This patch refactors common code to allow cse_final() function to send
EOP cmd if the SoC user selects `SOC_INTEL_CSE_SET_EOP` kconfig.
This patch helps cse_final_ready_to_boot() and
cse_final_end_of_firmware() function for being meaningful with its
operation and let cse_final() being that outer layer to perform three
operations based on the selected kconfig.
1. send cse eop command
2. perform cse_final_ready_to_boot() operations
3. perform cse_final_end_of_firmware() operations
Additionally, ensures the platform that choose to send EOP late
(like JSL and TGL) is not being impacted due to this code refactoring
hence, skip calling into CSE.final if SoC selects
`SOC_INTEL_CSE_SEND_EOP_LATE` config.
BUG=b:260041679
TEST=Able to send EOP command successfully for Google/Taeko.
Signed-off-by: Subrata Banik <subratabanik@google.com>
Change-Id: I412291c9378011509d3825f9b01e81bfced53303
Reviewed-on: https://review.coreboot.org/c/coreboot/+/69975
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
Reviewed-by: Lean Sheng Tan <sheng.tan@9elements.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Presently, coreboot supports two instances of sending EOP cmd to
the Intel CSE.
1. Sending EOP cmd to CSE during `.final` operation from cse pci driver.
2. Starting with Alder Lake, the recommendation was to send EOP to CSE
earlier than CSE `.final` operation. Since then it's referred to as
`Sending EOP Early`. This method helped to save the CSE EOP
response time significantly.
During Meteor Lake platform, CSE EOP response time has become
non-deterministic and we have figured that sending EOP command later
than CSE .final operation is actually helping to optimize the boot time
significantly (around ~150ms savings compared to sending from `.final`
ops and ~5sec compared to sending CSE early).
Hence, this patch intended to create yet another kconfig for sending
CSE late (specifically after `.final` operation). The idea for this
newer config is to use the boot state machine for sending CSE EOP cmd.
The patch train in this series would add the specific changes to allow
sending EOP late and perform other essential operations required prior
booting to OS as coreboot decided to skip calling into FSP Notify phase.
Starting with Jasper Lake, coreboot sends EOP before loading payload
hence, this config is applicable for those platforms.
The current plan is that Intel Jasper Lake, Tiger Lake and Meteor Lake
platform will select this newer config from SoC code.
BUG=b:260041679
TEST=Able to send EOP command successfully for Google/Taeko.
Signed-off-by: Subrata Banik <subratabanik@google.com>
Change-Id: Iea512cd5b79d61dd5d5a962079baf525027c831f
Reviewed-on: https://review.coreboot.org/c/coreboot/+/69976
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
Reviewed-by: Lean Sheng Tan <sheng.tan@9elements.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
This patch ensures dropping of the duplicate macro introduced with
'commit 9e4488ab06 ("soc/intel/{adl,cmn}: Add/Remove LTR
disqualification for UFS")'
`PCH_PWRM_BASE_SIZE` macro represents the size of the PMC MMIO range
which can be used as is even in ufs.asl file.
BUG=b:252975357
TEST=Build and boot nirwen and see no issues in PLT runs.
Signed-off-by: Subrata Banik <subratabanik@google.com>
Change-Id: Ic967c609e1330eca1b9e1143e7efd78db011f317
Reviewed-on: https://review.coreboot.org/c/coreboot/+/70180
Reviewed-by: Reka Norman <rekanorman@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Nick Vaccaro <nvaccaro@google.com>
The patch logs CSE RO's write protection information for Meteor Lake
platform. As part of write protection information, coreboot logs status
on CSE RO write protection and range. Also, logs error message if EOM
is disabled, and write protection for CSE RO is not enabled.
Port of commit abe0d810f0 ("soc/intel/alderlake: Log CSE RO write
protection info for ADL").
BUG=none
TEST=Verify the write protection details on google/rex.
