The patch stores the ISH in the CBMEM table. It verifies CSE has been
updated by comparing previous and current CSE versions. If it has, the
patch updates the previous CSE version with the current CSE version. It
then updates the CBMEM table with the current ISH version.
BUG=b:273661726
TEST=The current and old CSE and ISH versions are verified on the
google/nissa during cold and warm reboots.
Additionally, version updates are verified by a debug patch that
purposely updated the stored cse version.
Signed-off-by: Dinesh Gehlot <digehlot@google.com>
Change-Id: Ie5c5faf926c75b05d189fb1118020fff024fc3e0
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74208
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Kangheui Won <khwon@chromium.org>
The patch implements an API that stores the CSE firmware version in the
CBMEM table. The API will be called from RAMSTAGE based on boot state
machine BS_PRE_DEVICE/BS_ON_EXIT
Additionally, renamed ramstage_cse_fw_sync() to ramstage_cse_misc_ops()
in order to add more CSE related operations at ramstage.
This patch also adds a configuration option,
'SOC_INTEL_STORE_CSE_FPT_PARTITION_VERSION', which enables the storage
of firmware version information in CBMEM memory. This information can be
used to identify the firmware version that is currently installed on the
system. The option depends on the `DRIVERS_INTEL_ISH` config and
platform should be flexible enough to opt out from enabling this
feature.
The cost of sending HECI command to read the CSE FPT is significant
(~200ms) hence, the idea is to read the CSE RW version on every cold
reset (to cover the CSE update scenarios) and store into CBMEM to
avoid the cost of resending the HECI command in all consecutive warm
boots.
Later boot stages can just read the CBMEM ID to retrieve the ISH
version if required.
Finally, ensure this feature is platform specific hence, getting
enabled for the platform that would like to store the ISH version into
the CBMEM and parse to perform some additional work.
BUG=b:273661726
TEST=Able to build and boot google/marasov.
Signed-off-by: Dinesh Gehlot <digehlot@google.com>
Change-Id: I923049d2f1f589f87e1a29e1ac94af7f5fccc2c8
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74256
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Lean Sheng Tan <sheng.tan@9elements.com>
This change adds the VPU into the DMAR SATC table in order to support
the VPU IO virtualization.
BUG=None
TEST=Enabled the VPU, booted to kernel and verified that DMAR SATC table
includeded the VPU entry.
Change-Id: I6d4af7c9844e33483a1e616eaee061a90d0be6fc
Signed-off-by: John Zhao <john.zhao@intel.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74221
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
This patch refactors cse_fw_sync() function to include timestamp
associated with the CSE sync operation.This effort will ensure the
SoC code just makes a call into the cse_fw_sync() without bothering
about adding timestamp entries.
TEST=Able to build and boot google/marasov.
Signed-off-by: Subrata Banik <subratabanik@google.com>
Change-Id: Ib5e8fc2b8c3b605103f7b1238df5a8405e363f83
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74582
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
Reviewed-by: Kangheui Won <khwon@chromium.org>
To use generated build.h one should have had a pre-requisite in the
Makefile. Reference coreboot_build_date from lib/version.c instead.
Change-Id: Icd6fa2ddf8aa584b0f51ba130592f227bbdad975
Signed-off-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74447
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Nico Huber <nico.h@gmx.de>
Reviewed-by: Felix Held <felix-coreboot@felixheld.de>
After the obsoletion of Processor() it is necessary to provide
_CST package to define P_LVLx IO addresses for C2/C3 transitions.
The latency values from _CST will always replace those in FADT.
Change-Id: I3230be719659fe9cdf9ed6ae73bc91b05093ab97
Signed-off-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74430
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Felix Held <felix-coreboot@felixheld.de>
srcclk_pin is 0-based and '0' is a valid clock source number. If
srcclk_pin is set to -1, then the clock will not be disabled in D3.
Therefore, clock source gating method should not be generated.
BUG=b:271003060
BRANCH=firmware-brya-14505.B
TEST=Boot to OS and check that rtd3 ACPI entries are generated as
expected. For those PCI devices with RTD3 driver whose srcclk_pin to
0, the RTD3 entries should not be missing due to check error.
