* [PATCH] drm/amdgpu: fix AGP addressing when GART is not at 0
@ 2023-11-10 14:47 Alex Deucher
2023-11-10 14:50 ` Christian König
0 siblings, 1 reply; 12+ messages in thread
From: Alex Deucher @ 2023-11-10 14:47 UTC (permalink / raw)
To: amd-gfx; +Cc: Alex Deucher, Yifan Zhang, Jesse Zhang, christian.koenig
This worked by luck if the GART aperture ended up at 0. When
we ended up moving GART on some chips, the GART aperture ended
up offsetting the the AGP address since the resource->start is
a GART offset, not an MC address. Fix this by moving the AGP
address setup into amdgpu_bo_gpu_offset_no_check().
Reported-by: Jesse Zhang <Jesse.Zhang@amd.com>
Reported-by: Yifan Zhang <yifan1.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: christian.koenig@amd.com
---
drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 10 +++++++---
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 4 +---
2 files changed, 8 insertions(+), 6 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
index cef920a93924..1b3e97522838 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
@@ -1527,10 +1527,14 @@ u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo)
u64 amdgpu_bo_gpu_offset_no_check(struct amdgpu_bo *bo)
{
struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
- uint64_t offset;
+ uint64_t offset, addr;
- offset = (bo->tbo.resource->start << PAGE_SHIFT) +
- amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type);
+ addr = amdgpu_gmc_agp_addr(&bo->tbo);
+ if (addr != AMDGPU_BO_INVALID_OFFSET)
+ offset = addr;
+ else
+ offset = (bo->tbo.resource->start << PAGE_SHIFT) +
+ amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type);
return amdgpu_gmc_sign_extend(offset);
}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 05991c5c8ddb..ab4a762aed5b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -959,10 +959,8 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo)
return 0;
addr = amdgpu_gmc_agp_addr(bo);
- if (addr != AMDGPU_BO_INVALID_OFFSET) {
- bo->resource->start = addr >> PAGE_SHIFT;
+ if (addr != AMDGPU_BO_INVALID_OFFSET)
return 0;
- }
/* allocate GART space */
placement.num_placement = 1;
--
2.41.0
^ permalink raw reply related [flat|nested] 12+ messages in thread
* Re: [PATCH] drm/amdgpu: fix AGP addressing when GART is not at 0
2023-11-10 14:47 [PATCH] drm/amdgpu: fix AGP addressing when GART is not at 0 Alex Deucher
@ 2023-11-10 14:50 ` Christian König
2023-11-12 18:13 ` Alex Deucher
0 siblings, 1 reply; 12+ messages in thread
From: Christian König @ 2023-11-10 14:50 UTC (permalink / raw)
To: Alex Deucher, amd-gfx; +Cc: Yifan Zhang, Jesse Zhang
Am 10.11.23 um 15:47 schrieb Alex Deucher:
> This worked by luck if the GART aperture ended up at 0. When
> we ended up moving GART on some chips, the GART aperture ended
> up offsetting the the AGP address since the resource->start is
> a GART offset, not an MC address. Fix this by moving the AGP
> address setup into amdgpu_bo_gpu_offset_no_check().
>
> Reported-by: Jesse Zhang <Jesse.Zhang@amd.com>
> Reported-by: Yifan Zhang <yifan1.zhang@amd.com>
> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
> Cc: christian.koenig@amd.com
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 10 +++++++---
> drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 4 +---
> 2 files changed, 8 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> index cef920a93924..1b3e97522838 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> @@ -1527,10 +1527,14 @@ u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo)
> u64 amdgpu_bo_gpu_offset_no_check(struct amdgpu_bo *bo)
> {
> struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
> - uint64_t offset;
> + uint64_t offset, addr;
>
> - offset = (bo->tbo.resource->start << PAGE_SHIFT) +
> - amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type);
> + addr = amdgpu_gmc_agp_addr(&bo->tbo);
IIRC you must check bo->tbo.resource->mem_type before calling
amdgpu_gmc_agp_addr().
Regards,
Christian.
> + if (addr != AMDGPU_BO_INVALID_OFFSET)
> + offset = addr;
> + else
> + offset = (bo->tbo.resource->start << PAGE_SHIFT) +
> + amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type);
>
> return amdgpu_gmc_sign_extend(offset);
> }
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> index 05991c5c8ddb..ab4a762aed5b 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> @@ -959,10 +959,8 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo)
> return 0;
>
> addr = amdgpu_gmc_agp_addr(bo);
> - if (addr != AMDGPU_BO_INVALID_OFFSET) {
> - bo->resource->start = addr >> PAGE_SHIFT;
> + if (addr != AMDGPU_BO_INVALID_OFFSET)
> return 0;
> - }
>
> /* allocate GART space */
> placement.num_placement = 1;
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH] drm/amdgpu: fix AGP addressing when GART is not at 0
2023-11-10 14:50 ` Christian König
@ 2023-11-12 18:13 ` Alex Deucher
2023-11-16 9:37 ` Zhang, Yifan
0 siblings, 1 reply; 12+ messages in thread
From: Alex Deucher @ 2023-11-12 18:13 UTC (permalink / raw)
To: Christian König; +Cc: Alex Deucher, Yifan Zhang, Jesse Zhang, amd-gfx
On Sat, Nov 11, 2023 at 2:17 AM Christian König
<christian.koenig@amd.com> wrote:
>
> Am 10.11.23 um 15:47 schrieb Alex Deucher:
> > This worked by luck if the GART aperture ended up at 0. When
> > we ended up moving GART on some chips, the GART aperture ended
> > up offsetting the the AGP address since the resource->start is
> > a GART offset, not an MC address. Fix this by moving the AGP
> > address setup into amdgpu_bo_gpu_offset_no_check().
> >
> > Reported-by: Jesse Zhang <Jesse.Zhang@amd.com>
> > Reported-by: Yifan Zhang <yifan1.zhang@amd.com>
> > Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
> > Cc: christian.koenig@amd.com
> > ---
> > drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 10 +++++++---
> > drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 4 +---
> > 2 files changed, 8 insertions(+), 6 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> > index cef920a93924..1b3e97522838 100644
> > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> > @@ -1527,10 +1527,14 @@ u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo)
> > u64 amdgpu_bo_gpu_offset_no_check(struct amdgpu_bo *bo)
> > {
> > struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
> > - uint64_t offset;
> > + uint64_t offset, addr;
> >
> > - offset = (bo->tbo.resource->start << PAGE_SHIFT) +
> > - amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type);
> > + addr = amdgpu_gmc_agp_addr(&bo->tbo);
>
> IIRC you must check bo->tbo.resource->mem_type before calling
> amdgpu_gmc_agp_addr().
Yes, this was fixed in v2.
Alex
>
> Regards,
> Christian.
>
> > + if (addr != AMDGPU_BO_INVALID_OFFSET)
> > + offset = addr;
> > + else
> > + offset = (bo->tbo.resource->start << PAGE_SHIFT) +
> > + amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type);
> >
> > return amdgpu_gmc_sign_extend(offset);
> > }
> > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> > index 05991c5c8ddb..ab4a762aed5b 100644
> > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> > @@ -959,10 +959,8 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo)
> > return 0;
> >
> > addr = amdgpu_gmc_agp_addr(bo);
> > - if (addr != AMDGPU_BO_INVALID_OFFSET) {
> > - bo->resource->start = addr >> PAGE_SHIFT;
> > + if (addr != AMDGPU_BO_INVALID_OFFSET)
> > return 0;
> > - }
> >
> > /* allocate GART space */
> > placement.num_placement = 1;
>
^ permalink raw reply [flat|nested] 12+ messages in thread
* RE: [PATCH] drm/amdgpu: fix AGP addressing when GART is not at 0
2023-11-12 18:13 ` Alex Deucher
@ 2023-11-16 9:37 ` Zhang, Yifan
2023-11-16 14:25 ` Alex Deucher
0 siblings, 1 reply; 12+ messages in thread
From: Zhang, Yifan @ 2023-11-16 9:37 UTC (permalink / raw)
To: Alex Deucher, Koenig, Christian
Cc: Deucher, Alexander, Zhang, Jesse(Jie), amd-gfx
[AMD Official Use Only - General]
Ping... this patch seems still not merged.
