drm/prime: Ensure mmap offset is initialized
Subsystems (3)
PATCH 1/1 drm/prime: Ensure mmap offset is initialized
From: Rob Clark <[email protected]>
If a GEM object is allocated, and then exported as a dma-buf fd which is mmap'd before or without the GEM buffer being directly mmap'd, the vma_node could be unitialized. This leads to a situation where the CPU mapping is not correctly torn down in drm_vma_node_unmap().
---
Note, it's possible the issue existed in some related form prior to the commit tagged with Fixes.
drivers/gpu/drm/drm_prime.c | 5 +++++ 1 file changed, 5 insertions(+)
drivers/gpu/drm/drm_prime.c +5 -0
@@ -716,6 +716,11 @@ int drm_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)struct file *fil;int ret; +/* Ensure that the vma_node is initialized: */ + ret =drm_gem_create_mmap_offset (obj); +if (ret) +return ret; +/* Add the fake offset */ vma->vm_pgoff +=drm_vma_node_start (&obj->vma_node);
Signature · 1 line
2.35.3
Thomas Zimmermann
maintainerHi
Am 29.05.22 um 18:29 schrieb Rob Clark:
Quoted 6 lines
From: Rob Clark <[email protected]>
If a GEM object is allocated, and then exported as a dma-buf fd which is mmap'd before or without the GEM buffer being directly mmap'd, the
vma_nodecould be unitialized. This leads to a situation where the CPU mapping is not correctly torn down indrm_vma_node_unmap().
Which drivers are affected by this problem?
I checked several drivers and most appear to be initializing the offset during object construction, such as GEM SHMEM. [1] TTM-based drivers also seem unaffected. [2]
From a quick grep, only etnaviv, msm and omapdrm appear to be affected? They only seem to run drm_gem_create_mmap_offset() from their ioctl-handling code.
If so, I'd say it's preferable to fix these drivers and put a drm_WARN_ONCE() into drm_gem_prime_mmap().
Best regards
Thomas
[1] https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/drm_gem_shmem_helper.c#L85
[2] https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/ttm/ttm_bo.c#L1002
Quoted 26 lines
Fixes e5516553999f ("drm: call drm_gem_object_funcs.mmap with fake offset")Signed-off-by Rob Clark <[email protected]>---
Note, it's possible the issue existed in some related form prior to the commit tagged with Fixes.drivers/gpu/drm/drm_prime.c | 5 +++++ 1 file changed, 5 insertions(+)drivers/gpu/drm/drm_prime.c +5
-0@@ -716,6 +716,11 @@ int drm_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)struct file *fil;int ret; +/* Ensure that the vma_node is initialized: */ + ret =drm_gem_create_mmap_offset (obj); +if (ret) +return ret; +/* Add the fake offset */ vma->vm_pgoff +=drm_vma_node_start (&obj->vma_node);
Signature · 6 lines
Thomas Zimmermann Graphics Driver Developer SUSE Software Solutions Germany GmbH Maxfeldstr. 5, 90409 Nürnberg, Germany (HRB 36809, AG Nürnberg) Geschäftsführer: Ivo Totev
Quoted 17 lines · Thomas Zimmermann
Hi
Am 29.05.22 um 18:29 schrieb Rob Clark:
Quoted 6 lines
From: Rob Clark <[email protected]>
If a GEM object is allocated, and then exported as a dma-buf fd which is mmap'd before or without the GEM buffer being directly mmap'd, the
vma_nodecould be unitialized. This leads to a situation where the CPU mapping is not correctly torn down indrm_vma_node_unmap().Which drivers are affected by this problem?
I checked several drivers and most appear to be initializing the offset during object construction, such as GEM SHMEM. [1] TTM-based drivers also seem unaffected. [2]
Quoted 6 lines
From a quick grep, only etnaviv, msm and omapdrm appear to be affected?
They only seem to rundrm_gem_create_mmap_offset()from their ioctl-handling code.If so, I'd say it's preferable to fix these drivers and put a
drm_WARN_ONCE()intodrm_gem_prime_mmap().
That is good if fewer drivers are affected, however I disagree with your proposal. At least for freedreno userspace, a lot of bo's never get mmap'd (either directly of via dmabuf), so we should not be allocating a mmap offset unnecessarily.
