GNUBoot/coreboot-kgpe-d16
2
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity
coreboot-kgpe-d16/src/drivers/mrc_cache/mrc_cache.c

740 lines
19 KiB
C
Raw Normal View History

src/drivers: Use SPDX for GPL-2.0-only files Done with sed and God Lines. Only done for C-like code for now. Change-Id: I38eaffa391ed5971217ffad74a312b1641e431c9 Signed-off-by: Angel Pons <th3fanbus@gmail.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/40051 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Felix Held <felix-coreboot@felixheld.de>
2020-04-02 23:48:19 +02:00
/* SPDX-License-Identifier: GPL-2.0-only */
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
#include <string.h>
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
#include <boot_device.h>
bootstate: remove need for #ifdef ENV_RAMSTAGE The BOOT_STATE_INIT_ENTRY macro can only be used in ramstage, however the current state of the header meant bad build errors in non-ramstage. Therefore, people had to #ifdef in the source. Remove that requirement. Change-Id: I8755fc68bbaca6b72fbe8b4db4bcc1ccb35622bd Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: http://review.coreboot.org/11492 Tested-by: build bot (Jenkins) Reviewed-by: Martin Roth <martinroth@google.com>
2015-09-03 00:34:04 +02:00
#include <bootstate.h>
soc/intel/common: remove nvm headers and c file There's no sense in having the nvm abstraction in its own files. Put that support directly into mrc_cache.c. BUG=b:69614064 Change-Id: I0f1a801c6e1a8c35f70faf9e4318bdc45955047a Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/22883 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2017-12-14 21:53:14 +01:00
#include <bootmode.h>
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
#include <console/console.h>
#include <cbmem.h>
soc/intel/common: log event when MRC cache is updated Log when the MRC cache is attempted to be updated with status of success or failure. Just one slot is supported currently which is deemed 'normal'. This is because there are more slots anticipated in the future. BUG=chrome-os-partner:59395 Change-Id: I0f81458325697aff9924cc359a4173e0d35da5da Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17231 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins)
2016-11-04 16:07:14 +01:00
#include <elog.h>
fmap: new API using region_device Instead of being pointer based use the region infrastrucutre. Additionally, this removes the need for arch-specific compilation paths. The users of the new API can use the region APIs to memory map or read the region provided by the new fmap API. Change-Id: Ie36e9ff9cb554234ec394b921f029eeed6845aee Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: http://review.coreboot.org/9170 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2015-03-29 06:56:22 +02:00
#include <fmap.h>
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
#include <ip_checksum.h>
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
#include <region_file.h>
security/vboot: Make mrc_cache hash functions generic We need to extend the functionality of the mrc_cache hash functions to work for both recovery and normal mrc_cache data. Updating the API of these functions to pass in an index to identify the hash indices for recovery and normal mode. BUG=b:150502246 BRANCH=None TEST=make sure memory training still works on nami Change-Id: I9c0bb25eafc731ca9c7a95113ab940f55997fc0f Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46432 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2020-10-16 22:15:59 +02:00
#include <security/vboot/antirollback.h>
mrc_cache: Move mrc_cache_*_hash functions into mrc_cache driver This CL would remove these calls from fsp 2.0. Platforms that select MRC_STASH_TO_CBMEM, updating the TPM NVRAM space is moved from romstage (when data stashed to CBMEM) to ramstage (when data is written back to SPI flash. BUG=b:150502246 BRANCH=None TEST=make sure memory training still works on nami Change-Id: I3088ca6927c7dbc65386c13e868afa0462086937 Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46510 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com> Reviewed-by: Julius Werner <jwerner@chromium.org>
2020-10-16 21:30:05 +02:00
#include <security/vboot/mrc_cache_hash_tpm.h>
security/vboot: Move vboot2 to security kconfig section This commit just moves the vboot sources into the security directory and fixes kconfig/makefile paths. Fix vboot2 headers Change-Id: Icd87f95640186f7a625242a3937e1dd13347eb60 Signed-off-by: Philipp Deppenwiese <zaolin@das-labor.org> Reviewed-on: https://review.coreboot.org/22074 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Martin Roth <martinroth@google.com>
2017-10-17 17:02:29 +02:00
#include <security/vboot/vboot_common.h>
soc/intel/common: remove nvm headers and c file There's no sense in having the nvm abstraction in its own files. Put that support directly into mrc_cache.c. BUG=b:69614064 Change-Id: I0f1a801c6e1a8c35f70faf9e4318bdc45955047a Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/22883 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2017-12-14 21:53:14 +01:00
#include <spi_flash.h>
bootmode: Get rid of CONFIG_BOOTMODE_STRAPS With VBOOT_VERIFY_FIRMWARE separated from CHROMEOS, move recovery and developer mode check functions to vboot. Thus, get rid of the BOOTMODE_STRAPS option which controlled these functions under src/lib. BUG=chrome-os-partner:55639 Change-Id: Ia2571026ce8976856add01095cc6be415d2be22e Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/15868 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-07-25 22:02:36 +02:00
baytrail: Move MRC cache code to a common directory This common code can be shared across Intel SOCs. Change-Id: Id9ec4ccd3fc81cbab19a7d7e13bfa3975d9802d0 Signed-off-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/196263 Reviewed-by: Aaron Durbin <adurbin@chromium.org> (cherry picked from commit f9919e2551b02056b83918d2e7b515b25541c583) Signed-off-by: Isaac Christensen <isaac.christensen@se-eng.com> Reviewed-on: http://review.coreboot.org/6967 Tested-by: build bot (Jenkins) Reviewed-by: Marc Jones <marc.jones@se-eng.com>
2014-04-22 19:46:06 +02:00
#include "mrc_cache.h"
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
#define DEFAULT_MRC_CACHE "RW_MRC_CACHE"
#define VARIABLE_MRC_CACHE "RW_VAR_MRC_CACHE"
#define RECOVERY_MRC_CACHE "RECOVERY_MRC_CACHE"
#define UNIFIED_MRC_CACHE "UNIFIED_MRC_CACHE"
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
#define MRC_DATA_SIGNATURE (('M'<<0)|('R'<<8)|('C'<<16)|('D'<<24))
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
struct mrc_metadata {
uint32_t signature;
uint32_t data_size;
uint16_t data_checksum;
uint16_t header_checksum;
uint32_t version;
Rename __attribute__((packed)) --> __packed 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>
2017-07-13 02:20:27 +02:00
} __packed;
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
soc/intel/common: Add save/restore for variable MRC data Piggy-back on existing MRC cache infrastructure to store variable MRC data. Only one set of data can be valid at given point of time. Currently this magically happens because region alignment is forced to 0x1000 and region itself is of the same size. This needs to be somehow programmatically enforced. Change-Id: I8a660d356ca760b8ff9907396fb9b34cb16cf1db Signed-off-by: Andrey Petrov <andrey.petrov@intel.com> Reviewed-on: https://review.coreboot.org/17320 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-08 17:30:06 +01:00
enum result {
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
UPDATE_FAILURE = -1,
soc/intel/common: Add save/restore for variable MRC data Piggy-back on existing MRC cache infrastructure to store variable MRC data. Only one set of data can be valid at given point of time. Currently this magically happens because region alignment is forced to 0x1000 and region itself is of the same size. This needs to be somehow programmatically enforced. Change-Id: I8a660d356ca760b8ff9907396fb9b34cb16cf1db Signed-off-by: Andrey Petrov <andrey.petrov@intel.com> Reviewed-on: https://review.coreboot.org/17320 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-08 17:30:06 +01:00
UPDATE_SUCCESS = 0,
ALREADY_UPTODATE = 1
};
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
#define NORMAL_FLAG (1 << 0)
#define RECOVERY_FLAG (1 << 1)
fmap: new API using region_device Instead of being pointer based use the region infrastrucutre. Additionally, this removes the need for arch-specific compilation paths. The users of the new API can use the region APIs to memory map or read the region provided by the new fmap API. Change-Id: Ie36e9ff9cb554234ec394b921f029eeed6845aee Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: http://review.coreboot.org/9170 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2015-03-29 06:56:22 +02:00
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
struct cache_region {
const char *name;
uint32_t cbmem_id;
int type;
int elog_slot;
mrc_cache: Add tpm_hash_index field to cache_region struct Pull selection of tpm hash index logic into cache_region struct. This CL also enables the storing of the MRC hash into the TPM NVRAM space for both recovery and non-recovery cases. This will affect all platforms with TPM2 enabled and use the MRC_CACHE driver. BUG=b:150502246 BRANCH=None TEST=make sure memory training still works on nami and lazor Change-Id: I1a744d6f40f062ca3aab6157b3747e6c1f6977f9 Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46514 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2020-10-20 19:42:26 +02:00
uint32_t tpm_hash_index;
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
int flags;
};
soc/intel/common/mrc_cache: Don't assume FMAP is tied to CHROMEOS The old code only checked for an RW_MRC_CACHE region when CONFIG_CHROMEOS was selected. This assumption is not necessarily true, as one can have FMAP without a CHROMEOS build. As a result, always search FMAP first before falling back on CBFS for locating the MRC cache region. The old logic where CHROMEOS builds would fail when RW_MRC_CACHE was not found is preserved, such that behavior does not change. Change-Id: I3596ef3235eff661af055968ea641f3e9671cdcd Signed-off-by: Alexandru Gagniuc <alexandrux.gagniuc@intel.com> Reviewed-on: https://review.coreboot.org/14757 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-04-12 23:52:29 +02:00
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
static const struct cache_region recovery_training = {
.name = RECOVERY_MRC_CACHE,
.cbmem_id = CBMEM_ID_MRCDATA,
.type = MRC_TRAINING_DATA,
.elog_slot = ELOG_MEM_CACHE_UPDATE_SLOT_RECOVERY,
mrc_cache: Add tpm_hash_index field to cache_region struct Pull selection of tpm hash index logic into cache_region struct. This CL also enables the storing of the MRC hash into the TPM NVRAM space for both recovery and non-recovery cases. This will affect all platforms with TPM2 enabled and use the MRC_CACHE driver. BUG=b:150502246 BRANCH=None TEST=make sure memory training still works on nami and lazor Change-Id: I1a744d6f40f062ca3aab6157b3747e6c1f6977f9 Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46514 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2020-10-20 19:42:26 +02:00
.tpm_hash_index = MRC_REC_HASH_NV_INDEX,
coreboot: Replace all IS_ENABLED(CONFIG_XXX) with CONFIG(XXX) This patch is a raw application of find src/ -type f | xargs sed -i -e 's/IS_ENABLED\s*(CONFIG_/CONFIG(/g' Change-Id: I6262d6d5c23cabe23c242b4f38d446b74fe16b88 Signed-off-by: Julius Werner <jwerner@chromium.org> Reviewed-on: https://review.coreboot.org/c/coreboot/+/31774 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2019-03-06 01:53:33 +01:00
#if CONFIG(HAS_RECOVERY_MRC_CACHE)
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
.flags = RECOVERY_FLAG,
#else
.flags = 0,
#endif
};
fmap: new API using region_device Instead of being pointer based use the region infrastrucutre. Additionally, this removes the need for arch-specific compilation paths. The users of the new API can use the region APIs to memory map or read the region provided by the new fmap API. Change-Id: Ie36e9ff9cb554234ec394b921f029eeed6845aee Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: http://review.coreboot.org/9170 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2015-03-29 06:56:22 +02:00
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
static const struct cache_region normal_training = {
.name = DEFAULT_MRC_CACHE,
.cbmem_id = CBMEM_ID_MRCDATA,
.type = MRC_TRAINING_DATA,
.elog_slot = ELOG_MEM_CACHE_UPDATE_SLOT_NORMAL,
mrc_cache: Add tpm_hash_index field to cache_region struct Pull selection of tpm hash index logic into cache_region struct. This CL also enables the storing of the MRC hash into the TPM NVRAM space for both recovery and non-recovery cases. This will affect all platforms with TPM2 enabled and use the MRC_CACHE driver. BUG=b:150502246 BRANCH=None TEST=make sure memory training still works on nami and lazor Change-Id: I1a744d6f40f062ca3aab6157b3747e6c1f6977f9 Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46514 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2020-10-20 19:42:26 +02:00
.tpm_hash_index = MRC_RW_HASH_NV_INDEX,
mrc_cache: Move code for triggering memory training into mrc_cache Currently the decision of whether or not to use mrc_cache in recovery mode is made within the individual platforms' drivers (ie: fsp2.0, fsp1.1, etc.). As this is not platform specific, but uses common vboot infrastructure, the code can be unified and moved into mrc_cache. The conditions are as follows: 1. If HAS_RECOVERY_MRC_CACHE, use mrc_cache data (unless retrain switch is true) 2. If !HAS_RECOVERY_MRC_CACHE && VBOOT_STARTS_IN_BOOTBLOCK, this means that memory training will occur after verified boot, meaning that mrc_cache will be filled with data from executing RW code. So in this case, we never want to use the training data in the mrc_cache for recovery mode. 3. If !HAS_RECOVERY_MRC_CACHE && VBOOT_STARTS_IN_ROMSTAGE, this means that memory training happens before verfied boot, meaning that the mrc_cache data is generated by RO code, so it is safe to use for a recovery boot. 4. Any platform that does not use vboot should be unaffected. Additionally, we have removed the MRC_CLEAR_NORMAL_CACHE_ON_RECOVERY_RETRAIN config because the mrc_cache driver takes care of invalidating the mrc_cache data for normal mode. If the platform: 1. !HAS_RECOVERY_MRC_CACHE, always invalidate mrc_cache data 2. HAS_RECOVERY_MRC_CACHE, only invalidate if retrain switch is set BUG=b:150502246 BRANCH=None TEST=1. run dut-control power_state:rec_force_mrc twice on lazor ensure that memory retraining happens both times run dut-control power_state:rec twice on lazor ensure that memory retraining happens only first time 2. remove HAS_RECOVERY_MRC_CACHE from lazor Kconfig boot twice to ensure caching of memory training occurred on each boot. Change-Id: I3875a7b4a4ba3c1aa8a3c1507b3993036a7155fc Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46855 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-10-27 23:58:31 +01:00
#if CONFIG(VBOOT_STARTS_IN_ROMSTAGE)
/*
* If VBOOT_STARTS_IN_ROMSTAGE is selected, this means that
* memory training happens before vboot (in RO) and the
* mrc_cache data is always safe to use.
*/
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
.flags = NORMAL_FLAG | RECOVERY_FLAG,
mrc_cache: Move code for triggering memory training into mrc_cache Currently the decision of whether or not to use mrc_cache in recovery mode is made within the individual platforms' drivers (ie: fsp2.0, fsp1.1, etc.). As this is not platform specific, but uses common vboot infrastructure, the code can be unified and moved into mrc_cache. The conditions are as follows: 1. If HAS_RECOVERY_MRC_CACHE, use mrc_cache data (unless retrain switch is true) 2. If !HAS_RECOVERY_MRC_CACHE && VBOOT_STARTS_IN_BOOTBLOCK, this means that memory training will occur after verified boot, meaning that mrc_cache will be filled with data from executing RW code. So in this case, we never want to use the training data in the mrc_cache for recovery mode. 3. If !HAS_RECOVERY_MRC_CACHE && VBOOT_STARTS_IN_ROMSTAGE, this means that memory training happens before verfied boot, meaning that the mrc_cache data is generated by RO code, so it is safe to use for a recovery boot. 4. Any platform that does not use vboot should be unaffected. Additionally, we have removed the MRC_CLEAR_NORMAL_CACHE_ON_RECOVERY_RETRAIN config because the mrc_cache driver takes care of invalidating the mrc_cache data for normal mode. If the platform: 1. !HAS_RECOVERY_MRC_CACHE, always invalidate mrc_cache data 2. HAS_RECOVERY_MRC_CACHE, only invalidate if retrain switch is set BUG=b:150502246 BRANCH=None TEST=1. run dut-control power_state:rec_force_mrc twice on lazor ensure that memory retraining happens both times run dut-control power_state:rec twice on lazor ensure that memory retraining happens only first time 2. remove HAS_RECOVERY_MRC_CACHE from lazor Kconfig boot twice to ensure caching of memory training occurred on each boot. Change-Id: I3875a7b4a4ba3c1aa8a3c1507b3993036a7155fc Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46855 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-10-27 23:58:31 +01:00
#else
/*
* If !VBOOT_STARTS_IN_ROMSTAGE, this means that memory training happens after
* vboot (in RW code) and is never safe to use in recovery.
