github-mirror/linux
1
0
Fork 0
mirror of https://github.com/torvalds/linux.git synced 2026-07-28 10:09:10 +02:00
Code Issues Actions Packages Projects Releases Wiki Activity Graph

drm/nouveau/gsp: read MMU_LOCK to fix WPR placement on GA100

Browse Source 
On GA100, the row remapper hardware reserves a small amount of DRAM at
the end of framebuffer for spare rows used to repair memory errors at
runtime.  When an uncorrectable ECC error is detected in a DRAM row,
the row remapper redirects accesses to a spare row, transparently
repairing the fault.

The LOCAL_MEMORY_RANGE register (0x100ce0) reports the GPU's FB address
range, but its encoding rounds to 1GB boundaries.  On GA100, VBIOS
originally rounded this value down, which could lose up to ~1GB of
usable FB.  As a workaround, newer VBIOS instead rounds up to the next
1GB boundary and programs MMU_LOCK (registers 0x1fa82c/0x1fa830) to
mark the gap between the actual usable FB and the rounded-up range as
reserved.

OpenRM's kgspCalculateFbLayout_TU102() handles this by reading the
MMU_LOCK registers and computing the WPR top boundary as:

    vbiosReservedOffset = min(mmuLockLo, vgaWorkspaceOffset)

Without this, the WPR region is placed at the top of LOCAL_MEMORY_RANGE,
which overlaps the reserved region.  The booter firmware detects this
and rejects the WPR layout.

Add ga100_gsp_mmu_lock_lo() to read the MMU_LOCK range and clamp
gsp->fb.bios.addr accordingly, mirroring OpenRM's behavior.

This is a GA100-only issue.  GA102 and later add the
NV_USABLE_FB_SIZE_IN_MB register which reports the correct usable FB
size directly, eliminating the need for the MMU_LOCK workaround.

Signed-off-by: Timur Tabi <ttabi@nvidia.com>
Link: https://patch.msgid.link/20260430223838.2530778-5-ttabi@nvidia.com
Signed-off-by: Danilo Krummrich <dakr@kernel.org>
This commit is contained in:
Timur Tabi 2026-04-30 17:38:32 -05:00 committed by Danilo Krummrich
parent 3cee1a10b8
commit 0094a7a95d
1 changed files with 49 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

49
drivers/gpu/drm/nouveau/nvkm/subdev/gsp/tu102.c
View File

@ -291,6 +291,42 @@ tu102_gsp_vga_workspace_addr(struct nvkm_gsp *gsp, u64 fb_size)
return addr;
}
/*
* Read the MMU_LOCK range programmed by VBIOS.
*
* The row remapper reserves a small amount of DRAM at the end of FB for
* spare rows used to repair memory errors. The register used to report
* FB size rounds to 1GB boundaries, and VBIOS rounds up rather than
* down -- reporting a larger size than is actually usable. To
* compensate, VBIOS programs MMU_LOCK to fence off the unusable region
* at the top. We must read this and keep WPR below it.
*
* Returns the low address of the locked region, or 0 if not set.
*/
static u64
ga100_gsp_mmu_lock_lo(struct nvkm_gsp *gsp)
{
struct nvkm_device *device = gsp->subdev.device;
u32 lo, hi;
u64 addr_lo, addr_hi;
/* NV_PFB_PRI_MMU_LOCK_CFG_PRIV_LEVEL_MASK */
if (!(nvkm_rd32(device, 0x1fa7c8) & 0x00000001))
return 0;
lo = nvkm_rd32(device, 0x1fa82c); /* NV_PFB_PRI_MMU_LOCK_ADDR_LO */
hi = nvkm_rd32(device, 0x1fa830); /* NV_PFB_PRI_MMU_LOCK_ADDR_HI */
addr_lo = (u64)(lo >> 4) << 12;
addr_hi = (u64)(hi >> 4) << 12;
if (addr_hi < addr_lo)
return 0;
nvkm_debug(&gsp->subdev, "MMU_LOCK range: 0x%llx - 0x%llx\n", addr_lo, addr_hi);
return addr_lo;
}
int
tu102_gsp_oneinit(struct nvkm_gsp *gsp)
{
@ -304,6 +340,19 @@ tu102_gsp_oneinit(struct nvkm_gsp *gsp)
gsp->fb.bios.addr = gsp->fb.bios.vga_workspace.addr;
gsp->fb.bios.size = gsp->fb.bios.vga_workspace.size;
/*
* On GA100, VBIOS may lock the top of FB via MMU_LOCK to reserve
* space for the row remapper. Mirror OpenRM's kgspCalculateFbLayout
* by clamping bios.addr to min(mmuLockLo, vgaWorkspace) so that the
* entire WPR2 layout and gspFwWprEnd stay within usable FB.
*/
if (device->chipset == 0x170) {
u64 lock_lo = ga100_gsp_mmu_lock_lo(gsp);
if (lock_lo)
gsp->fb.bios.addr = min(gsp->fb.bios.addr, lock_lo);
}
ret = gsp->func->booter.ctor(gsp, "booter-load", gsp->fws.booter.load,
&device->sec2->falcon, &gsp->booter.load);
if (ret)