This patch makes the SoC specific callback code more readable by adding
`soc_` prefix into the `get_dptf_platform_info()`.
In nutshell this patch renames `get_dptf_platform_info()` to
`soc_get_dptf_platform_info()`.
TEST=Able to build Google/Rex without any compilation issue.
Signed-off-by: Subrata Banik <subratabanik@google.com>
Change-Id: I27d6a146d5928e1742f82f85f51ad42656f46344
Reviewed-on: https://review.coreboot.org/c/coreboot/+/71096
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Sumeet R Pawnikar <sumeet.r.pawnikar@intel.com>
Adapted from WIP (and now abandoned) patches CB:25334, 26308, 26309.
Update the nhlt_soc_add_*() methods for max98373, max98927, and rt5514
codecs to program the render and feedback slot numbers as appropriate.
TEST=boot Windows on google/eve, atlas, nocturne, and rammus. Verify
audio functional with both Google project campfire drivers as well as
coolstar's AVS audio drivers.
Change-Id: Ib8c6e24ba539e205bd5bbd856ecff43b2c016c2e
Signed-off-by: Matt DeVillier <matt.devillier@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/70394
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Raul Rangel <rrangel@chromium.org>
Reviewed-by: CoolStar <coolstarorganization@gmail.com>
Adapted from WIP (and abandoned) patch CB:25334, this patch:
1. Ensures SSP endpoint InstanceId is 0
2. Adds capability_size parameter at the end of the nhlt
3. Adsd more config_type enum values to accommodate feedback stream
4. Programs virtual_slot values for max98373, max98927,
and rt5514 nhlt files
5. Adds NHLT feedback_config parameters
Default feedback configs are added here to the max98373, max98927, and
rt5514 codecs; in a follow-on patch, these will be overridden at the
board level.
TEST=tested with subsequent patch
Change-Id: I59285e332de09bb448b0d67ad56c72a208588d47
Signed-off-by: Matt DeVillier <matt.devillier@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/70393
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-by: Raul Rangel <rrangel@chromium.org>
Reviewed-by: CoolStar <coolstarorganization@gmail.com>
smm_relocation_handler is run for each thread but IA32_SMRR_PHYS_BASE
and IA32_SMRR_PHYS_MASK are core scope, need to avoid writing the
same MSR that has been locked by another thread.
Tested=On OCP Crater Lake, rdmsr -a 0x1f3 can see all cores set the lock
bit.
Change-Id: I9cf5a6761c9a9e1578c6132ef83e288540d41176
Signed-off-by: Johnny Lin <johnny_lin@wiwynn.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/70870
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-by: Jonathan Zhang <jonzhang@fb.com>
This patch fixes a hidden issue present inside FSP-S while coreboot
decides to skip performing MP initialization by overriding FSP-S UPDs
as below:
1. CpuMpPpi ------> Passing `NULL` as coreboot assume FSP don't need
to use coreboot wrapper for performing any
operation over APs.
2. SkipMpInit -----> Set `1` to let FSP know that coreboot decided
to skip FSP running CPU feature programming.
Unfortunately, the assumption of coreboot is not aligned with FSP when
it comes to the behaviour of `CpuMpPpi` UPD. FSP assumes ownership of
the APs (Application Processors) upon passing `NULL` pointer to the
`CpuMpPpi` FSP-S UPD.
FSP-S creates its own infrastructure code after seeing the CpuMpPpi
UPD is set to `NULL`. FSP requires the CpuMpPei module, file name
`UefiCpuPkg/CpuMpPei/CpuMpPei.c`, function name `InitializeCpuMpWorker`
to perform those additional initialization which is not relevant for
the coreboot upon selecting the SkipMpInit UPD to 1 (a.k.a avoid
running CPU feature programming on APs).
Additionally, FSP-S binary size has increased by ~30KB (irrespective of
being compressed) with the inclusion of the CpuMpPei module, which is
eventually not meaningful for coreboot.
Hence, this patch selects `MP_SERVICES_PPI_V2_NOOP` config
unconditionally to ensure pass a valid pointer to the `CpuMpPpi` UPD
and avoid APs getting hijacked by FSP while coreboot decides to set
SkipMpInit UPD.
Ideally, FSP should have avoided all AP related operations when
coreboot requested FSP to skip MP init by overriding required UPDs.
TEST=Able to drop CpuMpPei Module from FSP and boot to Chrome OS on
Google/Redrix, Kano, Taeko devices with SkipMpInit=1.
Without this patch:
Here is the CPU AP logs coming from the EDK2 (open-source)
[UefiCpuPkg/CpuMpPei/CpuMpPei.c] when coreboot sets `NULL` to the
CpuMpPpi UPD.