*/
.flags = NORMAL_FLAG,
#endif
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
};
fmap: new API using region_device Instead of being pointer based use the region infrastrucutre. Additionally, this removes the need for arch-specific compilation paths. The users of the new API can use the region APIs to memory map or read the region provided by the new fmap API. Change-Id: Ie36e9ff9cb554234ec394b921f029eeed6845aee Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: http://review.coreboot.org/9170 Tested-by: build bot (Jenkins) Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2015-03-29 06:56:22 +02:00
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
static const struct cache_region variable_data = {
.name = VARIABLE_MRC_CACHE,
.cbmem_id = CBMEM_ID_VAR_MRCDATA,
.type = MRC_VARIABLE_DATA,
.elog_slot = ELOG_MEM_CACHE_UPDATE_SLOT_VARIABLE,
mrc_cache: Add tpm_hash_index field to cache_region struct Pull selection of tpm hash index logic into cache_region struct. This CL also enables the storing of the MRC hash into the TPM NVRAM space for both recovery and non-recovery cases. This will affect all platforms with TPM2 enabled and use the MRC_CACHE driver. BUG=b:150502246 BRANCH=None TEST=make sure memory training still works on nami and lazor Change-Id: I1a744d6f40f062ca3aab6157b3747e6c1f6977f9 Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46514 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2020-10-20 19:42:26 +02:00
.tpm_hash_index = 0,
mrc_cache: Move code for triggering memory training into mrc_cache Currently the decision of whether or not to use mrc_cache in recovery mode is made within the individual platforms' drivers (ie: fsp2.0, fsp1.1, etc.). As this is not platform specific, but uses common vboot infrastructure, the code can be unified and moved into mrc_cache. The conditions are as follows: 1. If HAS_RECOVERY_MRC_CACHE, use mrc_cache data (unless retrain switch is true) 2. If !HAS_RECOVERY_MRC_CACHE && VBOOT_STARTS_IN_BOOTBLOCK, this means that memory training will occur after verified boot, meaning that mrc_cache will be filled with data from executing RW code. So in this case, we never want to use the training data in the mrc_cache for recovery mode. 3. If !HAS_RECOVERY_MRC_CACHE && VBOOT_STARTS_IN_ROMSTAGE, this means that memory training happens before verfied boot, meaning that the mrc_cache data is generated by RO code, so it is safe to use for a recovery boot. 4. Any platform that does not use vboot should be unaffected. Additionally, we have removed the MRC_CLEAR_NORMAL_CACHE_ON_RECOVERY_RETRAIN config because the mrc_cache driver takes care of invalidating the mrc_cache data for normal mode. If the platform: 1. !HAS_RECOVERY_MRC_CACHE, always invalidate mrc_cache data 2. HAS_RECOVERY_MRC_CACHE, only invalidate if retrain switch is set BUG=b:150502246 BRANCH=None TEST=1. run dut-control power_state:rec_force_mrc twice on lazor ensure that memory retraining happens both times run dut-control power_state:rec twice on lazor ensure that memory retraining happens only first time 2. remove HAS_RECOVERY_MRC_CACHE from lazor Kconfig boot twice to ensure caching of memory training occurred on each boot. Change-Id: I3875a7b4a4ba3c1aa8a3c1507b3993036a7155fc Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46855 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-10-27 23:58:31 +01:00
#if CONFIG(VBOOT_STARTS_IN_ROMSTAGE)
/*
* If VBOOT_STARTS_IN_ROMSTAGE is selected, this means that
* memory training happens before vboot (in RO) and the
* mrc_cache data is always safe to use.
*/
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
.flags = NORMAL_FLAG | RECOVERY_FLAG,
mrc_cache: Move code for triggering memory training into mrc_cache Currently the decision of whether or not to use mrc_cache in recovery mode is made within the individual platforms' drivers (ie: fsp2.0, fsp1.1, etc.). As this is not platform specific, but uses common vboot infrastructure, the code can be unified and moved into mrc_cache. The conditions are as follows: 1. If HAS_RECOVERY_MRC_CACHE, use mrc_cache data (unless retrain switch is true) 2. If !HAS_RECOVERY_MRC_CACHE && VBOOT_STARTS_IN_BOOTBLOCK, this means that memory training will occur after verified boot, meaning that mrc_cache will be filled with data from executing RW code. So in this case, we never want to use the training data in the mrc_cache for recovery mode. 3. If !HAS_RECOVERY_MRC_CACHE && VBOOT_STARTS_IN_ROMSTAGE, this means that memory training happens before verfied boot, meaning that the mrc_cache data is generated by RO code, so it is safe to use for a recovery boot. 4. Any platform that does not use vboot should be unaffected. Additionally, we have removed the MRC_CLEAR_NORMAL_CACHE_ON_RECOVERY_RETRAIN config because the mrc_cache driver takes care of invalidating the mrc_cache data for normal mode. If the platform: 1. !HAS_RECOVERY_MRC_CACHE, always invalidate mrc_cache data 2. HAS_RECOVERY_MRC_CACHE, only invalidate if retrain switch is set BUG=b:150502246 BRANCH=None TEST=1. run dut-control power_state:rec_force_mrc twice on lazor ensure that memory retraining happens both times run dut-control power_state:rec twice on lazor ensure that memory retraining happens only first time 2. remove HAS_RECOVERY_MRC_CACHE from lazor Kconfig boot twice to ensure caching of memory training occurred on each boot. Change-Id: I3875a7b4a4ba3c1aa8a3c1507b3993036a7155fc Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46855 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-10-27 23:58:31 +01:00
#else
/*
* If !VBOOT_STARTS_IN_ROMSTAGE, this means that memory training happens after
* vboot (in RW code) and is never safe to use in recovery.
*/
.flags = NORMAL_FLAG,
#endif
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
};
soc/intel/common/mrc_cache: Don't assume FMAP is tied to CHROMEOS The old code only checked for an RW_MRC_CACHE region when CONFIG_CHROMEOS was selected. This assumption is not necessarily true, as one can have FMAP without a CHROMEOS build. As a result, always search FMAP first before falling back on CBFS for locating the MRC cache region. The old logic where CHROMEOS builds would fail when RW_MRC_CACHE was not found is preserved, such that behavior does not change. Change-Id: I3596ef3235eff661af055968ea641f3e9671cdcd Signed-off-by: Alexandru Gagniuc <alexandrux.gagniuc@intel.com> Reviewed-on: https://review.coreboot.org/14757 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-04-12 23:52:29 +02:00
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
/* Order matters here for priority in matching. */
static const struct cache_region *cache_regions[] = {
&recovery_training,
&normal_training,
&variable_data,
};
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
mrc_cache: Move mrc_cache_*_hash functions into mrc_cache driver This CL would remove these calls from fsp 2.0. Platforms that select MRC_STASH_TO_CBMEM, updating the TPM NVRAM space is moved from romstage (when data stashed to CBMEM) to ramstage (when data is written back to SPI flash. BUG=b:150502246 BRANCH=None TEST=make sure memory training still works on nami Change-Id: I3088ca6927c7dbc65386c13e868afa0462086937 Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46510 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com> Reviewed-by: Julius Werner <jwerner@chromium.org>
2020-10-16 21:30:05 +02:00
/* TPM MRC hash functionality depends on vboot starting before memory init. */
_Static_assert(!CONFIG(MRC_SAVE_HASH_IN_TPM) ||
CONFIG(VBOOT_STARTS_IN_BOOTBLOCK),
"for TPM MRC hash functionality, vboot must start in bootblock");
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
static int lookup_region_by_name(const char *name, struct region *r)
mrc: Add support for separate training cache in recovery mode 1. Re-factor MRC cache driver to properly select RW_MRC_CACHE or RECOVERY_MRC_CACHE based on the boot mode. - If normal mode boot, use RW_MRC_CACHE, if available. - If recovery mode boot: - Retrain memory if RECOVERY_MRC_CACHE not present, or recovery is requested explicity with retrain memory request. - Use RECOVERY_MRC_CACHE otherwise. 2. Protect RW and RECOVERY mrc caches in recovery and non-recovery boot modes. Check if both are present under one unified region and protect that region as a whole. Else try protecting individual regions. 3. Update training data in appropriate cache: - Use RW_MRC_CACHE if normal mode. - Use RECOVERY_MRC_CACHE if present in recovery mode. Else use RW_MRC_CACHE. 4. Add proper debug logs to indicate which training data cache is used at any point. BUG=chrome-os-partner:59352 BRANCH=None TEST=Verified that correct cache is used in both normal and recovery mode on reef. Change-Id: Ie79737a1450bd1ff71543e44a5a3e16950e70fb3 Signed-off-by: Furquan Shaikh <furquan@chromium.org> Reviewed-on: https://review.coreboot.org/17242 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-06 07:57:02 +01:00
{
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
if (fmap_locate_area(name, r) == 0)
mrc: Add support for separate training cache in recovery mode 1. Re-factor MRC cache driver to properly select RW_MRC_CACHE or RECOVERY_MRC_CACHE based on the boot mode. - If normal mode boot, use RW_MRC_CACHE, if available. - If recovery mode boot: - Retrain memory if RECOVERY_MRC_CACHE not present, or recovery is requested explicity with retrain memory request. - Use RECOVERY_MRC_CACHE otherwise. 2. Protect RW and RECOVERY mrc caches in recovery and non-recovery boot modes. Check if both are present under one unified region and protect that region as a whole. Else try protecting individual regions. 3. Update training data in appropriate cache: - Use RW_MRC_CACHE if normal mode. - Use RECOVERY_MRC_CACHE if present in recovery mode. Else use RW_MRC_CACHE. 4. Add proper debug logs to indicate which training data cache is used at any point. BUG=chrome-os-partner:59352 BRANCH=None TEST=Verified that correct cache is used in both normal and recovery mode on reef. Change-Id: Ie79737a1450bd1ff71543e44a5a3e16950e70fb3 Signed-off-by: Furquan Shaikh <furquan@chromium.org> Reviewed-on: https://review.coreboot.org/17242 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-06 07:57:02 +01:00
return 0;
drivers/mrc_cache: Always generate an FMAP region This automatically generates an FMAP region for the MRC_CACHE driver which is easier to handle than a cbfsfile. Adds some spaces and more comments to Makefile.inc to improve readability. Tested on Thinkpad x200 with some proof of concept patches. Change-Id: Iaaca36b1123b094ec1bbe5df4fb25660919173ca Signed-off-by: Arthur Heymans <arthur@aheymans.xyz> Reviewed-on: https://review.coreboot.org/23150 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
2018-01-07 23:12:49 +01:00
return -1;
mrc: Add support for separate training cache in recovery mode 1. Re-factor MRC cache driver to properly select RW_MRC_CACHE or RECOVERY_MRC_CACHE based on the boot mode. - If normal mode boot, use RW_MRC_CACHE, if available. - If recovery mode boot: - Retrain memory if RECOVERY_MRC_CACHE not present, or recovery is requested explicity with retrain memory request. - Use RECOVERY_MRC_CACHE otherwise. 2. Protect RW and RECOVERY mrc caches in recovery and non-recovery boot modes. Check if both are present under one unified region and protect that region as a whole. Else try protecting individual regions. 3. Update training data in appropriate cache: - Use RW_MRC_CACHE if normal mode. - Use RECOVERY_MRC_CACHE if present in recovery mode. Else use RW_MRC_CACHE. 4. Add proper debug logs to indicate which training data cache is used at any point. BUG=chrome-os-partner:59352 BRANCH=None TEST=Verified that correct cache is used in both normal and recovery mode on reef. Change-Id: Ie79737a1450bd1ff71543e44a5a3e16950e70fb3 Signed-off-by: Furquan Shaikh <furquan@chromium.org> Reviewed-on: https://review.coreboot.org/17242 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-06 07:57:02 +01:00
}
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
static const struct cache_region *lookup_region_type(int type)
mrc: Add support for separate training cache in recovery mode 1. Re-factor MRC cache driver to properly select RW_MRC_CACHE or RECOVERY_MRC_CACHE based on the boot mode. - If normal mode boot, use RW_MRC_CACHE, if available. - If recovery mode boot: - Retrain memory if RECOVERY_MRC_CACHE not present, or recovery is requested explicity with retrain memory request. - Use RECOVERY_MRC_CACHE otherwise. 2. Protect RW and RECOVERY mrc caches in recovery and non-recovery boot modes. Check if both are present under one unified region and protect that region as a whole. Else try protecting individual regions. 3. Update training data in appropriate cache: - Use RW_MRC_CACHE if normal mode. - Use RECOVERY_MRC_CACHE if present in recovery mode. Else use RW_MRC_CACHE. 4. Add proper debug logs to indicate which training data cache is used at any point. BUG=chrome-os-partner:59352 BRANCH=None TEST=Verified that correct cache is used in both normal and recovery mode on reef. Change-Id: Ie79737a1450bd1ff71543e44a5a3e16950e70fb3 Signed-off-by: Furquan Shaikh <furquan@chromium.org> Reviewed-on: https://review.coreboot.org/17242 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-06 07:57:02 +01:00
{
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
int i;
int flags;
vboot: Clean up pre-RAM use of vboot_recovery_mode_enabled() vboot_recovery_mode_enabled() was recently changed to assert() when it is called before vboot logic has run, because we cannot determine whether we're going to be in recovery mode at that point and we wanted to flush out existing uses that pretended that we could. Turns out there are a bunch of uses like that, and there is some code that is shared across configurations that can and those that can't. This patch cleans them up to either remove checks that cannot return true, or add explicit Kconfig guards to clarify that the code is shared. This means that using a separate recovery MRC cache is no longer supported on boards that use VBOOT_STARTS_IN_ROMSTAGE (this has already been broken with CB:38780, but with this patch those boards will boot again using their normal MRC caches rather than just die). Skipping the MRC cache and always regenerating from scratch in recovery mode is likewise no longer supported for VBOOT_STARTS_IN_ROMSTAGE. For FSP1.1 boards, none of them support VBOOT_STARTS_IN_BOOTBLOCK and that is unlikely to change in the future so we will just hardcode that fact in Kconfig (otherwise, fsp1.1 raminit would also have to be fixed to work around this issue). Signed-off-by: Julius Werner <jwerner@chromium.org> Change-Id: I31bfc7663724fdacab9955224dcaf650d1ec1c3c Reviewed-on: https://review.coreboot.org/c/coreboot/+/39221 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2020-03-03 00:54:43 +01:00
if (CONFIG(VBOOT_STARTS_IN_BOOTBLOCK) && vboot_recovery_mode_enabled())
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
flags = RECOVERY_FLAG;
else
flags = NORMAL_FLAG;
for (i = 0; i < ARRAY_SIZE(cache_regions); i++) {
if (cache_regions[i]->type != type)
continue;
if ((cache_regions[i]->flags & flags) == flags)
return cache_regions[i];
mrc: Add support for separate training cache in recovery mode 1. Re-factor MRC cache driver to properly select RW_MRC_CACHE or RECOVERY_MRC_CACHE based on the boot mode. - If normal mode boot, use RW_MRC_CACHE, if available. - If recovery mode boot: - Retrain memory if RECOVERY_MRC_CACHE not present, or recovery is requested explicity with retrain memory request. - Use RECOVERY_MRC_CACHE otherwise. 2. Protect RW and RECOVERY mrc caches in recovery and non-recovery boot modes. Check if both are present under one unified region and protect that region as a whole. Else try protecting individual regions. 3. Update training data in appropriate cache: - Use RW_MRC_CACHE if normal mode. - Use RECOVERY_MRC_CACHE if present in recovery mode. Else use RW_MRC_CACHE. 4. Add proper debug logs to indicate which training data cache is used at any point. BUG=chrome-os-partner:59352 BRANCH=None TEST=Verified that correct cache is used in both normal and recovery mode on reef. Change-Id: Ie79737a1450bd1ff71543e44a5a3e16950e70fb3 Signed-off-by: Furquan Shaikh <furquan@chromium.org> Reviewed-on: https://review.coreboot.org/17242 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-06 07:57:02 +01:00
}
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
return NULL;
mrc: Add support for separate training cache in recovery mode 1. Re-factor MRC cache driver to properly select RW_MRC_CACHE or RECOVERY_MRC_CACHE based on the boot mode. - If normal mode boot, use RW_MRC_CACHE, if available. - If recovery mode boot: - Retrain memory if RECOVERY_MRC_CACHE not present, or recovery is requested explicity with retrain memory request. - Use RECOVERY_MRC_CACHE otherwise. 2. Protect RW and RECOVERY mrc caches in recovery and non-recovery boot modes. Check if both are present under one unified region and protect that region as a whole. Else try protecting individual regions. 3. Update training data in appropriate cache: - Use RW_MRC_CACHE if normal mode. - Use RECOVERY_MRC_CACHE if present in recovery mode. Else use RW_MRC_CACHE. 4. Add proper debug logs to indicate which training data cache is used at any point. BUG=chrome-os-partner:59352 BRANCH=None TEST=Verified that correct cache is used in both normal and recovery mode on reef. Change-Id: Ie79737a1450bd1ff71543e44a5a3e16950e70fb3 Signed-off-by: Furquan Shaikh <furquan@chromium.org> Reviewed-on: https://review.coreboot.org/17242 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-06 07:57:02 +01:00
}
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
static const struct cache_region *lookup_region(struct region *r, int type)
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
{
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
const struct cache_region *cr;
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
cr = lookup_region_type(type);
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
if (cr == NULL) {
printk(BIOS_ERR, "MRC: failed to locate region type %d.\n",
mrc_cache: Update mrc_cache data in romstage Previously, we were writing to cbmem after memory training and then writing the training data from cbmem to mrc_cache in ramstage. We were doing this because we were unable to read/write to SPI simultaneously on older x86 chips. Now that newer chips allow for simultaneously reads and writes, we can move the mrc_cache update into romstage. This is beneficial if there is a reboot for some reason after memory training but before the previous mrc_cache_stash_data call originally in ramstage. If this happens, we would lose all the mrc_cache training data in the next boot even though we've already performed the memory training. Added new config BOOT_DEVICE_SPI_FLASH_NO_EARLY_WRITES to accomodate older x86 platforms that don't do mmapping but still want to use the cbmem to store the mrc_cache data in order to write the mrc_cache data back at a later time. We are maintaining the use of cbmem for these older platforms because we have no way of validating the earlier write back to mrc_cache at this time. BUG=b:150502246 BRANCH=None TEST=reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I3430bda45484cb8c2b01ab9614508039dfaac9a3 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44196 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-05-02 02:00:31 +02:00
type);
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
return NULL;
}
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
if (lookup_region_by_name(cr->name, r) < 0)
return NULL;
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
return cr;
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
}
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
static int mrc_header_valid(struct region_device *rdev, struct mrc_metadata *md)
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
{
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
