diff options
| author | Felix Held <felix-coreboot@felixheld.de> | 2022-08-13 00:29:23 +0200 |
|---|---|---|
| committer | Felix Held <felix-coreboot@felixheld.de> | 2022-08-25 19:49:10 +0000 |
| commit | 199b10fc21ee094af88abb2b5856502565c40ae7 (patch) | |
| tree | c1966c7b7fce41327b49daa7d2cc8dc099b52739 /src/soc/amd/mendocino | |
| parent | b22bac893bd9e4718808c189a74353088cd840f7 (diff) | |
soc/amd: rework SPI flash MMIO region handling
Only 16 MByte of the SPI flash can be mapped right below the 4 GB
boundary.
In case of a larger SPI flash size, still only the 16 MByte region
starting at 0xff000000 can be configured as WRPROT and be reserved for
the MMIO mapped SPI flash region. The next 16 MByte MMIO region starting
at address 0xfe000000 contain for example the LAPIC MMIO region, the
ACPIMMIO region and the UART/I2C controller MMIO regions which shouldn't
be configured as WRPROT. Reserving this region for the MMIO mapped SPI
flash would also result in an overlap with the MMIO resources mentioned
above.
In the case of a smaller SPI flash, reserving the full 16 MByte flash
MMIO region makes sure that the resource allocator won't try to put
anything else in the lower parts of the 16 MByte SPI mapping region.
To avoid the issues described above, always reserve/cache the maximum
amount of 16 MBytes of flash that can be mapped below 4 GB.
TEST=On boards with 16 MByte SPI flash chips, the resulting image of a
timeless build doesn't change with this patch. Verified this on Chausie
(Mendocino), Majolica (Cezanne), Cereme (Picasso) and Google/Careena
(Stoneyridge). On Mandolin (Picasso) with an 8 MByte flash, the
resulting image of a timeless build is different, but neither the
coreboot console output nor the Linux dmesg output shows any errors that
might be related to this change.
Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: Ie12bd48e48e267a84dc494f67e8e0c7a4a01a320
Reviewed-on: https://review.coreboot.org/c/coreboot/+/66700
Reviewed-by: Martin Roth <martin.roth@amd.corp-partner.google.com>
Reviewed-by: Raul Rangel <rrangel@chromium.org>
Reviewed-by: Fred Reitberger <reitbergerfred@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Diffstat (limited to 'src/soc/amd/mendocino')
| -rw-r--r-- | src/soc/amd/mendocino/cpu.c | 4 |
1 files changed, 3 insertions, 1 deletions
diff --git a/src/soc/amd/mendocino/cpu.c b/src/soc/amd/mendocino/cpu.c index 0867595eb6..a0aa05ccbd 100644 --- a/src/soc/amd/mendocino/cpu.c +++ b/src/soc/amd/mendocino/cpu.c @@ -4,6 +4,7 @@ #include <acpi/acpi.h> #include <amdblocks/cpu.h> +#include <amdblocks/iomap.h> #include <amdblocks/mca.h> #include <amdblocks/reset.h> #include <amdblocks/smm.h> @@ -58,7 +59,8 @@ void mp_init_cpus(struct bus *cpu_bus) "mp_init_with_smm failed. Halting.\n"); /* pre_mp_init made the flash not cacheable. Reset to WP for performance. */ - mtrr_use_temp_range(FLASH_BASE_ADDR, CONFIG_ROM_SIZE, MTRR_TYPE_WRPROT); + mtrr_use_temp_range(FLASH_BELOW_4GB_MAPPING_REGION_BASE, + FLASH_BELOW_4GB_MAPPING_REGION_SIZE, MTRR_TYPE_WRPROT); /* SMMINFO only needs to be set up when booting from S5 */ if (!acpi_is_wakeup_s3()) |