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, and make it constant.
In the case of Intel model 2065x the mca_get_bank_count() call replaces
a magic number.
Change-Id: I245b15f57e77edca179e9e28965383a227617174
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/56244
Reviewed-by: Raul Rangel <rrangel@chromium.org>
Reviewed-by: Patrick Rudolph <siro@das-labor.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.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>
Commit 1aa60a95bd broke microcode loading for chipsets that have a
microcode blob with a total_size field set to 0. This appears to be
support for older chipsets, where the size was set to 0 and assumed to
be 2048 bytes. The fix is to change the result of the subtraction to a
signed type, and ensure the following comparison is done without
promoting the signed type to an unsigned one.
Resolves: https://ticket.coreboot.org/issues/313
Change-Id: I62def8014fd3f3bbf607b4d58ddc4dca4c695622
Signed-off-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/56153
Reviewed-by: Furquan Shaikh <furquan@google.com>
Reviewed-by: Nico Huber <nico.h@gmx.de>
Reviewed-by: Stefan Ott <coreboot@desire.ch>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Introduce `USE_EXP_X86_64_SUPPORT` in `src/arch/x86/Kconfig` and guard
it with `HAVE_EXP_X86_64_SUPPORT`. Replace the per-CPU implementations
of the same functionality with the newly-added Kconfig options. Update
documentation and the config file for QEMU accordingly.
Change-Id: I550216fd2a8323342d6b605306b0b95ffd5dcd1c
Signed-off-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/55760
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Patrick Rudolph <siro@das-labor.org>
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>
IO MWAIT redirection is disabled, which means reads to the P_LVL2 and
P_LVL3 "registers" will never produce any C-state transition requests.
Moreover, the register resource descriptors for all reported C-states
use the FFixedHW address space, not I/O.
Change-Id: I026835dd24d7ac1e1bae2d851e011e1670abaad4
Signed-off-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/55215
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Patrick Rudolph <siro@das-labor.org>
Even if IO MWAIT redirection were enabled, the base address is wrong.
Moreover, the register resource descriptors for all reported C-states
use the FFixedHW address space, not I/O.
Change-Id: Ic2faaafbe4928994aeeab8098d8e0fb6703d203d
Signed-off-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/55214
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Patrick Rudolph <siro@das-labor.org>
Microcode header supports advertising support for only one CPU
signature and processor flags. If there are multiple processor
families supported by this microcode blob, they are mentioned in
the extended signature table.
Add support to parse the extended processor signature table to
determine if the microcode blob supports the currently running CPU.
BUG=b:182234962
TEST=Booted ADL brya system with a processor whose signature/pf are
in the extended signature table of a microcode patch. Was able to
match and load the patch appropriately.
Signed-off-by: Rizwan Qureshi <rizwan.qureshi@intel.com>
Change-Id: I1466caf4a4ba1f9a0214bdde19cce57dd65dacbd
Reviewed-on: https://review.coreboot.org/c/coreboot/+/54734
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
Reviewed-by: Furquan Shaikh <furquan@google.com>
Some platforms which have large amounts of RAM and also write-combining
regions may decide to drop the WC regions in favor of the default when
preserving MTRRs for the OS. From a data safety perspective, this is
safe to do, but if, say, the graphics framebuffer is the region that is
changed from WC to UC/WB, then the performance of writing to the
framebuffer will decrease dramatically.
Modern OSes typically use Page Attribute Tables (PAT) to determine the
cacheability on a page level and usually do not touch the MTRRs. Thus,
it is believed to be safe to stop reserving MTRRs for the OS, in
general; PentiumII is the exception here in that OSes that still
support that may still require MTRRs to be available. In any case, if
the OS wants to reprogram all of the MTRRs, it is of course still free
to do so (after consulting the e820 table).
BUG=b:185452338
TEST=Verify MTRR programming on a brya (where `sa_add_dram_resources`
was faked to think it had 32 GiB of DRAM installed) and variable MTRR
map includes a WC entry for the framebuffer (and all the RAM):
MTRR: default type WB/UC MTRR counts: 13/9.
MTRR: UC selected as default type.