uint16_t checksum;
uint16_t checksum_result;
size_t size;
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
if (rdev_readat(rdev, md, 0, sizeof(*md)) < 0) {
printk(BIOS_ERR, "MRC: couldn't read metadata\n");
return -1;
}
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
if (md->signature != MRC_DATA_SIGNATURE) {
printk(BIOS_ERR, "MRC: invalid header signature\n");
return -1;
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
}
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
/* Compute checksum over header with 0 as the value. */
checksum = md->header_checksum;
md->header_checksum = 0;
checksum_result = compute_ip_checksum(md, sizeof(*md));
soc/intel/common: Store MRC data in next available slot in the cache Currently, coreboot performs an erase of the entire MRC cache region on flash if there is a version mismatch for the MRC data. Instead of doing that, store the new MRC data in the next available slot, if there is enough space available in the cache region. BUG=chrome-os-partner:55699 Change-Id: Ib24a94f0a47c79941ed9f60095360ae3aad5540b Signed-off-by: Furquan Shaikh <furquan@google.com> Reviewed-on: https://review.coreboot.org/15915 Tested-by: build bot (Jenkins) Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-07-27 17:04:54 +02:00
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
if (checksum != checksum_result) {
printk(BIOS_ERR, "MRC: header checksum mismatch: %x vs %x\n",
checksum, checksum_result);
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
return -1;
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
}
/* Put back original. */
md->header_checksum = checksum;
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
/* Re-size the region device according to the metadata as a region_file
* does block allocation. */
size = sizeof(*md) + md->data_size;
if (rdev_chain(rdev, rdev, 0, size) < 0) {
printk(BIOS_ERR, "MRC: size exceeds rdev size: %zx vs %zx\n",
size, region_device_sz(rdev));
mrc_cache: add version field In order to allow for updateable memory init code on intel x86 platforms one needs to ensure the saved mrc data matches the code consuming the data. To that end add a version field to the saved data structure. BUG=chrome-os-partner:46050 BRANCH=None TEST=Built and booted on glados. Suspended and resumed. Also verified version mismatch path. Change-Id: Ie86db1750af5d9bff6446999b0d04b60612f8d29 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/12700 Tested-by: build bot (Jenkins) Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org> Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2015-12-09 22:04:49 +01:00
return -1;
}
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
return 0;
}
mrc_cache: Add tpm_hash_index field to cache_region struct Pull selection of tpm hash index logic into cache_region struct. This CL also enables the storing of the MRC hash into the TPM NVRAM space for both recovery and non-recovery cases. This will affect all platforms with TPM2 enabled and use the MRC_CACHE driver. BUG=b:150502246 BRANCH=None TEST=make sure memory training still works on nami and lazor Change-Id: I1a744d6f40f062ca3aab6157b3747e6c1f6977f9 Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46514 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2020-10-20 19:42:26 +02:00
static int mrc_data_valid(int type, const struct mrc_metadata *md,
mrc_cache: Add mrc_cache fetch functions to support non-x86 platforms Create two new functions to fetch mrc_cache data (replacing mrc_cache_get_current): - mrc_cache_load_current: fetches the mrc_cache data and drops it into the given buffer. This is useful for ARM platforms where the mmap operation is very expensive. - mrc_cache_mmap_leak: fetch the mrc_cache data and puts it into a given buffer. This is useful for platforms where the mmap operation is a no-op (like x86 platforms). As the name mentions, we are not freeing the memory that we allocated with the mmap, so it is the caller's responsibility to do so. Additionally, we are replacing mrc_cache_latest with mrc_cache_get_latest_slot_info, which does not check the validity of the data when retrieving the current mrc_cache slot. This allows the caller some flexibility in deciding where they want the mrc_cache data stored (either in an mmaped region or at a given address). BUG=b:150502246 BRANCH=None TEST=Testing on a nami (x86) device: reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I259dd4f550719d821bbafa2d445cbae6ea22e988 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44006 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-07-24 01:10:52 +02:00
void *data, size_t data_size)
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
{
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
uint16_t checksum;
mrc_cache: Add tpm_hash_index field to cache_region struct Pull selection of tpm hash index logic into cache_region struct. This CL also enables the storing of the MRC hash into the TPM NVRAM space for both recovery and non-recovery cases. This will affect all platforms with TPM2 enabled and use the MRC_CACHE driver. BUG=b:150502246 BRANCH=None TEST=make sure memory training still works on nami and lazor Change-Id: I1a744d6f40f062ca3aab6157b3747e6c1f6977f9 Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46514 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2020-10-20 19:42:26 +02:00
const struct cache_region *cr = lookup_region_type(type);
uint32_t hash_idx;
if (cr == NULL)
return -1;
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
mrc_cache: Add mrc_cache fetch functions to support non-x86 platforms Create two new functions to fetch mrc_cache data (replacing mrc_cache_get_current): - mrc_cache_load_current: fetches the mrc_cache data and drops it into the given buffer. This is useful for ARM platforms where the mmap operation is very expensive. - mrc_cache_mmap_leak: fetch the mrc_cache data and puts it into a given buffer. This is useful for platforms where the mmap operation is a no-op (like x86 platforms). As the name mentions, we are not freeing the memory that we allocated with the mmap, so it is the caller's responsibility to do so. Additionally, we are replacing mrc_cache_latest with mrc_cache_get_latest_slot_info, which does not check the validity of the data when retrieving the current mrc_cache slot. This allows the caller some flexibility in deciding where they want the mrc_cache data stored (either in an mmaped region or at a given address). BUG=b:150502246 BRANCH=None TEST=Testing on a nami (x86) device: reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I259dd4f550719d821bbafa2d445cbae6ea22e988 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44006 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-07-24 01:10:52 +02:00
if (md->data_size != data_size)
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
return -1;
mrc_cache: Add tpm_hash_index field to cache_region struct Pull selection of tpm hash index logic into cache_region struct. This CL also enables the storing of the MRC hash into the TPM NVRAM space for both recovery and non-recovery cases. This will affect all platforms with TPM2 enabled and use the MRC_CACHE driver. BUG=b:150502246 BRANCH=None TEST=make sure memory training still works on nami and lazor Change-Id: I1a744d6f40f062ca3aab6157b3747e6c1f6977f9 Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46514 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2020-10-20 19:42:26 +02:00
hash_idx = cr->tpm_hash_index;
mrc_cache: Remove unnecessary data checksum calculation When MRC_SAVE_HASH_IN_TPM is selected, we can just use the TPM hash to verify the MRC_CACHE data. Thus, we don't need to calculate the checksum anymore in this case. BUG=b:150502246 BRANCH=None TEST=make sure memory training still works on nami Change-Id: I1db4469da49755805b541f50c7ef2f9cdb749425 Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46515 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2020-10-17 09:53:41 +02:00
if (hash_idx && CONFIG(MRC_SAVE_HASH_IN_TPM)) {
if (!mrc_cache_verify_hash(hash_idx, data, data_size))
return -1;
} else {
checksum = compute_ip_checksum(data, data_size);
if (md->data_checksum != checksum) {
printk(BIOS_ERR, "MRC: data checksum mismatch: %x vs %x\n",
md->data_checksum, checksum);
return -1;
}
}
mrc_cache: Move mrc_cache_*_hash functions into mrc_cache driver This CL would remove these calls from fsp 2.0. Platforms that select MRC_STASH_TO_CBMEM, updating the TPM NVRAM space is moved from romstage (when data stashed to CBMEM) to ramstage (when data is written back to SPI flash. BUG=b:150502246 BRANCH=None TEST=make sure memory training still works on nami Change-Id: I3088ca6927c7dbc65386c13e868afa0462086937 Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46510 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com> Reviewed-by: Julius Werner <jwerner@chromium.org>
2020-10-16 21:30:05 +02:00
mrc: Add support for separate training cache in recovery mode 1. Re-factor MRC cache driver to properly select RW_MRC_CACHE or RECOVERY_MRC_CACHE based on the boot mode. - If normal mode boot, use RW_MRC_CACHE, if available. - If recovery mode boot: - Retrain memory if RECOVERY_MRC_CACHE not present, or recovery is requested explicity with retrain memory request. - Use RECOVERY_MRC_CACHE otherwise. 2. Protect RW and RECOVERY mrc caches in recovery and non-recovery boot modes. Check if both are present under one unified region and protect that region as a whole. Else try protecting individual regions. 3. Update training data in appropriate cache: - Use RW_MRC_CACHE if normal mode. - Use RECOVERY_MRC_CACHE if present in recovery mode. Else use RW_MRC_CACHE. 4. Add proper debug logs to indicate which training data cache is used at any point. BUG=chrome-os-partner:59352 BRANCH=None TEST=Verified that correct cache is used in both normal and recovery mode on reef. Change-Id: Ie79737a1450bd1ff71543e44a5a3e16950e70fb3 Signed-off-by: Furquan Shaikh <furquan@chromium.org> Reviewed-on: https://review.coreboot.org/17242 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-06 07:57:02 +01:00
return 0;
}
mrc_cache: Add mrc_cache fetch functions to support non-x86 platforms Create two new functions to fetch mrc_cache data (replacing mrc_cache_get_current): - mrc_cache_load_current: fetches the mrc_cache data and drops it into the given buffer. This is useful for ARM platforms where the mmap operation is very expensive. - mrc_cache_mmap_leak: fetch the mrc_cache data and puts it into a given buffer. This is useful for platforms where the mmap operation is a no-op (like x86 platforms). As the name mentions, we are not freeing the memory that we allocated with the mmap, so it is the caller's responsibility to do so. Additionally, we are replacing mrc_cache_latest with mrc_cache_get_latest_slot_info, which does not check the validity of the data when retrieving the current mrc_cache slot. This allows the caller some flexibility in deciding where they want the mrc_cache data stored (either in an mmaped region or at a given address). BUG=b:150502246 BRANCH=None TEST=Testing on a nami (x86) device: reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I259dd4f550719d821bbafa2d445cbae6ea22e988 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44006 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-07-24 01:10:52 +02:00
static int mrc_cache_get_latest_slot_info(const char *name,
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
const struct region_device *backing_rdev,
struct mrc_metadata *md,
struct region_file *cache_file,
struct region_device *rdev,
bool fail_bad_data)
mrc_cache: add version field In order to allow for updateable memory init code on intel x86 platforms one needs to ensure the saved mrc data matches the code consuming the data. To that end add a version field to the saved data structure. BUG=chrome-os-partner:46050 BRANCH=None TEST=Built and booted on glados. Suspended and resumed. Also verified version mismatch path. Change-Id: Ie86db1750af5d9bff6446999b0d04b60612f8d29 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/12700 Tested-by: build bot (Jenkins) Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org> Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2015-12-09 22:04:49 +01:00
{
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
/* Init and obtain a handle to the file data. */
if (region_file_init(cache_file, backing_rdev) < 0) {
printk(BIOS_ERR, "MRC: region file invalid in '%s'\n", name);
soc/intel/common: Add save/restore for variable MRC data Piggy-back on existing MRC cache infrastructure to store variable MRC data. Only one set of data can be valid at given point of time. Currently this magically happens because region alignment is forced to 0x1000 and region itself is of the same size. This needs to be somehow programmatically enforced. Change-Id: I8a660d356ca760b8ff9907396fb9b34cb16cf1db Signed-off-by: Andrey Petrov <andrey.petrov@intel.com> Reviewed-on: https://review.coreboot.org/17320 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-08 17:30:06 +01:00
return -1;
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
}
soc/intel/common: Add save/restore for variable MRC data Piggy-back on existing MRC cache infrastructure to store variable MRC data. Only one set of data can be valid at given point of time. Currently this magically happens because region alignment is forced to 0x1000 and region itself is of the same size. This needs to be somehow programmatically enforced. Change-Id: I8a660d356ca760b8ff9907396fb9b34cb16cf1db Signed-off-by: Andrey Petrov <andrey.petrov@intel.com> Reviewed-on: https://review.coreboot.org/17320 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-08 17:30:06 +01:00
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
/* Provide a 0 sized region_device from here on out so the caller
* has a valid yet unusable region_device. */
rdev_chain(rdev, backing_rdev, 0, 0);
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
/* No data to return. */
if (region_file_data(cache_file, rdev) < 0) {
printk(BIOS_ERR, "MRC: no data in '%s'\n", name);
return fail_bad_data ? -1 : 0;
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
}
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
/* Validate header and resize region to reflect actual usage on the
* saved medium (including metadata and data). */
if (mrc_header_valid(rdev, md) < 0) {
printk(BIOS_ERR, "MRC: invalid header in '%s'\n", name);
return fail_bad_data ? -1 : 0;
}
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
return 0;
}
mrc_cache: Add mrc_cache fetch functions to support non-x86 platforms Create two new functions to fetch mrc_cache data (replacing mrc_cache_get_current): - mrc_cache_load_current: fetches the mrc_cache data and drops it into the given buffer. This is useful for ARM platforms where the mmap operation is very expensive. - mrc_cache_mmap_leak: fetch the mrc_cache data and puts it into a given buffer. This is useful for platforms where the mmap operation is a no-op (like x86 platforms). As the name mentions, we are not freeing the memory that we allocated with the mmap, so it is the caller's responsibility to do so. Additionally, we are replacing mrc_cache_latest with mrc_cache_get_latest_slot_info, which does not check the validity of the data when retrieving the current mrc_cache slot. This allows the caller some flexibility in deciding where they want the mrc_cache data stored (either in an mmaped region or at a given address). BUG=b:150502246 BRANCH=None TEST=Testing on a nami (x86) device: reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I259dd4f550719d821bbafa2d445cbae6ea22e988 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44006 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-07-24 01:10:52 +02:00
static int mrc_cache_find_current(int type, uint32_t version,
struct region_device *rdev,
struct mrc_metadata *md)
soc/intel/common: Add save/restore for variable MRC data Piggy-back on existing MRC cache infrastructure to store variable MRC data. Only one set of data can be valid at given point of time. Currently this magically happens because region alignment is forced to 0x1000 and region itself is of the same size. This needs to be somehow programmatically enforced. Change-Id: I8a660d356ca760b8ff9907396fb9b34cb16cf1db Signed-off-by: Andrey Petrov <andrey.petrov@intel.com> Reviewed-on: https://review.coreboot.org/17320 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-08 17:30:06 +01:00
{
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
const struct cache_region *cr;
struct region region;
struct region_device read_rdev;
struct region_file cache_file;
size_t data_size;
mrc_cache: Add mrc_cache fetch functions to support non-x86 platforms Create two new functions to fetch mrc_cache data (replacing mrc_cache_get_current): - mrc_cache_load_current: fetches the mrc_cache data and drops it into the given buffer. This is useful for ARM platforms where the mmap operation is very expensive. - mrc_cache_mmap_leak: fetch the mrc_cache data and puts it into a given buffer. This is useful for platforms where the mmap operation is a no-op (like x86 platforms). As the name mentions, we are not freeing the memory that we allocated with the mmap, so it is the caller's responsibility to do so. Additionally, we are replacing mrc_cache_latest with mrc_cache_get_latest_slot_info, which does not check the validity of the data when retrieving the current mrc_cache slot. This allows the caller some flexibility in deciding where they want the mrc_cache data stored (either in an mmaped region or at a given address). BUG=b:150502246 BRANCH=None TEST=Testing on a nami (x86) device: reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I259dd4f550719d821bbafa2d445cbae6ea22e988 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44006 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-07-24 01:10:52 +02:00
const size_t md_size = sizeof(*md);
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
const bool fail_bad_data = true;
mrc_cache: Move code for triggering memory training into mrc_cache Currently the decision of whether or not to use mrc_cache in recovery mode is made within the individual platforms' drivers (ie: fsp2.0, fsp1.1, etc.). As this is not platform specific, but uses common vboot infrastructure, the code can be unified and moved into mrc_cache. The conditions are as follows: 1. If HAS_RECOVERY_MRC_CACHE, use mrc_cache data (unless retrain switch is true) 2. If !HAS_RECOVERY_MRC_CACHE && VBOOT_STARTS_IN_BOOTBLOCK, this means that memory training will occur after verified boot, meaning that mrc_cache will be filled with data from executing RW code. So in this case, we never want to use the training data in the mrc_cache for recovery mode. 3. If !HAS_RECOVERY_MRC_CACHE && VBOOT_STARTS_IN_ROMSTAGE, this means that memory training happens before verfied boot, meaning that the mrc_cache data is generated by RO code, so it is safe to use for a recovery boot. 4. Any platform that does not use vboot should be unaffected. Additionally, we have removed the MRC_CLEAR_NORMAL_CACHE_ON_RECOVERY_RETRAIN config because the mrc_cache driver takes care of invalidating the mrc_cache data for normal mode. If the platform: 1. !HAS_RECOVERY_MRC_CACHE, always invalidate mrc_cache data 2. HAS_RECOVERY_MRC_CACHE, only invalidate if retrain switch is set BUG=b:150502246 BRANCH=None TEST=1. run dut-control power_state:rec_force_mrc twice on lazor ensure that memory retraining happens both times run dut-control power_state:rec twice on lazor ensure that memory retraining happens only first time 2. remove HAS_RECOVERY_MRC_CACHE from lazor Kconfig boot twice to ensure caching of memory training occurred on each boot. Change-Id: I3875a7b4a4ba3c1aa8a3c1507b3993036a7155fc Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46855 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-10-27 23:58:31 +01:00
/*
* In recovery mode, force retraining if the memory retrain
* switch is set.
