There were some issues with the current Linuxboot Makefiles.
- multithreaded compilation didn't work, because some prerequisites
were missing
- initramfs wasn't added for x86 qemu boot.
- riscv support was incomplete
It began with separate patches, but resulted in a clean up patch, that
is hard to separate. The most important changes are the following:
- Instead of phony targets, actual files are now used as prerequisites
- riscv can now be used as target
- initramfs works now also for x86
- instead of querying the most recent version from the internet, I set a
known working version (because I tested it) that can be customized
and/or upgraded in the future. The reasons:
- querying the version from the internet requires a constant
connection to the internet even after linux kernel is already
build (aka subsequent builds).
- one usually wants to use a known working version, but optionally
still have the posibillity to choose a custom one. This patch
introduces this possibility in its most simple form.
- I removed as much ifeq statements as possible and moved that
responsibility to Kconfig, because they tend to make the
Makefile less readable.
Signed-off-by: Maximilian Brune <maximilian.brune@9elements.com>
Change-Id: I25e757108e0dd473969fe5a192ad0733f1fe6286
Reviewed-on: https://review.coreboot.org/c/coreboot/+/76150
Reviewed-by: Lean Sheng Tan <sheng.tan@9elements.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
This patch increases the `tcc_offset` to reduce the TCC
(Thermal Control Circuit) activation temperature to avoid running
into abrupt power off during power cycle tests.
On Intel processors, the core frequency can be by an HW agent when
the current temperature reaches the TCC activation temperature.
The default TCC activation temperature is specified by MSR_IA32_TEMPERATURE_TARGET (which is 90°C for google/rex variants).
However, this patch adjusted the TCC by specifying an offset in
degrees C (i.e., using `tcc_offset` from variant override device tree).
Note: The bigger the TCC offset is, the lower the effective TCC activation temperature would be, to ensure that processors can be throttled earlier before the system critical overheats.
BUG=b:283008762
TEST=Able to perform power cycle on google/screebo w/o any crash/shutdown.
Change-Id: Ib19703877dbbfc26b2d9f538dda4f10c27cf872d
Signed-off-by: Subrata Banik <subratabanik@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/76658
Reviewed-by: Paul Menzel <paulepanter@mailbox.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
Hook the PchHdaSdiEnable UPD so that mainboard can change the
settings via devicetree. PchHdaSdiEnable UPD enable HDA SDI lanes.
BUG=b:268546941
TEST=Verified the settings on google/brya using debug FSP logs.
Signed-off-by: Bora Guvendik <bora.guvendik@intel.com>
Change-Id: I82bbfa5442936aefa53f8826e395b7ce75c895a3
Reviewed-on: https://review.coreboot.org/c/coreboot/+/76449
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Nick Vaccaro <nvaccaro@google.com>
Adding EC_HOST_EVENT_PANIC to SCI mask allows the EC to interrupt the
Kernel when an EC panic occurs. If system safe mode is also enabled
on the EC, the kernel will have a short period to extract and save info
about the EC panic.
BUG=b:290985698
BRANCH=firmware-volteer-13672.B
TEST=Observe kernel ec panic handler run when ec panics
Change-Id: I87173f93d0e47baa816d15dad0777007342b4fdb
Signed-off-by: Rob Barnes <robbarnes@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/76610
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
This patch creates a new variant rex4es. The new variant will support ESx samples. The existing rex variant will support the QS samples.
BUG=b:290732344
TEST=Able to build google/rex4es board and boot on target hardware.
Change-Id: I25dd1f42ee812f47289da0c2ef7aa79d6f340d48
Signed-off-by: Dinesh Gehlot <digehlot@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/76605
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
Reviewed-by: Subrata Banik <subratabanik@google.com>
This patch creates a rex model so that other variants developed using
`rex` baseboard are easy to land without duplicating the config
selection.
So far, `rex0` and `rex_ec_ish` are developed using the `rex` model.