Excerpt from google/rex coreboot log:
[DEBUG] ME: WP for RO is enabled : YES
[DEBUG] ME: RO write protection scope - Start=0x4000, End=0x396FFF
Signed-off-by: Kapil Porwal <kapilporwal@google.com>
Change-Id: Idb072a873a8b8323532799f5fc64f995c9f0a604
Reviewed-on: https://review.coreboot.org/c/coreboot/+/69571
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Subrata Banik <subratabanik@google.com>
Reviewed-by: Sridhar Siricilla <sridhar.siricilla@intel.corp-partner.google.com>
Reviewed-by: Sridhar Siricilla <sridhar.siricilla@intel.com>
This reverts commit 095c931cf1.
Previously cpu_info() was implemented with a struct on top of an
aligned stack. As FSP changed the stack value cpu_info() could not be
used in FSP context (which PPI is). Now cpu_info() uses GDT segments,
which FSP does not touch so it can be used.
This also exports cpu_infos from cpu.c as it's a convenient way to get
the struct device * for a certain index.
TESTED on aldrvp: FSP-S works and is able to run code on APs.
Change-Id: I3a40156ba275b572d7d1913d8c17c24b4c8f6d78
Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/69509
Reviewed-by: Subrata Banik <subratabanik@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Name a variable based on its utility. `is_external` variable adds
`ExternalFacingPort` _DSD property to an ACPI device hence
rename it to `add_acpi_external_facing_port`.
BUG=b:259716145
TEST=Build google/rex with this flag and verify it in SSDT at
runtime.
SSDT snippet:
Name (_DSD, Package (0x04) // _DSD: Device-Specific Data
{
ToUUID ("6211e2c0-58a3-4af3-90e1-927a4e0c55a4"),
Package (0x01)
{
Package (0x02)
{
"HotPlugSupportInD3",
One
}
},
ToUUID ("efcc06cc-73ac-4bc3-bff0-76143807c389"),
Package (0x01)
{
Package (0x02)
{
"ExternalFacingPort",
One
}
}
})
Signed-off-by: Kapil Porwal <kapilporwal@google.com>
Change-Id: I65100283ed9b65037c9890f28ecab41fcfa25d83
Reviewed-on: https://review.coreboot.org/c/coreboot/+/69970
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>
Later soc/intel/common/smbus addresses TCO2_STS as a separate
16-bit register, while baytrail and braswell assumes 32-bit
wide TCO1_STS to extend as TCO2_STS.
In src/soc/intel/denverton_ns:
#define TCO2_STS_SECOND_TO 0x02
In soc/intel/baytrail,braswell:
#define SECOND_TO_STS (1 << 17)
Elsewehere
#define SECOND_TO_STS (1 << 1)
It's expected that we remove the first (1 << 17) case and only
access TCO2_STS as a separate 16-bit register. For now, use
unique names to avoid confusion.
Change-Id: I07cc46a9d600b2bf2f23588b26891268e9ce4de0
Signed-off-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/70044
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Elyes Haouas <ehaouas@noos.fr>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-by: Subrata Banik <subratabanik@google.com>
Tree is inconsistent with the use of TCO register space offsets and
related preprocessor defines. The legacy space was offset from ACPI
PM base by 0x60, but this changed with later platforms. The convenient
way is to define the TCO registers relative to its base address and
subtract 0x60 here, but this change cannot be easily done tree-wide or
in one go.
For the transient period, apply TCO_SPACE_NOT_YET_SPLIT flag until
all platforms use a clean style of tco_{read,write} accessor functions
instead of {read,write}_pmbase16(), or worse, inw/outl().
Change-Id: I16213cdb13f98fccb261004b31e81a9a44cb6e3b
Signed-off-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/70043
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Elyes Haouas <ehaouas@noos.fr>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
This patch creates an API for CSE-Lite specific SKU to retrieve the
Write Protect (WP) information (`cse_log_ro_write_protection_info`)
like WP range and limit, if the region is write-protected or not etc.
BUG=none
TEST=Able to compile the cse_lite.c file for google/kano without
any error.
Signed-off-by: Subrata Banik <subratabanik@google.com>
Change-Id: I8f4b7880534ded5401b6f8d601ded88019c636c8
Reviewed-on: https://review.coreboot.org/c/coreboot/+/69968
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
A previous CL ("Add missing ACPI device path names",
commit d22500f0c61f8c8e10d8f4a24e3e2bf031163c07) caused some errors
from the Kernel on Brya devices (see Tim's comment on patchset 8):
> ACPI Error: AE_NOT_FOUND, While resolving a named reference
> package element - \_SB_.PCI0.FSPI
FSPI is defined in src/soc/intel/alderlake/chipset.cb:
device pci 1f.5 alias fast_spi on end
This CL adds the corresponding FSPI device to the DSDT to prevent
the error mentioned above.