Signed-off-by: Cliff Huang <cliff.huang@intel.com>
Change-Id: Ia831b8fd17572cc35765bd226d1db470f12ddd41
Reviewed-on: https://review.coreboot.org/c/coreboot/+/73889
Reviewed-by: Jérémy Compostella <jeremy.compostella@intel.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Bora Guvendik <bora.guvendik@intel.com>
This patch sends the CSE EOP command asynchronous implementation early
as part of `soc_init_pre_device`.
Without this patch the duration between asynchronous CSE EOP send and
receive commands is not ample which causes idle delay while waiting
for EOP response.
The goal of the CSE async implementation is to avoid idle delay while
capturing the response from CSE EOP cmd. This patch helps to create
ample duration between CSE EOP command being sent and response being
captured.
TEST=Able to boot google/rex sku to ChromeOS and observed ~100ms of
boot time savings (across warm and cold reset scenarios)
Change-Id: I91ed38edbd5a31d61d4888e1466169a3494d635a
Signed-off-by: Subrata Banik <subratabanik@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74561
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
Reviewed-by: Tarun Tuli <taruntuli@google.com>
Remove the `TcssD3ColdDisable` option in devicetree, as it exists
in Kconfig. The setting is currently unused.
Signed-off-by: Sean Rhodes <sean@starlabs.systems>
Change-Id: I08930ef84438140a13df74900570b126088bd1cb
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74478
Reviewed-by: Nico Huber <nico.h@gmx.de>
Reviewed-by: Michael Niewöhner <foss@mniewoehner.de>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Remove the `TcssD3ColdDisable` option in devicetree, as it exists
in Kconfig. The setting is currently unused.
Signed-off-by: Sean Rhodes <sean@starlabs.systems>
Change-Id: I2590e8dec0a308e0dc3d467cb3dd2bb97e877492
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74477
Reviewed-by: Nico Huber <nico.h@gmx.de>
Reviewed-by: Michael Niewöhner <foss@mniewoehner.de>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Remove the `TcssD3ColdDisable` option in devicetree, as it exists
in Kconfig. The setting is only used on `starlabs/starbook` which
selects D3COLD_SUPPORT so the UPDs will not change.
Signed-off-by: Sean Rhodes <sean@starlabs.systems>
Change-Id: I50e49e900c96748edd5b678765e47cc0e0d9b280
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74476
Reviewed-by: Nico Huber <nico.h@gmx.de>
Reviewed-by: Michael Niewöhner <foss@mniewoehner.de>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Use D3COLD_SUPPORT Kconfig option to set the maximum support sleep
state. Report `4` in `_S0W` only when D3COLD_SUPPORT is enabled, as
if it is not, it will break S3 exit.
When D3COLD_SUPPORT is not enabled, return `3` (D3Hot).
This fixed S3 exit on both TGL and ADL. Tested on StarBook
Mk V and Mk VI.
Signed-off-by: Sean Rhodes <sean@starlabs.systems>
Change-Id: I578d4933b6144aec79fe0b2eb168338ef82c0b9d
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74406
Reviewed-by: Nico Huber <nico.h@gmx.de>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Michael Niewöhner <foss@mniewoehner.de>
The Kconfig option SOC_INTEL_TIGERLAKE_S3 suggests that it's doing
something with S3, but it's actually disabling D3Cold support.
Remove it, and instead use D3COLD_SUPPORT so it's clear what the
option is doing.
Signed-off-by: Sean Rhodes <sean@starlabs.systems>
Change-Id: Id43f3e5c8620d474831cc02fcecebd8aac961687
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74405
Reviewed-by: Nico Huber <nico.h@gmx.de>
Reviewed-by: Michael Niewöhner <foss@mniewoehner.de>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Move D3COLD_SUPPORT to device, so it can be used by multiple
SOCs.
