From: Tvrtko Ursulin tvrtko.ursulin@intel.com
This is an alternative proposed fix for the below references bug report where dma fence error propagation is causing undesirable change in behaviour post GPU hang/reset.
Approach in this patch is to simply stop propagating all dma fence errors by default since that seems to be the upstream ask.
To handle the case where i915 needs error propagation for security, I add a new dma fence flag DMA_FENCE_FLAG_PROPAGATE_ERROR and make use of it in the command parsing chain only.
It sounds a plausible argument that fence propagation could be useful in which case a core flag to enable opt-in should be universally useful.
Signed-off-by: Tvrtko Ursulin tvrtko.ursulin@intel.com Reported-by: Marcin Slusarz marcin.slusarz@intel.com Reported-by: Miroslav Bendik References: 9e31c1fe45d5 ("drm/i915: Propagate errors on awaiting already signaled fences") References: https://gitlab.freedesktop.org/drm/intel/-/issues/3080 Cc: Jason Ekstrand jason.ekstrand@intel.com Cc: Daniel Vetter daniel.vetter@ffwll.ch --- drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c | 2 ++ drivers/gpu/drm/i915/i915_sw_fence.c | 8 ++++---- drivers/gpu/drm/i915/i915_sw_fence.h | 8 ++++++++ include/linux/dma-fence.h | 1 + 4 files changed, 15 insertions(+), 4 deletions(-)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c index 297143511f99..6a516d1261d0 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c @@ -2522,6 +2522,8 @@ static int eb_parse_pipeline(struct i915_execbuffer *eb, }
dma_fence_work_init(&pw->base, &eb_parse_ops); + /* Propagate errors for security. */ + __set_bit(DMA_FENCE_FLAG_PROPAGATE_ERROR, &pw->base.dma.flags);
pw->engine = eb->engine; pw->batch = eb->batch->vma; diff --git a/drivers/gpu/drm/i915/i915_sw_fence.c b/drivers/gpu/drm/i915/i915_sw_fence.c index 2744558f3050..2ee917932ccf 100644 --- a/drivers/gpu/drm/i915/i915_sw_fence.c +++ b/drivers/gpu/drm/i915/i915_sw_fence.c @@ -449,7 +449,7 @@ static void dma_i915_sw_fence_wake_timer(struct dma_fence *dma,
fence = xchg(&cb->base.fence, NULL); if (fence) { - i915_sw_fence_set_error_once(fence, dma->error); + i915_sw_fence_propagate_dma_fence_error(fence, dma); i915_sw_fence_complete(fence); }
@@ -480,7 +480,7 @@ int i915_sw_fence_await_dma_fence(struct i915_sw_fence *fence, might_sleep_if(gfpflags_allow_blocking(gfp));
if (dma_fence_is_signaled(dma)) { - i915_sw_fence_set_error_once(fence, dma->error); + i915_sw_fence_propagate_dma_fence_error(fence, dma); return 0; }
@@ -496,7 +496,7 @@ int i915_sw_fence_await_dma_fence(struct i915_sw_fence *fence, if (ret) return ret;
- i915_sw_fence_set_error_once(fence, dma->error); + i915_sw_fence_propagate_dma_fence_error(fence, dma); return 0; }
@@ -548,7 +548,7 @@ int __i915_sw_fence_await_dma_fence(struct i915_sw_fence *fence, debug_fence_assert(fence);
if (dma_fence_is_signaled(dma)) { - i915_sw_fence_set_error_once(fence, dma->error); + i915_sw_fence_propagate_dma_fence_error(fence, dma); return 0; }
diff --git a/drivers/gpu/drm/i915/i915_sw_fence.h b/drivers/gpu/drm/i915/i915_sw_fence.h index 30a863353ee6..872ef80ebd10 100644 --- a/drivers/gpu/drm/i915/i915_sw_fence.h +++ b/drivers/gpu/drm/i915/i915_sw_fence.h @@ -116,4 +116,12 @@ i915_sw_fence_set_error_once(struct i915_sw_fence *fence, int error) cmpxchg(&fence->error, 0, error); }
