diff options
| author | Leah Rowe <leah@libreboot.org> | 2024-11-01 15:59:30 +0000 |
|---|---|---|
| committer | Leah Rowe <leah@libreboot.org> | 2024-11-01 15:59:30 +0000 |
| commit | 14b4838d49556eb544c0f6a96d72128c3480ec2c (patch) | |
| tree | 71cc7fb04cd7871ddfc4377b9653990a3d712578 /config/coreboot/default/patches/0026-nb-haswell-lock-policy-regs-when-disabling-IOMMU.patch | |
| parent | 67c92889a866d67132bcc37de6444e0bc56f548c (diff) | |
There were a lot of unnecessary patches, such as the VRAM
patches; as Nicholas Chin has explained to me, the drivers
for these machines will just allocate what RAM they want
anyway, so in a lot of cases the extra allocated Video RAM
simply reduces the total amount of memory for other uses.
In general, we have a lot of patches that have existed for
years. A much more aggressive sweep will be done in the next
major audit, especially when the revisions are updated again.
Signed-off-by: Leah Rowe <leah@libreboot.org>
Diffstat (limited to 'config/coreboot/default/patches/0026-nb-haswell-lock-policy-regs-when-disabling-IOMMU.patch')
| -rw-r--r-- | config/coreboot/default/patches/0026-nb-haswell-lock-policy-regs-when-disabling-IOMMU.patch | 55 |
1 files changed, 55 insertions, 0 deletions
diff --git a/config/coreboot/default/patches/0026-nb-haswell-lock-policy-regs-when-disabling-IOMMU.patch b/config/coreboot/default/patches/0026-nb-haswell-lock-policy-regs-when-disabling-IOMMU.patch new file mode 100644 index 00000000..fe9034b0 --- /dev/null +++ b/config/coreboot/default/patches/0026-nb-haswell-lock-policy-regs-when-disabling-IOMMU.patch @@ -0,0 +1,55 @@ +From ce7d65790b9b8656ebbaa0ca715adff6a9c25588 Mon Sep 17 00:00:00 2001 +From: Leah Rowe <info@minifree.org> +Date: Sat, 4 May 2024 02:00:53 +0100 +Subject: [PATCH 26/51] nb/haswell: lock policy regs when disabling IOMMU + +Angel Pons told me I should do it. See comments here: +https://review.coreboot.org/c/coreboot/+/81016 + +I see no harm in complying with the request. I'll merge +this into the main patch at a later date and try to +get this upstreamed. + +Just a reminder: on Optiplex 9020 variants, Xorg locks up +under Linux when tested with a graphics card; disabling +IOMMU works around the issue. Intel graphics work just fine +with IOMMU turned on. Libreboot disables IOMMU by default, +on the 9020, so that users can install graphics cards easily. + +Signed-off-by: Leah Rowe <info@minifree.org> +--- + src/northbridge/intel/haswell/early_init.c | 15 +++++++-------- + 1 file changed, 7 insertions(+), 8 deletions(-) + +diff --git a/src/northbridge/intel/haswell/early_init.c b/src/northbridge/intel/haswell/early_init.c +index 1a7e0b1076..e9506ee830 100644 +--- a/src/northbridge/intel/haswell/early_init.c ++++ b/src/northbridge/intel/haswell/early_init.c +@@ -160,17 +160,16 @@ static void haswell_setup_iommu(void) + const u32 capid0_a = pci_read_config32(HOST_BRIDGE, CAPID0_A); + u8 enable_iommu = get_uint_option("iommu", 1); + +- if (!enable_iommu) +- return; +- + if (capid0_a & VTD_DISABLE) + return; + +- /* Setup BARs: zeroize top 32 bits; set enable bit */ +- mchbar_write32(GFXVTBAR + 4, GFXVT_BASE_ADDRESS >> 32); +- mchbar_write32(GFXVTBAR + 0, GFXVT_BASE_ADDRESS | 1); +- mchbar_write32(VTVC0BAR + 4, VTVC0_BASE_ADDRESS >> 32); +- mchbar_write32(VTVC0BAR + 0, VTVC0_BASE_ADDRESS | 1); ++ if (enable_iommu) { ++ /* Setup BARs: zeroize top 32 bits; set enable bit */ ++ mchbar_write32(GFXVTBAR + 4, GFXVT_BASE_ADDRESS >> 32); ++ mchbar_write32(GFXVTBAR + 0, GFXVT_BASE_ADDRESS | 1); ++ mchbar_write32(VTVC0BAR + 4, VTVC0_BASE_ADDRESS >> 32); ++ mchbar_write32(VTVC0BAR + 0, VTVC0_BASE_ADDRESS | 1); ++ } + + /* Set L3HIT2PEND_DIS, lock GFXVTBAR policy config registers */ + u32 reg32; +-- +2.39.5 + |