Rather than explicitly checking for Recovery or Developer mode via
vboot, use display_init_required() so that vboot is not required, and
other instances where the display is needed pre-OS (such as when
applying a critical system update) are covered as well.
With this change, SoCs implementing selective GOP init will need to
select VBOOT_MUST_REQUEST_DISPLAY in order for display_init_required()
to not assert on compilation.
BUG=b:255812886
TEST=build/boot skyrim
Change-Id: Iac7e06863764a9f21c8a50fc19050cb5a6627df2
Signed-off-by: Matt DeVillier <matt.devillier@amd.corp-partner.google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/73046
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martin.roth@amd.corp-partner.google.com>
Rather than explicitly checking for Recovery or Developer mode via
vboot, use display_init_required() so that vboot is not required, and
other instances where the display is needed pre-OS (such as when
applying a critical system update) are covered as well.
Select VBOOT_MUST_REQUEST_DISPLAY in order for display_init_required()
to function properly (and not assert on compilation).
BUG=b:255812886
TEST=build/boot skyrim
Change-Id: If2fee71bcc11468fd2db0abaafe4ea35e2953993
Signed-off-by: Matt DeVillier <matt.devillier@amd.corp-partner.google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/73047
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martin.roth@amd.corp-partner.google.com>
The coreboot-common acpi_create_fadt writes a pointer to the FACS table
into both firmware_ctrl and x_firmware_ctl_l FADT fields and sets
x_firmware_ctl_h to zero. When x_firmware_ctl_[l,h] is non-zero, the
pointer in firmware_ctrl will be ignored, but that's what is already
done on Cezanne and newer.
TEST=Linux doesn't complain about any new ACPI problem on Mandolin.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: Ib9eab4dcf828f28a60c6312ec96872aac4cfb266
Reviewed-on: https://review.coreboot.org/c/coreboot/+/73190
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Elyes Haouas <ehaouas@noos.fr>
Reviewed-by: Fred Reitberger <reitbergerfred@gmail.com>
The FADT data structure is zero-initialized in acpi_create_fadt which
then calls the SoC-specific acpi_fill_fadt function, therefore it's not
needed to assign 0 to the ARM_boot_arch FADT field in acpi_fill_fadt.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: Ica968db1228a2d63e83f2b6c4ea57c5f02bf1504
Reviewed-on: https://review.coreboot.org/c/coreboot/+/73187
Reviewed-by: Elyes Haouas <ehaouas@noos.fr>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Fred Reitberger <reitbergerfred@gmail.com>
The EFS entry at offset 0x14 can point to either the first level PSP
directory table or to the PSP combo directory structure that was used
before the introduction of the AMD A/B recovery scheme. This scheme is
not to be confused with the VBOOT scheme. The PSP verstage code checks
if the header this entry points to begins with the PSP_COOKIE, which
indicates the entry is a first level PSP directory table. Due to that,
the EFS entry at offset 0x14 is always expected to point to a PSP
directory table, so rename combo_psp_directory to new_psp_directory to
match the actual usage. This EFS entry that points to the PSP directory
table is called new_psp_directory, since the entry at EFS offset 0x10
was used on some early AMD chips to point to the older PSP directory
table and that one is already called psp_directory. amdfwtool uses the
same naming scheme for those two PSP directory table pointers.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I10f19ee63f8d422433dba64402d84fd6bb9e0f9e
Reviewed-on: https://review.coreboot.org/c/coreboot/+/73083
Reviewed-by: Fred Reitberger <reitbergerfred@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Add the correct defaults for VGA_BIOS_ID and VGA_BIOS_FILE in
Mendocino's Kconfig instead of relying on the board's .config files
providing the correct settings. Those settings are per-SoC and not
per-board, so this is valid for all boards using the Mendocino APU.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I09c537d3801123e7ffc01608171918b0396b7a5f
Reviewed-on: https://review.coreboot.org/c/coreboot/+/73051
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Fred Reitberger <reitbergerfred@gmail.com>
Add the correct defaults for VGA_BIOS_ID and VGA_BIOS_FILE in Cezanne's
Kconfig instead of relying on the board's .config files providing the
correct settings. Those settings are per-SoC and not per-board, so this
is valid for all boards using the Cezanne APU.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I67101d518c6b873ad89932ae39c2deb2ed6a4c29
Reviewed-on: https://review.coreboot.org/c/coreboot/+/73050
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Fred Reitberger <reitbergerfred@gmail.com>
The way the PSP_APCB_FILES list is created will always insert at least a
space into it. When tested by the if, this space will prevent the else
clause from ever running and never generate a build error.