TEST=Built and tested on brya by verifying the error is gone.
BUG=b:231582182
Change-Id: I11e89ad2a5d47f6b579f755b0a41399ee3cb856c
Reviewed-on: https://review.coreboot.org/c/coreboot/+/69920
Reviewed-by: Tarun Tuli <taruntuli@google.com>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
This patch refactors the MDIO access for the TSN GbE device by placing
the MDIO read and write functions into mdio_bus_operations struct which
is assigned to the .ops_mdio member of the PCI device struct. In this
way the MDIO interface of the TSN GbE device is exposed and can be used
by other drivers if needed.
Change-Id: I5d1b9dd2f2ba8c18291fff314c13f0c3851784aa
Signed-off-by: Mario Scheithauer <mario.scheithauer@siemens.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/69383
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Werner Zeh <werner.zeh@siemens.com>
This patch skips setting D0I3 bit for all HECI devices by FSP.
The learning being made from Alder Lake platform showed that the CSE
EOP cmd response time is highly nondeterministic and letting the EOP
cmd issued by FSP makes the response time even worse.
The idea being pursued during Alder Lake platform is to let FSP skip sending the EOP cmd and coreboot sends it at the last minute
(late sending of EOP) to ensure there is ample time for CSE to come
to a state where the response to the EOP is almost immediate.
There were a number of refactoring being done to ensure the EOP cmd
can be sent at the later stage.
#1: Ensure FSP is not putting those HECI devices into the D0i3. (SoC specific change)
#2: Modify the CSE related boot state based operation to allow a
proper window for sending late EOP cmd. (Common Code Specific change)
The entire refactoring helps us to save ~60ms of boot time.
Without those code change EOP sending timestamp as below:
943:after sending EOP to ME 1,248,328(61,954))
With those code change EOP sending timestamp as below:
943:after sending EOP to ME 1,231,660 (2,754)
Port of commit d6da4ef69e ("soc/intel/alderlake: Skip setting D0I3
bit for HECI devices") to incorporate the #1 which is a SoC specific
code change.
BUG=none
TEST=FSP-S UPD dump suggested `DisableD0I3SettingForHeci` UPD is
set to `1`.
Excerpt from google/rex coreboot log:
[SPEW ] DisableD0I3SettingForHeci : 0x1
Signed-off-by: Kapil Porwal <kapilporwal@google.com>
Change-Id: I1c3765ce41f192ab5f5ff176e0a2b49b312d18d2
Reviewed-on: https://review.coreboot.org/c/coreboot/+/69680
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Subrata Banik <subratabanik@google.com>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
This patch helps to save 10.200ms of booting time without any issue
seen during MP Init. All cores are out from reset and alive.
Port the Alder Lake 'commit 6526e78967 ("soc/intel/alderlake: Select X86_INIT_NEED_1_SIPI Kconfig for RPL")' also to Meteor Lake.
Additionally, no performance degradation is observed while running
benchmarks.
BUG=b:211770003
TEST=Able to boot Google, Rex to ChromeOS with all cores enabled.
Without this patch:
30:device enumeration 1,480,217 (28,232)
With this patch:
30:device enumeration 1,472,466 (18,334)
Signed-off-by: Subrata Banik <subratabanik@google.com>
Change-Id: Iec21470b9b34514169789c39bdc3be4e4ff6c7b5
Reviewed-on: https://review.coreboot.org/c/coreboot/+/69851
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
Reviewed-by: Tarun Tuli <taruntuli@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
For the most part, this just moves the existing post codes into macros
so that they're not just bare numbers.
cache_as_ram.S:
Post code 0x28 was previously pointless with just a single jump between
it and post code 0x29, car_init_done. This code was removed, and the
0x28 value was used to differentiate the car_nem_enhanced subroutine
from the other 0x26 post codes used before calling the clear_car
subroutine.
All other post codes remain identical.
