This will allow factoring out and moving check_mca() to soc/amd/common.
Change-Id: I92c7657baef17c248a5aef1eda268e9647502837
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/56280
Reviewed-by: Raul Rangel <rrangel@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
This patch changes the way how the not implemented MCA bank 3 gets
skipped. For the not implemented bank 3 the name gets set to NULL
resulting in mca_is_valid_bank returning false causing the bank to get
skipped. This is a preparation for commonizing the MCA(X) handing in the
soc/amd sub-tree.
Change-Id: I40d6a6752504d804c45b445fce7e763e80161211
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/56277
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martinroth@google.com>
In mca_check_all_banks only check valid MCA banks for errors. This
aligns the Picasso code a bit more with the Stoneyridge code base which
will be updated in a follow-up patch. This is a preparation for
commonizing the MCA(X) handing in the soc/amd sub-tree.
Change-Id: I0c7f3066afd220e6b8bf8308a321189d7a2679f6
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/56275
Reviewed-by: Raul Rangel <rrangel@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
The length of mca_bank_name should match the return value of
mca_get_bank_count which gets the number of MCA banks from an MSR.
TEST=No error message on serial console on amd/mandolin
Change-Id: Ibdad51a7ef27266e110dfbb43188361952618342
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/56274
Reviewed-by: Raul Rangel <rrangel@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Also use array indices for the initialization.
TEST=Checked with the public Picasso PPR #55570-B1 Rev 3.16
Change-Id: I10a65210da73e64b67d613609fcc0f9a245a81fb
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/56273
Reviewed-by: Raul Rangel <rrangel@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
According to the AMD FCH architects, we should be using the default
value for the NO_HOG bit. This fixes a problem where the SPI DMA no
longer functions after the LPC init runs.
BUG=b:179699789, b:192373221
TEST=Boot guybrush and see SPI DMA working
Signed-off-by: Raul E Rangel <rrangel@chromium.org>
Change-Id: If015869657f36d3533f4ab9ebd1f54b0d4eb283a
Reviewed-on: https://review.coreboot.org/c/coreboot/+/56233
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Marshall Dawson <marshalldawson3rd@gmail.com>
There is no technical reason this needs to be done in romstage. Moving
it into ramstage allow us (in future CLs) to use threads to pre-load
the apob from SPI.
BUG=b:179699789
TEST=Boot and Ezkinil and Guybrush and verify APOB update still work
Signed-off-by: Raul E Rangel <rrangel@chromium.org>
Change-Id: I960437ff4400645de5a3e7447fcdbc52de85943e
Reviewed-on: https://review.coreboot.org/c/coreboot/+/56227
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Karthik Ramasubramanian <kramasub@google.com>
The %p format specifier already prints out 0x, so remove the 0x from the
string. I also updated the other format specifiers to use the %# syntax
to print out the 0x.
BUG=b:179699789
TEST=see correct format.
Signed-off-by: Raul E Rangel <rrangel@chromium.org>
Change-Id: I5b00d2c06687e549f69486eb5e18f7bed560b2ec
Reviewed-on: https://review.coreboot.org/c/coreboot/+/56225
Reviewed-by: Karthik Ramasubramanian <kramasub@google.com>
Reviewed-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Currently the MCA support for Cezanne only clears the MCA status
registers. The MCA error handling and BERT table generation will be
added in subsequent patches.
Change-Id: Ib9b5174186c28c8c82f57ffd8936c8dad4e63c5b
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/56262
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Raul Rangel <rrangel@chromium.org>
Add this macro to be able to conveniently access the MC_CTL_MASK
register for each MCA bank. Also drop the unused definitions for
MC1_CTL_MASK and MC4_CTL_MASK.
Change-Id: I23ce1eac2ffce35a2b45387ee86aa77b52da5494
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/56261
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Raul Rangel <rrangel@chromium.org>
This MSR isn't an architectural MSR, so it shouldn't be in the common
x86 MSR definition header file. From family 17h on this register has
moved to a different location.
Change-Id: Id11d942876da217034e6f912b1058f00bd15c22c
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/56246
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Raul Rangel <rrangel@chromium.org>
Since we can use both the old MCA registers and the new MCAX registers
to access the MCA status registers, we can use the common
mca_clear_status function here.