*/
if (vboot_recovery_mode_enabled() && get_recovery_mode_retrain_switch())
return -1;
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
cr = lookup_region(&region, type);
if (cr == NULL)
return -1;
soc/intel/common: Add save/restore for variable MRC data Piggy-back on existing MRC cache infrastructure to store variable MRC data. Only one set of data can be valid at given point of time. Currently this magically happens because region alignment is forced to 0x1000 and region itself is of the same size. This needs to be somehow programmatically enforced. Change-Id: I8a660d356ca760b8ff9907396fb9b34cb16cf1db Signed-off-by: Andrey Petrov <andrey.petrov@intel.com> Reviewed-on: https://review.coreboot.org/17320 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-08 17:30:06 +01:00
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
if (boot_device_ro_subregion(&region, &read_rdev) < 0)
return -1;
soc/intel/common: Add save/restore for variable MRC data Piggy-back on existing MRC cache infrastructure to store variable MRC data. Only one set of data can be valid at given point of time. Currently this magically happens because region alignment is forced to 0x1000 and region itself is of the same size. This needs to be somehow programmatically enforced. Change-Id: I8a660d356ca760b8ff9907396fb9b34cb16cf1db Signed-off-by: Andrey Petrov <andrey.petrov@intel.com> Reviewed-on: https://review.coreboot.org/17320 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-08 17:30:06 +01:00
mrc_cache: Add mrc_cache fetch functions to support non-x86 platforms Create two new functions to fetch mrc_cache data (replacing mrc_cache_get_current): - mrc_cache_load_current: fetches the mrc_cache data and drops it into the given buffer. This is useful for ARM platforms where the mmap operation is very expensive. - mrc_cache_mmap_leak: fetch the mrc_cache data and puts it into a given buffer. This is useful for platforms where the mmap operation is a no-op (like x86 platforms). As the name mentions, we are not freeing the memory that we allocated with the mmap, so it is the caller's responsibility to do so. Additionally, we are replacing mrc_cache_latest with mrc_cache_get_latest_slot_info, which does not check the validity of the data when retrieving the current mrc_cache slot. This allows the caller some flexibility in deciding where they want the mrc_cache data stored (either in an mmaped region or at a given address). BUG=b:150502246 BRANCH=None TEST=Testing on a nami (x86) device: reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I259dd4f550719d821bbafa2d445cbae6ea22e988 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44006 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-07-24 01:10:52 +02:00
if (mrc_cache_get_latest_slot_info(cr->name,
&read_rdev,
md,
&cache_file,
rdev,
fail_bad_data) < 0)
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
return -1;
soc/intel/common: Add save/restore for variable MRC data Piggy-back on existing MRC cache infrastructure to store variable MRC data. Only one set of data can be valid at given point of time. Currently this magically happens because region alignment is forced to 0x1000 and region itself is of the same size. This needs to be somehow programmatically enforced. Change-Id: I8a660d356ca760b8ff9907396fb9b34cb16cf1db Signed-off-by: Andrey Petrov <andrey.petrov@intel.com> Reviewed-on: https://review.coreboot.org/17320 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-08 17:30:06 +01:00
mrc_cache: Add mrc_cache fetch functions to support non-x86 platforms Create two new functions to fetch mrc_cache data (replacing mrc_cache_get_current): - mrc_cache_load_current: fetches the mrc_cache data and drops it into the given buffer. This is useful for ARM platforms where the mmap operation is very expensive. - mrc_cache_mmap_leak: fetch the mrc_cache data and puts it into a given buffer. This is useful for platforms where the mmap operation is a no-op (like x86 platforms). As the name mentions, we are not freeing the memory that we allocated with the mmap, so it is the caller's responsibility to do so. Additionally, we are replacing mrc_cache_latest with mrc_cache_get_latest_slot_info, which does not check the validity of the data when retrieving the current mrc_cache slot. This allows the caller some flexibility in deciding where they want the mrc_cache data stored (either in an mmaped region or at a given address). BUG=b:150502246 BRANCH=None TEST=Testing on a nami (x86) device: reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I259dd4f550719d821bbafa2d445cbae6ea22e988 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44006 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-07-24 01:10:52 +02:00
if (version != md->version) {
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
printk(BIOS_INFO, "MRC: version mismatch: %x vs %x\n",
mrc_cache: Add mrc_cache fetch functions to support non-x86 platforms Create two new functions to fetch mrc_cache data (replacing mrc_cache_get_current): - mrc_cache_load_current: fetches the mrc_cache data and drops it into the given buffer. This is useful for ARM platforms where the mmap operation is very expensive. - mrc_cache_mmap_leak: fetch the mrc_cache data and puts it into a given buffer. This is useful for platforms where the mmap operation is a no-op (like x86 platforms). As the name mentions, we are not freeing the memory that we allocated with the mmap, so it is the caller's responsibility to do so. Additionally, we are replacing mrc_cache_latest with mrc_cache_get_latest_slot_info, which does not check the validity of the data when retrieving the current mrc_cache slot. This allows the caller some flexibility in deciding where they want the mrc_cache data stored (either in an mmaped region or at a given address). BUG=b:150502246 BRANCH=None TEST=Testing on a nami (x86) device: reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I259dd4f550719d821bbafa2d445cbae6ea22e988 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44006 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-07-24 01:10:52 +02:00
md->version, version);
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
return -1;
}
/* Re-size rdev to only contain the data. i.e. remove metadata. */
mrc_cache: Add mrc_cache fetch functions to support non-x86 platforms Create two new functions to fetch mrc_cache data (replacing mrc_cache_get_current): - mrc_cache_load_current: fetches the mrc_cache data and drops it into the given buffer. This is useful for ARM platforms where the mmap operation is very expensive. - mrc_cache_mmap_leak: fetch the mrc_cache data and puts it into a given buffer. This is useful for platforms where the mmap operation is a no-op (like x86 platforms). As the name mentions, we are not freeing the memory that we allocated with the mmap, so it is the caller's responsibility to do so. Additionally, we are replacing mrc_cache_latest with mrc_cache_get_latest_slot_info, which does not check the validity of the data when retrieving the current mrc_cache slot. This allows the caller some flexibility in deciding where they want the mrc_cache data stored (either in an mmaped region or at a given address). BUG=b:150502246 BRANCH=None TEST=Testing on a nami (x86) device: reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I259dd4f550719d821bbafa2d445cbae6ea22e988 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44006 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-07-24 01:10:52 +02:00
data_size = md->data_size;
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
return rdev_chain(rdev, rdev, md_size, data_size);
mrc_cache: add version field In order to allow for updateable memory init code on intel x86 platforms one needs to ensure the saved mrc data matches the code consuming the data. To that end add a version field to the saved data structure. BUG=chrome-os-partner:46050 BRANCH=None TEST=Built and booted on glados. Suspended and resumed. Also verified version mismatch path. Change-Id: Ie86db1750af5d9bff6446999b0d04b60612f8d29 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/12700 Tested-by: build bot (Jenkins) Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org> Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2015-12-09 22:04:49 +01:00
}
mrc_cache: Change mrc_cache_load_current to return size of entry Modify mrc_cache_load current to return the size of the mrc_cache entry so that caller will know what the actual size of the data returned is. This is needed for ARM devices like trogdor, which need to know the size of the training data when populating the QcLib interface table. BUG=b:150502246 BRANCH=None TEST=./util/abuild/abuild -p none -t GOOGLE_NAMI -x -a Change-Id: Ia314717ad2a7d5232b37a19951c1aecd7f843c27 Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46110 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2020-10-07 00:14:24 +02:00
ssize_t mrc_cache_load_current(int type, uint32_t version, void *buffer,
size_t buffer_size)
mrc_cache: Add mrc_cache fetch functions to support non-x86 platforms Create two new functions to fetch mrc_cache data (replacing mrc_cache_get_current): - mrc_cache_load_current: fetches the mrc_cache data and drops it into the given buffer. This is useful for ARM platforms where the mmap operation is very expensive. - mrc_cache_mmap_leak: fetch the mrc_cache data and puts it into a given buffer. This is useful for platforms where the mmap operation is a no-op (like x86 platforms). As the name mentions, we are not freeing the memory that we allocated with the mmap, so it is the caller's responsibility to do so. Additionally, we are replacing mrc_cache_latest with mrc_cache_get_latest_slot_info, which does not check the validity of the data when retrieving the current mrc_cache slot. This allows the caller some flexibility in deciding where they want the mrc_cache data stored (either in an mmaped region or at a given address). BUG=b:150502246 BRANCH=None TEST=Testing on a nami (x86) device: reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I259dd4f550719d821bbafa2d445cbae6ea22e988 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44006 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-07-24 01:10:52 +02:00
{
struct region_device rdev;
struct mrc_metadata md;
mrc_cache: Change mrc_cache_load_current to return size of entry Modify mrc_cache_load current to return the size of the mrc_cache entry so that caller will know what the actual size of the data returned is. This is needed for ARM devices like trogdor, which need to know the size of the training data when populating the QcLib interface table. BUG=b:150502246 BRANCH=None TEST=./util/abuild/abuild -p none -t GOOGLE_NAMI -x -a Change-Id: Ia314717ad2a7d5232b37a19951c1aecd7f843c27 Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46110 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2020-10-07 00:14:24 +02:00
ssize_t data_size;
mrc_cache: Add mrc_cache fetch functions to support non-x86 platforms Create two new functions to fetch mrc_cache data (replacing mrc_cache_get_current): - mrc_cache_load_current: fetches the mrc_cache data and drops it into the given buffer. This is useful for ARM platforms where the mmap operation is very expensive. - mrc_cache_mmap_leak: fetch the mrc_cache data and puts it into a given buffer. This is useful for platforms where the mmap operation is a no-op (like x86 platforms). As the name mentions, we are not freeing the memory that we allocated with the mmap, so it is the caller's responsibility to do so. Additionally, we are replacing mrc_cache_latest with mrc_cache_get_latest_slot_info, which does not check the validity of the data when retrieving the current mrc_cache slot. This allows the caller some flexibility in deciding where they want the mrc_cache data stored (either in an mmaped region or at a given address). BUG=b:150502246 BRANCH=None TEST=Testing on a nami (x86) device: reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I259dd4f550719d821bbafa2d445cbae6ea22e988 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44006 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-07-24 01:10:52 +02:00
if (mrc_cache_find_current(type, version, &rdev, &md) < 0)
return -1;
data_size = region_device_sz(&rdev);
if (buffer_size < data_size)
return -1;
if (rdev_readat(&rdev, buffer, 0, data_size) != data_size)
return -1;
mrc_cache: Add tpm_hash_index field to cache_region struct Pull selection of tpm hash index logic into cache_region struct. This CL also enables the storing of the MRC hash into the TPM NVRAM space for both recovery and non-recovery cases. This will affect all platforms with TPM2 enabled and use the MRC_CACHE driver. BUG=b:150502246 BRANCH=None TEST=make sure memory training still works on nami and lazor Change-Id: I1a744d6f40f062ca3aab6157b3747e6c1f6977f9 Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46514 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2020-10-20 19:42:26 +02:00
if (mrc_data_valid(type, &md, buffer, data_size) < 0)
mrc_cache: Add mrc_cache fetch functions to support non-x86 platforms Create two new functions to fetch mrc_cache data (replacing mrc_cache_get_current): - mrc_cache_load_current: fetches the mrc_cache data and drops it into the given buffer. This is useful for ARM platforms where the mmap operation is very expensive. - mrc_cache_mmap_leak: fetch the mrc_cache data and puts it into a given buffer. This is useful for platforms where the mmap operation is a no-op (like x86 platforms). As the name mentions, we are not freeing the memory that we allocated with the mmap, so it is the caller's responsibility to do so. Additionally, we are replacing mrc_cache_latest with mrc_cache_get_latest_slot_info, which does not check the validity of the data when retrieving the current mrc_cache slot. This allows the caller some flexibility in deciding where they want the mrc_cache data stored (either in an mmaped region or at a given address). BUG=b:150502246 BRANCH=None TEST=Testing on a nami (x86) device: reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I259dd4f550719d821bbafa2d445cbae6ea22e988 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44006 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-07-24 01:10:52 +02:00
return -1;
mrc_cache: Change mrc_cache_load_current to return size of entry Modify mrc_cache_load current to return the size of the mrc_cache entry so that caller will know what the actual size of the data returned is. This is needed for ARM devices like trogdor, which need to know the size of the training data when populating the QcLib interface table. BUG=b:150502246 BRANCH=None TEST=./util/abuild/abuild -p none -t GOOGLE_NAMI -x -a Change-Id: Ia314717ad2a7d5232b37a19951c1aecd7f843c27 Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46110 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2020-10-07 00:14:24 +02:00
return data_size;
mrc_cache: Add mrc_cache fetch functions to support non-x86 platforms Create two new functions to fetch mrc_cache data (replacing mrc_cache_get_current): - mrc_cache_load_current: fetches the mrc_cache data and drops it into the given buffer. This is useful for ARM platforms where the mmap operation is very expensive. - mrc_cache_mmap_leak: fetch the mrc_cache data and puts it into a given buffer. This is useful for platforms where the mmap operation is a no-op (like x86 platforms). As the name mentions, we are not freeing the memory that we allocated with the mmap, so it is the caller's responsibility to do so. Additionally, we are replacing mrc_cache_latest with mrc_cache_get_latest_slot_info, which does not check the validity of the data when retrieving the current mrc_cache slot. This allows the caller some flexibility in deciding where they want the mrc_cache data stored (either in an mmaped region or at a given address). BUG=b:150502246 BRANCH=None TEST=Testing on a nami (x86) device: reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I259dd4f550719d821bbafa2d445cbae6ea22e988 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44006 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-07-24 01:10:52 +02:00
}
void *mrc_cache_current_mmap_leak(int type, uint32_t version,
size_t *data_size)
{
struct region_device rdev;
void *data;
size_t region_device_size;
struct mrc_metadata md;
if (mrc_cache_find_current(type, version, &rdev, &md) < 0)
return NULL;
region_device_size = region_device_sz(&rdev);
if (data_size)
*data_size = region_device_size;
data = rdev_mmap_full(&rdev);
if (data == NULL) {
printk(BIOS_INFO, "MRC: mmap failure.\n");
return NULL;
}
mrc_cache: Add tpm_hash_index field to cache_region struct Pull selection of tpm hash index logic into cache_region struct. This CL also enables the storing of the MRC hash into the TPM NVRAM space for both recovery and non-recovery cases. This will affect all platforms with TPM2 enabled and use the MRC_CACHE driver. BUG=b:150502246 BRANCH=None TEST=make sure memory training still works on nami and lazor Change-Id: I1a744d6f40f062ca3aab6157b3747e6c1f6977f9 Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46514 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2020-10-20 19:42:26 +02:00
if (mrc_data_valid(type, &md, data, region_device_size) < 0)
mrc_cache: Add mrc_cache fetch functions to support non-x86 platforms Create two new functions to fetch mrc_cache data (replacing mrc_cache_get_current): - mrc_cache_load_current: fetches the mrc_cache data and drops it into the given buffer. This is useful for ARM platforms where the mmap operation is very expensive. - mrc_cache_mmap_leak: fetch the mrc_cache data and puts it into a given buffer. This is useful for platforms where the mmap operation is a no-op (like x86 platforms). As the name mentions, we are not freeing the memory that we allocated with the mmap, so it is the caller's responsibility to do so. Additionally, we are replacing mrc_cache_latest with mrc_cache_get_latest_slot_info, which does not check the validity of the data when retrieving the current mrc_cache slot. This allows the caller some flexibility in deciding where they want the mrc_cache data stored (either in an mmaped region or at a given address). BUG=b:150502246 BRANCH=None TEST=Testing on a nami (x86) device: reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I259dd4f550719d821bbafa2d445cbae6ea22e988 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44006 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-07-24 01:10:52 +02:00
return NULL;
return data;
}
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
static bool mrc_cache_needs_update(const struct region_device *rdev,