POST_BOOTBLOCK and POST_CODE_ZERO are expected to become global, whereas
the POST_SOC codes are expected to be Intel only.
Signed-off-by: Martin Roth <gaumless@gmail.com>
Change-Id: I82a34960ae73fc263359e4519234ee78e7e3daab
Reviewed-on: https://review.coreboot.org/c/coreboot/+/69865
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 parses CPU and PCH Trace Hub modes from the debug area in the
Descriptor Region. The modes can be updated in the debug area in order
to configure the CPU and PCH Trace Hub modes. The debug area's offset
starts from the SPI Flash offset:0xf00.
For runtime debugging, the OEM Section in the Descriptor Region is being
used as debug area. The OEM Section details are documented in the SPI
Programmer Guide of CSE Lite kit.
TEST=Build code for Gimble
Signed-off-by: Sridhar Siricilla <sridhar.siricilla@intel.com>
Change-Id: I61241c5c1981ddc4b21581bb3ed9f531da5f41b2
Reviewed-on: https://review.coreboot.org/c/coreboot/+/64437
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Tim Wawrzynczak <inforichland@gmail.com>
After DisplayPort is plugged into type-C port, its hpd signal
instantly presents and EC has mux_info for dp and hpd. This change
fixes the DP detection flow to avoid the 1 second delay while no DP
is connected. If DP is present, there will be requests towards PMC
through the sequence of connect, safe mode, dp and hpd mode.
BUG=b:247670186
TEST=Built image and validated the DisplayPort preboot feature on Rex.
Change-Id: I7cb95ec7fcc7e1a86e86466e6d45390eedcc4531
Signed-off-by: zhaojohn <john.zhao@intel.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/69694
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Set `EnableTcoTimer=1` in order to keep FSP from
1) enabling ACPI Timer emulation in uCode.
2) disabling the PM ACPI Timer.
Both actions are now done in coreboot.
`EnableTcoTimer=1` makes FSP skip these steps in any possible case
including `SkipMpInit=0`, `SkipMpInit=1`, use of the MP PPI or FSP
Multiphase Init. This way full control is left to coreboot.
Port of commit 0e905801f8 ("soc/intel: transition full control over PM
Timer from FSP to coreboot").
NOTE: This will have a huge power impact when it's enabled. If TCO timer
is disabled, uCode ACPI timer emulation must be enabled, and WDAT table
must not be exposed to the OS.
BUG=none
TEST=Boot to OS on google/rex.
Excerpt from google/rex coreboot log:
[SPEW ] EnableTcoTimer = 1
Signed-off-by: Kapil Porwal <kapilporwal@google.com>
Change-Id: I2693f0390e6c9fa92fec366ab87589c3bcea9027
Reviewed-on: https://review.coreboot.org/c/coreboot/+/69613
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Subrata Banik <subratabanik@google.com>
The dimm_info structure (defined in src/include/memory_info.h)
currently does not hold information about the DIMM's
node/controller ID.
This patch extends the dimm_info structure by adding a new field for
the node ID, called node_num. Also, adapt the dimm_info_fill()
function accordingly to populate the newly-added field.
Background: These changes are necessary for the Atlas mainboard, where
we are currently experiencing issues with the DIMMs device/bank
locator. Our 2 DIMMs share the same CHANNEL and DIMM ID but have a
distinct NODE ID. By looking at the smbios table we see
Channel-0-DIMM-0 for both DIMMs. Thus, we need their NODE IDs in order
to distinguish them.
This patch was tested by building and booting for the Alderlake-P
RVP board, which has the same DIMM slot configuration as the
Prodrive Atlas mainboard.
Signed-off-by: David Milosevic <David.Milosevic@9elements.com>
Change-Id: I6ffa5bdff0ba0e3c4a4a51f2419291fd1278cd68
Reviewed-on: https://review.coreboot.org/c/coreboot/+/68525
Reviewed-by: Jonathan Zhang <jonzhang@fb.com>
Reviewed-by: Lean Sheng Tan <sheng.tan@9elements.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
As per Intel doc #627331 Section 3.6.1 "Intel CSME Production Machine
Determination", from ADL onwards there are three criteria which
determine whether a device is in production mode:
1. Fuses are programmed
2. SPI descriptor is locked
3. Manufacturing variables are locked
When logging whether the device is in manufacturing mode, 1 and 2 are
already checked. Add a check for 3 as well.