BR,
-R
Quoted 42 lines
Best regards
Thomas[1]
https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/drm_gem_shmem_helper.c#L85
[2]
https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/ttm/ttm_bo.c#L1002Quoted 16 lines
Fixes e5516553999f ("drm: call drm_gem_object_funcs.mmap with fake offset")Signed-off-by Rob Clark <[email protected]>---
Note, it's possible the issue existed in some related form prior to the commit tagged with Fixes.drivers/gpu/drm/drm_prime.c | 5 +++++ 1 file changed, 5 insertions(+)drivers/gpu/drm/drm_prime.c +0
-0@@ -716,6 +716,11 @@ int drm_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)struct file *fil;drivers/gpu/drm/drm_prime.c int drm_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma) · line 726
Quoted 10 lines
drivers/gpu/drm/drm_prime.c +5
-0int ret; +/* Ensure that the vma_node is initialized: */ + ret =drm_gem_create_mmap_offset (obj); +if (ret) +return ret; +/* Add the fake offset */ vma->vm_pgoff +=drm_vma_node_start (&obj->vma_node);--
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Maxfeldstr. 5, 90409 Nürnberg, Germany
(HRB 36809, AG Nürnberg)
Geschäftsführer: Ivo Totev
Thomas Zimmermann
maintainerHi
Am 30.05.22 um 15:47 schrieb Rob Clark:
Quoted 22 lines · Thomas Zimmermann
Quoted 21 lines
Hi
Am 29.05.22 um 18:29 schrieb Rob Clark:
Quoted 6 lines
From: Rob Clark <[email protected]>
If a GEM object is allocated, and then exported as a dma-buf fd which is mmap'd before or without the GEM buffer being directly mmap'd, the
vma_nodecould be unitialized. This leads to a situation where the CPU mapping is not correctly torn down indrm_vma_node_unmap().Which drivers are affected by this problem?
I checked several drivers and most appear to be initializing the offset during object construction, such as GEM SHMEM. [1] TTM-based drivers also seem unaffected. [2]
From a quick grep, only etnaviv, msm and omapdrm appear to be affected?
They only seem to rundrm_gem_create_mmap_offset()from their ioctl-handling code.
Quoted 7 lines
Quoted 2 lines
If so, I'd say it's preferable to fix these drivers and put a
drm_WARN_ONCE()intodrm_gem_prime_mmap().That is good if fewer drivers are affected, however I disagree with your proposal. At least for freedreno userspace, a lot of bo's never get mmap'd (either directly of via dmabuf), so we should not be allocating a mmap offset unnecessarily.
I see.
I the reason I'm arguing against the current patch is that the fix appears like a workaround and 6 months from now, few will remember why it's there. Especially since most drivers initialize the offset correctly. (Not too long ago, I refactored the handling of these mmap calls throughout DRM drivers and it was confusing at times.)
So here's another suggestion: I further looked at the 3 drivers that I mentioned. etnaviv and msm can easily wrap the call to drm_gem_prime_mmap() and init the offset first. [1][2] omapdrm doesn't actually use drm_gem_prime_mmap(). The offset can instead be initialized at the top of the driver's dmabuf mmap function. [3]
Best regards
Thomas
[1] https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/etnaviv/etnaviv_drv.c#L480
[2] https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/msm/msm_drv.c#L961
[3] https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/omapdrm/omap_gem_dmabuf.c#L66
Quoted 46 lines
BR,
-RQuoted 42 lines
Best regards
Thomas[1]
https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/drm_gem_shmem_helper.c#L85
[2]
https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/ttm/ttm_bo.c#L1002Quoted 16 lines
Fixes e5516553999f ("drm: call drm_gem_object_funcs.mmap with fake offset")Signed-off-by Rob Clark <[email protected]>---
Note, it's possible the issue existed in some related form prior to the commit tagged with Fixes.drivers/gpu/drm/drm_prime.c | 5 +++++ 1 file changed, 5 insertions(+)drivers/gpu/drm/drm_prime.c +0
-0@@ -716,6 +716,11 @@ int drm_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)struct file *fil;drivers/gpu/drm/drm_prime.c int drm_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma) · line 726
Quoted 10 lines
drivers/gpu/drm/drm_prime.c +5
-0int ret; +/* Ensure that the vma_node is initialized: */ + ret =drm_gem_create_mmap_offset (obj); +if (ret) +return ret; +/* Add the fake offset */ vma->vm_pgoff +=drm_vma_node_start (&obj->vma_node);--
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Maxfeldstr. 5, 90409 Nürnberg, Germany
(HRB 36809, AG Nürnberg)
Geschäftsführer: Ivo Totev
Signature · 6 lines
Thomas Zimmermann Graphics Driver Developer SUSE Software Solutions Germany GmbH Maxfeldstr. 5, 90409 Nürnberg, Germany (HRB 36809, AG Nürnberg) Geschäftsführer: Ivo Totev
Quoted 34 lines · Thomas Zimmermann
Hi
Am 30.05.22 um 15:47 schrieb Rob Clark:
Quoted 22 lines · Thomas Zimmermann
Quoted 21 lines
Hi
Am 29.05.22 um 18:29 schrieb Rob Clark:
Quoted 6 lines
From: Rob Clark <[email protected]>
If a GEM object is allocated, and then exported as a dma-buf fd which is mmap'd before or without the GEM buffer being directly mmap'd, the
vma_nodecould be unitialized. This leads to a situation where the CPU mapping is not correctly torn down indrm_vma_node_unmap().Which drivers are affected by this problem?