mrc_cache: Update mrc_cache data in romstage Previously, we were writing to cbmem after memory training and then writing the training data from cbmem to mrc_cache in ramstage. We were doing this because we were unable to read/write to SPI simultaneously on older x86 chips. Now that newer chips allow for simultaneously reads and writes, we can move the mrc_cache update into romstage. This is beneficial if there is a reboot for some reason after memory training but before the previous mrc_cache_stash_data call originally in ramstage. If this happens, we would lose all the mrc_cache training data in the next boot even though we've already performed the memory training. Added new config BOOT_DEVICE_SPI_FLASH_NO_EARLY_WRITES to accomodate older x86 platforms that don't do mmapping but still want to use the cbmem to store the mrc_cache data in order to write the mrc_cache data back at a later time. We are maintaining the use of cbmem for these older platforms because we have no way of validating the earlier write back to mrc_cache at this time. BUG=b:150502246 BRANCH=None TEST=reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I3430bda45484cb8c2b01ab9614508039dfaac9a3 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44196 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-05-02 02:00:31 +02:00
const struct mrc_metadata *new_md,
const void *new_data, size_t new_data_size)
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
{
mrc_cache: Update mrc_cache data in romstage Previously, we were writing to cbmem after memory training and then writing the training data from cbmem to mrc_cache in ramstage. We were doing this because we were unable to read/write to SPI simultaneously on older x86 chips. Now that newer chips allow for simultaneously reads and writes, we can move the mrc_cache update into romstage. This is beneficial if there is a reboot for some reason after memory training but before the previous mrc_cache_stash_data call originally in ramstage. If this happens, we would lose all the mrc_cache training data in the next boot even though we've already performed the memory training. Added new config BOOT_DEVICE_SPI_FLASH_NO_EARLY_WRITES to accomodate older x86 platforms that don't do mmapping but still want to use the cbmem to store the mrc_cache data in order to write the mrc_cache data back at a later time. We are maintaining the use of cbmem for these older platforms because we have no way of validating the earlier write back to mrc_cache at this time. BUG=b:150502246 BRANCH=None TEST=reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I3430bda45484cb8c2b01ab9614508039dfaac9a3 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44196 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-05-02 02:00:31 +02:00
void *mapping, *data_mapping;
drivers/mrc_cache: Fix size comparison in mrc_cache update `mrc_cache_needs_update` is comparing the "new size" of the MRC data (minus metadata size) to the size including the metadata, which causes the driver to think the data has changed, and so it will rewrite the MRC cache on every boot. This patch removes the metadata size from the comparison. BUG=b:171513942 BRANCH=volteer TEST=1) Memory training data gets written the on a boot where the data was wiped out. 2) Memory training data does not get written back on every subsequent boot. Signed-off-by: Tim Wawrzynczak <twawrzynczak@chromium.org> Change-Id: I7280276f71fdaa492c327b2b7ade8e53e7c59f51 Reviewed-on: https://review.coreboot.org/c/coreboot/+/46824 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Shelley Chen <shchen@google.com> Reviewed-by: Furquan Shaikh <furquan@google.com>
2020-10-26 18:40:41 +01:00
size_t old_data_size = region_device_sz(rdev) - sizeof(struct mrc_metadata);
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
bool need_update = false;
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
drivers/mrc_cache: Fix size comparison in mrc_cache update `mrc_cache_needs_update` is comparing the "new size" of the MRC data (minus metadata size) to the size including the metadata, which causes the driver to think the data has changed, and so it will rewrite the MRC cache on every boot. This patch removes the metadata size from the comparison. BUG=b:171513942 BRANCH=volteer TEST=1) Memory training data gets written the on a boot where the data was wiped out. 2) Memory training data does not get written back on every subsequent boot. Signed-off-by: Tim Wawrzynczak <twawrzynczak@chromium.org> Change-Id: I7280276f71fdaa492c327b2b7ade8e53e7c59f51 Reviewed-on: https://review.coreboot.org/c/coreboot/+/46824 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Shelley Chen <shchen@google.com> Reviewed-by: Furquan Shaikh <furquan@google.com>
2020-10-26 18:40:41 +01:00
if (new_data_size != old_data_size)
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
return true;
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
mapping = rdev_mmap_full(rdev);
mrc_cache: Update mrc_cache data in romstage Previously, we were writing to cbmem after memory training and then writing the training data from cbmem to mrc_cache in ramstage. We were doing this because we were unable to read/write to SPI simultaneously on older x86 chips. Now that newer chips allow for simultaneously reads and writes, we can move the mrc_cache update into romstage. This is beneficial if there is a reboot for some reason after memory training but before the previous mrc_cache_stash_data call originally in ramstage. If this happens, we would lose all the mrc_cache training data in the next boot even though we've already performed the memory training. Added new config BOOT_DEVICE_SPI_FLASH_NO_EARLY_WRITES to accomodate older x86 platforms that don't do mmapping but still want to use the cbmem to store the mrc_cache data in order to write the mrc_cache data back at a later time. We are maintaining the use of cbmem for these older platforms because we have no way of validating the earlier write back to mrc_cache at this time. BUG=b:150502246 BRANCH=None TEST=reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I3430bda45484cb8c2b01ab9614508039dfaac9a3 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44196 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-05-02 02:00:31 +02:00
if (mapping == NULL) {
printk(BIOS_ERR, "MRC: cannot mmap existing cache.\n");
return true;
}
data_mapping = mapping + sizeof(struct mrc_metadata);
/* we need to compare the md and the data separately */
/* check the mrc_metadata */
if (memcmp(new_md, mapping, sizeof(struct mrc_metadata)))
need_update = true;
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
mrc_cache: Update mrc_cache data in romstage Previously, we were writing to cbmem after memory training and then writing the training data from cbmem to mrc_cache in ramstage. We were doing this because we were unable to read/write to SPI simultaneously on older x86 chips. Now that newer chips allow for simultaneously reads and writes, we can move the mrc_cache update into romstage. This is beneficial if there is a reboot for some reason after memory training but before the previous mrc_cache_stash_data call originally in ramstage. If this happens, we would lose all the mrc_cache training data in the next boot even though we've already performed the memory training. Added new config BOOT_DEVICE_SPI_FLASH_NO_EARLY_WRITES to accomodate older x86 platforms that don't do mmapping but still want to use the cbmem to store the mrc_cache data in order to write the mrc_cache data back at a later time. We are maintaining the use of cbmem for these older platforms because we have no way of validating the earlier write back to mrc_cache at this time. BUG=b:150502246 BRANCH=None TEST=reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I3430bda45484cb8c2b01ab9614508039dfaac9a3 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44196 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-05-02 02:00:31 +02:00
/* check the data */
if (!need_update && memcmp(new_data, data_mapping, new_data_size))
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
need_update = true;
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
rdev_munmap(rdev, mapping);
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
return need_update;
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
}
soc/intel/common: Add save/restore for variable MRC data Piggy-back on existing MRC cache infrastructure to store variable MRC data. Only one set of data can be valid at given point of time. Currently this magically happens because region alignment is forced to 0x1000 and region itself is of the same size. This needs to be somehow programmatically enforced. Change-Id: I8a660d356ca760b8ff9907396fb9b34cb16cf1db Signed-off-by: Andrey Petrov <andrey.petrov@intel.com> Reviewed-on: https://review.coreboot.org/17320 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-08 17:30:06 +01:00
static void log_event_cache_update(uint8_t slot, enum result res)
soc/intel/common: log event when MRC cache is updated Log when the MRC cache is attempted to be updated with status of success or failure. Just one slot is supported currently which is deemed 'normal'. This is because there are more slots anticipated in the future. BUG=chrome-os-partner:59395 Change-Id: I0f81458325697aff9924cc359a4173e0d35da5da Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17231 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins)
2016-11-04 16:07:14 +01:00
{
const int type = ELOG_TYPE_MEM_CACHE_UPDATE;
struct elog_event_mem_cache_update event = {
soc/intel/common: Add save/restore for variable MRC data Piggy-back on existing MRC cache infrastructure to store variable MRC data. Only one set of data can be valid at given point of time. Currently this magically happens because region alignment is forced to 0x1000 and region itself is of the same size. This needs to be somehow programmatically enforced. Change-Id: I8a660d356ca760b8ff9907396fb9b34cb16cf1db Signed-off-by: Andrey Petrov <andrey.petrov@intel.com> Reviewed-on: https://review.coreboot.org/17320 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-08 17:30:06 +01:00
.slot = slot
soc/intel/common: log event when MRC cache is updated Log when the MRC cache is attempted to be updated with status of success or failure. Just one slot is supported currently which is deemed 'normal'. This is because there are more slots anticipated in the future. BUG=chrome-os-partner:59395 Change-Id: I0f81458325697aff9924cc359a4173e0d35da5da Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17231 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins)
2016-11-04 16:07:14 +01:00
};
soc/intel/common: Add save/restore for variable MRC data Piggy-back on existing MRC cache infrastructure to store variable MRC data. Only one set of data can be valid at given point of time. Currently this magically happens because region alignment is forced to 0x1000 and region itself is of the same size. This needs to be somehow programmatically enforced. Change-Id: I8a660d356ca760b8ff9907396fb9b34cb16cf1db Signed-off-by: Andrey Petrov <andrey.petrov@intel.com> Reviewed-on: https://review.coreboot.org/17320 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-08 17:30:06 +01:00
/* Filter through interesting events only */
switch (res) {
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
case UPDATE_FAILURE:
soc/intel/common: Add save/restore for variable MRC data Piggy-back on existing MRC cache infrastructure to store variable MRC data. Only one set of data can be valid at given point of time. Currently this magically happens because region alignment is forced to 0x1000 and region itself is of the same size. This needs to be somehow programmatically enforced. Change-Id: I8a660d356ca760b8ff9907396fb9b34cb16cf1db Signed-off-by: Andrey Petrov <andrey.petrov@intel.com> Reviewed-on: https://review.coreboot.org/17320 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-08 17:30:06 +01:00
event.status = ELOG_MEM_CACHE_UPDATE_STATUS_FAIL;
break;
case UPDATE_SUCCESS:
event.status = ELOG_MEM_CACHE_UPDATE_STATUS_SUCCESS;
break;
default:
return;
}
soc/intel/common: log event when MRC cache is updated Log when the MRC cache is attempted to be updated with status of success or failure. Just one slot is supported currently which is deemed 'normal'. This is because there are more slots anticipated in the future. BUG=chrome-os-partner:59395 Change-Id: I0f81458325697aff9924cc359a4173e0d35da5da Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17231 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins)
2016-11-04 16:07:14 +01:00
if (elog_add_event_raw(type, &event, sizeof(event)) < 0)
printk(BIOS_ERR, "Failed to log mem cache update event.\n");
}
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
/* During ramstage this code purposefully uses incoherent transactions between
* read and write. The read assumes a memory-mapped boot device that can be used
* to quickly locate and compare the up-to-date data. However, when an update
* is required it uses the writeable region access to perform the update. */
mrc_cache: Update mrc_cache data in romstage Previously, we were writing to cbmem after memory training and then writing the training data from cbmem to mrc_cache in ramstage. We were doing this because we were unable to read/write to SPI simultaneously on older x86 chips. Now that newer chips allow for simultaneously reads and writes, we can move the mrc_cache update into romstage. This is beneficial if there is a reboot for some reason after memory training but before the previous mrc_cache_stash_data call originally in ramstage. If this happens, we would lose all the mrc_cache training data in the next boot even though we've already performed the memory training. Added new config BOOT_DEVICE_SPI_FLASH_NO_EARLY_WRITES to accomodate older x86 platforms that don't do mmapping but still want to use the cbmem to store the mrc_cache data in order to write the mrc_cache data back at a later time. We are maintaining the use of cbmem for these older platforms because we have no way of validating the earlier write back to mrc_cache at this time. BUG=b:150502246 BRANCH=None TEST=reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I3430bda45484cb8c2b01ab9614508039dfaac9a3 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44196 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-05-02 02:00:31 +02:00
static void update_mrc_cache_by_type(int type,
struct mrc_metadata *new_md,
const void *new_data,
size_t new_data_size)
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
{
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
const struct cache_region *cr;
struct region region;
struct region_device read_rdev;
struct region_device write_rdev;
struct region_file cache_file;
struct mrc_metadata md;
struct incoherent_rdev backing_irdev;
const struct region_device *backing_rdev;
struct region_device latest_rdev;
const bool fail_bad_data = false;
mrc_cache: Add tpm_hash_index field to cache_region struct Pull selection of tpm hash index logic into cache_region struct. This CL also enables the storing of the MRC hash into the TPM NVRAM space for both recovery and non-recovery cases. This will affect all platforms with TPM2 enabled and use the MRC_CACHE driver. BUG=b:150502246 BRANCH=None TEST=make sure memory training still works on nami and lazor Change-Id: I1a744d6f40f062ca3aab6157b3747e6c1f6977f9 Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46514 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2020-10-20 19:42:26 +02:00
uint32_t hash_idx;
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
cr = lookup_region(&region, type);
if (cr == NULL)
return;
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
printk(BIOS_DEBUG, "MRC: Checking cached data update for '%s'.\n",
cr->name);
mrc: Add support for separate training cache in recovery mode 1. Re-factor MRC cache driver to properly select RW_MRC_CACHE or RECOVERY_MRC_CACHE based on the boot mode. - If normal mode boot, use RW_MRC_CACHE, if available. - If recovery mode boot: - Retrain memory if RECOVERY_MRC_CACHE not present, or recovery is requested explicity with retrain memory request. - Use RECOVERY_MRC_CACHE otherwise. 2. Protect RW and RECOVERY mrc caches in recovery and non-recovery boot modes. Check if both are present under one unified region and protect that region as a whole. Else try protecting individual regions. 3. Update training data in appropriate cache: - Use RW_MRC_CACHE if normal mode. - Use RECOVERY_MRC_CACHE if present in recovery mode. Else use RW_MRC_CACHE. 4. Add proper debug logs to indicate which training data cache is used at any point. BUG=chrome-os-partner:59352 BRANCH=None TEST=Verified that correct cache is used in both normal and recovery mode on reef. Change-Id: Ie79737a1450bd1ff71543e44a5a3e16950e70fb3 Signed-off-by: Furquan Shaikh <furquan@chromium.org> Reviewed-on: https://review.coreboot.org/17242 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-06 07:57:02 +01:00
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
if (boot_device_ro_subregion(&region, &read_rdev) < 0)
return;
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
if (boot_device_rw_subregion(&region, &write_rdev) < 0)
return;
backing_rdev = incoherent_rdev_init(&backing_irdev, &region, &read_rdev,
&write_rdev);
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
if (backing_rdev == NULL)
return;
mrc_cache: Add mrc_cache fetch functions to support non-x86 platforms Create two new functions to fetch mrc_cache data (replacing mrc_cache_get_current): - mrc_cache_load_current: fetches the mrc_cache data and drops it into the given buffer. This is useful for ARM platforms where the mmap operation is very expensive. - mrc_cache_mmap_leak: fetch the mrc_cache data and puts it into a given buffer. This is useful for platforms where the mmap operation is a no-op (like x86 platforms). As the name mentions, we are not freeing the memory that we allocated with the mmap, so it is the caller's responsibility to do so. Additionally, we are replacing mrc_cache_latest with mrc_cache_get_latest_slot_info, which does not check the validity of the data when retrieving the current mrc_cache slot. This allows the caller some flexibility in deciding where they want the mrc_cache data stored (either in an mmaped region or at a given address). BUG=b:150502246 BRANCH=None TEST=Testing on a nami (x86) device: reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I259dd4f550719d821bbafa2d445cbae6ea22e988 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44006 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-07-24 01:10:52 +02:00
/* Note that mrc_cache_get_latest_slot_info doesn't check the
* validity of the current slot. If the slot is invalid,
* we'll overwrite it anyway when we update the mrc_cache.
