Older versions of gcc (observed with gcc 4.8.5 on CentOS 7) complain
about having the label "err_ioremap" at the end of a compound
statement in bios_mp_start_all(). The label is correctly placed,
since it immediately follows the iounmap() that would be required to
undo a successful ioremap() in the non-error case.
Fix by adding an explicit "return" immediately after the label.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Intel and AMD distribute microcode updates, which are typically
applied by the BIOS and/or the booted operating system.
BIOS updates can be difficult to obtain and cumbersome to apply, and
are often neglected. Operating system updates may be subject to
strict change control processes, particularly for production
workloads. There is therefore value in being able to update the
microcode at boot time using a freshly downloaded microcode update
file, particularly in scenarios where the physical hardware and the
installed operating system are controlled by different parties (such
as in a public cloud infrastructure).
Add support for parsing Intel and AMD microcode update images, and for
applying the updates to all CPUs in the system.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Provide an implementation of the iPXE multiprocessor API for BIOS,
based on sending broadcast INIT and SIPI interprocessor interrupts to
start up all application processors.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Application processors are started via INIT and SIPI interprocessor
interrupts: the INIT places the processor into a "wait for SIPI"
state, and the SIPI then starts the processor in real mode at a
page-aligned address derived from the SIPI vector number.
Add support for installing a real-mode SIPI handler that will switch
the CPU into protected mode with flat physical addressing, load
initial register contents, and then jump to the address of a
protected-mode SIPI handler. No stack pointer is set up, to avoid the
need to allocate stack space for each available processor.
We use 32-bit physical addressing in order to minimise the changes
required for a 64-bit build. The existing long mode transition code
relies on the existence of the stack, so we cannot easily switch the
application processor into long mode. We could use 32-bit virtual
addressing, but this runtime environment does not currently exist
outside of librm.S itself in a 64-bit build, and using it would
complicate the implementation of the protected-mode SIPI handler.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Define an API for executing very limited functions on application
processors in a multiprocessor system, along with an x86-only
implementation.
The normal iPXE runtime environment is effectively non-existent on
application processors. There is no ability to make firmware calls
(e.g. to write to a console), and there may be no stack space
available.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
The drive specification alone does not necessarily contain enough
information to perform a SAN boot (or local disk boot) under UEFI. If
the next-stage bootloader is installed in the EFI system partition
under a non-standard name (e.g. "\EFI\debian\grubx64.efi") then this
explicit boot filename must also be specified.
Generalise this concept to use a "SAN boot configuration parameters"
structure (currently containing only the optional explicit boot
filename), to allow for easy expansion to provide other parameters
such as the partition UUID or volume label.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
We currently use the SAN device pointer as the debug message stream
identifier. This pointer is not always available: for example, when
booting from a local disk there is no underlying SAN device.
Switch to using the drive number as the debug message colour stream
identifier, so that all block device debug messages may be colourised
consistently.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
When ExitBootServices() invokes efi_shutdown_hook(), there may be
nothing to generate an interrupt since the timer is disabled in the
first step of ExitBootServices(). Additionally, for VMs OVMF masks
everything from the PIC (except the timer) by default. This means
that calling cpu_nap() may hang indefinitely. This was seen in
practice in netfront_reset() when running in a VM on XenServer.
Fix this by skipping the halt if an EFI shutdown is in progress.
Signed-off-by: Ross Lagerwall <ross.lagerwall@citrix.com>
Modified-by: Michael Brown <mcb30@ipxe.org>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
The slightly incomprehensible LoongArch64 implementation for
bigint_subtract() is observed to produce incorrect results for some
input values.
Replace the suspicious LoongArch64 implementations of bigint_add(),
bigint_subtract(), bigint_rol() and bigint_ror(), and add a test case
for a subtraction that was producing an incorrect result with the
previous implementation.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Big integer multiplication is currently used only as part of modular
exponentiation, where both multiplicand and multiplier will be the
same size.
Relax this requirement to allow for the use of big integer
multiplication in other contexts.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Newer versions of the GNU assembler (observed with binutils 2.41) will
complain about the ".arch i386" in files assembled with "as --64",
with the message "Error: 64bit mode not supported on 'i386'".
In files such as stack.S that contain no instructions to be assembled,
the ".arch i386" is redundant and may be removed entirely.
In the remaining files, fix by moving ".arch i386" below the relevant
".code16" or ".code32" directive, so that the assembler is no longer
expecting 64-bit instructions to be used by the time that the ".arch
i386" directive is encountered.
