;
+
+	do {
+		value_i = *(--value);
+		reference_i = *(--reference);
+		if ( value_i != reference_i )
+			break;
+	} while ( --size );
+
+	return ( value_i >= reference_i );
+}
+
+/**
+ * Test if bit is set in big integer
+ *
+ * @v value0		Element 0 of big integer
+ * @v size		Number of elements
+ * @v bit		Bit to test
+ * @ret is_set		Bit is set
+ */
+static inline __attribute__ (( always_inline )) int
+bigint_bit_is_set_raw ( const uint64_t *value0, unsigned int size,
+			unsigned int bit ) {
+	const bigint_t ( size ) __attribute__ (( may_alias )) *value =
+		( ( const void * ) value0 );
+	unsigned int index = ( bit / ( 8 * sizeof ( value->element[0] ) ) );
+	unsigned int subindex = ( bit % ( 8 * sizeof ( value->element[0] ) ) );
+
+	return ( !! ( value->element[index] & ( 1UL << subindex ) ) );
+}
+
+/**
+ * Find highest bit set in big integer
+ *
+ * @v value0		Element 0 of big integer
+ * @v size		Number of elements
+ * @ret max_bit		Highest bit set + 1 (or 0 if no bits set)
+ */
+static inline __attribute__ (( always_inline )) int
+bigint_max_set_bit_raw ( const uint64_t *value0, unsigned int size ) {
+	const uint64_t *value = ( value0 + size );
+	int max_bit = ( 8 * sizeof ( bigint_t ( size ) ) );
+	uint64_t value_i;
+
+	do {
+		value_i = *(--value);
+		max_bit -= ( 64 - fls ( value_i ) );
+		if ( value_i )
+			break;
+	} while ( --size );
+
+	return max_bit;
+}
+
+/**
+ * Grow big integer
+ *
+ * @v source0		Element 0 of source big integer
+ * @v source_size	Number of elements in source big integer
+ * @v dest0		Element 0 of destination big integer
+ * @v dest_size		Number of elements in destination big integer
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_grow_raw ( const uint64_t *source0, unsigned int source_size,
+		  uint64_t *dest0, unsigned int dest_size ) {
+	unsigned int pad_size = ( dest_size - source_size );
+
+	memcpy ( dest0, source0, sizeof ( bigint_t ( source_size ) ) );
+	memset ( ( dest0 + source_size ), 0, sizeof ( bigint_t ( pad_size ) ) );
+}
+
+/**
+ * Shrink big integer
+ *
+ * @v source0		Element 0 of source big integer
+ * @v source_size	Number of elements in source big integer
+ * @v dest0		Element 0 of destination big integer
+ * @v dest_size		Number of elements in destination big integer
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_shrink_raw ( const uint64_t *source0, unsigned int source_size __unused,
+		    uint64_t *dest0, unsigned int dest_size ) {
+
+	memcpy ( dest0, source0, sizeof ( bigint_t ( dest_size ) ) );
+}
+
+/**
+ * Finalise big integer
+ *
+ * @v value0		Element 0 of big integer to finalise
+ * @v size		Number of elements
+ * @v out		Output buffer
+ * @v len		Length of output buffer
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_done_raw ( const uint64_t *value0, unsigned int size __unused,
+		  void *out, size_t len ) {
+	const uint8_t *value_byte = ( ( const void * ) value0 );
+	uint8_t *out_byte = ( out + len );
+
+	/* Copy raw data in reverse order */
+	while ( len-- )
+		*(--out_byte) = *(value_byte++);
+}
+
+extern void bigint_multiply_raw ( const uint64_t *multiplicand0,
+				  const uint64_t *multiplier0,
+				  uint64_t *value0, unsigned int size );
+
+#endif /* _BITS_BIGINT_H */

src/arch/arm64/include/bits/bitops.h

@@ -0,0 +1,100 @@
+#ifndef _BITS_BITOPS_H
+#define _BITS_BITOPS_H
+
+/** @file
+ *
+ * ARM bit operations
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+
+/**
+ * Test and set bit atomically
+ *
+ * @v bit		Bit to set
+ * @v bits		Bit field
+ * @ret old		Old value of bit (zero or non-zero)
+ */
+static inline __attribute__ (( always_inline )) int
+test_and_set_bit ( unsigned int bit, volatile void *bits ) {
+	unsigned int index = ( bit / 64 );
+	unsigned int offset = ( bit % 64 );
+	volatile uint64_t *qword = ( ( ( volatile uint64_t * ) bits ) + index );
+	uint64_t mask = ( 1UL << offset );
+	uint64_t old;
+	uint64_t new;
+	uint32_t flag;
+
+	__asm__ __volatile__ ( "\n1:\n\t"
+			       "ldxr %0, %3\n\t"
+			       "orr %1, %0, %4\n\t"
+			       "stxr %w2, %1, %3\n\t"
+			       "tst %w2, %w2\n\t"
+			       "bne 1b\n\t"
+			       : "=&r" ( old ), "=&r" ( new ), "=&r" ( flag ),
+				 "+Q" ( *qword )
+			       : "r" ( mask )
+			       : "cc" );
+
+	return ( !! ( old & mask ) );
+}
+
+/**
+ * Test and clear bit atomically
+ *
+ * @v bit		Bit to set
+ * @v bits		Bit field
+ * @ret old		Old value of bit (zero or non-zero)
+ */
+static inline __attribute__ (( always_inline )) int
+test_and_clear_bit ( unsigned int bit, volatile void *bits ) {
+	unsigned int index = ( bit / 64 );
+	unsigned int offset = ( bit % 64 );
+	volatile uint64_t *qword = ( ( ( volatile uint64_t * ) bits ) + index );
+	uint64_t mask = ( 1UL << offset );
+	uint64_t old;
+	uint64_t new;
+	uint32_t flag;
+
+	__asm__ __volatile__ ( "\n1:\n\t"
+			       "ldxr %0, %3\n\t"
+			       "bic %1, %0, %4\n\t"
+			       "stxr %w2, %1, %3\n\t"
+			       "tst %w2, %w2\n\t"
+			       "bne 1b\n\t"
+			       : "=&r" ( old ), "=&r" ( new ), "=&r" ( flag ),
+				 "+Q" ( *qword )
+			       : "r" ( mask )
+			       : "cc" );
+
+	return ( !! ( old & mask ) );
+}
+
+/**
+ * Set bit atomically
+ *
+ * @v bit		Bit to set
+ * @v bits		Bit field
+ */
+static inline __attribute__ (( always_inline )) void
+set_bit ( unsigned int bit, volatile void *bits ) {
+
+	test_and_set_bit ( bit, bits );
+}
+
+/**
+ * Clear bit atomically
+ *
+ * @v bit		Bit to set
+ * @v bits		Bit field
+ */
+static inline __attribute__ (( always_inline )) void
+clear_bit ( unsigned int bit, volatile void *bits ) {
+
+	test_and_clear_bit ( bit, bits );
+}
+
+#endif /* _BITS_BITOPS_H */

src/arch/arm64/include/bits/byteswap.h

@@ -0,0 +1,47 @@
+#ifndef _BITS_BYTESWAP_H
+#define _BITS_BYTESWAP_H
+
+/** @file
+ *
+ * Byte-order swapping functions
+ *
+ */
+
+#include <stdint.h>
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+static inline __attribute__ (( always_inline, const )) uint16_t
+__bswap_variable_16 ( uint16_t x ) {
+	__asm__ ( "rev16 %0, %1" : "=r" ( x ) : "r" ( x ) );
+	return x;
+}
+
+static inline __attribute__ (( always_inline )) void
+__bswap_16s ( uint16_t *x ) {
+	*x = __bswap_variable_16 ( *x );
+}
+
+static inline __attribute__ (( always_inline, const )) uint32_t
+__bswap_variable_32 ( uint32_t x ) {
+	__asm__ ( "rev32 %0, %1" : "=r" ( x ) : "r" ( x ) );
+	return x;
+}
+
+static inline __attribute__ (( always_inline )) void
+__bswap_32s ( uint32_t *x ) {
+	*x = __bswap_variable_32 ( *x );
+}
+
+static inline __attribute__ (( always_inline, const )) uint64_t
+__bswap_variable_64 ( uint64_t x ) {
+	__asm__ ( "rev %0, %1" : "=r" ( x ) : "r" ( x ) );
+	return x;
+}
+
+static inline __attribute__ (( always_inline )) void
+__bswap_64s ( uint64_t *x ) {
+	*x = __bswap_variable_64 ( *x );
+}
+
+#endif

src/arch/arm64/include/bits/compiler.h

@@ -0,0 +1,16 @@
+#ifndef _BITS_COMPILER_H
+#define _BITS_COMPILER_H
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/** Dummy relocation type */
+#define RELOC_TYPE_NONE R_AARCH64_NULL
+
+#ifndef ASSEMBLY
+
+#define __asmcall
+#define __libgcc
+
+#endif /* ASSEMBLY */
+
+#endif /*_BITS_COMPILER_H */

src/arch/arm64/include/bits/profile.h

@@ -0,0 +1,28 @@
+#ifndef _BITS_PROFILE_H
+#define _BITS_PROFILE_H
+
+/** @file
+ *
+ * Profiling
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+
+/**
+ * Get profiling timestamp
+ *
+ * @ret timestamp	Timestamp
+ */
+static inline __attribute__ (( always_inline )) uint64_t
+profile_timestamp ( void ) {
+	uint64_t cycles;
+
+	/* Read cycle counter */
+	__asm__ __volatile__ ( "mrs %0, CNTVCT_EL0\n\t" : "=r" ( cycles ) );
+	return cycles;
+}
+
+#endif /* _BITS_PROFILE_H */

src/arch/arm64/include/bits/stdint.h

@@ -0,0 +1,21 @@
+#ifndef _BITS_STDINT_H
+#define _BITS_STDINT_H
+
+typedef __SIZE_TYPE__		size_t;
+typedef signed long		ssize_t;
+typedef signed long		off_t;
+
+typedef unsigned char		uint8_t;
+typedef unsigned short		uint16_t;
+typedef unsigned int		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;
+
+typedef unsigned long		physaddr_t;
+typedef unsigned long		intptr_t;
+
+#endif /* _BITS_STDINT_H */

src/arch/arm64/include/bits/string.h

@@ -0,0 +1,106 @@
+#ifndef BITS_STRING_H
+#define BITS_STRING_H
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/** @file
+ *
+ * String functions
+ *
+ */
+
+extern void arm64_bzero ( void *dest, size_t len );
+extern void arm64_memset ( void *dest, size_t len, int character );
+extern void arm64_memcpy ( void *dest, const void *src, size_t len );
+extern void arm64_memmove_forwards ( void *dest, const void *src, size_t len );
+extern void arm64_memmove_backwards ( void *dest, const void *src, size_t len );
+extern void arm64_memmove ( void *dest, const void *src, size_t len );
+
+/**
+ * Fill memory region
+ *
+ * @v dest		Destination region
+ * @v character		Fill character
+ * @v len		Length
+ * @ret dest		Destination region
+ */
+static inline __attribute__ (( always_inline )) void *
+memset ( void *dest, int character, size_t len ) {
+
+	/* Allow gcc to generate inline "stX xzr" instructions for
+	 * small, constant lengths.
+	 */
+	if ( __builtin_constant_p ( character ) && ( character == 0 ) &&
+	     __builtin_constant_p ( len ) && ( len <= 64 ) ) {
+		__builtin_memset ( dest, 0, len );
+		return dest;
+	}
+
+	/* For zeroing larger or non-constant lengths, use the
+	 * optimised variable-length zeroing code.
+	 */
+	if ( __builtin_constant_p ( character ) && ( character == 0 ) ) {
+		arm64_bzero ( dest, len );
+		return dest;
+	}
+
+	/* Not necessarily zeroing: use basic variable-length code */
+	arm64_memset ( dest, len, character );
+	return dest;
+}
+
+/**
+ * Copy memory region
+ *
+ * @v dest		Destination region
+ * @v src		Source region
+ * @v len		Length
+ * @ret dest		Destination region
+ */
+static inline __attribute__ (( always_inline )) void *
+memcpy ( void *dest, const void *src, size_t len ) {
+
+	/* Allow gcc to generate inline "ldX"/"stX" instructions for
+	 * small, constant lengths.
+	 */
+	if ( __builtin_constant_p ( len ) && ( len <= 64 ) ) {
+		__builtin_memcpy ( dest, src, len );
+		return dest;
+	}
+
+	/* Otherwise, use variable-length code */
+	arm64_memcpy ( dest, src, len );
+	return dest;
+}
+
+/**
+ * Copy (possibly overlapping) memory region
+ *
+ * @v dest		Destination region
+ * @v src		Source region
+ * @v len		Length
+ * @ret dest		Destination region
+ */
+static inline __attribute__ (( always_inline )) void *
+memmove ( void *dest, const void *src, size_t len ) {
+	ssize_t offset = ( dest - src );
+
+	/* If required direction of copy is known at build time, then
+	 * use the appropriate forwards/backwards copy directly.
+	 */
+	if ( __builtin_constant_p ( offset ) ) {
+		if ( offset <= 0 ) {
+			arm64_memmove_forwards ( dest, src, len );
+			return dest;
+		} else {
+			arm64_memmove_backwards ( dest, src, len );
+			return dest;
+		}
+	}
+
+	/* Otherwise, use ambidirectional copy */
+	arm64_memmove ( dest, src, len );
+	return dest;
+}
+
+#endif /* BITS_STRING_H */

src/arch/arm64/include/bits/strings.h

@@ -0,0 +1,69 @@
+#ifndef _BITS_STRINGS_H
+#define _BITS_STRINGS_H
+
+/** @file
+ *
+ * String functions
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/**
+ * Find first (i.e. least significant) set bit
+ *
+ * @v value		Value
+ * @ret lsb		Least significant bit set in value (LSB=1), or zero
+ */
+static inline __attribute__ (( always_inline )) int __ffsll ( long long value ){
+	unsigned long long bits = value;
+	unsigned long long lsb;
+	unsigned int lz;
+
+	/* Extract least significant set bit */
+	lsb = ( bits & -bits );
+
+	/* Count number of leading zeroes before LSB */
+	__asm__ ( "clz %0, %1" : "=r" ( lz ) : "r" ( lsb ) );
+
+	return ( 64 - lz );
+}
+
+/**
+ * Find first (i.e. least significant) set bit
+ *
+ * @v value		Value
+ * @ret lsb		Least significant bit set in value (LSB=1), or zero
+ */
+static inline __attribute__ (( always_inline )) int __ffsl ( long value ) {
+
+	return __ffsll ( value );
+}
+
+/**
+ * Find last (i.e. most significant) set bit
+ *
+ * @v value		Value
+ * @ret msb		Most significant bit set in value (LSB=1), or zero
+ */
+static inline __attribute__ (( always_inline )) int __flsll ( long long value ){
+	unsigned int lz;
+
+	/* Count number of leading zeroes */
+	__asm__ ( "clz %0, %1" : "=r" ( lz ) : "r" ( value ) );
+
+	return ( 64 - lz );
+}
+
+/**
+ * Find last (i.e. most significant) set bit
+ *
+ * @v value		Value
+ * @ret msb		Most significant bit set in value (LSB=1), or zero
+ */
+static inline __attribute__ (( always_inline )) int __flsl ( long value ) {
+
+	return __flsll ( value );
+}
+
+#endif /* _BITS_STRINGS_H */

src/arch/arm64/include/bits/tcpip.h

@@ -0,0 +1,15 @@
+#ifndef _BITS_TCPIP_H
+#define _BITS_TCPIP_H
+
+/** @file
+ *
+ * Transport-network layer interface
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+extern uint16_t tcpip_continue_chksum ( uint16_t sum, const void *data,
+					size_t len );
+
+#endif /* _BITS_TCPIP_H */

src/arch/arm64/include/efi/ipxe/dhcp_arch.h

@@ -0,0 +1,40 @@
+/*
+ * Copyright (C) 2015 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * 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 of the
+ * License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+#ifndef _DHCP_ARCH_H
+#define _DHCP_ARCH_H
+
+/** @file
+ *
+ * Architecture-specific DHCP options
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <ipxe/dhcp.h>
+
+#define DHCP_ARCH_CLIENT_ARCHITECTURE DHCP_CLIENT_ARCHITECTURE_ARM64
+
+#define DHCP_ARCH_CLIENT_NDI 1 /* UNDI */ , 3, 10 /* v3.10 */
+
+#endif

