Select the common block graphics driver for Stoneyridge.
Drop Stoney's ACPI stub for the iGPU as the device will now be
generated by the common block acpigen and put into the SSDT.
TEST=tested with rest of patch train
Change-Id: I260b964be59c1a208ff907c474243a9ace03f206
Signed-off-by: CoolStar <coolstarorganization@gmail.com>
Signed-off-by: Matt DeVillier <matt.devillier@amd.corp-partner.google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78428
Reviewed-by: Felix Held <felix-coreboot@felixheld.de>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Factor out the FSP-dependent graphics init call and header into a
separate file, so that the common graphics init can be used by non-FSP
platforms (eg Stoneyridge) without any preprocessor guards.
TEST=build google/skyrim
Change-Id: Ib025ad3adec0945b4454892d78c30b4cc79e57a0
Signed-off-by: Matt DeVillier <matt.devillier@amd.corp-partner.google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78599
Reviewed-by: Felix Held <felix-coreboot@felixheld.de>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
In EC versions older than 1.18, if the mirror flag was enabled, the
EC would mirror once the system reached S5.
When a mirror is successful, the system will automatically power
on, as it acts like it's been in G3. This led to machines turning on
when the intention was them to be off.
In 1.18 and later, they're installed when turning on. The result was
slower boot times when mirroring, but no unwanted powering on.
Because of this, coreboot no longer needs to power off when setting
the mirror flag.
Change-Id: I973c1ecd59f32d3353ca392769b44aadf5fcc9c3
Signed-off-by: Sean Rhodes <sean@starlabs.systems>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78200
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Matt DeVillier <matt.devillier@gmail.com>
Disable the GpioOverride UPD in FSP M, and comment out the Clock Request
GPIOs to ensure that coreboot doesn't touch them.
This solves behaviour that can only be described as weird:
* Devices connected to Root Ports don't initialise
* Hang seen when entering S5
* Hang when edk2 is reached
Change-Id: Idf8d2112a1c44064af73bb54fd3e1a1a429e0649
Signed-off-by: Sean Rhodes <sean@starlabs.systems>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78199
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Matt DeVillier <matt.devillier@gmail.com>
Default SD card interface (PCI 14.5) to off in the baseboard, and have
all variants which use it enable it in their override tree. This will
allow for simplification when moving to using the chipset devicetree
references in a later patch.
Change-Id: I6e1230045f54e0fee376f5eeeca9da4fb9d5f6c4
Signed-off-by: Matt DeVillier <matt.devillier@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78550
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Felix Singer <service+coreboot-gerrit@felixsinger.de>
Reviewed-by: Yuchen He <yuchenhe126@gmail.com>
Default I2C3 (proximity sensor) to off in baseboard, since all variants
which use one already enable it in their override tree. This allows
variants which do not use it (the majority) to drop it from their
override trees.
Change-Id: If17cb4538a7f64d019e4e28285fb8977de72252f
Signed-off-by: Matt DeVillier <matt.devillier@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78549
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Felix Singer <service+coreboot-gerrit@felixsinger.de>
Reviewed-by: Yuchen He <yuchenhe126@gmail.com>
Default I2C2 (digitizer) to off in the baseboard, since all variants
which use one already enable it in their override tree. This allows
variants which do not use it (the majority) to drop it from their
override trees.
Change-Id: Ife42a6b849278362c1951b80b7a95363e68a2541
Signed-off-by: Matt DeVillier <matt.devillier@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78548
Reviewed-by: Yuchen He <yuchenhe126@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Felix Singer <service+coreboot-gerrit@felixsinger.de>
Default GSPI1 (fingerprint reader) to off in baseboard, since all
variants which use one already enable it in their override tree.
This allows variants which do not use it to drop it from their
override trees.
Change-Id: I07979e35b67635ceadd3906e37de177dd081d35a
Signed-off-by: Matt DeVillier <matt.devillier@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78547
Reviewed-by: Yuchen He <yuchenhe126@gmail.com>
Reviewed-by: Felix Singer <service+coreboot-gerrit@felixsinger.de>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Create an event handler for the PEWAKE# GPIO and notify the device
driver to wake up the device.
