| Age | Commit message (Collapse) | Author |
|---|
|
To bring genoa_poc more in line with the other AMD SoCs, move the
reporting of the memory map up to cbmem_top from the openSIL-specific
add_opensil_memmap function to read_soc_memmap_resources. This is a
preparation for making this code common for all newer AMD SoCs.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: Ic06282baa3bb9a65d297b5717697a12d08605d2f
Reviewed-on: https://review.coreboot.org/c/coreboot/+/81388
Reviewed-by: Matt DeVillier <matt.devillier@amd.corp-partner.google.com>
Reviewed-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
|
|
Move the gpp_bridge_* device functions that are bridges to the external
PCIe ports below the corresponding mpio chip. This avoids the need for
dummy devices and does things in a slightly more coreboot-native way.
TEST=PCIe lane config reported by openSIL is identical
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Tested-by: Varshit Pandya <pandyavarshit@gmail.com>
Change-Id: I7e39bf68d30d7d00b16f943953e8207d6fe9ef41
Reviewed-on: https://review.coreboot.org/c/coreboot/+/81340
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-by: Varshit Pandya <pandyavarshit@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
|
|
Even though it has an 'amd_' prefix, the amd_pci_domain_fill_ssdt
implementation doesn't contain any AMD-specific code and can also be
used by other SoCs. So factor it out, move the implementation to
src/acpi/acpigen_pci_root_resource_producer.c, and rename it to
pci_domain_fill_ssdt. When a SoC now assigns pci_domain_fill_ssdt to its
domain operation's acpi_fill_ssdt function pointer, the PCI domain
resource producer information will be added to the SSDT.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I7bd8568cf0b7051c74adbedfe0e416a0938ccb99
Reviewed-on: https://review.coreboot.org/c/coreboot/+/80464
Reviewed-by: Patrick Rudolph <patrick.rudolph@9elements.com>
Reviewed-by: Matt DeVillier <matt.devillier@amd.corp-partner.google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
|
|
Change-Id: Ie7bc4f3ae00bb9601001dbb71e7c3c84fd4f759a
Signed-off-by: Martin Roth <gaumless@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/80596
Reviewed-by: Subrata Banik <subratabanik@google.com>
Reviewed-by: Yidi Lin <yidilin@google.com>
Reviewed-by: Felix Singer <service+coreboot-gerrit@felixsinger.de>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
|
|
Printing the data fabric MMIO decode window configuration might be
useful and it also aligns this SoC more with the other AMD family 17h+
SoCs.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I52f6655a5c63e31165549dcb6f5f95d4e74bad3d
Reviewed-on: https://review.coreboot.org/c/coreboot/+/80356
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Matt DeVillier <matt.devillier@amd.corp-partner.google.com>
|
|
Make the initialization of the IOAPIC(s) in the PCI root(s) common
across all AMD family 17h+ SoCs. For this the more general
implementation from the Genoa code that supports multiple PC roots is
moved to the common AMD code. All other family 17h+ SoCs are then
adapted to use the common code. For those non-Genoa SoCs, the
initialization of this second IOAPIC is moved from the northbridge
device to the domain device above to match Genoa.
Test=Both the FCH IOAPIC and the PCIe root IOAPIC are still initialized
on Mandolin
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I7c0ec6ac2f11cb11e46248cceec96c1fd2a49c16
Reviewed-on: https://review.coreboot.org/c/coreboot/+/80286
Reviewed-by: Matt DeVillier <matt.devillier@amd.corp-partner.google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
|
|
Provide 3 separate functions for each openSIL time point instead of one,
so that we don't need the xSIM-api header file to be included in
opensil.h to decouple the coreboot code more form the openSIL code. This
will allow to create an openSIL stub implementation to already get most
of the coreboot-side SoC code in place before the openSIL source code is
done and released.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I969bc0862560b7254c48f04e9a03387417f328bc
Reviewed-on: https://review.coreboot.org/c/coreboot/+/80287
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-by: Varshit Pandya <pandyavarshit@gmail.com>
|
|
To make add_opensil_memmap match the other function that are directly or
indirectly called by amd_pci_domain_read_resources, pass the resource
index as a pointer instead of passing it by value and then returning the
new resource index.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I6a17e488a01cc52b2dab5dd3e3d58bdf3acb554d
Reviewed-on: https://review.coreboot.org/c/coreboot/+/80269
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
|
|
Introduce read_soc_memmap_resources which gets called by
amd_pci_domain_read_resources for the first domain of the SoC to report
the DRAM and PCI config space access resources to the allocator. For
Genoa this allows to use amd_pci_domain_read_resources as read_resources
in the genoa_pci_domain_ops instead of needing to wrap that call to be
able to call add_opensil_memmap for the first domain. For the other
family 17h+ SoCs the moves the reporting of the DRAM resources and the
PCI config space access resources from the northbridge device to the
domain device.