Remove the non-functional check. Instead, mainboards should select
warn_no_apcb or die_no_apcb to generate a warning message or build error
if the APCB is missing.
Signed-off-by: Fred Reitberger <reitbergerfred@gmail.com>
Change-Id: Ic96846d74df2dc279e13b22f2a83b6f893954fe8
Reviewed-on: https://review.coreboot.org/c/coreboot/+/73009
Reviewed-by: Martin Roth <martin.roth@amd.corp-partner.google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
The way the PSP_APCB_FILES list is created will always insert at least a
space into it. When tested by the if, this space will prevent the else
clause from ever running and never generate a build error.
Remove the non-functional check. Instead, mainboards should select
warn_no_apcb or die_no_apcb to generate a warning message or build error
if the APCB is missing.
Signed-off-by: Fred Reitberger <reitbergerfred@gmail.com>
Change-Id: I26b96966495dc35a8b4a0cb7d5a841f3812f2a70
Reviewed-on: https://review.coreboot.org/c/coreboot/+/73007
Reviewed-by: Martin L Roth <gaumless@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
The way the PSP_APCB_FILES list is created will always insert at least a
space into it. When tested by the if, this space will prevent the else
clause from ever running and never generate a build error.
Remove the non-functional check. Instead, mainboards should select
warn_no_apcb or die_no_apcb to generate a warning message or build error
if the APCB is missing.
Signed-off-by: Fred Reitberger <reitbergerfred@gmail.com>
Change-Id: Ib9fe0f05739fb19da2494629dc1d5aaa0ca6431f
Reviewed-on: https://review.coreboot.org/c/coreboot/+/73006
Reviewed-by: Martin L Roth <gaumless@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Since it actually depends on the SoC type whether the old PSP
directory table pointer or the new comboable PSP directory table
pointer is used in EFS, get this information from the SoC ID instead
of passing the comboable flag for the SoCs that need to use the new
comboable PSP directory table pointer.
TEST=Binary identical on amd/majolica, pcengines/apu2, amd/gardenia
Change-Id: I0c3f21065939d1b13c2607aba16cbef74dd8d389
Signed-off-by: Zheng Bao <fishbaozi@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/73020
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Felix Held <felix-coreboot@felixheld.de>
Add support for the selective GOP init feature by only running the FSP
GOP driver when necessary: if the FMAP cache is invalid, or if the
board is booted in either recovery or developer mode.
BUG=b:255812886
TEST=tested with rest of patch train
Change-Id: I7ddadc254e05aca0fdd7a9567160a9329cb0e15c
Signed-off-by: Matt DeVillier <matt.devillier@amd.corp-partner.google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/70899
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin L Roth <gaumless@gmail.com>
Write the SHA256 hash of the cached VBIOS data when saving to FMAP,
and use it to validate the data read from FMAP on subsequent boots.
Add TPM2 as a dependency to the selection of VBIOS_CACHE_IN_FMAP.
BUG=b:255812886
TEST=tested with rest of patch train
Change-Id: I9c8f23b000b90a1072aeb7a57d3b7b2b2bc626dc
Signed-off-by: Matt DeVillier <matt.devillier@amd.corp-partner.google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/72402
Reviewed-by: Martin L Roth <gaumless@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Instead of using PCI_DEVFN(SMBUS_DEV, SMBUS_FUNC), use the equivalent
SMBUS_DEVFN define.
Even though the FCH IOAPIC is in the LPC part of the FCH, it needs the
IVRS IOAPIC table's source_dev_id field set to SMBUS_DEVFN which is the
function 0 of the FCH PCI device. LPC is function 3 of the FCH device.
When assigning LPC_DEVFN to source_dev_id, the kernel from Ubuntu
2022.04 LTS complains about the IOAPIC part of the IVRS table being
wrong:
AMD-Vi: [Firmware Bug]: : No southbridge IOAPIC found
AMD-Vi: Disabling interrupt remapping
With SMBUS_DEVFN being used as source_dev_id, no such error is reported.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Suggested-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Change-Id: I8470d67b2513031e75fb422d4c1c181e017ace0a
Reviewed-on: https://review.coreboot.org/c/coreboot/+/70503
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martin.roth@amd.corp-partner.google.com>
Reviewed-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
APOB on Phoenix is larger, so expand the reserved DRAM and MRC_CACHE
regions to fit. This requires moving memory addresses around to prevent
overlapping memory linker errors.