The plan is to extend the support for `rex4es` and `rex4es_ec_ish`
variants.
TEST=Able to build and boot google/rex.
Change-Id: Id4e8d1162da93b7266ee1108f870e89b6d884ab9
Signed-off-by: Subrata Banik <subratabanik@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/76608
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Reviewed-by: Jakub Czapiga <jacz@semihalf.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
acpi.c contains architectural specific things like IOAPIC, legacy IRQ,
DMAR, HPET, ... all which require the presence of architectural headers.
Instead of littering the code with #if ENV_X86 move the functions to
different compilation units.
Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
Change-Id: I5083b26c0d4cc6764b4e3cb0ff586797cae7e3af
Reviewed-on: https://review.coreboot.org/c/coreboot/+/76008
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-by: Lean Sheng Tan <sheng.tan@9elements.com>
With arm64 -Wstack-usage= is enabled which is triggered on any use of
alloca(). Since this function basically works on x86 without wrecking
things and causing massive stack consumption it's unlikely to cause
problems on arm64.
Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
Change-Id: I5d445d151db5e6cc7b6e13bf74ce81007d819f1d
Reviewed-on: https://review.coreboot.org/c/coreboot/+/76007
Reviewed-by: Nico Huber <nico.h@gmx.de>
Reviewed-by: Tim Wawrzynczak <inforichland@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Lean Sheng Tan <sheng.tan@9elements.com>
AGESA S3 restore needs to occur before SMM finalization/locking,
but it's a crapshoot as to which runs first since both use the same
BS_OS_RESUME/BS_ON_ENTRY boot state callback, and there's no way
to prioritize/force ordering.
To work around this, move the AGESA S3 resume call to the preceding
boot state (BS_OS_RESUME_CHECK) to ensure it runs first, and guard it
to ensure it only runs on the S3 resume path.
BUG=none
TEST=build/boot google/liara, verify S3 resume successful.
Change-Id: I765db140c6708a0b129f79fb7d3dc8a4ab3095bd
Signed-off-by: Matt DeVillier <matt.devillier@amd.corp-partner.google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/76592
Reviewed-by: Fred Reitberger <reitbergerfred@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Marshall Dawson <marshalldawson3rd@gmail.com>
This supports a brand new I2C driver that is designed specifically
for the Pixel 2013 chromebook (LINK). The GMBus interface on the IGPU
is an i2c-compatible interface, but AFAIK only Link has touch devices
attached in this way.
On Windows, the PCIe device for the IGP is owned by the Intel
proprietary driver, hence a separate ACPI device has to be added for
the I2C driver arbitrator to attach to. The MMIO method is used instead
of _CRS so that Windows does not try to assign ownership of the
resource to our device (even though we're using the MMIO registers at
the same time as the IGP driver).
Even though in theory 2 drivers accessing the same MMIO may cause
problems, in testing, there has been no issues with
sleep/wake/hibernate, updating/installing/uninstalling the IGP driver,
or changing display resolutions with the i2c driver attached.
The arbitrator is necessary as well, since even though there are
multiple i2c buses, the MMIO registers are shared. Hence a shared lock
is required for i2c access across the buses.
The original Sleep Button devices are preserved for Linux due to the
completely custom and non-standard implementation of the Windows driver
in order to work around the non-standard nature of Link's hardware.
Change-Id: If7ee05d15bc17d335cf8c1a8e80bea62800de475
Signed-off-by: CoolStar <coolstarorganization@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/76159
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Nico Huber <nico.h@gmx.de>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Reviewed-by: Matt DeVillier <matt.devillier@gmail.com>
From the Linux documentation (Documentation/PCI/acpi-info.rst):
[6] PCI Firmware 3.2, sec 4.1.2:
If the operating system does not natively comprehend reserving the
MMCFG region, the MMCFG region must be reserved by firmware. The
address range reported in the MCFG table or by _CBA method (see Section
4.1.3) must be reserved by declaring a motherboard resource. For most
systems, the motherboard resource would appear at the root of the ACPI
namespace (under \_SB) in a node with a _HID of EISAID (PNP0C02), and
the resources in this case should not be claimed in the root PCI bus’s
_CRS. The resources can optionally be returned in Int15 E820 or
EFIGetMemoryMap as reserved memory but must always be reported through
ACPI as a motherboard resource.