src/arch/arm64/include/gdbmach.h

@@ -0,0 +1,45 @@
+#ifndef GDBMACH_H
+#define GDBMACH_H
+
+/** @file
+ *
+ * GDB architecture specifics
+ *
+ * This file declares functions for manipulating the machine state and
+ * debugging context.
+ *
+ */
+
+#include <stdint.h>
+
+typedef unsigned long gdbreg_t;
+
+/* Register snapshot */
+enum {
+	/* Not yet implemented */
+	GDBMACH_NREGS,
+};
+
+#define GDBMACH_SIZEOF_REGS ( GDBMACH_NREGS * sizeof ( gdbreg_t ) )
+
+static inline void gdbmach_set_pc ( gdbreg_t *regs, gdbreg_t pc ) {
+	/* Not yet implemented */
+	( void ) regs;
+	( void ) pc;
+}
+
+static inline void gdbmach_set_single_step ( gdbreg_t *regs, int step ) {
+	/* Not yet implemented */
+	( void ) regs;
+	( void ) step;
+}
+
+static inline void gdbmach_breakpoint ( void ) {
+	/* Not yet implemented */
+}
+
+extern int gdbmach_set_breakpoint ( int type, unsigned long addr, size_t len,
+				    int enable );
+extern void gdbmach_init ( void );
+
+#endif /* GDBMACH_H */

src/arch/arm64/include/limits.h

@@ -0,0 +1,59 @@
+#ifndef LIMITS_H
+#define LIMITS_H	1
+
+/* Number of bits in a `char' */
+#define CHAR_BIT	8
+
+/* Minimum and maximum values a `signed char' can hold */
+#define SCHAR_MIN	(-128)
+#define SCHAR_MAX	127
+
+/* Maximum value an `unsigned char' can hold. (Minimum is 0.) */
+#define UCHAR_MAX	255
+
+/* Minimum and maximum values a `char' can hold */
+#define CHAR_MIN	SCHAR_MIN
+#define CHAR_MAX	SCHAR_MAX
+
+/* Minimum and maximum values a `signed short int' can hold */
+#define SHRT_MIN	(-32768)
+#define SHRT_MAX	32767
+
+/* Maximum value an `unsigned short' can hold. (Minimum is 0.) */
+#define USHRT_MAX	65535
+
+
+/* Minimum and maximum values a `signed int' can hold */
+#define INT_MIN		(-INT_MAX - 1)
+#define INT_MAX		2147483647
+
+/* Maximum value an `unsigned int' can hold. (Minimum is 0.) */
+#define UINT_MAX	4294967295U
+
+
+/* Minimum and maximum values a `signed int' can hold */
+#define INT_MAX		2147483647
+#define INT_MIN		(-INT_MAX - 1)
+
+
+/* Maximum value an `unsigned int' can hold. (Minimum is 0.) */
+#define UINT_MAX	4294967295U
+
+
+/* Minimum and maximum values a `signed long' can hold */
+#define LONG_MAX	9223372036854775807L
+#define LONG_MIN	(-LONG_MAX - 1L)
+
+/* Maximum value an `unsigned long' can hold. (Minimum is 0.) */
+#define ULONG_MAX	18446744073709551615UL
+
+/* Minimum and maximum values a `signed long long' can hold */
+#define LLONG_MAX	9223372036854775807LL
+#define LLONG_MIN	(-LONG_MAX - 1LL)
+
+
+/* Maximum value an `unsigned long long' can hold. (Minimum is 0.) */
+#define ULLONG_MAX	18446744073709551615ULL
+
+
+#endif /* LIMITS_H */

src/arch/arm64/include/setjmp.h

@@ -0,0 +1,44 @@
+#ifndef _SETJMP_H
+#define _SETJMP_H
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+
+/** A jump buffer */
+typedef struct {
+	/** Saved x19 */
+	uint64_t x19;
+	/** Saved x20 */
+	uint64_t x20;
+	/** Saved x21 */
+	uint64_t x21;
+	/** Saved x22 */
+	uint64_t x22;
+	/** Saved x23 */
+	uint64_t x23;
+	/** Saved x24 */
+	uint64_t x24;
+	/** Saved x25 */
+	uint64_t x25;
+	/** Saved x26 */
+	uint64_t x26;
+	/** Saved x27 */
+	uint64_t x27;
+	/** Saved x28 */
+	uint64_t x28;
+	/** Saved frame pointer (x29) */
+	uint64_t x29;
+	/** Saved link register (x30) */
+	uint64_t x30;
+	/** Saved stack pointer (x31) */
+	uint64_t sp;
+} jmp_buf[1];
+
+extern int __asmcall __attribute__ (( returns_twice ))
+setjmp ( jmp_buf env );
+
+extern void __asmcall __attribute__ (( noreturn ))
+longjmp ( jmp_buf env, int val );
+
+#endif /* _SETJMP_H */

src/arch/i386/Makefile

@@ -74,49 +74,21 @@ CFLAGS			+= -Ui386
 # recognise an option that starts with "no", so we have to test for
 # output on stderr instead of checking the exit status.
 #
+# Current versions of gcc require -no-pie; older versions require
+# -nopie.  We therefore test for both.
+#
 ifeq ($(CCTYPE),gcc)
-PIE_TEST = [ -z "`$(CC) -fno-PIE -nopie -x c -c /dev/null -o /dev/null 2>&1`" ]
-PIE_FLAGS := $(shell $(PIE_TEST) && $(ECHO) '-fno-PIE -nopie')
-WORKAROUND_CFLAGS += $(PIE_FLAGS)
+PIE_TEST = [ -z "`$(CC) -fno-PIE -no-pie -x c -c /dev/null -o /dev/null 2>&1`" ]
+PIE_FLAGS := $(shell $(PIE_TEST) && $(ECHO) '-fno-PIE -no-pie')
+PIE_TEST2 = [ -z "`$(CC) -fno-PIE -nopie -x c -c /dev/null -o /dev/null 2>&1`" ]
+PIE_FLAGS2 := $(shell $(PIE_TEST2) && $(ECHO) '-fno-PIE -nopie')
+WORKAROUND_CFLAGS += $(PIE_FLAGS) $(PIE_FLAGS2)
 endif
 
-# Define version string for lkrnprefix.S
-#
-CFLAGS_lkrnprefix	+= -DVERSION="\"$(VERSION)\""
-
-# Locations of isolinux files
-#
-SYSLINUX_DIR_LIST	:= \
-	/usr/lib/syslinux \
-	/usr/lib/syslinux/bios \
-	/usr/lib/syslinux/modules/bios \
-	/usr/share/syslinux \
-	/usr/share/syslinux/bios \
-	/usr/share/syslinux/modules/bios \
-	/usr/local/share/syslinux \
-	/usr/local/share/syslinux/bios \
-	/usr/local/share/syslinux/modules/bios \
-	/usr/lib/ISOLINUX
-ISOLINUX_BIN_LIST	:= \
-	$(ISOLINUX_BIN) \
-	$(patsubst %,%/isolinux.bin,$(SYSLINUX_DIR_LIST))
-LDLINUX_C32_LIST	:= \
-	$(LDLINUX_C32) \
-	$(patsubst %,%/ldlinux.c32,$(SYSLINUX_DIR_LIST))
-ISOLINUX_BIN	= $(firstword $(wildcard $(ISOLINUX_BIN_LIST)))
-LDLINUX_C32	= $(firstword $(wildcard $(LDLINUX_C32_LIST)))
-
 # i386-specific directories containing source files
 #
-SRCDIRS		+= arch/i386/core arch/i386/transitions arch/i386/prefix
-SRCDIRS		+= arch/i386/firmware/pcbios
-SRCDIRS		+= arch/i386/image
-SRCDIRS		+= arch/i386/interface/pcbios
-SRCDIRS		+= arch/i386/interface/pxe
-SRCDIRS		+= arch/i386/interface/pxeparent
-SRCDIRS 	+= arch/i386/interface/syslinux
-SRCDIRS		+= arch/i386/interface/vmware
-SRCDIRS		+= arch/i386/hci/commands
+SRCDIRS		+= arch/i386/core
+SRCDIRS		+= arch/i386/tests
 
 # Include common x86 Makefile
 #

src/arch/i386/Makefile.efi

@@ -8,6 +8,10 @@ ELF2EFI		= $(ELF2EFI32)
 #
 CFLAGS		+= -malign-double
 
+# Specify EFI boot file
+#
+EFI_BOOT_FILE	= bootia32.efi
+
 # Include generic EFI Makefile
 #
 MAKEDEPS	+= arch/x86/Makefile.efi

src/arch/i386/Makefile.pcbios

@@ -1,101 +1,6 @@
 # -*- makefile -*- : Force emacs to use Makefile mode
 
-# The i386 linker script
+# Include generic BIOS Makefile
 #
-LDSCRIPT	= arch/i386/scripts/i386.lds
-
-# Stop ld from complaining about our customised linker script
-#
-LDFLAGS		+= -N --no-check-sections
-
-# pcbios specific drivers
-SRCDIRS		+= arch/i386/drivers
-SRCDIRS		+= arch/i386/drivers/net
-
-# Media types.
-#
-MEDIA		+= rom
-MEDIA		+= mrom
-MEDIA		+= pcirom
-MEDIA		+= isarom
-MEDIA		+= pxe
-MEDIA		+= kpxe
-MEDIA		+= kkpxe
-MEDIA		+= kkkpxe
-MEDIA		+= lkrn
-MEDIA		+= dsk
-MEDIA		+= nbi
-MEDIA		+= hd
-MEDIA		+= raw
-MEDIA		+= exe
-
-# Padding rules
-#
-PAD_rom		= $(PERL) $(PADIMG) --blksize=512 --byte=0xff
-PAD_mrom	= $(PAD_rom)
-PAD_pcirom	= $(PAD_rom)
-PAD_isarom	= $(PAD_rom)
-PAD_dsk		= $(PERL) $(PADIMG) --blksize=512
-PAD_hd		= $(PERL) $(PADIMG) --blksize=32768
-PAD_exe		= $(PERL) $(PADIMG) --blksize=512
-
-# Finalisation rules
-#
-FINALISE_rom	= $(PERL) $(FIXROM)
-FINALISE_mrom	= $(FINALISE_rom)
-FINALISE_pcirom	= $(FINALISE_rom)
-FINALISE_isarom	= $(FINALISE_rom)
-
-# Use $(ROMS) rather than $(DRIVERS) for "allroms", "allmroms", etc.
-#
-LIST_NAME_rom := ROMS
-LIST_NAME_mrom := ROMS
-LIST_NAME_pcirom := ROMS
-LIST_NAME_isarom := ROMS
-
-# rule to make a non-emulation ISO boot image
-NON_AUTO_MEDIA	+= iso
-%iso:	%lkrn util/geniso
-	$(QM)$(ECHO) "  [GENISO] $@"
-	$(Q)ISOLINUX_BIN=$(ISOLINUX_BIN) LDLINUX_C32=$(LDLINUX_C32) \
-	    VERSION="$(VERSION)" bash util/geniso -o $@ $<
-
-# rule to make a floppy emulation ISO boot image
-NON_AUTO_MEDIA	+= liso
-%liso:	%lkrn util/geniso
-	$(QM)$(ECHO) "  [GENISO] $@"
-	$(Q)VERSION="$(VERSION)" bash util/geniso -l -o $@ $<
-
-# rule to make a syslinux floppy image (mountable, bootable)
-NON_AUTO_MEDIA	+= sdsk
-%sdsk:	%lkrn util/gensdsk
-	$(QM)$(ECHO) "  [GENSDSK] $@"
-	$(Q)bash util/gensdsk $@ $<
-
-# rule to write disk images to /dev/fd0
-NON_AUTO_MEDIA	+= fd0
-%fd0 : %dsk
-	$(QM)$(ECHO) "  [DD] $@"
-	$(Q)dd if=$< bs=512 conv=sync of=/dev/fd0
-	$(Q)sync
-
-# Special target for building Master Boot Record binary
-$(BIN)/mbr.bin : $(BIN)/mbr.o
-	$(QM)$(ECHO) "  [OBJCOPY] $@"
-	$(Q)$(OBJCOPY) -O binary $< $@
-
-# rule to make a USB disk image
-$(BIN)/usbdisk.bin : $(BIN)/usbdisk.o
-	$(QM)$(ECHO) "  [OBJCOPY] $@"
-	$(Q)$(OBJCOPY) -O binary $< $@
-
-NON_AUTO_MEDIA	+= usb
-%usb: $(BIN)/usbdisk.bin %hd
-	$(QM)$(ECHO) "  [FINISH] $@"
-	$(Q)cat $^ > $@
-
-# Padded floppy image (e.g. for iLO)
-NON_AUTO_MEDIA += pdsk
-%pdsk : %dsk
-	$(Q)cp $< $@
-	$(Q)$(PADIMG) --blksize=1474560 $@
+MAKEDEPS	+= arch/x86/Makefile.pcbios
+include arch/x86/Makefile.pcbios

src/arch/i386/core/gdbidt.S

@@ -15,41 +15,29 @@
 /* POSIX signal numbers for reporting traps to GDB */
 #define SIGILL 4
 #define SIGTRAP 5
-#define SIGBUS 7
 #define SIGFPE 8
-#define SIGSEGV 11
 #define SIGSTKFLT 16
 
-	.globl	gdbmach_nocode_sigfpe
-gdbmach_nocode_sigfpe:
+	.globl	gdbmach_sigfpe
+gdbmach_sigfpe:
 	pushl	$SIGFPE
 	jmp	gdbmach_interrupt
 
-	.globl	gdbmach_nocode_sigtrap
-gdbmach_nocode_sigtrap:
+	.globl	gdbmach_sigtrap
+gdbmach_sigtrap:
 	pushl	$SIGTRAP
 	jmp	gdbmach_interrupt
 
-	.globl	gdbmach_nocode_sigstkflt
-gdbmach_nocode_sigstkflt:
+	.globl	gdbmach_sigstkflt
+gdbmach_sigstkflt:
 	pushl	$SIGSTKFLT
 	jmp	gdbmach_interrupt
 
-	.globl	gdbmach_nocode_sigill
-gdbmach_nocode_sigill:
+	.globl	gdbmach_sigill
+gdbmach_sigill:
 	pushl	$SIGILL
 	jmp	gdbmach_interrupt
 
