All AMD SoCs with Zen-based CPU cores are already using timestamps based
on the TSC counter, so use the existing common infrastructure instead of
reimplementing it in a similar way.
The behavior of the code changes slightly, but results in identical
timestamps. The timestamp_get implementation in soc/amd/common/block/cpu
divided the result of rdtscll() in timestamp_get by the result of
tsc_freq_mhz() and didn't override the weak timestamp_tick_freq_mhz
implementation that returns 1. The non AMD specific code returns the
result of rdtscll() in timestamp_get, but returns tsc_freq_mhz() instead
of 1 in timestamp_tick_freq_mhz, so we still get the correct timestamps.
TEST=The raw timestamps printed on the serial console are now multiplied
by the expected factor of the TSC frequency in MHz.
TEST=Normalized timestamps printed on the serial console by the x86 code
don't change significantly on Mandolin when comparing before and after
this patch. A slight variation in the timestamps is expected. An example
would be:
Before: CPU_CLUSTER: 0 init finished in 630 msecs
After: CPU_CLUSTER: 0 init finished in 629 msecs
TEST=The calculations of the time spent in verstage on PSP before
entering the bootblock on Guybrush result in similar times when
multiplying the value before the patch with the TSC frequency in the
case with the patch applied. The raw values printed on the serial
console by the verstage on PSP use the 1us time base, but the timestamp
logs that end up in CBMEM will be fixed up to use the same time base as
the x86 part of coreboot.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I57b732e5c78222d278d3328b26bb8decb8f4783e
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74016
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martin.roth@amd.corp-partner.google.com>
Reviewed-by: Paul Menzel <paulepanter@mailbox.org>
Having L1.2 enabled on the SD port increases the kernel resume times by
between 30 & 40ms. This patch disables L1.2 on SD to get that time
back.
As with needing to have hotplug enabled on the SD card, this seems like
a driver issue, so hopefully that will get sorted out and this patch
can be reverted.
BUG=b:274025743
TEST=resume times are decreased.
BRANCH=skyrim
Signed-off-by: Martin Roth <gaumless@gmail.com>
Change-Id: I2c409fa2cd66c712c5ba7104635499d63fa0d2be
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74118
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Reviewed-by: Karthik Ramasubramanian <kramasub@google.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'.
Change-Id: I81e9dc66f952c14cb14f513955d3fe853396b21c
Signed-off-by: Jeremy Compostella <jeremy.compostella@intel.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/73922
Reviewed-by: Nick Vaccaro <nvaccaro@google.com>
Reviewed-by: Tarun Tuli <taruntuli@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
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.
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.
Change-Id: Ib09dcf9140eb8a00807a09e2af711021df4b416f
Signed-off-by: Jeremy Compostella <jeremy.compostella@intel.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/73619
Reviewed-by: Tarun Tuli <taruntuli@google.com>
Reviewed-by: Subrata Banik <subratabanik@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Nick Vaccaro <nvaccaro@google.com>
In order for the code to find the correct VBIOS file in CBFS, remap the
revision ID in the RAVEN2_VBIOS_VID_DID case to the one that matches the
CBFS file name. This will make the code work as expected on devices with
the PCI ID RAVEN2_VBIOS_VID_DID and a revision != RAVEN2_VBIOS_REV.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I94412dc2e778e7c4f74e475cd49114a00a81b2ce
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74045
Reviewed-by: Matt DeVillier <matt.devillier@amd.corp-partner.google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
This is a new port for the Intel DQ67SW desktop board. It is
microATX-sized with an LGA1155 socket and four DIMM sockets for DDR3
SDRAM.
A list of tested working and non-working features is in the
documentation page.
Change-Id: Ifc703f2d0ad45495e71d3f7799347430f5196791
Signed-off-by: Michael Büchler <michael.buechler@posteo.net>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/73087
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
The Realtek RTL8111E NIC is currently not defined as a child device,
resulting in the on_board flag not being set to 1. This means that
Linux / udev will call the device enp4s0 rather than eno0, as is
appropriate for on-board ethernet devices.
This patch defines the NIC as a child device of PCIe port 6, so that
it's properly defined as an on-board device.
Change-Id: I2e1b65e4d27852297a739e332c52c15a8c81b858
Signed-off-by: Kevin Keijzer <kevin@quietlife.nl>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74090
Reviewed-by: Felix Singer <felixsinger@posteo.net>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
To avoid having to calculate the length of a struct separately, rework
the code to give the struct a tag name, so that `sizeof()` can be used
instead. This involves refactoring the `get_emi_eeprom_vpd()` function
to return a struct instead of a union, so callers can no longer access
the EEPROM data as an array of bytes without additional code, but this
array view is only used inside `get_emi_eeprom_vpd()` when reading the
data from EMI.
Change-Id: Id1bc40939631baa131b5f60eadbfe42838294ebe
Signed-off-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/73983
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Maximilian Brune <maximilian.brune@9elements.com>
Reviewed-by: Lean Sheng Tan <sheng.tan@9elements.com>
This patch avoids the redundant programming of SRAM BAR when
the SRAM PCI device is enabled. Rather read the PCH SRAM Base
Address Register while enabling crashlog feature.
Additionally, this patch relies on PCI enumeration to get the
SRAM BAR rather than hijacking the SPI temporary base address
which might have resulted in problems if SPI is disabled on
some platform with BAR being implemented.