BUG=b:301150499
TEST=Compiled and tested on google/redrix:
1. Enable runtime suspend for linux mtk_t7xx driver
2. Wait for device to enter suspended state
3. Modem should be able to wake up driver, e.g. on SIM card insert/eject
The interrupts should show up under /proc/interrupts as ACPI:Event
Signed-off-by: Paweł Anikiel <panikiel@google.com>
Change-Id: I32257689da85ea71f9de781093b3ede0cfe70a0e
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78297
Reviewed-by: Jakub Czapiga <czapiga@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
This signal gets deasserted by the WWAN modem to reactivate the PCIe
link when in low power mode. In order to handle this efficiently, the
kernel needs to set up an interrupt.
BUG=b:301150499
TEST=Compiled and tested on google/redrix
Signed-off-by: Paweł Anikiel <panikiel@google.com>
Change-Id: I37f6836aefe4a374eaff3e4bc11358be274cf563
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78416
Reviewed-by: Paul Menzel <paulepanter@mailbox.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Nick Vaccaro <nvaccaro@google.com>
Add ACPI devices for these components so that generated LPI constraints
for them have valid device references.
TEST=tested with rest of patch train
Change-Id: I3b85fec3de8f33d338425a417cc8b0f5290a5e4f
Signed-off-by: Matt DeVillier <matt.devillier@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78520
Reviewed-by: Felix Singer <service+coreboot-gerrit@felixsinger.de>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Subrata Banik <subratabanik@google.com>
Add ACPI devices for these components so that generated LPI constraints
for them have valid device references.
TEST=tested with rest of patch train
Change-Id: Ib70dc29f54d28ec1fe7b630ab3fab24bcdd08154
Signed-off-by: Matt DeVillier <matt.devillier@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78519
Reviewed-by: Felix Singer <service+coreboot-gerrit@felixsinger.de>
Reviewed-by: Subrata Banik <subratabanik@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
When walking the devicetree to generate the list of devices and minimum
sleep states, skip any devices which have the disable or hidden flags
set. This prevents adding entries for devices which are not present,
which are hidden (and likely to not have a min sleep state entry), or
generating duplicate entries in the case of PCIe remapping.
Any of these conditions are considered invalid by Windows and will
result in a BSOD with an INTERNAL_POWER_ERROR.
TEST=tested with rest of patch train
Change-Id: I06f64a72c82b9e03dc8af18700d24b3d10b7d3a7
Signed-off-by: Matt DeVillier <matt.devillier@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78518
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Subrata Banik <subratabanik@google.com>
Reviewed-by: Paul Menzel <paulepanter@mailbox.org>
Reviewed-by: Felix Singer <service+coreboot-gerrit@felixsinger.de>
If a root port is not present but was enabled in the devicetree, mark
it disabled so that no ACPI references will be generated by any
function which walks the devicetree (eg, LPI constraints).
TEST=tested with rest of patch train
Change-Id: I52e23fb1c0148a599ed736fc294e593ebbd27860
Signed-off-by: Matt DeVillier <matt.devillier@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78517
Reviewed-by: Felix Singer <service+coreboot-gerrit@felixsinger.de>
Reviewed-by: Paul Menzel <paulepanter@mailbox.org>
Reviewed-by: Subrata Banik <subratabanik@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Users have reported audio cutting in and out when playing through the
speakers on bonw15 and oryp11. This issue originally only affected
serw13 and was fixed before upstreaming. Apply the updated HDA verb
provided by Clevo to fix speaker output on these units as well.
Change-Id: I105bf165227456593863faa9bb8c4f152e49796b
Signed-off-by: Levi Portenier <levi@system76.com>
Signed-off-by: Tim Crawford <tcrawford@system76.com>
Tested-by: Daniel Sutton <daniel@system76.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78511
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Jeremy Soller <jeremy@system76.com>
Follow thermal team request, modify tcc_offset from 20 to 10.
BUG=b:306548525
TEST=Build and verified by thermal team
Change-Id: I7537e103be4cd1196c934ca72dbd61e064aed371
Signed-off-by: Tyler Wang <tyler.wang@quanta.corp-partner.google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78490
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Subrata Banik <subratabanik@google.com>
Reviewed-by: David Wu <david_wu@quanta.corp-partner.google.com>
Some Type-C monitors do not immediately assert HPD. If we enter FSP-S
before HDP is asserted, display initialisation may fail. So wait for
HPD.
This is similar to commit b40c600914 ("mainboard/hatch: Fix puff DP
output on cold boots") on puff, except we don't use
google_chromeec_wait_for_displayport() since that EC command was removed
for TCPMv2 (https://crrev.com/c/4221975). Instead we use the HPD signals
only. By waiting for any HPD signal (Type-C or HDMI), we skip waiting if
HDMI is connected, which is the same behaviour as puff and fizz.