TEST='./util/scripts/testsoc -K PHOENIX -K GLINDA' successfully builds
all boards
Signed-off-by: Fred Reitberger <reitbergerfred@gmail.com>
Change-Id: I42af7230ca5f09ba66b2b3c4f99ac3feac7feeea
Reviewed-on: https://review.coreboot.org/c/coreboot/+/72905
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
For MDN, PHX, & Glinda platforms, the Keyboard Reset functionality has
been moved from GPIO 129 to GPIO 21.
Additionally, the issue where the system would reset when the KBDRST_L
pin went low even when not configured for Keyboard reset seems to have
been fixed, so remove that text.
Signed-off-by: Martin Roth <gaumless@gmail.com>
Change-Id: Iefe7e00d63777577b59ee98cb974b07afea1fd12
Reviewed-on: https://review.coreboot.org/c/coreboot/+/72912
Reviewed-by: Jon Murphy <jpmurphy@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Eric Lai <eric_lai@quanta.corp-partner.google.com>
One of the main functions performed by the FSP GOP driver is to modify
the ATOMBIOS tables (part of the VBIOS) in memory based on the display
output configuration. This device-specific modified VBIOS can be cached
in a FMAP region specific for that purpose, then loaded into memory
instead of the "generic" VBIOS, saving the ~130ms execution time of the
GOP driver.
As this approach only works when no pre-OS display output is needed,
limit its use to ChromeOS builds, with the GOP driver enabled, and
not booting in either recovery or developer modes.
SoCs supporting this feature will need to selectively run the FSP GOP
driver as needed, using the same criteria used here to determine
whether to load the VBIOS from CBFS or from the FMAP cache.
Boards utilizing this feature will need to add a dedicated FMAP region
with the appropriate name/size, and select the required Kconfig options.
BUG=b:255812886
TEST=tested with rest of patch train
Change-Id: Ib9cfd192500d411655a3c8fa436098897428109e
Signed-off-by: Matt DeVillier <matt.devillier@amd.corp-partner.google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/70858
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Jon Murphy <jpmurphy@google.com>
Instead of having a magic entry in the CPU device ID table list to tell
find_cpu_driver that it has reached the end of the list, introduce and
use CPU_TABLE_END. Since the vendor entry in the CPU device ID struct is
compared against X86_VENDOR_INVALID which is 0, use X86_VENDOR_INVALID
instead of the 0 in the CPU_TABLE_END definition.
TEST=Timeless build for Mandolin results in identical image.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Suggested-by: Angel Pons <th3fanbus@gmail.com>
Change-Id: I0cae6d65b2265cf5ebf90fe1a9d885d0c489eb92
Reviewed-on: https://review.coreboot.org/c/coreboot/+/72888
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Fred Reitberger <reitbergerfred@gmail.com>
For Carrizo, the soc name was set as UNKNOWN.
The change is supposed to be binary unmodified, except the SPI
settings. According to the spec, the Stoneyridge and Carrizo have the
same definition of SPI setting in EFS.
Change-Id: I9704a44773b2f541f650451ed883a51e2939e12a
Signed-off-by: Zheng Bao <fishbaozi@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/66823
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Fred Reitberger <reitbergerfred@gmail.com>
Port over the remaining AMD SoCs to use CPUID_FROM_FMS. The Glinda CPUID
still needs to be updated to the actual CPUID, but for now just change
it to use CPUID_FROM_FMS.
TEST=Resulting image of timeless build for Gardenia (Stoneyridge),
Majolica (Cezanne), Chausie (Mendocino), Mayan (Phoenix) and Birman
(Glinda) don't change.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: Ia508f857d06f3c15e3ac9f813302471348ce3d89
Reviewed-on: https://review.coreboot.org/c/coreboot/+/72862
Reviewed-by: Fred Reitberger <reitbergerfred@gmail.com>
Reviewed-by: Martin Roth <martin.roth@amd.corp-partner.google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Use CPUID_ALL_STEPPINGS_MASK to only need one CPU device ID table entry
per family & model combination and not one per stepping.
TEST=Mandolin with a Picasso APU with PICASSO_B1_CPUID (0x00810f81)
still finished mpinit and boots successfully even though now only
PICASSO_B0_CPUID (0x00810f80) with CPUID_ALL_STEPPINGS_MASK specified as
device match mask. When commenting out the line with PICASSO_B0_CPUID
as a negative test, mpinit fails as expected.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I00ba43834ad86ecffa09d60599b17d122acd0b99
Reviewed-on: https://review.coreboot.org/c/coreboot/+/72848
Reviewed-by: Fred Reitberger <reitbergerfred@gmail.com>
Reviewed-by: Martin Roth <martin.roth@amd.corp-partner.google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Instead of always doing exact matches between the CPUID read in
identify_cpu and the device entries of the CPU device ID table,
offer the possibility to use a bit mask in the CPUID matching. This
allows covering all steppings of a CPU family/model with one entry and
avoids that case of a missing new stepping causing the CPUs not being
properly initialized.