Signed-off-by: Sean Rhodes <sean@starlabs.systems>
Change-Id: Ie92736458ab95374c51346107665dc0fd1e653a4
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74404
Reviewed-by: Nico Huber <nico.h@gmx.de>
Reviewed-by: Michael Niewöhner <foss@mniewoehner.de>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
The Intel FSP used on ChromeOS platform has dropped the
`CpuFeaturesPei.ffs` module to opt for coreboot running this
additional feature programming on BSP and APs.
TEST=Able to build and boot google/rex without any boot regression.
Please refer to the boot time and SPI flash savings after dropping
the FSP feature programming:
Boot time savings=10ms
SPI Flash size savings=34KB
Signed-off-by: Subrata Banik <subratabanik@google.com>
Change-Id: Iaed0a009813098610190b2a3a985b0748c0d51de
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74168
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
The Kconfig option SOC_INTEL_ALDERLAKE_S3 suggests that it's doing
something with S3, but it's actually disabling D3Cold support.
Rename it to D3COLD_SUPPORT to make it clear what it's doing.
Signed-off-by: Sean Rhodes <sean@starlabs.systems>
Change-Id: Ifc3f19912ac7ee55be8ec7a491598140f9532675
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74403
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Nico Huber <nico.h@gmx.de>
Reviewed-by: Lean Sheng Tan <sheng.tan@9elements.com>
Reviewed-by: Michael Niewöhner <foss@mniewoehner.de>
Add function needed to generate ACPI backlight control SSDT, along with
Kconfig values for accessing the registers.
Tested by adding gfx register on google/magpie. Backlight controls
work on Windows 10 and Linux 6.1.
Change-Id: Iaa9872cd590c3b1298667cc80354ed3efd91c6c8
Signed-off-by: Matt DeVillier <matt.devillier@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74324
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Sean Rhodes <sean@starlabs.systems>
The patch moves API that gets the CSE FW partition information into
CSE Lite specific file aka cse_lite.c because the consumer of this API
is the cse_lite specific ChromeOS devices hence, it's meaningful to
move the cse lite specific implementation inside cse_lite.c file.
BUG=b:273661726
TEST=Able to build and boot google/marasov with this code change.
Signed-off-by: Subrata Banik <subratabanik@google.com>
Change-Id: I49ffaec467f6fb24327de3b2882e37bf31eeb7cf
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74382
Reviewed-by: Kangheui Won <khwon@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Enable early caching of the TOM region to optimize the boot time by
selecting `SOC_INTEL_COMMON_BASECODE_RAMTOP` config.
Purpose of this feature is to cache the TOM (with a fixed size of
16MB) for all consecutive boots even before calling into the FSP.
Otherwise, this range remains un-cached until postcar boot stage
updates the MTRR programming. FSP-M and late romstage uses this
uncached TOM range for various purposes (like relocating services
between SPI mapped cached memory to DRAM based uncache memory) hence
having the ability to cache this range beforehand would help to
optimize the boot time (more than 50ms as applicable).
Signed-off-by: Lean Sheng Tan <sheng.tan@9elements.com>
Change-Id: I3b68d13aa414e69c0a80122021e6755352db32fd
Reviewed-on: https://review.coreboot.org/c/coreboot/+/73738
Reviewed-by: Subrata Banik <subratabanik@google.com>
Reviewed-by: Sean Rhodes <sean@starlabs.systems>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Enable early caching of the TOM region to optimize the boot time by
selecting `SOC_INTEL_COMMON_BASECODE_RAMTOP` config.
Purpose of this feature is to cache the TOM (with a fixed size of
16MB) for all consecutive boots even before calling into the FSP.
Otherwise, this range remains un-cached until postcar boot stage
updates the MTRR programming. FSP-M and late romstage uses this
uncached TOM range for various purposes (like relocating services
between SPI mapped cached memory to DRAM based uncache memory) hence
having the ability to cache this range beforehand would help to
optimize the boot time (more than 50ms as applicable).
TEST=Able to build and boot Starlab ADL laptop to OS.
Signed-off-by: Lean Sheng Tan <sheng.tan@9elements.com>
Change-Id: Iba554af4ff0896e133d20860ff72dd1a10ebd1e3
Reviewed-on: https://review.coreboot.org/c/coreboot/+/73736
Reviewed-by: Subrata Banik <subratabanik@google.com>
Reviewed-by: Sean Rhodes <sean@starlabs.systems>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Maximilian Brune <maximilian.brune@9elements.com>
This patch adds CPU ID for B0 stepping (aka ES2).
