Re: [Nouveau] [PATCH drm-misc-next 0/5] Nouveau VM_BIND uAPI Fixes
For the series: Reviewed-by: Dave Airlie On Tue, 8 Aug 2023 at 02:32, Danilo Krummrich wrote: > > The patch series provides a few fixes for the recently merged VM_BIND uAPI > mostly addressing a couple of warnings. > > It also contains one patch to slightly reduce the memory footprint of > struct nouveau_uvma. > > Danilo Krummrich (5): > nouveau/dmem: fix copy-paste error in nouveau_dmem_migrate_chunk() > drm/nouveau: nvkm: vmm: silence warning from cast > drm/nouveau: remove incorrect __user annotations > drm/nouveau: uvmm: remove incorrect calls to mas_unlock() > drm/nouveau: uvmm: remove dedicated VM pointer from VMAs > > drivers/gpu/drm/nouveau/nouveau_dmem.c| 2 +- > drivers/gpu/drm/nouveau/nouveau_exec.c| 6 ++--- > drivers/gpu/drm/nouveau/nouveau_exec.h| 2 +- > drivers/gpu/drm/nouveau/nouveau_uvmm.c| 23 --- > drivers/gpu/drm/nouveau/nouveau_uvmm.h| 14 +-- > .../gpu/drm/nouveau/nvkm/subdev/mmu/uvmm.c| 5 ++-- > 6 files changed, 24 insertions(+), 28 deletions(-) > > > base-commit: 82d750e9d2f5d0594c8f7057ce59127e701af781 > -- > 2.41.0 >
Re: [Nouveau] [PATCH] drm/nouveau/sched: Don't pass user flags to drm_syncobj_find_fence()
On Mon, Aug 07, 2023 at 06:41:44PM -0500, Faith Ekstrand wrote:
> The flags field in drm_syncobj_find_fence() takes SYNCOBJ_WAIT flags
> from the syncobj UAPI whereas sync->flags is from the nouveau UAPI. What
> we actually want is 0 flags which tells it to just try to find the
> fence and then return without waiting.
Good catch!
Reviewed-by: Danilo Krummrich
>
> Signed-off-by: Faith Ekstrand
> Fixes: b88baab82871 ("drm/nouveau: implement new VM_BIND uAPI")
> Cc: Danilo Krummrich
> Cc: Dave Airlie
> ---
> drivers/gpu/drm/nouveau/nouveau_sched.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/nouveau/nouveau_sched.c
> b/drivers/gpu/drm/nouveau/nouveau_sched.c
> index b3b59fbec291..3424a1bf6af3 100644
> --- a/drivers/gpu/drm/nouveau/nouveau_sched.c
> +++ b/drivers/gpu/drm/nouveau/nouveau_sched.c
> @@ -142,7 +142,7 @@ sync_find_fence(struct nouveau_job *job,
>
> ret = drm_syncobj_find_fence(job->file_priv,
>sync->handle, point,
> - sync->flags, fence);
> + 0 /* flags */, fence);
> if (ret)
> return ret;
>
> --
> 2.41.0
>
[Nouveau] [PATCH] drm/nouveau/sched: Don't pass user flags to drm_syncobj_find_fence()
The flags field in drm_syncobj_find_fence() takes SYNCOBJ_WAIT flags
from the syncobj UAPI whereas sync->flags is from the nouveau UAPI. What
we actually want is 0 flags which tells it to just try to find the
fence and then return without waiting.
Signed-off-by: Faith Ekstrand
Fixes: b88baab82871 ("drm/nouveau: implement new VM_BIND uAPI")
Cc: Danilo Krummrich
Cc: Dave Airlie
---
drivers/gpu/drm/nouveau/nouveau_sched.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/nouveau/nouveau_sched.c
b/drivers/gpu/drm/nouveau/nouveau_sched.c
index b3b59fbec291..3424a1bf6af3 100644
--- a/drivers/gpu/drm/nouveau/nouveau_sched.c
+++ b/drivers/gpu/drm/nouveau/nouveau_sched.c
@@ -142,7 +142,7 @@ sync_find_fence(struct nouveau_job *job,
ret = drm_syncobj_find_fence(job->file_priv,
sync->handle, point,
-sync->flags, fence);
+0 /* flags */, fence);
if (ret)
return ret;
--
2.41.0
Re: [Nouveau] [PATCH] drm/nouveau/disp: Revert a NULL check inside nouveau_connector_get_modes
Ugh, thanks for catching this!
Reviewed-by: Lyude Paul
On Sat, 2023-08-05 at 12:18 +0200, Karol Herbst wrote:
> The original commit adding that check tried to protect the kenrel against
> a potential invalid NULL pointer access.
>
> However we call nouveau_connector_detect_depth once without a native_mode
> set on purpose for non LVDS connectors and this broke DP support in a few
> cases.
>
> Cc: Olaf Skibbe
> Cc: Lyude Paul
> Closes: https://gitlab.freedesktop.org/drm/nouveau/-/issues/238
> Closes: https://gitlab.freedesktop.org/drm/nouveau/-/issues/245
> Fixes: 20a2ce87fbaf8 ("drm/nouveau/dp: check for NULL
> nv_connector->native_mode")
> Signed-off-by: Karol Herbst
> ---
> drivers/gpu/drm/nouveau/nouveau_connector.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/nouveau/nouveau_connector.c
> b/drivers/gpu/drm/nouveau/nouveau_connector.c
> index f75c6f09dd2af..a2e0033e8a260 100644
> --- a/drivers/gpu/drm/nouveau/nouveau_connector.c
> +++ b/drivers/gpu/drm/nouveau/nouveau_connector.c
> @@ -967,7 +967,7 @@ nouveau_connector_get_modes(struct drm_connector
> *connector)
> /* Determine display colour depth for everything except LVDS now,
>* DP requires this before mode_valid() is called.