[SPEW ] Loading PEIM EDADEB9D-DDBA-48BD-9D22-C1C169C8C5C6
[SPEW ] Loading PEIM at 0x00076F9A000 EntryPoint=0x00076FA24E2
CpuMpPei.efi PROGRESS CODE: V03020002 I0
[SPEW ] Register PPI Notify: F894643D-C449-42D1-8EA8-85BDD8C65BDE
[SPEW ] Notify: PPI Guid: F894643D-C449-42D1-8EA8-85BDD8C65BDE,
Peim notify entry point: 76FA0239
AP Loop Mode is 2
GetMicrocodePatchInfoFromHob: Microcode patch cache HOB is not found.
CPU[0000]: Microcode revision = 00000000, expected = 00000000
[SPEW ] Register PPI Notify: 8F9D4825-797D-48FC-8471-845025792EF6
Does not find any stored CPU BIST information from PPI!
APICID - 0x00000000, BIST - 0x00000000
[SPEW ] Install PPI: 9E9F374B-8F16-4230-9824-5846EE766A97
[SPEW ] Install PPI: 5CB9CB3D-31A4-480C-9498-29D269BACFBA
[SPEW ] Install PPI: EE16160A-E8BE-47A6-820A-C6900DB0250A
PROGRESS CODE: V03020003 I0
With this patch:
No instance of `CpuMpPei` has been found in the AP UART log with FSP
debug enabled.
This patch is backported from
commit 8409f156d5 (soc/intel/alderlake:
Remove dependency of FSP-S CpuMpPei Module)
Change-Id: I7d9fb37ca1cd4bf325edc951ee7293e459fa2ea4
Signed-off-by: Subrata Banik <subratabanik@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/70600
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
Reviewed-by: Ivy Jian <ivy.jian@quanta.corp-partner.google.com>
The details about how the CPU multiprocessor init (MP) has migrated
from coreboot to FSP can be found in
https://doc.coreboot.org/soc/intel/mp_init/mp_init.html.
The major reason behind this migration is to support the Intel
proprietary and restricted CPU feature programming which can't be
performed if coreboot sets the BIOS_DONE or BIOS Reset CPL as part
of coreboot MP Init flow (prior to calling FSP-S). Hence, the new
flow introduced with Tiger Lake platform forced having monolithic
MP Init peformed by FSP (using coreboot MP PPI wrapper code).
The last 3-4 years of FSP doing MP Init has demonstrated ample
issues during platform bringup which is specific to UEFI MP Service
implementation and not relevant to open source coreboot. This new
flow makes the debug and validation aspect complicated where
any FSP MP Init code changes should have been validated with coreboot
MP PPI wrapper else might cause some failure, unfortunately,
the validation commitment has never been met, hence, issue debugging
is the only solution that remains in practice.
Most importantly, the restricted feature programming which demanded
closed source MP Init (for features like SGX and C6DRAM) has never
been enabled in coreboot (starting with Alder Lake, the SGX feature
has been dropped).
This patch attempts to decouple FSP-S doing MP Init from the rest
of the FSP-S silicon init and introduces 2nd MultiPhase SI init
which allows bootloader to perform the mandatory SoC programming
before FSP-S has done with PM programming (a.k.a set the reset CPL).
The core/uncore BWG suggests the minimum SoC programming before
BIOS Reset CPL is set. coreboot uses the MultiPhaseSI Init Index 2
to perform the required CPU programming before enabling the BIOS
Reset CPL.
This implementation would allow us to get rid of FSP running CPU
feature programming and additionally make several EDK2 MP service
modules optional (those are packed to create FSP-S blob).
In summary, this change would allow coreboot to utilize open source
MP init without running into FSP-S related code blocks.
Note: At present, Intel Meteor Lake FSP doesn't have support for
MultiPhase SI Init, Index 2 (submitted a FSP code changes over
chrome-internal to enable this feature to decouple MP Init from
FSP-S init).
This patch is backported from
commit b6c3a0325b (soc/intel/alderlake:
Implement MultiPhase SI Init Index 2 callback).
BUG=b:233199592
TEST=Build and boot google/taeko to ChromeOS.
Perform several thousands cycles of suspend test and power cycle
without running into any issue.
Change-Id: I2ea1a8bb2b142e39c2bc9d248b7fd0041366c0db
Signed-off-by: Subrata Banik <subratabanik@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/70558
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Christian Walter <christian.walter@9elements.com>
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
According to Intel doc#630774, BIOS should clear Host Interrupt Status
if it has read all the slots of the message from the ME circular buffer.
Since this is not found in client ME document, add a Kconfig
SOC_INTEL_CSE_SERVER_SKU that only clears interrupt status for Server
ME SKU.