DOC=#723567
TEST=Able to boot on B0 rvp and get correct CPU Name in coreboot log.
Signed-off-by: Musse Abdullahi <musse.abdullahi@intel.com>
Change-Id: I8b939ccc8b05e3648c55f8f2a0a391cb08f04184
Signed-off-by: Musse Abdullahi <musse.abdullahi@intel.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74300
Reviewed-by: Paul Menzel <paulepanter@mailbox.org>
Reviewed-by: Subrata Banik <subratabanik@google.com>
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Boards with SOC_INTEL_COMMON_BLOCK_ACPI_CPU_HYBRID have
special handling for the time being.
Change of aopen/dxplplusu is coupled with sb/intel/i82801dx.
Change of emulation/qemu-i440fx is coupled with intel/i82371eb.
For asus/p2b, this adds MADT LAPIC entries, even though platform
has ACPI_NO_MADT selected. Even previously ACPI_NO_MADT creates
the MADT, including an entry for LAPIC address.
Change-Id: I1f8d7ee9891553742d73a92b55a87c04fa95a132
Signed-off-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74316
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Nico Huber <nico.h@gmx.de>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-by: Felix Held <felix-coreboot@felixheld.de>
This fixes the Jenkins build error when building INTEL_ARCHERCITY_CRB
that was caused by the API change in commit 36e6f9bc04. This patch removes the
broken API function and also adds package_id log print same as previous
commit mentioned above.
Signed-off-by: Lean Sheng Tan <sheng.tan@9elements.com>
Change-Id: I89e14b40186007ab0290b24cd6bd58015be376b6
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74436
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Sean Rhodes <sean@starlabs.systems>
This patch avoids cannonlake base config to select eNEM for CAR by
default. Rather allow other SoC config to choose the applicable CAR
mode between eNEM and NEM.
CML and WHL select eNEM whereas CFL decided to use NEM for CAR setup.
Here is some background about why CFL SoC platform decided to choose
NEM over eNEM:
It was found that some coffeelake CPUs like Intel i3 9100E fail to enter
CAR mode because some MSR used by NEM enhanced are lacking. According to
the Intel SDM CPUID.EAX=07h.ECX=0 reg EBX[12 or 15] should indicate the
presence of IA32_PAR_ASSOC and CPUID.EAX=10h.ECX[1 or 2] reg ECX[2]
should indicate IA32_L3_QOS_CFG and IA32_L2_QOS_CFG respectively but
even on a Intel coffeelake CPU that works with the NEM_ENHANCED these
CPUID bits are all 0 so there is no way of knowing whether NEM_ENHANCED
will work at runtime. Instead just always use regular NEM.
Signed-off-by: Subrata Banik <subratabanik@google.com>
Change-Id: Ibeaa4d53279ff9cbcd0b2ac5f2ad71925872355b
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74377
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Currently the xeon_sp code reassigns struct devices apic_id so that srat
entries can be added in a certain order.
This is not a good idea as it breaks thread local storage which contains
a pointer to its struct device cpu.
This moves the sorting of the lapic_ids to the srat table generation
and adds the numa node id in each core init entry. Now it is done in
parallel too as a bonus.
Change-Id: I372bcea1932d28e9bf712cc712f19a76fe3199b1
Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/68912
Reviewed-by: Patrick Rudolph <siro@das-labor.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Avoid asserts related to CNVi UPDs which are not boot critical.
Instead, add error messages which are more helpful in identifying
the issue.
BUG=none
TEST=Boot to the OS on google/rex
Signed-off-by: Kapil Porwal <kapilporwal@google.com>
Change-Id: I49a988b7eda009456d438ba7be0d2918826e1c36
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74370
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Subrata Banik <subratabanik@google.com>
This patch retrieves details of a specified firmware partition table.
The information retrieved includes the current firmware version and
other information about the firmware partition. The patch communicates
with the ME using the HECI command to acquire this information.