-	.globl	gdbmach_withcode_sigbus
-gdbmach_withcode_sigbus:
-	movl	$SIGBUS, (%esp)
-	jmp	gdbmach_interrupt
-
-	.globl	gdbmach_withcode_sigsegv
-gdbmach_withcode_sigsegv:
-	movl	$SIGSEGV, (%esp)
-	jmp	gdbmach_interrupt
-
 /* When invoked, the stack contains: eflags, cs, eip, signo. */
 #define IH_OFFSET_GDB_REGS ( 0 )
 #define IH_OFFSET_GDB_EIP ( IH_OFFSET_GDB_REGS + SIZEOF_I386_REGS )

src/arch/i386/core/gdbmach.c

@@ -1,184 +0,0 @@
-/*
- * Copyright (C) 2008 Stefan Hajnoczi <stefanha@gmail.com>.
- *
- * 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 of the
- * License, or any later version.
- *
- * This program is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License for more details.
- *
- * You should have received a copy of the GNU General Public License
- * along with this program; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
- * 02110-1301, USA.
- *
- * You can also choose to distribute this program under the terms of
- * the Unmodified Binary Distribution Licence (as given in the file
- * COPYING.UBDL), provided that you have satisfied its requirements.
- */
-
-FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
-
-#include <stddef.h>
-#include <stdio.h>
-#include <assert.h>
-#include <ipxe/uaccess.h>
-#include <ipxe/gdbstub.h>
-#include <librm.h>
-#include <gdbmach.h>
-
-/** @file
- *
- * GDB architecture-specific bits for i386
- *
- */
-
-enum {
-	DR7_CLEAR = 0x00000400,    /* disable hardware breakpoints */
-	DR6_CLEAR = 0xffff0ff0,    /* clear breakpoint status */
-};
-
-/** Hardware breakpoint, fields stored in x86 bit pattern form */
-struct hwbp {
-	int type;           /* type (1=write watchpoint, 3=access watchpoint) */
-	unsigned long addr; /* linear address */
-	size_t len;         /* length (0=1-byte, 1=2-byte, 3=4-byte) */
-	int enabled;
-};
-
-static struct hwbp hwbps [ 4 ];
-static gdbreg_t dr7 = DR7_CLEAR;
-
-static struct hwbp *gdbmach_find_hwbp ( int type, unsigned long addr, size_t len ) {
-	struct hwbp *available = NULL;
-	unsigned int i;
-	for ( i = 0; i < sizeof hwbps / sizeof hwbps [ 0 ]; i++ ) {
-		if ( hwbps [ i ].type == type && hwbps [ i ].addr == addr && hwbps [ i ].len == len ) {
-			return &hwbps [ i ];
-		}
-		if ( !hwbps [ i ].enabled ) {
-			available = &hwbps [ i ];
-		}
-	}
-	return available;
-}
-
-static void gdbmach_commit_hwbp ( struct hwbp *bp ) {
-	unsigned int regnum = bp - hwbps;
-
-	/* Set breakpoint address */
-	assert ( regnum < ( sizeof hwbps / sizeof hwbps [ 0 ] ) );
-	switch ( regnum ) {
-		case 0:
-			__asm__ __volatile__ ( "movl %0, %%dr0\n" : : "r" ( bp->addr ) );
-			break;
-		case 1:
-			__asm__ __volatile__ ( "movl %0, %%dr1\n" : : "r" ( bp->addr ) );
-			break;
-		case 2:
-			__asm__ __volatile__ ( "movl %0, %%dr2\n" : : "r" ( bp->addr ) );
-			break;
-		case 3:
-			__asm__ __volatile__ ( "movl %0, %%dr3\n" : : "r" ( bp->addr ) );
-			break;
-	}
-
-	/* Set type */
-	dr7 &= ~( 0x3 << ( 16 + 4 * regnum ) );
-	dr7 |= bp->type << ( 16 + 4 * regnum );
-
-	/* Set length */
-	dr7 &= ~( 0x3 << ( 18 + 4 * regnum ) );
-	dr7 |= bp->len << ( 18 + 4 * regnum );
-
-	/* Set/clear local enable bit */
-	dr7 &= ~( 0x3 << 2 * regnum );
- 	dr7 |= bp->enabled << 2 * regnum;
-}
-
-int gdbmach_set_breakpoint ( int type, unsigned long addr, size_t len, int enable ) {
-	struct hwbp *bp;
-	
-	/* Check and convert breakpoint type to x86 type */
-	switch ( type ) {
-		case GDBMACH_WATCH:
-			type = 0x1;
-			break;
-		case GDBMACH_AWATCH:
-			type = 0x3;
-			break;
-		default:
-			return 0; /* unsupported breakpoint type */
-	}
-
-	/* Only lengths 1, 2, and 4 are supported */
-	if ( len != 2 && len != 4 ) {
-		len = 1;
-	}
-	len--; /* convert to x86 breakpoint length bit pattern */
-
-	/* Calculate linear address by adding segment base */
-	addr += virt_offset;
-
-	/* Set up the breakpoint */
-	bp = gdbmach_find_hwbp ( type, addr, len );
-	if ( !bp ) {
-		return 0; /* ran out of hardware breakpoints */
-	}
-	bp->type = type;
-	bp->addr = addr;
-	bp->len = len;
-	bp->enabled = enable;
-	gdbmach_commit_hwbp ( bp );
-	return 1;
-}
-
-static void gdbmach_disable_hwbps ( void ) {
-	/* Store and clear hardware breakpoints */
-	__asm__ __volatile__ ( "movl %0, %%dr7\n" : : "r" ( DR7_CLEAR ) );
-}
-
-static void gdbmach_enable_hwbps ( void ) {
-	/* Clear breakpoint status register */
-	__asm__ __volatile__ ( "movl %0, %%dr6\n" : : "r" ( DR6_CLEAR ) );
-
-	/* Restore hardware breakpoints */
-	__asm__ __volatile__ ( "movl %0, %%dr7\n" : : "r" ( dr7 ) );
-}
-
-__asmcall void gdbmach_handler ( int signo, gdbreg_t *regs ) {
-	gdbmach_disable_hwbps();
-	gdbstub_handler ( signo, regs );
-	gdbmach_enable_hwbps();
-}
-
-static void * gdbmach_interrupt_vectors[] = {
-	gdbmach_nocode_sigfpe,		/* Divide by zero */
-	gdbmach_nocode_sigtrap,		/* Debug trap */
-	NULL,				/* Non-maskable interrupt */
-	gdbmach_nocode_sigtrap,		/* Breakpoint */
-	gdbmach_nocode_sigstkflt,	/* Overflow */
-	gdbmach_nocode_sigstkflt,	/* Bound range exceeded */
-	gdbmach_nocode_sigill,		/* Invalid opcode */
-	NULL,				/* Device not available */
-	gdbmach_withcode_sigbus,	/* Double fault */
-	NULL,				/* Coprocessor segment overrun */
-	gdbmach_withcode_sigsegv,	/* Invalid TSS */
-	gdbmach_withcode_sigsegv,	/* Segment not present */
-	gdbmach_withcode_sigsegv,	/* Stack segment fault */
-	gdbmach_withcode_sigsegv,	/* General protection fault */
-	gdbmach_withcode_sigsegv,	/* Page fault */
-};
-
-void gdbmach_init ( void ) {
-	unsigned int i;
-
-	for ( i = 0 ; i < ( sizeof ( gdbmach_interrupt_vectors ) /
-			    sizeof ( gdbmach_interrupt_vectors[0] ) ) ; i++ ) {
-		set_interrupt_vector ( i, gdbmach_interrupt_vectors[i] );
-	}
-}

src/arch/i386/core/rdtsc_timer.c

@@ -1,94 +0,0 @@
-/*
- * Copyright (C) 2008 Michael Brown <mbrown@fensystems.co.uk>.
- *
- * 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 of the
- * License, or any later version.
- *
- * This program is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License for more details.
- *
- * You should have received a copy of the GNU General Public License
- * along with this program; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
- * 02110-1301, USA.
- *
- * You can also choose to distribute this program under the terms of
- * the Unmodified Binary Distribution Licence (as given in the file
- * COPYING.UBDL), provided that you have satisfied its requirements.
- */
-
-FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
-
-/** @file
- *
- * RDTSC timer
- *
- */
-
-#include <assert.h>
-#include <ipxe/timer.h>
-#include <ipxe/pit8254.h>
-
-/**
- * Number of TSC ticks per microsecond
- *
- * This is calibrated on the first use of the timer.
- */
-static unsigned long rdtsc_ticks_per_usec;
-
-/**
- * Delay for a fixed number of microseconds
- *
- * @v usecs		Number of microseconds for which to delay
- */
-static void rdtsc_udelay ( unsigned long usecs ) {
-	unsigned long start;
-	unsigned long elapsed;
-
-	/* Sanity guard, since we may divide by this */
-	if ( ! usecs )
-		usecs = 1;
-
-	start = currticks();
-	if ( rdtsc_ticks_per_usec ) {
-		/* Already calibrated; busy-wait until done */
-		do {
-			elapsed = ( currticks() - start );
-		} while ( elapsed < ( usecs * rdtsc_ticks_per_usec ) );
-	} else {
-		/* Not yet calibrated; use 8254 PIT and calibrate
-		 * based on result.
-		 */
-		pit8254_udelay ( usecs );
-		elapsed = ( currticks() - start );
-		rdtsc_ticks_per_usec = ( elapsed / usecs );
-		DBG ( "RDTSC timer calibrated: %ld ticks in %ld usecs "
-		      "(%ld MHz)\n", elapsed, usecs,
-		      ( rdtsc_ticks_per_usec << TSC_SHIFT ) );
-	}
-}
-
-/**
- * Get number of ticks per second
- *
- * @ret ticks_per_sec	Number of ticks per second
- */
-static unsigned long rdtsc_ticks_per_sec ( void ) {
-
-	/* Calibrate timer, if not already done */
-	if ( ! rdtsc_ticks_per_usec )
-		udelay ( 1 );
-
-	/* Sanity check */
-	assert ( rdtsc_ticks_per_usec != 0 );
-
-	return ( rdtsc_ticks_per_usec * 1000 * 1000 );
-}
-
-PROVIDE_TIMER ( rdtsc, udelay, rdtsc_udelay );
-PROVIDE_TIMER_INLINE ( rdtsc, currticks );
-PROVIDE_TIMER ( rdtsc, ticks_per_sec, rdtsc_ticks_per_sec );

src/arch/i386/core/virtaddr.S

@@ -1,145 +0,0 @@
-/*
- * Functions to support the virtual addressing method of relocation
- * that Etherboot uses.
- *
- */
-
-FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL )
-
-#include "librm.h"
-		
-	.arch i386
-	.text
-	.code32
-	
-/****************************************************************************
- * _virt_to_phys (virtual addressing)
- *
- * Switch from virtual to flat physical addresses.  %esp is adjusted
- * to a physical value.  Segment registers are set to flat physical
- * selectors.  All other registers are preserved.  Flags are
- * preserved.
- *
- * Parameters: none
- * Returns: none
- ****************************************************************************
- */
-	.globl _virt_to_phys
-_virt_to_phys:
-	/* Preserve registers and flags */
-	pushfl
-	pushl	%eax
-	pushl	%ebp
-
-	/* Change return address to a physical address */
-	movl	virt_offset, %ebp
-	addl	%ebp, 12(%esp)
-
-	/* Switch to physical code segment */
-	cli
-	pushl	$PHYSICAL_CS
-	leal	1f(%ebp), %eax
-	pushl	%eax
-	lret
-1:
-	/* Reload other segment registers and adjust %esp */
-	movl	$PHYSICAL_DS, %eax
-	movl	%eax, %ds
-	movl	%eax, %es
-	movl	%eax, %fs
-	movl	%eax, %gs
-	movl	%eax, %ss
-	addl	%ebp, %esp
-
-	/* Restore registers and flags, and return */
-	popl	%ebp
-	popl	%eax
-	popfl
-	ret
-
-/****************************************************************************
- * _phys_to_virt (flat physical addressing)
- *
- * Switch from flat physical to virtual addresses.  %esp is adjusted
- * to a virtual value.  Segment registers are set to virtual
- * selectors.  All other registers are preserved.  Flags are
- * preserved.
- *
- * Parameters: none
- * Returns: none
- ****************************************************************************
- */
-	.globl _phys_to_virt
-_phys_to_virt:
-	/* Preserve registers and flags */
-	pushfl
-	pushl	%eax
-	pushl	%ebp
-
-	/* Switch to virtual code segment */
-	cli
-	ljmp	$VIRTUAL_CS, $1f
-1:
-	/* Reload data segment registers */
-	movl	$VIRTUAL_DS, %eax
-	movl	%eax, %ds
-	movl	%eax, %es
-	movl	%eax, %fs
-	movl	%eax, %gs
-
-	/* Reload stack segment and adjust %esp */
-	movl	virt_offset, %ebp
-	movl	%eax, %ss
-	subl	%ebp, %esp
-
-	/* Change the return address to a virtual address */
-	subl	%ebp, 12(%esp)
-
-	/* Restore registers and flags, and return */
-	popl	%ebp
-	popl	%eax
-	popfl
-	ret
-
-/****************************************************************************
- * _intr_to_virt (virtual code segment, virtual or physical stack segment)
- *
- * Switch from virtual code segment with either a virtual or physical
- * stack segment to using virtual addressing.  %esp is adjusted if
- * necessary to a virtual value.  Segment registers are set to virtual
- * selectors.  All other registers are preserved.  Flags are
- * preserved.
- *
- * Parameters: none
- * Returns: none
- ****************************************************************************
- */
-	.globl _intr_to_virt
-_intr_to_virt:
-	/* Preserve registers and flags */
-	pushfl
-	pushl	%eax
-	pushl	%ebp
-
-	/* Check whether stack segment is physical or virtual */
-	movl	%ss, %eax
-	cmpw	$VIRTUAL_DS, %ax
-	movl	$VIRTUAL_DS, %eax
-
-	/* Reload data segment registers */
-	movl	%eax, %ds
-	movl	%eax, %es
-	movl	%eax, %fs
-	movl	%eax, %gs
-
-	/* Reload stack segment and adjust %esp if necessary */
-	je	1f
-	movl	virt_offset, %ebp
-	movl	%eax, %ss
-	subl	%ebp, %esp
-1:
-	/* Restore registers and flags, and return */
-	popl	%ebp
-	popl	%eax
-	popfl
-	ret

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

@@ -9,7 +9,7 @@ FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
 #ifndef ASSEMBLY
 
 /** Declare a function with standard calling conventions */
-#define __asmcall __attribute__ (( cdecl, regparm(0) ))
+#define __asmcall __attribute__ (( used, cdecl, regparm(0) ))
 
 /**
  * Declare a function with libgcc implicit linkage

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

@@ -46,27 +46,4 @@ hv_call ( struct hv_hypervisor *hv, unsigned int code, const void *in,
 	return result;
 }
 
-/**
- * Set bit atomically
- *
- * @v bits		Bit field
- * @v bit		Bit to set
- */
-static inline __attribute__ (( always_inline )) void
-hv_set_bit ( void *bits, unsigned int bit ) {
-	struct {
-		uint32_t dword[ ( bit / 32 ) + 1 ];
-	} *dwords = bits;
-
-	/* Set bit using "lock bts".  Inform compiler that any memory
-	 * from the start of the bit field up to and including the
-	 * dword containing this bit may be modified.  (This is
-	 * overkill but shouldn't matter in practice since we're
-	 * unlikely to subsequently read other bits from the same bit
-	 * field.)
-	 */
-	__asm__ __volatile__ ( "lock bts %1, %0"
-			       : "+m" ( *dwords ) : "Ir" ( bit ) );
-}
-
 #endif /* _BITS_HYPERV_H */

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

@@ -1,15 +0,0 @@
-#ifndef _BITS_TIMER_H
-#define _BITS_TIMER_H
-
-/** @file
- *
- * i386-specific timer API implementations
- *
- */
-
-FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
-
-#include <ipxe/bios_timer.h>
-#include <ipxe/rdtsc_timer.h>
-
-#endif /* _BITS_TIMER_H */

src/arch/i386/include/efi/ipxe/dhcp_arch.h

@@ -33,14 +33,8 @@ FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
 
 #include <ipxe/dhcp.h>
 
-#define DHCP_ARCH_VENDOR_CLASS_ID \
-	DHCP_STRING ( 'P', 'X', 'E', 'C', 'l', 'i', 'e', 'n', 't', ':',      \
-		      'A', 'r', 'c', 'h', ':', '0', '0', '0', '0', '6', ':', \
-		      'U', 'N', 'D', 'I', ':', '0', '0', '3', '0', '1', '0' )
+#define DHCP_ARCH_CLIENT_ARCHITECTURE DHCP_CLIENT_ARCHITECTURE_IA32
 
-#define DHCP_ARCH_CLIENT_ARCHITECTURE \
-	DHCP_WORD ( DHCP_CLIENT_ARCHITECTURE_IA32 )
-
-#define DHCP_ARCH_CLIENT_NDI DHCP_OPTION ( 1 /* UNDI */ , 3, 10 /* v3.10 */ )
+#define DHCP_ARCH_CLIENT_NDI 1 /* UNDI */ , 3, 10 /* v3.10 */
 
 #endif

src/arch/i386/include/gdbmach.h

@@ -47,12 +47,10 @@ enum {
 };
 