Some of the CPU device ID tables can now be deduplicated using the
CPUID_ALL_STEPPINGS_MASK define, but that's outside of the scope of this
patch.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I0540b514ca42591c0d3468307a82b5612585f614
Reviewed-on: https://review.coreboot.org/c/coreboot/+/72847
Reviewed-by: Fred Reitberger <reitbergerfred@gmail.com>
Reviewed-by: Martin Roth <martin.roth@amd.corp-partner.google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Since all SoCs define the df_mmio_control union for the bits used in the
code, data_fabric_print_mmio_conf can take advantage of that and also
print a decoded version of those bits.
Output on Mandolin before the patch:
=== Data Fabric MMIO configuration registers ===
idx control base limit
0 93 fc000000 febfffff
1 93 10000000000 ffffffffffff
2 93 d0000000 f7ffffff
3 1093 fed00000 fedfffff
4 90 0 ffff
5 90 0 ffff
6 90 0 ffff
7 90 0 ffff
Output on Mandolin with the patch:
=== Data Fabric MMIO configuration registers ===
idx base limit control R W NP F-ID
0 fc000000 febfffff 93 x x 9
1 10000000000 ffffffffffff 93 x x 9
2 d0000000 f7ffffff 93 x x 9
3 fed00000 fedfffff 1093 x x x 9
4 0 ffff 90 9
5 0 ffff 90 9
6 0 ffff 90 9
7 0 ffff 90 9
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I06e1d3a3e9abd664f59f2bb852394e7f723f2b30
Reviewed-on: https://review.coreboot.org/c/coreboot/+/72880
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Fred Reitberger <reitbergerfred@gmail.com>
In contrast to Mendocino and all other AMD SoCs in the coreboot tree,
Rembrandt, on which Mendocino is based on, has a DF_MMIO_REG_SET_SIZE of
3 instead of 4, so the next data fabric MMIO register is 3 DWORDs after
the last one instead of the 4 DWORDs on the other SoCs. This was checked
against PPR #56558 Rev 3.04.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I454ad5d182f0040db93c9b3a83941333392c6061
Reviewed-on: https://review.coreboot.org/c/coreboot/+/72879
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Fred Reitberger <reitbergerfred@gmail.com>
To be able to handle a special case, add a per-SoC define for
DF_MMIO_REG_SET_SIZE instead of having this hard-coded as 4 in the
DF_MMIO_* macros. To avoid some duplication, also introduce the
DF_MMIO_REG_OFFSET macro.
TEST=Output from data_fabric_print_mmio_conf doesn't change on Mandolin.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I67420a2973c8ef9a7f0ce19ddc0013de69731689
Reviewed-on: https://review.coreboot.org/c/coreboot/+/72878
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Fred Reitberger <reitbergerfred@gmail.com>
Since things are done a bit differently on Stoneyridge, it's probably
safer to run a test instead of assuming that the test on Picasso was
sufficient to be reasonably sure that this will also work as expected on
Stoneyridge.
TEST=No change of ACPI-related messages in dmesg with this patch.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I432752fae8be08d3cbd7d30215b350c4528c7206
Reviewed-on: https://review.coreboot.org/c/coreboot/+/72495
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Elyes Haouas <ehaouas@noos.fr>
Instead of just printing the register contents, normalize the contents
of the base and limit registers to actual MMIO addresses and then print
those. This will hopefully avoid some confusion caused by the shifted
addresses.
Output on Mandolin before the patch:
=== Data Fabric MMIO configuration registers ===
Addresses are shifted to the right by 16 bits.
idx control base limit
0 93 fc00 febf
1 93 1000000 ffffffff
2 93 d000 f7ff
3 1093 fed0 fedf
4 90 0 0
5 90 0 0
6 90 0 0
7 90 0 0
Output on Mandolin after the patch:
=== Data Fabric MMIO configuration registers ===
idx control base limit
0 93 fc000000 febfffff
1 93 10000000000 ffffffffffff
2 93 d0000000 f7ffffff
3 1093 fed00000 fedfffff
4 90 0 ffff
5 90 0 ffff
6 90 0 ffff
7 90 0 ffff
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I62eeb88ddac6a7a421fccc8e433523459117976a
Reviewed-on: https://review.coreboot.org/c/coreboot/+/72739
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Fred Reitberger <reitbergerfred@gmail.com>