BUG=b:303533815
BRANCH=dedede
TEST=On dexi, connect a display via a Type-C to HDMI dongle and check
the dev and recovery screens are now displayed correctly. Also check the
logs in the following cases:
Cold reboot in dev mode, Type-C to HDMI dongle:
HPD ready after 800 ms
Warm reboot in dev mode, Type-C to HDMI dongle:
HPD ready after 0 ms
Cold/warm reboot in dev mode, direct Type-C:
HPD ready after 0 ms
Cold/warm reboot in dev mode, direct HDMI:
HPD ready after 0 ms
Cold/warm reboot in dev mode, no display:
HPD not ready after 3000 ms. Abort.
Change-Id: Ib4fc071cac98a542072ffbeb6943bff4c988554c
Signed-off-by: Kenneth Chan <kenneth.chan@quanta.corp-partner.google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78450
Reviewed-by: Sheng-Liang Pan <sheng-liang.pan@quanta.corp-partner.google.com>
Reviewed-by: Ren Kuo <ren.kuo@quanta.corp-partner.google.com>
Reviewed-by: Eric Lai <ericllai@google.com>
Reviewed-by: David Wu <david_wu@quanta.corp-partner.google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
After commit e12b313844 ("drivers/pc80/rtc/option.c: Allow CMOS
defaults to extend to bank 1"), Thinkpad X200 with
CONFIG(STATIC_OPTION_TABLE) can no longer resume from s3 (detected via
bisect).
Further inspection shows that DRAM training result of GM45 is stored
in CMOS above 128 bytes in raminit_read_write_training.c, for s3 resume
to restore, but it will be erased by sanitize_cmos(), which now clears
both bank 0 and bank 1, leaving only "untrained" result restored, so s3
resume will fail.
However, resetting CMOS seems unnecessary during s3 resume. Now,
cmos_need_reset will be negated when acpi_is_wakeup_s3() returns true.
Tested: Thinkpad X200 with CONFIG(STATIC_OPTION_TABLE) can resume from
s3 again with these changes.
Change-Id: I533e83f3b95f327b0e24f4d750f8812325b7770b
Signed-off-by: Bill XIE <persmule@hardenedlinux.org>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78288
Reviewed-by: Jonathon Hall <jonathon.hall@puri.sm>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Clevo had apparently swapped the Realtek card reader for the O2 Micro
card reader for newer batches of all TGL models. Enable the BayHub
driver on everything (except bonw15, which doesn't have a card reader)
to fix LTR programming, as was done for other in commit 3d7a5bdf58
("mb/system76: Enable DRIVERS_GENERIC_BAYHUB_LV2 to fix LTR issue").
Tested on system76/galp5: CPU reaches C-states deeper than C2 when idle.
Change-Id: I3667e08acd23c12638159a2f7d2592737a34e63d
Signed-off-by: Tim Crawford <tcrawford@system76.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78298
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Jeremy Soller <jeremy@system76.com>
Reviewed-by: Eric Lai <ericllai@google.com>
Update Goodix touchpad HID to GDIX0000 for GXTP7288 and GXTP7863.
BUG=b:305118852
BRANCH=firmware-dedede-13606.B
TEST=Build and touchpads are workable
# evtest for GXTP7863
No device specified, trying to scan all of /dev/input/event*
Available devices:
/dev/input/event0: Lid Switch
/dev/input/event1: Power Button
/dev/input/event2: AT Translated Set 2 keyboard
/dev/input/event3: cros_ec_buttons
/dev/input/event4: Elan Touchscreen
/dev/input/event5: GDIX0000:00 27C6:0D51 Mouse
/dev/input/event6: GDIX0000:00 27C6:0D51 Touchpad
# evtest for GXTP7288
No device specified, trying to scan all of /dev/input/event*
Available devices:
/dev/input/event0: Lid Switch
/dev/input/event1: Power Button
/dev/input/event10: GDIX0000:00 27C6:01F5 Touchpad
/dev/input/event11: sof-da7219max98360a Headset Jack
/dev/input/event12: sof-da7219max98360a HDMI/DP,pcm=2
/dev/input/event13: sof-da7219max98360a HDMI/DP,pcm=3
/dev/input/event14: sof-da7219max98360a HDMI/DP,pcm=4
/dev/input/event2: AT Translated Set 2 keyboard
/dev/input/event3: cros_ec_buttons
/dev/input/event4: ELAN900C:00 04F3:2E5D
/dev/input/event5: ELAN900C:00 04F3:2E5D UNKNOWN
/dev/input/event6: ELAN900C:00 04F3:2E5D UNKNOWN
/dev/input/event7: ELAN900C:00 04F3:2E5D Stylus
/dev/input/event8: ELAN900C:00 04F3:2E5D Stylus
/dev/input/event9: GDIX0000:00 27C6:01F5 Mouse
Change-Id: Id2a6223bdbb2f0693149136baa853ca2efb57815
Signed-off-by: Dtrain Hsu <dtrain_hsu@compal.corp-partner.google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78503
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Eric Lai <ericllai@google.com>
When CBFS verification is enabled, add amdfw_a/b.rom at offset 128 bytes
to account for CBFS file header with hash attribute. When CBFS
verification is disabled, add amdfw_a/b.rom at offset 64 bytes to
account for CBFS file header without hash attribute.