TEST=Resources still get reported on Mandolin, but now under the domain
instead of the northbridge PCI device
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: Ib19fd94e06fa3a1d95ade7fafe22db013045a942
Reviewed-on: https://review.coreboot.org/c/coreboot/+/80268
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
|
|
Previously the code checked if the first downstream bus of the domain
was bus 0 in segment group 0 to only run certain code for the first
domain. Instead check if the domain number is 0 which should make the
code a bit easier to understand.
TEST=add_opensil_memmap still gets called exactly once on Onyx
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: Id8cc0078843e5e0361a53ba897cde508cee16aad
Reviewed-on: https://review.coreboot.org/c/coreboot/+/79996
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
|
|
This renames bus to upstream and link_list to downstream.
Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
Change-Id: I80a81b6b8606e450ff180add9439481ec28c2420
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78330
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Felix Held <felix-coreboot@felixheld.de>
|
|
Macros can be confusing on their own; hiding commas make things worse.
This can sometimes be downright misleading. A "good" example would be
the code in soc/intel/xeon_sp/spr/chip.c:
CHIP_NAME("Intel SapphireRapids-SP").enable_dev = chip_enable_dev,
This appears as CHIP_NAME() being some struct when in fact these are
defining 2 separate members of the same struct.
It was decided to remove this macro altogether, as it does not do
anything special and incurs a maintenance burden.
Change-Id: Iaed6dfb144bddcf5c43634b0c955c19afce388f0
Signed-off-by: Nicholas Sudsgaard <devel+coreboot@nsudsgaard.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/80239
Reviewed-by: Yidi Lin <yidilin@google.com>
Reviewed-by: Felix Singer <service+coreboot-gerrit@felixsinger.de>
Reviewed-by: Jakub Czapiga <czapiga@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Matt DeVillier <matt.devillier@amd.corp-partner.google.com>
|
|
ACPI_SCI_IRQ is defined as 9 for all AMD SoCs, so move the definition to
the common amdblocks/acpi.h. Since all but Stoneyridge's soc/acpi.h are
now empty, delete those files too.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I8210c98dc4cf2c6001d5273d132053278ff7fea5
Reviewed-on: https://review.coreboot.org/c/coreboot/+/80222
Reviewed-by: Matt DeVillier <matt.devillier@amd.corp-partner.google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
|
|
Since the definition is the same for all SoCs, move it to the common
amdblock/acpi.h header. Since the Stoneyridge northbridge.c file also
includes this prototype, remove the static attribute of the function
there.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: Ib9aa215f2b4ba58f43fed2c751d989f1719e0a17
Reviewed-on: https://review.coreboot.org/c/coreboot/+/80221
Reviewed-by: Matt DeVillier <matt.devillier@amd.corp-partner.google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
|
|
Instead of open-coding this functionality in all AMD SoCs, factor it out
into a common implementation.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: Idb65c398b747e70ec67107e0a1d4bd6551501347
Reviewed-on: https://review.coreboot.org/c/coreboot/+/80208
Reviewed-by: Matt DeVillier <matt.devillier@amd.corp-partner.google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Varshit Pandya <pandyavarshit@gmail.com>
Reviewed-by: Marshall Dawson <marshalldawson3rd@gmail.com>
|
|
Call setup_opensil, opensil_entry, and fch_init in the right order from
the init method of the SoC's chip operations. This brings this SoC both
more in line with the other SoCs and avoids using boot state hooks for
this which also makes the sequence in which those functions are called
easier to understand. Previously the boot states were used so that
setup_opensil was run before configure_mpio which was run before
opensil_entry(SIL_TP1), but since configure_mpio is called from
setup_opensil, this is no longer necessary.
TEST=Onyx still boots to the payload and the MPIO configuration reported
from the openSIL code is still the same. The FCH init code now runs
before the resource allocation like on the AMD SoCs that rely on FSP.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: Ic752635da5eaa9e333cfb927836f0d260d2ac049
Reviewed-on: https://review.coreboot.org/c/coreboot/+/79985
Reviewed-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
|
|
The .inc suffix is confusing to various tools as it's not specific to
Makefiles. This means that editors don't recognize the files, and don't
open them with highlighting and any other specific editor functionality.
This issue is also seen in the release notes generation script where
Makefiles get renamed before running cloc.
Signed-off-by: Martin Roth <gaumless@gmail.com>
Change-Id: Ie449267fe4fdd75110f577e1b9f748cd06140950
Reviewed-on: https://review.coreboot.org/c/coreboot/+/80071
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-by: Felix Singer <service+coreboot-gerrit@felixsinger.de>
|
|
Since the romstage code is very similar between all AMD non-CAR SoCs,
factor out a common romstage implementation. All SoCs that select
SOC_AMD_COMMON_BLOCK_PM_CHIPSET_STATE_SAVE call fill_chipset_state, so
this Kconfig option can be used to determine whether to make that call.
