Enable SSE2 (and SSE) when compiling for x86_64. Compilers often assume
SSE2 is present and enabled when targeting x86_64.
This fixes:
- lzma decompression code is compiled with the -Ofast flag
- 'everything' when compiling with clang.
This mostly affects qemu targets, which did not have this flag selected
yet.
TESTED on qemu.
Change-Id: I3cdc584c97016e15513df663a54a7bdb549a73e4
Signed-off-by: Patrick Rudolph <patrick.rudolph@9elements.com>
Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/44869
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Elyes Haouas <ehaouas@noos.fr>
the "x86 PIC code ebx" workaround done previously
by commit 689e31d18b ("Make cpuid functions usable
when compiled with PIC") does not work for x86_64
(the upper dword of rbx is set to 0)
the GCC bug that needed the workaround was fixed
in version 5 (see GCC bug 54232)
Change-Id: Iff1dd72c7423a3b385a000457bcd065cf7ed6b95
Signed-off-by: Matei Dibu <matdibu@protonmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/66345
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Elyes Haouas <ehaouas@noos.fr>
Some platform run early stages like romstage and verstage from CAR
instead of XIP. This allows to link them like other arch inside the
_program region. This make in place LZ4 decompression possible as it
needs a bit of extra place to extract the code which is now provided by
the .bss.
Tested on up/squared (Intel APL).
Change-Id: I6cf51f943dde5f642d75ba4c5d3be520dc56370a
Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/48156
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Sean Rhodes <sean@starlabs.systems>
Reviewed-by: Nico Huber <nico.h@gmx.de>
Fix the warning below when building GA-945GCM-S2L with 64-bit:
src/arch/x86/idt.S:216: Warning: no instruction mnemonic suffix given and no register operands; using default for `iret'
Change-Id: Ibbc106714e25293951a71d84fea0a660f41f9c02
Signed-off-by: Elyes HAOUAS <ehaouas@noos.fr>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/61336
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
gcc12+ will require riscv architecture selection to come not only with
featurei suffixd charactersa, it also comes with feature_ful suffix_ed
words_mith. Much creative, very appreciate.
To accommodate for this madness, enable the already existing (but off by
default) support for that in our gcc11 build, support using by detecting
the compiler's behavior in xcompile and pass that knowledge along to our
build system.
Then cross our fingers and hope for the best!
Change-Id: I5dfeed766626e78d4f8378d9d857b7a4d61510fd
Signed-off-by: Patrick Georgi <patrick@coreboot.org>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/67457
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Elyes Haouas <ehaouas@noos.fr>
The header file `rules.h` is automatically included in the build by the
top level makefile using the command:
`-include src/soc/intel/common/block/scs/early_mmc.c`.
Similar to `config.h` and 'kconfig.h`, this file does not need to be
included manually, so remove it.
Signed-off-by: Martin Roth <gaumless@gmail.com>
Change-Id: I23a1876b4b671d8565cf9b391d3babf800c074db
Reviewed-on: https://review.coreboot.org/c/coreboot/+/67348
Reviewed-by: Elyes Haouas <ehaouas@noos.fr>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Clear the SMBIOS region before writing SMBIOS tables.
On librem_mini and librem_mini_v2, CBMEM allocations are offset by 4K
for reboots relative to the cold boot. This means the unused SMBIOS
region could contain the first 4K of the ACPI tables from the last boot
(including the signature), which prevents Linux from booting.
The CBMEM 4K offset appears to be due to FSP allocating memory
differently between cold boot and reboot, this appears to be normal and
causes the CBMEM base address to change.
It is not clear why Linux examines an ACPI signature found in this
region, but boot logs over serial confirm that it sees the corrupt
table. The table is supposed to be found just below 1M, and kernel
source appears to look in this region, but it is definitely finding the
corrupt table in CBMEM.
Normal cold boot:
[ 0.008615] ACPI: RSDP 0x00000000000F6190 000024 (v02 COREv4)
[ 0.008619] ACPI: XSDT 0x0000000099B480E0 00005C (v01 COREv4 COREBOOT 00000000 CORE 20220331)
[ 0.008624] ACPI: FACP 0x0000000099B4A2A0 000114 (v06 COREv4 COREBOOT 00000000 CORE 20220331)
[ 0.008634] ACPI: DSDT 0x0000000099B48280 00201F (v02 COREv4 COREBOOT 20110725 INTL 20220331)
...
Reboot with corrupt table:
[ 0.008820] ACPI: RSDP 0x00000000000F6190 000024 (v02 COREv4)
[ 0.008823] ACPI: XSDT 0x0000000099B480E0 00005C (v01 COREv4 COREBOOT 00000000 CORE 20220331)
[ 0.008828] ACPI: ???G 0x0000000099B4A2A0 20002001 (v00 ?G?$ 47020100 ?, 47020100)
[ 0.008831] ACPI: �y 0x0000000099B4A3C0 54523882 (v67 ?_HID? A�? 65520D4E al T 20656D69)
...
There are no specific errors but it returns to the firmware soon after,
presumably due to a fault. This appears to be so early in the boot
that panic=0 on the kernel command line has no effect.
Test: build/boot Librem Mini, Librem Mini v2 and reboot.
Change-Id: Ia20d0b30160e89e8d96add34d7e0e881f070ec61
Signed-off-by: Jonathon Hall <jonathon.hall@puri.sm>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/66377
Reviewed-by: Matt DeVillier <matt.devillier@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
fix the following checkpatch errors:
WARNING:BRACES: braces {} are not necessary for any arm of this statement
354: FILE: src/arch/x86/smbios.c:354:
+ if (CONFIG_ROM_SIZE >= 1 * GiB) {
[...]
+ } else {
[...]
WARNING:BRACES: braces {} are not necessary for single statement blocks
561: FILE: src/arch/x86/smbios.c:561:
+ if (leaf_b_threads == 0) {
+ leaf_b_threads = 1;
+ }
Signed-off-by: Fabio Aiuto <fabioaiuto83@gmail.com>
Change-Id: I14c29e4358cad4cd5ef169ebab7079db2129d8fa
Reviewed-on: https://review.coreboot.org/c/coreboot/+/66925
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martin.roth@amd.corp-partner.google.com>
Since there are many identifiers whose name contain "__unused" in
headers of musl libc, introducing a macro which expands "__unused" to
the source of a util may have disastrous effect during its compiling
under a musl-based platform.
However, it is hard to detect musl at build time as musl is notorious
for having explicitly been refusing to add a macro like "__MUSL__" to
announce its own presence.
Using __always_unused and __maybe_unused for everything may be a good
idea. This is how it works in the Linux kernel, so that would at least
make us match some other standard rather than doing our own thing
(especially since the other compiler.h shorthand macros are also
inspired by Linux).
Signed-off-by: Bill XIE <persmule@hardenedlinux.org>
Change-Id: I547ae3371d7568f5aed732ceefe0130a339716a9
Reviewed-on: https://review.coreboot.org/c/coreboot/+/65717
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Felix Singer <felixsinger@posteo.net>
Reviewed-by: Julius Werner <jwerner@chromium.org>
This patch contains several minor cleanups related to compiler.h:
- Replace __always_unused() (which is a Linux-specific concept that
doesn't make sense without also having __maybe_unused(), and had zero
uses in the codebase) with __unused() which moves here from helpers.h
- Add __underscores__ to the names of all attributes in the compiler
attribute shorthand macros. This is necessary to make them work in
files where the same name was already used for an identifier (e.g.
cbfstool/cbfs.h's `unused` array of file types).
- Remove libpayload's own copy of compiler.h and make it directly pull
in the commonlib/bsd copy.
Signed-off-by: Julius Werner <jwerner@chromium.org>
Change-Id: I9644da594bb69133843c6b7f12ce50b2e45fd24b
Reviewed-on: https://review.coreboot.org/c/coreboot/+/64737
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Yu-Ping Wu <yupingso@google.com>
Reviewed-by: Elyes Haouas <ehaouas@noos.fr>
There are too many "FIT" in firmware land. In order to reduce possible
confusion of CBFS_TYPE_FIT with the Intel Firmware Interface Table, this
patch renames it to CBFS_TYPE_FIT_PAYLOAD (including the cbfstool
argument, so calling scripts will now need to replace `-t fit` with `-t
fit_payload`).
Signed-off-by: Julius Werner <jwerner@chromium.org>
Change-Id: I826cefce54ade06c6612c8a7bb53e02092e7b11a
Reviewed-on: https://review.coreboot.org/c/coreboot/+/64735
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Yu-Ping Wu <yupingso@google.com>
This adds 2 flags:
* invisible opt-in flag for platforms on which clang seems to work
* visible opt-in flag to allow experimenting
Clang seems to work rather well on x86_32 so it makes sense to start
adding that to Jenkins buildtesting, which this allows.
This allows abuild to differentiate between targets that are known to
build with clang. This makes buildtesting just those targets easier.
Change-Id: I46f1bad59bda94f60f4a141237ede11f6eb93cc2
Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/63081
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin L Roth <gaumless@tutanota.com>
Systems have a lot more cores now and 4KiB is not cutting it. E.g.
for a system with 255 cores more than 16KiB is needed.
We could also make this a Kconfig parameter but it's probably not
worth having such micro optimizations to save a few KiB.
Change-Id: Idd47e55d8d679cc70eae996ee1af3ad7eaa1d0cc
Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/63484
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-by: Paul Menzel <paulepanter@mailbox.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
With CBFS verification, cbfstool (CB:41121) needs bootblock to be
present in coreboot.pre in order to locate the metadata hash stored in
it. Therefore we have to ensure that bootblock is added to CBFS before
other CBFS files are added.
To solve the problem, create the 'add_bootblock' function, and call it
in the coreboot.pre recipe. Because bootblock.bin is now a prerequisite
of coreboot.pre, it will get built even if CONFIG_BOOTBLOCK_IN_CBFS=n.
BUG=b:233263447
TEST=emerge-guybrush coreboot
TEST=emerge-corsola coreboot chromeos-bootimage
TEST=cbfstool image-kingler.bin print -v
TEST=Kingler booted successfully
BRANCH=none
Change-Id: I385deb8231e44310ee139c3f69f449e75b92b2be
Signed-off-by: Yu-Ping Wu <yupingso@chromium.org>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/64547
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Julius Werner <jwerner@chromium.org>
AGESA has a lot of code in the .data section (initialized data). However
there is no such section in CAR stages as the code runs in XIP mode and
CAR is too small to contain the data section. When the linker can not
match code to a section it will just append it, which is why AGESA
worked at all.
Follow-up patches will attempt to fix AGESA and set Kconfig parameter to
'n'. After all AGESA sources have been fixed, this can be removed.
Change-Id: I311ee17e3c0bd283692194fcee63af4449583d74
Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/64387
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