+static inline void +i915_sw_fence_propagate_dma_fence_error(struct i915_sw_fence *fence, + struct dma_fence *dma) +{ + if (unlikely(test_bit(DMA_FENCE_FLAG_PROPAGATE_ERROR, &dma->flags))) + i915_sw_fence_set_error_once(fence, dma->error); +} + #endif /* _I915_SW_FENCE_H_ */ diff --git a/include/linux/dma-fence.h b/include/linux/dma-fence.h index 6ffb4b2c6371..8dabe1650f11 100644 --- a/include/linux/dma-fence.h +++ b/include/linux/dma-fence.h @@ -99,6 +99,7 @@ enum dma_fence_flag_bits { DMA_FENCE_FLAG_SIGNALED_BIT, DMA_FENCE_FLAG_TIMESTAMP_BIT, DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT, + DMA_FENCE_FLAG_PROPAGATE_ERROR, DMA_FENCE_FLAG_USER_BITS, /* must always be last member */ };
On May 7, 2021 03:35:33 Tvrtko Ursulin tvrtko.ursulin@linux.intel.com wrote:
From: Tvrtko Ursulin tvrtko.ursulin@intel.com
This is an alternative proposed fix for the below references bug report where dma fence error propagation is causing undesirable change in behaviour post GPU hang/reset.
Approach in this patch is to simply stop propagating all dma fence errors by default since that seems to be the upstream ask.
To handle the case where i915 needs error propagation for security, I add a new dma fence flag DMA_FENCE_FLAG_PROPAGATE_ERROR and make use of it in the command parsing chain only.
This sounds plausible. I can't review in full without doing a thorough audit of the surrounding code, though. I'll try to get to that next week if Daniel doesn't beat me to it. Thanks for working on this!
--Jason
It sounds a plausible argument that fence propagation could be useful in which case a core flag to enable opt-in should be universally useful.
Signed-off-by: Tvrtko Ursulin tvrtko.ursulin@intel.com Reported-by: Marcin Slusarz marcin.slusarz@intel.com Reported-by: Miroslav Bendik References: 9e31c1fe45d5 ("drm/i915: Propagate errors on awaiting already signaled fences") References: https://gitlab.freedesktop.org/drm/intel/-/issues/3080 Cc: Jason Ekstrand jason.ekstrand@intel.com Cc: Daniel Vetter daniel.vetter@ffwll.ch
drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c | 2 ++ drivers/gpu/drm/i915/i915_sw_fence.c | 8 ++++---- drivers/gpu/drm/i915/i915_sw_fence.h | 8 ++++++++ include/linux/dma-fence.h | 1 + 4 files changed, 15 insertions(+), 4 deletions(-)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c index 297143511f99..6a516d1261d0 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c @@ -2522,6 +2522,8 @@ static int eb_parse_pipeline(struct i915_execbuffer *eb, }
dma_fence_work_init(&pw->base, &eb_parse_ops);
- /* Propagate errors for security. */
- __set_bit(DMA_FENCE_FLAG_PROPAGATE_ERROR, &pw->base.dma.flags);
pw->engine = eb->engine; pw->batch = eb->batch->vma; diff --git a/drivers/gpu/drm/i915/i915_sw_fence.c b/drivers/gpu/drm/i915/i915_sw_fence.c index 2744558f3050..2ee917932ccf 100644 --- a/drivers/gpu/drm/i915/i915_sw_fence.c +++ b/drivers/gpu/drm/i915/i915_sw_fence.c @@ -449,7 +449,7 @@ static void dma_i915_sw_fence_wake_timer(struct dma_fence *dma,
fence = xchg(&cb->base.fence, NULL); if (fence) {
- i915_sw_fence_set_error_once(fence, dma->error);
- i915_sw_fence_propagate_dma_fence_error(fence, dma);
i915_sw_fence_complete(fence); }
@@ -480,7 +480,7 @@ int i915_sw_fence_await_dma_fence(struct i915_sw_fence *fence, might_sleep_if(gfpflags_allow_blocking(gfp));
if (dma_fence_is_signaled(dma)) {
- i915_sw_fence_set_error_once(fence, dma->error);
- i915_sw_fence_propagate_dma_fence_error(fence, dma);
return 0; }