BUG=None
TEST=Build Skyrim, Myst BIOS images with and without CBFS verification
enabled.
Change-Id: Ic374ac41df0c8fb8ce59488881ce5846e9058915
Signed-off-by: Karthikeyan Ramasubramanian <kramasub@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78425
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Matt DeVillier <matt.devillier@amd.corp-partner.google.com>
Fix the hash file names to be used to verify signed PSP binaries when
booting with VBOOT FW Slot B.
BUG=None
TEST=Build and boot to OS in Myst with PSP Verstage enabled using both
VBOOT slots A and B.
Change-Id: I89f02922bc901d8ac71d48bf5128fe6ecead43a0
Signed-off-by: Karthikeyan Ramasubramanian <kramasub@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78236
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Matt DeVillier <matt.devillier@amd.corp-partner.google.com>
Enable RO verification by GSC and RO/RW CBFS verification.
BUG=b:277087492
TEST=Build and boot to OS in Myst with CBFS verification enabled using
PSP verstage.
Change-Id: I2dd3ce59f331f89660185309ccf60c53d50e4fad
Signed-off-by: Karthikeyan Ramasubramanian <kramasub@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78235
Reviewed-by: Tim Van Patten <timvp@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Eric Lai <ericllai@google.com>
Some stalls are observed while using CCP DMA in PSP verstage -
especially with CBFS verification enabled. Also with RW CBFS
verification enabled, the entire firmware body is not loaded during
verstage for verification. Instead the files are verified as and when
they are loaded from CBFS. Hence the impact to boot time is reduced
since only few files are loaded during PSP verstage. Hence disable CCP
DMA in PSP verstage until the root cause is identified.
BUG=None
TEST=Build and boot to OS in Myst with CBFS verification enabled.
Change-Id: I22ac108b08abcfe432dfd175644393e384888e11
Signed-off-by: Karthikeyan Ramasubramanian <kramasub@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78234
Reviewed-by: Martin L Roth <gaumless@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Add SPI flash RO ranges to be verified by GSC in order to enable CBFS
verification. Also with CBFS verification enabled, CBFS metadata is
more than 64 bytes. So configure the offset of amdfw_a/b to 128 bytes -
next address aligned to 64 bytes.
BUG=b:277087492
TEST=Build and boot to OS in Myst with and without CBFS verification
enabled.
Change-Id: Ibfffd3d6fce8b80ec156a7b13b387e1df8c43347
Signed-off-by: Karthikeyan Ramasubramanian <kramasub@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78233
Reviewed-by: Tim Van Patten <timvp@google.com>
Reviewed-by: Matt DeVillier <matt.devillier@amd.corp-partner.google.com>
Reviewed-by: Martin L Roth <gaumless@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Meteor Lake TME bits [42-45] are reserved regardless of if the part
supports TME or not.
On a device with TME fused off, we noticed some reboot hangs which
have been narrowed down to internal IP routing issues when the IA
accesses the Input Output Manager (IOM) which is mapped at
0x3fff0aa0000 (0x3ff upper 32 bits).
It turns out since TME is fused off, coreboot uses the full physical
address size reported by CPUID MAXPHYADDR (46 bits). Therefore, it
allocates thunderbolt memory range on 46 bits (0x3fff upper 32 bits).
Since 4 of these bits are actually reserved, it seems that this
address range is "stripped down" to 42 bits (=> 0x3ff upper 32 bits)
resulting in potential conflict with other devices such as IOM.
