| Message ID | 20210706101209.3034092-6-daniel.vetter@ffwll.ch (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | dma-resv fence DAG fixes | expand |
On Tue, 6 Jul 2021 at 11:12, Daniel Vetter <daniel.vetter@ffwll.ch> wrote: > > We're lifting, or well, clarifying that the restriction that shared > fences have to be strictly after the exclusive one doesn't apply > anymore. > > So adjust the code to always also wait for the exclusive fence. > > Signed-off-by: Daniel Vetter <daniel.vetter@intel.com> > Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> > Cc: "Thomas Hellström" <thomas.hellstrom@linux.intel.com> > Cc: Jason Ekstrand <jason@jlekstrand.net> > --- > drivers/gpu/drm/i915/i915_sw_fence.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/drivers/gpu/drm/i915/i915_sw_fence.c b/drivers/gpu/drm/i915/i915_sw_fence.c > index 91711a46b1c7..271d321cea83 100644 > --- a/drivers/gpu/drm/i915/i915_sw_fence.c > +++ b/drivers/gpu/drm/i915/i915_sw_fence.c > @@ -601,10 +601,10 @@ int i915_sw_fence_await_reservation(struct i915_sw_fence *fence, > for (i = 0; i < count; i++) > dma_fence_put(shared[i]); > kfree(shared); > - } else { > - excl = dma_resv_get_excl_unlocked(resv); > } > > + excl = dma_resv_get_excl_unlocked(resv); > + The dma_resv_get_fences() call looks like it already fishes out the exclusive fence. Does this not leak the extra ref now? > if (ret >= 0 && excl) { > pending = i915_sw_fence_await_dma_fence(fence, > excl, > -- > 2.32.0 > > _______________________________________________ > Intel-gfx mailing list > Intel-gfx@lists.freedesktop.org > https://lists.freedesktop.org/mailman/listinfo/intel-gfx
On Tue, Jul 6, 2021 at 2:47 PM Matthew Auld <matthew.william.auld@gmail.com> wrote: > > On Tue, 6 Jul 2021 at 11:12, Daniel Vetter <daniel.vetter@ffwll.ch> wrote: > > > > We're lifting, or well, clarifying that the restriction that shared > > fences have to be strictly after the exclusive one doesn't apply > > anymore. > > > > So adjust the code to always also wait for the exclusive fence. > > > > Signed-off-by: Daniel Vetter <daniel.vetter@intel.com> > > Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> > > Cc: "Thomas Hellström" <thomas.hellstrom@linux.intel.com> > > Cc: Jason Ekstrand <jason@jlekstrand.net> > > --- > > drivers/gpu/drm/i915/i915_sw_fence.c | 4 ++-- > > 1 file changed, 2 insertions(+), 2 deletions(-) > > > > diff --git a/drivers/gpu/drm/i915/i915_sw_fence.c b/drivers/gpu/drm/i915/i915_sw_fence.c > > index 91711a46b1c7..271d321cea83 100644 > > --- a/drivers/gpu/drm/i915/i915_sw_fence.c > > +++ b/drivers/gpu/drm/i915/i915_sw_fence.c > > @@ -601,10 +601,10 @@ int i915_sw_fence_await_reservation(struct i915_sw_fence *fence, > > for (i = 0; i < count; i++) > > dma_fence_put(shared[i]); > > kfree(shared); > > - } else { > > - excl = dma_resv_get_excl_unlocked(resv); > > } > > > > + excl = dma_resv_get_excl_unlocked(resv); > > + > > The dma_resv_get_fences() call looks like it already fishes out the > exclusive fence. Does this not leak the extra ref now? Oh right I overlooked this, we already pick up the exclusive fence unconditionally. Control flow here was a bit too clever for my parser. I'll drop this patch in the next round. -Daniel > > > if (ret >= 0 && excl) { > > pending = i915_sw_fence_await_dma_fence(fence, > > excl, > > -- > > 2.32.0 > > > > _______________________________________________ > > Intel-gfx mailing list > > Intel-gfx@lists.freedesktop.org > > https://lists.freedesktop.org/mailman/listinfo/intel-gfx
diff --git a/drivers/gpu/drm/i915/i915_sw_fence.c b/drivers/gpu/drm/i915/i915_sw_fence.c index 91711a46b1c7..271d321cea83 100644 --- a/drivers/gpu/drm/i915/i915_sw_fence.c +++ b/drivers/gpu/drm/i915/i915_sw_fence.c @@ -601,10 +601,10 @@ int i915_sw_fence_await_reservation(struct i915_sw_fence *fence, for (i = 0; i < count; i++) dma_fence_put(shared[i]); kfree(shared); - } else { - excl = dma_resv_get_excl_unlocked(resv); } + excl = dma_resv_get_excl_unlocked(resv); + if (ret >= 0 && excl) { pending = i915_sw_fence_await_dma_fence(fence, excl,
We're lifting, or well, clarifying that the restriction that shared fences have to be strictly after the exclusive one doesn't apply anymore. So adjust the code to always also wait for the exclusive fence. Signed-off-by: Daniel Vetter <daniel.vetter@intel.com> Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Cc: "Thomas Hellström" <thomas.hellstrom@linux.intel.com> Cc: Jason Ekstrand <jason@jlekstrand.net> --- drivers/gpu/drm/i915/i915_sw_fence.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)