On SPR-SP, if mainboard calls get_me_fw_version via HECI-1, with the
change can avoid seeing below Linux warning during boot with Linux
v5.12:
[ 17.868929] irq 16: nobody cared (try booting with the "irqpoll" option)
[ 17.883819] CPU: 10 PID: 0 Comm: swapper/10 Not tainted 5.12.0
[ 17.902412] Hardware name: Wiwynn Crater Lake EVT2/Crater Lake-Class1
[ 17.922327] Call Trace:
[ 17.927780] <IRQ>
[ 17.932253] dump_stack+0x64/0x7c
[ 17.939640] __report_bad_irq+0x37/0xb1
[ 17.948206] note_interrupt.cold.11+0xa/0x63
[ 17.957713] handle_irq_event_percpu+0x6a/0x80
[ 17.967626] handle_irq_event+0x2a/0x50
[ 17.976163] handle_fasteoi_irq+0x9e/0x140
[ 17.985305] __common_interrupt+0x38/0x90
[ 17.994255] common_interrupt+0x7a/0xa0
[ 18.002821] </IRQ>
[ 18.007514] asm_common_interrupt+0x1e/0x40
Change-Id: I1cf21112870e53a11134d43e461b735ead239717
Signed-off-by: Johnny Lin <johnny_lin@wiwynn.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/66579
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Tim Wawrzynczak <inforichland@gmail.com>
On nissa, sending EOP late improves boot time by about 57ms.
Before (SOC_INTEL_CSE_SEND_EOP_EARLY):
943:after sending EOP to ME 931,206 (58,431)
943:after sending EOP to ME 932,911 (58,427)
943:after sending EOP to ME 930,908 (58,429)
943:after sending EOP to ME 941,357 (61,748)
943:after sending EOP to ME 933,289 (62,050)
943:after sending EOP to ME 939,578 (62,453)
943:after sending EOP to ME 932,491 (62,050)
943:after sending EOP to ME 929,693 (62,655)
943:after sending EOP to ME 942,247 (62,654)
943:after sending EOP to ME 936,984 (61,751)
After (SOC_INTEL_CSE_SEND_EOP_LATE):
943:after sending EOP to ME 1,107,816 (3,498)
943:after sending EOP to ME 1,053,286 (25,212)
943:after sending EOP to ME 1,124,095 (3,511)
943:after sending EOP to ME 1,098,591 (3,498)
943:after sending EOP to ME 1,107,772 (3,499)
943:after sending EOP to ME 1,080,008 (45,969)
943:after sending EOP to ME 1,081,754 (8,024)
943:after sending EOP to ME 1,109,193 (4,102)
943:after sending EOP to ME 1,088,866 (4,201)
943:after sending EOP to ME 1,081,684 (4,203)
BUG=b:247902068
TEST=EOP time is improved on nissa (measurements above).
Change-Id: I2389831b4ab62f247193b5b0c5ec201e12eaa3db
Signed-off-by: Reka Norman <rekanorman@chromium.org>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/70849
Reviewed-by: Kangheui Won <khwon@chromium.org>
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 original value of the `DevIntConfigPtr` is unknown, and there's no
way to be absolutely certain that it actually points to usable memory.
Instead of copying the data, update the pointer's address to reference
the global variable directly. It is assumed that FSP does not write to
the memory pointed by `DevIntConfigPtr`. Confirming this assumption is
pointless; one might as well reimplement FSP instead.
Change-Id: I90594cc09e3fa2aef98658441c323a44a869635b
Signed-off-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/65217
Reviewed-by: Paul Menzel <paulepanter@mailbox.org>
Reviewed-by: Nico Huber <nico.h@gmx.de>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Matt DeVillier <matt.devillier@gmail.com>
Reviewed-by: Michael Niewöhner <foss@mniewoehner.de>
Clang generates R_X86_64_32S symbols that get truncated.
TESTED:
- prodrive/hermes boots with GCC and clang
- MTRR are properly cleared (tested by filling in both
MTRR_FIX_64K_00000 and MTRR_FIX_4K_F8000 before clearing)
Change-Id: I6a5139f7029b6f35b44377f105dded06f6d9cbf9
Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/69388
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Drops the `FIXME` comment and relevant code as this patch enables
VMX and VTD.
This patch also fixes the problem of additional reboot on every warm
boot due to overriding the CPU soft-strap.
TEST=No extra reboot seen while issuing warm reset from kernel
console.
without this patch:
950:calling FspMemoryInit 1,225,259 (20,537)
951:returning from FspMemoryInit 10,334,707 (9, 109,447)
with this patch:
950:calling FspMemoryInit 1,225,259 (20,537)
951:returning from FspMemoryInit 1,334,707 (109,447)
Change-Id: Ib130698e7255876c5a12abc93dd7d8a34dfae968
Signed-off-by: Subrata Banik <subratabanik@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/70553
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
This patch skips sending CONN IPC command to PMC if system is resuming
from S3.
Sending CONN IPC command as part of `tcss_configure_dp_mode()` function
results into ERROR while system is resuming from S3.
Additionally, skip `configure_aux_bias_pads()` during S3 resume.
BUG=b:260984500
TEST=Able to test on Google/Rex.
Without this patch:
[ERROR] pmc_send_ipc_cmd status: fatal
[ERROR] Port 1 connect request failed
[SPEW ] [TCSS] TcssInit() - End
With this patch:
No error seen during S3 resume.
Signed-off-by: Subrata Banik <subratabanik@google.com>
Change-Id: I1dab7dc8b4ad76ca0c9630456803c1b9a320fe40
Reviewed-on: https://review.coreboot.org/c/coreboot/+/70222
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