BUG=b:288978352
TEST=No reboot issue on rex with TME fused off
Change-Id: I96ba23ab304257003c0413243d3ac8129ce31743
Signed-off-by: Jeremy Compostella <jeremy.compostella@intel.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78452
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
Reviewed-by: Subrata Banik <subratabanik@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
According the Intel Software Developer Manual,
CPUID.80000008H:EAX[15:8] reports the physical-address width supported
by the processor. Unfortunately, it does not necessarily reflect the
physical-address space the system can actulally use as some of those
bits can be reserved for internal hardware use.
It is critical for coreboot to know the actual physical address size.
Overestimating this size can lead to device resource overlaps due to
the hardware ignoring upper reserved bits. On rex for instance, it
creates some reboot hangs due to an overlap between thunderbolt and
Input Output Manager (IOM) address space.
As some SoCs, such as Meteor Lake, have physical address reserved bits
which cannot be probed at runtime, this commit introduces
`CPU_INTEL_COMMON_RESERVED_PHYS_ADDR_BITS' Kconfig to set the number
of physical address reserved bits at compilation time for those SoCs.
A runtime detection by hardware probing will be attempted if the value
is 0 (default).
BUG=b:288978352
Change-Id: I8748fa3e5bdfd339e973d562c5a201d5616f813e
Signed-off-by: Jeremy Compostella <jeremy.compostella@intel.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78451
Reviewed-by: Subrata Banik <subratabanik@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
This reverts commit 55b7dee278.
Reason for revert: accidentally submitted out of order / breaks tree
Change-Id: Ic15d0e3688cd54f7d678998341263e7bd30e75f2
Signed-off-by: Matt DeVillier <matt.devillier@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78525
Tested-by: Martin L Roth <gaumless@gmail.com>
Reviewed-by: Martin L Roth <gaumless@gmail.com>
This reverts commit bb5fa6419d.
Reason for revert: accidentally committed out of order; reverting to
unbreak tree
Change-Id: I36aa1fd3a0befe49b7e9e34198676f16fb08cf73
Signed-off-by: Matt DeVillier <matt.devillier@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78524
Reviewed-by: Martin L Roth <gaumless@gmail.com>
Tested-by: Martin L Roth <gaumless@gmail.com>
Having a CBFS cache scratchpad offers a generic way to decompress CBFS
files through the cbfs_map() function without having to reserve a
per-file specific memory region.
This commit introduces the x86 `RAMSTAGE_CBFS_CACHE_SIZE' Kconfig to
set a ramstage CBFS cache size. A cache size of zero disables the
CBFS cache feature. The default size is 16 KB which seems a
reasonable minimal value large enough to satisfy basic needs such as
the decompression of a small configuration file. This setting can be
adjusted depending on the platform needs and capabilities.
To support S3 suspend/resume use-case, the CBFS cache memory cannot be
released to the operating system. There are two options to meet this
requirement:
1. Define a static CBFS cache buffer (located in the .bss section)
2. Create a new CBMEM entry
Option #2 seems more powerful but considering that:
1. The CBFS cache is actually not a cache but just a scratch pad
designed to be isolated between stages
2. postcar is a very short stage not really needing CBFS cache
3. The static initialization of the `cbfs_cache' global
variable (cf. src/lib/cbfs.c) offers a simple and robust design
=> It is simpler to use a static buffer and limit the support to
ramstage.
Since some AMD SoCs (cf. `SOC_AMD_COMMON_BLOCK_NONCAR' Kconfig) define
a `_cbfs_cache' region, an extra `POSTRAM_CBFS_CACHE_IN_BSS' Kconfig
must be set to enable the use of a static buffer as the CBFS cache
scratchpad.
TEST=Decompression of vbt.bin in ramstage on rex using cbfs_map()
Change-Id: I7fbb1b51cda9f84842992e365b16c5ced1010b89
Signed-off-by: Jeremy Compostella <jeremy.compostella@intel.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/77885
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Having a CBFS cache scratchpad offers a generic way to decompress CBFS
files through the cbfs_map() function without having to reserve a
per-file specific memory region.
This commit introduces the x86 `PRERAM_CBFS_CACHE_SIZE' Kconfig to set
the pre-memory stages CBFS cache size. A cache size of zero disables
the CBFS cache feature. The default value is 16 KB which seems a
reasonable minimal value enough to satisfy basic needs such as the
decompression of a small configuration file. This setting can be
adjusted depending on the platform needs and capabilities.
We have set this size to zero for all the platforms without enough
space in Cache-As-RAM to accommodate the default size.