@@ -496,7 +496,7 @@ int i915_sw_fence_await_dma_fence(struct i915_sw_fence *fence, if (ret) return ret;
- i915_sw_fence_set_error_once(fence, dma->error);
- i915_sw_fence_propagate_dma_fence_error(fence, dma);
return 0; }
@@ -548,7 +548,7 @@ int __i915_sw_fence_await_dma_fence(struct i915_sw_fence *fence, debug_fence_assert(fence);
if (dma_fence_is_signaled(dma)) {
- i915_sw_fence_set_error_once(fence, dma->error);
- i915_sw_fence_propagate_dma_fence_error(fence, dma);
return 0; }
diff --git a/drivers/gpu/drm/i915/i915_sw_fence.h b/drivers/gpu/drm/i915/i915_sw_fence.h index 30a863353ee6..872ef80ebd10 100644 --- a/drivers/gpu/drm/i915/i915_sw_fence.h +++ b/drivers/gpu/drm/i915/i915_sw_fence.h @@ -116,4 +116,12 @@ i915_sw_fence_set_error_once(struct i915_sw_fence *fence, int error) cmpxchg(&fence->error, 0, error); }
+static inline void +i915_sw_fence_propagate_dma_fence_error(struct i915_sw_fence *fence,
- struct dma_fence *dma)
+{
- if (unlikely(test_bit(DMA_FENCE_FLAG_PROPAGATE_ERROR, &dma->flags)))
- i915_sw_fence_set_error_once(fence, dma->error);
+}
#endif /* _I915_SW_FENCE_H_ */ diff --git a/include/linux/dma-fence.h b/include/linux/dma-fence.h index 6ffb4b2c6371..8dabe1650f11 100644 --- a/include/linux/dma-fence.h +++ b/include/linux/dma-fence.h @@ -99,6 +99,7 @@ enum dma_fence_flag_bits { DMA_FENCE_FLAG_SIGNALED_BIT, DMA_FENCE_FLAG_TIMESTAMP_BIT, DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT,
- DMA_FENCE_FLAG_PROPAGATE_ERROR,
DMA_FENCE_FLAG_USER_BITS, /* must always be last member */ };
-- 2.30.2
Intel-gfx mailing list Intel-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/intel-gfx
On Fri, May 07, 2021 at 09:35:21AM +0100, Tvrtko Ursulin wrote:
From: Tvrtko Ursulin tvrtko.ursulin@intel.com
This is an alternative proposed fix for the below references bug report where dma fence error propagation is causing undesirable change in behaviour post GPU hang/reset.
Approach in this patch is to simply stop propagating all dma fence errors by default since that seems to be the upstream ask.
To handle the case where i915 needs error propagation for security, I add a new dma fence flag DMA_FENCE_FLAG_PROPAGATE_ERROR and make use of it in the command parsing chain only.
It sounds a plausible argument that fence propagation could be useful in which case a core flag to enable opt-in should be universally useful.
Signed-off-by: Tvrtko Ursulin tvrtko.ursulin@intel.com Reported-by: Marcin Slusarz marcin.slusarz@intel.com Reported-by: Miroslav Bendik References: 9e31c1fe45d5 ("drm/i915: Propagate errors on awaiting already signaled fences") References: https://gitlab.freedesktop.org/drm/intel/-/issues/3080 Cc: Jason Ekstrand jason.ekstrand@intel.com Cc: Daniel Vetter daniel.vetter@ffwll.ch
drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c | 2 ++ drivers/gpu/drm/i915/i915_sw_fence.c | 8 ++++---- drivers/gpu/drm/i915/i915_sw_fence.h | 8 ++++++++ include/linux/dma-fence.h | 1 +
I still don't like this, least because we still introduce the concept of error propagation to dma-fence (but hey only in i915 code, which is exactly the kind of not-really-upstream approach we got a major chiding for).
The only thing this does is make it explicitly opt-in instead opt-out, like the first fix. The right approach is imo still to just throw it out, and instead make the one error propagation we really need very, very explicit. Instead of hiding it behind lots of magic.