Change-Id: I9ddcc119eca2659361b1496fd7ffe124fb323d26
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/56260
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Raul Rangel <rrangel@chromium.org>
When accessing the MCA MSRs, the MCA bank number gets multiplied by 4
and added to the IA32_MC0_* define to get the MSR number. Add a macro
that already does this calculation to avoid open coding this repeatedly.
Change-Id: I2de753b8c8ac8dcff5a94d5bba43aa13bbf94b99
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/56243
Reviewed-by: Raul Rangel <rrangel@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Only the fields bank and sts from the mca_bank struct were used outside
a local scope, so remove the rest. Also rename the struct that now only
contains the bank number and the status MSR content to mca_bank_status.
Change-Id: I925347dff950ac2bd021635ca988c02fba48df7f
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/56237
Reviewed-by: Raul Rangel <rrangel@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Those registers are architectural MSR and this also gets them in line
with IA32_MC0_CTL and IA32_MC0_STATUS. Also move them below the
definitions for IA32_MC0_STATUS, so that the numbers of the MSRs are
ascending.
Change-Id: Icef6526c896720248f5b648ddf1a271bdf46917c
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/56235
Reviewed-by: Raul Rangel <rrangel@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Since initial_lapicid() returns an unsigned int, change the type of the
local variables the return value gets assigned to to unsigned int as
well if applicable. Also change the printk format strings for printing
the variable's contents to %u where it was %d before.
Change-Id: I289015b81b2a9d915c4cab9b0544fc19b85df7a3
Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/55063
Reviewed-by: Raul Rangel <rrangel@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Use the common mca_get_bank_count function instead of open-coding the
functionality to get the MCA bank number. Also re-type the num_banks
variable from signed in to unsigned int, since the number of MCA bank is
always positive.
Change-Id: I126767cf9ad468cab6d6537dd73e9b2dc377b5c4
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/56185
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Marshall Dawson <marshalldawson3rd@gmail.com>
This change only enables the IOMMU device. We still require the IVRS
table to take advantage of the IOMMU. This will happen when the picasso
IVRS code is moved into common.
BUG=b:190515051
TEST=lspci shows IOMMU device
00:00.2 IOMMU: Advanced Micro Devices, Inc. [AMD] Device 1631
Signed-off-by: Raul E Rangel <rrangel@chromium.org>
Change-Id: I5c7cae3d25af5a45d48658ffa948a2856adc4346
Reviewed-on: https://review.coreboot.org/c/coreboot/+/55737
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Jason Glenesk <jason.glenesk@gmail.com>
Reviewed-by: Felix Held <felix-coreboot@felixheld.de>
The Majolica UEFI ACPI tables have this listed as shared. It's already a
level interrupt, so no reason it shouldn't be shared.
This change makes it so Windows can correctly initialize the GPIO
controller.
BUG=b:186212501
TEST=Boot guybrush to windows and see GPIO controller functional. Also
boot guybrush to windows and verify GPIO controller still works.
Signed-off-by: Raul E Rangel <rrangel@chromium.org>
Change-Id: I48c6d548a2a8d67599f25e37eeafc90764d9e2d2
Reviewed-on: https://review.coreboot.org/c/coreboot/+/56177
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Felix Held <felix-coreboot@felixheld.de>
Allow hot plug devices to subscribe to IOMMU services. Currently the
IOMMU end range is limited to device B:0 D:1f F:6. This prevents the
devices on bus 1 and higher to subscribe to IOMMU services. As per AMD
IOMMU spec v3 section 5.2.2.1 all possible device IDs must be defined,
whether the device ID is actually populated or not. Device entries are
used to report ranges when hot-plug and SR-IOV devices are possible.
With this change the hot plug devices can now bind to IOMMU services
(as tested on kernel v5.4), and below errors are not seen in dmesg.
AMD-Vi: Event logged [IO_PAGE_FAULT device=04:00.3 domain=0x0000]
AMD-Vi: Event logged [IO_PAGE_FAULT device=05:00.0 domain=0x0000]
AMD-Vi: Event logged [IO_PAGE_FAULT device=04:00.4 domain=0x0000]
TEST= Verify dGPU can enumerate on hotplug. No IO page fault errors seen.
The hot plug devices can successfully bind to IOMMU services in
kernel.