 /* Interrupt vectors */
-extern void gdbmach_nocode_sigfpe ( void );
-extern void gdbmach_nocode_sigtrap ( void );
-extern void gdbmach_nocode_sigstkflt ( void );
-extern void gdbmach_nocode_sigill ( void );
-extern void gdbmach_withcode_sigbus ( void );
-extern void gdbmach_withcode_sigsegv ( void );
+extern void gdbmach_sigfpe ( void );
+extern void gdbmach_sigtrap ( void );
+extern void gdbmach_sigstkflt ( void );
+extern void gdbmach_sigill ( void );
 
 static inline void gdbmach_set_pc ( gdbreg_t *regs, gdbreg_t pc ) {
 	regs [ GDBMACH_EIP ] = pc;

src/arch/i386/include/ipxe/bios_timer.h

@@ -1,44 +0,0 @@
-#ifndef _IPXE_BIOS_TIMER_H
-#define _IPXE_BIOS_TIMER_H
-
-/** @file
- *
- * BIOS timer
- *
- */
-
-FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
-
-#ifdef TIMER_PCBIOS
-#define TIMER_PREFIX_pcbios
-#else
-#define TIMER_PREFIX_pcbios __pcbios_
-#endif
-
-#include <ipxe/pit8254.h>
-
-/**
- * Delay for a fixed number of microseconds
- *
- * @v usecs		Number of microseconds for which to delay
- */
-static inline __always_inline void
-TIMER_INLINE ( pcbios, udelay ) ( unsigned long usecs ) {
-	/* BIOS timer is not high-resolution enough for udelay(), so
-	 * we use the 8254 Programmable Interval Timer.
-	 */
-	pit8254_udelay ( usecs );
-}
-
-/**
- * Get number of ticks per second
- *
- * @ret ticks_per_sec	Number of ticks per second
- */
-static inline __always_inline unsigned long
-TIMER_INLINE ( pcbios, ticks_per_sec ) ( void ) {
-	/* BIOS timer ticks over at 18.2 ticks per second */
-	return 18;
-}
-
-#endif /* _IPXE_BIOS_TIMER_H */

src/arch/i386/include/ipxe/rdtsc_timer.h

@@ -1,39 +0,0 @@
-#ifndef _IPXE_RDTSC_TIMER_H
-#define _IPXE_RDTSC_TIMER_H
-
-/** @file
- *
- * RDTSC timer
- *
- */
-
-FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
-
-#ifdef TIMER_RDTSC
-#define TIMER_PREFIX_rdtsc
-#else
-#define TIMER_PREFIX_rdtsc __rdtsc_
-#endif
-
-/**
- * RDTSC values can easily overflow an unsigned long.  We discard the
- * low-order bits in order to obtain sensibly-scaled values.
- */
-#define TSC_SHIFT 8
-
-/**
- * Get current system time in ticks
- *
- * @ret ticks		Current time, in ticks
- */
-static inline __always_inline unsigned long
-TIMER_INLINE ( rdtsc, currticks ) ( void ) {
-	unsigned long ticks;
-
-	__asm__ __volatile__ ( "rdtsc\n\t"
-			       "shrdl %1, %%edx, %%eax\n\t"
-			       : "=a" ( ticks ) : "i" ( TSC_SHIFT ) : "edx" );
-	return ticks;
-}
-
-#endif /* _IPXE_RDTSC_TIMER_H */

src/arch/i386/include/pcbios/ipxe/dhcp_arch.h

@@ -33,14 +33,8 @@ FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
 
 #include <ipxe/dhcp.h>
 
-#define DHCP_ARCH_VENDOR_CLASS_ID \
-	DHCP_STRING ( 'P', 'X', 'E', 'C', 'l', 'i', 'e', 'n', 't', ':',      \
-		      'A', 'r', 'c', 'h', ':', '0', '0', '0', '0', '0', ':', \
-		      'U', 'N', 'D', 'I', ':', '0', '0', '2', '0', '0', '1' )
+#define DHCP_ARCH_CLIENT_ARCHITECTURE DHCP_CLIENT_ARCHITECTURE_X86
 
-#define DHCP_ARCH_CLIENT_ARCHITECTURE \
-	DHCP_WORD ( DHCP_CLIENT_ARCHITECTURE_X86 )
-
-#define DHCP_ARCH_CLIENT_NDI DHCP_OPTION ( 1 /* UNDI */ , 2, 1 /* v2.1 */ )
+#define DHCP_ARCH_CLIENT_NDI 1 /* UNDI */ , 2, 1 /* v2.1 */
 
 #endif

src/arch/i386/include/pxeparent.h

@@ -1,11 +0,0 @@
-#ifndef PXEPARENT_H
-#define PXEPARENT_H
-
-FILE_LICENCE ( GPL2_OR_LATER );
-
-#include <pxe_types.h>
-
-extern int pxeparent_call ( SEGOFF16_t entry, unsigned int function,
-			    void *params, size_t params_len );
-
-#endif

src/arch/i386/include/setjmp.h

@@ -4,7 +4,6 @@
 FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
 
 #include <stdint.h>
-#include <realmode.h>
 
 /** A jump buffer */
 typedef struct {
@@ -22,29 +21,10 @@ typedef struct {
 	uint32_t ebp;
 } jmp_buf[1];
 
-/** A real-mode-extended jump buffer */
-typedef struct {
-	/** Jump buffer */
-	jmp_buf env;
-	/** Real-mode stack pointer */
-	segoff_t rm_stack;
-} rmjmp_buf[1];
-
 extern int __asmcall __attribute__ (( returns_twice ))
 setjmp ( jmp_buf env );
 
 extern void __asmcall __attribute__ (( noreturn ))
 longjmp ( jmp_buf env, int val );
 
-#define rmsetjmp( _env ) ( {					\
-	(_env)->rm_stack.segment = rm_ss;			\
-	(_env)->rm_stack.offset = rm_sp;			\
-	setjmp ( (_env)->env ); } )				\
-
-#define rmlongjmp( _env, _val ) do {				\
-	rm_ss = (_env)->rm_stack.segment;			\
-	rm_sp = (_env)->rm_stack.offset;			\
-	longjmp ( (_env)->env, (_val) );			\
-	} while ( 0 )
-
 #endif /* _SETJMP_H */

src/arch/i386/interface/pxeparent/pxeparent.c

@@ -1,283 +0,0 @@
-/*
- * Copyright (C) 2007 Michael Brown <mbrown@fensystems.co.uk>.
- *
- * 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 of the
- * License, or any later version.
- *
- * This program is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License for more details.
- *
- * You should have received a copy of the GNU General Public License
- * along with this program; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
- * 02110-1301, USA.
- */
-
-FILE_LICENCE ( GPL2_OR_LATER );
-
-#include <ipxe/dhcp.h>
-#include <ipxe/profile.h>
-#include <pxeparent.h>
-#include <pxe_api.h>
-#include <pxe_types.h>
-#include <pxe.h>
-
-/** @file
- *
- * Call interface to parent PXE stack
- *
- */
-
-/* Disambiguate the various error causes */
-#define EINFO_EPXECALL							\
-	__einfo_uniqify ( EINFO_EPLATFORM, 0x01,			\
-			  "External PXE API error" )
-#define EPXECALL( status ) EPLATFORM ( EINFO_EPXECALL, status )
-
-/** A parent PXE API call profiler */
-struct pxeparent_profiler {
-	/** Total time spent performing REAL_CALL() */
-	struct profiler total;
-	/** Time spent transitioning to real mode */
-	struct profiler p2r;
-	/** Time spent in external code */
-	struct profiler ext;
-	/** Time spent transitioning back to protected mode */
-	struct profiler r2p;
-};
-
-/** PXENV_UNDI_TRANSMIT profiler */
-static struct pxeparent_profiler pxeparent_tx_profiler __profiler = {
-	{ .name = "pxeparent.tx" },
-	{ .name = "pxeparent.tx_p2r" },
-	{ .name = "pxeparent.tx_ext" },
-	{ .name = "pxeparent.tx_r2p" },
-};
-
-/** PXENV_UNDI_ISR profiler
- *
- * Note that this profiler will not see calls to
- * PXENV_UNDI_ISR_IN_START, which are handled by the UNDI ISR and do
- * not go via pxeparent_call().
- */
-static struct pxeparent_profiler pxeparent_isr_profiler __profiler = {
-	{ .name = "pxeparent.isr" },
-	{ .name = "pxeparent.isr_p2r" },
-	{ .name = "pxeparent.isr_ext" },
-	{ .name = "pxeparent.isr_r2p" },
-};
-
-/** PXE unknown API call profiler
- *
- * This profiler can be used to measure the overhead of a dummy PXE
- * API call.
- */
-static struct pxeparent_profiler pxeparent_unknown_profiler __profiler = {
-	{ .name = "pxeparent.unknown" },
-	{ .name = "pxeparent.unknown_p2r" },
-	{ .name = "pxeparent.unknown_ext" },
-	{ .name = "pxeparent.unknown_r2p" },
-};
-
-/** Miscellaneous PXE API call profiler */
-static struct pxeparent_profiler pxeparent_misc_profiler __profiler = {
-	{ .name = "pxeparent.misc" },
-	{ .name = "pxeparent.misc_p2r" },
-	{ .name = "pxeparent.misc_ext" },
-	{ .name = "pxeparent.misc_r2p" },
-};
-
-/**
- * Name PXE API call
- *
- * @v function		API call number
- * @ret name		API call name
- */
-static inline __attribute__ (( always_inline )) const char *
-pxeparent_function_name ( unsigned int function ) {
-	switch ( function ) {
-	case PXENV_START_UNDI:
-		return "PXENV_START_UNDI";
-	case PXENV_STOP_UNDI:
-		return "PXENV_STOP_UNDI";
-	case PXENV_UNDI_STARTUP:
-		return "PXENV_UNDI_STARTUP";
-	case PXENV_UNDI_CLEANUP:
-		return "PXENV_UNDI_CLEANUP";
-	case PXENV_UNDI_INITIALIZE:
-		return "PXENV_UNDI_INITIALIZE";
-	case PXENV_UNDI_RESET_ADAPTER:
-		return "PXENV_UNDI_RESET_ADAPTER";
-	case PXENV_UNDI_SHUTDOWN:
-		return "PXENV_UNDI_SHUTDOWN";
-	case PXENV_UNDI_OPEN:
-		return "PXENV_UNDI_OPEN";
-	case PXENV_UNDI_CLOSE:
-		return "PXENV_UNDI_CLOSE";
-	case PXENV_UNDI_TRANSMIT:
-		return "PXENV_UNDI_TRANSMIT";
-	case PXENV_UNDI_SET_MCAST_ADDRESS:
-		return "PXENV_UNDI_SET_MCAST_ADDRESS";
-	case PXENV_UNDI_SET_STATION_ADDRESS:
-		return "PXENV_UNDI_SET_STATION_ADDRESS";
-	case PXENV_UNDI_SET_PACKET_FILTER:
-		return "PXENV_UNDI_SET_PACKET_FILTER";
-	case PXENV_UNDI_GET_INFORMATION:
-		return "PXENV_UNDI_GET_INFORMATION";
-	case PXENV_UNDI_GET_STATISTICS:
-		return "PXENV_UNDI_GET_STATISTICS";
-	case PXENV_UNDI_CLEAR_STATISTICS:
-		return "PXENV_UNDI_CLEAR_STATISTICS";
-	case PXENV_UNDI_INITIATE_DIAGS:
-		return "PXENV_UNDI_INITIATE_DIAGS";
-	case PXENV_UNDI_FORCE_INTERRUPT:
-		return "PXENV_UNDI_FORCE_INTERRUPT";
-	case PXENV_UNDI_GET_MCAST_ADDRESS:
-		return "PXENV_UNDI_GET_MCAST_ADDRESS";
-	case PXENV_UNDI_GET_NIC_TYPE:
-		return "PXENV_UNDI_GET_NIC_TYPE";
-	case PXENV_UNDI_GET_IFACE_INFO:
-		return "PXENV_UNDI_GET_IFACE_INFO";
-	/*
-	 * Duplicate case value; this is a bug in the PXE specification.
-	 *
-	 *	case PXENV_UNDI_GET_STATE:
-	 *		return "PXENV_UNDI_GET_STATE";
-	 */
-	case PXENV_UNDI_ISR:
-		return "PXENV_UNDI_ISR";
-	case PXENV_GET_CACHED_INFO:
-		return "PXENV_GET_CACHED_INFO";
-	default:
-		return "UNKNOWN API CALL";
-	}
-}
-
-/**
- * Determine applicable profiler pair (for debugging)
- *
- * @v function		API call number
- * @ret profiler	Profiler
- */
-static struct pxeparent_profiler * pxeparent_profiler ( unsigned int function ){
-
-	/* Determine applicable profiler */
-	switch ( function ) {
-	case PXENV_UNDI_TRANSMIT:
-		return &pxeparent_tx_profiler;
-	case PXENV_UNDI_ISR:
-		return &pxeparent_isr_profiler;
-	case PXENV_UNKNOWN:
-		return &pxeparent_unknown_profiler;
-	default:
-		return &pxeparent_misc_profiler;
-	}
-}
-
-/**
- * PXE parent parameter block
- *
- * Used as the parameter block for all parent PXE API calls.  Resides
- * in base memory.
- */
-static union u_PXENV_ANY __bss16 ( pxeparent_params );
-#define pxeparent_params __use_data16 ( pxeparent_params )
-
-/** PXE parent entry point
- *
- * Used as the indirection vector for all parent PXE API calls.  Resides in
- * base memory.
- */
-SEGOFF16_t __bss16 ( pxeparent_entry_point );
-#define pxeparent_entry_point __use_data16 ( pxeparent_entry_point )
-
-/**
- * Issue parent PXE API call
- *
- * @v entry		Parent PXE stack entry point
- * @v function		API call number
- * @v params		PXE parameter block
- * @v params_len	Length of PXE parameter block
- * @ret rc		Return status code
- */
-int pxeparent_call ( SEGOFF16_t entry, unsigned int function,
-		     void *params, size_t params_len ) {
-	struct pxeparent_profiler *profiler = pxeparent_profiler ( function );
-	PXENV_EXIT_t exit;
-	unsigned long started;
-	unsigned long stopped;
-	int discard_D;
-	int rc;
-
-	/* Copy parameter block and entry point */
-	assert ( params_len <= sizeof ( pxeparent_params ) );
-	memcpy ( &pxeparent_params, params, params_len );
-	memcpy ( &pxeparent_entry_point, &entry, sizeof ( entry ) );
-
-	/* Call real-mode entry point.  This calling convention will
-	 * work with both the !PXE and the PXENV+ entry points.
-	 */
-	profile_start ( &profiler->total );
-	__asm__ __volatile__ ( REAL_CODE ( "pushl %%ebp\n\t" /* gcc bug */
-					   "rdtsc\n\t"
-					   "pushl %%eax\n\t"
-					   "pushw %%es\n\t"
-					   "pushw %%di\n\t"
-					   "pushw %%bx\n\t"
-					   "lcall *pxeparent_entry_point\n\t"
-					   "movw %%ax, %%bx\n\t"
-					   "rdtsc\n\t"
-					   "addw $6, %%sp\n\t"
-					   "popl %%edx\n\t"
-					   "popl %%ebp\n\t" /* gcc bug */ )
-			       : "=a" ( stopped ), "=d" ( started ),
-				 "=b" ( exit ), "=D" ( discard_D )
-			       : "b" ( function ),
-			         "D" ( __from_data16 ( &pxeparent_params ) )
-			       : "ecx", "esi" );
-	profile_stop ( &profiler->total );
-	profile_start_at ( &profiler->p2r, profile_started ( &profiler->total));
-	profile_stop_at ( &profiler->p2r, started );
-	profile_start_at ( &profiler->ext, started );
-	profile_stop_at ( &profiler->ext, stopped );
-	profile_start_at ( &profiler->r2p, stopped );
-	profile_stop_at ( &profiler->r2p, profile_stopped ( &profiler->total ));
-
-	/* Determine return status code based on PXENV_EXIT and
-	 * PXENV_STATUS
-	 */
-	rc = ( ( exit == PXENV_EXIT_SUCCESS ) ?
-	       0 : -EPXECALL ( pxeparent_params.Status ) );
-
-	/* If anything goes wrong, print as much debug information as
-	 * it's possible to give.
-	 */
-	if ( rc != 0 ) {
-		SEGOFF16_t rm_params = {
-			.segment = rm_ds,
-			.offset = __from_data16 ( &pxeparent_params ),
-		};
-
-		DBG ( "PXEPARENT %s failed: %s\n",
-		       pxeparent_function_name ( function ), strerror ( rc ) );
-		DBG ( "PXEPARENT parameters at %04x:%04x length "
-		       "%#02zx, entry point at %04x:%04x\n",
-		       rm_params.segment, rm_params.offset, params_len,
-		       pxeparent_entry_point.segment,
-		       pxeparent_entry_point.offset );
-		DBG ( "PXEPARENT parameters provided:\n" );
-		DBG_HDA ( rm_params, params, params_len );
-		DBG ( "PXEPARENT parameters returned:\n" );
-		DBG_HDA ( rm_params, &pxeparent_params, params_len );
-	}
-
-	/* Copy parameter block back */
-	memcpy ( params, &pxeparent_params, params_len );
-
-	return rc;
-}
-