*/
if (mrc_cache_get_latest_slot_info(cr->name,
backing_rdev,
&md,
&cache_file,
&latest_rdev,
fail_bad_data) < 0)
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
return;
mrc_cache: Update mrc_cache data in romstage Previously, we were writing to cbmem after memory training and then writing the training data from cbmem to mrc_cache in ramstage. We were doing this because we were unable to read/write to SPI simultaneously on older x86 chips. Now that newer chips allow for simultaneously reads and writes, we can move the mrc_cache update into romstage. This is beneficial if there is a reboot for some reason after memory training but before the previous mrc_cache_stash_data call originally in ramstage. If this happens, we would lose all the mrc_cache training data in the next boot even though we've already performed the memory training. Added new config BOOT_DEVICE_SPI_FLASH_NO_EARLY_WRITES to accomodate older x86 platforms that don't do mmapping but still want to use the cbmem to store the mrc_cache data in order to write the mrc_cache data back at a later time. We are maintaining the use of cbmem for these older platforms because we have no way of validating the earlier write back to mrc_cache at this time. BUG=b:150502246 BRANCH=None TEST=reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I3430bda45484cb8c2b01ab9614508039dfaac9a3 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44196 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-05-02 02:00:31 +02:00
if (!mrc_cache_needs_update(&latest_rdev,
new_md, new_data, new_data_size)) {
drivers/mrc_cache: generate debug messages for MRC cache update When MRC cached data update is performed, messages are written to event log, which is flash based. For system that does not have flash based event log, the messages are lost. Added corresponding BIOS_DEBUG messages. Signed-off-by: Jonathan Zhang <jonzhang@fb.com> Change-Id: I1ef4794151fea7213c8317ddc898b0e37da280b5 Reviewed-on: https://review.coreboot.org/c/coreboot/+/41981 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-by: Michael Niewöhner Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-06-02 00:39:59 +02:00
printk(BIOS_DEBUG, "MRC: '%s' does not need update.\n", cr->name);
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
log_event_cache_update(cr->elog_slot, ALREADY_UPTODATE);
return;
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
}
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
printk(BIOS_DEBUG, "MRC: cache data '%s' needs update.\n", cr->name);
mrc_cache: Update mrc_cache data in romstage Previously, we were writing to cbmem after memory training and then writing the training data from cbmem to mrc_cache in ramstage. We were doing this because we were unable to read/write to SPI simultaneously on older x86 chips. Now that newer chips allow for simultaneously reads and writes, we can move the mrc_cache update into romstage. This is beneficial if there is a reboot for some reason after memory training but before the previous mrc_cache_stash_data call originally in ramstage. If this happens, we would lose all the mrc_cache training data in the next boot even though we've already performed the memory training. Added new config BOOT_DEVICE_SPI_FLASH_NO_EARLY_WRITES to accomodate older x86 platforms that don't do mmapping but still want to use the cbmem to store the mrc_cache data in order to write the mrc_cache data back at a later time. We are maintaining the use of cbmem for these older platforms because we have no way of validating the earlier write back to mrc_cache at this time. BUG=b:150502246 BRANCH=None TEST=reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I3430bda45484cb8c2b01ab9614508039dfaac9a3 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44196 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-05-02 02:00:31 +02:00
struct update_region_file_entry entries[] = {
[0] = {
.size = sizeof(struct mrc_metadata),
.data = new_md,
},
[1] = {
.size = new_data_size,
.data = new_data,
},
};
if (region_file_update_data_arr(&cache_file, entries, ARRAY_SIZE(entries)) < 0) {
printk(BIOS_ERR, "MRC: failed to update '%s'.\n", cr->name);
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
log_event_cache_update(cr->elog_slot, UPDATE_FAILURE);
drivers/mrc_cache: generate debug messages for MRC cache update When MRC cached data update is performed, messages are written to event log, which is flash based. For system that does not have flash based event log, the messages are lost. Added corresponding BIOS_DEBUG messages. Signed-off-by: Jonathan Zhang <jonzhang@fb.com> Change-Id: I1ef4794151fea7213c8317ddc898b0e37da280b5 Reviewed-on: https://review.coreboot.org/c/coreboot/+/41981 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-by: Michael Niewöhner Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-06-02 00:39:59 +02:00
} else {
printk(BIOS_DEBUG, "MRC: updated '%s'.\n", cr->name);
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
log_event_cache_update(cr->elog_slot, UPDATE_SUCCESS);
mrc_cache: Add tpm_hash_index field to cache_region struct Pull selection of tpm hash index logic into cache_region struct. This CL also enables the storing of the MRC hash into the TPM NVRAM space for both recovery and non-recovery cases. This will affect all platforms with TPM2 enabled and use the MRC_CACHE driver. BUG=b:150502246 BRANCH=None TEST=make sure memory training still works on nami and lazor Change-Id: I1a744d6f40f062ca3aab6157b3747e6c1f6977f9 Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46514 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2020-10-20 19:42:26 +02:00
hash_idx = cr->tpm_hash_index;
if (hash_idx && CONFIG(MRC_SAVE_HASH_IN_TPM))
security/vboot: Make mrc_cache hash functions generic We need to extend the functionality of the mrc_cache hash functions to work for both recovery and normal mrc_cache data. Updating the API of these functions to pass in an index to identify the hash indices for recovery and normal mode. BUG=b:150502246 BRANCH=None TEST=make sure memory training still works on nami Change-Id: I9c0bb25eafc731ca9c7a95113ab940f55997fc0f Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46432 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2020-10-16 22:15:59 +02:00
mrc_cache_update_hash(hash_idx, new_data, new_data_size);
drivers/mrc_cache: generate debug messages for MRC cache update When MRC cached data update is performed, messages are written to event log, which is flash based. For system that does not have flash based event log, the messages are lost. Added corresponding BIOS_DEBUG messages. Signed-off-by: Jonathan Zhang <jonzhang@fb.com> Change-Id: I1ef4794151fea7213c8317ddc898b0e37da280b5 Reviewed-on: https://review.coreboot.org/c/coreboot/+/41981 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-by: Michael Niewöhner Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-06-02 00:39:59 +02:00
}
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
}
soc/intel/common: remove nvm headers and c file There's no sense in having the nvm abstraction in its own files. Put that support directly into mrc_cache.c. BUG=b:69614064 Change-Id: I0f1a801c6e1a8c35f70faf9e4318bdc45955047a Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/22883 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2017-12-14 21:53:14 +01:00
/* Read flash status register to determine if write protect is active */
static int nvm_is_write_protected(void)
{
u8 sr1;
u8 wp_gpio;
u8 wp_spi;
coreboot: Replace all IS_ENABLED(CONFIG_XXX) with CONFIG(XXX) This patch is a raw application of find src/ -type f | xargs sed -i -e 's/IS_ENABLED\s*(CONFIG_/CONFIG(/g' Change-Id: I6262d6d5c23cabe23c242b4f38d446b74fe16b88 Signed-off-by: Julius Werner <jwerner@chromium.org> Reviewed-on: https://review.coreboot.org/c/coreboot/+/31774 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2019-03-06 01:53:33 +01:00
if (!CONFIG(CHROMEOS))
soc/intel/common: remove nvm headers and c file There's no sense in having the nvm abstraction in its own files. Put that support directly into mrc_cache.c. BUG=b:69614064 Change-Id: I0f1a801c6e1a8c35f70faf9e4318bdc45955047a Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/22883 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2017-12-14 21:53:14 +01:00
return 0;
coreboot: Replace all IS_ENABLED(CONFIG_XXX) with CONFIG(XXX) This patch is a raw application of find src/ -type f | xargs sed -i -e 's/IS_ENABLED\s*(CONFIG_/CONFIG(/g' Change-Id: I6262d6d5c23cabe23c242b4f38d446b74fe16b88 Signed-off-by: Julius Werner <jwerner@chromium.org> Reviewed-on: https://review.coreboot.org/c/coreboot/+/31774 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2019-03-06 01:53:33 +01:00
if (!CONFIG(BOOT_DEVICE_SPI_FLASH))
soc/intel/common: remove nvm headers and c file There's no sense in having the nvm abstraction in its own files. Put that support directly into mrc_cache.c. BUG=b:69614064 Change-Id: I0f1a801c6e1a8c35f70faf9e4318bdc45955047a Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/22883 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2017-12-14 21:53:14 +01:00
return 0;
/* Read Write Protect GPIO if available */
wp_gpio = get_write_protect_state();
/* Read Status Register 1 */
if (spi_flash_status(boot_device_spi_flash(), &sr1) < 0) {
printk(BIOS_ERR, "Failed to read SPI status register 1\n");
return -1;
}
wp_spi = !!(sr1 & 0x80);
printk(BIOS_DEBUG, "SPI flash protection: WPSW=%d SRP0=%d\n",
wp_gpio, wp_spi);
return wp_gpio && wp_spi;
}
/* Apply protection to a range of flash */
static int nvm_protect(const struct region *r)
{
coreboot: Replace all IS_ENABLED(CONFIG_XXX) with CONFIG(XXX) This patch is a raw application of find src/ -type f | xargs sed -i -e 's/IS_ENABLED\s*(CONFIG_/CONFIG(/g' Change-Id: I6262d6d5c23cabe23c242b4f38d446b74fe16b88 Signed-off-by: Julius Werner <jwerner@chromium.org> Reviewed-on: https://review.coreboot.org/c/coreboot/+/31774 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2019-03-06 01:53:33 +01:00
if (!CONFIG(MRC_SETTINGS_PROTECT))
soc/intel/common: remove nvm headers and c file There's no sense in having the nvm abstraction in its own files. Put that support directly into mrc_cache.c. BUG=b:69614064 Change-Id: I0f1a801c6e1a8c35f70faf9e4318bdc45955047a Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/22883 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2017-12-14 21:53:14 +01:00
return 0;
coreboot: Replace all IS_ENABLED(CONFIG_XXX) with CONFIG(XXX) This patch is a raw application of find src/ -type f | xargs sed -i -e 's/IS_ENABLED\s*(CONFIG_/CONFIG(/g' Change-Id: I6262d6d5c23cabe23c242b4f38d446b74fe16b88 Signed-off-by: Julius Werner <jwerner@chromium.org> Reviewed-on: https://review.coreboot.org/c/coreboot/+/31774 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2019-03-06 01:53:33 +01:00
if (!CONFIG(BOOT_DEVICE_SPI_FLASH))
soc/intel/common: remove nvm headers and c file There's no sense in having the nvm abstraction in its own files. Put that support directly into mrc_cache.c. BUG=b:69614064 Change-Id: I0f1a801c6e1a8c35f70faf9e4318bdc45955047a Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/22883 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2017-12-14 21:53:14 +01:00
return 0;
drivers/mrc_cache: Avoid unused variable assignment Fix the scan-build warning below: CC romstage/drivers/mrc_cache/mrc_cache.o src/drivers/mrc_cache/mrc_cache.c:450:26: warning: Value stored to 'flash' during its initialization is never read const struct spi_flash *flash = boot_device_spi_flash(); ^~~~~ ~~~~~~~~~~~~~~~~~~~~~~~ 1 warning generated. The function can return early before the value is read. Fix this, by getting rid of the variable, as the value is only read once. Change-Id: I3c94b123f4994eed9d7568b63971fd5b1d94bc09 Found-by: scan-build (clang-tools-9 1:9.0.1-12) Signed-off-by: Paul Menzel <pmenzel@molgen.mpg.de> Reviewed-on: https://review.coreboot.org/c/coreboot/+/42798 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Patrick Rudolph <siro@das-labor.org>
2020-06-25 09:11:56 +02:00
return spi_flash_ctrlr_protect_region(boot_device_spi_flash(), r, WRITE_PROTECT);
soc/intel/common: remove nvm headers and c file There's no sense in having the nvm abstraction in its own files. Put that support directly into mrc_cache.c. BUG=b:69614064 Change-Id: I0f1a801c6e1a8c35f70faf9e4318bdc45955047a Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/22883 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2017-12-14 21:53:14 +01:00
}
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
/* Protect mrc region with a Protected Range Register */
static int protect_mrc_cache(const char *name)
{
struct region region;
coreboot: Replace all IS_ENABLED(CONFIG_XXX) with CONFIG(XXX) This patch is a raw application of find src/ -type f | xargs sed -i -e 's/IS_ENABLED\s*(CONFIG_/CONFIG(/g' Change-Id: I6262d6d5c23cabe23c242b4f38d446b74fe16b88 Signed-off-by: Julius Werner <jwerner@chromium.org> Reviewed-on: https://review.coreboot.org/c/coreboot/+/31774 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2019-03-06 01:53:33 +01:00
if (!CONFIG(MRC_SETTINGS_PROTECT))
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
return 0;
if (lookup_region_by_name(name, &region) < 0) {
drivers/mrc_cache: Reduce severity of `region not found` log message On autogenerated FMAPs, there's no `UNIFIED_MRC_CACHE` region. The current code will print a spurious error message about it, though. Reduce the log level to BIOS_INFO to avoid confusion. Change-Id: I0961bb2a7d2d81dc5c0d28f6e6c29b320421fc3e Signed-off-by: Angel Pons <th3fanbus@gmail.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/45076 Reviewed-by: Furquan Shaikh <furquan@google.com> Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net> Reviewed-by: Tim Wawrzynczak <twawrzynczak@chromium.org> Reviewed-by: Christian Walter <christian.walter@9elements.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-09-03 16:04:27 +02:00
printk(BIOS_INFO, "MRC: Could not find region '%s'\n", name);
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
return -1;
}
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
if (nvm_is_write_protected() <= 0) {
printk(BIOS_INFO, "MRC: NOT enabling PRR for '%s'.\n", name);
return 0;
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
}
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
if (nvm_protect(&region) < 0) {
printk(BIOS_ERR, "MRC: ERROR setting PRR for '%s'.\n", name);
return -1;
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
}
soc/intel/common: remove mrc cache assumptions Update the mrc cache implementation to use region_file. Instead of relying on memory-mapped access and pointer arithmetic use the region_devices and region_file to obtain the latest data associated with the region. This removes the need for the nvm wrapper as the region_devices can be used directly. Thus, the library is more generic and can be extended to work on different boot mediums. BUG=chrome-os-partner:56151 Change-Id: Ic14e2d2f7339e50256b4a3a297fc33991861ca44 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17717 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-04 05:08:20 +01:00
printk(BIOS_INFO, "MRC: Enabled Protected Range on '%s'.\n", name);
return 0;
mrc: Add support for separate training cache in recovery mode 1. Re-factor MRC cache driver to properly select RW_MRC_CACHE or RECOVERY_MRC_CACHE based on the boot mode. - If normal mode boot, use RW_MRC_CACHE, if available. - If recovery mode boot: - Retrain memory if RECOVERY_MRC_CACHE not present, or recovery is requested explicity with retrain memory request. - Use RECOVERY_MRC_CACHE otherwise. 2. Protect RW and RECOVERY mrc caches in recovery and non-recovery boot modes. Check if both are present under one unified region and protect that region as a whole. Else try protecting individual regions. 3. Update training data in appropriate cache: - Use RW_MRC_CACHE if normal mode. - Use RECOVERY_MRC_CACHE if present in recovery mode. Else use RW_MRC_CACHE. 4. Add proper debug logs to indicate which training data cache is used at any point. BUG=chrome-os-partner:59352 BRANCH=None TEST=Verified that correct cache is used in both normal and recovery mode on reef. Change-Id: Ie79737a1450bd1ff71543e44a5a3e16950e70fb3 Signed-off-by: Furquan Shaikh <furquan@chromium.org> Reviewed-on: https://review.coreboot.org/17242 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-06 07:57:02 +01:00
}
soc/intel/common: log event when MRC cache is updated Log when the MRC cache is attempted to be updated with status of success or failure. Just one slot is supported currently which is deemed 'normal'. This is because there are more slots anticipated in the future. BUG=chrome-os-partner:59395 Change-Id: I0f81458325697aff9924cc359a4173e0d35da5da Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17231 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins)
2016-11-04 16:07:14 +01:00
mrc: Add support for separate training cache in recovery mode 1. Re-factor MRC cache driver to properly select RW_MRC_CACHE or RECOVERY_MRC_CACHE based on the boot mode. - If normal mode boot, use RW_MRC_CACHE, if available. - If recovery mode boot: - Retrain memory if RECOVERY_MRC_CACHE not present, or recovery is requested explicity with retrain memory request. - Use RECOVERY_MRC_CACHE otherwise. 2. Protect RW and RECOVERY mrc caches in recovery and non-recovery boot modes. Check if both are present under one unified region and protect that region as a whole. Else try protecting individual regions. 3. Update training data in appropriate cache: - Use RW_MRC_CACHE if normal mode. - Use RECOVERY_MRC_CACHE if present in recovery mode. Else use RW_MRC_CACHE. 4. Add proper debug logs to indicate which training data cache is used at any point. BUG=chrome-os-partner:59352 BRANCH=None TEST=Verified that correct cache is used in both normal and recovery mode on reef. Change-Id: Ie79737a1450bd1ff71543e44a5a3e16950e70fb3 Signed-off-by: Furquan Shaikh <furquan@chromium.org> Reviewed-on: https://review.coreboot.org/17242 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-06 07:57:02 +01:00
static void protect_mrc_region(void)
{
/*
* Check if there is a single unified region that encompasses both
* RECOVERY_MRC_CACHE and DEFAULT_MRC_CACHE. In that case protect the
* entire region using a single PRR.
*
* If we are not able to protect the entire region, try protecting
* individual regions next.
*/
if (protect_mrc_cache(UNIFIED_MRC_CACHE) == 0)
return;
coreboot: Replace all IS_ENABLED(CONFIG_XXX) with CONFIG(XXX) This patch is a raw application of find src/ -type f | xargs sed -i -e 's/IS_ENABLED\s*(CONFIG_/CONFIG(/g' Change-Id: I6262d6d5c23cabe23c242b4f38d446b74fe16b88 Signed-off-by: Julius Werner <jwerner@chromium.org> Reviewed-on: https://review.coreboot.org/c/coreboot/+/31774 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2019-03-06 01:53:33 +01:00
if (CONFIG(HAS_RECOVERY_MRC_CACHE))
mrc: Add support for separate training cache in recovery mode 1. Re-factor MRC cache driver to properly select RW_MRC_CACHE or RECOVERY_MRC_CACHE based on the boot mode. - If normal mode boot, use RW_MRC_CACHE, if available. - If recovery mode boot: - Retrain memory if RECOVERY_MRC_CACHE not present, or recovery is requested explicity with retrain memory request. - Use RECOVERY_MRC_CACHE otherwise. 2. Protect RW and RECOVERY mrc caches in recovery and non-recovery boot modes. Check if both are present under one unified region and protect that region as a whole. Else try protecting individual regions. 3. Update training data in appropriate cache: - Use RW_MRC_CACHE if normal mode. - Use RECOVERY_MRC_CACHE if present in recovery mode. Else use RW_MRC_CACHE. 4. Add proper debug logs to indicate which training data cache is used at any point. BUG=chrome-os-partner:59352 BRANCH=None TEST=Verified that correct cache is used in both normal and recovery mode on reef. Change-Id: Ie79737a1450bd1ff71543e44a5a3e16950e70fb3 Signed-off-by: Furquan Shaikh <furquan@chromium.org> Reviewed-on: https://review.coreboot.org/17242 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-06 07:57:02 +01:00
protect_mrc_cache(RECOVERY_MRC_CACHE);
protect_mrc_cache(DEFAULT_MRC_CACHE);
}
soc/intel/common: provide option to invalide MRC cache on recovery Allow a board/platform to invalidate the normal MRC cache when hardware retraining is requested in recovery mode. A small 4 byte payload is used to update the latest data written. It will of course fail on MRC cache retreival on next usage. BUG=chrome-os-partner:60592 BRANCH=reef Change-Id: Ic88cd66802664813562d003f50d54e4a4ea375ef Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17870 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-14 22:36:56 +01:00
static void invalidate_normal_cache(void)
{
struct region_file cache_file;
struct region_device rdev;
const char *name = DEFAULT_MRC_CACHE;
const uint32_t invalid = ~MRC_DATA_SIGNATURE;
mrc_cache: Move code for triggering memory training into mrc_cache Currently the decision of whether or not to use mrc_cache in recovery mode is made within the individual platforms' drivers (ie: fsp2.0, fsp1.1, etc.). As this is not platform specific, but uses common vboot infrastructure, the code can be unified and moved into mrc_cache. The conditions are as follows: 1. If HAS_RECOVERY_MRC_CACHE, use mrc_cache data (unless retrain switch is true) 2. If !HAS_RECOVERY_MRC_CACHE && VBOOT_STARTS_IN_BOOTBLOCK, this means that memory training will occur after verified boot, meaning that mrc_cache will be filled with data from executing RW code. So in this case, we never want to use the training data in the mrc_cache for recovery mode. 3. If !HAS_RECOVERY_MRC_CACHE && VBOOT_STARTS_IN_ROMSTAGE, this means that memory training happens before verfied boot, meaning that the mrc_cache data is generated by RO code, so it is safe to use for a recovery boot. 4. Any platform that does not use vboot should be unaffected. Additionally, we have removed the MRC_CLEAR_NORMAL_CACHE_ON_RECOVERY_RETRAIN config because the mrc_cache driver takes care of invalidating the mrc_cache data for normal mode. If the platform: 1. !HAS_RECOVERY_MRC_CACHE, always invalidate mrc_cache data 2. HAS_RECOVERY_MRC_CACHE, only invalidate if retrain switch is set BUG=b:150502246 BRANCH=None TEST=1. run dut-control power_state:rec_force_mrc twice on lazor ensure that memory retraining happens both times run dut-control power_state:rec twice on lazor ensure that memory retraining happens only first time 2. remove HAS_RECOVERY_MRC_CACHE from lazor Kconfig boot twice to ensure caching of memory training occurred on each boot. Change-Id: I3875a7b4a4ba3c1aa8a3c1507b3993036a7155fc Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46855 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-10-27 23:58:31 +01:00
/*
* If !HAS_RECOVERY_MRC_CACHE and VBOOT_STARTS_IN_ROMSTAGE is
* selected, this means that memory training occurs before
* verified boot (in RO), so normal mode cache does not need
* to be invalidated.
