With earlier flow, a chunk of CBMEM region was allocated for each SRAM
e.g., PUNIT SRAM, SOC PMC SRAM and IOE PMC SRAM. Then entire SRAM
content was copied to dedicated CBMEM region. Later in acpi_bert.c, the
BERT table was getting created for each chunk of CBMEM. This flow was
not considering creating separate entries for each region of crashlog
records. It resulted in only the first entry getting decoded from each
SRAM.
New flow aims to fix this issue. With new flow, a simple singly linked
list is created to store each region of crashlog records from all
SRAMs. The crashlog data is not copied to CBMEM. The nodes are
allocated dynamically and then copied to ACPI BERT table and then
freed. This flow also makes the overall crashlog code much simpler.
BUG=b:298234592
TEST=With this change decoding crashlog show comprehensive details,
tested on REX.
Change-Id: I43bb61485b77d786647900ca284b7f492f412aee
Signed-off-by: Pratikkumar Prajapati <pratikkumar.v.prajapati@intel.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78257
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
cpu_cl_cleanup() function checks if the SOC supports storage-off
feature. This feature allows to turn off PUNIT SSRAM to save power.
Enable the storage-off if it's supported. Enabling it also clears the
crashlog records from PUNIT SSRAM.
cpu_cl_rearm() function rearms the CPU crashlog.
BUG=b:262501347
TEST=Able to build google/rex. Verified both features get asserted.
Change-Id: Id9ba0f5db0b5d2bd57a7a21f178ef1e86ca63fae
Signed-off-by: Pratikkumar Prajapati <pratikkumar.v.prajapati@intel.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/77239
Reviewed-by: Kapil Porwal <kapilporwal@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Region with metadata tag contains information about BDF entry for
SOC PMC SRAM and IOE SRAM. We don't need to parse this as we already
define BDFs in soc/pci_devs.h for these SRAMs. Also we need to skip
to region as it does not contain any crashlog data.
BUG=b:262501347
TEST=Able to build google/rex. Able to trigger crashlog and decode
correctly.
Change-Id: Id8ed40b865cde8e89045f5c9e713398fcbff5890
Signed-off-by: Pratikkumar Prajapati <pratikkumar.v.prajapati@intel.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/76834
Reviewed-by: Subrata Banik <subratabanik@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
CPU crashlog discovery table and crashlog record is considered
invalid if first 32bits of the table is either 0x0 (no crashlog)
or 0xdeadbeef (invalid crashlog).
Crashlog record is considered consumed if bit 31 is set. So in this
case stop processing the subsequent records.
BUG=b:289600699
TEST=Able to build and verified invalid records are skipped on
google/rex.
Change-Id: Ia81bd293a533217425e44473ae85b2115c85faf6
Signed-off-by: Pratikkumar Prajapati <pratikkumar.v.prajapati@intel.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/76333
Reviewed-by: Subrata Banik <subratabanik@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>