Reported-by: Ali Mustakim <alim@forwardcomputers.com>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
The .text directive is entirely redundant when followed by a .section
directive giving an explicit section name and attributes.
Remove these unnecessary directives to simplify the code.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Support scanning for the 64-bit SMBIOS3 entry point in addition to the
32-bit SMBIOS2 entry point.
Prefer use of the 32-bit entry point if present, since this is
guaranteed to be within accessible memory.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
The PCI cloud API (PCIAPI_CLOUD) currently selects the first PCI API
that successfully discovers a PCI device address range. The ECAM API
may discover an address range but subsequently be unable to map the
configuration space region, which would result in the selected PCI API
being unusable.
Fix by instead selecting the first PCI API that can be successfully
used to discover a PCI device.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Simplify the per-netdevice GuestInfo settings code by using
driver-private data to hold the settings block, instead of using a
separate allocation.
The settings block (if existent) will be automatically unregistered
when the parent network device settings block is unregistered, and no
longer needs to be separately freed. The guestinfo_net_remove()
function may therefore be omitted completely.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Allow network upper-layer drivers (such as LLDP, which attaches to
each network device in order to provide a corresponding LLDP settings
block) to specify a size for private data, which will be allocated as
part of the network device structure (as with the existing private
data allocated for the underlying device driver).
This will allow network upper-layer drivers to be simplified by
omitting memory allocation and freeing code. If the upper-layer
driver requires a reference counter (e.g. for interface
initialisation), then it may use the network device's existing
reference counter, since this is now the reference counter for the
containing block of memory.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
We currently use "push $1f" within inline assembly to push the address
of the real-mode code fragment, relying on the assembler to treat this
as "pushl" for 32-bit code or "pushq" for 64-bit code.
As of binutils commit 5cc0077 ("x86: further adjust extend-to-32bit-
address conditions"), first included in binutils-2.41, this implicit
operand size is no longer calculated as expected and 64-bit builds
will fail with
Error: operand size mismatch for `push'
Fix by adding an explicit operand size to the "push" instruction.
Originally-fixed-by: Justin Cano <jstncno@gmail.com>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
The special key range (from KEY_MIN upwards) currently overlaps with
the valid range for Unicode characters, and therefore prohibits the
use of Unicode key values outside the ASCII range.
Create space for Unicode key values by moving the special keys to the
range immediately above the maximum valid Unicode character. This
allows the existing encoding of special keys as an efficiently packed
representation of the equivalent ANSI escape sequence to be maintained
almost as-is.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Some versions of ld will complain that the automatically created (and
unused by our build process) ELF program headers include a "LOAD
segment with RWX permissions".
Silence this warning by adding "-z separate-code" to the linker
options, where supported.
For BIOS builds, where the prefix will generally require writable
access to its own (tiny) code segment, simply inhibit the warning
completely via "--no-warn-rwx-segments".
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Multiple target patterns in pattern rules are treated as grouped
targets regardless of the separator character. Newer verions of make
will generate "warning: pattern recipe did not update peer target" to
warn that the rule was expected to update all of the (implicitly)
grouped targets.
Fix by splitting all multiple target pattern rules into single target
pattern rules.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
There is no common standard for I/O-space access for non-x86 CPU
families, and non-MMIO peripherals are vanishingly rare.
Generalise the existing ARM definitions for dummy PIO to allow for
reuse by other CPU architectures.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
The PCI I/O API (supporting accesses to PCI configuration space) is
not related to the general I/O API (supporting accesses to
memory-mapped I/O peripherals).
Remove the spurious inclusion of ipxe/io.h from the PCI I/O header.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
When invoking a kernel via the UEFI shim, the kernel (and potentially
also a helper binary such as GRUB) must be accessible via the virtual
filesystem exposed via EFI_SIMPLE_FILE_SYSTEM_PROTOCOL but must not be
present in the magic initrd constructed from all registered images.
Allow for images to be flagged as hidden, which will cause them to be
excluded from API-level lists of all images such as the virtual
filesystem directory contents, the magic initrd, or the Multiboot
module list. Hidden images remain visible to iPXE commands including
"imgstat", which will show a "[HIDDEN]" flag for such images.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
We unregister script images during their execution, to prevent a
"boot" command from re-executing the containing script. This also has
the side effect of preventing executing scripts from showing up within
the Linux magic initrd image (or the Multiboot module list).
Additional logic in bzimage.c and efi_file.c prevents a currently
executing kernel from showing up within the magic initrd image.
Similar logic in multiboot.c prevents the Multiboot kernel from
showing up as a Multiboot module.
