| Message ID | 20180926071703.15257-1-jason.ekstrand@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | drm/syncobj: Don't leak fences when WAIT_FOR_SUBMIT is set | expand |
Quoting Jason Ekstrand (2018-09-26 08:17:03) > We attempt to get fences earlier in the hopes that everything will > already have fences and no callbacks will be needed. If we do succeed > in getting a fence, getting one a second time will result in a duplicate > ref with no unref. This is causing memory leaks in Vulkan applications > that create a lot of fences; playing for a few hours can, apparently, > bring down the system. > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107899 > Signed-off-by: Jason Ekstrand <jason@jlekstrand.net> > Cc: stable@vger.kernel.org > --- > drivers/gpu/drm/drm_syncobj.c | 5 +++++ > 1 file changed, 5 insertions(+) > > diff --git a/drivers/gpu/drm/drm_syncobj.c b/drivers/gpu/drm/drm_syncobj.c > index adb3cb27d31e..759278fef35a 100644 > --- a/drivers/gpu/drm/drm_syncobj.c > +++ b/drivers/gpu/drm/drm_syncobj.c > @@ -97,6 +97,8 @@ static int drm_syncobj_fence_get_or_add_callback(struct drm_syncobj *syncobj, > { > int ret; > > + WARN_ON(*fence); I would have just put if (*fence) return; since the function is tied to the array_wait implementation. Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> -Chris
On Wed, Sep 26, 2018 at 3:18 AM Chris Wilson <chris@chris-wilson.co.uk> wrote: > Quoting Jason Ekstrand (2018-09-26 08:17:03) > > We attempt to get fences earlier in the hopes that everything will > > already have fences and no callbacks will be needed. If we do succeed > > in getting a fence, getting one a second time will result in a duplicate > > ref with no unref. This is causing memory leaks in Vulkan applications > > that create a lot of fences; playing for a few hours can, apparently, > > bring down the system. > > > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107899 > > Signed-off-by: Jason Ekstrand <jason@jlekstrand.net> > > Cc: stable@vger.kernel.org > > --- > > drivers/gpu/drm/drm_syncobj.c | 5 +++++ > > 1 file changed, 5 insertions(+) > > > > diff --git a/drivers/gpu/drm/drm_syncobj.c > b/drivers/gpu/drm/drm_syncobj.c > > index adb3cb27d31e..759278fef35a 100644 > > --- a/drivers/gpu/drm/drm_syncobj.c > > +++ b/drivers/gpu/drm/drm_syncobj.c > > @@ -97,6 +97,8 @@ static int > drm_syncobj_fence_get_or_add_callback(struct drm_syncobj *syncobj, > > { > > int ret; > > > > + WARN_ON(*fence); > > I would have just put if (*fence) return; since the function is tied to > the array_wait implementation. > I considered doing that but marginally liked this better. If you have a preference, I'm happy to change itl. --Jason > Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> > -Chris > <div dir="ltr"><div class="gmail_quote"><div dir="ltr">On Wed, Sep 26, 2018 at 3:18 AM Chris Wilson <<a href="mailto:chris@chris-wilson.co.uk">chris@chris-wilson.co.uk</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Quoting Jason Ekstrand (2018-09-26 08:17:03)<br> > We attempt to get fences earlier in the hopes that everything will<br> > already have fences and no callbacks will be needed. If we do succeed<br> > in getting a fence, getting one a second time will result in a duplicate<br> > ref with no unref. This is causing memory leaks in Vulkan applications<br> > that create a lot of fences; playing for a few hours can, apparently,<br> > bring down the system.<br> > <br> > Bugzilla: <a href="https://bugs.freedesktop.org/show_bug.cgi?id=107899" rel="noreferrer" target="_blank">https://bugs.freedesktop.org/show_bug.cgi?id=107899</a><br> > Signed-off-by: Jason Ekstrand <<a href="mailto:jason@jlekstrand.net" target="_blank">jason@jlekstrand.net</a>><br> > Cc: <a href="mailto:stable@vger.kernel.org" target="_blank">stable@vger.kernel.org</a><br> > ---<br> > drivers/gpu/drm/drm_syncobj.c | 5 +++++<br> > 1 file changed, 5 insertions(+)<br> > <br> > diff --git a/drivers/gpu/drm/drm_syncobj.c b/drivers/gpu/drm/drm_syncobj.c<br> > index adb3cb27d31e..759278fef35a 100644<br> > --- a/drivers/gpu/drm/drm_syncobj.c<br> > +++ b/drivers/gpu/drm/drm_syncobj.c<br> > @@ -97,6 +97,8 @@ static int drm_syncobj_fence_get_or_add_callback(struct drm_syncobj *syncobj,<br> > {<br> > int ret;<br> > <br> > + WARN_ON(*fence);<br> <br> I would have just put if (*fence) return; since the function is tied to<br> the array_wait implementation.