The one error propagation we need is when the cmd parser work fails, it must cancel it's corresponding request to make sure the batchbuffer doesn't run. This should require about 2 lines in total:
- one line to store the request so that the cmd parser work can access it. No refcounting needed, because the the request cannot even start (much less get freed) before the cmd parser has singalled its fence
- one line to kill the request if the parsing fails. Maybe 2 if you include the if condition. I have no idea how that's done since I'm honestly lost how the i915 scheduler decides whether to run a batch or not. I'm guessing we have a version of this for the ringbuffer and the execlist backend (if not maybe gen7 cmdparser is broken?)
I don't see any need for magic behind-the-scenes propagation of such a security critical error. Especially when that error propagation thing caused security bugs of its own, is an i915-only feature, and not motivated by any userspace/uapi requirements at all.
Thanks, Daniel
4 files changed, 15 insertions(+), 4 deletions(-)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c index 297143511f99..6a516d1261d0 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c @@ -2522,6 +2522,8 @@ static int eb_parse_pipeline(struct i915_execbuffer *eb, }
dma_fence_work_init(&pw->base, &eb_parse_ops);
/* Propagate errors for security. */
__set_bit(DMA_FENCE_FLAG_PROPAGATE_ERROR, &pw->base.dma.flags);
pw->engine = eb->engine; pw->batch = eb->batch->vma;
diff --git a/drivers/gpu/drm/i915/i915_sw_fence.c b/drivers/gpu/drm/i915/i915_sw_fence.c index 2744558f3050..2ee917932ccf 100644 --- a/drivers/gpu/drm/i915/i915_sw_fence.c +++ b/drivers/gpu/drm/i915/i915_sw_fence.c @@ -449,7 +449,7 @@ static void dma_i915_sw_fence_wake_timer(struct dma_fence *dma,
fence = xchg(&cb->base.fence, NULL); if (fence) {
i915_sw_fence_set_error_once(fence, dma->error);
i915_sw_fence_propagate_dma_fence_error(fence, dma);@@ -480,7 +480,7 @@ int i915_sw_fence_await_dma_fence(struct i915_sw_fence *fence, might_sleep_if(gfpflags_allow_blocking(gfp));
if (dma_fence_is_signaled(dma)) {
i915_sw_fence_set_error_once(fence, dma->error);
i915_sw_fence_propagate_dma_fence_error(fence, dma);@@ -496,7 +496,7 @@ int i915_sw_fence_await_dma_fence(struct i915_sw_fence *fence, if (ret) return ret;
i915_sw_fence_set_error_once(fence, dma->error);
i915_sw_fence_propagate_dma_fence_error(fence, dma);@@ -548,7 +548,7 @@ int __i915_sw_fence_await_dma_fence(struct i915_sw_fence *fence, debug_fence_assert(fence);
if (dma_fence_is_signaled(dma)) {
i915_sw_fence_set_error_once(fence, dma->error);
i915_sw_fence_propagate_dma_fence_error(fence, dma);diff --git a/drivers/gpu/drm/i915/i915_sw_fence.h b/drivers/gpu/drm/i915/i915_sw_fence.h index 30a863353ee6..872ef80ebd10 100644 --- a/drivers/gpu/drm/i915/i915_sw_fence.h +++ b/drivers/gpu/drm/i915/i915_sw_fence.h @@ -116,4 +116,12 @@ i915_sw_fence_set_error_once(struct i915_sw_fence *fence, int error) cmpxchg(&fence->error, 0, error); }
+static inline void +i915_sw_fence_propagate_dma_fence_error(struct i915_sw_fence *fence,
struct dma_fence *dma)+{
- if (unlikely(test_bit(DMA_FENCE_FLAG_PROPAGATE_ERROR, &dma->flags)))
i915_sw_fence_set_error_once(fence, dma->error);+}
#endif /* _I915_SW_FENCE_H_ */ diff --git a/include/linux/dma-fence.h b/include/linux/dma-fence.h index 6ffb4b2c6371..8dabe1650f11 100644 --- a/include/linux/dma-fence.h +++ b/include/linux/dma-fence.h @@ -99,6 +99,7 @@ enum dma_fence_flag_bits { DMA_FENCE_FLAG_SIGNALED_BIT, DMA_FENCE_FLAG_TIMESTAMP_BIT, DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT,
- DMA_FENCE_FLAG_PROPAGATE_ERROR, DMA_FENCE_FLAG_USER_BITS, /* must always be last member */
};
-- 2.30.2
On 10/05/2021 16:55, Daniel Vetter wrote:
On Fri, May 07, 2021 at 09:35:21AM +0100, Tvrtko Ursulin wrote:
From: Tvrtko Ursulin tvrtko.ursulin@intel.com
This is an alternative proposed fix for the below references bug report where dma fence error propagation is causing undesirable change in behaviour post GPU hang/reset.