I checked several drivers and most appear to be initializing the offset during object construction, such as GEM SHMEM. [1] TTM-based drivers also seem unaffected. [2]
From a quick grep, only etnaviv, msm and omapdrm appear to be affected?
They only seem to rundrm_gem_create_mmap_offset()from their ioctl-handling code.Quoted 7 lines
Quoted 2 lines
If so, I'd say it's preferable to fix these drivers and put a
drm_WARN_ONCE()intodrm_gem_prime_mmap().That is good if fewer drivers are affected, however I disagree with your proposal. At least for freedreno userspace, a lot of bo's never get mmap'd (either directly of via dmabuf), so we should not be allocating a mmap offset unnecessarily.
Quoted 7 lines
I see.
I the reason I'm arguing against the current patch is that the fix appears like a workaround and 6 months from now, few will remember why it's there. Especially since most drivers initialize the offset correctly. (Not too long ago, I refactored the handling of these mmap calls throughout DRM drivers and it was confusing at times.)
I dispute the "correctly" part.. and that this is a workaround ;-)
But I can send a v2 with the addition of a comment explaining the reason, so git-blame archeology isn't required to understand the reasoning
BR,
-R
Quoted 70 lines
So here's another suggestion: I further looked at the 3 drivers that I mentioned. etnaviv and msm can easily wrap the call to
drm_gem_prime_mmap()and init the offset first. [1][2] omapdrm doesn't actually usedrm_gem_prime_mmap(). The offset can instead be initialized at the top of the driver's dmabuf mmap function. [3]Best regards
Thomas[1]
https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/etnaviv/etnaviv_drv.c#L480
[2]
https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/msm/msm_drv.c#L961
[3]
https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/omapdrm/omap_gem_dmabuf.c#L66Quoted 46 lines
BR,
-RQuoted 42 lines
Best regards
Thomas[1]
https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/drm_gem_shmem_helper.c#L85
[2]
https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/ttm/ttm_bo.c#L1002Quoted 16 lines
Fixes e5516553999f ("drm: call drm_gem_object_funcs.mmap with fake offset")Signed-off-by Rob Clark <[email protected]>---
Note, it's possible the issue existed in some related form prior to the commit tagged with Fixes.drivers/gpu/drm/drm_prime.c | 5 +++++ 1 file changed, 5 insertions(+)drivers/gpu/drm/drm_prime.c +0
-0@@ -716,6 +716,11 @@ int drm_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)struct file *fil;drivers/gpu/drm/drm_prime.c int drm_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma) · line 726
Quoted 10 lines
drivers/gpu/drm/drm_prime.c +5
-0int ret; +/* Ensure that the vma_node is initialized: */ + ret =drm_gem_create_mmap_offset (obj); +if (ret) +return ret; +/* Add the fake offset */ vma->vm_pgoff +=drm_vma_node_start (&obj->vma_node);--
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Maxfeldstr. 5, 90409 Nürnberg, Germany
(HRB 36809, AG Nürnberg)
Geschäftsführer: Ivo Totev--
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Maxfeldstr. 5, 90409 Nürnberg, Germany
(HRB 36809, AG Nürnberg)
Geschäftsführer: Ivo Totev
Quoted 23 lines · Rob Clark
Quoted 21 lines · Thomas Zimmermann
Hi
Am 29.05.22 um 18:29 schrieb Rob Clark:
Quoted 6 lines
From: Rob Clark <[email protected]>
If a GEM object is allocated, and then exported as a dma-buf fd which is mmap'd before or without the GEM buffer being directly mmap'd, the
vma_nodecould be unitialized. This leads to a situation where the CPU mapping is not correctly torn down indrm_vma_node_unmap().Which drivers are affected by this problem?
I checked several drivers and most appear to be initializing the offset during object construction, such as GEM SHMEM. [1] TTM-based drivers also seem unaffected. [2]
From a quick grep, only etnaviv, msm and omapdrm appear to be affected?