BUG=b:273661726
Test=Verified the changes for ISH partition on nissa board.
Signed-off-by: Dinesh Gehlot <digehlot@google.com>
Change-Id: I0582010bbb836bd4734f843a8c74dee49d203fd8
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74005
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Sridhar Siricilla <sridhar.siricilla@intel.com>
Hook the newly created/exposed CnviWifiCore UPD up as a chip driver.
Enable this option by default to maintain the existing behavior.
BUG=b:270985197
TEST=Verified by enabling/disabling the UPD on google/rex
Signed-off-by: Kapil Porwal <kapilporwal@google.com>
Change-Id: I5b4662c2a064f7c9074797c8a2541dcf1dd686fe
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74306
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Reviewed-by: Subrata Banik <subratabanik@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
The patch updates total cpu count variable and total P-core count in
cpu_apic_info_type structure to `unsigned short int` to address more
cores.
TEST=Verify the build on Rex
Signed-off-by: Sridhar Siricilla <sridhar.siricilla@intel.com>
Change-Id: I46239cc7ad9870e7134955af56b9f6625be2b002
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74305
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-by: Subrata Banik <subratabanik@google.com>
Processor attached memory should not use reserved_ram_from_to and
treat the calculation of gi_mem_size size as 64MB.
By default SOC_INTEL_HAS_CXL is enabled for Sapphire Rapids platforms,
this should fix small total memory issue. Before the fix running
command 'free -g -h' under Linux shows the total memory is only 1.4Gi,
after the fix it's showing the expected total memory size 15Gi.
Tested=On AC without attaching CXL memory, the total memory size is
the same as de-selecting SOC_INTEL_HAS_CXL.
On OCP Crater Lake with CXL memory attached, CXL memory can be recognized
in NUMA node 1:
numactl -H
available: 2 nodes (0-1)
node 0 cpus: 0 1 2 3 4 5 6 .. 59
node 0 size: 95854 MB
node 0 free: 93860 MB
node 1 cpus:
node 1 size: 63488 MB
node 1 free: 63488 MB
node distances:
node 0 1
0: 10 14
1: 14 10
Change-Id: I38e9d138fd284620ac616a65f444e943f1774869
Signed-off-by: Johnny Lin <johnny_lin@wiwynn.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74296
Reviewed-by: Patrick Rudolph <patrick.rudolph@9elements.com>
Reviewed-by: Shuo Liu <shuo.liu@intel.com>
Reviewed-by: Johnny Lin <Johnny_Lin@wiwynn.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Lean Sheng Tan <sheng.tan@9elements.com>
The Kconfig is only used in common code to gather the build time
maximum socket number FSP support. The same information is available
in FSP header as MAX_SOCKET, thus use the FSP as truth of source.
Currently MAX_SOCKET is 4.
Change-Id: I10282c79dbf5d612c37b7e45b900af105bb83c36
Signed-off-by: Patrick Rudolph <patrick.rudolph@9elements.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74339
Reviewed-by: Paul Menzel <paulepanter@mailbox.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-by: Johnny Lin <Johnny_Lin@wiwynn.com>
Reviewed-by: Lean Sheng Tan <sheng.tan@9elements.com>
Set the P2SB device as hidden as FSP-S is hiding the PCI configuration
space from coreboot on Alder Lake systems.
Signed-off-by: Michał Żygowski <michal.zygowski@3mdeb.com>
Change-Id: I5cfde7c1f6791578a03d73e89bcde31af608f12d
Reviewed-on: https://review.coreboot.org/c/coreboot/+/69950
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Reviewed-by: Subrata Banik <subratabanik@google.com>
P2SB device is being hidden from coreboot by FSP-S. This breaks the
resource allocator which does not report P2SB BAR via intel common
block P2SB driver. Hook up the common block P2SB driver ops to
soc_enable function so that the resources will be reported. The P2SB
device must be set as hidden in the devicetree.
This fixes the silent resource allocation conflicts on machines with
devices having big BARs which accidentally overlapped P2SB BAR.