Approach in this patch is to simply stop propagating all dma fence errors by default since that seems to be the upstream ask.
To handle the case where i915 needs error propagation for security, I add a new dma fence flag DMA_FENCE_FLAG_PROPAGATE_ERROR and make use of it in the command parsing chain only.
It sounds a plausible argument that fence propagation could be useful in which case a core flag to enable opt-in should be universally useful.
Signed-off-by: Tvrtko Ursulin tvrtko.ursulin@intel.com Reported-by: Marcin Slusarz marcin.slusarz@intel.com Reported-by: Miroslav Bendik References: 9e31c1fe45d5 ("drm/i915: Propagate errors on awaiting already signaled fences") References: https://gitlab.freedesktop.org/drm/intel/-/issues/3080 Cc: Jason Ekstrand jason.ekstrand@intel.com Cc: Daniel Vetter daniel.vetter@ffwll.ch
drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c | 2 ++ drivers/gpu/drm/i915/i915_sw_fence.c | 8 ++++---- drivers/gpu/drm/i915/i915_sw_fence.h | 8 ++++++++ include/linux/dma-fence.h | 1 +
I still don't like this, least because we still introduce the concept of error propagation to dma-fence (but hey only in i915 code, which is exactly the kind of not-really-upstream approach we got a major chiding for).
The only thing this does is make it explicitly opt-in instead opt-out, like the first fix. The right approach is imo still to just throw it out, and instead make the one error propagation we really need very, very explicit. Instead of hiding it behind lots of magic.
The one error propagation we need is when the cmd parser work fails, it must cancel it's corresponding request to make sure the batchbuffer doesn't run. This should require about 2 lines in total:
one line to store the request so that the cmd parser work can access it. No refcounting needed, because the the request cannot even start (much less get freed) before the cmd parser has singalled its fence
one line to kill the request if the parsing fails. Maybe 2 if you include the if condition. I have no idea how that's done since I'm honestly lost how the i915 scheduler decides whether to run a batch or not. I'm guessing we have a version of this for the ringbuffer and the execlist backend (if not maybe gen7 cmdparser is broken?)
I don't see any need for magic behind-the-scenes propagation of such a security critical error. Especially when that error propagation thing caused security bugs of its own, is an i915-only feature, and not motivated by any userspace/uapi requirements at all.
I took this approach because to me propagating errors sounds more logical than ignoring them and I was arguing in the commit message that the infrastructure to enable that could be put in place as opt-in.
I also do not see a lot of magic in this patch. Only thing, potentially the logic should be inverted so that the waiter marks itself as interested in receiving errors. That would probably make even more sense as a core concept.
Has there been a wider discussion on this topic in the past? I am curious to know, even if propagation currently is i915 only, could other drivers be interested.
Note that it adds almost nothing to the dma-buf common code about a single flag, and at some point (currently missing) documentation on the very flag.
Regards,
Tvrtko
On Tue, May 11, 2021 at 10:05:27AM +0100, Tvrtko Ursulin wrote:
On 10/05/2021 16:55, Daniel Vetter wrote:
On Fri, May 07, 2021 at 09:35:21AM +0100, Tvrtko Ursulin wrote:
From: Tvrtko Ursulin tvrtko.ursulin@intel.com
This is an alternative proposed fix for the below references bug report where dma fence error propagation is causing undesirable change in behaviour post GPU hang/reset.
Approach in this patch is to simply stop propagating all dma fence errors by default since that seems to be the upstream ask.