They only seem to rundrm_gem_create_mmap_offset()from their ioctl-handling code.
Quoted 7 lines
Quoted 2 lines
If so, I'd say it's preferable to fix these drivers and put a
drm_WARN_ONCE()intodrm_gem_prime_mmap().That is good if fewer drivers are affected, however I disagree with your proposal. At least for freedreno userspace, a lot of bo's never get mmap'd (either directly of via dmabuf), so we should not be allocating a mmap offset unnecessarily.
Does this actually matter in the grand scheme of things? We originally allocated mmap offset only on demand because userspace only had 32bit loff_t support and so simply couldn't mmap anything if the offset ended up above 32bit (even if there was still va space available).
But those days are long gone (about 10 years or so) and the allocation overhead for an mmap offset is tiny. So I think unless you can benchmark an impact allocating it at bo alloc seems like the simplest design overall, and hence what we should be doing. And if the vma offset allocation every gets too slow due to fragmentation we can lift the hole tree from i915 into drm_mm and the job should be done. At that point we could also allocate the offset unconditionally in the gem_init function and be done with it.
Iow I concur with Thomas here, unless there's hard data contrary simplicity imo trumps here.
-Daniel
Quoted 46 lines
BR,
-RQuoted 42 lines
Best regards
Thomas[1]
https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/drm_gem_shmem_helper.c#L85
[2]
https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/ttm/ttm_bo.c#L1002Quoted 16 lines
Fixes e5516553999f ("drm: call drm_gem_object_funcs.mmap with fake offset")Signed-off-by Rob Clark <[email protected]>---
Note, it's possible the issue existed in some related form prior to the commit tagged with Fixes.drivers/gpu/drm/drm_prime.c | 5 +++++ 1 file changed, 5 insertions(+)drivers/gpu/drm/drm_prime.c +0
-0@@ -716,6 +716,11 @@ int drm_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)struct file *fil;drivers/gpu/drm/drm_prime.c int drm_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma) · line 726
Quoted 10 lines
drivers/gpu/drm/drm_prime.c +5
-0int ret; +/* Ensure that the vma_node is initialized: */ + ret =drm_gem_create_mmap_offset (obj); +if (ret) +return ret; +/* Add the fake offset */ vma->vm_pgoff +=drm_vma_node_start (&obj->vma_node);--
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Maxfeldstr. 5, 90409 Nürnberg, Germany
(HRB 36809, AG Nürnberg)
Geschäftsführer: Ivo Totev
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch
Quoted 47 lines · Daniel Vetter
Quoted 23 lines · Rob Clark
Quoted 21 lines · Thomas Zimmermann
Hi
Am 29.05.22 um 18:29 schrieb Rob Clark:
Quoted 6 lines
From: Rob Clark <[email protected]>
If a GEM object is allocated, and then exported as a dma-buf fd which is mmap'd before or without the GEM buffer being directly mmap'd, the
vma_nodecould be unitialized. This leads to a situation where the CPU mapping is not correctly torn down indrm_vma_node_unmap().Which drivers are affected by this problem?
I checked several drivers and most appear to be initializing the offset during object construction, such as GEM SHMEM. [1] TTM-based drivers also seem unaffected. [2]
From a quick grep, only etnaviv, msm and omapdrm appear to be affected?
They only seem to rundrm_gem_create_mmap_offset()from their ioctl-handling code.Quoted 7 lines
Quoted 2 lines
If so, I'd say it's preferable to fix these drivers and put a
drm_WARN_ONCE()intodrm_gem_prime_mmap().That is good if fewer drivers are affected, however I disagree with your proposal. At least for freedreno userspace, a lot of bo's never get mmap'd (either directly of via dmabuf), so we should not be allocating a mmap offset unnecessarily.
Does this actually matter in the grand scheme of things? We originally allocated mmap offset only on demand because userspace only had 32bit
loff_tsupport and so simply couldn't mmap anything if the offset ended up above 32bit (even if there was still va space available).But those days are long gone (about 10 years or so) and the allocation overhead for an mmap offset is tiny. So I think unless you can benchmark an impact allocating it at bo alloc seems like the simplest design overall, and hence what we should be doing. And if the vma offset allocation every gets too slow due to fragmentation we can lift the hole tree from i915 into
drm_mmand the job should be done. At that point we could also allocate the offset unconditionally in thegem_initfunction and be done with it.
Quoted 2 lines
Iow I concur with Thomas here, unless there's hard data contrary simplicity imo trumps here.
