Dave Airlie
2024-Mar-28  02:43 UTC
[PATCH] nouveau/uvmm: fix addr/range calcs for remap operations
From: Dave Airlie <airlied at redhat.com>
dEQP-VK.sparse_resources.image_rebind.2d_array.r64i.128_128_8
was causing a remap operation like the below.
op_remap: prev: 0000003fffed0000 00000000000f0000 00000000a5abd18a
0000000000000000
op_remap: next:
op_remap: unmap: 0000003fffed0000 0000000000100000 0
op_map: map: 0000003ffffc0000 0000000000010000 000000005b1ba33c 00000000000e0000
This was resulting in an unmap operation from 0x3fffed0000+0xf0000, 0x100000
which was corrupting the pagetables and oopsing the kernel.
Fixes the prev + unmap range calcs to use start/end and map back to addr/range.
Signed-off-by: Dave Airlie <airlied at redhat.com>
Fixes: b88baab82871 ("drm/nouveau: implement new VM_BIND uAPI")
Cc: Danilo Krummrich <dakr at redhat.com>
---
 drivers/gpu/drm/nouveau/nouveau_uvmm.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/drivers/gpu/drm/nouveau/nouveau_uvmm.c
b/drivers/gpu/drm/nouveau/nouveau_uvmm.c
index 9675ef25b16d..87bce1a9d073 100644
--- a/drivers/gpu/drm/nouveau/nouveau_uvmm.c
+++ b/drivers/gpu/drm/nouveau/nouveau_uvmm.c
@@ -813,15 +813,15 @@ op_remap(struct drm_gpuva_op_remap *r,
 	struct drm_gpuva_op_unmap *u = r->unmap;
 	struct nouveau_uvma *uvma = uvma_from_va(u->va);
 	u64 addr = uvma->va.va.addr;
-	u64 range = uvma->va.va.range;
+	u64 end = uvma->va.va.addr + uvma->va.va.range;
 
 	if (r->prev)
 		addr = r->prev->va.addr + r->prev->va.range;
 
 	if (r->next)
-		range = r->next->va.addr - addr;
+		end = r->next->va.addr;
 
-	op_unmap_range(u, addr, range);
+	op_unmap_range(u, addr, end - addr);
 }
 
 static int
-- 
2.43.2
Danilo Krummrich
2024-Mar-28  17:07 UTC
[PATCH] nouveau/uvmm: fix addr/range calcs for remap operations
On 3/28/24 03:43, Dave Airlie wrote:> From: Dave Airlie <airlied at redhat.com> > > dEQP-VK.sparse_resources.image_rebind.2d_array.r64i.128_128_8 > was causing a remap operation like the below. > > op_remap: prev: 0000003fffed0000 00000000000f0000 00000000a5abd18a 0000000000000000 > op_remap: next: > op_remap: unmap: 0000003fffed0000 0000000000100000 0 > op_map: map: 0000003ffffc0000 0000000000010000 000000005b1ba33c 00000000000e0000 > > This was resulting in an unmap operation from 0x3fffed0000+0xf0000, 0x100000 > which was corrupting the pagetables and oopsing the kernel.Good catch, thanks for looking into that.> > Fixes the prev + unmap range calcs to use start/end and map back to addr/range.I like how using start/end instead fixes the issue and keeps it simple.> > Signed-off-by: Dave Airlie <airlied at redhat.com> > Fixes: b88baab82871 ("drm/nouveau: implement new VM_BIND uAPI") > Cc: Danilo Krummrich <dakr at redhat.com>Applied the patch to drm-misc-fixes.> --- > drivers/gpu/drm/nouveau/nouveau_uvmm.c | 6 +++--- > 1 file changed, 3 insertions(+), 3 deletions(-) > > diff --git a/drivers/gpu/drm/nouveau/nouveau_uvmm.c b/drivers/gpu/drm/nouveau/nouveau_uvmm.c > index 9675ef25b16d..87bce1a9d073 100644 > --- a/drivers/gpu/drm/nouveau/nouveau_uvmm.c > +++ b/drivers/gpu/drm/nouveau/nouveau_uvmm.c > @@ -813,15 +813,15 @@ op_remap(struct drm_gpuva_op_remap *r, > struct drm_gpuva_op_unmap *u = r->unmap; > struct nouveau_uvma *uvma = uvma_from_va(u->va); > u64 addr = uvma->va.va.addr; > - u64 range = uvma->va.va.range; > + u64 end = uvma->va.va.addr + uvma->va.va.range; > > if (r->prev) > addr = r->prev->va.addr + r->prev->va.range; > > if (r->next) > - range = r->next->va.addr - addr; > + end = r->next->va.addr; > > - op_unmap_range(u, addr, range); > + op_unmap_range(u, addr, end - addr); > } > > static int
Seemingly Similar Threads
- [PATCH drm-misc-next 0/3] [RFC] DRM GPUVA Manager GPU-VM features
- [PATCH] nouveau: set placement to original placement on uvmm validate.
- [PATCH drm-misc-next v9 09/12] drm/gpuvm: reference count drm_gpuvm structures
- [PATCH 1/2] drm/nouveau: don't fini scheduler if not initialized
- [PATCH] drm/nouveau: Constify struct nouveau_job_ops