Age | Commit message (Collapse) | Author |
|
Change the cache-as-ram teardown to use invd instead of wbinvd.
Save the return and recover the call's return address in
chipset_teardown_car.
CAR teardown had been modified to use wbinvd to send CAR contents
to DRAM backing prior to teardown. This allowed CAR variables,
stack, and local variables to be preserved while running the
AMD_DISABLE_STACK macro.
Using the wbinvd instruction has the side effect of sending all
dirty cache contents to DRAM and not only our CAR data. This
would likely cause corruption, e.g. during S3 resume.
Stoney Ridge now uses a postcar stage and this is no longer a
requirement.
BUG=b:64768556
Change-Id: I8e6bcb3947f508b1db1a42fd0714bba70074837a
Signed-off-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Reviewed-on: https://review.coreboot.org/20967
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
|
|
Insert a postcar stage for Stoney Ridge and move romstage's CAR
teardown there.
The AMD cache-as-ram teardown procedure currently uses a wbinvd
instruction to send CAR contents to DRAM backing. This allows
preserving stack contents and CAR globals after the teardown
happens, but likely results in memory corruption during S3 resume.
Due to the current base of the DCACHE region, reverting to an
invd instruction will break the detection mechanism for CAR
migrated variables. Using postcar avoids this problem.
The current behavior of AGESA is to set up all cores' MTRRs during
the AmdInitPost() entry point. This implementation takes control
back and causes postcar's _start to clear all settings and set
attributes only for the BIOS flash device, TSEG, and enough space
below cbmem_top to load and run ramstage.
BUG=b:64768556
Change-Id: I1045446655b81b806d75903d75288ab17b5e77d1
Signed-off-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Reviewed-on: https://review.coreboot.org/20966
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
|
|
The guards in the header files were inconsistent. Some had no leading or
trailing underscores, some had one, some had both leading and trailing.
Change all to double leading & trailing underscores.
Change all comments to have a space before them instead of tabs
BUG=b:62235990
Test=Build Kahlee
Change-Id: I4466df529ab201c922096a31d7438381778b582f
Signed-off-by: Martin Roth <martinroth@google.com>
Reviewed-on: https://review.coreboot.org/21073
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Marc Jones <marc@marcjonesconsulting.com>
Reviewed-by: Marshall Dawson <marshalldawson3rd@gmail.com>
|
|
Add dedicated CAR setup and teardown functions and Kconfig
options to force their inclusion into the build. The .S files
are mostly duplicated code from the old cache_as_ram.inc file.
The .S files use global proc names in anticipation for use with
the Kconfig symbols C_ENVIRONMENT_BOOTBLOCK and POSTCAR_STAGE.
Move the mainboard romstage functionality into the soc directory
and change the function name to be compatible with the call
from assembly_entry.S. Drop the BIST check like other devices.
Move InitReset and InitEarly to bootblock. These AGESA entry
points set some default settings, and release/recapture the
AP cores. There are currently some early dependencies on
InitReset. Future work should include:
* Pull the necessary functionality from InitReset into bootblock
* Move InitReset and InitEarly to car_stage_entry() and out of
bootblock
- Add a mechanism for the BSP to give the APs an address
to call and skip most of bootblock and verstage (when
available) (1)
- Reunify BiosCallOuts.c and OemCustomize.c
(1) During the InitReset call, the BSP enables the APs by setting
core enable bits in F18F0x1DC and APs begin fetching/executing
from the reset vector. The BSP waits for all APs to also
reach InitReset, where they enter an endless loop. The BSP
sends a command to them to execute a HLT instruction and the
BSP eventually returns from InitReset. The goal would be to
preserve this process but prevent APs from rerunning early
code.
Change-Id: I811c7ef875b980874f3c4b1f234f969ae5618c44
Signed-off-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Reviewed-on: https://review.coreboot.org/19755
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
Reviewed-by: Martin Roth <martinroth@google.com>
|
|
Also unify __attribute__ ((..)) to __attribute__((..)) and
handle ((__packed__)) like ((packed))
Change-Id: Ie60a51c3fa92b5009724a5b7c2932e361bf3490c
Signed-off-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Reviewed-on: https://review.coreboot.org/15921
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
|
|
Add files for supporting the BIOS->PSP communication not
covered by AGESA. The first command implemented notifies the
PSP that DRAM is ready.
This patch also introduces the amd/common/block directory
structure similar to intel/common/block.
Change-Id: I34b2744b071aa3dfb1071b2aabde32ddb662ab87
Signed-off-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Reviewed-on: https://review.coreboot.org/19753
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martinroth@google.com>
|