32b userspace is still alive and well, at least on arm chromebooks ;-)
BR,
-R
Quoted 55 lines
-Daniel
Quoted 46 lines
BR,
-RQuoted 42 lines
Best regards
Thomas[1]
https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/drm_gem_shmem_helper.c#L85
[2]
https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/ttm/ttm_bo.c#L1002Quoted 16 lines
Fixes e5516553999f ("drm: call drm_gem_object_funcs.mmap with fake offset")Signed-off-by Rob Clark <[email protected]>---
Note, it's possible the issue existed in some related form prior to the commit tagged with Fixes.drivers/gpu/drm/drm_prime.c | 5 +++++ 1 file changed, 5 insertions(+)drivers/gpu/drm/drm_prime.c +0
-0@@ -716,6 +716,11 @@ int drm_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)struct file *fil;drivers/gpu/drm/drm_prime.c int drm_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma) · line 726
Quoted 10 lines
drivers/gpu/drm/drm_prime.c +5
-0int ret; +/* Ensure that the vma_node is initialized: */ + ret =drm_gem_create_mmap_offset (obj); +if (ret) +return ret; +/* Add the fake offset */ vma->vm_pgoff +=drm_vma_node_start (&obj->vma_node);--
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Maxfeldstr. 5, 90409 Nürnberg, Germany
(HRB 36809, AG Nürnberg)
Geschäftsführer: Ivo Totev--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
Thomas Zimmermann
maintainerHi
Am 30.05.22 um 17:41 schrieb Rob Clark:
Quoted 48 lines · Daniel Vetter
Quoted 47 lines · Rob Clark
Quoted 23 lines · Thomas Zimmermann
Quoted 21 lines
Hi
Am 29.05.22 um 18:29 schrieb Rob Clark:
Quoted 6 lines
From: Rob Clark <[email protected]>
If a GEM object is allocated, and then exported as a dma-buf fd which is mmap'd before or without the GEM buffer being directly mmap'd, the
vma_nodecould be unitialized. This leads to a situation where the CPU mapping is not correctly torn down indrm_vma_node_unmap().Which drivers are affected by this problem?
I checked several drivers and most appear to be initializing the offset during object construction, such as GEM SHMEM. [1] TTM-based drivers also seem unaffected. [2]
From a quick grep, only etnaviv, msm and omapdrm appear to be affected?
They only seem to rundrm_gem_create_mmap_offset()from their ioctl-handling code.Quoted 7 lines
Quoted 2 lines
If so, I'd say it's preferable to fix these drivers and put a
drm_WARN_ONCE()intodrm_gem_prime_mmap().That is good if fewer drivers are affected, however I disagree with your proposal. At least for freedreno userspace, a lot of bo's never get mmap'd (either directly of via dmabuf), so we should not be allocating a mmap offset unnecessarily.
Does this actually matter in the grand scheme of things? We originally allocated mmap offset only on demand because userspace only had 32bit
loff_tsupport and so simply couldn't mmap anything if the offset ended up above 32bit (even if there was still va space available).But those days are long gone (about 10 years or so) and the allocation overhead for an mmap offset is tiny. So I think unless you can benchmark an impact allocating it at bo alloc seems like the simplest design overall, and hence what we should be doing. And if the vma offset allocation every gets too slow due to fragmentation we can lift the hole tree from i915 into
drm_mmand the job should be done. At that point we could also allocate the offset unconditionally in thegem_initfunction and be done with it.
Quoted 4 lines
Quoted 2 lines
Iow I concur with Thomas here, unless there's hard data contrary simplicity imo trumps here.
32b userspace is still alive and well, at least on arm chromebooks ;-)
I mostly dislike the inconsistency among drivers. If we want to create the offset on-demand in the DRM helpers, we should do so for all drivers. At least our generic GEM helpers and TTM should implement this pattern.