Also add logs for each individual criteria so it's easy to tell why the
overall Manufacturing Mode is set or not.
BUG=b:255462682
TEST=On a nivviks which has not gone through EOM:
Before:
[DEBUG] ME: Manufacturing Mode : YES
[DEBUG] ME: SPI Protection Mode Enabled : NO
After:
[DEBUG] ME: Manufacturing Mode : YES
[DEBUG] ME: SPI Protection Mode Enabled : NO
[DEBUG] ME: FPFs Committed : NO
[DEBUG] ME: Manufacturing Vars Locked : NO
On an anahera which has gone through EOM:
Before:
[DEBUG] ME: Manufacturing Mode : NO
[DEBUG] ME: SPI Protection Mode Enabled : YES
After:
[DEBUG] ME: Manufacturing Mode : NO
[DEBUG] ME: SPI Protection Mode Enabled : YES
[DEBUG] ME: FPFs Committed : YES
[DEBUG] ME: Manufacturing Vars Locked : YES
Change-Id: Iac605baa291ab5cc5f28464006f4828c12c748fe
Signed-off-by: Reka Norman <rekanorman@chromium.org>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/69324
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
Reviewed-by: Kangheui Won <khwon@chromium.org>
Reviewed-by: Subrata Banik <subratabanik@google.com>
Reviewed-by: Michał Żygowski <michal.zygowski@3mdeb.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
As per Intel doc #729124 Section 3.6.1 "Intel CSME Production Machine
Determination", from ADL onwards there are three criteria which
determine whether a device is in production mode:
1. Fuses are programmed
2. SPI descriptor is locked
3. Manufacturing variables are locked
When logging whether the device is in manufacturing mode, #1 and #2 are
already checked. Add a check for #3 as well.
TEST=Build and boot MTL RVP
Snippet from coreboot log:
[DEBUG] ME: Manufacturing Mode : YES
Signed-off-by: Sridhar Siricilla <sridhar.siricilla@intel.com>
Change-Id: I495a7d8730716fc92e8c57b2caef73e8bb44d30b
Reviewed-on: https://review.coreboot.org/c/coreboot/+/69578
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Subrata Banik <subratabanik@google.com>
Reviewed-by: Jamie Ryu <jamie.m.ryu@intel.com>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Reviewed-by: Ivy Jian <ivy.jian@quanta.corp-partner.google.com>
Update header files for FSP for Meteor Lake platform to
version 2404_00, previous version being 2364_00.
FSPM:
1. Address offset changes
2. Rename `PlatformDebugConsent` to `PlatformDebugOption`
FSPS:
1. Address offset changes
Additionally, incorporate the UPD name change for MTL romstage.
BUG=b:255481471
TEST=Able to build and boot Google, Rex to ChromeOS.
Signed-off-by: vjadeja-intel <vikrant.l.jadeja@intel.com>
Signed-off-by: Subrata Banik <subratabanik@google.com>
Change-Id: I63ef4ecb6569141542a3b9bf4ee8cbcd2946582e
Reviewed-on: https://review.coreboot.org/c/coreboot/+/69182
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
Reviewed-by: Tarun Tuli <taruntuli@google.com>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
This patch uses index 0x10 for IOE P2SB memory resource allocation
instead of static 0.
Additionally, switches to `mmio_resource` from `mmio_resource_kb`.
TEST=Able to build and boot Google/Rex and observed log as below.
Without the code change:
[SPEW ] PCI: 00:13.0 resource base 3fff0aa0000 size 1400 align 0
gran 0 limit 0 flags f0000200 index 0
With the code change:
[SPEW ] PCI: 00:13.0 resource base 3fff0aa0000 size 1400 align 0
gran 0 limit 0 flags f0000200 index 10
Signed-off-by: Subrata Banik <subratabanik@google.com>
Change-Id: I44caac73e245f536f3a22baafa1a6a0370e1dd37
Reviewed-on: https://review.coreboot.org/c/coreboot/+/69041
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>
Reviewed-by: Ivy Jian <ivy.jian@quanta.corp-partner.google.com>