MTRR: 0 base 0x0000000000000000 mask 0x00003fff80000000 type 6
MTRR: 1 base 0x0000000077000000 mask 0x00003fffff000000 type 0
MTRR: 2 base 0x0000000078000000 mask 0x00003ffff8000000 type 0
MTRR: 3 base 0x0000000090000000 mask 0x00003ffff0000000 type 1
MTRR: 4 base 0x0000000100000000 mask 0x00003fff00000000 type 6
MTRR: 5 base 0x0000000200000000 mask 0x00003ffe00000000 type 6
MTRR: 6 base 0x0000000400000000 mask 0x00003ffc00000000 type 6
MTRR: 7 base 0x0000000800000000 mask 0x00003fff80000000 type 6
MTRR: 8 base 0x000000087fc00000 mask 0x00003fffffc00000 type 0
ADL has 9 variable-range MTRRs, previously 8 of them were used, and
there was no separate entry for the framebuffer, thus leaving the
default MTRR in place of uncached.
Change-Id: I2ae2851248c95fd516627b101ebcb36ec59c29c3
Signed-off-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/52522
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-by: Furquan Shaikh <furquan@google.com>
With CBnT a digest needs to be made of the IBB, Initial BootBlock, in
this case the bootblock. After that a pointer to the BPM, Boot Policy
Manifest, containing the IBB digest needs to be added to the FIT
table.
If the fit table is inside the IBB, updating it with a pointer to the
BPM, would make the digest invalid.
The proper solution is to move the FIT table out of the bootblock.
The FIT table itself does not need to be covered by the digest as it
just contains pointers to structures that can by verified by the
hardware itself, such as microcode and ACMs (Authenticated Code
Modules).
Change-Id: I352e11d5f7717147a877be16a87e9ae35ae14856
Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/50926
Reviewed-by: Patrick Rudolph <patrick.rudolph@9elements.com>
Reviewed-by: Christian Walter <christian.walter@9elements.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
CB:49896 added support in `intel_microcode_find()` to cache the found
microcode for faster subsequent accesses. This works okay when the
function succeeds in finding the microcode on BSP. However, if for any
reason, `cpu_microcode_blob.bin` does not contain a valid microcode
for the given processor, then the logic ends up attempting to find
microcode again and again every time it is called (because
`ucode_updates` is set to NULL on failed find, thus retriggering the
whole find sequence every time). This leads to a weird race condition
when multiple APs are running in parallel and executing this
function.
A snippet of the issues observed in the scenario described above:
```
...
microcode: Update skipped, already up-to-date
...
Microcode header corrupted!
...
```
1. AP reports that microcode update is being skipped since the current
version matches the version in CBFS (even though there is no matching
microcode update in CBFS).
2. AP reports microcode header is corrupted because it thinks that the
data size reported in the microcode is larger than the file read from
CBFS.
Above issues occur because each time an AP calls
`intel_microcode_find()`, it might end up seeing some intermittent
state of `ucode_updates` and taking incorrect action.
This change fixes this race condition by separating the logic for
finding microcode into an internal function `find_cbfs_microcode()`
and maintaining the caching logic in `intel_microcode_find()` using a
boolean flag `microcode_checked`.
BUG=b:182232187
TEST=Verified that `intel_microcode_find()` no longer makes repeated
attempts to find microcode from CBFS if it failed the first time.
Change-Id: I8600c830ba029e5cb9c0d7e0f1af18d87c61ad3a
Signed-off-by: Furquan Shaikh <furquan@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/51371
Reviewed-by: Patrick Rudolph
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
Reviewed-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Leverage the existing `acpigen_write_CST_package` function.
Yes, bad devicetree values can trigger undefined behavior. The old code
already had this issue, and will be addressed in subsequent commits.
Change-Id: Icec5431987d91242930efcea0c8ea4e3df3182fd
Signed-off-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/49093
Reviewed-by: Nico Huber <nico.h@gmx.de>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
CPUID result does not change when HyperThreading is disabled on
HT-enabled CPUs, which breaks `generate_cpu_entries`. Use MSR 0x35
instead, which returns the currently-enabled core and thread count.
Also rename the function to `get_logical_cores_per_package, which is
more accurate. Based on commit 920d2b77f2 (cpu/intel/206ax/acpi.c: Fix
get_cores_per_package). The MSR definition is the same for Sandy Bridge
and Haswell.
Change-Id: I5e1789d3037780b4285c9e367ff0e2b0d4365b39
Signed-off-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/49099
Reviewed-by: Nico Huber <nico.h@gmx.de>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