So in order for the OS to use ECAM MMCONF over legacy PCI IO
configuration, a PNP0C02 HID device needs to reserve this region.
As no AMD platform has this defined in DSDT this fixes Linux using
legacy PCI IO configuration over MMCONF. Tianocore messes with e820
table in such a way that it prevents Linux from using PCIe ECAM. This
change fixes that problem.
Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
Change-Id: I852e393726a1b086cf582f4d2d707e7cde05cbf4
Reviewed-on: https://review.coreboot.org/c/coreboot/+/75729
Reviewed-by: Felix Held <felix-coreboot@felixheld.de>
Reviewed-by: Sean Rhodes <sean@starlabs.systems>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
There is a problem of screen shake on the old panel[1]. So increase the
panel GOP component pull-down circuit size in hardware, and update the
initialization code at the same time. The new initialization code is
mainly adjusted for GOP timing. When Display sleep in, raise all GOP
signals to VGHO and then drop to GND. In order to be consistent with
the current panel model, let's rename this file.
[1]: INX old panel product number is HJ110IZ-01A-B1, and the new
panel product number is HJ110IZ-01A-B2. We have recalled the shipment
old panel.
BUG=b:270276344
BRANCH=trogdor
TEST= test firmware display pass
Change-Id: I2b2534afee1ed700c39d3c360aafd685b63ccbfb
Signed-off-by: Cong Yang <yangcong5@huaqin.corp-partner.google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74973
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Julius Werner <jwerner@chromium.org>
This patch updates the Type-C USB2/3 port mapping to reflect the mux
connection change as mentioned in previous patch
commit ee3f796200 (mb/google/rex/var/ovis: Fix mux
change as per schematics).
Here is the correct port mapping after considering the mux swap:
+--------------------------------+-------------+---------------+
| TCSS-USB Mapping | Port C0 | Port C1 | Port C2 |
+------------------+-------------+-------------+---------------+
| USB2-Port | 2 | 3 | 1 |
| USB3-Port | 0 | 2 | 1 |
+------------------+-------------+-------------+---------------+
BUG=b:289300284
TEST=Able to build and boot google/ovis to get display over Type-C1
and Type-C2 port.
Change-Id: I460004842dd8fcdc03fca6639d03e422259380ca
Signed-off-by: Subrata Banik <subratabanik@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/76464
Reviewed-by: Paul Menzel <paulepanter@mailbox.org>
Reviewed-by: Nick Vaccaro <nvaccaro@google.com>
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Patch to increase CONSOLE_CBMEM_BUFFER_SIZE to contain FSP debug serial log.
The existing implementation uses larger cbmem size irrespective of FSP debug is enabled or
not. Ideally. larger cbmem size is required only if FSP debug is enabled.
Bug=b:284124701
TEST=Able to build and boot google/marasov.
Change-Id: I9a9e660f2738813808e0dd65d2783424b49f9a5e
Signed-off-by: Subrata Banik <subratabanik@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/76541
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Reduces the CAR (Cache-as-RAM) variable usage while using FSP debug binaries, which can prevent the CAR from becoming too full and unable
to integrate other CAR global variables.
This change has the following downsides:
- FSP debug output into the cbmem buffer will be partial.
To test this change, you can:
Build and boot google/rex without any function impact with non-serial
and serial FSP debug image (unless what has been documented here).
Bug=b:284124701
TEST=Able to build and boot google/rex.
Change-Id: I16a1aa25fd32327d03a37381a696c86c95014ba0
Signed-off-by: Subrata Banik <subratabanik@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/76540
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
According to Intel doc#763797 to overcome early command training hang
issue, the CsPiStartHighinEct needs to be enabled for hynix memory.
BUG=b:281643325
BRANCH=firmware-brya-14505.B
TEST=Built and booted into OS.
Change-Id: I95702e675fa3b73c7e8ee0c8625c7828d8129ea8
Signed-off-by: Nick Vaccaro <nvaccaro@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/76355
Reviewed-by: Subrata Banik <subratabanik@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
This commit provides option for board to set CsPiStartHighinEct
FSP UPD using a new cs_pi_start_high_in_ect mb_cfg field.
BUG=b:279835630
BRANCH=none
TEST=CsPiStartHighinEct UPD is set properly
Change-Id: I7d0d5f3c782e29fb047ea421e1a5fdfc30bcc26d
Signed-off-by: Nick Vaccaro <nvaccaro@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/76354
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Subrata Banik <subratabanik@google.com>
Disable the CSME by default now that S3 is used instead of S0ix.
The CSME will not go into a low power state during S0ix when it is
disabled. This prevents the CPU from reaching C10 and so increases the
power usage during suspend compared to leaving CSME enabled. (This was
measured to be a ~2W different on TGL-U.) In S3, the state of the CSME
doesn't matter because the CPU will be off.
Change-Id: I88c0aebdcc977f3ba9dd8f46a6abfaa7a4ae8eb6
Signed-off-by: Tim Crawford <tcrawford@system76.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/73354
Reviewed-by: Paul Menzel <paulepanter@mailbox.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Jeremy Soller <jeremy@system76.com>
After fixing TPM logs clobbering other regions in CB:73297, S3 no longer
causes cache issues resulting in power off after multiple suspends.
This is required for disabling Intel CSME by default.
Change-Id: I7eef4c883fd65db93dae81adabd895b2de90496a
Signed-off-by: Tim Crawford <tcrawford@system76.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/73353
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Jeremy Soller <jeremy@system76.com>
The preram TPM log was being copied to the end of the CBMEM TPM log no
matter what the size of the CBMEM TPM log was. Eventually, it would
overwrite anything else in CBMEM beyond the TPM log.
This can currently be reproduced by enabling TPM_MEASURED_BOOT and
performing multiple S3 suspends, as coreboot is incorrectly performing
TPM measurements on S3 resume.
Change-Id: If76299e68eb5ed2ed20c947be35cea46c51fcdec
Signed-off-by: Jeremy Soller <jeremy@system76.com>
Signed-off-by: Tim Crawford <tcrawford@system76.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/73297
Reviewed-by: Martin L Roth <gaumless@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
The original default, minimum abbreviated hash length was 7. It dif-
fers on newer systems, however. This breaks reproducibility, so set
an explicit length. 12 hex digits should be good enough.
Note: This sets only a minimum. With a high enough number of commit
objects in the repository, Git could still decide to use a longer
hash, again breaking reproducibility. 12 digits will hopefully pro-
vide enough margin.
Change-Id: Ia86e9cc41e27a0a57d498dcb13aec954c4ea0f04
Signed-off-by: Nico Huber <nico.h@gmx.de>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/76560
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martin.roth@amd.corp-partner.google.com>
The current Sapphire Rapids code assumes that all sockets have working
CPUs. On multi-socket platforms a CPU might be missing or was disabled
due to an error. The variable PlatformData.numofIIO and the variable
SystemStatus.numCpus reflect the working CPUs, but not the actual
socket count.
Update the code to iterate over sockets until PlatformData.numofIIO
IIOs have been found. This is required as FSP doesn't sort IIOs by
working/non working status.
This resolves invalid ACPI table generation and it fixes a crash
as commands were sent to a disabled CPU.
TEST: Disabled Socket1 on IBM/SBP1.
Change-Id: I237b6392764bbdb3b96013f577a10a4394ba9c6e
Signed-off-by: Patrick Rudolph <patrick.rudolph@9elements.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/76559
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