To handle the case where i915 needs error propagation for security, I add a new dma fence flag DMA_FENCE_FLAG_PROPAGATE_ERROR and make use of it in the command parsing chain only.
It sounds a plausible argument that fence propagation could be useful in which case a core flag to enable opt-in should be universally useful.
Signed-off-by: Tvrtko Ursulin tvrtko.ursulin@intel.com Reported-by: Marcin Slusarz marcin.slusarz@intel.com Reported-by: Miroslav Bendik References: 9e31c1fe45d5 ("drm/i915: Propagate errors on awaiting already signaled fences") References: https://gitlab.freedesktop.org/drm/intel/-/issues/3080 Cc: Jason Ekstrand jason.ekstrand@intel.com Cc: Daniel Vetter daniel.vetter@ffwll.ch
drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c | 2 ++ drivers/gpu/drm/i915/i915_sw_fence.c | 8 ++++---- drivers/gpu/drm/i915/i915_sw_fence.h | 8 ++++++++ include/linux/dma-fence.h | 1 +
I still don't like this, least because we still introduce the concept of error propagation to dma-fence (but hey only in i915 code, which is exactly the kind of not-really-upstream approach we got a major chiding for).
The only thing this does is make it explicitly opt-in instead opt-out, like the first fix. The right approach is imo still to just throw it out, and instead make the one error propagation we really need very, very explicit. Instead of hiding it behind lots of magic.
The one error propagation we need is when the cmd parser work fails, it must cancel it's corresponding request to make sure the batchbuffer doesn't run. This should require about 2 lines in total:
one line to store the request so that the cmd parser work can access it. No refcounting needed, because the the request cannot even start (much less get freed) before the cmd parser has singalled its fence
one line to kill the request if the parsing fails. Maybe 2 if you include the if condition. I have no idea how that's done since I'm honestly lost how the i915 scheduler decides whether to run a batch or not. I'm guessing we have a version of this for the ringbuffer and the execlist backend (if not maybe gen7 cmdparser is broken?)
I don't see any need for magic behind-the-scenes propagation of such a security critical error. Especially when that error propagation thing caused security bugs of its own, is an i915-only feature, and not motivated by any userspace/uapi requirements at all.
I took this approach because to me propagating errors sounds more logical than ignoring them and I was arguing in the commit message that the infrastructure to enable that could be put in place as opt-in.
I also do not see a lot of magic in this patch. Only thing, potentially the logic should be inverted so that the waiter marks itself as interested in receiving errors. That would probably make even more sense as a core concept.
Has there been a wider discussion on this topic in the past? I am curious to know, even if propagation currently is i915 only, could other drivers be interested.
There hasn't been. i915-gem team decided "this is a cool concept", which resulted in a security bug. Now we're a few months in arguing whether a cool-looking concept that leads to a security bug is maybe a good idea, and whether we should sneak it in as a core concept to dma-buf.h without any wider discussion on the concept.
Note that it adds almost nothing to the dma-buf common code about a single flag, and at some point (currently missing) documentation on the very flag.
This is really not how upstream collaboration works, and it needs to stop.
If you want this, start another thread arguing why this is a good idea, fully decoupled from the security fix here. -Daniel
On 17/05/2021 16:12, Daniel Vetter wrote:
On Tue, May 11, 2021 at 10:05:27AM +0100, Tvrtko Ursulin wrote:
On 10/05/2021 16:55, Daniel Vetter wrote:
On Fri, May 07, 2021 at 09:35:21AM +0100, Tvrtko Ursulin wrote:
From: Tvrtko Ursulin tvrtko.ursulin@intel.com
This is an alternative proposed fix for the below references bug report where dma fence error propagation is causing undesirable change in behaviour post GPU hang/reset.
Approach in this patch is to simply stop propagating all dma fence errors by default since that seems to be the upstream ask.
To handle the case where i915 needs error propagation for security, I add a new dma fence flag DMA_FENCE_FLAG_PROPAGATE_ERROR and make use of it in the command parsing chain only.
It sounds a plausible argument that fence propagation could be useful in which case a core flag to enable opt-in should be universally useful.