TEST=Boot MSI PRO Z690-A with multiple PCIe devices/dGPUs with big
BARs and see resource conflicts no longer occur.
Signed-off-by: Michał Żygowski <michal.zygowski@3mdeb.com>
Change-Id: I7c59441268676a8aab075abbc036e651b9426057
Reviewed-on: https://review.coreboot.org/c/coreboot/+/69949
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Reviewed-by: Subrata Banik <subratabanik@google.com>
Fix a couple of bugs in the _OSC method for handling
"PCI Host Bridge Device" on Xeon-SP.
- Drop the Sleep. The code doesn't write to hardware at all, so
there's no need to sleep here.
- Make sure that the number of DWORD passed in Arg2 is at least 3.
The existing check was useless as it would not create the
DWordField, but then use it anyways.
- Add check for CXL 2 device method calls which provide a 5 DWORD
long buffer to prevent buffer overflows when invoking the
"PCI Host Bridge Device" method.
Test:
Boot on Archer City and confirm that no ACPI errors are reported
for _OSC.
Change-Id: Ide598e386c30ced24e4f96c37f2b4a609ac33441
Signed-off-by: Patrick Rudolph <patrick.rudolph@9elements.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74231
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Jonathan Zhang <jon.zhixiong.zhang@gmail.com>
According to ADL PCH BIOS specification (DOC# 630603) ADL-S PCH
uses a fixed SBREG_BAR of 256MiB starting at 0xe0000000.
Signed-off-by: Michał Żygowski <michal.zygowski@3mdeb.com>
Change-Id: Ied59a6dad8fb065dc3aeb6281bd32074aaa5e3b8
Reviewed-on: https://review.coreboot.org/c/coreboot/+/69948
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Now the CPU topology is filled in struct device during mp_init.
Change-Id: I7322b43f5b95dda5fbe81e7427f5269c9d6f8755
Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/69223
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Lean Sheng Tan <sheng.tan@9elements.com>
This patch sends the CSE EOP command asynchronous implementation early
as part of `soc_init_pre_device`.
Without this patch the duration between asynchronous CSE EOP send and
receive commands is not ample whichcauses idle delay while waiting
for EOP response.
The goal of the CSE async implementation is to avoid idle delay while
capturing the response from CSE EOP cmd.
This patch helps to create ample duration between CSE EOP command
being sent and response being captured.
TEST=Able to boot google/marasov EVT sku to ChromeOS and observed
~30ms of boot time savings (across warm and cold reset scenarios).
Without this patch:
963:returning from FspMultiPhaseSiInit 907,326 (97,293)
...
...
115:finished elog init 967,343 (2,581)
942:before sending EOP to ME 967,821 (478)
…
16:finished LZMA decompress (ignore for x86) 1,017,937 (12,135)
943:after sending EOP to ME 1,067,799 (49,861)
…
…
1101:jumping to kernel 1,144,587 (13,734)
Total Time: 1,144,549
With this patch:
963:returning from FspMultiPhaseSiInit 918,291 (97,320)
942:before sending EOP to ME 918,522 (230)
...
...
16:finished LZMA decompress (ignore for x86) 1,029,476 (12,483)
943:after sending EOP to ME 1,033,456 (3,980)
...
...