*/
if (!CONFIG(HAS_RECOVERY_MRC_CACHE) && CONFIG(VBOOT_STARTS_IN_ROMSTAGE))
return;
/* We only invalidate the normal cache in recovery mode. */
soc/intel/common: provide option to invalide MRC cache on recovery Allow a board/platform to invalidate the normal MRC cache when hardware retraining is requested in recovery mode. A small 4 byte payload is used to update the latest data written. It will of course fail on MRC cache retreival on next usage. BUG=chrome-os-partner:60592 BRANCH=reef Change-Id: Ic88cd66802664813562d003f50d54e4a4ea375ef Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17870 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-14 22:36:56 +01:00
if (!vboot_recovery_mode_enabled())
return;
mrc_cache: Move code for triggering memory training into mrc_cache Currently the decision of whether or not to use mrc_cache in recovery mode is made within the individual platforms' drivers (ie: fsp2.0, fsp1.1, etc.). As this is not platform specific, but uses common vboot infrastructure, the code can be unified and moved into mrc_cache. The conditions are as follows: 1. If HAS_RECOVERY_MRC_CACHE, use mrc_cache data (unless retrain switch is true) 2. If !HAS_RECOVERY_MRC_CACHE && VBOOT_STARTS_IN_BOOTBLOCK, this means that memory training will occur after verified boot, meaning that mrc_cache will be filled with data from executing RW code. So in this case, we never want to use the training data in the mrc_cache for recovery mode. 3. If !HAS_RECOVERY_MRC_CACHE && VBOOT_STARTS_IN_ROMSTAGE, this means that memory training happens before verfied boot, meaning that the mrc_cache data is generated by RO code, so it is safe to use for a recovery boot. 4. Any platform that does not use vboot should be unaffected. Additionally, we have removed the MRC_CLEAR_NORMAL_CACHE_ON_RECOVERY_RETRAIN config because the mrc_cache driver takes care of invalidating the mrc_cache data for normal mode. If the platform: 1. !HAS_RECOVERY_MRC_CACHE, always invalidate mrc_cache data 2. HAS_RECOVERY_MRC_CACHE, only invalidate if retrain switch is set BUG=b:150502246 BRANCH=None TEST=1. run dut-control power_state:rec_force_mrc twice on lazor ensure that memory retraining happens both times run dut-control power_state:rec twice on lazor ensure that memory retraining happens only first time 2. remove HAS_RECOVERY_MRC_CACHE from lazor Kconfig boot twice to ensure caching of memory training occurred on each boot. Change-Id: I3875a7b4a4ba3c1aa8a3c1507b3993036a7155fc Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46855 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-10-27 23:58:31 +01:00
/*
* For platforms with a recovery mrc_cache, no need to
* invalidate when retrain switch is not set.
*/
if (CONFIG(HAS_RECOVERY_MRC_CACHE) && !get_recovery_mode_retrain_switch())
soc/intel/common: provide option to invalide MRC cache on recovery Allow a board/platform to invalidate the normal MRC cache when hardware retraining is requested in recovery mode. A small 4 byte payload is used to update the latest data written. It will of course fail on MRC cache retreival on next usage. BUG=chrome-os-partner:60592 BRANCH=reef Change-Id: Ic88cd66802664813562d003f50d54e4a4ea375ef Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17870 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-14 22:36:56 +01:00
return;
if (fmap_locate_area_as_rdev_rw(name, &rdev) < 0) {
printk(BIOS_ERR, "MRC: Couldn't find '%s' region. Invalidation failed\n",
name);
return;
}
if (region_file_init(&cache_file, &rdev) < 0) {
printk(BIOS_ERR, "MRC: region file invalid for '%s'. Invalidation failed\n",
name);
return;
}
/* Push an update that consists of 4 bytes that is smaller than the
* MRC metadata as well as an invalid signature. */
if (region_file_update_data(&cache_file, &invalid, sizeof(invalid)) < 0)
printk(BIOS_ERR, "MRC: invalidation failed for '%s'.\n", name);
}
mrc_cache: Update mrc_cache data in romstage Previously, we were writing to cbmem after memory training and then writing the training data from cbmem to mrc_cache in ramstage. We were doing this because we were unable to read/write to SPI simultaneously on older x86 chips. Now that newer chips allow for simultaneously reads and writes, we can move the mrc_cache update into romstage. This is beneficial if there is a reboot for some reason after memory training but before the previous mrc_cache_stash_data call originally in ramstage. If this happens, we would lose all the mrc_cache training data in the next boot even though we've already performed the memory training. Added new config BOOT_DEVICE_SPI_FLASH_NO_EARLY_WRITES to accomodate older x86 platforms that don't do mmapping but still want to use the cbmem to store the mrc_cache data in order to write the mrc_cache data back at a later time. We are maintaining the use of cbmem for these older platforms because we have no way of validating the earlier write back to mrc_cache at this time. BUG=b:150502246 BRANCH=None TEST=reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I3430bda45484cb8c2b01ab9614508039dfaac9a3 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44196 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-05-02 02:00:31 +02:00
static void update_mrc_cache_from_cbmem(int type)
{
const struct cache_region *cr;
struct region region;
const struct cbmem_entry *to_be_updated;
cr = lookup_region(&region, type);
if (cr == NULL) {
mrc_cache: Move code for triggering memory training into mrc_cache Currently the decision of whether or not to use mrc_cache in recovery mode is made within the individual platforms' drivers (ie: fsp2.0, fsp1.1, etc.). As this is not platform specific, but uses common vboot infrastructure, the code can be unified and moved into mrc_cache. The conditions are as follows: 1. If HAS_RECOVERY_MRC_CACHE, use mrc_cache data (unless retrain switch is true) 2. If !HAS_RECOVERY_MRC_CACHE && VBOOT_STARTS_IN_BOOTBLOCK, this means that memory training will occur after verified boot, meaning that mrc_cache will be filled with data from executing RW code. So in this case, we never want to use the training data in the mrc_cache for recovery mode. 3. If !HAS_RECOVERY_MRC_CACHE && VBOOT_STARTS_IN_ROMSTAGE, this means that memory training happens before verfied boot, meaning that the mrc_cache data is generated by RO code, so it is safe to use for a recovery boot. 4. Any platform that does not use vboot should be unaffected. Additionally, we have removed the MRC_CLEAR_NORMAL_CACHE_ON_RECOVERY_RETRAIN config because the mrc_cache driver takes care of invalidating the mrc_cache data for normal mode. If the platform: 1. !HAS_RECOVERY_MRC_CACHE, always invalidate mrc_cache data 2. HAS_RECOVERY_MRC_CACHE, only invalidate if retrain switch is set BUG=b:150502246 BRANCH=None TEST=1. run dut-control power_state:rec_force_mrc twice on lazor ensure that memory retraining happens both times run dut-control power_state:rec twice on lazor ensure that memory retraining happens only first time 2. remove HAS_RECOVERY_MRC_CACHE from lazor Kconfig boot twice to ensure caching of memory training occurred on each boot. Change-Id: I3875a7b4a4ba3c1aa8a3c1507b3993036a7155fc Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46855 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-10-27 23:58:31 +01:00
printk(BIOS_INFO, "MRC: could not find cache_region type %d\n", type);
mrc_cache: Update mrc_cache data in romstage Previously, we were writing to cbmem after memory training and then writing the training data from cbmem to mrc_cache in ramstage. We were doing this because we were unable to read/write to SPI simultaneously on older x86 chips. Now that newer chips allow for simultaneously reads and writes, we can move the mrc_cache update into romstage. This is beneficial if there is a reboot for some reason after memory training but before the previous mrc_cache_stash_data call originally in ramstage. If this happens, we would lose all the mrc_cache training data in the next boot even though we've already performed the memory training. Added new config BOOT_DEVICE_SPI_FLASH_NO_EARLY_WRITES to accomodate older x86 platforms that don't do mmapping but still want to use the cbmem to store the mrc_cache data in order to write the mrc_cache data back at a later time. We are maintaining the use of cbmem for these older platforms because we have no way of validating the earlier write back to mrc_cache at this time. BUG=b:150502246 BRANCH=None TEST=reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I3430bda45484cb8c2b01ab9614508039dfaac9a3 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44196 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-05-02 02:00:31 +02:00
return;
}
to_be_updated = cbmem_entry_find(cr->cbmem_id);
if (to_be_updated == NULL) {
printk(BIOS_INFO, "MRC: No data in cbmem for '%s'.\n",
cr->name);
return;
}
update_mrc_cache_by_type(type,
/* pointer to mrc_cache entry metadata header */
cbmem_entry_start(to_be_updated),
/* pointer to start of mrc_cache entry data */
cbmem_entry_start(to_be_updated) +
sizeof(struct mrc_metadata),
/* size of just data portion of the entry */
cbmem_entry_size(to_be_updated) -
sizeof(struct mrc_metadata));
}
static void finalize_mrc_cache(void *unused)
mrc: Add support for separate training cache in recovery mode 1. Re-factor MRC cache driver to properly select RW_MRC_CACHE or RECOVERY_MRC_CACHE based on the boot mode. - If normal mode boot, use RW_MRC_CACHE, if available. - If recovery mode boot: - Retrain memory if RECOVERY_MRC_CACHE not present, or recovery is requested explicity with retrain memory request. - Use RECOVERY_MRC_CACHE otherwise. 2. Protect RW and RECOVERY mrc caches in recovery and non-recovery boot modes. Check if both are present under one unified region and protect that region as a whole. Else try protecting individual regions. 3. Update training data in appropriate cache: - Use RW_MRC_CACHE if normal mode. - Use RECOVERY_MRC_CACHE if present in recovery mode. Else use RW_MRC_CACHE. 4. Add proper debug logs to indicate which training data cache is used at any point. BUG=chrome-os-partner:59352 BRANCH=None TEST=Verified that correct cache is used in both normal and recovery mode on reef. Change-Id: Ie79737a1450bd1ff71543e44a5a3e16950e70fb3 Signed-off-by: Furquan Shaikh <furquan@chromium.org> Reviewed-on: https://review.coreboot.org/17242 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-06 07:57:02 +01:00
{
mrc_cache: Update mrc_cache data in romstage Previously, we were writing to cbmem after memory training and then writing the training data from cbmem to mrc_cache in ramstage. We were doing this because we were unable to read/write to SPI simultaneously on older x86 chips. Now that newer chips allow for simultaneously reads and writes, we can move the mrc_cache update into romstage. This is beneficial if there is a reboot for some reason after memory training but before the previous mrc_cache_stash_data call originally in ramstage. If this happens, we would lose all the mrc_cache training data in the next boot even though we've already performed the memory training. Added new config BOOT_DEVICE_SPI_FLASH_NO_EARLY_WRITES to accomodate older x86 platforms that don't do mmapping but still want to use the cbmem to store the mrc_cache data in order to write the mrc_cache data back at a later time. We are maintaining the use of cbmem for these older platforms because we have no way of validating the earlier write back to mrc_cache at this time. BUG=b:150502246 BRANCH=None TEST=reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I3430bda45484cb8c2b01ab9614508039dfaac9a3 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44196 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-05-02 02:00:31 +02:00
if (CONFIG(MRC_STASH_TO_CBMEM)) {
update_mrc_cache_from_cbmem(MRC_TRAINING_DATA);
soc/intel/common: Add save/restore for variable MRC data Piggy-back on existing MRC cache infrastructure to store variable MRC data. Only one set of data can be valid at given point of time. Currently this magically happens because region alignment is forced to 0x1000 and region itself is of the same size. This needs to be somehow programmatically enforced. Change-Id: I8a660d356ca760b8ff9907396fb9b34cb16cf1db Signed-off-by: Andrey Petrov <andrey.petrov@intel.com> Reviewed-on: https://review.coreboot.org/17320 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-08 17:30:06 +01:00
mrc_cache: Update mrc_cache data in romstage Previously, we were writing to cbmem after memory training and then writing the training data from cbmem to mrc_cache in ramstage. We were doing this because we were unable to read/write to SPI simultaneously on older x86 chips. Now that newer chips allow for simultaneously reads and writes, we can move the mrc_cache update into romstage. This is beneficial if there is a reboot for some reason after memory training but before the previous mrc_cache_stash_data call originally in ramstage. If this happens, we would lose all the mrc_cache training data in the next boot even though we've already performed the memory training. Added new config BOOT_DEVICE_SPI_FLASH_NO_EARLY_WRITES to accomodate older x86 platforms that don't do mmapping but still want to use the cbmem to store the mrc_cache data in order to write the mrc_cache data back at a later time. We are maintaining the use of cbmem for these older platforms because we have no way of validating the earlier write back to mrc_cache at this time. BUG=b:150502246 BRANCH=None TEST=reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I3430bda45484cb8c2b01ab9614508039dfaac9a3 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44196 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-05-02 02:00:31 +02:00
if (CONFIG(MRC_SETTINGS_VARIABLE_DATA))
update_mrc_cache_from_cbmem(MRC_VARIABLE_DATA);
}
soc/intel/common: Add save/restore for variable MRC data Piggy-back on existing MRC cache infrastructure to store variable MRC data. Only one set of data can be valid at given point of time. Currently this magically happens because region alignment is forced to 0x1000 and region itself is of the same size. This needs to be somehow programmatically enforced. Change-Id: I8a660d356ca760b8ff9907396fb9b34cb16cf1db Signed-off-by: Andrey Petrov <andrey.petrov@intel.com> Reviewed-on: https://review.coreboot.org/17320 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-08 17:30:06 +01:00
mrc_cache: Move code for triggering memory training into mrc_cache Currently the decision of whether or not to use mrc_cache in recovery mode is made within the individual platforms' drivers (ie: fsp2.0, fsp1.1, etc.). As this is not platform specific, but uses common vboot infrastructure, the code can be unified and moved into mrc_cache. The conditions are as follows: 1. If HAS_RECOVERY_MRC_CACHE, use mrc_cache data (unless retrain switch is true) 2. If !HAS_RECOVERY_MRC_CACHE && VBOOT_STARTS_IN_BOOTBLOCK, this means that memory training will occur after verified boot, meaning that mrc_cache will be filled with data from executing RW code. So in this case, we never want to use the training data in the mrc_cache for recovery mode. 3. If !HAS_RECOVERY_MRC_CACHE && VBOOT_STARTS_IN_ROMSTAGE, this means that memory training happens before verfied boot, meaning that the mrc_cache data is generated by RO code, so it is safe to use for a recovery boot. 4. Any platform that does not use vboot should be unaffected. Additionally, we have removed the MRC_CLEAR_NORMAL_CACHE_ON_RECOVERY_RETRAIN config because the mrc_cache driver takes care of invalidating the mrc_cache data for normal mode. If the platform: 1. !HAS_RECOVERY_MRC_CACHE, always invalidate mrc_cache data 2. HAS_RECOVERY_MRC_CACHE, only invalidate if retrain switch is set BUG=b:150502246 BRANCH=None TEST=1. run dut-control power_state:rec_force_mrc twice on lazor ensure that memory retraining happens both times run dut-control power_state:rec twice on lazor ensure that memory retraining happens only first time 2. remove HAS_RECOVERY_MRC_CACHE from lazor Kconfig boot twice to ensure caching of memory training occurred on each boot. Change-Id: I3875a7b4a4ba3c1aa8a3c1507b3993036a7155fc Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46855 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-10-27 23:58:31 +01:00
invalidate_normal_cache();
soc/intel/common: provide option to invalide MRC cache on recovery Allow a board/platform to invalidate the normal MRC cache when hardware retraining is requested in recovery mode. A small 4 byte payload is used to update the latest data written. It will of course fail on MRC cache retreival on next usage. BUG=chrome-os-partner:60592 BRANCH=reef Change-Id: Ic88cd66802664813562d003f50d54e4a4ea375ef Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://review.coreboot.org/17870 Tested-by: build bot (Jenkins) Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
2016-12-14 22:36:56 +01:00
mrc: Add support for separate training cache in recovery mode 1. Re-factor MRC cache driver to properly select RW_MRC_CACHE or RECOVERY_MRC_CACHE based on the boot mode. - If normal mode boot, use RW_MRC_CACHE, if available. - If recovery mode boot: - Retrain memory if RECOVERY_MRC_CACHE not present, or recovery is requested explicity with retrain memory request. - Use RECOVERY_MRC_CACHE otherwise. 2. Protect RW and RECOVERY mrc caches in recovery and non-recovery boot modes. Check if both are present under one unified region and protect that region as a whole. Else try protecting individual regions. 3. Update training data in appropriate cache: - Use RW_MRC_CACHE if normal mode. - Use RECOVERY_MRC_CACHE if present in recovery mode. Else use RW_MRC_CACHE. 4. Add proper debug logs to indicate which training data cache is used at any point. BUG=chrome-os-partner:59352 BRANCH=None TEST=Verified that correct cache is used in both normal and recovery mode on reef. Change-Id: Ie79737a1450bd1ff71543e44a5a3e16950e70fb3 Signed-off-by: Furquan Shaikh <furquan@chromium.org> Reviewed-on: https://review.coreboot.org/17242 Tested-by: build bot (Jenkins) Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2016-11-06 07:57:02 +01:00