Best Regards,
Yifan
-----Original Message-----
From: Alex Deucher <alexdeucher@gmail.com>
Sent: Monday, November 13, 2023 2:13 AM
To: Koenig, Christian <Christian.Koenig@amd.com>
Cc: Deucher, Alexander <Alexander.Deucher@amd.com>; amd-gfx@lists.freedesktop.org; Zhang, Yifan <Yifan1.Zhang@amd.com>; Zhang, Jesse(Jie) <Jesse.Zhang@amd.com>
Subject: Re: [PATCH] drm/amdgpu: fix AGP addressing when GART is not at 0
On Sat, Nov 11, 2023 at 2:17 AM Christian König <christian.koenig@amd.com> wrote:
>
> Am 10.11.23 um 15:47 schrieb Alex Deucher:
> > This worked by luck if the GART aperture ended up at 0. When we
> > ended up moving GART on some chips, the GART aperture ended up
> > offsetting the the AGP address since the resource->start is a GART
> > offset, not an MC address. Fix this by moving the AGP address setup
> > into amdgpu_bo_gpu_offset_no_check().
> >
> > Reported-by: Jesse Zhang <Jesse.Zhang@amd.com>
> > Reported-by: Yifan Zhang <yifan1.zhang@amd.com>
> > Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
> > Cc: christian.koenig@amd.com
> > ---
> > drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 10 +++++++---
> > drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 4 +---
> > 2 files changed, 8 insertions(+), 6 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> > b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> > index cef920a93924..1b3e97522838 100644
> > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> > @@ -1527,10 +1527,14 @@ u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo)
> > u64 amdgpu_bo_gpu_offset_no_check(struct amdgpu_bo *bo)
> > {
> > struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
> > - uint64_t offset;
> > + uint64_t offset, addr;
> >
> > - offset = (bo->tbo.resource->start << PAGE_SHIFT) +
> > - amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type);
> > + addr = amdgpu_gmc_agp_addr(&bo->tbo);
>
> IIRC you must check bo->tbo.resource->mem_type before calling
> amdgpu_gmc_agp_addr().
Yes, this was fixed in v2.
Alex
>
> Regards,
> Christian.
>
> > + if (addr != AMDGPU_BO_INVALID_OFFSET)
> > + offset = addr;
> > + else
> > + offset = (bo->tbo.resource->start << PAGE_SHIFT) +
> > + amdgpu_ttm_domain_start(adev,
> > + bo->tbo.resource->mem_type);
> >
> > return amdgpu_gmc_sign_extend(offset);
> > }
> > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> > b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> > index 05991c5c8ddb..ab4a762aed5b 100644
> > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> > @@ -959,10 +959,8 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo)
> > return 0;
> >
> > addr = amdgpu_gmc_agp_addr(bo);
> > - if (addr != AMDGPU_BO_INVALID_OFFSET) {
> > - bo->resource->start = addr >> PAGE_SHIFT;
> > + if (addr != AMDGPU_BO_INVALID_OFFSET)
> > return 0;
> > - }
> >
> > /* allocate GART space */
> > placement.num_placement = 1;
>
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH] drm/amdgpu: fix AGP addressing when GART is not at 0
2023-11-16 9:37 ` Zhang, Yifan
@ 2023-11-16 14:25 ` Alex Deucher
2023-11-16 15:05 ` Zhang, Yifan
0 siblings, 1 reply; 12+ messages in thread
From: Alex Deucher @ 2023-11-16 14:25 UTC (permalink / raw)
To: Zhang, Yifan
Cc: Deucher, Alexander, Zhang, Jesse(Jie), Koenig, Christian, amd-gfx
On Thu, Nov 16, 2023 at 4:37 AM Zhang, Yifan <Yifan1.Zhang@amd.com> wrote:
>
> [AMD Official Use Only - General]
>
> Ping... this patch seems still not merged.
>
Can you confirm it fixes the AGP issues you saw?
Alex
> Best Regards,
> Yifan
>
> -----Original Message-----
> From: Alex Deucher <alexdeucher@gmail.com>
> Sent: Monday, November 13, 2023 2:13 AM
> To: Koenig, Christian <Christian.Koenig@amd.com>
> Cc: Deucher, Alexander <Alexander.Deucher@amd.com>; amd-gfx@lists.freedesktop.org; Zhang, Yifan <Yifan1.Zhang@amd.com>; Zhang, Jesse(Jie) <Jesse.Zhang@amd.com>
> Subject: Re: [PATCH] drm/amdgpu: fix AGP addressing when GART is not at 0
>
> On Sat, Nov 11, 2023 at 2:17 AM Christian König <christian.koenig@amd.com> wrote:
> >
> > Am 10.11.23 um 15:47 schrieb Alex Deucher:
> > > This worked by luck if the GART aperture ended up at 0. When we
> > > ended up moving GART on some chips, the GART aperture ended up
> > > offsetting the the AGP address since the resource->start is a GART
> > > offset, not an MC address. Fix this by moving the AGP address setup
> > > into amdgpu_bo_gpu_offset_no_check().
> > >
> > > Reported-by: Jesse Zhang <Jesse.Zhang@amd.com>
> > > Reported-by: Yifan Zhang <yifan1.zhang@amd.com>
> > > Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
> > > Cc: christian.koenig@amd.com
> > > ---
> > > drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 10 +++++++---
> > > drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 4 +---
> > > 2 files changed, 8 insertions(+), 6 deletions(-)
> > >
> > > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> > > b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> > > index cef920a93924..1b3e97522838 100644
> > > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> > > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> > > @@ -1527,10 +1527,14 @@ u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo)
> > > u64 amdgpu_bo_gpu_offset_no_check(struct amdgpu_bo *bo)
> > > {
> > > struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
> > > - uint64_t offset;
> > > + uint64_t offset, addr;
> > >
> > > - offset = (bo->tbo.resource->start << PAGE_SHIFT) +
> > > - amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type);
> > > + addr = amdgpu_gmc_agp_addr(&bo->tbo);
> >
> > IIRC you must check bo->tbo.resource->mem_type before calling
> > amdgpu_gmc_agp_addr().
>
> Yes, this was fixed in v2.
>
> Alex
>
> >
> > Regards,
> > Christian.
> >
> > > + if (addr != AMDGPU_BO_INVALID_OFFSET)
> > > + offset = addr;
> > > + else
> > > + offset = (bo->tbo.resource->start << PAGE_SHIFT) +
> > > + amdgpu_ttm_domain_start(adev,
> > > + bo->tbo.resource->mem_type);
> > >
> > > return amdgpu_gmc_sign_extend(offset);
> > > }
> > > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> > > b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> > > index 05991c5c8ddb..ab4a762aed5b 100644
> > > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> > > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> > > @@ -959,10 +959,8 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo)
> > > return 0;
> > >
> > > addr = amdgpu_gmc_agp_addr(bo);
> > > - if (addr != AMDGPU_BO_INVALID_OFFSET) {
> > > - bo->resource->start = addr >> PAGE_SHIFT;
> > > + if (addr != AMDGPU_BO_INVALID_OFFSET)
> > > return 0;
> > > - }
> > >
> > > /* allocate GART space */
> > > placement.num_placement = 1;
> >
^ permalink raw reply [flat|nested] 12+ messages in thread
* RE: [PATCH] drm/amdgpu: fix AGP addressing when GART is not at 0
2023-11-16 14:25 ` Alex Deucher
@ 2023-11-16 15:05 ` Zhang, Yifan
0 siblings, 0 replies; 12+ messages in thread
From: Zhang, Yifan @ 2023-11-16 15:05 UTC (permalink / raw)
To: Alex Deucher
Cc: Deucher, Alexander, Zhang, Jesse(Jie), Koenig, Christian, amd-gfx
[AMD Official Use Only - General]
Yes, it fixes regressions in KFDTest introduced by this commit ("b93ed51c32ca drm/amdgpu: fix AGP init order"),
e.g. KFDMemoryTest.MemoryRegister failure:
fault addr 0x00008084575a6000 is calculated by (gart_start + AGP aperture mc addr) wrongly.
[ 46.662856] amdgpu 0000:c2:00.0: amdgpu: [gfxhub] page fault (src_id:0 ring:169 vmid:0 pasid:0, for process pid 0 thread pid 0)
[ 46.662890] amdgpu 0000:c2:00.0: amdgpu: in page starting at address 0x00008084575a6000 from client 10
[ 46.662909] amdgpu 0000:c2:00.0: amdgpu: GCVM_L2_PROTECTION_FAULT_STATUS:0x00040B52
[ 46.662923] amdgpu 0000:c2:00.0: amdgpu: Faulty UTCL2 client ID: CPC (0x5)
[ 46.662936] amdgpu 0000:c2:00.0: amdgpu: MORE_FAULTS: 0x0
[ 46.662947] amdgpu 0000:c2:00.0: amdgpu: WALKER_ERROR: 0x1
[ 46.662957] amdgpu 0000:c2:00.0: amdgpu: PERMISSION_FAULTS: 0x5
[ 46.662968] amdgpu 0000:c2:00.0: amdgpu: MAPPING_ERROR: 0x1
[ 46.662979] amdgpu 0000:c2:00.0: amdgpu: RW: 0x1
-----Original Message-----
From: Alex Deucher <alexdeucher@gmail.com>
Sent: Thursday, November 16, 2023 10:26 PM
To: Zhang, Yifan <Yifan1.Zhang@amd.com>
Cc: Koenig, Christian <Christian.Koenig@amd.com>; Deucher, Alexander <Alexander.Deucher@amd.com>; amd-gfx@lists.freedesktop.org; Zhang, Jesse(Jie) <Jesse.Zhang@amd.com>
Subject: Re: [PATCH] drm/amdgpu: fix AGP addressing when GART is not at 0
On Thu, Nov 16, 2023 at 4:37 AM Zhang, Yifan <Yifan1.Zhang@amd.com> wrote:
>
> [AMD Official Use Only - General]
>
> Ping... this patch seems still not merged.
>
Can you confirm it fixes the AGP issues you saw?
Alex
> Best Regards,
> Yifan
>
> -----Original Message-----
> From: Alex Deucher <alexdeucher@gmail.com>
> Sent: Monday, November 13, 2023 2:13 AM
> To: Koenig, Christian <Christian.Koenig@amd.com>
> Cc: Deucher, Alexander <Alexander.Deucher@amd.com>;
> amd-gfx@lists.freedesktop.org; Zhang, Yifan <Yifan1.Zhang@amd.com>;
> Zhang, Jesse(Jie) <Jesse.Zhang@amd.com>
> Subject: Re: [PATCH] drm/amdgpu: fix AGP addressing when GART is not
> at 0
>
> On Sat, Nov 11, 2023 at 2:17 AM Christian König <christian.koenig@amd.com> wrote:
> >
> > Am 10.11.23 um 15:47 schrieb Alex Deucher:
> > > This worked by luck if the GART aperture ended up at 0. When we
> > > ended up moving GART on some chips, the GART aperture ended up
> > > offsetting the the AGP address since the resource->start is a GART
> > > offset, not an MC address. Fix this by moving the AGP address
> > > setup into amdgpu_bo_gpu_offset_no_check().
> > >
> > > Reported-by: Jesse Zhang <Jesse.Zhang@amd.com>
> > > Reported-by: Yifan Zhang <yifan1.zhang@amd.com>
> > > Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
> > > Cc: christian.koenig@amd.com
> > > ---
> > > drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 10 +++++++---
> > > drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 4 +---
> > > 2 files changed, 8 insertions(+), 6 deletions(-)
> > >
> > > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> > > b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> > > index cef920a93924..1b3e97522838 100644
> > > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> > > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> > > @@ -1527,10 +1527,14 @@ u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo)
> > > u64 amdgpu_bo_gpu_offset_no_check(struct amdgpu_bo *bo)
> > > {
> > > struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
> > > - uint64_t offset;
> > > + uint64_t offset, addr;
> > >
> > > - offset = (bo->tbo.resource->start << PAGE_SHIFT) +
> > > - amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type);
> > > + addr = amdgpu_gmc_agp_addr(&bo->tbo);
> >
> > IIRC you must check bo->tbo.resource->mem_type before calling
> > amdgpu_gmc_agp_addr().
>
> Yes, this was fixed in v2.
>
> Alex
>
> >
> > Regards,
> > Christian.
> >
> > > + if (addr != AMDGPU_BO_INVALID_OFFSET)
> > > + offset = addr;
> > > + else
> > > + offset = (bo->tbo.resource->start << PAGE_SHIFT) +
> > > + amdgpu_ttm_domain_start(adev,
> > > + bo->tbo.resource->mem_type);
> > >
> > > return amdgpu_gmc_sign_extend(offset);
> > > }
> > > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> > > b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> > > index 05991c5c8ddb..ab4a762aed5b 100644
> > > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> > > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> > > @@ -959,10 +959,8 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo)
> > > return 0;
> > >
> > > addr = amdgpu_gmc_agp_addr(bo);
> > > - if (addr != AMDGPU_BO_INVALID_OFFSET) {
> > > - bo->resource->start = addr >> PAGE_SHIFT;
> > > + if (addr != AMDGPU_BO_INVALID_OFFSET)
> > > return 0;
> > > - }
> > >
> > > /* allocate GART space */
> > > placement.num_placement = 1;
> >
^ permalink raw reply [flat|nested] 12+ messages in thread
* [PATCH] drm/amdgpu: fix AGP addressing when GART is not at 0
@ 2023-11-20 18:49 Alex Deucher
0 siblings, 0 replies; 12+ messages in thread
From: Alex Deucher @ 2023-11-20 18:49 UTC (permalink / raw)
To: amd-gfx
Cc: Alex Deucher, Yifan Zhang, Jesse Zhang, Christian König,
mario.limonciello
This worked by luck if the GART aperture ended up at 0. When
we ended up moving GART on some chips, the GART aperture ended
up offsetting the the AGP address since the resource->start is
a GART offset, not an MC address. Fix this by moving the AGP
address setup into amdgpu_bo_gpu_offset_no_check().
v2: check mem_type before checking agp
v3: check if the ttm bo has a ttm_tt allocated yet
Fixes: 67318cb84341 ("drm/amdgpu/gmc11: set gart placement GC11")
Reported-by: Jesse Zhang <Jesse.Zhang@amd.com>
Reported-by: Yifan Zhang <yifan1.zhang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: christian.koenig@amd.com
Cc: mario.limonciello@amd.com
---
drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 3 +++
drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 10 +++++++---
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 4 +---
3 files changed, 11 insertions(+), 6 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
index 5f71414190e9..d2f273d77e59 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
@@ -181,6 +181,9 @@ uint64_t amdgpu_gmc_agp_addr(struct ttm_buffer_object *bo)
{
struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev);
+ if (!bo->ttm)
+ return AMDGPU_BO_INVALID_OFFSET;
+
if (bo->ttm->num_pages != 1 || bo->ttm->caching == ttm_cached)
return AMDGPU_BO_INVALID_OFFSET;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
index c938648de853..c24f7b2c04c1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
@@ -1538,10 +1538,14 @@ u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo)
u64 amdgpu_bo_gpu_offset_no_check(struct amdgpu_bo *bo)
{
struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
- uint64_t offset;
+ uint64_t offset = AMDGPU_BO_INVALID_OFFSET;
- offset = (bo->tbo.resource->start << PAGE_SHIFT) +
- amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type);
+ if (bo->tbo.resource->mem_type == TTM_PL_TT)
+ offset = amdgpu_gmc_agp_addr(&bo->tbo);
+
+ if (offset == AMDGPU_BO_INVALID_OFFSET)
+ offset = (bo->tbo.resource->start << PAGE_SHIFT) +
+ amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type);
return amdgpu_gmc_sign_extend(offset);
}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 05991c5c8ddb..ab4a762aed5b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -959,10 +959,8 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo)
return 0;
addr = amdgpu_gmc_agp_addr(bo);
- if (addr != AMDGPU_BO_INVALID_OFFSET) {
- bo->resource->start = addr >> PAGE_SHIFT;
+ if (addr != AMDGPU_BO_INVALID_OFFSET)
return 0;
- }
/* allocate GART space */
placement.num_placement = 1;
--
2.41.0
^ permalink raw reply related [flat|nested] 12+ messages in thread
* Re: [PATCH] drm/amdgpu: fix AGP addressing when GART is not at 0
2023-11-17 19:24 ` Alex Deucher
@ 2023-11-20 16:16 ` Christian König
0 siblings, 0 replies; 12+ messages in thread
From: Christian König @ 2023-11-20 16:16 UTC (permalink / raw)
To: Alex Deucher, Christian König
Cc: Alex Deucher, Yifan Zhang, Jesse Zhang, amd-gfx
Am 17.11.23 um 20:24 schrieb Alex Deucher:
> On Fri, Nov 10, 2023 at 10:22 AM Christian König
> <christian.koenig@amd.com> wrote:
>> Am 10.11.23 um 16:02 schrieb Alex Deucher:
>>> This worked by luck if the GART aperture ended up at 0. When
>>> we ended up moving GART on some chips, the GART aperture ended
>>> up offsetting the the AGP address since the resource->start is
>>> a GART offset, not an MC address. Fix this by moving the AGP
>>> address setup into amdgpu_bo_gpu_offset_no_check().
>>>
>>> v2: check mem_type before checking agp
>>>
>>> Fixes: 67318cb84341 ("drm/amdgpu/gmc11: set gart placement GC11")
>>> Reported-by: Jesse Zhang <Jesse.Zhang@amd.com>
>>> Reported-by: Yifan Zhang <yifan1.zhang@amd.com>
>>> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
>>> Cc: christian.koenig@amd.com
>> Reviewed-by: Christian König <christian.koenig@amd.com>
> Mario is getting a segfault with this patch on PHX. Any ideas?
No idea, what does amdgpu_gmc_agp_addr+0x16 point to?
What could be is that the BO isn't backed by anything, e.g. GART not
initialized yet or something like that.
In this case we probably need to add some checks to amdgpu_gmc_agp_addr().
Regards,
Christian.
> I
> don't see how this could happen off hand.
>
> [ 28.980823] [drm] GART: num cpu pages 131072, num gpu pages 131072
> [ 28.980846] BUG: kernel NULL pointer dereference, address:
> 000000000000000c
> [ 28.981424] #PF: supervisor read access in kernel mode
> [ 28.981849] #PF: error_code(0x0000) - not-present page
> [ 28.982259] PGD 0 P4D 0
> [ 28.982469] Oops: 0000 [#1] PREEMPT SMP NOPTI
> [ 28.982817] CPU: 10 PID: 547 Comm: (udev-worker) Not tainted
> 6.7.0-rc1-00006-ge5e258131973 #175
> [ 28.984060] RIP: 0010:amdgpu_gmc_agp_addr+0x16/
> 0x60 [amdgpu]
> [ 28.984828] Code: 00 00 00 00 90 90 90 90 90 90 90 90 90 90 90 90 90
> 90 90 90 0f 1f 44 00 00 48 8b 87 88 01 00 00 49 b8 ff ff ff ff ff ff ff
> 7f <83> 78 0c 01 75 2e 83 78 28 02 74 28 48 8b 40 18 48 8b 97 60 01 00
> [ 28.984829] RSP: 0018:ffffc90000ff7998 EFLAGS: 00010282
> [ 28.984831] RAX: 0000000000000000 RBX: ffff888103f83800 RCX:
> 0000000000000000
> [ 28.984832] RDX: 7fffffffffffffff RSI: ffff888103f83858 RDI:
> ffff888103f83858
> [ 28.984832] RBP: ffff888102f0f020 R08: 7fffffffffffffff R09:
> 0000000000000000
> [ 28.984833] R10: 0000000000001000 R11: 0000000000001000 R12:
> ffff888103f83800
> [ 28.984834] R13: ffff888102f0fdf8 R14: ffff888102f0fe00 R15:
> 0000000000000000
> [ 28.984835] FS: 00007fe6aa26a8c0(0000) GS:ffff88844e880000(0000)
> knlGS:0000000000000000
> [ 28.984836] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> [ 28.984837] CR2: 000000000000000c CR3: 000000011556e000 CR4:
> 0000000000750ef0
> [ 28.984838] PKRU: 55555554
> [ 28.984839] Call Trace:
> [ 28.984842] <TASK>
> [ 28.984845] ? __die+0x20/0x70
> [ 28.984850] ? page_fault_oops+0x151/0x4b0
> [ 28.984854] ? srso_alias_return_thunk+0x5/0xfbef5
> [ 28.992663] ? do_user_addr_fault+0x65/0x6b0
> [ 28.992672] ? exc_page_fault+0x74/0x170
> [ 28.992676] ? asm_exc_page_fault+0x22/0x30
> [ 28.993714] ? amdgpu_gmc_agp_addr+0x16/0x60 [amdgpu]
> [ 28.994455] amdgpu_bo_gpu_offset_no_check+0x1a/0x70 [amdgpu]
> [ 28.995110] amdgpu_bo_create_reserved.part.0+0x109/0x290 [amdgpu]
> [ 28.995786] ? __pfx_amdgpu_bo_destroy+0x10/0x10 [amdgpu]
> [ 28.996400] amdgpu_bo_create_kernel+0x3f/0xa0 [amdgpu]
> [ 28.996992] amdgpu_device_init+0x15fa/0x2b60 [amdgpu]
> [ 28.997591] ? pci_bus_read_config_word+0x46/0x80
> [ 28.997598] ? srso_alias_return_thunk+0x5/0xfbef5
> [ 28.998385] ? do_pci_enable_device+0xd4/0x100
> [ 28.998390] amdgpu_driver_load_kms+0x15/0x190 [amdgpu]
> [ 28.999194] amdgpu_pci_probe+0x180/0x570 [amdgpu]
> [ 28.999781] ? srso_alias_return_thunk+0x5/0xfbef5
> [ 29.000232] local_pci_probe+0x43/0xa0
> [ 29.000552] pci_device_probe+0xc5/0x270
> [ 29.000883] really_probe+0x1b4/0x420
> [ 29.001194] __driver_probe_device+0x84/0x170
> [ 29.001558] driver_probe_device+0x1e/0xb0
> [ 29.001901] __driver_attach+0xe5/0x1f0
> [ 29.002224] ? __pfx___driver_attach+0x10/0x10
> [ 29.002594] bus_for_each_dev+0x75/0xd0
> [ 29.002919] bus_add_driver+0x112/0x220
> [ 29.003243] driver_register+0x5c/0x120
> [ 29.003569] ? __pfx_amdgpu_init+0x10/0x10 [amdgpu]
> [ 29.004148] do_one_initcall+0x41/0x300
> [ 29.004471] ? srso_alias_return_thunk+0x5/0xfbef5
> [ 29.004876] ? kmalloc_trace+0x25/0x90
> [ 29.005197] do_init_module+0x64/0x250
> [ 29.005201] init_module_from_file+0x8b/0xd0
> [ 29.005207] idempotent_init_module+0x181/0x240
> [ 29.006235] __x64_sys_finit_module+0x5a/0xb0
> [ 29.006238] do_syscall_64+0x5c/0xe0
> [ 29.006868] ? srso_alias_return_thunk+0x5/0xfbef5
> [ 29.006870] ? syscall_exit_to_user_mode+0x27/0x40
> [ 29.006872] ? srso_alias_return_thunk+0x5/0xfbef5
> [ 29.008000] ? do_syscall_64+0x6b/0xe0
> [ 29.008004] entry_SYSCALL_64_after_hwframe+0x6e/0x76
> [ 29.008703] RIP: 0033:0x7fe6aa125c7d
> [ 29.008991] Code: ff c3 66 2e 0f 1f 84 00 00 00 00 00 90 f3 0f 1e fa
> 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f
> 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 6b 81 0d 00 f7 d8 64 89 01 48
> [ 29.008993] RSP: 002b:00007ffe22be7618 EFLAGS: 00000246 ORIG_RAX:
> 0000000000000139
> [ 29.010434] RAX: ffffffffffffffda RBX: 000055b2f8dee0f0 RCX:
> 00007fe6aa125c7d
> [ 29.010435] RDX: 0000000000000000 RSI: 00007fe6aa33544a RDI:
> 0000000000000017
> [ 29.010436] RBP: 00007fe6aa33544a R08: 0000000000000040 R09:
> fffffffffffffde0
> [ 29.010438] R10: fffffffffffffe18 R11: 0000000000000246 R12:
> 0000000000020000
> [ 29.010439] R13: 000055b2f8f12ba0 R14: 0000000000000000 R15:
> 000055b2f8f49ef0
> [ 29.010443] </TASK>
> [ 29.010444] Modules linked in: binfmt_misc squashfs vfat fat
> snd_hda_codec_hdmi intel_rapl_msr intel_rapl_common amdgpu(+)
> snd_soc_dmic snd_ps_pdm_dma snd_soc_ps_mach snd_hda_intel
> snd_intel_dspcfg snd_soc_core snd_hda_codec snd_compress ac97_bus
> snd_pcm_dmaengine snd_hda_core snd_seq_dummy uvcvideo snd_hwdep uvc
> hid_sensor_magn_3d hid_sensor_gyro_3d snd_seq_oss videobuf2_vmalloc
> snd_pci_ps hid_sensor_trigger videobuf2_memops hid_sensor_iio_common
> snd_seq_midi snd_rpl_pci_acp6x snd_seq_midi_event
> industrialio_triggered_buffer kfifo_buf videobuf2_v4l2 snd_acp_pci
> edac_mce_amd industrialio crct10dif_pclmul videobuf2_common snd_rawmidi
> snd_acp_legacy_common polyval_clmulni amdxcp gpu_sched polyval_generic
> snd_pci_acp6x joydev i2c_algo_bit videodev drm_suballoc_helper
> drm_ttm_helper snd_seq snd_pcm ghash_clmulni_intel ttm sha512_ssse3
> snd_seq_device drm_exec sha256_ssse3 mc snd_timer snd_pci_acp5x
> drm_display_helper sha1_ssse3 snd_rn_pci_acp3x snd_acp_config wmi_bmof
> snd rapl cec snd_soc_acpi drm_buddy soundcore i2c_piix4
> [ 29.014048] snd_pci_acp3x k10temp amd_pmf platform_profile
> hid_sensor_hub amd_pmc hid_multitouch serio_raw kvm_amd ccp kvm
> irqbypass iptable_filter ip6table_filter ip6_tables br_netfilter bridge
> stp llc arp_tables parport_pc ppdev lp parport loop fuse ip_tables
> rtsx_pci_sdmmc mmc_core video nvme crc32_pclmul crc32c_intel thunderbolt
> nvme_core amd_sfh r8169 rtsx_pci wmi i2c_hid_acpi i2c_hid
> [ 29.023590] CR2: 000000000000000c
> [ 29.023856] ---[ end trace 0000000000000000 ]---
> [ 29.024220] RIP: 0010:amdgpu_gmc_agp_addr+0x16/0x60 [amdgpu]
> [ 29.024886] Code: 00 00 00 00 90 90 90 90 90 90 90 90 90 90 90 90 90
> 90 90 90 0f 1f 44 00 00 48 8b 87 88 01 00 00 49 b8 ff ff ff ff ff ff ff
> 7f <83> 78 0c 01 75 2e 83 78 28 02 74 28 48 8b 40 18 48 8b 97 60 01 00
> [ 29.026324] RSP: 0018:ffffc90000ff7998 EFLAGS: 00010282
> [ 29.026737] RAX: 0000000000000000 RBX: ffff888103f83800 RCX:
> 0000000000000000
> [ 29.027290] RDX: 7fffffffffffffff RSI: ffff888103f83858 RDI:
> ffff888103f83858
> [ 29.027848] RBP: ffff888102f0f020 R08: 7fffffffffffffff R09:
> 0000000000000000
> [ 29.028401] R10: 0000000000001000 R11: 0000000000001000 R12:
> ffff888103f83800
> [ 29.028960] R13: ffff888102f0fdf8 R14: ffff888102f0fe00 R15:
> 0000000000000000
> [ 29.029521] FS: 00007fe6aa26a8c0(0000) GS:ffff88844e880000(0000)
> knlGS:0000000000000000
> [ 29.030147] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> [ 29.030600] CR2: 000000000000000c CR3: 000000011556e000 CR4:
> 0000000000750ef0
>
>>> ---
>>> drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 10 +++++++---
>>> drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 4 +---
>>> 2 files changed, 8 insertions(+), 6 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
>>> index cef920a93924..d79b4ca1ecfc 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
>>> @@ -1527,10 +1527,14 @@ u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo)
>>> u64 amdgpu_bo_gpu_offset_no_check(struct amdgpu_bo *bo)
>>> {
>>> struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
>>> - uint64_t offset;
>>> + uint64_t offset = AMDGPU_BO_INVALID_OFFSET;
>>>
>>> - offset = (bo->tbo.resource->start << PAGE_SHIFT) +
>>> - amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type);
>>> + if (bo->tbo.resource->mem_type == TTM_PL_TT)
>>> + offset = amdgpu_gmc_agp_addr(&bo->tbo);
>>> +
>>> + if (offset == AMDGPU_BO_INVALID_OFFSET)
>>> + offset = (bo->tbo.resource->start << PAGE_SHIFT) +
>>> + amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type);
>>>
>>> return amdgpu_gmc_sign_extend(offset);
>>> }
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>>> index 05991c5c8ddb..ab4a762aed5b 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
>>> @@ -959,10 +959,8 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo)
>>> return 0;
>>>
>>> addr = amdgpu_gmc_agp_addr(bo);
>>> - if (addr != AMDGPU_BO_INVALID_OFFSET) {
>>> - bo->resource->start = addr >> PAGE_SHIFT;
>>> + if (addr != AMDGPU_BO_INVALID_OFFSET)
>>> return 0;
>>> - }
>>>
>>> /* allocate GART space */
>>> placement.num_placement = 1;
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH] drm/amdgpu: fix AGP addressing when GART is not at 0
2023-11-10 15:03 ` Christian König
@ 2023-11-17 19:24 ` Alex Deucher
2023-11-20 16:16 ` Christian König
0 siblings, 1 reply; 12+ messages in thread
From: Alex Deucher @ 2023-11-17 19:24 UTC (permalink / raw)
To: Christian König; +Cc: Alex Deucher, Yifan Zhang, Jesse Zhang, amd-gfx
On Fri, Nov 10, 2023 at 10:22 AM Christian König
<christian.koenig@amd.com> wrote:
>
> Am 10.11.23 um 16:02 schrieb Alex Deucher:
> > This worked by luck if the GART aperture ended up at 0. When
> > we ended up moving GART on some chips, the GART aperture ended
> > up offsetting the the AGP address since the resource->start is
> > a GART offset, not an MC address. Fix this by moving the AGP
> > address setup into amdgpu_bo_gpu_offset_no_check().
> >
> > v2: check mem_type before checking agp
> >
> > Fixes: 67318cb84341 ("drm/amdgpu/gmc11: set gart placement GC11")
> > Reported-by: Jesse Zhang <Jesse.Zhang@amd.com>
> > Reported-by: Yifan Zhang <yifan1.zhang@amd.com>
> > Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
> > Cc: christian.koenig@amd.com
>
> Reviewed-by: Christian König <christian.koenig@amd.com>
Mario is getting a segfault with this patch on PHX. Any ideas? I
don't see how this could happen off hand.
[ 28.980823] [drm] GART: num cpu pages 131072, num gpu pages 131072
[ 28.980846] BUG: kernel NULL pointer dereference, address:
000000000000000c
[ 28.981424] #PF: supervisor read access in kernel mode
[ 28.981849] #PF: error_code(0x0000) - not-present page
[ 28.982259] PGD 0 P4D 0
[ 28.982469] Oops: 0000 [#1] PREEMPT SMP NOPTI
[ 28.982817] CPU: 10 PID: 547 Comm: (udev-worker) Not tainted
6.7.0-rc1-00006-ge5e258131973 #175
[ 28.984060] RIP: 0010:amdgpu_gmc_agp_addr+0x16/
0x60 [amdgpu]
[ 28.984828] Code: 00 00 00 00 90 90 90 90 90 90 90 90 90 90 90 90 90
90 90 90 0f 1f 44 00 00 48 8b 87 88 01 00 00 49 b8 ff ff ff ff ff ff ff
7f <83> 78 0c 01 75 2e 83 78 28 02 74 28 48 8b 40 18 48 8b 97 60 01 00
[ 28.984829] RSP: 0018:ffffc90000ff7998 EFLAGS: 00010282
[ 28.984831] RAX: 0000000000000000 RBX: ffff888103f83800 RCX:
0000000000000000
[ 28.984832] RDX: 7fffffffffffffff RSI: ffff888103f83858 RDI:
ffff888103f83858
[ 28.984832] RBP: ffff888102f0f020 R08: 7fffffffffffffff R09:
0000000000000000
[ 28.984833] R10: 0000000000001000 R11: 0000000000001000 R12:
ffff888103f83800
[ 28.984834] R13: ffff888102f0fdf8 R14: ffff888102f0fe00 R15:
0000000000000000
[ 28.984835] FS: 00007fe6aa26a8c0(0000) GS:ffff88844e880000(0000)
knlGS:0000000000000000
[ 28.984836] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 28.984837] CR2: 000000000000000c CR3: 000000011556e000 CR4:
0000000000750ef0
[ 28.984838] PKRU: 55555554
[ 28.984839] Call Trace:
[ 28.984842] <TASK>
[ 28.984845] ? __die+0x20/0x70
[ 28.984850] ? page_fault_oops+0x151/0x4b0
[ 28.984854] ? srso_alias_return_thunk+0x5/0xfbef5
[ 28.992663] ? do_user_addr_fault+0x65/0x6b0
[ 28.992672] ? exc_page_fault+0x74/0x170
[ 28.992676] ? asm_exc_page_fault+0x22/0x30
[ 28.993714] ? amdgpu_gmc_agp_addr+0x16/0x60 [amdgpu]
[ 28.994455] amdgpu_bo_gpu_offset_no_check+0x1a/0x70 [amdgpu]
[ 28.995110] amdgpu_bo_create_reserved.part.0+0x109/0x290 [amdgpu]
[ 28.995786] ? __pfx_amdgpu_bo_destroy+0x10/0x10 [amdgpu]
[ 28.996400] amdgpu_bo_create_kernel+0x3f/0xa0 [amdgpu]
[ 28.996992] amdgpu_device_init+0x15fa/0x2b60 [amdgpu]
[ 28.997591] ? pci_bus_read_config_word+0x46/0x80
[ 28.997598] ? srso_alias_return_thunk+0x5/0xfbef5
[ 28.998385] ? do_pci_enable_device+0xd4/0x100
[ 28.998390] amdgpu_driver_load_kms+0x15/0x190 [amdgpu]
[ 28.999194] amdgpu_pci_probe+0x180/0x570 [amdgpu]
[ 28.999781] ? srso_alias_return_thunk+0x5/0xfbef5
[ 29.000232] local_pci_probe+0x43/0xa0
[ 29.000552] pci_device_probe+0xc5/0x270
[ 29.000883] really_probe+0x1b4/0x420
[ 29.001194] __driver_probe_device+0x84/0x170
[ 29.001558] driver_probe_device+0x1e/0xb0
[ 29.001901] __driver_attach+0xe5/0x1f0
[ 29.002224] ? __pfx___driver_attach+0x10/0x10
[ 29.002594] bus_for_each_dev+0x75/0xd0
[ 29.002919] bus_add_driver+0x112/0x220
[ 29.003243] driver_register+0x5c/0x120
[ 29.003569] ? __pfx_amdgpu_init+0x10/0x10 [amdgpu]
[ 29.004148] do_one_initcall+0x41/0x300
[ 29.004471] ? srso_alias_return_thunk+0x5/0xfbef5
[ 29.004876] ? kmalloc_trace+0x25/0x90
[ 29.005197] do_init_module+0x64/0x250
[ 29.005201] init_module_from_file+0x8b/0xd0
[ 29.005207] idempotent_init_module+0x181/0x240
[ 29.006235] __x64_sys_finit_module+0x5a/0xb0
[ 29.006238] do_syscall_64+0x5c/0xe0
[ 29.006868] ? srso_alias_return_thunk+0x5/0xfbef5
[ 29.006870] ? syscall_exit_to_user_mode+0x27/0x40
[ 29.006872] ? srso_alias_return_thunk+0x5/0xfbef5
[ 29.008000] ? do_syscall_64+0x6b/0xe0
[ 29.008004] entry_SYSCALL_64_after_hwframe+0x6e/0x76
[ 29.008703] RIP: 0033:0x7fe6aa125c7d
[ 29.008991] Code: ff c3 66 2e 0f 1f 84 00 00 00 00 00 90 f3 0f 1e fa
48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f
05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 6b 81 0d 00 f7 d8 64 89 01 48
[ 29.008993] RSP: 002b:00007ffe22be7618 EFLAGS: 00000246 ORIG_RAX:
0000000000000139
[ 29.010434] RAX: ffffffffffffffda RBX: 000055b2f8dee0f0 RCX:
00007fe6aa125c7d
[ 29.010435] RDX: 0000000000000000 RSI: 00007fe6aa33544a RDI:
0000000000000017
[ 29.010436] RBP: 00007fe6aa33544a R08: 0000000000000040 R09:
fffffffffffffde0
[ 29.010438] R10: fffffffffffffe18 R11: 0000000000000246 R12:
0000000000020000
[ 29.010439] R13: 000055b2f8f12ba0 R14: 0000000000000000 R15:
000055b2f8f49ef0
[ 29.010443] </TASK>
[ 29.010444] Modules linked in: binfmt_misc squashfs vfat fat
snd_hda_codec_hdmi intel_rapl_msr intel_rapl_common amdgpu(+)
snd_soc_dmic snd_ps_pdm_dma snd_soc_ps_mach snd_hda_intel
snd_intel_dspcfg snd_soc_core snd_hda_codec snd_compress ac97_bus
snd_pcm_dmaengine snd_hda_core snd_seq_dummy uvcvideo snd_hwdep uvc
hid_sensor_magn_3d hid_sensor_gyro_3d snd_seq_oss videobuf2_vmalloc
snd_pci_ps hid_sensor_trigger videobuf2_memops hid_sensor_iio_common
snd_seq_midi snd_rpl_pci_acp6x snd_seq_midi_event
industrialio_triggered_buffer kfifo_buf videobuf2_v4l2 snd_acp_pci
edac_mce_amd industrialio crct10dif_pclmul videobuf2_common snd_rawmidi
snd_acp_legacy_common polyval_clmulni amdxcp gpu_sched polyval_generic
snd_pci_acp6x joydev i2c_algo_bit videodev drm_suballoc_helper
drm_ttm_helper snd_seq snd_pcm ghash_clmulni_intel ttm sha512_ssse3
snd_seq_device drm_exec sha256_ssse3 mc snd_timer snd_pci_acp5x
drm_display_helper sha1_ssse3 snd_rn_pci_acp3x snd_acp_config wmi_bmof
snd rapl cec snd_soc_acpi drm_buddy soundcore i2c_piix4
[ 29.014048] snd_pci_acp3x k10temp amd_pmf platform_profile
hid_sensor_hub amd_pmc hid_multitouch serio_raw kvm_amd ccp kvm
irqbypass iptable_filter ip6table_filter ip6_tables br_netfilter bridge
stp llc arp_tables parport_pc ppdev lp parport loop fuse ip_tables
rtsx_pci_sdmmc mmc_core video nvme crc32_pclmul crc32c_intel thunderbolt
nvme_core amd_sfh r8169 rtsx_pci wmi i2c_hid_acpi i2c_hid
[ 29.023590] CR2: 000000000000000c
[ 29.023856] ---[ end trace 0000000000000000 ]---
[ 29.024220] RIP: 0010:amdgpu_gmc_agp_addr+0x16/0x60 [amdgpu]
[ 29.024886] Code: 00 00 00 00 90 90 90 90 90 90 90 90 90 90 90 90 90
90 90 90 0f 1f 44 00 00 48 8b 87 88 01 00 00 49 b8 ff ff ff ff ff ff ff
7f <83> 78 0c 01 75 2e 83 78 28 02 74 28 48 8b 40 18 48 8b 97 60 01 00
[ 29.026324] RSP: 0018:ffffc90000ff7998 EFLAGS: 00010282
[ 29.026737] RAX: 0000000000000000 RBX: ffff888103f83800 RCX:
0000000000000000
[ 29.027290] RDX: 7fffffffffffffff RSI: ffff888103f83858 RDI:
ffff888103f83858
[ 29.027848] RBP: ffff888102f0f020 R08: 7fffffffffffffff R09:
0000000000000000
[ 29.028401] R10: 0000000000001000 R11: 0000000000001000 R12:
ffff888103f83800
[ 29.028960] R13: ffff888102f0fdf8 R14: ffff888102f0fe00 R15:
0000000000000000
[ 29.029521] FS: 00007fe6aa26a8c0(0000) GS:ffff88844e880000(0000)
knlGS:0000000000000000
[ 29.030147] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 29.030600] CR2: 000000000000000c CR3: 000000011556e000 CR4:
0000000000750ef0
>
> > ---
> > drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 10 +++++++---
> > drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 4 +---
> > 2 files changed, 8 insertions(+), 6 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> > index cef920a93924..d79b4ca1ecfc 100644
> > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> > @@ -1527,10 +1527,14 @@ u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo)
> > u64 amdgpu_bo_gpu_offset_no_check(struct amdgpu_bo *bo)
> > {
> > struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
> > - uint64_t offset;
> > + uint64_t offset = AMDGPU_BO_INVALID_OFFSET;
> >
> > - offset = (bo->tbo.resource->start << PAGE_SHIFT) +
> > - amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type);
> > + if (bo->tbo.resource->mem_type == TTM_PL_TT)
> > + offset = amdgpu_gmc_agp_addr(&bo->tbo);
> > +
> > + if (offset == AMDGPU_BO_INVALID_OFFSET)
> > + offset = (bo->tbo.resource->start << PAGE_SHIFT) +
> > + amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type);
> >
> > return amdgpu_gmc_sign_extend(offset);
> > }
> > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> > index 05991c5c8ddb..ab4a762aed5b 100644
> > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> > @@ -959,10 +959,8 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo)
> > return 0;
> >
> > addr = amdgpu_gmc_agp_addr(bo);
> > - if (addr != AMDGPU_BO_INVALID_OFFSET) {
> > - bo->resource->start = addr >> PAGE_SHIFT;
> > + if (addr != AMDGPU_BO_INVALID_OFFSET)
> > return 0;
> > - }
> >
> > /* allocate GART space */
> > placement.num_placement = 1;
>
^ permalink raw reply [flat|nested] 12+ messages in thread
* RE: [PATCH] drm/amdgpu: fix AGP addressing when GART is not at 0
2023-11-10 15:02 Alex Deucher
2023-11-10 15:03 ` Christian König
@ 2023-11-11 3:18 ` Zhang, Yifan
1 sibling, 0 replies; 12+ messages in thread
From: Zhang, Yifan @ 2023-11-11 3:18 UTC (permalink / raw)
To: Deucher, Alexander, amd-gfx; +Cc: Zhang, Jesse(Jie), Koenig, Christian
[AMD Official Use Only - General]
Reviewed-by: Yifan Zhang <yifan1.zhang@amd.com>
-----Original Message-----
From: Deucher, Alexander <Alexander.Deucher@amd.com>
Sent: Friday, November 10, 2023 11:02 PM
To: amd-gfx@lists.freedesktop.org
Cc: Deucher, Alexander <Alexander.Deucher@amd.com>; Zhang, Jesse(Jie) <Jesse.Zhang@amd.com>; Zhang, Yifan <Yifan1.Zhang@amd.com>; Koenig, Christian <Christian.Koenig@amd.com>
Subject: [PATCH] drm/amdgpu: fix AGP addressing when GART is not at 0
This worked by luck if the GART aperture ended up at 0. When we ended up moving GART on some chips, the GART aperture ended up offsetting the the AGP address since the resource->start is a GART offset, not an MC address. Fix this by moving the AGP address setup into amdgpu_bo_gpu_offset_no_check().
v2: check mem_type before checking agp
Fixes: 67318cb84341 ("drm/amdgpu/gmc11: set gart placement GC11")
Reported-by: Jesse Zhang <Jesse.Zhang@amd.com>
Reported-by: Yifan Zhang <yifan1.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: christian.koenig@amd.com
---
drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 10 +++++++---
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 4 +---
2 files changed, 8 insertions(+), 6 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
index cef920a93924..d79b4ca1ecfc 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
@@ -1527,10 +1527,14 @@ u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo)
u64 amdgpu_bo_gpu_offset_no_check(struct amdgpu_bo *bo) {
struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
- uint64_t offset;
+ uint64_t offset = AMDGPU_BO_INVALID_OFFSET;
- offset = (bo->tbo.resource->start << PAGE_SHIFT) +
- amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type);
+ if (bo->tbo.resource->mem_type == TTM_PL_TT)
+ offset = amdgpu_gmc_agp_addr(&bo->tbo);
+
+ if (offset == AMDGPU_BO_INVALID_OFFSET)
+ offset = (bo->tbo.resource->start << PAGE_SHIFT) +
+ amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type);
return amdgpu_gmc_sign_extend(offset); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 05991c5c8ddb..ab4a762aed5b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -959,10 +959,8 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo)
return 0;
addr = amdgpu_gmc_agp_addr(bo);
- if (addr != AMDGPU_BO_INVALID_OFFSET) {
- bo->resource->start = addr >> PAGE_SHIFT;
+ if (addr != AMDGPU_BO_INVALID_OFFSET)
return 0;
- }
/* allocate GART space */
placement.num_placement = 1;
--
2.41.0
^ permalink raw reply related [flat|nested] 12+ messages in thread
* Re: [PATCH] drm/amdgpu: fix AGP addressing when GART is not at 0
2023-11-10 15:02 Alex Deucher
@ 2023-11-10 15:03 ` Christian König
2023-11-17 19:24 ` Alex Deucher
2023-11-11 3:18 ` Zhang, Yifan
1 sibling, 1 reply; 12+ messages in thread
From: Christian König @ 2023-11-10 15:03 UTC (permalink / raw)
To: Alex Deucher, amd-gfx; +Cc: Yifan Zhang, Jesse Zhang
Am 10.11.23 um 16:02 schrieb Alex Deucher:
> This worked by luck if the GART aperture ended up at 0. When
> we ended up moving GART on some chips, the GART aperture ended
> up offsetting the the AGP address since the resource->start is
> a GART offset, not an MC address. Fix this by moving the AGP
> address setup into amdgpu_bo_gpu_offset_no_check().
>
> v2: check mem_type before checking agp
>
> Fixes: 67318cb84341 ("drm/amdgpu/gmc11: set gart placement GC11")
> Reported-by: Jesse Zhang <Jesse.Zhang@amd.com>
> Reported-by: Yifan Zhang <yifan1.zhang@amd.com>
> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
> Cc: christian.koenig@amd.com
Reviewed-by: Christian König <christian.koenig@amd.com>
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 10 +++++++---
> drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 4 +---
> 2 files changed, 8 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> index cef920a93924..d79b4ca1ecfc 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> @@ -1527,10 +1527,14 @@ u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo)
> u64 amdgpu_bo_gpu_offset_no_check(struct amdgpu_bo *bo)
> {
> struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
> - uint64_t offset;
> + uint64_t offset = AMDGPU_BO_INVALID_OFFSET;
>
> - offset = (bo->tbo.resource->start << PAGE_SHIFT) +
> - amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type);
> + if (bo->tbo.resource->mem_type == TTM_PL_TT)
> + offset = amdgpu_gmc_agp_addr(&bo->tbo);
> +
> + if (offset == AMDGPU_BO_INVALID_OFFSET)
> + offset = (bo->tbo.resource->start << PAGE_SHIFT) +
> + amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type);
>
> return amdgpu_gmc_sign_extend(offset);
> }
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> index 05991c5c8ddb..ab4a762aed5b 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> @@ -959,10 +959,8 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo)
> return 0;
>
> addr = amdgpu_gmc_agp_addr(bo);
> - if (addr != AMDGPU_BO_INVALID_OFFSET) {
> - bo->resource->start = addr >> PAGE_SHIFT;
> + if (addr != AMDGPU_BO_INVALID_OFFSET)
> return 0;
> - }
>
> /* allocate GART space */
> placement.num_placement = 1;
^ permalink raw reply [flat|nested] 12+ messages in thread
* [PATCH] drm/amdgpu: fix AGP addressing when GART is not at 0
@ 2023-11-10 15:02 Alex Deucher
2023-11-10 15:03 ` Christian König
2023-11-11 3:18 ` Zhang, Yifan
0 siblings, 2 replies; 12+ messages in thread
From: Alex Deucher @ 2023-11-10 15:02 UTC (permalink / raw)
To: amd-gfx; +Cc: Alex Deucher, Yifan Zhang, Jesse Zhang, christian.koenig
This worked by luck if the GART aperture ended up at 0. When
we ended up moving GART on some chips, the GART aperture ended
up offsetting the the AGP address since the resource->start is
a GART offset, not an MC address. Fix this by moving the AGP
address setup into amdgpu_bo_gpu_offset_no_check().
v2: check mem_type before checking agp
Fixes: 67318cb84341 ("drm/amdgpu/gmc11: set gart placement GC11")
Reported-by: Jesse Zhang <Jesse.Zhang@amd.com>
Reported-by: Yifan Zhang <yifan1.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: christian.koenig@amd.com
---
drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 10 +++++++---
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 4 +---
2 files changed, 8 insertions(+), 6 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
index cef920a93924..d79b4ca1ecfc 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
@@ -1527,10 +1527,14 @@ u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo)
u64 amdgpu_bo_gpu_offset_no_check(struct amdgpu_bo *bo)
{
struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
- uint64_t offset;
+ uint64_t offset = AMDGPU_BO_INVALID_OFFSET;
- offset = (bo->tbo.resource->start << PAGE_SHIFT) +
- amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type);
+ if (bo->tbo.resource->mem_type == TTM_PL_TT)
+ offset = amdgpu_gmc_agp_addr(&bo->tbo);
+
+ if (offset == AMDGPU_BO_INVALID_OFFSET)
+ offset = (bo->tbo.resource->start << PAGE_SHIFT) +
+ amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type);
return amdgpu_gmc_sign_extend(offset);
}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 05991c5c8ddb..ab4a762aed5b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -959,10 +959,8 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo)
return 0;
addr = amdgpu_gmc_agp_addr(bo);
- if (addr != AMDGPU_BO_INVALID_OFFSET) {
- bo->resource->start = addr >> PAGE_SHIFT;
+ if (addr != AMDGPU_BO_INVALID_OFFSET)
return 0;
- }
/* allocate GART space */
placement.num_placement = 1;
--
2.41.0
^ permalink raw reply related [flat|nested] 12+ messages in thread
end of thread, other threads:[~2023-11-20 18:49 UTC | newest]
Thread overview: 12+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-11-10 14:47 [PATCH] drm/amdgpu: fix AGP addressing when GART is not at 0 Alex Deucher
2023-11-10 14:50 ` Christian König
2023-11-12 18:13 ` Alex Deucher
2023-11-16 9:37 ` Zhang, Yifan
2023-11-16 14:25 ` Alex Deucher
2023-11-16 15:05 ` Zhang, Yifan
2023-11-10 15:02 Alex Deucher
2023-11-10 15:03 ` Christian König
2023-11-17 19:24 ` Alex Deucher
2023-11-20 16:16 ` Christian König
2023-11-11 3:18 ` Zhang, Yifan
2023-11-20 18:49 Alex Deucher
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.