TEST=Decompression of vbt.bin in romstage on rex using cbfs_map()
Change-Id: Iee493f9947fddcc57576f04c3d6a2d58c7368e09
Signed-off-by: Jeremy Compostella <jeremy.compostella@intel.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/77290
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
The linker can make relocation entries of a symbol which has a value
of zero point to the undefined symbol entry. It is permitted since
when the symbol value is zero as the documentation of the relocation
entry `r_info' field states:
"If the index is STN_UNDEF, the undefined symbol index, the relocation
uses 0 as the symbol value."
The ELF binary does not really have any missing symbols. It is an
optimization as the symbol points to the undefined symbol because its
value is zero.
A typical way to hit this cbfstool limitation is to define an empty
region using the REGION macro in the linker script. Here is an
example if we assume `CONFIG_MY_REGION' is set to 0:
.car.data {
[...]
REGION(my_region, CONFIG_MY_REGION_SIZE)
[...]
}
A region is defined as follow:
#define REGION_SIZE(name) ((size_t)_##name##_size)
#define DECLARE_REGION(name) \
extern u8 _##name[]; \
extern u8 _e##name[]; \
extern u8 _##name##_size[];
So the size of the region is actually the address of the
`_##name##_size' symbol. Therefore, the `_my_region_size' symbol
address is zero and the linker can make the relocation entry of this
symbol point to the undefined symbol index.
In such a situation, cbfstool hits a segmentation fault when it
attempts to relocate the symbol in `parse_elf_to_xip_stage()'
function. We resolves this issue by making cbfstool skips relocation
entries pointing to the undefined symbol similarly to the way it skips
relocation relative to absolute symbols. A symbol which value is zero
can be considered an absolute symbol and therefore should not be
relocated.
Of course, we could argue that we could just prevent the declaration
of an empty region as illustrated in the following example:
.car.data {
[...]
#if CONFIG_MY_REGION_SIZE > 0
REGION(my_region, CONFIG_MY_REGION_SIZE)
#endif
[...]
}
However, this is not a satisfying solution because:
1. It requires to add unnecessary code in the linker script as an empty
region is a valid declaration. Such a workaround requires the code
using it to mark the region symbols as weak symbols to handle the
situation where the region is not defined.
2. There could be other situations which have yet to be uncovered which
would lead the same cbfstool crash.
3. A binary with an empty region is a valid ELF file and cbfstool
should not crash when it is asked to create an eXecute-In-Place stage
out of it.
Change-Id: I2803fd3e96e7ff7a0b22d72d50bfbce7acaeb941
Signed-off-by: Jeremy Compostella <jeremy.compostella@intel.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/77699
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
This fixes assembling with clang which complains about fpu instructions.
TEST: BUILD_TIMELESS=1 remains the same.
Change-Id: I175b8e749fafde5fb7ffb8101fc0dc892d9b4e0d
Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/74539
Reviewed-by: Patrick Rudolph <patrick.rudolph@9elements.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
As per PPR, Genoa supports up to 96 core, that is 192 threads.
It also supports dual socket.
Change-Id: I817fea7c41477f476794e9e5c16451037d01f912
Signed-off-by: Varshit Pandya <pandyavarshit@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78402
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martin.roth@amd.corp-partner.google.com>
Reviewed-by: Matt DeVillier <matt.devillier@amd.corp-partner.google.com>
Update SaGv gears and frequency values as per recommendation
from power and performance team. This change doesn't cause
negative impact on firmware boot time performance.
BUG=b:274137879
TEST=Verified the settings on google/rex using debug FSP logs.
Signed-off-by: Bora Guvendik <bora.guvendik@intel.com>
Change-Id: Ie8a81c05f25b1cdab1008d09c606d1debea6e6e4
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78268
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
Reviewed-by: Subrata Banik <subratabanik@google.com>
Crypto Engine in PSP prefers the buffer from Static RAM (SRAM). Hence if
a buffer comes from within SRAM address range, then it is passed
directly to Crypto Engine. Otherwise a bounce bufer from the stack is
used. But on SoCs like Picasso where PSP Verstage stack is mapped to a
virtual address space this check fails causing a bounce buffer to be
used and hence a stack overflow. Fix this issue by assuming that the
buffer comes from the SRAM always in such SoCs and pass the buffer
directly to crypto engine.
BUG=b:259649666
TEST=Build and boot to OS in Dalboz with unsigned PSP verstage.
Change-Id: I2161c8f0720c770efa5c05aece9584c3cbe7712a
Signed-off-by: Karthikeyan Ramasubramanian <kramasub@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78426
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Matt DeVillier <matt.devillier@amd.corp-partner.google.com>
CoolStar