protect_mrc_region();
baytrail: add initial support The initial Bay Trail code is intended to support the mobile and desktop version of Bay Trail. This support can train memory and execute through ramstage. However, the resource allocation is not curently handled correctly. The MRC cache parameters are successfully saved and reused after the initial cold boot. BUG=chrome-os-partner:22292 BRANCH=None TEST=Built and booted on a reference board through ramstage. Change-Id: I238ede326802aad272c6cca39d7ad4f161d813f5 Signed-off-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: https://chromium-review.googlesource.com/168387 Reviewed-by: Duncan Laurie <dlaurie@chromium.org> Reviewed-on: http://review.coreboot.org/4847 Tested-by: build bot (Jenkins) Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
2013-09-07 07:41:48 +02:00
}
mrc_cache: Move mrc_cache_stash_data to end of file We need to pull update_mrc_cache into mrc_cache_stash_data, so moving to end of the file to make sure update_mrc_cache is defined before. BUG=b:150502246 BRANCH=None TEST=Testing on a nami (x86) device: reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I9e14fec96e9dabceafc2f6f5663fc6f1023f0395 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44195 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Julius Werner <jwerner@chromium.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2020-08-04 23:43:28 +02:00
int mrc_cache_stash_data(int type, uint32_t version, const void *data,
mrc_cache: Update mrc_cache data in romstage Previously, we were writing to cbmem after memory training and then writing the training data from cbmem to mrc_cache in ramstage. We were doing this because we were unable to read/write to SPI simultaneously on older x86 chips. Now that newer chips allow for simultaneously reads and writes, we can move the mrc_cache update into romstage. This is beneficial if there is a reboot for some reason after memory training but before the previous mrc_cache_stash_data call originally in ramstage. If this happens, we would lose all the mrc_cache training data in the next boot even though we've already performed the memory training. Added new config BOOT_DEVICE_SPI_FLASH_NO_EARLY_WRITES to accomodate older x86 platforms that don't do mmapping but still want to use the cbmem to store the mrc_cache data in order to write the mrc_cache data back at a later time. We are maintaining the use of cbmem for these older platforms because we have no way of validating the earlier write back to mrc_cache at this time. BUG=b:150502246 BRANCH=None TEST=reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I3430bda45484cb8c2b01ab9614508039dfaac9a3 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44196 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-05-02 02:00:31 +02:00
size_t size)
mrc_cache: Move mrc_cache_stash_data to end of file We need to pull update_mrc_cache into mrc_cache_stash_data, so moving to end of the file to make sure update_mrc_cache is defined before. BUG=b:150502246 BRANCH=None TEST=Testing on a nami (x86) device: reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I9e14fec96e9dabceafc2f6f5663fc6f1023f0395 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44195 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Julius Werner <jwerner@chromium.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2020-08-04 23:43:28 +02:00
{
const struct cache_region *cr;
mrc_cache: Update mrc_cache data in romstage Previously, we were writing to cbmem after memory training and then writing the training data from cbmem to mrc_cache in ramstage. We were doing this because we were unable to read/write to SPI simultaneously on older x86 chips. Now that newer chips allow for simultaneously reads and writes, we can move the mrc_cache update into romstage. This is beneficial if there is a reboot for some reason after memory training but before the previous mrc_cache_stash_data call originally in ramstage. If this happens, we would lose all the mrc_cache training data in the next boot even though we've already performed the memory training. Added new config BOOT_DEVICE_SPI_FLASH_NO_EARLY_WRITES to accomodate older x86 platforms that don't do mmapping but still want to use the cbmem to store the mrc_cache data in order to write the mrc_cache data back at a later time. We are maintaining the use of cbmem for these older platforms because we have no way of validating the earlier write back to mrc_cache at this time. BUG=b:150502246 BRANCH=None TEST=reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I3430bda45484cb8c2b01ab9614508039dfaac9a3 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44196 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-05-02 02:00:31 +02:00
struct mrc_metadata md = {
.signature = MRC_DATA_SIGNATURE,
.data_size = size,
.version = version,
.data_checksum = compute_ip_checksum(data, size),
};
md.header_checksum =
compute_ip_checksum(&md, sizeof(struct mrc_metadata));
if (CONFIG(MRC_STASH_TO_CBMEM)) {
/* Store data in cbmem for use in ramstage */
struct mrc_metadata *cbmem_md;
size_t cbmem_size;
cbmem_size = sizeof(*cbmem_md) + size;
mrc_cache: Move code for triggering memory training into mrc_cache Currently the decision of whether or not to use mrc_cache in recovery mode is made within the individual platforms' drivers (ie: fsp2.0, fsp1.1, etc.). As this is not platform specific, but uses common vboot infrastructure, the code can be unified and moved into mrc_cache. The conditions are as follows: 1. If HAS_RECOVERY_MRC_CACHE, use mrc_cache data (unless retrain switch is true) 2. If !HAS_RECOVERY_MRC_CACHE && VBOOT_STARTS_IN_BOOTBLOCK, this means that memory training will occur after verified boot, meaning that mrc_cache will be filled with data from executing RW code. So in this case, we never want to use the training data in the mrc_cache for recovery mode. 3. If !HAS_RECOVERY_MRC_CACHE && VBOOT_STARTS_IN_ROMSTAGE, this means that memory training happens before verfied boot, meaning that the mrc_cache data is generated by RO code, so it is safe to use for a recovery boot. 4. Any platform that does not use vboot should be unaffected. Additionally, we have removed the MRC_CLEAR_NORMAL_CACHE_ON_RECOVERY_RETRAIN config because the mrc_cache driver takes care of invalidating the mrc_cache data for normal mode. If the platform: 1. !HAS_RECOVERY_MRC_CACHE, always invalidate mrc_cache data 2. HAS_RECOVERY_MRC_CACHE, only invalidate if retrain switch is set BUG=b:150502246 BRANCH=None TEST=1. run dut-control power_state:rec_force_mrc twice on lazor ensure that memory retraining happens both times run dut-control power_state:rec twice on lazor ensure that memory retraining happens only first time 2. remove HAS_RECOVERY_MRC_CACHE from lazor Kconfig boot twice to ensure caching of memory training occurred on each boot. Change-Id: I3875a7b4a4ba3c1aa8a3c1507b3993036a7155fc Signed-off-by: Shelley Chen <shchen@google.com> Reviewed-on: https://review.coreboot.org/c/coreboot/+/46855 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-10-27 23:58:31 +01:00
cr = lookup_region_type(type);
if (cr == NULL) {
printk(BIOS_INFO, "MRC: No region type found. Skip adding to cbmem for type %d.\n",
type);
return 0;
}
mrc_cache: Update mrc_cache data in romstage Previously, we were writing to cbmem after memory training and then writing the training data from cbmem to mrc_cache in ramstage. We were doing this because we were unable to read/write to SPI simultaneously on older x86 chips. Now that newer chips allow for simultaneously reads and writes, we can move the mrc_cache update into romstage. This is beneficial if there is a reboot for some reason after memory training but before the previous mrc_cache_stash_data call originally in ramstage. If this happens, we would lose all the mrc_cache training data in the next boot even though we've already performed the memory training. Added new config BOOT_DEVICE_SPI_FLASH_NO_EARLY_WRITES to accomodate older x86 platforms that don't do mmapping but still want to use the cbmem to store the mrc_cache data in order to write the mrc_cache data back at a later time. We are maintaining the use of cbmem for these older platforms because we have no way of validating the earlier write back to mrc_cache at this time. BUG=b:150502246 BRANCH=None TEST=reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I3430bda45484cb8c2b01ab9614508039dfaac9a3 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44196 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-05-02 02:00:31 +02:00
cbmem_md = cbmem_add(cr->cbmem_id, cbmem_size);
if (cbmem_md == NULL) {
printk(BIOS_ERR, "MRC: failed to add '%s' to cbmem.\n",
cr->name);
return -1;
}
memcpy(cbmem_md, &md, sizeof(*cbmem_md));
/* cbmem_md + 1 is the pointer to the mrc_cache data */
memcpy(cbmem_md + 1, data, size);
} else {
/* Otherwise store to mrc_cache right away */
update_mrc_cache_by_type(type, &md, data, size);
mrc_cache: Move mrc_cache_stash_data to end of file We need to pull update_mrc_cache into mrc_cache_stash_data, so moving to end of the file to make sure update_mrc_cache is defined before. BUG=b:150502246 BRANCH=None TEST=Testing on a nami (x86) device: reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I9e14fec96e9dabceafc2f6f5663fc6f1023f0395 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44195 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Julius Werner <jwerner@chromium.org> Reviewed-by: Furquan Shaikh <furquan@google.com>
2020-08-04 23:43:28 +02:00
}
return 0;
}
soc/intel/common: Move update_mrc_cache after BS_DEV_ENUMERATE This patch ensures that MRC cache data is already written into SPI chip before SPI protected regions are getting locked during BS_DEV_RESOURCES-BS_ON_EXIT. This requirements are intended to support platform security guideline where all required chipset registers are expected to be in lock down stage before launching any 3rd party code as in option rom etc. coreboot has to change its execution order to meet those requirements. Hence storing mrc cache data into SPI has been moved right after pci enumeration is done, so that SPI registers can be lock down before calling post pci enumeration FSP NotifyPhase() API which is targeted to be done in BS_DEV_ENABLE-BS_ON_ENTRY. TEST=Ensure MRC training data is stored into SPI chip and power_ Resume autotest is passing. Change-Id: I8ee26b5cc70433438cf4e45e707b8a54f89cf9b0 Signed-off-by: Subrata Banik <subrata.banik@intel.com> Reviewed-on: https://review.coreboot.org/21028 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2017-08-16 13:12:46 +02:00
/*
drivers/mrc_cache: Make bootstate for SPI writes variable The default time for writing MRC data to flash is at the exit of BS_DEV_ENUMERATE but this is too early for some implementations. Add an option to Kconfig for allowing a late option to be selected. The timing of the late option is at the entry of BS_OS_RESUME_CHECK. TEST=Select option and inspect console log on Kahlee BUG=b:69614064 Change-Id: Ie7ba9070829d98414ee788e14d1a768145d742ea Signed-off-by: Marshall Dawson <marshalldawson3rd@gmail.com> Reviewed-on: https://review.coreboot.org/22937 Reviewed-by: Martin Roth <martinroth@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2017-12-18 23:56:44 +01:00
* Ensures MRC training data is stored into SPI after PCI enumeration is done.
* Some implementations may require this to be later than others.
soc/intel/common: Move update_mrc_cache after BS_DEV_ENUMERATE This patch ensures that MRC cache data is already written into SPI chip before SPI protected regions are getting locked during BS_DEV_RESOURCES-BS_ON_EXIT. This requirements are intended to support platform security guideline where all required chipset registers are expected to be in lock down stage before launching any 3rd party code as in option rom etc. coreboot has to change its execution order to meet those requirements. Hence storing mrc cache data into SPI has been moved right after pci enumeration is done, so that SPI registers can be lock down before calling post pci enumeration FSP NotifyPhase() API which is targeted to be done in BS_DEV_ENABLE-BS_ON_ENTRY. TEST=Ensure MRC training data is stored into SPI chip and power_ Resume autotest is passing. Change-Id: I8ee26b5cc70433438cf4e45e707b8a54f89cf9b0 Signed-off-by: Subrata Banik <subrata.banik@intel.com> Reviewed-on: https://review.coreboot.org/21028 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Aaron Durbin <adurbin@chromium.org>
2017-08-16 13:12:46 +02:00
*/
coreboot: Replace all IS_ENABLED(CONFIG_XXX) with CONFIG(XXX) This patch is a raw application of find src/ -type f | xargs sed -i -e 's/IS_ENABLED\s*(CONFIG_/CONFIG(/g' Change-Id: I6262d6d5c23cabe23c242b4f38d446b74fe16b88 Signed-off-by: Julius Werner <jwerner@chromium.org> Reviewed-on: https://review.coreboot.org/c/coreboot/+/31774 Tested-by: build bot (Jenkins) <no-reply@coreboot.org> Reviewed-by: Patrick Georgi <pgeorgi@google.com>
2019-03-06 01:53:33 +01:00
#if CONFIG(MRC_WRITE_NV_LATE)
mrc_cache: Update mrc_cache data in romstage Previously, we were writing to cbmem after memory training and then writing the training data from cbmem to mrc_cache in ramstage. We were doing this because we were unable to read/write to SPI simultaneously on older x86 chips. Now that newer chips allow for simultaneously reads and writes, we can move the mrc_cache update into romstage. This is beneficial if there is a reboot for some reason after memory training but before the previous mrc_cache_stash_data call originally in ramstage. If this happens, we would lose all the mrc_cache training data in the next boot even though we've already performed the memory training. Added new config BOOT_DEVICE_SPI_FLASH_NO_EARLY_WRITES to accomodate older x86 platforms that don't do mmapping but still want to use the cbmem to store the mrc_cache data in order to write the mrc_cache data back at a later time. We are maintaining the use of cbmem for these older platforms because we have no way of validating the earlier write back to mrc_cache at this time. BUG=b:150502246 BRANCH=None TEST=reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I3430bda45484cb8c2b01ab9614508039dfaac9a3 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44196 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-05-02 02:00:31 +02:00
BOOT_STATE_INIT_ENTRY(BS_OS_RESUME_CHECK, BS_ON_ENTRY, finalize_mrc_cache, NULL);
drivers/mrc_cache: Make bootstate for SPI writes variable The default time for writing MRC data to flash is at the exit of BS_DEV_ENUMERATE but this is too early for some implementations. Add an option to Kconfig for allowing a late option to be selected. The timing of the late option is at the entry of BS_OS_RESUME_CHECK. TEST=Select option and inspect console log on Kahlee BUG=b:69614064 Change-Id: Ie7ba9070829d98414ee788e14d1a768145d742ea Signed-off-by: Marshall Dawson <marshalldawson3rd@gmail.com> Reviewed-on: https://review.coreboot.org/22937 Reviewed-by: Martin Roth <martinroth@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2017-12-18 23:56:44 +01:00
#else
mrc_cache: Update mrc_cache data in romstage Previously, we were writing to cbmem after memory training and then writing the training data from cbmem to mrc_cache in ramstage. We were doing this because we were unable to read/write to SPI simultaneously on older x86 chips. Now that newer chips allow for simultaneously reads and writes, we can move the mrc_cache update into romstage. This is beneficial if there is a reboot for some reason after memory training but before the previous mrc_cache_stash_data call originally in ramstage. If this happens, we would lose all the mrc_cache training data in the next boot even though we've already performed the memory training. Added new config BOOT_DEVICE_SPI_FLASH_NO_EARLY_WRITES to accomodate older x86 platforms that don't do mmapping but still want to use the cbmem to store the mrc_cache data in order to write the mrc_cache data back at a later time. We are maintaining the use of cbmem for these older platforms because we have no way of validating the earlier write back to mrc_cache at this time. BUG=b:150502246 BRANCH=None TEST=reboot from ec console. Make sure memory training happens. reboot from ec console. Make sure that we don't do training again. Signed-off-by: Shelley Chen <shchen@google.com> Change-Id: I3430bda45484cb8c2b01ab9614508039dfaac9a3 Reviewed-on: https://review.coreboot.org/c/coreboot/+/44196 Reviewed-by: Furquan Shaikh <furquan@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2020-05-02 02:00:31 +02:00
BOOT_STATE_INIT_ENTRY(BS_DEV_ENUMERATE, BS_ON_EXIT, finalize_mrc_cache, NULL);
drivers/mrc_cache: Make bootstate for SPI writes variable The default time for writing MRC data to flash is at the exit of BS_DEV_ENUMERATE but this is too early for some implementations. Add an option to Kconfig for allowing a late option to be selected. The timing of the late option is at the entry of BS_OS_RESUME_CHECK. TEST=Select option and inspect console log on Kahlee BUG=b:69614064 Change-Id: Ie7ba9070829d98414ee788e14d1a768145d742ea Signed-off-by: Marshall Dawson <marshalldawson3rd@gmail.com> Reviewed-on: https://review.coreboot.org/22937 Reviewed-by: Martin Roth <martinroth@google.com> Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
2017-12-18 23:56:44 +01:00
#endif
Reference in a new issue Copy permalink