In the FSP case, amd_fsp_early_init gets called, while in the case of an
implementation that doesn't rely on an FSP to do the initialization,
cbmem_initialize_empty gets called to set up CBMEM which otherwise would
be done inside the FSP driver code. Since only some SoCs call
fch_disable_legacy_dma_io again in romstage right after
amd_fsp_early_init, introduce the new
SOC_AMD_COMMON_ROMSTAGE_LEGACY_DMA_FIXUP Kconfig option, so that the
SoCs can specify if this call is needed or not.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I4a0695714ba08b13a58b12a490da50cb7f5a1ca9
Reviewed-on: https://review.coreboot.org/c/coreboot/+/80083
Reviewed-by: Matt DeVillier <matt.devillier@amd.corp-partner.google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martin.roth@amd.corp-partner.google.com>
|
|
This reverts commit acbc4912375085a099c2427def464d6e481f2a90.
Reason for revert: CB:79525 fixes the issue that led to the revert
by not maintaining the heap in the SMM-stored copy of ramstage at all.
Change-Id: I3c8ef785486d275c9341859d34fce12253bd2bb9
Signed-off-by: Patrick Georgi <patrick@coreboot.org>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/80023
Reviewed-by: Sean Rhodes <sean@starlabs.systems>
Reviewed-by: Subrata Banik <subratabanik@google.com>
Reviewed-by: Julius Werner <jwerner@chromium.org>
Reviewed-by: Matt DeVillier <matt.devillier@amd.corp-partner.google.com>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
|
|
Add initial support for multiple PCI segment groups. Instead of
modifying secondary in the bus struct introduce a new segment_group
struct element and keep existing common code.
Since all platforms currently only use 1 segment this is not a
functional change. On platforms that support more than 1 segment the
segment has to be set when creating the PCI domain.
Signed-off-by: Patrick Rudolph <patrick.rudolph@9elements.com>
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: Ied3313c41896362dd989ee2ab1b1bcdced840aa8
Reviewed-on: https://review.coreboot.org/c/coreboot/+/79927
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Nico Huber <nico.h@gmx.de>
Reviewed-by: Martin Roth <martin.roth@amd.corp-partner.google.com>
|
|
Move the IOMMU_IOAPIC_IDX define from amdblocks/data_fabric.h to
amdblocks/ioapic.h which is both a more logical place for it to be and
this is also a preparation to use the common AMD MADT code for the
Stoneyridge SoC.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: Iaa20e802cf5ed93f0d05842abb1aea0d43b1cac4
Reviewed-on: https://review.coreboot.org/c/coreboot/+/79862
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
|
|
The acpi_fill_madt implementation from the Genoa PoC also works for the
other AMD SoCs that select SOC_AMD_COMMON_BLOCK_DATA_FABRIC_DOMAIN, so
factor out this function to the common AMD ACPI code and change those
other SoCs to use the new common functionality instead of having their
own implementations.
The old code on the single-domain SoCs used the GNB_IO_APIC_ADDR base
address to create the MADT entry for the additional IOAPIC in the root
complex. The new code iterates over all domains and looks for a resource
with the IOMMU_IOAPIC_IDX index in each domain and if it finds it, it
creates an MADT entry for that IOAPIC. This resource is created earlier
in the boot process when the non-PCI resources are read from the IOHC
registers and reported to the allocator.
TEST=The resulting MADT doesn't change on Mandolin
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: I4cc0d3f30b4e6ba29542dcfde84ccac90820d258
Reviewed-on: https://review.coreboot.org/c/coreboot/+/79861
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
|
|
Allow SoC code to set LAPIC access mode to X2APIC
Change-Id: I208cca35c328e1566a57aaaa8ee7809e0760261c
Signed-off-by: Varshit Pandya <pandyavarshit@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/79644
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Felix Held <felix-coreboot@felixheld.de>
|
|
The openSIL code for the Genoa SoC is only a proof of concept, so change
the name of the Kconfig option to include this code in the build from
SOC_AMD_OPENSIL_GENOA to SOC_AMD_OPENSIL_GENOA_POC to clarify that this
is code that isn't intended or ready to be productized.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: If91cdaa7c324426964bba2de2109b6c38482fab8
Reviewed-on: https://review.coreboot.org/c/coreboot/+/79574
Reviewed-by: Varshit Pandya <pandyavarshit@gmail.com>
Reviewed-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin L Roth <gaumless@gmail.com>
|
|
Even though this SoC is called 'Genoa', the openSIL implementation and
the corresponding coreboot integration is only a proof of concept that
isn't fully featured, has known limitations and bugs, and is not meant
for or ready to being productized. Adding the proof of concept suffix to
the name should point this out clearly enough so that no potential
customer could infer that this might be a fully functional and supported
implementation which it is not.
Change-Id: Ia459b1e007dcfd8e8710c12e252b2f9a4ae19b72
Signed-off-by: Varshit Pandya <pandyavarshit@gmail.com>
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/77894
Reviewed-by: Martin Roth <martin.roth@amd.corp-partner.google.com>
Reviewed-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
|