src/arch/i386/transitions/librm.S

@@ -1,671 +0,0 @@
-/*
- * librm: a library for interfacing to real-mode code
- *
- * Michael Brown <mbrown@fensystems.co.uk>
- *
- */
-
-FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL )
-
-/* Drag in local definitions */
-#include "librm.h"
-
-/* For switches to/from protected mode */
-#define CR0_PE 1
-
-/* Size of various C data structures */
-#define SIZEOF_I386_SEG_REGS	12
-#define SIZEOF_I386_REGS	32
-#define SIZEOF_REAL_MODE_REGS	( SIZEOF_I386_SEG_REGS + SIZEOF_I386_REGS )
-#define SIZEOF_I386_FLAGS	4
-#define SIZEOF_I386_ALL_REGS	( SIZEOF_REAL_MODE_REGS + SIZEOF_I386_FLAGS )
-	
-	.arch i386
-
-/****************************************************************************
- * Global descriptor table
- *
- * Call init_librm to set up the GDT before attempting to use any
- * protected-mode code.
- *
- * NOTE: This must be located before prot_to_real, otherwise gas
- * throws a "can't handle non absolute segment in `ljmp'" error due to
- * not knowing the value of REAL_CS when the ljmp is encountered.
- *
- * Note also that putting ".word gdt_end - gdt - 1" directly into
- * gdt_limit, rather than going via gdt_length, will also produce the
- * "non absolute segment" error.  This is most probably a bug in gas.
- ****************************************************************************
- */
-	.section ".data16", "aw", @progbits
-	.align 16
-gdt:
-gdtr:		/* The first GDT entry is unused, the GDTR can fit here. */
-gdt_limit:		.word gdt_length - 1
-gdt_base:		.long 0
-			.word 0 /* padding */
-
-	.org	gdt + VIRTUAL_CS, 0
-virtual_cs:	/* 32 bit protected mode code segment, virtual addresses */
-	.word	0xffff, 0
-	.byte	0, 0x9f, 0xcf, 0
-
-	.org	gdt + VIRTUAL_DS, 0
-virtual_ds:	/* 32 bit protected mode data segment, virtual addresses */
-	.word	0xffff, 0
-	.byte	0, 0x93, 0xcf, 0
-	
-	.org	gdt + PHYSICAL_CS, 0
-physical_cs:	/* 32 bit protected mode code segment, physical addresses */
-	.word	0xffff, 0
-	.byte	0, 0x9f, 0xcf, 0
-
-	.org	gdt + PHYSICAL_DS, 0
-physical_ds:	/* 32 bit protected mode data segment, physical addresses */
-	.word	0xffff, 0
-	.byte	0, 0x93, 0xcf, 0	
-
-	.org	gdt + REAL_CS, 0
-real_cs: 	/* 16 bit real mode code segment */
-	.word	0xffff, 0
-	.byte	0, 0x9b, 0x00, 0
-
-	.org	gdt + REAL_DS	
-real_ds:	/* 16 bit real mode data segment */
-	.word	0xffff, ( REAL_DS << 4 )
-	.byte	0, 0x93, 0x00, 0
-
-gdt_end:
-	.equ	gdt_length, gdt_end - gdt
-
-/****************************************************************************
- * init_librm (real-mode far call, 16-bit real-mode far return address)
- *
- * Initialise the GDT ready for transitions to protected mode.
- *
- * Parameters:
- *   %cs : .text16 segment
- *   %ds : .data16 segment
- *   %edi : Physical base of protected-mode code (virt_offset)
- ****************************************************************************
- */
-	.section ".text16", "ax", @progbits
-	.code16
-	.globl init_librm
-init_librm:
-	/* Preserve registers */
-	pushl	%eax
-	pushl	%ebx
-
-	/* Store virt_offset and set up virtual_cs and virtual_ds segments */
-	movl	%edi, %eax
-	movw	$virtual_cs, %bx
-	call	set_seg_base
-	movw	$virtual_ds, %bx
-	call	set_seg_base	
-	movl	%edi, rm_virt_offset
-
-	/* Negate virt_offset */
-	negl	%edi
-		
-	/* Store rm_cs and text16, set up real_cs segment */
-	xorl	%eax, %eax
-	movw	%cs, %ax
-	movw	%ax, %cs:rm_cs
-	shll	$4, %eax
-	movw	$real_cs, %bx
-	call	set_seg_base
-	addr32 leal	(%eax, %edi), %ebx
-	movl	%ebx, rm_text16
-
-	/* Store rm_ds and data16 */
-	xorl	%eax, %eax
-	movw	%ds, %ax
-	movw	%ax, %cs:rm_ds
-	shll	$4, %eax
-	addr32 leal	(%eax, %edi), %ebx
-	movl	%ebx, rm_data16
-
-	/* Set GDT base */
-	movl	%eax, gdt_base
-	addl	$gdt, gdt_base
-
-	/* Initialise IDT */
-	pushl	$init_idt
-	pushw	%cs
-	call	prot_call
-	popl	%eax /* discard */
-
-	/* Restore registers */
-	negl	%edi
-	popl	%ebx
-	popl	%eax
-	lret
-
-	.section ".text16", "ax", @progbits
-	.code16
-set_seg_base:
-1:	movw	%ax, 2(%bx)
-	rorl	$16, %eax
-	movb	%al, 4(%bx)
-	movb	%ah, 7(%bx)
-	roll	$16, %eax
-	ret
-
-/****************************************************************************
- * real_to_prot (real-mode near call, 32-bit virtual return address)
- *
- * Switch from 16-bit real-mode to 32-bit protected mode with virtual
- * addresses.  The real-mode %ss:sp is stored in rm_ss and rm_sp, and
- * the protected-mode %esp is restored from the saved pm_esp.
- * Interrupts are disabled.  All other registers may be destroyed.
- *
- * The return address for this function should be a 32-bit virtual
- * address.
- *
- * Parameters: 
- *   %ecx : number of bytes to move from RM stack to PM stack
- *
- ****************************************************************************
- */
-	.section ".text16", "ax", @progbits
-	.code16
-real_to_prot:
-	/* Enable A20 line */
-	call	enable_a20
-	/* A failure at this point is fatal, and there's nothing we
-	 * can do about it other than lock the machine to make the
-	 * problem immediately visible.
-	 */
-1:	jc	1b
-
-	/* Make sure we have our data segment available */
-	movw	%cs:rm_ds, %ax
-	movw	%ax, %ds
-
-	/* Add virt_offset, text16 and data16 to stack to be
-	 * copied, and also copy the return address.
-	 */
-	pushl	rm_virt_offset
-	pushl	rm_text16
-	pushl	rm_data16
-	addw	$16, %cx /* %ecx must be less than 64kB anyway */
-
-	/* Real-mode %ss:%sp => %ebp:%edx and virtual address => %esi */
-	xorl	%ebp, %ebp
-	movw	%ss, %bp
-	movzwl	%sp, %edx
-	movl	%ebp, %eax
-	shll	$4, %eax
-	addr32 leal (%eax,%edx), %esi
-	subl	rm_virt_offset, %esi
-
-	/* Load protected-mode global descriptor table */
-	data32 lgdt gdtr
-
-	/* Zero segment registers.  This wastes around 12 cycles on
-	 * real hardware, but saves a substantial number of emulated
-	 * instructions under KVM.
-	 */
-	xorw	%ax, %ax
-	movw	%ax, %ds
-	movw	%ax, %es
-	movw	%ax, %fs
-	movw	%ax, %gs
-	movw	%ax, %ss
-
-	/* Switch to protected mode */
-	cli
-	movl	%cr0, %eax
-	orb	$CR0_PE, %al
-	movl	%eax, %cr0
-	data32 ljmp	$VIRTUAL_CS, $r2p_pmode
-	.section ".text", "ax", @progbits
-	.code32
-r2p_pmode:
-	/* Set up protected-mode data segments and stack pointer */
-	movw	$VIRTUAL_DS, %ax
-	movw	%ax, %ds
-	movw	%ax, %es
-	movw	%ax, %fs
-	movw	%ax, %gs
-	movw	%ax, %ss
-	movl	pm_esp, %esp
-
-	/* Load protected-mode interrupt descriptor table */
-	lidt	idtr
-
-	/* Record real-mode %ss:sp (after removal of data) */
-	movw	%bp, rm_ss
-	addl	%ecx, %edx
-	movw	%dx, rm_sp
-
-	/* Move data from RM stack to PM stack */
-	subl	%ecx, %esp
-	movl	%esp, %edi
-	rep movsb
-
-	/* Publish virt_offset, text16 and data16 for PM code to use */
-	popl	data16
-	popl	text16
-	popl	virt_offset
-
-	/* Return to virtual address */
-	ret
-
-/****************************************************************************
- * prot_to_real (protected-mode near call, 32-bit real-mode return address)
- *
- * Switch from 32-bit protected mode with virtual addresses to 16-bit
- * real mode.  The protected-mode %esp is stored in pm_esp and the
- * real-mode %ss:sp is restored from the saved rm_ss and rm_sp.  The
- * high word of the real-mode %esp is set to zero.  All real-mode data
- * segment registers are loaded from the saved rm_ds.  Interrupts are
- * *not* enabled, since we want to be able to use prot_to_real in an
- * ISR.  All other registers may be destroyed.
- *
- * The return address for this function should be a 32-bit (sic)
- * real-mode offset within .code16.
- *
- * Parameters: 
- *   %ecx : number of bytes to move from PM stack to RM stack
- *   %esi : real-mode global and interrupt descriptor table registers
- *
- ****************************************************************************
- */
-	.section ".text", "ax", @progbits
-	.code32
-prot_to_real:
-	/* Copy real-mode global descriptor table register to RM code segment */
-	movl	text16, %edi
-	leal	rm_gdtr(%edi), %edi
-	movsw
-	movsl
-
-	/* Load real-mode interrupt descriptor table register */
-	lidt	(%esi)
-
-	/* Add return address to data to be moved to RM stack */
-	addl	$4, %ecx
-	
-	/* Real-mode %ss:sp => %ebp:edx and virtual address => %edi */
-	movzwl	rm_ss, %ebp
-	movzwl	rm_sp, %edx
-	subl	%ecx, %edx
-	movl	%ebp, %eax
-	shll	$4, %eax
-	leal	(%eax,%edx), %edi
-	subl	virt_offset, %edi
-	
-	/* Move data from PM stack to RM stack */
-	movl	%esp, %esi
-	rep movsb
-	
-	/* Record protected-mode %esp (after removal of data) */
-	movl	%esi, pm_esp
-
-	/* Load real-mode segment limits */
-	movw	$REAL_DS, %ax
-	movw	%ax, %ds
-	movw	%ax, %es
-	movw	%ax, %fs
-	movw	%ax, %gs
-	movw	%ax, %ss
-	ljmp	$REAL_CS, $p2r_rmode
-	.section ".text16", "ax", @progbits
-	.code16
-p2r_rmode:
-	/* Load real-mode GDT */
-	data32 lgdt %cs:rm_gdtr
-	/* Switch to real mode */
-	movl	%cr0, %eax
-	andb	$0!CR0_PE, %al
-	movl	%eax, %cr0
-p2r_ljmp_rm_cs:
-	ljmp	$0, $1f
-1:
-	/* Set up real-mode data segments and stack pointer */
-	movw	%cs:rm_ds, %ax
-	movw	%ax, %ds
-	movw	%ax, %es
-	movw	%ax, %fs
-	movw	%ax, %gs
-	movw	%bp, %ss
-	movl	%edx, %esp
-
-	/* Return to real-mode address */
-	data32 ret
-
-
-	/* Real-mode code and data segments.  Assigned by the call to
-	 * init_librm.  rm_cs doubles as the segment part of the jump
-	 * instruction used by prot_to_real.  Both are located in
-	 * .text16 rather than .data16: rm_cs since it forms part of
-	 * the jump instruction within the code segment, and rm_ds
-	 * since real-mode code needs to be able to locate the data
-	 * segment with no other reference available.
-	 */
-	.globl rm_cs
-	.equ	rm_cs, ( p2r_ljmp_rm_cs + 3 )
-
-	.section ".text16.data", "aw", @progbits
-	.globl rm_ds
-rm_ds:	.word 0
-
-	/* Real-mode global and interrupt descriptor table registers */
-	.section ".text16.data", "aw", @progbits
-rm_gdtr:
-	.word 0 /* Limit */
-	.long 0 /* Base */
-
-/****************************************************************************
- * prot_call (real-mode far call, 16-bit real-mode far return address)
- *
- * Call a specific C function in the protected-mode code.  The
- * prototype of the C function must be
- *   void function ( struct i386_all_regs *ix86 ); 
- * ix86 will point to a struct containing the real-mode registers
- * at entry to prot_call.  
- *
- * All registers will be preserved across prot_call(), unless the C
- * function explicitly overwrites values in ix86.  Interrupt status
- * and GDT will also be preserved.  Gate A20 will be enabled.
- *
- * Note that prot_call() does not rely on the real-mode stack
- * remaining intact in order to return, since everything relevant is
- * copied to the protected-mode stack for the duration of the call.
- * In particular, this means that a real-mode prefix can make a call
- * to main() which will return correctly even if the prefix's stack
- * gets vapourised during the Etherboot run.  (The prefix cannot rely
- * on anything else on the stack being preserved, so should move any
- * critical data to registers before calling main()).
- *
- * Parameters:
- *   function : virtual address of protected-mode function to call
- *
- * Example usage:
- *	pushl	$pxe_api_call
- *	call	prot_call
- *	addw	$4, %sp
- * to call in to the C function
- *      void pxe_api_call ( struct i386_all_regs *ix86 );
- ****************************************************************************
- */
-
-#define PC_OFFSET_GDT ( 0 )
-#define PC_OFFSET_IDT ( PC_OFFSET_GDT + 6 )
-#define PC_OFFSET_IX86 ( PC_OFFSET_IDT + 6 )
-#define PC_OFFSET_RETADDR ( PC_OFFSET_IX86 + SIZEOF_I386_ALL_REGS )
-#define PC_OFFSET_FUNCTION ( PC_OFFSET_RETADDR + 4 )
-#define PC_OFFSET_END ( PC_OFFSET_FUNCTION + 4 )
-
-	.section ".text16", "ax", @progbits
-	.code16
-	.globl prot_call
-prot_call:
-	/* Preserve registers, flags and GDT on external RM stack */
-	pushfl
-	pushal
-	pushw	%gs
-	pushw	%fs
-	pushw	%es
-	pushw	%ds
-	pushw	%ss
-	pushw	%cs
-	subw	$PC_OFFSET_IX86, %sp
-	movw	%sp, %bp
-	sidt	PC_OFFSET_IDT(%bp)
-	sgdt	PC_OFFSET_GDT(%bp)
-
-	/* For sanity's sake, clear the direction flag as soon as possible */
-	cld
-
-	/* Switch to protected mode and move register dump to PM stack */
-	movl	$PC_OFFSET_END, %ecx
-	pushl	$pc_pmode
-	jmp	real_to_prot
-	.section ".text", "ax", @progbits
-	.code32
-pc_pmode:
-	/* Call function */
-	leal	PC_OFFSET_IX86(%esp), %eax
-	pushl	%eax
-	call	*(PC_OFFSET_FUNCTION+4)(%esp)
-	popl	%eax /* discard */
-
-	/* Switch to real mode and move register dump back to RM stack */
-	movl	$PC_OFFSET_END, %ecx
-	movl	%esp, %esi
-	pushl	$pc_rmode
-	jmp	prot_to_real
-	.section ".text16", "ax", @progbits
-	.code16
-pc_rmode:
-	/* Restore registers and flags and return */
-	addw	$( PC_OFFSET_IX86 + 4 /* also skip %cs and %ss */ ), %sp
-	popw	%ds
-	popw	%es
-	popw	%fs
-	popw	%gs
-	popal
-	/* popal skips %esp.  We therefore want to do "movl -20(%sp),
-	 * %esp", but -20(%sp) is not a valid 80386 expression.
-	 * Fortunately, prot_to_real() zeroes the high word of %esp, so
-	 * we can just use -20(%esp) instead.
-	 */
-	addr32 movl -20(%esp), %esp
-	popfl
-	lret
-
-/****************************************************************************
- * real_call (protected-mode near call, 32-bit virtual return address)
- *
- * Call a real-mode function from protected-mode code.
- *
- * The non-segment register values will be passed directly to the
- * real-mode code.  The segment registers will be set as per
- * prot_to_real.  The non-segment register values set by the real-mode
- * function will be passed back to the protected-mode caller.  A
- * result of this is that this routine cannot be called directly from
- * C code, since it clobbers registers that the C ABI expects the
- * callee to preserve.
- *
- * librm.h defines a convenient macro REAL_CODE() for using real_call.
- * See librm.h and realmode.h for details and examples.
- *
- * Parameters:
- *   (32-bit) near pointer to real-mode function to call
- *
- * Returns: none
- ****************************************************************************
- */
-
-#define RC_OFFSET_PRESERVE_REGS ( 0 )
-#define RC_OFFSET_RETADDR ( RC_OFFSET_PRESERVE_REGS + SIZEOF_I386_REGS )
-#define RC_OFFSET_FUNCTION ( RC_OFFSET_RETADDR + 4 )
-#define RC_OFFSET_END ( RC_OFFSET_FUNCTION + 4 )
-
-	.section ".text", "ax", @progbits
-	.code32
-	.globl real_call
-real_call:
-	/* Create register dump and function pointer copy on PM stack */
-	pushal
-	pushl	RC_OFFSET_FUNCTION(%esp)
-
-	/* Switch to real mode and move register dump to RM stack  */
-	movl	$( RC_OFFSET_RETADDR + 4 /* function pointer copy */ ), %ecx
-	pushl	$rc_rmode
-	movl	$rm_default_gdtr_idtr, %esi
-	jmp	prot_to_real
-	.section ".text16", "ax", @progbits
-	.code16
-rc_rmode:
-	/* Call real-mode function */
-	popl	rc_function
-	popal
-	call	*rc_function
-	pushal
-
-	/* For sanity's sake, clear the direction flag as soon as possible */
-	cld
-
-	/* Switch to protected mode and move register dump back to PM stack */
-	movl	$RC_OFFSET_RETADDR, %ecx
-	pushl	$rc_pmode
-	jmp	real_to_prot
-	.section ".text", "ax", @progbits
-	.code32
-rc_pmode:
-	/* Restore registers and return */
-	popal
-	ret
-
-
-	/* Function vector, used because "call xx(%sp)" is not a valid
-	 * 16-bit expression.
-	 */
-	.section ".data16", "aw", @progbits
-rc_function:	.word 0, 0
-
-	/* Default real-mode global and interrupt descriptor table registers */
-	.section ".data", "aw", @progbits
-rm_default_gdtr_idtr:
-	.word 0		/* Global descriptor table limit */
-	.long 0		/* Global descriptor table base */
-	.word 0x03ff	/* Interrupt descriptor table limit */
-	.long 0		/* Interrupt descriptor table base */
-
-/****************************************************************************
- * flatten_real_mode (real-mode near call)
- *
- * Switch to flat real mode
- *
- ****************************************************************************
- */
-	.section ".text16", "ax", @progbits
-	.code16
-	.globl flatten_real_mode
-flatten_real_mode:
-	/* Modify GDT to use flat real mode */
-	movb	$0x8f, real_cs + 6
-	movb	$0x8f, real_ds + 6
-	/* Call dummy protected-mode function */
-	pushl	$flatten_dummy
-	pushw	%cs
-	call	prot_call
-	addw	$4, %sp
-	/* Restore GDT */
-	movb	$0x00, real_cs + 6
-	movb	$0x00, real_ds + 6
-	/* Return */
-	ret
-
-	.section ".text", "ax", @progbits
-	.code32
-flatten_dummy:
-	ret
-
-/****************************************************************************
- * Interrupt wrapper
- *
- * Used by the protected-mode interrupt vectors to call the
- * interrupt() function.
- *
- * May be entered with either physical or virtual stack segment.
- ****************************************************************************
- */
-	.globl interrupt_wrapper
-interrupt_wrapper:
-	/* Preserve segment registers and original %esp */
-	pushl	%ds
-	pushl	%es
-	pushl	%fs
-	pushl	%gs
-	pushl	%ss
-	pushl	%esp
-
-	/* Switch to virtual addressing */
-	call	_intr_to_virt
-
-	/* Expand IRQ number to whole %eax register */
-	movzbl	%al, %eax
-
-	/* Call interrupt handler */
-	call	interrupt
-
-	/* Restore original stack and segment registers */
-	lss	(%esp), %esp
-	popl	%ss
-	popl	%gs
-	popl	%fs
-	popl	%es
-	popl	%ds
-
-	/* Restore registers and return */
-	popal
-	iret
-
-/****************************************************************************
- * Stored real-mode and protected-mode stack pointers
- *
- * The real-mode stack pointer is stored here whenever real_to_prot
- * is called and restored whenever prot_to_real is called.  The
- * converse happens for the protected-mode stack pointer.
- *
- * Despite initial appearances this scheme is, in fact re-entrant,
- * because program flow dictates that we always return via the point
- * we left by.  For example:
- *    PXE API call entry
- *  1   real => prot
- *        ...
- *        Print a text string
- *	    ...
- *  2       prot => real
- *            INT 10
- *  3       real => prot
- *	    ...
- *        ...
- *  4   prot => real
- *    PXE API call exit
- *
- * At point 1, the RM mode stack value, say RPXE, is stored in
- * rm_ss,sp.  We want this value to still be present in rm_ss,sp when
- * we reach point 4.
- *
- * At point 2, the RM stack value is restored from RPXE.  At point 3,
- * the RM stack value is again stored in rm_ss,sp.  This *does*
- * overwrite the RPXE that we have stored there, but it's the same
- * value, since the code between points 2 and 3 has managed to return
- * to us.
- ****************************************************************************
- */
-	.section ".data", "aw", @progbits
-	.globl rm_sp
-rm_sp:	.word 0
-	.globl rm_ss
-rm_ss:	.word 0
-pm_esp:	.long _estack
-
-/****************************************************************************
- * Virtual address offsets
- *
- * These are used by the protected-mode code to map between virtual
- * and physical addresses, and to access variables in the .text16 or
- * .data16 segments.
- ****************************************************************************
- */
-	/* Internal copies, created by init_librm (which runs in real mode) */
-	.section ".data16", "aw", @progbits
-rm_virt_offset:	.long 0
-rm_text16:	.long 0
-rm_data16:	.long 0
-
-	/* Externally-visible copies, created by real_to_prot */
-	.section ".data", "aw", @progbits
-	.globl virt_offset
-virt_offset:	.long 0	
-	.globl text16
-text16:		.long 0
-	.globl data16
-data16:		.long 0