Signed-off-by: Tvrtko Ursulin tvrtko.ursulin@intel.com Reported-by: Marcin Slusarz marcin.slusarz@intel.com Reported-by: Miroslav Bendik References: 9e31c1fe45d5 ("drm/i915: Propagate errors on awaiting already signaled fences") References: https://gitlab.freedesktop.org/drm/intel/-/issues/3080 Cc: Jason Ekstrand jason.ekstrand@intel.com Cc: Daniel Vetter daniel.vetter@ffwll.ch
drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c | 2 ++ drivers/gpu/drm/i915/i915_sw_fence.c | 8 ++++---- drivers/gpu/drm/i915/i915_sw_fence.h | 8 ++++++++ include/linux/dma-fence.h | 1 +
I still don't like this, least because we still introduce the concept of error propagation to dma-fence (but hey only in i915 code, which is exactly the kind of not-really-upstream approach we got a major chiding for).
The only thing this does is make it explicitly opt-in instead opt-out, like the first fix. The right approach is imo still to just throw it out, and instead make the one error propagation we really need very, very explicit. Instead of hiding it behind lots of magic.
The one error propagation we need is when the cmd parser work fails, it must cancel it's corresponding request to make sure the batchbuffer doesn't run. This should require about 2 lines in total:
one line to store the request so that the cmd parser work can access it. No refcounting needed, because the the request cannot even start (much less get freed) before the cmd parser has singalled its fence
one line to kill the request if the parsing fails. Maybe 2 if you include the if condition. I have no idea how that's done since I'm honestly lost how the i915 scheduler decides whether to run a batch or not. I'm guessing we have a version of this for the ringbuffer and the execlist backend (if not maybe gen7 cmdparser is broken?)
I don't see any need for magic behind-the-scenes propagation of such a security critical error. Especially when that error propagation thing caused security bugs of its own, is an i915-only feature, and not motivated by any userspace/uapi requirements at all.
I took this approach because to me propagating errors sounds more logical than ignoring them and I was arguing in the commit message that the infrastructure to enable that could be put in place as opt-in.
I also do not see a lot of magic in this patch. Only thing, potentially the logic should be inverted so that the waiter marks itself as interested in receiving errors. That would probably make even more sense as a core concept.
Has there been a wider discussion on this topic in the past? I am curious to know, even if propagation currently is i915 only, could other drivers be interested.
There hasn't been. i915-gem team decided "this is a cool concept", which resulted in a security bug. Now we're a few months in arguing whether a cool-looking concept that leads to a security bug is maybe a good idea, and whether we should sneak it in as a core concept to dma-buf.h without any wider discussion on the concept.
Note that it adds almost nothing to the dma-buf common code about a single flag, and at some point (currently missing) documentation on the very flag.
This is really not how upstream collaboration works, and it needs to stop.
If you want this, start another thread arguing why this is a good idea, fully decoupled from the security fix here.
When I asked you whether you know there were past discussions on this topic, clearly the point of that was to figure out whether a new discussion needs to be started, or I need to go and read an existing one to get up to speed.
I don't know how you interpreted that as an attempt to sneak anything in. And I don't know how I could have reliably figured out the answer to that question without asking. So colour me confused.
To clarify on the security issue part - are you talking about https://gitlab.freedesktop.org/drm/intel/-/issues/3080, or the other security issue, the one which would be caused by simply reverting the error propagation in i915?
Regards,
Tvrtko
On Mon, May 17, 2021 at 5:33 PM Tvrtko Ursulin tvrtko.ursulin@linux.intel.com wrote:
On 17/05/2021 16:12, Daniel Vetter wrote:
On Tue, May 11, 2021 at 10:05:27AM +0100, Tvrtko Ursulin wrote:
On 10/05/2021 16:55, Daniel Vetter wrote:
On Fri, May 07, 2021 at 09:35:21AM +0100, Tvrtko Ursulin wrote:
From: Tvrtko Ursulin tvrtko.ursulin@intel.com
This is an alternative proposed fix for the below references bug report where dma fence error propagation is causing undesirable change in behaviour post GPU hang/reset.