Signed-off-by: Aamir Bohra <aamirbohra@gmail.com>
Change-Id: I256c0f8032662674a4d75746de49c250e341c579
Reviewed-on: https://review.coreboot.org/c/coreboot/+/55816
Reviewed-by: Raul Rangel <rrangel@chromium.org>
Reviewed-by: Jason Glenesk <jason.glenesk@amd.corp-partner.google.com>
Reviewed-by: ritul guru <ritul.bits@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
BUG=none
TEST=Boot with CONSOLE_LOGLEVEL_3 and no longer see the message printed.
Signed-off-by: Raul E Rangel <rrangel@chromium.org>
Change-Id: I0bdb92f547ceb8be624521211f4a3b94a91dae22
Reviewed-on: https://review.coreboot.org/c/coreboot/+/55972
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martinroth@google.com>
We are currently reading the uCode for each CPU. This is unnecessary
since the uCode never changes.
BUG=b:177909625
TEST=Boot guybrush and see "microcode: being updated to patch id" for
each CPU. I no longer see CBFS access for each CPU. This drops device
initialization time by 32 ms.
Also boot Ezkinil and verify microcode was also updated.
Signed-off-by: Raul E Rangel <rrangel@chromium.org>
Change-Id: I98b9d4ce8290a1f08063176809e903e671663208
Reviewed-on: https://review.coreboot.org/c/coreboot/+/55987
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Furquan Shaikh <furquan@google.com>
Reviewed-by: Karthik Ramasubramanian <kramasub@google.com>
Start using the custom boot device.
BUG=b:179699789
TEST=Boot guybrush to OS
Signed-off-by: Raul E Rangel <rrangel@chromium.org>
Change-Id: I4ae7272677f563e8827ba154fe5177c8c01155c0
Reviewed-on: https://review.coreboot.org/c/coreboot/+/55855
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martinroth@google.com>
This is a copy of mmap_boot.c and mem_rdev_ro_ops. I split it up so it
was easier to review.
The next patches will add support for the SPI DMA controller. This will
provide a minor speed up vs using mmap reads. It will also provide the
facilities to perform asynchronous SPI loading.
BUG=b:179699789
TEST=Boot guybrush to OS
Signed-off-by: Raul E Rangel <rrangel@chromium.org>
Change-Id: Id26e2a69601d0f31e256d0010008904a447c8e21
Reviewed-on: https://review.coreboot.org/c/coreboot/+/55854
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martinroth@google.com>
Introduce the `ARCH_ALL_STAGES_X86` Kconfig symbol to automatically
select the per-stage arch options. Subsequent commits will leverage
this to allow choosing between 32-bit and 64-bit coreboot where all
stages are x86. AMD Picasso and AMD Cezanne are the only exceptions
to this rule: they disable `ARCH_ALL_STAGES_X86` and explicitly set
the per-stage arch options accordingly.
Change-Id: Ia2ddbae8c0dfb5301352d725032f6ebd370428c9
Signed-off-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/55759
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Patrick Rudolph <siro@das-labor.org>
To generalise the choice of 32-bit or 64-bit coreboot on x86 hardware,
have platforms select `ARCH_X86` directly instead of through per-stage
Kconfig options, effectively reversing the dependency order.
Change-Id: If15436817ba664398055e9efc6c7c656de3bf3e4
Signed-off-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/55758
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Patrick Rudolph <siro@das-labor.org>
With two versions of *speed_mhz_to_reported_mts() we need to call the
correct one based on the reported memory type.
BUG=b:184124605
TEST="dmidecode --type 17" in OS on Guybrush
Signed-off-by: Nikolai Vyssotski <nikolai.vyssotski@amd.corp-partner.google.com>
Change-Id: I92e834097546e3ef7130830444a80f818bdea3d5
Reviewed-on: https://review.coreboot.org/c/coreboot/+/55852
Reviewed-by: Rob Barnes <robbarnes@google.com>
Reviewed-by: Raul Rangel <rrangel@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
When initializing espi early, there may be mainboard requirements to
configure the bus properly. This allows the mainboard to do that.
BUG=192100564
TEST=Build along with next patch, eSPI works on guybrush
Change-Id: Icc02877a09b8f8ed20fd1b04f3cee0509f1a85c5
Signed-off-by: Martin Roth <martinroth@chromium.org>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/55863
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Nikolai Vyssotski <nikolai.vyssotski@amd.corp-partner.google.com>
Reviewed-by: Jason Glenesk <jason.glenesk@gmail.com>
Reviewed-by: Raul Rangel <rrangel@chromium.org>
In order to support Audio Co-processor (ACP) DMIC hardware runtime
detection on the platform, ACPI _WOV method is populated on the
concerned ACP device. This method returns the ACPI Integer value as 1
if ACP DMIC exists on the platform.