>*/
> - if (connector->connector_type != DRM_MODE_CONNECTOR_LVDS &&
> nv_connector->native_mode)
> + if (connector->connector_type != DRM_MODE_CONNECTOR_LVDS)
> nouveau_connector_detect_depth(connector);
>
> /* Find the native mode if this is a digital panel, if we didn't
--
Cheers,
Lyude Paul (she/her)
Software Engineer at Red Hat
Re: [Nouveau] [PATCH drm-misc-next v9 06/11] drm/nouveau: fence: separate fence alloc and emit
Hi Christian,
On 8/7/23 20:07, Christian König wrote:
Am 03.08.23 um 18:52 schrieb Danilo Krummrich:
The new (VM_BIND) UAPI exports DMA fences through DRM syncobjs. Hence,
in order to emit fences within DMA fence signalling critical sections
(e.g. as typically done in the DRM GPU schedulers run_job() callback) we
need to separate fence allocation and fence emitting.
At least from the description that sounds like it might be illegal.
Daniel can you take a look as well.
What exactly are you doing here?
I'm basically doing exactly the same as amdgpu_fence_emit() does in
amdgpu_ib_schedule() called by amdgpu_job_run().
The difference - and this is what this patch is for - is that I separate
the fence allocation from emitting the fence, such that the fence
structure is allocated before the job is submitted to the GPU scheduler.
amdgpu solves this with GFP_ATOMIC within amdgpu_fence_emit() to
allocate the fence structure in this case.
- Danilo
Regards,
Christian.
Signed-off-by: Danilo Krummrich
---
drivers/gpu/drm/nouveau/dispnv04/crtc.c | 9 -
drivers/gpu/drm/nouveau/nouveau_bo.c | 52 +++--
drivers/gpu/drm/nouveau/nouveau_chan.c | 6 ++-
drivers/gpu/drm/nouveau/nouveau_dmem.c | 9 +++--
drivers/gpu/drm/nouveau/nouveau_fence.c | 16 +++-
drivers/gpu/drm/nouveau/nouveau_fence.h | 3 +-
drivers/gpu/drm/nouveau/nouveau_gem.c | 5 ++-
7 files changed, 59 insertions(+), 41 deletions(-)
diff --git a/drivers/gpu/drm/nouveau/dispnv04/crtc.c
b/drivers/gpu/drm/nouveau/dispnv04/crtc.c
index a6f2e681bde9..a34924523133 100644
--- a/drivers/gpu/drm/nouveau/dispnv04/crtc.c
+++ b/drivers/gpu/drm/nouveau/dispnv04/crtc.c
@@ -1122,11 +1122,18 @@ nv04_page_flip_emit(struct nouveau_channel *chan,
PUSH_NVSQ(push, NV_SW, NV_SW_PAGE_FLIP, 0x);
PUSH_KICK(push);
- ret = nouveau_fence_new(chan, false, pfence);
+ ret = nouveau_fence_new(pfence);
if (ret)
goto fail;
+ ret = nouveau_fence_emit(*pfence, chan);
+ if (ret)
+ goto fail_fence_unref;
+
return 0;
+
+fail_fence_unref:
+ nouveau_fence_unref(pfence);
fail:
spin_lock_irqsave(>event_lock, flags);
list_del(>head);
diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c
b/drivers/gpu/drm/nouveau/nouveau_bo.c
index 057bc995f19b..e9cbbf594e6f 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bo.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -820,29 +820,39 @@ nouveau_bo_move_m2mf(struct ttm_buffer_object
*bo, int evict,
mutex_lock(>mutex);
else
mutex_lock_nested(>mutex, SINGLE_DEPTH_NESTING);
+
ret = nouveau_fence_sync(nouveau_bo(bo), chan, true,
ctx->interruptible);
- if (ret == 0) {
- ret = drm->ttm.move(chan, bo, bo->resource, new_reg);
- if (ret == 0) {
- ret = nouveau_fence_new(chan, false, );
- if (ret == 0) {
- /* TODO: figure out a better solution here
- *
- * wait on the fence here explicitly as going through
- * ttm_bo_move_accel_cleanup somehow doesn't seem to
do it.
- *
- * Without this the operation can timeout and we'll
fallback to a
- * software copy, which might take several minutes to
finish.
- */
- nouveau_fence_wait(fence, false, false);
- ret = ttm_bo_move_accel_cleanup(bo,
- >base,
- evict, false,
- new_reg);
- nouveau_fence_unref();
- }
- }
+ if (ret)
+ goto out_unlock;
+
+ ret = drm->ttm.move(chan, bo, bo->resource, new_reg);
+ if (ret)
+ goto out_unlock;
+
+ ret = nouveau_fence_new();
+ if (ret)
+ goto out_unlock;
+
+ ret = nouveau_fence_emit(fence, chan);
+ if (ret) {
+ nouveau_fence_unref();
+ goto out_unlock;
}
+
+ /* TODO: figure out a better solution here
+ *
+ * wait on the fence here explicitly as going through
+ * ttm_bo_move_accel_cleanup somehow doesn't seem to do it.
+ *
+ * Without this the operation can timeout and we'll fallback to a
+ * software copy, which might take several minutes to finish.
+ */
+ nouveau_fence_wait(fence, false, false);
+ ret = ttm_bo_move_accel_cleanup(bo, >base, evict, false,
+ new_reg);
+ nouveau_fence_unref();
+
+out_unlock:
mutex_unlock(>mutex);
return ret;
}
diff --git a/drivers/gpu/drm/nouveau/nouveau_chan.c
b/drivers/gpu/drm/nouveau/nouveau_chan.c
index 6d639314250a..f69be4c8f9f2 100644
--- a/drivers/gpu/drm/nouveau/nouveau_chan.c
+++ b/drivers/gpu/drm/nouveau/nouveau_chan.c
@@ -62,9 +62,11 @@ nouveau_channel_idle(struct nouveau_channel *chan)
struct nouveau_fence *fence = NULL;
int ret;
- ret = nouveau_fence_new(chan, false, );
+ ret =
Re: [Nouveau] [PATCH drm-misc-next v9 06/11] drm/nouveau: fence: separate fence alloc and emit
Am 03.08.23 um 18:52 schrieb Danilo Krummrich:
The new (VM_BIND) UAPI exports DMA fences through DRM syncobjs. Hence,
in order to emit fences within DMA fence signalling critical sections
(e.g. as typically done in the DRM GPU schedulers run_job() callback) we
need to separate fence allocation and fence emitting.