1101:jumping to kernel 1,111,410 (14,007)
Total Time: 1,111,375
Change-Id: Idaf45ef28747bebc02347f0faa77cc858a4a8ef1
Signed-off-by: Subrata Banik <subratabanik@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74293
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Tarun Tuli <taruntuli@google.com>
Currently coreboot presents the BSP core first, then efficient cores and
Performance cores as indicated below:
```
/sys/devices/system/cpu/cpu0/topology/thread_siblings_list:0-1
/sys/devices/system/cpu/cpu4/topology/thread_siblings_list:4
/sys/devices/system/cpu/cpu5/topology/thread_siblings_list:5
/sys/devices/system/cpu/cpu6/topology/thread_siblings_list:6
/sys/devices/system/cpu/cpu7/topology/thread_siblings_list:7
/sys/devices/system/cpu/cpu1/topology/thread_siblings_list:0-1
/sys/devices/system/cpu/cpu2/topology/thread_siblings_list:2-3
/sys/devices/system/cpu/cpu3/topology/thread_siblings_list:2-3
```
Existing code presents mix of different cores to OS and causes CPU load
balancing and power/performance impact. So, the patch fixes this
disorder by ordering the Performance cores first, compute die efficient
cores next, and finally SOC efficient cores if they are present. This
is done to run the media applications in a power efficient manner,
please refer the ChromeOS patches for details:
https://chromium-review.googlesource.com/c/chromiumos/platform2/+/3963893
BUG=b:262886449
TEST=Verified the code on Rex system
After the fix:
```
/sys/devices/system/cpu/cpu0/topology/thread_siblings_list:0-1
/sys/devices/system/cpu/cpu1/topology/thread_siblings_list:0-1
/sys/devices/system/cpu/cpu2/topology/thread_siblings_list:2-3
/sys/devices/system/cpu/cpu3/topology/thread_siblings_list:2-3
/sys/devices/system/cpu/cpu4/topology/thread_siblings_list:4
/sys/devices/system/cpu/cpu5/topology/thread_siblings_list:5
/sys/devices/system/cpu/cpu6/topology/thread_siblings_list:6
/sys/devices/system/cpu/cpu7/topology/thread_siblings_list:7
```
Change-Id: I21487a5eb0439ea0cb5976787d1769ee94777469
Signed-off-by: Sridhar Siricilla <sridhar.siricilla@intel.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/72132
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Jan Samek <jan.samek@siemens.com>
Reviewed-by: Sukumar Ghorai <sukumar.ghorai@intel.com>
Reviewed-by: Ronak Kanabar <ronak.kanabar@intel.com>
Set the `SOC_INTEL_CSE_SEND_EOP_ASYNC' flag to request End-Of-Post
right after PCI enumeration and handle the command response at
`BS_PAYLOAD_BOOT'.
With these settings we have observed a boot time reduction of about 20
to 30 ms on brya0.
BUG=b:268546941
BRANCH=firmware-brya-14505.B
TEST=Tests on brya0 with `SOC_INTEL_CSE_SEND_EOP_ASYNC' show
End-Of-Post after PCI initialization and EOP message received at
`BS_PAYLOAD_BOOT'.
Signed-off-by: Jeremy Compostella <jeremy.compostella@intel.com>
Change-Id: Ib850330fbb9e84839eb1093db054332cbcb59b41
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74215
Reviewed-by: Tarun Tuli <taruntuli@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Nick Vaccaro <nvaccaro@google.com>
coreboot supports three instances of sending EOP:
1. At CSE `.final' device operation
2. Early as with Alder Lake in chip_operations.init if
`SOC_INTEL_CSE_SEND_EOP_EARLY' is selected
3. At BS_PAYLOAD_BOOT as designed for Meteor Lake if
`SOC_INTEL_CSE_SEND_EOP_LATE' is selected
Currently, Alder Lake uses #3 as it results in better and more stable
boot time. However, what would deliver even better result is to not
actively wait for CSE completion.
This patch introduces a new `SOC_INTEL_CSE_SEND_EOP_ASYNC' Kconfig
which split the action of sending EOP request and receiving EOP
completion response from the CSE.
This patch used in conjunction with #1 can significantly
improves the overall boot time on a Raptor Lake design. For example
`SOC_INTEL_CSE_SEND_EOP_ASYNC' on a skolas board can deliver up to 36
ms boot time improvement as illustrated below.
| # | Late EOP | Async EOP |
|----------+----------+-----------|
| 1 | 1020.052 | 971.272 |
| 2 | 1015.911 | 971.821 |
| 3 | 1038.415 | 1021.841 |
| 4 | 1020.657 | 993.751 |
| 5 | 1065.128 | 1020.951 |
| 6 | 1037.859 | 1023.326 |
| 7 | 1042.010 | 984.412 |
|----------+----------+-----------|
| Mean | 1034.29 | 998.20 |
| Variance | 4.76 % | 5.21 % |
The improvement is not stable but comparing coreboot and FSP
performance timestamps demonstrate that the slowness is caused by a
lower memory frequency (SaGv point) at early boot which is not an
issue addressed by this patch.
