SMBIOS has a field to display the cache size, which is currently
set to UNKNOWN unconditionally, multiply the cache size of L1 and L2
by the number of cores.
TEST=Execute "dmidecode -t 7" to check if the cache information
is correct for Deltalake platform
Change-Id: Ieeb5d3346454ffb2291613dc2aa24b31d10c2e04
Signed-off-by: Morgan Jang <Morgan_Jang@wiwynn.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/46068
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-by: Jonathan Zhang <jonzhang@fb.com>
Reviewed-by: Nico Huber <nico.h@gmx.de>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
This fixes non-emulation platforms as those are using 32bit code
after the bootblock_crt0 entry, like setting up CAR and updating
microcode, which isn't yet converted to support long mode.
This is a noop for the only supported x86_64 platform and all
x86_32 platforms.
Change-Id: I45e56ed8db9a44c00cd61e962bb82f27926eb23f
Signed-off-by: Patrick Rudolph <siro@das-labor.org>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/37370
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Add Thread Count in SMBIOS type 4 "Processor Information".
Modify Thread Count 2 according to SMBIOS spec, it should
be the number of threads per processor socket.
TEST="dmidecode -t4" to check.
Signed-off-by: Francois Toguo <francois.toguo.fotso@intel.com>
Change-Id: I0e00ba706eecdc850a2c6a4d876a7732dcc8f755
Reviewed-on: https://review.coreboot.org/c/coreboot/+/45564
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
In case of 64K bootblock the pagetables don't fit, as the CBFS header
also needs a few bytes.
Fixes build error on platforms that use 64KiB bootblock.
Tested on Lenovo T410 with additional x86_64 patches.
Change-Id: I854c5f575e2376827a366cca8d25682c4d90bc8f
Signed-off-by: Patrick Rudolph <patrick.rudolph@9elements.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/37394
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Nearly every x86 platform uses the same arch for all stages. The only
exception is Picasso. So, factor out redundant symbols from the rest.
Alder Lake is not yet complete, so it has been skipped for now.
Change-Id: I7cff9efbc44546807d9af089292c69fb0acc7bad
Signed-off-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/45731
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Nico Huber <nico.h@gmx.de>
Adds support for loading Linux kernels through FIT payloads. This has
been implemented as an assembly function in order to simplify dealing
with some of the intricacies of loading a kernel (such as needing to
jump to the kernel in ARM mode and the kernel ABI).
TEST: Booted a FIT image containing a 5.4 kernel and initramfs on the
Beaglebone Black.
Change-Id: I7dbf9467665ec17447cec73676763844b4be4764
Signed-off-by: Sam Lewis <sam.vr.lewis@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/45335
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Julius Werner <jwerner@chromium.org>
ddr_frequency is ambiguous and is interpreted differently in several
places. Instead of renaming this field, this deprecates it and adds
two new fields with unambiguous naming, max_speed_mts and
configured_speed_mts. smbios.c falls back to using ddr_frequency
when either of these fields are 0.
The same value was being used for both configured memory speed and
max memory speed in SMBIOS type 17, which is not accurate when
configured speed is not the max speed.
BUG=b:167218112
TEST=Boot ezkinil, no change to dmidecode -t17
Change-Id: Iaa75401f9fc33642dbdce6c69bd9b20f96d1cc25
Signed-off-by: Rob Barnes <robbarnes@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/44549
Reviewed-by: Furquan Shaikh <furquan@google.com>
Reviewed-by: Raul Rangel <rrangel@chromium.org>
Reviewed-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
Reviewed-by: Jack Rosenthal <jrosenth@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
It seems that GCC's LTO doesn't like the way we implement
DECLARE_OPTIONAL_REGION(). This patch changes it so that rather than
having a normal DECLARE_REGION() in <symbols.h> and then an extra
DECLARE_OPTIONAL_REGION() in the C file using it, you just say
DECLARE_OPTIONAL_REGION() directly in <symbols.h> (in place and instead
of the usual DECLARE_REGION()). This basically looks the same way in the
resulting object file but somehow LTO seems to like it better.
Signed-off-by: Julius Werner <jwerner@chromium.org>
Change-Id: I6096207b311d70c8e9956cd9406bec45be04a4a2
Reviewed-on: https://review.coreboot.org/c/coreboot/+/44791
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Jacob Garber <jgarber1@ualberta.ca>
Reviewed-by: Hung-Te Lin <hungte@chromium.org>
Reviewed-by: HAOUAS Elyes <ehaouas@noos.fr>
The code can only be compiled as x86_32. Mark it as such to fix errors
in the x86_64 assembler. The caller has to make sure to call this code
in protected mode only.
Tested on HP Z220:
* Still boots on x86_32.
Change-Id: I4c0221fb3886b586c22fe05e36109fcdc20b7eed
Signed-off-by: Patrick Rudolph <patrick.rudolph@9elements.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/44674
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
The GDT loading did work fine on x86_64 a few months ago, but today it
only works in QEMU, but not on real hardware or KVM-enabled QEMU. This
might be related to toolchain changes.
Use 64bit GDT loading on x86_64 and force the assembler to generate a
64bit address load on the GDT. This will make sure no 32bit (signed)
displacement op is being generated, which points to the wrong address
in longmode.
Verified using readelf and made sure no R_X86_64_32S relocation symbol
is emitted. Disassembled the romstage ELF and made sure the GDT address
is 64bit in size.
Tested on QEMU and KVM-enabled QEMU: Doesn't crash any more on KVM.
Signed-off-by: Patrick Rudolph <siro@das-labor.org>
Change-Id: Ia824f90d9611e6e8db09bd62a05e6f990581f09a
Reviewed-on: https://review.coreboot.org/c/coreboot/+/43136
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Compiler's instrumentation cannot insert asan memory checks in
case of memory functions like memset, memcpy and memmove as they
are written in assembly.
So, we need to manually check the memory state before performing
each of these operations to ensure that ASan is triggered in case
of bad access.
Change-Id: I2030437636c77aea7cccda8efe050df4b77c15c7
Signed-off-by: Harshit Sharma <harshitsharmajs@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/44307
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Werner Zeh <werner.zeh@siemens.com>
This patch adds ASan support to romstage on x86 architecture.
A Kconfig option is added to enable ASan in romstage. Compiler
flags are updated. A memory space representing the shadow region
is reserved in linker section. And a function call to asan_init()
is added to initialize shadow region when romstage loads.
Change-Id: I67ebfb5e8d602e865b1f5c874860861ae4e54381
Signed-off-by: Harshit Sharma <harshitsharmajs@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/43604
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Werner Zeh <werner.zeh@siemens.com>
This patch adds address sanitizer module to the library and reserves
a linker section representing the shadow region for ramstage. Also,
it adds an instruction to initialize shadow region on x86
architecture when ramstage is loaded.
Change-Id: Ica06bd2be78fcfc79fa888721ed920d4e8248f3b
Signed-off-by: Harshit Sharma <harshitsharmajs@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/42496
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Werner Zeh <werner.zeh@siemens.com>
* On ARCH_RAMSTAGE_X86_64 jump to the payload in protected mode.
* Add a helper function to jump to arbitrary code in protected mode,
similar to the real mode call handler.
* Doesn't affect existing x86_32 code.
* Add a macro to cast pointer to uint32_t that dies if it would overflow
on conversion
Tested on QEMU Q35 using SeaBIOS as payload.
Tested on Lenovo T410 with additional x86_64 patches.
Change-Id: I6552ac30f1b6205e08e16d251328e01ce3fbfd14
Signed-off-by: Patrick Rudolph <siro@das-labor.org>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/30118
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Implements fit_payload_arch for the arm (aarch32) architecture, so that
FIT images can be used. The implementation is very similar to the
existing implementations for arm64 and riscv, and has mostly been
lifted from these other ports.
TEST: Booted Beaglebone Black (in progress port, to be submitted soon!)
with a FIT image containing a 5.4 kernel, dtb and initramfs.
Change-Id: I6b50c6f06b83c00a5b3622b5bbafe67130b6d233
Signed-off-by: Sam Lewis <sam.vr.lewis@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/44377
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Julius Werner <jwerner@chromium.org>
* Add support for loading GDT on x86_64.
* Add x86_64 assembly code to do the same as the x86_32 code.
* Separate x86_32 and x86_64 code.
Tested on qemu x86_32 and x86_64 using additional MTRRs.
Tested on Lenovo T410 with additional x86_64 patches.
Change-Id: I1c190627f5f0ed6f82738cb99423892382899d7b
Signed-off-by: Patrick Rudolph <siro@das-labor.org>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/30500
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Fix multiple issues allowing to boot until "Payload not loaded":
* The FMAP_CACHE was placed in memory mapped flash
- Place the FMAP_CACHE in DRAM.
* The FMAP_CACHE was overlapping the BOOTBLOCK, which has a default size
of 128KiB.
- Increase the bootblock size in memlayout to 128KiB to match the FMAP.
* The heap in bootblock wasn't usable.
- Add a linking check in armv7 common bootblock to relocate itself to
the linked address.
* A FIT payload couldn't be compiled in as the POSTRAM_CBFS_CACHE was
missing.
- Add the POSTRAM_CBFS_CACHE to memlayout.
* The coreboot log is spammed with missing timestamp table error messages
- Add TIMESTAMP table to memlayout.
Tested on QEMU armv7 vexpress.
Change-Id: Ib9357a5c059ca179826c5a7e7616a5c688ec2e95
Signed-off-by: Patrick Rudolph <siro@das-labor.org>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/39187
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Julius Werner <jwerner@chromium.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Put `__weak` at the beginning of functions and reflow lines to leverage
the increased line width of 96 characters.
Tested with BUILD_TIMELESS=1, Asrock B85M Pro4 does not change.
Change-Id: I3a5fd2d4344b83e09f89053c083ec80aa297061e
Signed-off-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/44020
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
The `GetPhysicallyInstalledSystemMemory` API call, at least on Windows
10, returns an error if SMBIOS tables are invalid. Various tools use
this API call and don't operate correctly if this fails. For example,
the "Intel Processor Diagnostic Tool" program is affected.
Windows then guesses the physical memory size by accumulating entries
from the firmware-provided memory map, which results in a total memory
size that is slightly lower than the actual installed memory capacity.
To fix this issue, add the handle to a type 16 entry to all type 17
entries.
Add new fields to struct memory_info and fill them in Intel common code.
Use the introduced variables to fill type 16 in smbios.c and provide
a handle to type 17 entries.
Besides keeping the current behaviour on intel/soc/common platforms, the
type 16 table is also emitted on platforms that don't explicitly fill
it, by using the existing fields of struct memory_info.
Tested on Windows 10:
The GetPhysicallyInstalledSystemMemory API call doesn't return an error
anymore and the installed memory is now being reported as 8192 MiB.
Change-Id: Idc3a363cbc3d0654dafd4176c4f4af9005210f42
Signed-off-by: Patrick Rudolph <patrick.rudolph@9elements.com>
Signed-off-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/43969
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Marcello Sylvester Bauer <sylv@sylv.io>
Reviewed-by: Christian Walter <christian.walter@9elements.com>
Reviewed-by: Patrick Rudolph <siro@das-labor.org>
Fill in the new fields introduced with version 3.0 and install the new
entry point structure identified by _SM3_.
Tested on Linux 5.6 using tianocore as payload:
Still able to decode the tables without errors.
Change-Id: Iba7a54e9de0b315f8072e6fd2880582355132a81
Signed-off-by: Patrick Rudolph <patrick.rudolph@9elements.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/43719
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Remove the obscure path in source code, where ACPI S3 resume
was prohibited and acpi_resume() would return and continue
to BS_WRITE_TABLES.
The condition when ACPI S3 would be prohibited needs to be
checked early in romstage already. For the time being, there
has been little interest to have CMOS option to disable
ACPI S3 resume feature.
Change-Id: If5105912759427f94f84d46d1a3141aa75cbd6ef
Signed-off-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/42498
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Implement type 19 by accumulating the DRAM dimm size found in cbmem's
CBMEM_ID_MEMINFO structure. This seems common on x86 where the
address space always starts at 0.
At least EDK2 uses this table in the UI and shows 0 MB DRAM if not
present.
Change-Id: Idee8b8cd0b155e14d62d4c12893ff01878ef3f1c
Signed-off-by: Patrick Rudolph <patrick.rudolph@9elements.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/43672
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-by: Christian Walter <christian.walter@9elements.com>
It's not related to spinlocks and the actual implementation
was also guarded by CONFIG(SMP).
With a single call-site in x86-specific code, empty stubs
for other arch are currently not necessary.
Also drop an unused included on a nearby line.
Change-Id: I00439e9c1d10c943ab5e404f5d687d316768fa16
Signed-off-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/43808
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
All supported x86 chips select HAVE_CF9_RESET, and also use 0xcf9 as
reset register in FADT. How unsurprising. We might as well use that
information to automatically fill in the FADT accordingly. So, do it.
To avoid having x86-specific code under arch-agnostic `acpi/`, create a
new optional `arch_fill_fadt` function, and override it for x86 systems.
Tested on Asus P8Z77-V LX2 with Linux 5.7.6 and Windows 10 at the end of
the patch train, both operating systems are able to boot successfully.
Change-Id: Ib436b04aafd66c3ddfa205b870c1e95afb3e846d
Signed-off-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/43389
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Patrick Rudolph <siro@das-labor.org>
Reviewed-by: Frans Hendriks <fhendriks@eltan.com>