This still leaves some corner cases that are not covered correctly.
For example: when using a gzip-compressed kernel image, nothing will
currently hide the original compressed image from the magic initrd.
Fix by moving the logic that temporarily unregisters the current image
from script_exec() to image_exec(), so that it applies to all image
types, and simplify the magic initrd and Multiboot module list
construction logic on the basis that no further filtering of the
registered image list is necessary.
This change has the side effect of hiding currently executing EFI
images from the virtual filesystem exposed by iPXE. For example, when
using iPXE to boot wimboot, the wimboot binary itself will no longer
be visible within the virtual filesystem.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
As noted in commit 3c83843 ("[rng] Check for several functioning RTC
interrupts"), experimentation shows that Hyper-V cannot be trusted to
reliably generate RTC interrupts. (As noted in commit f3ba0fb
("[hyperv] Provide timer based on the 10MHz time reference count
MSR"), Hyper-V appears to suffer from a general problem in reliably
generating any legacy interrupts.) An alternative entropy source is
therefore required for an image that may be used in a Hyper-V Gen1
virtual machine.
The x86 RDRAND instruction provides a suitable alternative entropy
source, but may not be supported by all CPUs. We must therefore allow
for multiple entropy sources to be compiled in, with the single active
entropy source selected only at runtime.
Restructure the internal entropy API to allow a working entropy source
to be detected and chosen at runtime.
Enable the RDRAND entropy source for all x86 builds, since it is
likely to be substantially faster than any other source.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
The Linux kernel bzImage image format and the CPIO archive constructor
will parse the image command line for certain arguments of the form
"key=value". This parsing is currently implemented using strstr() in
a way that can cause a false positive suffix match. For example, a
command line containing "highmem=<n>" would erroneously be treated as
containing a value for "mem=<n>".
Fix by centralising the logic used for parsing such arguments, and
including a check that the argument immediately follows a whitespace
delimiter (or is at the start of the string).
Reported-by: Filippo Giunchedi <filippo@esaurito.net>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Commit 74222cd ("[rng] Check for functioning RTC interrupt") added a
check that the RTC is capable of generating interrupts via the legacy
PIC, since this mechanism appears to be broken in some Hyper-V virtual
machines.
Experimentation shows that the RTC is sometimes capable of generating
a single interrupt, but will then generate no subsequent interrupts.
This currently causes rtc_entropy_check() to falsely detect that the
entropy gathering mechanism is functional.
Fix by checking for several RTC interrupts before declaring that it is
a functional entropy source.
Reported-by: Andreas Hammarskjöld <junior@2PintSoftware.com>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Add support for building a LoongArch64 Linux userspace binary.
Signed-off-by: Xiaotian Wu <wuxiaotian@loongson.cn>
Modified-by: Michael Brown <mcb30@ipxe.org>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
The PAGE_SHIFT definition is an architectural property, rather than an
aspect of a particular I/O API implementation (of which, in theory,
there may be more than one per architecture).
Reflect this by moving the definition to the top-level bits/io.h for
each architecture.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Over the years, the undocumented operand modifier used to produce the
unprefixed constant values in __einfo_error() has varied from "%c0" to
"%a0" in commit 1a77466 ("[build] Fix use of inline assembly on GCC
4.8 ARM64 builds") and back to "%c0" in commit 3fb3ffc ("[build] Fix
use of inline assembly on GCC 8 ARM64 builds"), according to the
evolving demands of the toolchain.
LoongArch64 suffers from a similar issue: GCC 13 will allow either,
but the currently released GCC 12 allows only the "%a0" form.
Introduce a macro ASM_NO_PREFIX, defined in bits/compiler.h, to
abstract away this difference and allow different architectures to use
different operand modifiers.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
The PXE UDP receive queue may grow without limit if the PXE NBP does
not call PXENV_UDP_READ sufficiently frequently.
Fix by implementing a cache discarder for received PXE UDP packets
(similar to the TCP cache discarder).
Reported-by: Tal Shorer <shorer@amazon.com>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Some versions of the 32-bit ARM linker seem to treat the absence of a
.note.GNU-stack section as implying an executable stack, and will
print a warning that this is deprecated behaviour.
Silence the warning by adding a .note.GNU-stack section to each
assembly file and retaining the sections in the Linux linker script.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
The EFI ABI requires the use of -mfloat-abi=soft, but other platforms
may require -mfloat-abi=hard.
Allow for this by using -mfloat-abi=soft only for EFI builds.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
The EFI ABI requires the use of -fno-short-enums, and the EDK2 headers
will perform a compile-time check that enums are 32 bits.