BUG=b:182960979
TEST=Build and boot to OS in guybrush. Ensure that the _WOV ACPI method
is populated under the scope of ACP device.
Scope (\_SB.PCI0.GP41.ACPD)
{
Method (_WOV, 0, NotSerialized)
{
Return (One)
}
}
Change-Id: Ide84f45f5ea2ae42d5efe71ac6d1595886157045
Signed-off-by: Karthikeyan Ramasubramanian <kramasub@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/55029
Reviewed-by: Furquan Shaikh <furquan@google.com>
Reviewed-by: Raul Rangel <rrangel@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
If the NO_EARLY_BOOTBLOCK_POSTCODES config option is enabled, configure
eSPI as early as possible in the x86 boot sequence.
We found that there are situations that can cause the system to hang if
there are any port80h postcodes sent out before eSPI is initialized.
BUG=b:191370340
TEST=Build & Boot with and without NO_EARLY_BOOTBLOCK_POSTCODES enabled.
Signed-off-by: Martin Roth <martinroth@chromium.org>
Change-Id: I0badb1c529e96ee4f81134287db53ce32473de6e
Reviewed-on: https://review.coreboot.org/c/coreboot/+/55732
Reviewed-by: Raul Rangel <rrangel@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
The original reason the BERT table was moved out of CBMEM was because
the OS was not able to access the region. This happened because the
CBMEM region was marked as type 16 in the e820 table. The OS isn't aware
of this type, so it prevents any drivers from accessing it. Depthcharge
now correctly labels the CBMEM region as reserved in the e820 table so
we can move the BERT table into CBMEM.
TEST=BERT ACPI table generation still works on AMD/Mandolin with SeaBIOS
as payload and BERT region inside CBMEM is inside a BIOS-e820 reserved
range. BERT generation also works on Zork with depthcharge.
Link: https://crrev.com/c/2939677
Signed-off-by: Raul E Rangel <rrangel@chromium.org>
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: Ie640e91c19ae5f9b275cc333284b4be34211fbf6
Reviewed-on: https://review.coreboot.org/c/coreboot/+/55279
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Our existing native function gpio configuration macro (PAD_NF) only sets
the pull. For PCIe reset, we now need to be able to set it to its
native function (PCIE_RST_L), and drive it low, then high.
BUG=b:182805349
TEST=Configure GPIO, see correct behavior.
Signed-off-by: Martin Roth <martinroth@chromium.org>
Change-Id: I636371517c99f94f76834abc4575795d51aa0368
Reviewed-on: https://review.coreboot.org/c/coreboot/+/55652
Reviewed-by: Raul Rangel <rrangel@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Pass the info if the non-graphics HD audio controller device is enabled
or disabled in the board's devicetree via a UPD to the FSP so that it
knows if it should enable or disable the corresponding device.
TEST=When adding "device ref hda on end" to the devicetree of
amd/majolica the non-graphics HD Audio controller shows up in lspci and
when that line isn't added the PCIe device doesn't show up.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I9f5e164d308906bfc788e5c2674c13c7b2ebf471
Reviewed-on: https://review.coreboot.org/c/coreboot/+/55680
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Currently the FSP only has one switch to disable both AHCI controllers.
If at least one of the two AHCI controller devices is enabled in the
board's devicetree, set the SATA enable UPD to 1 and otherwise set it to
0. Setting the UPD value to 0 when both AHCI controllers are disabled
saves around 60ms in boot time.
BUG=b:191385289
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I84e7c8bf2ab08c8254271ddfefd2e4e7d8c2e87b
Reviewed-on: https://review.coreboot.org/c/coreboot/+/55669
Reviewed-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Reviewed-by: Matt Papageorge <matthewpapa07@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Using existing defines instead of magic values improves readability of
the code. Also add comments to the MADT IRQ overrides to make it clearer
what those actually do.
TEST=Timeless build results in identical binary for amd/gardenia
(Stoneyridge), amd/mandolin (Picasso) and amd/majolica (Cezanne)
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I224ffbe8eb65bcdd5fc70c0ff8b15d55b3f6be01
Reviewed-on: https://review.coreboot.org/c/coreboot/+/55613
Reviewed-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Reviewed-by: Raul Rangel <rrangel@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