Best regards
Thomas
Quoted 59 lines
BR,
-RQuoted 55 lines
-Daniel
Quoted 46 lines
BR,
-RQuoted 42 lines
Best regards
Thomas[1]
https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/drm_gem_shmem_helper.c#L85
[2]
https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/ttm/ttm_bo.c#L1002Quoted 16 lines
Fixes e5516553999f ("drm: call drm_gem_object_funcs.mmap with fake offset")Signed-off-by Rob Clark <[email protected]>---
Note, it's possible the issue existed in some related form prior to the commit tagged with Fixes.drivers/gpu/drm/drm_prime.c | 5 +++++ 1 file changed, 5 insertions(+)drivers/gpu/drm/drm_prime.c +0
-0@@ -716,6 +716,11 @@ int drm_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)struct file *fil;drivers/gpu/drm/drm_prime.c int drm_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma) · line 726
Quoted 10 lines
drivers/gpu/drm/drm_prime.c +5
-0int ret; +/* Ensure that the vma_node is initialized: */ + ret =drm_gem_create_mmap_offset (obj); +if (ret) +return ret; +/* Add the fake offset */ vma->vm_pgoff +=drm_vma_node_start (&obj->vma_node);--
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Maxfeldstr. 5, 90409 Nürnberg, Germany
(HRB 36809, AG Nürnberg)
Geschäftsführer: Ivo Totev--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
Signature · 6 lines
Thomas Zimmermann Graphics Driver Developer SUSE Software Solutions Germany GmbH Maxfeldstr. 5, 90409 Nürnberg, Germany (HRB 36809, AG Nürnberg) Geschäftsführer: Ivo Totev
Quoted 55 lines · Thomas Zimmermann
Hi
Am 30.05.22 um 17:41 schrieb Rob Clark:
Quoted 51 lines · Daniel Vetter
Quoted 49 lines · Rob Clark
Quoted 23 lines · Thomas Zimmermann
Quoted 21 lines
Hi
Am 29.05.22 um 18:29 schrieb Rob Clark:
Quoted 6 lines
From: Rob Clark <[email protected]>
If a GEM object is allocated, and then exported as a dma-buf fd which is mmap'd before or without the GEM buffer being directly mmap'd, the
vma_nodecould be unitialized. This leads to a situation where the CPU mapping is not correctly torn down indrm_vma_node_unmap().Which drivers are affected by this problem?
I checked several drivers and most appear to be initializing the offset during object construction, such as GEM SHMEM. [1] TTM-based drivers also seem unaffected. [2]
From a quick grep, only etnaviv, msm and omapdrm appear to be affected?
They only seem to rundrm_gem_create_mmap_offset()from their ioctl-handling code.Quoted 7 lines
Quoted 2 lines
If so, I'd say it's preferable to fix these drivers and put a
drm_WARN_ONCE()intodrm_gem_prime_mmap().That is good if fewer drivers are affected, however I disagree with your proposal. At least for freedreno userspace, a lot of bo's never get mmap'd (either directly of via dmabuf), so we should not be allocating a mmap offset unnecessarily.
Does this actually matter in the grand scheme of things? We originally allocated mmap offset only on demand because userspace only had 32bit
loff_tsupport and so simply couldn't mmap anything if the offset ended up above 32bit (even if there was still va space available).But those days are long gone (about 10 years or so) and the allocation overhead for an mmap offset is tiny. So I think unless you can benchmark an impact allocating it at bo alloc seems like the simplest design overall, and hence what we should be doing. And if the vma offset allocation every gets too slow due to fragmentation we can lift the hole tree from i915 into
drm_mmand the job should be done. At that point we could also allocate the offset unconditionally in thegem_initfunction and be done with it.Iow I concur with Thomas here, unless there's hard data contrary simplicity imo trumps here.
Quoted 6 lines
Quoted 1 line
32b userspace is still alive and well, at least on arm chromebooks ;-)
I mostly dislike the inconsistency among drivers. If we want to create the offset on-demand in the DRM helpers, we should do so for all drivers. At least our generic GEM helpers and TTM should implement this pattern.
Possibly we should have drm_gem_get_mmap_offset() which combines drm_gem_create_mmap_offset() and drm_vma_node_start() calls, and use that everywhere.