We also observe some improvement on an Alder Lake design. For example,
the same configuration on a kano board can deliver up to 10 ms boot time
improvement as illustrated below.
| # | Late EOP | Async EOP |
|----------+----------+-----------|
| 0 | 1067.719 | 1050.106 |
| 1 | 1058.263 | 1056.836 |
| 2 | 1064.091 | 1056.709 |
| 3 | 1068.614 | 1055.042 |
| 4 | 1065.749 | 1056.732 |
| 5 | 1069.838 | 1057.846 |
| 6 | 1066.897 | 1053.548 |
| 7 | 1060.850 | 1051.911 |
|----------+----------+-----------|
| Mean | 1065.25 | 1054.84 |
The improvement is more limited on kano because a longer PCIe
initialization delays EOP in the Late EOP configuration which make it
faster to complete.
CSME team confirms that:
1. End-Of-Post is a blocking command in the sense that BIOS is
requested to wait for the command completion before loading the OS or
second stage bootloader.
2. The BIOS is not required to actively wait for completion of the
command and can perform other operations in the meantime as long as
they do not involve HECI commands.
On Raptor Lake, coreboot does not send any HECI command after
End-Of-Post. FSP-s code review did not reveal any HECI command being
sent as part of the `AFTER_PCI_ENUM', `READY_TO_BOOT' or
`END_OF_FIRMWARE' notifications.
If any HECI send and receive command has been sent the extra code
added in `cse_receive_eop()' should catch it.
According to commit 387ec919d9 ("soc/intel/alderlake: Select
SOC_INTEL_CSE_SEND_EOP_LATE"), FSP-silicon can sometimes (on the first
boot after flashing of a Marasov board for instance) request coreboot
to perform a global request out of AFTER_PCI_ENUM notification. Global
request relies on a HECI command. Even though, we tested that it does
not create any issue, `SOC_INTEL_CSE_SEND_EOP_ASYNC' flag should not
be associated to the `SOC_INTEL_CSE_SEND_EOP_EARLY' flag to prevent
potential a global reset command to "conflict" with the EOP command.
This patch also introduces a new code logic to detect if CSE is in the
right state to handle the EOP command. Otherwise, it uses the
prescribed method to make the CSE function disable. The typical
scenario is the ChromeOS recovery boot where CSE stays in RO partition
and therefore EOP command should be avoided.
[DEBUG] BS: BS_PAYLOAD_LOAD exit times (exec / console): 0 / 14 ms
[INFO ] HECI: coreboot in recovery mode; found CSE in expected
SOFT TEMP DISABLE state, skipping EOP
[INFO ] Disabling Heci using PMC IPC
[WARN ] HECI: CSE device 16.0 is hidden
[WARN ] HECI: CSE device 16.1 is disabled
[WARN ] HECI: CSE device 16.2 is disabled
[WARN ] HECI: CSE device 16.3 is disabled
[WARN ] HECI: CSE device 16.4 is disabled
[WARN ] HECI: CSE device 16.5 is disabled
BUG=b:276339544
BRANCH=firmware-brya-14505.B
TEST=Tests on brya0 with and `SOC_INTEL_CSE_SEND_EOP_ASYNC' show
End-Of-Post sent soon after FSP-s and EOP message receive at
`BS_PAYLOAD_BOOT'. Verify robustness by injecting a
`GET_BOOT_STATE' HECI command with or without `heci_reset'. The
implementation always successfully completed the EOP before
moving to the payload. As expected, the boot time benefit of the
asynchronous solution was under some injection scenario
undermined by this unexpected HECI command.
Signed-off-by: Jeremy Compostella <jeremy.compostella@intel.com>
Signed-off-by: Subrata Banik <subratabanik@google.com>
Change-Id: I01a56bfe3f6c37ffb5e51a527d9fe74785441c5a
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74214
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Nick Vaccaro <nvaccaro@google.com>
Reviewed-by: Tarun Tuli <taruntuli@google.com>