The EDK2 headers may be included even in builds for non-EFI platforms,
and so the -fno-short-enums flag must be used in all 32-bit ARM
builds. Fortunately, nothing else currently cares about enum sizes.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Add support for building as a Linux userspace binary for AArch64.
This allows the self-test suite to be more easily run for the 64-bit
ARM code. For example:
# On a native AArch64 system:
#
make bin-arm64-efi/tests.linux && ./bin-arm64-efi/tests.linux
# On a non-AArch64 system (e.g. x86_64) via cross-compilation,
# assuming that kernel and glibc headers are present within
# /usr/aarch64-linux-gnu/sys-root/:
#
make bin-arm64-linux/tests.linux CROSS=aarch64-linux-gnu- && \
qemu-aarch64 -L /usr/aarch64-linux-gnu/sys-root/ \
./bin-arm64-linux/tests.linux
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Move the platform-specific DHCP client architecture definitions to
header files of the form <ipxe/$(PLATFORM)/dhcparch.h>. This
simplifies the directory structure and allows the otherwise unused
arch/$(ARCH)/include/$(PLATFORM) to be removed from the include
directory search path, which avoids the confusing situation in which a
header file may potentially be accessed through more than one path.
For Linux userspace binaries on any architecture, use the EFI values
for that architecture by delegating to the EFI header file. This
avoids the need to explicitly select values for Linux userspace
binaries for each architecture.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
The requirement to undo the implicit "-Dlinux" is not specific to the
x86 architecture. Move this out of the x86-specific Makefile.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Reduce duplication between i386 and x86_64 by providing a single
shared linker script that both architectures can include.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
When chainloading iPXE from a VLAN device, the MAC address within the
cached DHCPACK will match the MAC address of the trunk device created
by iPXE, and the cached DHCPACK will then end up being erroneously
applied to the trunk device. This tends to break outbound IPv4
routing, since both the trunk and VLAN devices will have the same
assigned IPv4 address.
Fix by recording the VLAN tag along with the cached DHCPACK, and
treating the VLAN tag as part of the filter used to match the cached
DHCPACK against candidate network devices.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
bzimage_parse_cmdline() uses strcmp() to identify the named "vga=..."
kernel command line option values, which will give a false negative if
the option is not last on the command line.
Fix by temporarily changing the relevant command line separator (if
any) to a NUL terminator.
Debugged-by: Simon Rettberg <simon.rettberg@rz.uni-freiburg.de>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Pretty much all physical machines and off-the-shelf virtual machines
will provide a functional PCI BIOS. We therefore default to using
only the PCI BIOS, with no fallback to an alternative mechanism if the
PCI BIOS fails.
AWS EC2 provides the opportunity to experience some exceptions to this
rule. For example, the t3a.nano instances in eu-west-1 have no
functional PCI BIOS at all. As of commit 83516ba ("[cloud] Use
PCIAPI_DIRECT for cloud images") we therefore use direct Type 1
configuration space accesses in the images built and published for use
in the cloud.
Recent experience has discovered yet more variation in AWS EC2
instances. For example, some of the metal instance types have
multiple PCI host bridges and the direct Type 1 accesses therefore
see only a subset of the PCI devices.
Attempt to accommodate future such variations by making the PCI I/O
API selectable at runtime and choosing ECAM (if available), falling
back to the PCI BIOS (if available), then finally falling back to
direct Type 1 accesses.
This is implemented as a dedicated PCIAPI_CLOUD API, rather than by
having the PCI core select a suitable API at runtime (as was done for
timers in commit 302f1ee ("[time] Allow timer to be selected at
runtime"). The common case will remain that only the PCI BIOS API is
required, and we would prefer to retain the optimisations that come
from inlining the configuration space accesses in this common case.
Cloud images are (at present) disk images rather than ROM images, and
so the increased code size required for this design approach in the
PCIAPI_CLOUD case is acceptable.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Allow pcibios_discover() to return an empty range if the INT 1A,B101
PCI BIOS installation check call fails.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Allow pci_find_next() to discover devices beyond the first PCI
segment, by generalising pci_num_bus() (which implicitly assumes that
there is only a single PCI segment) with pci_discover() (which has the
ability to return an arbitrary contiguous chunk of PCI bus:dev.fn
address space).
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Allow for linked-in code to override the mechanism used to locate an
ACPI table, thereby opening up the possibility of ACPI self-tests.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Michael Brown
Ross Lagerwall
Geert Stappers
Xiaotian Wu