But I think we should fix this issue first, and then refactor on top, so that a fix can be backported to stable kernels ;-)
BR,
-R
Quoted 70 lines
Best regards
ThomasQuoted 59 lines
BR,
-RQuoted 55 lines
-Daniel
Quoted 46 lines
BR,
-RQuoted 42 lines
Best regards
Thomas[1]
https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/drm_gem_shmem_helper.c#L85
[2]
https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/ttm/ttm_bo.c#L1002Quoted 16 lines
Fixes e5516553999f ("drm: call drm_gem_object_funcs.mmap with fake offset")Signed-off-by Rob Clark <[email protected]>---
Note, it's possible the issue existed in some related form prior to the commit tagged with Fixes.drivers/gpu/drm/drm_prime.c | 5 +++++ 1 file changed, 5 insertions(+)drivers/gpu/drm/drm_prime.c +0
-0@@ -716,6 +716,11 @@ int drm_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)struct file *fil;drivers/gpu/drm/drm_prime.c int drm_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma) · line 726
Quoted 10 lines
drivers/gpu/drm/drm_prime.c +5
-0int ret; +/* Ensure that the vma_node is initialized: */ + ret =drm_gem_create_mmap_offset (obj); +if (ret) +return ret; +/* Add the fake offset */ vma->vm_pgoff +=drm_vma_node_start (&obj->vma_node);--
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Maxfeldstr. 5, 90409 Nürnberg, Germany
(HRB 36809, AG Nürnberg)
Geschäftsführer: Ivo Totev--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch--
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Maxfeldstr. 5, 90409 Nürnberg, Germany
(HRB 36809, AG Nürnberg)
Geschäftsführer: Ivo Totev
Quoted 49 lines · Rob Clark
Quoted 47 lines · Daniel Vetter
Quoted 23 lines · Rob Clark
Quoted 21 lines · Thomas Zimmermann
Hi
Am 29.05.22 um 18:29 schrieb Rob Clark:
Quoted 6 lines
From: Rob Clark <[email protected]>
If a GEM object is allocated, and then exported as a dma-buf fd which is mmap'd before or without the GEM buffer being directly mmap'd, the
vma_nodecould be unitialized. This leads to a situation where the CPU mapping is not correctly torn down indrm_vma_node_unmap().Which drivers are affected by this problem?
I checked several drivers and most appear to be initializing the offset during object construction, such as GEM SHMEM. [1] TTM-based drivers also seem unaffected. [2]
From a quick grep, only etnaviv, msm and omapdrm appear to be affected?
They only seem to rundrm_gem_create_mmap_offset()from their ioctl-handling code.Quoted 7 lines
Quoted 2 lines
If so, I'd say it's preferable to fix these drivers and put a
drm_WARN_ONCE()intodrm_gem_prime_mmap().That is good if fewer drivers are affected, however I disagree with your proposal. At least for freedreno userspace, a lot of bo's never get mmap'd (either directly of via dmabuf), so we should not be allocating a mmap offset unnecessarily.
Does this actually matter in the grand scheme of things? We originally allocated mmap offset only on demand because userspace only had 32bit
loff_tsupport and so simply couldn't mmap anything if the offset ended up above 32bit (even if there was still va space available).But those days are long gone (about 10 years or so) and the allocation overhead for an mmap offset is tiny. So I think unless you can benchmark an impact allocating it at bo alloc seems like the simplest design overall, and hence what we should be doing. And if the vma offset allocation every gets too slow due to fragmentation we can lift the hole tree from i915 into
drm_mmand the job should be done. At that point we could also allocate the offset unconditionally in thegem_initfunction and be done with it.
Quoted 4 lines
Quoted 2 lines
Iow I concur with Thomas here, unless there's hard data contrary simplicity imo trumps here.
32b userspace is still alive and well, at least on arm chromebooks ;-)
There's lots of different 32b userspace. The old thing was about userspace which didn't use mmap64, but only mmap. Which could only mmap the lower 4GB of a file, and so if you ended up with mmap_offset above 4G then you'd blow up.
But mmap64 is a thing since forever, and if you compile with the right glibc switch (loff_t is the magic thing iirc) it all works even with default mmap. So I really don't think you should have this problem anymore (except when cros is doing something really, really silly). -Daniel
Quoted 59 lines
BR,
-RQuoted 55 lines
-Daniel
Quoted 46 lines
BR,
-RQuoted 42 lines
Best regards
Thomas[1]
https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/drm_gem_shmem_helper.c#L85
[2]
https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/ttm/ttm_bo.c#L1002Quoted 16 lines
Fixes e5516553999f ("drm: call drm_gem_object_funcs.mmap with fake offset")Signed-off-by Rob Clark <[email protected]>---
Note, it's possible the issue existed in some related form prior to the commit tagged with Fixes.drivers/gpu/drm/drm_prime.c | 5 +++++ 1 file changed, 5 insertions(+)drivers/gpu/drm/drm_prime.c +0
-0@@ -716,6 +716,11 @@ int drm_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)struct file *fil;drivers/gpu/drm/drm_prime.c int drm_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma) · line 726
Quoted 10 lines
drivers/gpu/drm/drm_prime.c +5
-0int ret; +/* Ensure that the vma_node is initialized: */ + ret =drm_gem_create_mmap_offset (obj); +if (ret) +return ret; +/* Add the fake offset */ vma->vm_pgoff +=drm_vma_node_start (&obj->vma_node);--
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Maxfeldstr. 5, 90409 Nürnberg, Germany
(HRB 36809, AG Nürnberg)
Geschäftsführer: Ivo Totev--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch
Quoted 61 lines · Daniel Vetter
Quoted 49 lines · Rob Clark
Quoted 47 lines · Daniel Vetter
Quoted 23 lines · Rob Clark
Quoted 21 lines · Thomas Zimmermann
Hi
Am 29.05.22 um 18:29 schrieb Rob Clark:
Quoted 6 lines
From: Rob Clark <[email protected]>
If a GEM object is allocated, and then exported as a dma-buf fd which is mmap'd before or without the GEM buffer being directly mmap'd, the
vma_nodecould be unitialized. This leads to a situation where the CPU mapping is not correctly torn down indrm_vma_node_unmap().Which drivers are affected by this problem?