At least from the description that sounds like it might be illegal.
Daniel can you take a look as well.
What exactly are you doing here?
Regards,
Christian.
Signed-off-by: Danilo Krummrich
---
drivers/gpu/drm/nouveau/dispnv04/crtc.c | 9 -
drivers/gpu/drm/nouveau/nouveau_bo.c| 52 +++--
drivers/gpu/drm/nouveau/nouveau_chan.c | 6 ++-
drivers/gpu/drm/nouveau/nouveau_dmem.c | 9 +++--
drivers/gpu/drm/nouveau/nouveau_fence.c | 16 +++-
drivers/gpu/drm/nouveau/nouveau_fence.h | 3 +-
drivers/gpu/drm/nouveau/nouveau_gem.c | 5 ++-
7 files changed, 59 insertions(+), 41 deletions(-)
diff --git a/drivers/gpu/drm/nouveau/dispnv04/crtc.c
b/drivers/gpu/drm/nouveau/dispnv04/crtc.c
index a6f2e681bde9..a34924523133 100644
--- a/drivers/gpu/drm/nouveau/dispnv04/crtc.c
+++ b/drivers/gpu/drm/nouveau/dispnv04/crtc.c
@@ -1122,11 +1122,18 @@ nv04_page_flip_emit(struct nouveau_channel *chan,
PUSH_NVSQ(push, NV_SW, NV_SW_PAGE_FLIP, 0x);
PUSH_KICK(push);
- ret = nouveau_fence_new(chan, false, pfence);
+ ret = nouveau_fence_new(pfence);
if (ret)
goto fail;
+ ret = nouveau_fence_emit(*pfence, chan);
+ if (ret)
+ goto fail_fence_unref;
+
return 0;
+
+fail_fence_unref:
+ nouveau_fence_unref(pfence);
fail:
spin_lock_irqsave(>event_lock, flags);
list_del(>head);
diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c
b/drivers/gpu/drm/nouveau/nouveau_bo.c
index 057bc995f19b..e9cbbf594e6f 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bo.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -820,29 +820,39 @@ nouveau_bo_move_m2mf(struct ttm_buffer_object *bo, int
evict,
mutex_lock(>mutex);
else
mutex_lock_nested(>mutex, SINGLE_DEPTH_NESTING);
+
ret = nouveau_fence_sync(nouveau_bo(bo), chan, true,
ctx->interruptible);
- if (ret == 0) {
- ret = drm->ttm.move(chan, bo, bo->resource, new_reg);
- if (ret == 0) {
- ret = nouveau_fence_new(chan, false, );
- if (ret == 0) {
- /* TODO: figure out a better solution here
-*
-* wait on the fence here explicitly as going
through
-* ttm_bo_move_accel_cleanup somehow doesn't
seem to do it.
-*
-* Without this the operation can timeout and
we'll fallback to a
-* software copy, which might take several
minutes to finish.
-*/
- nouveau_fence_wait(fence, false, false);
- ret = ttm_bo_move_accel_cleanup(bo,
- >base,
- evict, false,
- new_reg);
- nouveau_fence_unref();
- }
- }
+ if (ret)
+ goto out_unlock;
+
+ ret = drm->ttm.move(chan, bo, bo->resource, new_reg);
+ if (ret)
+ goto out_unlock;
+
+ ret = nouveau_fence_new();
+ if (ret)
+ goto out_unlock;
+
+ ret = nouveau_fence_emit(fence, chan);
+ if (ret) {
+ nouveau_fence_unref();
+ goto out_unlock;
}
+
+ /* TODO: figure out a better solution here
+*
+* wait on the fence here explicitly as going through
+* ttm_bo_move_accel_cleanup somehow doesn't seem to do it.
+*
+* Without this the operation can timeout and we'll fallback to a
+* software copy, which might take several minutes to finish.
+*/
+ nouveau_fence_wait(fence, false, false);
+ ret = ttm_bo_move_accel_cleanup(bo, >base, evict, false,
+ new_reg);
+ nouveau_fence_unref();
+
+out_unlock:
mutex_unlock(>mutex);
return ret;
}
diff --git a/drivers/gpu/drm/nouveau/nouveau_chan.c
b/drivers/gpu/drm/nouveau/nouveau_chan.c
index 6d639314250a..f69be4c8f9f2 100644
--- a/drivers/gpu/drm/nouveau/nouveau_chan.c
+++ b/drivers/gpu/drm/nouveau/nouveau_chan.c
@@ -62,9 +62,11 @@ nouveau_channel_idle(struct nouveau_channel *chan)
struct nouveau_fence *fence = NULL;
int ret;
- ret = nouveau_fence_new(chan, false, );
+ ret = nouveau_fence_new();
[Nouveau] [PATCH drm-misc-next 5/5] drm/nouveau: uvmm: remove dedicated VM pointer from VMAs
VMAs can find their corresponding VM through their embedded struct
drm_gpuva which already carries a pointer to a struct drm_gpuva_manager
which the VM is based on. Hence, remove the struct nouveau_uvmm pointer
from struct nouveau_uvma to save a couple of bytes per mapping.
Signed-off-by: Danilo Krummrich
---
drivers/gpu/drm/nouveau/nouveau_uvmm.c | 9 -
drivers/gpu/drm/nouveau/nouveau_uvmm.h | 10 +-
2 files changed, 9 insertions(+), 10 deletions(-)
diff --git a/drivers/gpu/drm/nouveau/nouveau_uvmm.c
b/drivers/gpu/drm/nouveau/nouveau_uvmm.c
index 91b964ef98b1..3a1e8538f205 100644
--- a/drivers/gpu/drm/nouveau/nouveau_uvmm.c
+++ b/drivers/gpu/drm/nouveau/nouveau_uvmm.c
@@ -181,7 +181,7 @@ nouveau_uvma_vmm_put(struct nouveau_uvma *uvma)
u64 addr = uvma->va.va.addr;
u64 range = uvma->va.va.range;
- return nouveau_uvmm_vmm_put(uvma->uvmm, addr, range);
+ return nouveau_uvmm_vmm_put(to_uvmm(uvma), addr, range);
}
static int
@@ -192,7 +192,7 @@ nouveau_uvma_map(struct nouveau_uvma *uvma,
u64 offset = uvma->va.gem.offset;
u64 range = uvma->va.va.range;
- return nouveau_uvmm_vmm_map(uvma->uvmm, addr, range,
+ return nouveau_uvmm_vmm_map(to_uvmm(uvma), addr, range,
offset, uvma->kind, mem);
}
@@ -206,7 +206,7 @@ nouveau_uvma_unmap(struct nouveau_uvma *uvma)
if (drm_gpuva_invalidated(>va))
return 0;
- return nouveau_uvmm_vmm_unmap(uvma->uvmm, addr, range, sparse);
+ return nouveau_uvmm_vmm_unmap(to_uvmm(uvma), addr, range, sparse);
}
static int
@@ -586,7 +586,6 @@ op_map_prepare(struct nouveau_uvmm *uvmm,
if (ret)
return ret;
- uvma->uvmm = uvmm;
uvma->region = args->region;
uvma->kind = args->kind;
@@ -794,7 +793,7 @@ op_unmap_range(struct drm_gpuva_op_unmap *u,
bool sparse = !!uvma->region;
if (!drm_gpuva_invalidated(u->va))
- nouveau_uvmm_vmm_unmap(uvma->uvmm, addr, range, sparse);
+ nouveau_uvmm_vmm_unmap(to_uvmm(uvma), addr, range, sparse);
}
static void
diff --git a/drivers/gpu/drm/nouveau/nouveau_uvmm.h
b/drivers/gpu/drm/nouveau/nouveau_uvmm.h
index 534baadc3bf7..fc7f6fd2a4e1 100644
--- a/drivers/gpu/drm/nouveau/nouveau_uvmm.h
+++ b/drivers/gpu/drm/nouveau/nouveau_uvmm.h
@@ -37,12 +37,15 @@ struct nouveau_uvma_region {
struct nouveau_uvma {
struct drm_gpuva va;
- struct nouveau_uvmm *uvmm;
struct nouveau_uvma_region *region;
-
u8 kind;
};
+#define uvmm_from_mgr(x) container_of((x), struct nouveau_uvmm, umgr)
+#define uvma_from_va(x) container_of((x), struct nouveau_uvma, va)
+
+#define to_uvmm(x) uvmm_from_mgr((x)->va.mgr)
+
struct nouveau_uvmm_bind_job {
struct nouveau_job base;
@@ -79,9 +82,6 @@ struct nouveau_uvmm_bind_job_args {
#define to_uvmm_bind_job(job) container_of((job), struct
nouveau_uvmm_bind_job, base)
-#define uvmm_from_mgr(x) container_of((x), struct nouveau_uvmm, umgr)
-#define uvma_from_va(x) container_of((x), struct nouveau_uvma, va)
-
int nouveau_uvmm_init(struct nouveau_uvmm *uvmm, struct nouveau_cli *cli,
u64 kernel_managed_addr, u64 kernel_managed_size);
void nouveau_uvmm_fini(struct nouveau_uvmm *uvmm);
--
2.41.0
[Nouveau] [PATCH drm-misc-next 3/5] drm/nouveau: remove incorrect __user annotations
Fix copy-paste error causing EXEC and VM_BIND syscalls data pointers
to carry incorrect __user annotations.
Fixes: b88baab82871 ("drm/nouveau: implement new VM_BIND uAPI")
Reported-by: kernel test robot
Signed-off-by: Danilo Krummrich
---
drivers/gpu/drm/nouveau/nouveau_exec.c | 6 +++---
drivers/gpu/drm/nouveau/nouveau_exec.h | 2 +-
drivers/gpu/drm/nouveau/nouveau_uvmm.c | 6 +++---
drivers/gpu/drm/nouveau/nouveau_uvmm.h | 4 ++--
4 files changed, 9 insertions(+), 9 deletions(-)
diff --git a/drivers/gpu/drm/nouveau/nouveau_exec.c
b/drivers/gpu/drm/nouveau/nouveau_exec.c
index 42d9dd43ea02..0f927adda4ed 100644
--- a/drivers/gpu/drm/nouveau/nouveau_exec.c
+++ b/drivers/gpu/drm/nouveau/nouveau_exec.c
@@ -293,7 +293,7 @@ nouveau_exec(struct nouveau_exec_job_args *args)
static int
nouveau_exec_ucopy(struct nouveau_exec_job_args *args,
- struct drm_nouveau_exec __user *req)
+ struct drm_nouveau_exec *req)
{
struct drm_nouveau_sync **s;
u32 inc = req->wait_count;
@@ -352,7 +352,7 @@ nouveau_exec_ufree(struct nouveau_exec_job_args *args)
int
nouveau_exec_ioctl_exec(struct drm_device *dev,
- void __user *data,
+ void *data,
struct drm_file *file_priv)
{
struct nouveau_abi16 *abi16 = nouveau_abi16_get(file_priv);
@@ -360,7 +360,7 @@ nouveau_exec_ioctl_exec(struct drm_device *dev,
struct nouveau_abi16_chan *chan16;
struct nouveau_channel *chan = NULL;
struct nouveau_exec_job_args args = {};
- struct drm_nouveau_exec __user *req = data;
+ struct drm_nouveau_exec *req = data;
int ret = 0;
if (unlikely(!abi16))
diff --git a/drivers/gpu/drm/nouveau/nouveau_exec.h
b/drivers/gpu/drm/nouveau/nouveau_exec.h
index 3032db27b8d7..778cacd90f65 100644
--- a/drivers/gpu/drm/nouveau/nouveau_exec.h
+++ b/drivers/gpu/drm/nouveau/nouveau_exec.h
@@ -48,7 +48,7 @@ struct nouveau_exec_job {
int nouveau_exec_job_init(struct nouveau_exec_job **job,
struct nouveau_exec_job_args *args);
-int nouveau_exec_ioctl_exec(struct drm_device *dev, void __user *data,
+int nouveau_exec_ioctl_exec(struct drm_device *dev, void *data,
struct drm_file *file_priv);
#endif
diff --git a/drivers/gpu/drm/nouveau/nouveau_uvmm.c
b/drivers/gpu/drm/nouveau/nouveau_uvmm.c
index b515a21aa08e..2acbac73e57a 100644
--- a/drivers/gpu/drm/nouveau/nouveau_uvmm.c
+++ b/drivers/gpu/drm/nouveau/nouveau_uvmm.c
@@ -1687,7 +1687,7 @@ nouveau_uvmm_vm_bind(struct nouveau_uvmm_bind_job_args
*args)
static int
nouveau_uvmm_vm_bind_ucopy(struct nouveau_uvmm_bind_job_args *args,
- struct drm_nouveau_vm_bind __user *req)
+ struct drm_nouveau_vm_bind *req)
{
struct drm_nouveau_sync **s;
u32 inc = req->wait_count;
@@ -1749,12 +1749,12 @@ nouveau_uvmm_vm_bind_ufree(struct
nouveau_uvmm_bind_job_args *args)
int
nouveau_uvmm_ioctl_vm_bind(struct drm_device *dev,
- void __user *data,
+ void *data,
struct drm_file *file_priv)
{
struct nouveau_cli *cli = nouveau_cli(file_priv);
struct nouveau_uvmm_bind_job_args args = {};
- struct drm_nouveau_vm_bind __user *req = data;
+ struct drm_nouveau_vm_bind *req = data;
int ret = 0;
if (unlikely(!nouveau_cli_uvmm_locked(cli)))
diff --git a/drivers/gpu/drm/nouveau/nouveau_uvmm.h
b/drivers/gpu/drm/nouveau/nouveau_uvmm.h
index 3923c03012f9..534baadc3bf7 100644
--- a/drivers/gpu/drm/nouveau/nouveau_uvmm.h
+++ b/drivers/gpu/drm/nouveau/nouveau_uvmm.h
@@ -89,10 +89,10 @@ void nouveau_uvmm_fini(struct nouveau_uvmm *uvmm);
void nouveau_uvmm_bo_map_all(struct nouveau_bo *nvbov, struct nouveau_mem
*mem);
void nouveau_uvmm_bo_unmap_all(struct nouveau_bo *nvbo);
-int nouveau_uvmm_ioctl_vm_init(struct drm_device *dev, void __user *data,
+int nouveau_uvmm_ioctl_vm_init(struct drm_device *dev, void *data,
struct drm_file *file_priv);
-int nouveau_uvmm_ioctl_vm_bind(struct drm_device *dev, void __user *data,
+int nouveau_uvmm_ioctl_vm_bind(struct drm_device *dev, void *data,
struct drm_file *file_priv);
static inline void nouveau_uvmm_lock(struct nouveau_uvmm *uvmm)
--
2.41.0
[Nouveau] [PATCH drm-misc-next 4/5] drm/nouveau: uvmm: remove incorrect calls to mas_unlock()
Remove incorrect calls to mas_unlock() in the unwind path of
__nouveau_uvma_region_insert(). The region maple tree uses an external
lock instead, namely the global uvmm lock.
Fixes: b88baab82871 ("drm/nouveau: implement new VM_BIND uAPI")
Reported-by: kernel test robot
Signed-off-by: Danilo Krummrich
---
drivers/gpu/drm/nouveau/nouveau_uvmm.c | 8 ++--
1 file changed, 2 insertions(+), 6 deletions(-)
diff --git a/drivers/gpu/drm/nouveau/nouveau_uvmm.c
b/drivers/gpu/drm/nouveau/nouveau_uvmm.c
index 2acbac73e57a..91b964ef98b1 100644
--- a/drivers/gpu/drm/nouveau/nouveau_uvmm.c
+++ b/drivers/gpu/drm/nouveau/nouveau_uvmm.c
@@ -279,15 +279,11 @@ __nouveau_uvma_region_insert(struct nouveau_uvmm *uvmm,
u64 last = addr + range - 1;
MA_STATE(mas, >region_mt, addr, addr);
- if (unlikely(mas_walk())) {
- mas_unlock();
+ if (unlikely(mas_walk()))
return -EEXIST;
- }
- if (unlikely(mas.last < last)) {
- mas_unlock();
+ if (unlikely(mas.last < last))
return -EEXIST;
- }
mas.index = addr;
mas.last = last;
--
2.41.0
[Nouveau] [PATCH drm-misc-next 2/5] drm/nouveau: nvkm: vmm: silence warning from cast
Cast the integer to a pointer-sized type first to keep the compiler
happy.
Fixes: 6b252cf42281 ("drm/nouveau: nvkm/vmm: implement raw ops to manage uvmm")
Reported-by: kernel test robot
Reported-by: Stephen Rothwell
Signed-off-by: Danilo Krummrich
---
drivers/gpu/drm/nouveau/nvkm/subdev/mmu/uvmm.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/uvmm.c
b/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/uvmm.c
index 38b7ced934b1..46cbd4cedb78 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/uvmm.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/uvmm.c
@@ -399,6 +399,8 @@ nvkm_uvmm_mthd_raw_map(struct nvkm_uvmm *uvmm, struct
nvif_vmm_raw_v0 *args)
.no_comp = true,
};
struct nvkm_memory *memory;
+ void *argv = (void *)(uintptr_t)args->argv;
+ int argc = args->argc;
u64 handle = args->memory;
u8 refd;
int ret;
@@ -418,8 +420,7 @@ nvkm_uvmm_mthd_raw_map(struct nvkm_uvmm *uvmm, struct
nvif_vmm_raw_v0 *args)
return PTR_ERR(memory);
}
- ret = nvkm_memory_map(memory, args->offset, vmm, ,
- (void *)args->argv, args->argc);
+ ret = nvkm_memory_map(memory, args->offset, vmm, , argv, argc);
nvkm_memory_unref();
nvkm_memory_unref();
--
2.41.0
[Nouveau] [PATCH drm-misc-next 1/5] nouveau/dmem: fix copy-paste error in nouveau_dmem_migrate_chunk()
Fix call to nouveau_fence_emit() with wrong channel parameter.
Fixes: 7f2a0b50b2b2 ("drm/nouveau: fence: separate fence alloc and emit")
Reported-by: kernel test robot
Reported-by: Stephen Rothwell
Reviewed-by: Karol Herbst
Signed-off-by: Danilo Krummrich
---
drivers/gpu/drm/nouveau/nouveau_dmem.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/nouveau/nouveau_dmem.c
b/drivers/gpu/drm/nouveau/nouveau_dmem.c
index 4ad40e42cae1..61e84562094a 100644
--- a/drivers/gpu/drm/nouveau/nouveau_dmem.c
+++ b/drivers/gpu/drm/nouveau/nouveau_dmem.c
@@ -678,7 +678,7 @@ static void nouveau_dmem_migrate_chunk(struct nouveau_drm
*drm,
}
if (!nouveau_fence_new())
- nouveau_fence_emit(fence, chunk->drm->dmem->migrate.chan);
+ nouveau_fence_emit(fence, drm->dmem->migrate.chan);
migrate_vma_pages(args);
nouveau_dmem_fence_done();
nouveau_pfns_map(svmm, args->vma->vm_mm, args->start, pfns, i);
--
2.41.0
[Nouveau] [PATCH drm-misc-next 0/5] Nouveau VM_BIND uAPI Fixes
The patch series provides a few fixes for the recently merged VM_BIND uAPI mostly addressing a couple of warnings. It also contains one patch to slightly reduce the memory footprint of struct nouveau_uvma. Danilo Krummrich (5): nouveau/dmem: fix copy-paste error in nouveau_dmem_migrate_chunk() drm/nouveau: nvkm: vmm: silence warning from cast drm/nouveau: remove incorrect __user annotations drm/nouveau: uvmm: remove incorrect calls to mas_unlock() drm/nouveau: uvmm: remove dedicated VM pointer from VMAs drivers/gpu/drm/nouveau/nouveau_dmem.c| 2 +- drivers/gpu/drm/nouveau/nouveau_exec.c| 6 ++--- drivers/gpu/drm/nouveau/nouveau_exec.h| 2 +- drivers/gpu/drm/nouveau/nouveau_uvmm.c| 23 --- drivers/gpu/drm/nouveau/nouveau_uvmm.h| 14 +-- .../gpu/drm/nouveau/nvkm/subdev/mmu/uvmm.c| 5 ++-- 6 files changed, 24 insertions(+), 28 deletions(-) base-commit: 82d750e9d2f5d0594c8f7057ce59127e701af781 -- 2.41.0
[Nouveau] [PATCH drm-misc-next 0/5] Nouveau VM_BIND uAPI Fixes
The patch series provides a few fixes for the recently merged VM_BIND uAPI mostly addressing a couple of warnings. It also contains one patch to slightly reduce the memory footprint of struct nouveau_uvma. Danilo Krummrich (5): nouveau/dmem: fix copy-paste error in nouveau_dmem_migrate_chunk() drm/nouveau: nvkm: vmm: silence warning from cast drm/nouveau: remove incorrect __user annotations drm/nouveau: uvmm: remove incorrect calls to mas_unlock() drm/nouveau: uvmm: remove dedicated VM pointer from VMAs drivers/gpu/drm/nouveau/nouveau_dmem.c| 2 +- drivers/gpu/drm/nouveau/nouveau_exec.c| 6 ++--- drivers/gpu/drm/nouveau/nouveau_exec.h| 2 +- drivers/gpu/drm/nouveau/nouveau_uvmm.c| 23 --- drivers/gpu/drm/nouveau/nouveau_uvmm.h| 14 +-- .../gpu/drm/nouveau/nvkm/subdev/mmu/uvmm.c| 5 ++-- 6 files changed, 24 insertions(+), 28 deletions(-) base-commit: 82d750e9d2f5d0594c8f7057ce59127e701af781 -- 2.41.0
Re: [Nouveau] 2b5d1c29f6c4 ("drm/nouveau/disp: PIOR DP uses GPIO for HPD, not PMGR AUX interrupts")
On Mon, Aug 07, 2023 at 01:49:42PM +0200, Karol Herbst wrote: > in what way does it stop? Just not progressing? That would be kinda > concerning. Mind tracing with what arguments `nvkm_uevent_add` is > called with and without that patch? Well, me dumping those args I guess made the box not freeze before catching a #PF over serial. Does that help? [3.410135] Unpacking initramfs... [3.416319] software IO TLB: mapped [mem 0xa877d000-0xac77d000] (64MB) [3.418227] Initialise system trusted keyrings [3.432273] workingset: timestamp_bits=56 max_order=22 bucket_order=0 [3.439006] ntfs: driver 2.1.32 [Flags: R/W]. [3.443368] fuse: init (API version 7.38) [3.447601] 9p: Installing v9fs 9p2000 file system support [3.453223] Key type asymmetric registered [3.457332] Asymmetric key parser 'x509' registered [3.462236] Block layer SCSI generic (bsg) driver version 0.4 loaded (major 250) [3.475865] efifb: probing for efifb [3.479458] efifb: framebuffer at 0xf900, using 1920k, total 1920k [3.485969] efifb: mode is 800x600x32, linelength=3200, pages=1 [3.491872] efifb: scrolling: redraw [3.495438] efifb: Truecolor: size=8:8:8:8, shift=24:16:8:0 [3.502349] Console: switching to colour frame buffer device 100x37 [3.509564] fb0: EFI VGA frame buffer device [3.514013] ACPI: \_PR_.CP00: Found 4 idle states [3.518850] ACPI: \_PR_.CP01: Found 4 idle states [3.523687] ACPI: \_PR_.CP02: Found 4 idle states [3.528515] ACPI: \_PR_.CP03: Found 4 idle states [3.533346] ACPI: \_PR_.CP04: Found 4 idle states [3.538173] ACPI: \_PR_.CP05: Found 4 idle states [3.543003] ACPI: \_PR_.CP06: Found 4 idle states [3.544219] Freeing initrd memory: 8196K [3.547844] ACPI: \_PR_.CP07: Found 4 idle states [3.609542] Serial: 8250/16550 driver, 4 ports, IRQ sharing enabled [3.616224] 00:05: ttyS0 at I/O 0x3f8 (irq = 4, base_baud = 115200) is a 16550A [3.625552] serial :00:16.3: enabling device ( -> 0003) [3.633034] :00:16.3: ttyS1 at I/O 0xf0a0 (irq = 17, base_baud = 115200) is a 16550A [3.642451] Linux agpgart interface v0.103 [3.647141] ACPI: bus type drm_connector registered [3.653261] Console: switching to colour dummy device 80x25 [3.659092] nouveau :03:00.0: vgaarb: deactivate vga console [3.665174] nouveau :03:00.0: NVIDIA GT218 (0a8c00b1) [3.784585] nouveau :03:00.0: bios: version 70.18.83.00.08 [3.792244] nouveau :03:00.0: fb: 512 MiB DDR3 [3.948786] nouveau :03:00.0: DRM: VRAM: 512 MiB [3.953755] nouveau :03:00.0: DRM: GART: 1048576 MiB [3.959073] nouveau :03:00.0: DRM: TMDS table version 2.0 [3.964808] nouveau :03:00.0: DRM: DCB version 4.0 [3.969938] nouveau :03:00.0: DRM: DCB outp 00: 02000360 [3.976367] nouveau :03:00.0: DRM: DCB outp 01: 02000362 00020010 [3.982792] nouveau :03:00.0: DRM: DCB outp 02: 028003a6 0f220010 [3.989223] nouveau :03:00.0: DRM: DCB outp 03: 01011380 [3.995647] nouveau :03:00.0: DRM: DCB outp 04: 08011382 00020010 [4.002076] nouveau :03:00.0: DRM: DCB outp 05: 088113c6 0f220010 [4.008511] nouveau :03:00.0: DRM: DCB conn 00: 00101064 [4.014151] nouveau :03:00.0: DRM: DCB conn 01: 00202165 [4.021710] nvkm_uevent_add: uevent: 0x888100242100, event: 0x8881022de1a0, id: 0x0, bits: 0x1, func: 0x [4.033680] nvkm_uevent_add: uevent: 0x888100242300, event: 0x8881022de1a0, id: 0x0, bits: 0x1, func: 0x [4.045429] nouveau :03:00.0: DRM: MM: using COPY for buffer copies [4.052059] stackdepot: allocating hash table of 1048576 entries via kvcalloc [4.067191] nvkm_uevent_add: uevent: 0x888100242800, event: 0x888104b3e260, id: 0x0, bits: 0x1, func: 0x [4.078936] nvkm_uevent_add: uevent: 0x888100242900, event: 0x888104b3e260, id: 0x1, bits: 0x1, func: 0x [4.090514] nvkm_uevent_add: uevent: 0x888100242a00, event: 0x888102091f28, id: 0x1, bits: 0x3, func: 0x8177b700 [4.102118] tsc: Refined TSC clocksource calibration: 3591.345 MHz [4.108342] clocksource: tsc: mask: 0x max_cycles: 0x33c4635c383, max_idle_ns: 440795314831 ns [4.108401] nvkm_uevent_add: uevent: 0x8881020b6000, event: 0x888102091f28, id: 0xf, bits: 0x3, func: 0x8177b700 [4.129864] clocksource: Switched to clocksource tsc [4.131478] [drm] Initialized nouveau 1.3.1 20120801 for :03:00.0 on minor 0 [4.143806] BUG: kernel NULL pointer dereference, address: 0020 [4.144676] #PF: supervisor read access in kernel mode [4.144676] #PF: error_code(0x) - not-present page [4.144676] PGD 0 P4D 0 [4.144676] Oops: [#1] PREEMPT SMP PTI [4.144676] CPU: 2 PID: 1 Comm: swapper/0 Not tainted 6.5.0-rc5-dirty #1 [4.144676]
Re: [Nouveau] 2b5d1c29f6c4 ("drm/nouveau/disp: PIOR DP uses GPIO for HPD, not PMGR AUX interrupts")
On Sun, Aug 6, 2023 at 11:40 PM Borislav Petkov wrote: > > Hi folks, > > the patch in $Subject breaks booting here on one of my test boxes, see > below. > > Reverting it ontop of -rc4 fixes the issue. > > Thx. > > [3.580535] ACPI: \_PR_.CP04: Found 4 idle states > [3.585694] ACPI: \_PR_.CP05: Found 4 idle states > [3.590852] ACPI: \_PR_.CP06: Found 4 idle states > [3.596037] ACPI: \_PR_.CP07: Found 4 idle states > [3.644065] Freeing initrd memory: 6740K > [3.742932] Serial: 8250/16550 driver, 4 ports, IRQ sharing enabled > [3.750409] 00:05: ttyS0 at I/O 0x3f8 (irq = 4, base_baud = 115200) is a > 16550A > [3.762111] serial :00:16.3: enabling device ( -> 0003) > [3.771589] :00:16.3: ttyS1 at I/O 0xf0a0 (irq = 17, base_baud = > 115200) is a 16550A > [3.782503] Linux agpgart interface v0.103 > [3.787805] ACPI: bus type drm_connector registered > > <--- boot stops here. > in what way does it stop? Just not progressing? That would be kinda concerning. Mind tracing with what arguments `nvkm_uevent_add` is called with and without that patch? Also a boot log with `nouveau.debug=trace` might be helpful here. > It should continue with this: > > [3.795491] Console: switching to colour dummy device 80x25 > [3.801933] nouveau :03:00.0: vgaarb: deactivate vga console > [3.808303] nouveau :03:00.0: NVIDIA GT218 (0a8c00b1) > [3.931002] nouveau :03:00.0: bios: version 70.18.83.00.08 > [3.941731] nouveau :03:00.0: fb: 512 MiB DDR3 > [4.110348] tsc: Refined TSC clocksource calibration: 3591.349 MHz > [4.116627] clocksource: tsc: mask: 0x max_cycles: > 0x33c466a1ab5, max_idle_ns: 440795209767 ns > [4.126871] clocksource: Switched to clocksource tsc > [4.252013] nouveau :03:00.0: DRM: VRAM: 512 MiB > [4.257088] nouveau :03:00.0: DRM: GART: 1048576 MiB > [4.262501] nouveau :03:00.0: DRM: TMDS table version 2.0 > [4.268333] nouveau :03:00.0: DRM: DCB version 4.0 > [4.273561] nouveau :03:00.0: DRM: DCB outp 00: 02000360 > [4.280104] nouveau :03:00.0: DRM: DCB outp 01: 02000362 00020010 > [4.286630] nouveau :03:00.0: DRM: DCB outp 02: 028003a6 0f220010 > [4.293176] nouveau :03:00.0: DRM: DCB outp 03: 01011380 > [4.299711] nouveau :03:00.0: DRM: DCB outp 04: 08011382 00020010 > [4.306243] nouveau :03:00.0: DRM: DCB outp 05: 088113c6 0f220010 > [4.312772] nouveau :03:00.0: DRM: DCB conn 00: 00101064 > [4.318520] nouveau :03:00.0: DRM: DCB conn 01: 00202165 > [4.329488] nouveau :03:00.0: DRM: MM: using COPY for buffer copies > [4.336261] stackdepot: allocating hash table of 1048576 entries via > kvcalloc > ... > > > -- > Regards/Gruss, > Boris. > > https://people.kernel.org/tglx/notes-about-netiquette >
Re: [Nouveau] 2b5d1c29f6c4 ("drm/nouveau/disp: PIOR DP uses GPIO for HPD, not PMGR AUX interrupts")
[CCing the regression list, as it should be in the loop for regressions: https://docs.kernel.org/admin-guide/reporting-regressions.html] [TLDR: I'm adding this report to the list of tracked Linux kernel regressions; the text you find below is based on a few templates paragraphs you might have encountered already in similar form. See link in footer if these mails annoy you.] On 06.08.23 23:31, Borislav Petkov wrote: > > the patch in $Subject Side note, in case anyone cares: it was also included in 6.4.7. > breaks booting here on one of my test boxes, see > below. > > Reverting it ontop of -rc4 fixes the issue. > > Thx. Thanks for the report. To be sure the issue doesn't fall through the cracks unnoticed, I'm adding it to regzbot, the Linux kernel regression tracking bot: #regzbot ^introduced 2b5d1c29f6c4 #regzbot title drm/nouveau: stopped booting #regzbot ignore-activity This isn't a regression? This issue or a fix for it are already discussed somewhere else? It was fixed already? You want to clarify when the regression started to happen? Or point out I got the title or something else totally wrong? Then just reply and tell me -- ideally while also telling regzbot about it, as explained by the page listed in the footer of this mail. Developers: When fixing the issue, remember to add 'Link:' tags pointing to the report (the parent of this mail). See page linked in footer for details. Ciao, Thorsten (wearing his 'the Linux kernel's regression tracker' hat) -- Everything you wanna know about Linux kernel regression tracking: https://linux-regtracking.leemhuis.info/about/#tldr That page also explains what to do if mails like this annoy you. > [3.580535] ACPI: \_PR_.CP04: Found 4 idle states > [3.585694] ACPI: \_PR_.CP05: Found 4 idle states > [3.590852] ACPI: \_PR_.CP06: Found 4 idle states > [3.596037] ACPI: \_PR_.CP07: Found 4 idle states > [3.644065] Freeing initrd memory: 6740K > [3.742932] Serial: 8250/16550 driver, 4 ports, IRQ sharing enabled > [3.750409] 00:05: ttyS0 at I/O 0x3f8 (irq = 4, base_baud = 115200) is a > 16550A > [3.762111] serial :00:16.3: enabling device ( -> 0003) > [3.771589] :00:16.3: ttyS1 at I/O 0xf0a0 (irq = 17, base_baud = > 115200) is a 16550A > [3.782503] Linux agpgart interface v0.103 > [3.787805] ACPI: bus type drm_connector registered > > <--- boot stops here. > > It should continue with this: > > [3.795491] Console: switching to colour dummy device 80x25 > [3.801933] nouveau :03:00.0: vgaarb: deactivate vga console > [3.808303] nouveau :03:00.0: NVIDIA GT218 (0a8c00b1) > [3.931002] nouveau :03:00.0: bios: version 70.18.83.00.08 > [3.941731] nouveau :03:00.0: fb: 512 MiB DDR3 > [4.110348] tsc: Refined TSC clocksource calibration: 3591.349 MHz > [4.116627] clocksource: tsc: mask: 0x max_cycles: > 0x33c466a1ab5, max_idle_ns: 440795209767 ns > [4.126871] clocksource: Switched to clocksource tsc > [4.252013] nouveau :03:00.0: DRM: VRAM: 512 MiB > [4.257088] nouveau :03:00.0: DRM: GART: 1048576 MiB > [4.262501] nouveau :03:00.0: DRM: TMDS table version 2.0 > [4.268333] nouveau :03:00.0: DRM: DCB version 4.0 > [4.273561] nouveau :03:00.0: DRM: DCB outp 00: 02000360 > [4.280104] nouveau :03:00.0: DRM: DCB outp 01: 02000362 00020010 > [4.286630] nouveau :03:00.0: DRM: DCB outp 02: 028003a6 0f220010 > [4.293176] nouveau :03:00.0: DRM: DCB outp 03: 01011380 > [4.299711] nouveau :03:00.0: DRM: DCB outp 04: 08011382 00020010 > [4.306243] nouveau :03:00.0: DRM: DCB outp 05: 088113c6 0f220010 > [4.312772] nouveau :03:00.0: DRM: DCB conn 00: 00101064 > [4.318520] nouveau :03:00.0: DRM: DCB conn 01: 00202165 > [4.329488] nouveau :03:00.0: DRM: MM: using COPY for buffer copies > [4.336261] stackdepot: allocating hash table of 1048576 entries via > kvcalloc > ... > >