Approach in this patch is to simply stop propagating all dma fence errors by default since that seems to be the upstream ask.
To handle the case where i915 needs error propagation for security, I add a new dma fence flag DMA_FENCE_FLAG_PROPAGATE_ERROR and make use of it in the command parsing chain only.
It sounds a plausible argument that fence propagation could be useful in which case a core flag to enable opt-in should be universally useful.
Signed-off-by: Tvrtko Ursulin tvrtko.ursulin@intel.com Reported-by: Marcin Slusarz marcin.slusarz@intel.com Reported-by: Miroslav Bendik References: 9e31c1fe45d5 ("drm/i915: Propagate errors on awaiting already signaled fences") References: https://gitlab.freedesktop.org/drm/intel/-/issues/3080 Cc: Jason Ekstrand jason.ekstrand@intel.com Cc: Daniel Vetter daniel.vetter@ffwll.ch
drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c | 2 ++ drivers/gpu/drm/i915/i915_sw_fence.c | 8 ++++---- drivers/gpu/drm/i915/i915_sw_fence.h | 8 ++++++++ include/linux/dma-fence.h | 1 +
I still don't like this, least because we still introduce the concept of error propagation to dma-fence (but hey only in i915 code, which is exactly the kind of not-really-upstream approach we got a major chiding for).
The only thing this does is make it explicitly opt-in instead opt-out, like the first fix. The right approach is imo still to just throw it out, and instead make the one error propagation we really need very, very explicit. Instead of hiding it behind lots of magic.
The one error propagation we need is when the cmd parser work fails, it must cancel it's corresponding request to make sure the batchbuffer doesn't run. This should require about 2 lines in total:
one line to store the request so that the cmd parser work can access it. No refcounting needed, because the the request cannot even start (much less get freed) before the cmd parser has singalled its fence
one line to kill the request if the parsing fails. Maybe 2 if you include the if condition. I have no idea how that's done since I'm honestly lost how the i915 scheduler decides whether to run a batch or not. I'm guessing we have a version of this for the ringbuffer and the execlist backend (if not maybe gen7 cmdparser is broken?)
I don't see any need for magic behind-the-scenes propagation of such a security critical error. Especially when that error propagation thing caused security bugs of its own, is an i915-only feature, and not motivated by any userspace/uapi requirements at all.
I took this approach because to me propagating errors sounds more logical than ignoring them and I was arguing in the commit message that the infrastructure to enable that could be put in place as opt-in.
I also do not see a lot of magic in this patch. Only thing, potentially the logic should be inverted so that the waiter marks itself as interested in receiving errors. That would probably make even more sense as a core concept.
Has there been a wider discussion on this topic in the past? I am curious to know, even if propagation currently is i915 only, could other drivers be interested.
There hasn't been. i915-gem team decided "this is a cool concept", which resulted in a security bug. Now we're a few months in arguing whether a cool-looking concept that leads to a security bug is maybe a good idea, and whether we should sneak it in as a core concept to dma-buf.h without any wider discussion on the concept.
Note that it adds almost nothing to the dma-buf common code about a single flag, and at some point (currently missing) documentation on the very flag.
This is really not how upstream collaboration works, and it needs to stop.
If you want this, start another thread arguing why this is a good idea, fully decoupled from the security fix here.
When I asked you whether you know there were past discussions on this topic, clearly the point of that was to figure out whether a new discussion needs to be started, or I need to go and read an existing one to get up to speed.
I don't know how you interpreted that as an attempt to sneak anything in. And I don't know how I could have reliably figured out the answer to that question without asking. So colour me confused.
To clarify on the security issue part - are you talking about https://gitlab.freedesktop.org/drm/intel/-/issues/3080, or the other security issue, the one which would be caused by simply reverting the error propagation in i915?
Both.
But what I really don't get is why we keep defending a "it's cool" type of concept that accidentally caused a security bug as a consequence that neither author nor reviewer managed to forsee.
If it accidentally causes a security bug and is otherwise not justified by anything, it's not cool. There isn't anything left to argue about here. -Daniel
dri-devel@lists.freedesktop.org
-
Daniel Vetter -
Jason Ekstrand -
Tvrtko Ursulin