I checked several drivers and most appear to be initializing the offset during object construction, such as GEM SHMEM. [1] TTM-based drivers also seem unaffected. [2]
From a quick grep, only etnaviv, msm and omapdrm appear to be affected?
They only seem to rundrm_gem_create_mmap_offset()from their ioctl-handling code.Quoted 7 lines
Quoted 2 lines
If so, I'd say it's preferable to fix these drivers and put a
drm_WARN_ONCE()intodrm_gem_prime_mmap().That is good if fewer drivers are affected, however I disagree with your proposal. At least for freedreno userspace, a lot of bo's never get mmap'd (either directly of via dmabuf), so we should not be allocating a mmap offset unnecessarily.
Does this actually matter in the grand scheme of things? We originally allocated mmap offset only on demand because userspace only had 32bit
loff_tsupport and so simply couldn't mmap anything if the offset ended up above 32bit (even if there was still va space available).But those days are long gone (about 10 years or so) and the allocation overhead for an mmap offset is tiny. So I think unless you can benchmark an impact allocating it at bo alloc seems like the simplest design overall, and hence what we should be doing. And if the vma offset allocation every gets too slow due to fragmentation we can lift the hole tree from i915 into
drm_mmand the job should be done. At that point we could also allocate the offset unconditionally in thegem_initfunction and be done with it.Quoted 4 lines
Quoted 2 lines
Iow I concur with Thomas here, unless there's hard data contrary simplicity imo trumps here.
32b userspace is still alive and well, at least on arm chromebooks ;-)
There's lots of different 32b userspace. The old thing was about userspace which didn't use mmap64, but only mmap. Which could only mmap the lower 4GB of a file, and so if you ended up with
mmap_offsetabove 4G then you'd blow up.
Quoted 4 lines
But mmap64 is a thing since forever, and if you compile with the right glibc switch (
loff_tis the magic thing iirc) it all works even with default mmap. So I really don't think you should have this problem anymore (except when cros is doing something really, really silly).
The other thing, not sure quite how much it matters, is the vma_offset_manager size is smaller with 32b kernels, which is still a thing on some devices.
But at any rate, not allocating a mmap offset when it isn't needed still seems like an eminently reasonable thing to do. And IMO my fix is quite reasonable too. But if you disagree I can workaround the core helpers in the driver.
BR,
-R
Quoted 68 lines
-Daniel
Quoted 59 lines
BR,
-RQuoted 55 lines
-Daniel
Quoted 46 lines
BR,
-RQuoted 42 lines
Best regards
Thomas[1]
https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/drm_gem_shmem_helper.c#L85
[2]
https://elixir.bootlin.com/linux/v5.18/source/drivers/gpu/drm/ttm/ttm_bo.c#L1002Quoted 16 lines
Fixes e5516553999f ("drm: call drm_gem_object_funcs.mmap with fake offset")Signed-off-by Rob Clark <[email protected]>---
Note, it's possible the issue existed in some related form prior to the commit tagged with Fixes.drivers/gpu/drm/drm_prime.c | 5 +++++ 1 file changed, 5 insertions(+)drivers/gpu/drm/drm_prime.c +0
-0@@ -716,6 +716,11 @@ int drm_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)struct file *fil;drivers/gpu/drm/drm_prime.c int drm_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma) · line 726
Quoted 10 lines
drivers/gpu/drm/drm_prime.c +5
-0int ret; +/* Ensure that the vma_node is initialized: */ + ret =drm_gem_create_mmap_offset (obj); +if (ret) +return ret; +/* Add the fake offset */ vma->vm_pgoff +=drm_vma_node_start (&obj->vma_node);--
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Maxfeldstr. 5, 90409 Nürnberg, Germany
(HRB 36809, AG Nürnberg)
Geschäftsführer: Ivo Totev--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch