Instead of defining NUM_FIXED_MTRRS in both cpu/x86/mp_init.h and
cpu/x86/mtrr/mtrr.c in two different ways that will evaluate to the same
value, define it once in include/cpu/x86/mtrr.h which is included in
both C files.
TEST=Timeless build for amd/mandolin results in identical firmware image
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I71cec61e22f5ce76baef21344c7427be29f193f8
Reviewed-on: https://review.coreboot.org/c/coreboot/+/67774
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-by: Martin Roth <martin.roth@amd.corp-partner.google.com>
Instead of redefining the register address in smihandler.c, use the
existing definitions from include/cpu/x86/lapic_def.h.
TEST=Timeless build for lenovo/g505s which includes this file in the
build results in identical firmware image.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: Id22f9b5ce53c7bced6bbcc3f5026d4c793b34f78
Reviewed-on: https://review.coreboot.org/c/coreboot/+/67776
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
CB:63475 inadvertently disabled the STM by moving its load point
off of the MSEG boundry, which is a hardware requirement. In
addition, the BIOS resource list cannot be located within the
MSEG. This patch fixes the issue by moving the STM load point
to the MSEG boundry and placing the bios resource list just below
the MSEG where the STM setup functions can find it.
Fixes: commit 5747f6c (cpu/x86/smm_module_loader.c Rewrite setup)
Signed-off-by: Eugene Myers <edmyers@tycho.nsa.gov>
Change-Id: I7359939063bb1a172fcb701551c099edebfbedd5
Reviewed-on: https://review.coreboot.org/c/coreboot/+/67665
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-by: Eugene Myers <cedarhouse1@comcast.net>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
The µcode updates for Broadwell come from coreboot's blobs submodule
and have not been updated in at least 7 years. Use the µcode updates
available in the intel-microcode submodule. This change forgoes some
µcode updates for old Broadwell ULT/ULX steppings with CPUID 0x306d2
and 0x306d3, as well as an old µcode update for Haswell ULT/ULX CPUs
with CPUID 0x40651 in favor of a newer intel-microcode revision that
was already being used: when the µcode updates are concatenated into
one file, the newer µcode update revision would be placed before the
older revision, so the latter would never be used.
Change-Id: I67f8a58552bd211095c183e6f7a219d60e3be162
Signed-off-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/67526
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Commit 27126f135d (cpu/intel/haswell: add
Crystal Well CPU IDs) introduced new Haswell CPUIDs but did not include
any µcode updates for them. It is unknown how this could have worked as
the initial µcode inside the CPU can be quite unstable. Intel CPUs with
support for FIT (Firmware Interface Table) can have their µcode updated
before the x86 reset vector is executed.
The µcode updates for Crystal Well CPUID 0x40661 can be found inside the
intel-microcode submodule. There are no publicly available µcode updates
for Crystal Well CPUID 0x40660 as it is a pre-production stepping, which
is not meant to be used anymore. Hook up the available µcode updates for
Crystal Well CPUs.
Change-Id: If5264f333e681171a2ca4a68be155ffd40a1043b
Signed-off-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/67525
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-by: Martin Roth <martin.roth@amd.corp-partner.google.com>
There are two types of Haswell/Broadwell platforms: Trad(itional) with
separate CPU and PCH packages, and ULT/ULX where the CPU and PCH share
one package. Mainboards can specify which platform type they are using
the `INTEL_LYNXPOINT_LP` Kconfig option. There are so many differences
between Trad and ULT/ULX that it's not worth doing runtime detection.
The CPUIDs are different for Trad and ULT/ULX platforms, and so are the
µcode updates. So, including Trad µcode updates in a coreboot image for
an ULT/ULX mainboard makes no sense, and vice versa.
Adapt the Makefile so that only relevant µcode updates are added. Also,
add a few comments to indicate which updates correspond to which CPUs.
TEST=Run binwalk on coreboot.rom to verify included µcode updates for:
- Asrock B85M Pro4 (Haswell Trad)
- HP Folio 9480M (Haswell ULT/ULX)
- Purism Librem BDW (Broadwell ULT/ULX)
Change-Id: I6dc9e94ce9fede15cbcbe6be577c48c197a9212a
Signed-off-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/67524
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
AMD CPUs have a convenient MSR that allows to set the SMBASE in the save
state without ever entering SMM (e.g. at the default 0x30000 address).
This has been a feature in all AMD CPUs since at least AMD K8. This
allows to do relocation in parallel in ramstage and without setting up a
relocation handler, which likely results in a speedup. The more cores
the higher the speedup as relocation was happening sequentially. On a 4
core AMD picasso system this results in 33ms boot speedup.
TESTED on google/vilboz (Picasso) with CONFIG_SMI_DEBUG: verify that SMM
is correctly relocated with the BSP correctly entering the smihandler.
Change-Id: I9729fb94ed5c18cfd57b8098c838c08a04490e4b
Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/64872
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Felix Held <felix-coreboot@felixheld.de>
Haswell and Broadwell platforms usually stitch six microcode
patches. It has worked so far with the default value of four thanks a
bug which is being fixed by `util/ifittool: Error out if microcodes do
not fit the FIT table' commit.
BUG=b:245380705
TEST=Jenkins build without failing on the FIT table size
Signed-off-by: Jeremy Compostella <jeremy.compostella@intel.com>
Change-Id: I23bf79a3e8918499f6c51e6ef829312d5872181a
Reviewed-on: https://review.coreboot.org/c/coreboot/+/67466
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Since mono_time is now 64-bit, the utility functions interfacing with
mono_time should also be 64-bit so precision isn't lost.
Fixed build errors related to printing the now int64_t result of
stopwatch_duration_[m|u]secs in various places.
BUG=b:237082996
BRANCH=All
TEST=Boot dewatt
Change-Id: I169588f5e14285557f2d03270f58f4c07c0154d5
Signed-off-by: Rob Barnes <robbarnes@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/66170
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Werner Zeh <werner.zeh@siemens.com>
Reviewed-by: Raul Rangel <rrangel@chromium.org>
Reviewed-by: Yu-Ping Wu <yupingso@google.com>
The X2APIC_LATE_WORKAROUND kconfig allows bringing APs in XAPIC mode initially hence, it won't work if LAPIC ID is > 0xff.
This patch ensures the MAX_CPUS logic is appropriate while selecting X2APIC_LATE_WORKAROUND kconfig from SoC.
BUG=b:219061518, b:219053812
TEST=Able to build Google/Rex.
Signed-off-by: Subrata Banik <subratabanik@google.com>
Change-Id: I466e6cc568024a9dea80af21e0ebf3572e74a1f1
Reviewed-on: https://review.coreboot.org/c/coreboot/+/66110
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-by: Felix Held <felix-coreboot@felixheld.de>
Intel Meteor Lake SoC expects to select late x2APIC enablement where
AP bring up will use xAPIC and later x2APIC gets enabled using CPU init.
This patch provides an option where SoC code choose the correct
LAPIC access mode using choice selection.
BUG=b:219061518, b:219053812
TEST=Able to build Google/Rex.
Signed-off-by: Subrata Banik <subratabanik@google.com>
Change-Id: I6b50a0f5e39a95c25cd2c72219d2b402550a6fad
Reviewed-on: https://review.coreboot.org/c/coreboot/+/65786
Reviewed-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
The do_smm struct element in the mp_state struct was an int even though
it only had two possible states, so change it to bool to make this more
obvious. Also change the return type of is_smm_enabled from int to bool.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I8d2d95f0497649d67565243d14a5ab9c9cdda412
Reviewed-on: https://review.coreboot.org/c/coreboot/+/65776
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
This patch refactors the microcode loading and reloading API with a
helper function that perform the actual MSR write operation after
taking the microcode pointer from the caller function.
Also, convert the microcode loading failure msg type from `BIOS_INFO`
to `BIOS_ERR` to catch the error in proper.
TEST=Able to perform microcode loading on google/kano.
Signed-off-by: Subrata Banik <subratabanik@google.com>
Change-Id: I9a7cdc2d2c9211f1e0c7921015126f7a1be87761
Reviewed-on: https://review.coreboot.org/c/coreboot/+/65249
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
This patch introduces a newer API to reload the microcode patch when
SoC selects RELOAD_MICROCODE_PATCH config.
Expected to call this API being independent of CPU MP Init regular
flow hence, doesn't regress the boot time.
BUG=b:233199592
TEST=Build and boot google/kano to ChromeOS.
Signed-off-by: Subrata Banik <subratabanik@google.com>
Change-Id: If480e44b88d04e5cb25d7104961b70f7be041a23
Reviewed-on: https://review.coreboot.org/c/coreboot/+/65156
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Werner Zeh <werner.zeh@siemens.com>
Reviewed-by: Lean Sheng Tan <sheng.tan@9elements.com>
Reviewed-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
Prior commit hash 0310d34c2 (cpu/intel/microcode: Have provision to
re-load microcode patch) introduces an option to reload the microcode
based on SoC selecting RELOAD_MICROCODE_PATCH config.
This patch might potentially introduce a boot time regression (~30ms)
when RELOAD_MICROCODE_PATCH kconfig is enabled as all cores might end up
reloading the microcode without the proper need.
Note: RELOAD_MICROCODE_PATCH kconfig is not yet selected by any SoC
hence, it doesn't impact any coreboot project.
The idea is reloading microcode depends on specific use case
(for example: Skip FSP doing MP Init from Alder Lake onwards) hence,
a follow up patch will create a newer API to allow reloading of
microcode when RELOAD_MICROCODE_PATCH config is enabled.
BUG=b:233199592
TEST=Build and boot google/kano to ChromeOS.
Signed-off-by: Subrata Banik <subratabanik@google.com>
Change-Id: Ie320153d25cefe153fc8a67db447384f1f20f31f
Reviewed-on: https://review.coreboot.org/c/coreboot/+/65155
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Lean Sheng Tan <sheng.tan@9elements.com>
When switching to different board, 'make clean' needs to happen because
not everything gets properly regenerated. Microcode updates are among
those. You could end up with the microcode updates from the previous
build which can be incorrect. Adding $(DOTCONFIG) as a dependency which
gets updated when you change something in Kconfig fixes this.
TESTED: swap between boards that use different microcode and see that
the size changes.
Change-Id: Id1edecc28d492838904e3659f1fe8c9df0a69134
Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/65148
Reviewed-by: Nico Huber <nico.h@gmx.de>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Paul Menzel <paulepanter@mailbox.org>
This fixes following errors when building GA-945GCM-S2L with clang 14.0.5.
CC ramstage/cpu/x86/smm/smm_module_loader.o
src/cpu/x86/smm/smm_module_loader.c:180:10: error: format specifies type 'unsigned long' but the argument has type 'size_t' (aka 'unsigned int') [-Werror,-Wformat]
region_offset(&cpus[i].stub_code), i);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
src/cpu/x86/smm/smm_module_loader.c:184:20: error: format specifies type 'unsigned long' but the argument has type 'size_t' (aka 'unsigned int') [-Werror,-Wformat]
__func__, region_offset(&cpus[0].stub_code),
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
src/cpu/x86/smm/smm_module_loader.c:185:10: error: format specifies type 'unsigned long' but the argument has type 'size_t' (aka 'unsigned int') [-Werror,-Wformat]
region_offset(&cpus[i].stub_code), size);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
src/cpu/x86/smm/smm_module_loader.c:349:52: error: format specifies type 'unsigned long' but the argument has type 'size_t' (aka 'unsigned int') [-Werror,-Wformat]
printk(BIOS_DEBUG, "%-12s [0x%lx-0x%lx]\n", name, region_offset(®ion),
~~~ ^~~~~~~~~~~~~~~~~~~~~~
%zx
src/cpu/x86/smm/smm_module_loader.c:350:9: error: format specifies type 'unsigned long' but the argument has type 'size_t' (aka 'unsigned int') [-Werror,-Wformat]
region_end(®ion));
Signed-off-by: Elyes Haouas <ehaouas@noos.fr>
Change-Id: I59f20aacf91cb50fb194a84082a643b34c6c1ae5
Reviewed-on: https://review.coreboot.org/c/coreboot/+/65154
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
This patch provides an option to reload the microcode patch a.k.a
second microcode patch if SoC selects the required
RELOAD_MICROCODE_PATCH config.
There is a new feature requirement starting with ADL to re-load the
microcode patch as per new Mcheck initialization flow.
BUG=b:233199592
TEST=Build and boot google/taeko to ChromeOS. Able to re-load
microcode patch as below:
[INFO ] microcode: Re-load microcode patch
[INFO ] microcode: updated to revision 0x41b date=2022-03-08
Signed-off-by: Subrata Banik <subratabanik@google.com>
Change-Id: I0a3c29b3c25fccd31280a2a5a8d4fb22a6cf53bf
Reviewed-on: https://review.coreboot.org/c/coreboot/+/64833
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Tarun Tuli <taruntuli@google.com>
Use the common code to save data for fast boot or S3 resume.
An notable improvement that comes with this, is that the same 4K page
is not rewritten all the time. This prolongs the hardware's life.
TESTED on pcengines/apu1 and lenovo/g505s: S3 resume works fine.
Change-Id: I0f4f36dcead52a6c550fb5e606772e0a99029872
Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/44295
Reviewed-by: Mike Banon <mikebdp2@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-by: Michał Żygowski <michal.zygowski@3mdeb.com>
Save the regular boot MTRRs that are restored on the S3 path during
the CPU init in cbmem instead of storing them to the SPI flash.
This was probably done because historically this code run with late
cbmem init (in ramstage).
TESTED on pcengines/apu1 and lenovo/g505s: S3 resume works fine.
Change-Id: Ia58e7cd1afb785ba0c379ba75ef6090b56cb9dc6
Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/44294
Reviewed-by: Mike Banon <mikebdp2@gmail.com>
Reviewed-by: Paul Menzel <paulepanter@mailbox.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-by: Michał Żygowski <michal.zygowski@3mdeb.com>
This allows for some runtime checks on all SMM elements and removes
the need for manual checks.
We can drop completely separate codepaths on SMM_TSEG & SMM_ASEG as the
only difference is where permanent handler gets placed.
TESTED on prodrive/hermes and qemu with SSM_ASEG with 4 cores & SMM_TSEG
with 128 cores. This code figured out quite some problems with
overlapping regions so I think this is the right approach.
Change-Id: Ib7e2e3ae16c223ecfd8d5bce6ff6c17c53496925
Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/63602
Reviewed-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Lean Sheng Tan <sheng.tan@9elements.com>
Starting from Intel Pentium 4, cpus featured SSE2.
This will be used in the follow-up patches to determine whether to use
mfence as this instruction was introduced with the SSE2 feature set.
Change-Id: I8ce37d855cf84a9fb9fe9e18d77b0c19be261407
Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/64666
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Elyes Haouas <ehaouas@noos.fr>
Reviewed-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
The Intel Firmware Interface Table (FIT) is a bit of an annoying outlier
among CBFS files because it gets manipulated by a separate utility
(ifittool) after cbfstool has already added it to the image. This will
break file hashes created for CBFS verification.
This is not actually a problem when booting, since coreboot never
actually loads the FIT from CBFS -- instead, it's only in the image for
use by platform-specific mechanisms that run before coreboot's
bootblock. But having an invalid file hash in the CBFS image is
confusing when you want to verify that the image is correctly built for
verification.
This patch adds a new CBFS file type "intel_fit" which is only used for
the intel_fit (and intel_fit_ts, if applicable) file containing the FIT.
cbfstool will avoid generating and verifying file hashes for this type,
like it already does for the "bootblock" and "cbfs header" types. (Note
that this means that any attempt to use the CBFS API to actually access
this file from coreboot will result in a verification error when CBFS
verification is enabled.)
Signed-off-by: Julius Werner <jwerner@chromium.org>
Change-Id: I1c1bb6dab0c9ccc6e78529758a42ad3194cd130c
Reviewed-on: https://review.coreboot.org/c/coreboot/+/64736
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