TEST=Able to build and boot google/marasov and crashlog is working.
Change-Id: I8eb256aa63bbf7222f67cd16a160e71cfb89875a
Signed-off-by: Subrata Banik <subratabanik@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74056
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Tarun Tuli <taruntuli@google.com>
Reviewed-by: Pratikkumar V Prajapati <pratikkumar.v.prajapati@intel.com>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
Disable devices in variant.c instead of adding probe statements to
devicetree because storage devices need to be enabled when fw_config is
unprovisioned, and devicetree does not currently support this (it
disables all probed devices when fw_config is unprovisioned).
BUG=b:273791621
TEST=emerge-nissa coreboot
Change-Id: I1a6013e0ad0c430d83bbbad4b92392c8c4815b0d
Signed-off-by: Tony Huang <tony-huang@quanta.corp-partner.google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74059
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
This sets the location of the skyrim MP2 firmware within the mainboard's
blobs directory, and adds the Kconfig option to the mainboard directory
so that it can be enabled in a saved .config file.
The skyrim MP2 firmware is skyrim specific, so it should not be placed
in the main PSP AMD_BLOBS directory.
We will also only want to enable the MP2 firmware for chromeos builds as
it's not useful for non-chromeos builds.
BUG=b:259554520
TEST=Build MP2 firmware into image, see that it gets loaded
BRANCH=skyrim
Signed-off-by: Martin Roth <gaumless@gmail.com>
Change-Id: I04be6f2d0b605d4eca37fd927a70310259dc106c
Reviewed-on: https://review.coreboot.org/c/coreboot/+/73659
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Robert Zieba <robertzieba@google.com>
Reviewed-by: Jon Murphy <jpmurphy@google.com>
Instead of using the PSTATE SSDT generated by binaryPI, use the common
AMD code by selecting SOC_AMD_COMMON_BLOCK_ACPI_CPU_POWER_STATE. To
match the SSDT from binaryPI, set ACPI_SSDT_PSD_INDEPENDENT to n. There
are two differences to the binaryPI SSDT: Now coreboot includes the C1
state in the _CST package instead of just having the kernel add this due
to the ACPI_FADT_C1_SUPPORTED bit being set and the address of the
PS_STS_REG P state status MSR is written to the corresponding field of
the _PCT package instead of being 0.
TEST=On Careena the new P and C state ACPI packages are nearly identical
to the ones from the SSDT from binaryPI with the two functional
differences mentioned above.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: Icdf6bc8f0e0363f185a294ab84edcb51322e7eb7
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74023
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martin.roth@amd.corp-partner.google.com>
The C state ACPI packages binaryPI generates and passes to coreboot in
the PSTATE SSDT only include the C2 state, but the kernel will add the
C1 state to its usable C states in this case. The native C state code
will generate both the C1 and C2 state packages to be more complete and
also to be more in line with the other AMD SoCs.
The code added in this commit isn't used yet, but will be used as soon
as Stoneyridge will be using the common AMD generate_cpu_entries by
selecting SOC_AMD_COMMON_BLOCK_ACPI_CPU_POWER_STATE once all needed
helper functions are implemented for Stoneyridge.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I06f90306ac196704e0102d0da6eab03f51513c29
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74001
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martin.roth@amd.corp-partner.google.com>
Use get_pstate_core_freq instead of open-coding the calculations in
tsc_freq_mhz. In the case of the CPU frequency divider being 0,
get_pstate_core_freq will return 0; in this case that shouldn't happen,
TSC_DEFAULT_FREQ_MHZ will be used as frequency, since for the TSC
frequency it's better to err on the end of the expected frequency being
too high which will cause longer than expected delays instead of too
short delays.
Now that the code is using get_pstate_core_freq, this code is valid for
Glinda too, so also remove the comment on the
SOC_AMD_COMMON_BLOCK_TSC_FAM17H_19H option being selected in the Glinda
Kconfig. This Kconfig option will be renamed in a follow-up patch.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I01168834d4018c92f44782eda0c65b1aa392030d
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74013
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Reviewed-by: Martin Roth <martin.roth@amd.corp-partner.google.com>
Factor out the get_pstate_core_freq function from the SoC's acpi.c files
to both avoid duplication and to also be able to use the same function
in the TSC frequency calculation in a follow-up patch. The family 17h
and 19h SoCs use the same frequency encoding in the P state MSRs while
the family 1Ah SoCs use a different encoding. The family 15h and 16h
SoCs use another encoding, but since this isn't implemented in
Stoneyridge's acpi.c, this will be added in a follow-up patch.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I8619822c2c61e06ae5db86896d5323c9b105b25b
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74010
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
Reviewed-by: Martin Roth <martin.roth@amd.corp-partner.google.com>
Due to a non-constant TSC rate before the microcode update is applied,
the Performance Time Stamp Counter is used instead. To clarify this, add
a comment to the timestamp_get implementation. See commit 24079323d4
("soc/amd/stoneyridge: provide alternate monotonic timer") and the
description of the TscInvariant bit in CPUID Fn8000_0007_EDX Advanced
Power Management Information in the public version of BKDG #55072 Rev
3.09 for more details.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I824b372c36fa6f3eb912469b235a9474f6a58ff5
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74018
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