src/arch/i386/transitions/librm_mgmt.c

@@ -1,158 +0,0 @@
-/*
- * librm: a library for interfacing to real-mode code
- *
- * Michael Brown <mbrown@fensystems.co.uk>
- *
- */
-
-FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
-
-#include <stdint.h>
-#include <ipxe/profile.h>
-#include <realmode.h>
-#include <pic8259.h>
-
-/*
- * This file provides functions for managing librm.
- *
- */
-
-/** The interrupt wrapper */
-extern char interrupt_wrapper[];
-
-/** The interrupt vectors */
-static struct interrupt_vector intr_vec[NUM_INT];
-
-/** The interrupt descriptor table */
-struct interrupt_descriptor idt[NUM_INT] __attribute__ (( aligned ( 16 ) ));
-
-/** The interrupt descriptor table register */
-struct idtr idtr = {
-	.limit = ( sizeof ( idt ) - 1 ),
-};
-
-/** Timer interrupt profiler */
-static struct profiler timer_irq_profiler __profiler = { .name = "irq.timer" };
-
-/** Other interrupt profiler */
-static struct profiler other_irq_profiler __profiler = { .name = "irq.other" };
-
-/**
- * Allocate space on the real-mode stack and copy data there from a
- * user buffer
- *
- * @v data		User buffer
- * @v size		Size of stack data
- * @ret sp		New value of real-mode stack pointer
- */
-uint16_t copy_user_to_rm_stack ( userptr_t data, size_t size ) {
-	userptr_t rm_stack;
-	rm_sp -= size;
-	rm_stack = real_to_user ( rm_ss, rm_sp );
-	memcpy_user ( rm_stack, 0, data, 0, size );
-	return rm_sp;
-};
-
-/**
- * Deallocate space on the real-mode stack, optionally copying back
- * data to a user buffer.
- *
- * @v data		User buffer
- * @v size		Size of stack data
- */
-void remove_user_from_rm_stack ( userptr_t data, size_t size ) {
-	if ( data ) {
-		userptr_t rm_stack = real_to_user ( rm_ss, rm_sp );
-		memcpy_user ( rm_stack, 0, data, 0, size );
-	}
-	rm_sp += size;
-};
-
-/**
- * Set interrupt vector
- *
- * @v intr		Interrupt number
- * @v vector		Interrupt vector, or NULL to disable
- */
-void set_interrupt_vector ( unsigned int intr, void *vector ) {
-	struct interrupt_descriptor *idte;
-
-	idte = &idt[intr];
-	idte->segment = VIRTUAL_CS;
-	idte->attr = ( vector ? ( IDTE_PRESENT | IDTE_TYPE_IRQ32 ) : 0 );
-	idte->low = ( ( ( uint32_t ) vector ) & 0xffff );
-	idte->high = ( ( ( uint32_t ) vector ) >> 16 );
-}
-
-/**
- * Initialise interrupt descriptor table
- *
- */
-void init_idt ( void ) {
-	struct interrupt_vector *vec;
-	unsigned int intr;
-
-	/* Initialise the interrupt descriptor table and interrupt vectors */
-	for ( intr = 0 ; intr < NUM_INT ; intr++ ) {
-		vec = &intr_vec[intr];
-		vec->pushal = PUSHAL_INSN;
-		vec->movb = MOVB_INSN;
-		vec->intr = intr;
-		vec->jmp = JMP_INSN;
-		vec->offset = ( ( uint32_t ) interrupt_wrapper -
-				( uint32_t ) vec->next );
-		set_interrupt_vector ( intr, vec );
-	}
-	DBGC ( &intr_vec[0], "INTn vector at %p+%zxn (phys %#lx+%zxn)\n",
-	       intr_vec, sizeof ( intr_vec[0] ),
-	       virt_to_phys ( intr_vec ), sizeof ( intr_vec[0] ) );
-
-	/* Initialise the interrupt descriptor table register */
-	idtr.base = virt_to_phys ( idt );
-}
-
-/**
- * Determine interrupt profiler (for debugging)
- *
- * @v intr		Interrupt number
- * @ret profiler	Profiler
- */
-static struct profiler * interrupt_profiler ( int intr ) {
-
-	switch ( intr ) {
-	case IRQ_INT ( 0 ) :
-		return &timer_irq_profiler;
-	default:
-		return &other_irq_profiler;
-	}
-}
-
-/**
- * Interrupt handler
- *
- * @v intr		Interrupt number
- */
-void __attribute__ (( cdecl, regparm ( 1 ) )) interrupt ( int intr ) {
-	struct profiler *profiler = interrupt_profiler ( intr );
-	uint32_t discard_eax;
-
-	/* Reissue interrupt in real mode */
-	profile_start ( profiler );
-	__asm__ __volatile__ ( REAL_CODE ( "movb %%al, %%cs:(1f + 1)\n\t"
-					   "\n1:\n\t"
-					   "int $0x00\n\t" )
-			       : "=a" ( discard_eax ) : "0" ( intr ) );
-	profile_stop ( profiler );
-	profile_exclude ( profiler );
-}
-
-PROVIDE_UACCESS_INLINE ( librm, phys_to_user );
-PROVIDE_UACCESS_INLINE ( librm, user_to_phys );
-PROVIDE_UACCESS_INLINE ( librm, virt_to_user );
-PROVIDE_UACCESS_INLINE ( librm, user_to_virt );
-PROVIDE_UACCESS_INLINE ( librm, userptr_add );
-PROVIDE_UACCESS_INLINE ( librm, memcpy_user );
-PROVIDE_UACCESS_INLINE ( librm, memmove_user );
-PROVIDE_UACCESS_INLINE ( librm, memset_user );
-PROVIDE_UACCESS_INLINE ( librm, strlen_user );
-PROVIDE_UACCESS_INLINE ( librm, memchr_user );

src/arch/x86/Makefile

@@ -1,3 +1,8 @@
+# Assembler section type character
+#
+ASM_TCHAR	:= @
+ASM_TCHAR_OPS	:= @
+
 # Include common x86 headers
 #
 INCDIRS		+= arch/x86/include
@@ -5,11 +10,17 @@ INCDIRS		+= arch/x86/include
 # x86-specific directories containing source files
 #
 SRCDIRS		+= arch/x86/core
+SRCDIRS		+= arch/x86/image
+SRCDIRS		+= arch/x86/interface/pcbios
+SRCDIRS		+= arch/x86/interface/pxe
 SRCDIRS		+= arch/x86/interface/efi
+SRCDIRS		+= arch/x86/interface/vmware
+SRCDIRS		+= arch/x86/interface/syslinux
 SRCDIRS		+= arch/x86/prefix
 SRCDIRS		+= arch/x86/hci/commands
 SRCDIRS		+= arch/x86/drivers/xen
 SRCDIRS		+= arch/x86/drivers/hyperv
+SRCDIRS		+= arch/x86/transitions
 
 # breaks building some of the linux-related objects
 CFLAGS		+= -Ulinux
@@ -17,6 +28,10 @@ CFLAGS		+= -Ulinux
 # disable valgrind
 CFLAGS		+= -DNVALGRIND
 
+# Define version string for lkrnprefix.S
+#
+CFLAGS_lkrnprefix	+= -DVERSION="\"$(VERSION)\""
+
 # Include Hyper-V driver in the all-drivers build
 #
 DRIVERS_hyperv += hyperv

src/arch/x86/Makefile.efi

@@ -1,42 +1,6 @@
 # -*- makefile -*- : Force emacs to use Makefile mode
 
-# The EFI linker script
+# Include generic EFI Makefile
 #
-LDSCRIPT	= arch/x86/scripts/efi.lds
-
-# Retain relocation information for elf2efi
-#
-LDFLAGS		+= -q -S
-
-# Media types.
-#
-NON_AUTO_MEDIA	+= efi
-NON_AUTO_MEDIA	+= efidrv
-NON_AUTO_MEDIA	+= drv.efi
-NON_AUTO_MEDIA	+= efirom
-
-# Include SNP driver in the all-drivers build
-#
-DRIVERS_net += snp
-
-# Rules for building EFI files
-#
-$(BIN)/%.efi : $(BIN)/%.efi.tmp $(ELF2EFI)
-	$(QM)$(ECHO) "  [FINISH] $@"
-	$(Q)$(ELF2EFI) --subsystem=10 $< $@
-
-$(BIN)/%.efidrv : $(BIN)/%.efidrv.tmp $(ELF2EFI)
-	$(QM)$(ECHO) "  [FINISH] $@"
-	$(Q)$(ELF2EFI) --subsystem=11 $< $@
-
-$(BIN)/%.drv.efi : $(BIN)/%.efidrv
-	$(QM)$(ECHO) "  [FINISH] $@"
-	$(Q)$(CP) $< $@
-
-$(BIN)/%.efirom : $(BIN)/%.efidrv $(EFIROM)
-	$(QM)$(ECHO) "  [FINISH] $@"
-	$(Q)$(EFIROM) -v $(TGT_PCI_VENDOR) -d $(TGT_PCI_DEVICE) $< $@
-
-$(BIN)/efidrv.cab : $(BIN)/alldrv.efis # $(ALL_drv.efi) is not yet defined
-	$(QM)$(ECHO) "  [CAB] $@"
-	$(Q)$(LCAB) -n -q $(ALL_drv.efi) $@
+MAKEDEPS	+= Makefile.efi
+include Makefile.efi

src/arch/x86/Makefile.pcbios

@@ -0,0 +1,132 @@
+# -*- makefile -*- : Force emacs to use Makefile mode
+
+# BIOS-specific directories containing source files
+#
+SRCDIRS		+= arch/x86/drivers/net
+
+# The i386 linker script
+#
+LDSCRIPT	= arch/x86/scripts/pcbios.lds
+
+# Stop ld from complaining about our customised linker script
+#
+LDFLAGS		+= -N --no-check-sections
+
+# Prefix always starts at address zero
+#
+LDFLAGS		+= --section-start=.prefix=0
+
+# Media types.
+#
+MEDIA		+= rom
+MEDIA		+= mrom
+MEDIA		+= pcirom
+MEDIA		+= isarom
+MEDIA		+= pxe
+MEDIA		+= kpxe
+MEDIA		+= kkpxe
+MEDIA		+= kkkpxe
+MEDIA		+= lkrn
+MEDIA		+= dsk
+MEDIA		+= nbi
+MEDIA		+= hd
+MEDIA		+= raw
+MEDIA		+= exe
+
+# Padding rules
+#
+PAD_rom		= $(PERL) $(PADIMG) --blksize=512 --byte=0xff
+PAD_mrom	= $(PAD_rom)
+PAD_pcirom	= $(PAD_rom)
+PAD_isarom	= $(PAD_rom)
+PAD_dsk		= $(PERL) $(PADIMG) --blksize=512
+PAD_hd		= $(PERL) $(PADIMG) --blksize=32768
+PAD_exe		= $(PERL) $(PADIMG) --blksize=512
+
+# Finalisation rules
+#
+FINALISE_rom	= $(PERL) $(FIXROM)
+FINALISE_mrom	= $(FINALISE_rom)
+FINALISE_pcirom	= $(FINALISE_rom)
+FINALISE_isarom	= $(FINALISE_rom)
+
+# Use $(ROMS) rather than $(DRIVERS) for "allroms", "allmroms", etc.
+#
+LIST_NAME_rom := ROMS
+LIST_NAME_mrom := ROMS
+LIST_NAME_pcirom := ROMS
+LIST_NAME_isarom := ROMS
+
+# Locations of isolinux files
+#
+SYSLINUX_DIR_LIST	:= \
+	/usr/lib/syslinux \
+	/usr/lib/syslinux/bios \
+	/usr/lib/syslinux/modules/bios \
+	/usr/share/syslinux \
+	/usr/share/syslinux/bios \
+	/usr/share/syslinux/modules/bios \
+	/usr/local/share/syslinux \
+	/usr/local/share/syslinux/bios \
+	/usr/local/share/syslinux/modules/bios \
+	/usr/lib/ISOLINUX
+ISOLINUX_BIN_LIST	:= \
+	$(ISOLINUX_BIN) \
+	$(patsubst %,%/isolinux.bin,$(SYSLINUX_DIR_LIST))
+LDLINUX_C32_LIST	:= \
+	$(LDLINUX_C32) \
+	$(patsubst %,%/ldlinux.c32,$(SYSLINUX_DIR_LIST))
+ISOLINUX_BIN	= $(firstword $(wildcard $(ISOLINUX_BIN_LIST)))
+LDLINUX_C32	= $(firstword $(wildcard $(LDLINUX_C32_LIST)))
+
+# rule to make a non-emulation ISO boot image
+NON_AUTO_MEDIA	+= iso
+%iso:	%lkrn util/geniso
+	$(QM)$(ECHO) "  [GENISO] $@"
+	$(Q)ISOLINUX_BIN=$(ISOLINUX_BIN) LDLINUX_C32=$(LDLINUX_C32) \
+	    VERSION="$(VERSION)" bash util/geniso -o $@ $<
+
+# rule to make a floppy emulation ISO boot image
+NON_AUTO_MEDIA	+= liso
+%liso:	%lkrn util/geniso
+	$(QM)$(ECHO) "  [GENISO] $@"
+	$(Q)VERSION="$(VERSION)" bash util/geniso -l -o $@ $<
+
+# rule to make a syslinux floppy image (mountable, bootable)
+NON_AUTO_MEDIA	+= sdsk
+%sdsk:	%lkrn util/gensdsk
+	$(QM)$(ECHO) "  [GENSDSK] $@"
+	$(Q)bash util/gensdsk $@ $<
+
+# rule to write disk images to /dev/fd0
+NON_AUTO_MEDIA	+= fd0
+%fd0 : %dsk
+	$(QM)$(ECHO) "  [DD] $@"
+	$(Q)dd if=$< bs=512 conv=sync of=/dev/fd0
+	$(Q)sync
+
+# Special target for building Master Boot Record binary
+$(BIN)/mbr.tmp : $(BIN)/mbr.o
+	$(QM)$(ECHO) "  [LD] $@"
+	$(Q)$(LD) $(LDFLAGS) -o $@ -e mbr $<
+
+# rule to make a USB disk image
+$(BIN)/usbdisk.tmp : $(BIN)/usbdisk.o
+	$(QM)$(ECHO) "  [LD] $@"
+	$(Q)$(LD) $(LDFLAGS) -o $@ -e mbr $<
+
+NON_AUTO_MEDIA	+= usb
+%usb: $(BIN)/usbdisk.bin %hd
+	$(QM)$(ECHO) "  [FINISH] $@"
+	$(Q)cat $^ > $@
+
+NON_AUTO_MEDIA	+= vhd
+%vhd: %usb
+	$(QM)$(ECHO) "  [FINISH] $@"
+	$(Q)$(QEMUIMG) convert -f raw -O vpc $< $@
+
+# Padded floppy image (e.g. for iLO)
+NON_AUTO_MEDIA += pdsk
+%pdsk : %dsk
+	$(Q)cp $< $@
+	$(Q)$(PADIMG) --blksize=1474560 $@

src/arch/x86/core/cachedhcp.c

@@ -58,6 +58,7 @@ static void cachedhcp_init ( void ) {
 	struct dhcp_packet *dhcppkt;
 	struct dhcp_packet *tmp;
 	struct dhcphdr *dhcphdr;
+	size_t max_len;
 	size_t len;
 
 	/* Do nothing if no cached DHCPACK is present */
@@ -69,23 +70,25 @@ static void cachedhcp_init ( void ) {
 	/* No reliable way to determine length before parsing packet;
 	 * start by assuming maximum length permitted by PXE.
 	 */
-	len = sizeof ( BOOTPLAYER_t );
+	max_len = sizeof ( BOOTPLAYER_t );
 
 	/* Allocate and populate DHCP packet */
-	dhcppkt = zalloc ( sizeof ( *dhcppkt ) + len );
+	dhcppkt = zalloc ( sizeof ( *dhcppkt ) + max_len );
 	if ( ! dhcppkt ) {
 		DBGC ( colour, "CACHEDHCP could not allocate copy\n" );
 		return;
 	}
 	dhcphdr = ( ( ( void * ) dhcppkt ) + sizeof ( *dhcppkt ) );
 	copy_from_user ( dhcphdr, phys_to_user ( cached_dhcpack_phys ), 0,
-			 len );
-	dhcppkt_init ( dhcppkt, dhcphdr, len );
+			 max_len );
+	dhcppkt_init ( dhcppkt, dhcphdr, max_len );
 
-	/* Resize packet to required length.  If reallocation fails,
-	 * just continue to use the original packet.
+	/* Shrink packet to required length.  If reallocation fails,
+	 * just continue to use the original packet and waste the
+	 * unused space.
 	 */
 	len = dhcppkt_len ( dhcppkt );
+	assert ( len <= max_len );
 	tmp = realloc ( dhcppkt, ( sizeof ( *dhcppkt ) + len ) );
 	if ( tmp )
 		dhcppkt = tmp;

src/arch/x86/core/cpuid.c

@@ -24,6 +24,7 @@
 FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
 
 #include <string.h>
+#include <errno.h>
 #include <ipxe/cpuid.h>
 
 /** @file
@@ -32,15 +33,19 @@ FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
  *
  */
 
+/** Colour for debug messages */
+#define colour 0x861d
+
 /**
  * Check whether or not CPUID instruction is supported
  *
- * @ret is_supported	CPUID instruction is supported
+ * @ret rc		Return status code
  */
-int cpuid_is_supported ( void ) {
+static int cpuid_instruction_supported ( void ) {
 	unsigned long original;
 	unsigned long inverted;
 
+	/* Check for instruction existence via flag modifiability */
 	__asm__ ( "pushf\n\t"
 		  "pushf\n\t"
 		  "pop %0\n\t"
@@ -53,7 +58,54 @@ int cpuid_is_supported ( void ) {
 		  "popf\n\t"
 		  : "=&r" ( original ), "=&r" ( inverted )
 		  : "ir" ( CPUID_FLAG ) );
-	return ( ( original ^ inverted ) & CPUID_FLAG );
+	if ( ! ( ( original ^ inverted ) & CPUID_FLAG ) ) {
+		DBGC ( colour, "CPUID instruction is not supported\n" );
+		return -ENOTSUP;
+	}
+
+	return 0;
+}
+
+/**
+ * Check whether or not CPUID function is supported
+ *
+ * @v function		CPUID function
+ * @ret rc		Return status code
+ */
+int cpuid_supported ( uint32_t function ) {
+	uint32_t max_function;
+	uint32_t discard_b;
+	uint32_t discard_c;
+	uint32_t discard_d;
+	int rc;
+
+	/* Check that CPUID instruction is available */
+	if ( ( rc = cpuid_instruction_supported() ) != 0 )
+		return rc;
+
+	/* Find highest supported function number within this family */
+	cpuid ( ( function & CPUID_EXTENDED ), 0, &max_function, &discard_b,
+		&discard_c, &discard_d );
+
+	/* Fail if maximum function number is meaningless (e.g. if we
+	 * are attempting to call an extended function on a CPU which
+	 * does not support them).
+	 */
+	if ( ( max_function & CPUID_AMD_CHECK_MASK ) !=
+	     ( function & CPUID_AMD_CHECK_MASK ) ) {
+		DBGC ( colour, "CPUID invalid maximum function %#08x\n",
+		       max_function );
+		return -EINVAL;
+	}
+
+	/* Fail if this function is not supported */
+	if ( function > max_function ) {
+		DBGC ( colour, "CPUID function %#08x not supported\n",
+		       function );
+		return -ENOTTY;
+	}
+
+	return 0;
 }
 
 /**
@@ -62,23 +114,18 @@ int cpuid_is_supported ( void ) {
  * @v features		x86 CPU features to fill in
  */
 static void x86_intel_features ( struct x86_features *features ) {
-	uint32_t max_level;
 	uint32_t discard_a;
 	uint32_t discard_b;
-	uint32_t discard_c;
-	uint32_t discard_d;
+	int rc;
 
 	/* Check that features are available via CPUID */
-	cpuid ( CPUID_VENDOR_ID, &max_level, &discard_b, &discard_c,
-		&discard_d );
-	if ( max_level < CPUID_FEATURES ) {
-		DBGC ( features, "CPUID has no Intel-defined features (max "
-		       "level %08x)\n", max_level );
+	if ( ( rc = cpuid_supported ( CPUID_FEATURES ) ) != 0 ) {
+		DBGC ( features, "CPUID has no Intel-defined features\n" );
 		return;
 	}
 
 	/* Get features */
-	cpuid ( CPUID_FEATURES, &discard_a, &discard_b,
+	cpuid ( CPUID_FEATURES, 0, &discard_a, &discard_b,
 		&features->intel.ecx, &features->intel.edx );
 	DBGC ( features, "CPUID Intel features: %%ecx=%08x, %%edx=%08x\n",
 	       features->intel.ecx, features->intel.edx );
@@ -91,27 +138,18 @@ static void x86_intel_features ( struct x86_features *features ) {
  * @v features		x86 CPU features to fill in
  */
 static void x86_amd_features ( struct x86_features *features ) {
-	uint32_t max_level;
 	uint32_t discard_a;
 	uint32_t discard_b;
-	uint32_t discard_c;
-	uint32_t discard_d;
+	int rc;
 
 	/* Check that features are available via CPUID */
-	cpuid ( CPUID_AMD_MAX_FN, &max_level, &discard_b, &discard_c,
-		&discard_d );
-	if ( ( max_level & CPUID_AMD_CHECK_MASK ) != CPUID_AMD_CHECK ) {
-		DBGC ( features, "CPUID has no extended functions\n" );
-		return;
-	}
-	if ( max_level < CPUID_AMD_FEATURES ) {
-		DBGC ( features, "CPUID has no AMD-defined features (max "
-		       "level %08x)\n", max_level );
+	if ( ( rc = cpuid_supported ( CPUID_AMD_FEATURES ) ) != 0 ) {
+		DBGC ( features, "CPUID has no AMD-defined features\n" );
 		return;
 	}
 
 	/* Get features */
-	cpuid ( CPUID_AMD_FEATURES, &discard_a, &discard_b,
+	cpuid ( CPUID_AMD_FEATURES, 0, &discard_a, &discard_b,
 		&features->amd.ecx, &features->amd.edx );
 	DBGC ( features, "CPUID AMD features: %%ecx=%08x, %%edx=%08x\n",
 	       features->amd.ecx, features->amd.edx );
@@ -127,12 +165,6 @@ void x86_features ( struct x86_features *features ) {
 	/* Clear all features */
 	memset ( features, 0, sizeof ( *features ) );
 
-	/* Check that CPUID instruction is available */
-	if ( ! cpuid_is_supported() ) {
-		DBGC ( features, "CPUID instruction is not supported\n" );
-		return;
-	}
-
 	/* Get Intel-defined features */
 	x86_intel_features ( features );
 

src/arch/x86/core/cpuid_settings.c

@@ -37,10 +37,11 @@ FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
  * CPUID settings are numerically encoded as:
  *
  *  Bit  31	Extended function
- *  Bits 30-28	Unused
- *  Bits 27-24	Number of consecutive functions to call, minus one
+ *  Bits 30-24	(bit 22 = 1) Subfunction number
+ *		(bit 22 = 0) Number of consecutive functions to call, minus one
  *  Bit  23	Return result as little-endian (used for strings)
- *  Bits 22-18	Unused
+ *  Bit  22	Interpret bits 30-24 as a subfunction number
+ *  Bits 21-18	Unused
  *  Bits 17-16	Number of registers in register array, minus one
  *  Bits 15-8	Array of register indices.  First entry in array is in
  *		bits 9-8.  Indices are 0-%eax, 1-%ebx, 2-%ecx, 3-%edx.
@@ -50,6 +51,11 @@ FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
  * extracting a single register from a single function to be encoded
  * using "cpuid/<register>.<function>", e.g. "cpuid/2.0x80000001" to
  * retrieve the value of %ecx from calling CPUID with %eax=0x80000001.
+ *
+ * A subfunction (i.e. an input value for %ecx) may be specified using
+ * "cpuid/<subfunction>.0x40.<register>.<function>".  This slightly
+ * cumbersome syntax is required in order to maintain backwards
+ * compatibility with older scripts.
  */
 
 /** CPUID setting tag register indices */
@@ -60,12 +66,18 @@ enum cpuid_registers {
 	CPUID_EDX = 3,
 };
 
+/** CPUID setting tag flags */
+enum cpuid_flags {
+	CPUID_LITTLE_ENDIAN = 0x00800000UL,
+	CPUID_USE_SUBFUNCTION = 0x00400000UL,
+};
+
 /**
  * Construct CPUID setting tag
  *
  * @v function		Starting function number
- * @v num_functions	Number of consecutive functions
- * @v little_endian	Return result as little-endian
+ * @v subfunction	Subfunction, or number of consecutive functions minus 1
+ * @v flags		Flags
  * @v num_registers	Number of registers in register array
  * @v register1		First register in register array (or zero, if empty)
  * @v register2		Second register in register array (or zero, if empty)
@@ -73,21 +85,13 @@ enum cpuid_registers {
  * @v register4		Fourth register in register array (or zero, if empty)
  * @ret tag		Setting tag
  */
-#define CPUID_TAG( function, num_functions, little_endian, num_registers, \
-		   register1, register2, register3, register4 )		  \
-	( (function) | ( ( (num_functions) - 1 ) << 24 ) |		  \
-	  ( (little_endian) << 23 ) | ( ( (num_registers) - 1) << 16 ) |  \
-	  ( (register1) << 8 ) | ( (register2) << 10 ) |		  \
+#define CPUID_TAG( function, subfunction, flags, num_registers,		\
+		   register1, register2, register3, register4 )		\
+	( (function) | ( (subfunction) << 24 ) | (flags) |		\
+	  ( ( (num_registers) - 1 ) << 16 ) |				\
+	  ( (register1) << 8 ) | ( (register2) << 10 ) |		\
 	  ( (register3) << 12 ) | ( (register4) << 14 ) )
 
-/**
- * Extract endianness from CPUID setting tag
- *
- * @v tag		Setting tag
- * @ret little_endian	Result should be returned as little-endian
- */
-#define CPUID_LITTLE_ENDIAN( tag ) ( (tag) & 0x00800000UL )
-
 /**
  * Extract starting function number from CPUID setting tag
  *
@@ -97,12 +101,12 @@ enum cpuid_registers {
 #define CPUID_FUNCTION( tag ) ( (tag) & 0x800000ffUL )
 
 /**
- * Extract number of consecutive functions from CPUID setting tag
+ * Extract subfunction number from CPUID setting tag
  *
  * @v tag		Setting tag
- * @ret num_functions	Number of consecutive functions
+ * @ret subfunction	Subfunction number
  */
-#define CPUID_NUM_FUNCTIONS( tag ) ( ( ( (tag) >> 24 ) & 0xf ) + 1 )
+#define CPUID_SUBFUNCTION( tag ) ( ( (tag) >> 24 ) & 0x7f )
 
 /**
  * Extract register array from CPUID setting tag
@@ -149,62 +153,46 @@ static int cpuid_settings_fetch ( struct settings *settings,
 				  struct setting *setting,
 				  void *data, size_t len ) {
 	uint32_t function;
-	uint32_t max_function;
+	uint32_t subfunction;
 	uint32_t num_functions;
 	uint32_t registers;
 	uint32_t num_registers;
 	uint32_t buf[4];
 	uint32_t output;
-	uint32_t discard_b;
-	uint32_t discard_c;
-	uint32_t discard_d;
 	size_t frag_len;
 	size_t result_len = 0;
-
-	/* Fail unless CPUID is supported */
-	if ( ! cpuid_is_supported() ) {
-		DBGC ( settings, "CPUID not supported\n" );
-		return -ENOTSUP;
-	}
-
-	/* Find highest supported function number within this set */
-	function = CPUID_FUNCTION ( setting->tag );
-	cpuid ( function & CPUID_EXTENDED, &max_function, &discard_b,
-		&discard_c, &discard_d );
-
-	/* Fail if maximum function number is meaningless (e.g. if we
-	 * are attempting to call an extended function on a CPU which
-	 * does not support them).
-	 */
-	if ( ( max_function & CPUID_AMD_CHECK_MASK ) !=
-	     ( function & CPUID_AMD_CHECK_MASK ) ) {
-		DBGC ( settings, "CPUID invalid maximum function\n" );
-		return -ENOTSUP;
-	}
+	int rc;
 
 	/* Call each function in turn */
-	num_functions = CPUID_NUM_FUNCTIONS ( setting->tag );
+	function = CPUID_FUNCTION ( setting->tag );
+	subfunction = CPUID_SUBFUNCTION ( setting->tag );
+	if ( setting->tag & CPUID_USE_SUBFUNCTION ) {
+		num_functions = 1;
+	} else {
+		num_functions = ( subfunction + 1 );
+		subfunction = 0;
+	}
 	for ( ; num_functions-- ; function++ ) {
 
 		/* Fail if this function is not supported */
-		if ( function > max_function ) {
-			DBGC ( settings, "CPUID function %#08x not supported\n",
-			       function );
-			return -ENOTSUP;
+		if ( ( rc = cpuid_supported ( function ) ) != 0 ) {
+			DBGC ( settings, "CPUID function %#08x not supported: "
+			       "%s\n", function, strerror ( rc ) );
+			return rc;
 		}
 
 		/* Issue CPUID */
-		cpuid ( function, &buf[CPUID_EAX], &buf[CPUID_EBX],
-			&buf[CPUID_ECX], &buf[CPUID_EDX] );
-		DBGC ( settings, "CPUID %#08x => %#08x:%#08x:%#08x:%#08x\n",
-		       function, buf[0], buf[1], buf[2], buf[3] );
+		cpuid ( function, subfunction, &buf[CPUID_EAX],
+			&buf[CPUID_EBX], &buf[CPUID_ECX], &buf[CPUID_EDX] );
+		DBGC ( settings, "CPUID %#08x:%x => %#08x:%#08x:%#08x:%#08x\n",
+		       function, subfunction, buf[0], buf[1], buf[2], buf[3] );
 
 		/* Copy results to buffer */
 		registers = CPUID_REGISTERS ( setting->tag );
 		num_registers = CPUID_NUM_REGISTERS ( setting->tag );
 		for ( ; num_registers-- ; registers >>= 2 ) {
 			output = buf[ registers & 0x3 ];
-			if ( ! CPUID_LITTLE_ENDIAN ( setting->tag ) )
+			if ( ! ( setting->tag & CPUID_LITTLE_ENDIAN ) )
 				output = cpu_to_be32 ( output );
 			frag_len = sizeof ( output );
 			if ( frag_len > len )
@@ -260,7 +248,7 @@ const struct setting cpuvendor_setting __setting ( SETTING_HOST_EXTRA,
 						   cpuvendor ) = {
 	.name = "cpuvendor",
 	.description = "CPU vendor",
-	.tag = CPUID_TAG ( CPUID_VENDOR_ID, 1, 1, 3,
+	.tag = CPUID_TAG ( CPUID_VENDOR_ID, 0, CPUID_LITTLE_ENDIAN, 3,
 			   CPUID_EBX, CPUID_EDX, CPUID_ECX, 0 ),
 	.type = &setting_type_string,
 	.scope = &cpuid_settings_scope,
@@ -271,7 +259,7 @@ const struct setting cpumodel_setting __setting ( SETTING_HOST_EXTRA,
 						  cpumodel ) = {
 	.name = "cpumodel",
 	.description = "CPU model",
-	.tag = CPUID_TAG ( CPUID_MODEL, 3, 1, 4,
+	.tag = CPUID_TAG ( CPUID_MODEL, 2, CPUID_LITTLE_ENDIAN, 4,
 			   CPUID_EAX, CPUID_EBX, CPUID_ECX, CPUID_EDX ),
 	.type = &setting_type_string,
 	.scope = &cpuid_settings_scope,

src/arch/x86/core/dumpregs.c

@@ -6,11 +6,8 @@ void __asmcall _dump_regs ( struct i386_all_regs *ix86 ) {
 	__asm__ __volatile__ (
 		TEXT16_CODE ( ".globl dump_regs\n\t"
 			      "\ndump_regs:\n\t"
-			      "pushl $_dump_regs\n\t"
-			      "pushw %%cs\n\t"
-			      "call prot_call\n\t"
-			      "addr32 leal 4(%%esp), %%esp\n\t"
-			      "ret\n\t" ) : : );
+			      VIRT_CALL ( _dump_regs )
+			      "ret\n\t" ) : );
 
 	printf ( "EAX=%08x EBX=%08x ECX=%08x EDX=%08x\n"
 		 "ESI=%08x EDI=%08x EBP=%08x ESP=%08x\n"

src/arch/x86/core/gdbmach.c

@@ -0,0 +1,251 @@
+/*
+ * Copyright (C) 2008 Stefan Hajnoczi <stefanha@gmail.com>.
+ * Copyright (C) 2016 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * 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 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stddef.h>
+#include <stdio.h>
+#include <errno.h>
+#include <assert.h>
+#include <ipxe/uaccess.h>
+#include <ipxe/gdbstub.h>
+#include <librm.h>
+#include <gdbmach.h>
+
+/** @file
+ *
+ * GDB architecture-specific bits for x86
+ *
+ */
+
+/** Number of hardware breakpoints */
+#define NUM_HWBP 4
+
+/** Debug register 7: Global breakpoint enable */
+#define DR7_G( bp ) ( 2 << ( 2 * (bp) ) )
+
+/** Debug register 7: Global exact breakpoint enable */
+#define DR7_GE ( 1 << 9 )
+
+/** Debug register 7: Break on data writes */
+#define DR7_RWLEN_WRITE 0x11110000
+
+/** Debug register 7: Break on data access */
+#define DR7_RWLEN_ACCESS 0x33330000
+
+/** Debug register 7: One-byte length */
+#define DR7_RWLEN_1 0x00000000
+
+/** Debug register 7: Two-byte length */
+#define DR7_RWLEN_2 0x44440000
+
+/** Debug register 7: Four-byte length */
+#define DR7_RWLEN_4 0xcccc0000
+
+/** Debug register 7: Eight-byte length */
+#define DR7_RWLEN_8 0x88880000
+
+/** Debug register 7: Breakpoint R/W and length mask */
+#define DR7_RWLEN_MASK( bp ) ( 0xf0000 << ( 4 * (bp) ) )
+
+/** Hardware breakpoint addresses (debug registers 0-3) */
+static unsigned long dr[NUM_HWBP];
+
+/** Active value of debug register 7 */
+static unsigned long dr7 = DR7_GE;
+
+/**
+ * Update debug registers
+ *
+ */
+static void gdbmach_update ( void ) {
+
+	/* Set debug registers */
+	__asm__ __volatile__ ( "mov %0, %%dr0" : : "r" ( dr[0] ) );
+	__asm__ __volatile__ ( "mov %0, %%dr1" : : "r" ( dr[1] ) );
+	__asm__ __volatile__ ( "mov %0, %%dr2" : : "r" ( dr[2] ) );
+	__asm__ __volatile__ ( "mov %0, %%dr3" : : "r" ( dr[3] ) );
+	__asm__ __volatile__ ( "mov %0, %%dr7" : : "r" ( dr7 ) );
+}
+
+/**
+ * Find reusable or available hardware breakpoint
+ *
+ * @v addr		Linear address
+ * @v rwlen		Control bits
+ * @ret bp		Hardware breakpoint, or negative error
+ */
+static int gdbmach_find ( unsigned long addr, unsigned int rwlen ) {
+	unsigned int i;
+	int bp = -ENOENT;
+
+	/* Look for a reusable or available breakpoint */
+	for ( i = 0 ; i < NUM_HWBP ; i++ ) {
+
+		/* If breakpoint is not enabled, then it is available */
+		if ( ! ( dr7 & DR7_G ( i ) ) ) {
+			bp = i;
+			continue;
+		}
+
+		/* If breakpoint is enabled and has the same address
+		 * and control bits, then reuse it.
+		 */
+		if ( ( dr[i] == addr ) &&
+		     ( ( ( dr7 ^ rwlen ) & DR7_RWLEN_MASK ( i ) ) == 0 ) ) {
+			bp = i;
+			break;
+		}
+	}
+
+	return bp;
+}
+
+/**
+ * Set hardware breakpoint
+ *
+ * @v type		GDB breakpoint type
+ * @v addr		Virtual address
+ * @v len		Length
+ * @v enable		Enable (not disable) breakpoint
+ * @ret rc		Return status code
+ */
+int gdbmach_set_breakpoint ( int type, unsigned long addr, size_t len,
+			     int enable ) {
+	unsigned int rwlen;
+	unsigned long mask;
+	int bp;
+
+	/* Parse breakpoint type */
+	switch ( type ) {
+	case GDBMACH_WATCH:
+		rwlen = DR7_RWLEN_WRITE;
+		break;
+	case GDBMACH_AWATCH:
+		rwlen = DR7_RWLEN_ACCESS;
+		break;
+	default:
+		return -ENOTSUP;
+	}
+
+	/* Parse breakpoint length */
+	switch ( len ) {
+	case 1:
+		rwlen |= DR7_RWLEN_1;
+		break;
+	case 2:
+		rwlen |= DR7_RWLEN_2;
+		break;
+	case 4:
+		rwlen |= DR7_RWLEN_4;
+		break;
+	case 8:
+		rwlen |= DR7_RWLEN_8;
+		break;
+	default:
+		return -ENOTSUP;
+	}
+
+	/* Convert to linear address */
+	if ( sizeof ( physaddr_t ) <= sizeof ( uint32_t ) )
+		addr = virt_to_phys ( ( void * ) addr );
+
+	/* Find reusable or available hardware breakpoint */
+	bp = gdbmach_find ( addr, rwlen );
+	if ( bp < 0 )
+		return ( enable ? -ENOBUFS : 0 );
+
+	/* Configure this breakpoint */
+	DBGC ( &dr[0], "GDB bp %d at %p+%zx type %d (%sabled)\n",
+	       bp, ( ( void * ) addr ), len, type, ( enable ? "en" : "dis" ) );
+	dr[bp] = addr;
+	mask = DR7_RWLEN_MASK ( bp );
+	dr7 = ( ( dr7 & ~mask ) | ( rwlen & mask ) );
+	mask = DR7_G ( bp );
+	dr7 &= ~mask;
+	if ( enable )
+		dr7 |= mask;
+
+	/* Update debug registers */
+	gdbmach_update();
+
+	return 0;
+}
+
+/**
+ * Handle exception
+ *
+ * @v signo		GDB signal number
+ * @v regs		Register dump
+ */
+__asmcall void gdbmach_handler ( int signo, gdbreg_t *regs ) {
+	unsigned long dr7_disabled = DR7_GE;
+	unsigned long dr6_clear = 0;
+
+	/* Temporarily disable breakpoints */
+	__asm__ __volatile__ ( "mov %0, %%dr7\n" : : "r" ( dr7_disabled ) );
+
+	/* Handle exception */
+	DBGC ( &dr[0], "GDB signal %d\n", signo );
+	DBGC2_HDA ( &dr[0], 0, regs, ( GDBMACH_NREGS * sizeof ( *regs ) ) );
+	gdbstub_handler ( signo, regs );
+	DBGC ( &dr[0], "GDB signal %d returning\n", signo );
+	DBGC2_HDA ( &dr[0], 0, regs, ( GDBMACH_NREGS * sizeof ( *regs ) ) );
+
+	/* Clear breakpoint status register */
+	__asm__ __volatile__ ( "mov %0, %%dr6\n" : : "r" ( dr6_clear ) );
+
+	/* Re-enable breakpoints */
+	__asm__ __volatile__ ( "mov %0, %%dr7\n" : : "r" ( dr7 ) );
+}
+
+/**
+ * CPU exception vectors
+ *
+ * Note that we cannot intercept anything from INT8 (double fault)
+ * upwards, since these overlap by default with IRQ0-7.
+ */
+static void * gdbmach_vectors[] = {
+	gdbmach_sigfpe,		/* Divide by zero */
+	gdbmach_sigtrap,	/* Debug trap */
+	NULL,			/* Non-maskable interrupt */
+	gdbmach_sigtrap,	/* Breakpoint */
+	gdbmach_sigstkflt,	/* Overflow */
+	gdbmach_sigstkflt,	/* Bound range exceeded */
+	gdbmach_sigill,		/* Invalid opcode */
+};
+
+/**
+ * Initialise GDB
+ */
+void gdbmach_init ( void ) {
+	unsigned int i;
+
+	/* Hook CPU exception vectors */
+	for ( i = 0 ; i < ( sizeof ( gdbmach_vectors ) /
+			    sizeof ( gdbmach_vectors[0] ) ) ; i++ ) {
+		if ( gdbmach_vectors[i] )
+			set_interrupt_vector ( i, gdbmach_vectors[i] );
+	}
+}

src/arch/x86/core/linux/linux_api.c

@@ -108,3 +108,42 @@ void * linux_mremap ( void *old_address, __kernel_size_t old_size,
 int linux_munmap ( void *addr, __kernel_size_t length ) {
 	return linux_syscall ( __NR_munmap, addr, length );
 }
+
+int linux_socket ( int domain, int type_, int protocol ) {
+#ifdef __NR_socket
+	return linux_syscall ( __NR_socket, domain, type_, protocol );
+#else
+#ifndef SOCKOP_socket
+# define SOCKOP_socket 1
+#endif
+	unsigned long sc_args[] = { domain, type_, protocol };
+	return linux_syscall ( __NR_socketcall, SOCKOP_socket, sc_args );
+#endif
+}
+
+int linux_bind ( int fd, const struct sockaddr *addr, socklen_t addrlen ) {
+#ifdef __NR_bind
+	return linux_syscall ( __NR_bind, fd, addr, addrlen );
+#else
+#ifndef SOCKOP_bind
+# define SOCKOP_bind 2
+#endif
+	unsigned long sc_args[] = { fd, (unsigned long)addr, addrlen };
+	return linux_syscall ( __NR_socketcall, SOCKOP_bind, sc_args );
+#endif
+}
+
+ssize_t linux_sendto ( int fd, const void *buf, size_t len, int flags,
+		       const struct sockaddr *daddr, socklen_t addrlen ) {
+#ifdef __NR_sendto
+	return linux_syscall ( __NR_sendto, fd, buf, len, flags,
+			       daddr, addrlen );
+#else
+#ifndef SOCKOP_sendto
+# define SOCKOP_sendto 11
+#endif
+	unsigned long sc_args[] = { fd, (unsigned long)buf, len,
+				    flags, (unsigned long)daddr, addrlen };
+	return linux_syscall ( __NR_socketcall, SOCKOP_sendto, sc_args );
+#endif
+}