<br></blockquote><div><br></div><div>I considered doing that but marginally liked this better. If you have a preference, I'm happy to change itl.</div><div><br></div><div>--Jason<br></div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"> Reviewed-by: Chris Wilson <<a href="mailto:chris@chris-wilson.co.uk" target="_blank">chris@chris-wilson.co.uk</a>><br> -Chris<br> </blockquote></div></div>
Quoting Jason Ekstrand (2018-09-26 10:43:59) > On Wed, Sep 26, 2018 at 3:18 AM Chris Wilson <chris@chris-wilson.co.uk> wrote: > > Quoting Jason Ekstrand (2018-09-26 08:17:03) > > We attempt to get fences earlier in the hopes that everything will > > already have fences and no callbacks will be needed. If we do succeed > > in getting a fence, getting one a second time will result in a duplicate > > ref with no unref. This is causing memory leaks in Vulkan applications > > that create a lot of fences; playing for a few hours can, apparently, > > bring down the system. > > > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107899 > > Signed-off-by: Jason Ekstrand <jason@jlekstrand.net> > > Cc: stable@vger.kernel.org > > --- > > drivers/gpu/drm/drm_syncobj.c | 5 +++++ > > 1 file changed, 5 insertions(+) > > > > diff --git a/drivers/gpu/drm/drm_syncobj.c b/drivers/gpu/drm/ > drm_syncobj.c > > index adb3cb27d31e..759278fef35a 100644 > > --- a/drivers/gpu/drm/drm_syncobj.c > > +++ b/drivers/gpu/drm/drm_syncobj.c > > @@ -97,6 +97,8 @@ static int drm_syncobj_fence_get_or_add_callback(struct > drm_syncobj *syncobj, > > { > > int ret; > > > > + WARN_ON(*fence); > > I would have just put if (*fence) return; since the function is tied to > the array_wait implementation. > > > I considered doing that but marginally liked this better. If you have a > preference, I'm happy to change itl. Patch works, the difference is insignificant, land the patch and close the bug. -Chris
On Wed, Sep 26, 2018 at 02:17:03AM -0500, Jason Ekstrand wrote: > We attempt to get fences earlier in the hopes that everything will > already have fences and no callbacks will be needed. If we do succeed > in getting a fence, getting one a second time will result in a duplicate > ref with no unref. This is causing memory leaks in Vulkan applications > that create a lot of fences; playing for a few hours can, apparently, > bring down the system. > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107899 > Signed-off-by: Jason Ekstrand <jason@jlekstrand.net> Pushed to drm-misc-fixes, thanks! Sean > Cc: stable@vger.kernel.org > --- > drivers/gpu/drm/drm_syncobj.c | 5 +++++ > 1 file changed, 5 insertions(+) > > diff --git a/drivers/gpu/drm/drm_syncobj.c b/drivers/gpu/drm/drm_syncobj.c > index adb3cb27d31e..759278fef35a 100644 > --- a/drivers/gpu/drm/drm_syncobj.c > +++ b/drivers/gpu/drm/drm_syncobj.c > @@ -97,6 +97,8 @@ static int drm_syncobj_fence_get_or_add_callback(struct drm_syncobj *syncobj, > { > int ret; > > + WARN_ON(*fence); > + > *fence = drm_syncobj_fence_get(syncobj); > if (*fence) > return 1; > @@ -743,6 +745,9 @@ static signed long drm_syncobj_array_wait_timeout(struct drm_syncobj **syncobjs, > > if (flags & DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT) { > for (i = 0; i < count; ++i) { > + if (entries[i].fence) > + continue; > + > drm_syncobj_fence_get_or_add_callback(syncobjs[i], > &entries[i].fence, > &entries[i].syncobj_cb, > -- > 2.17.1 > > _______________________________________________ > dri-devel mailing list > dri-devel@lists.freedesktop.org > https://lists.freedesktop.org/mailman/listinfo/dri-devel
diff --git a/drivers/gpu/drm/drm_syncobj.c b/drivers/gpu/drm/drm_syncobj.c index adb3cb27d31e..759278fef35a 100644 --- a/drivers/gpu/drm/drm_syncobj.c +++ b/drivers/gpu/drm/drm_syncobj.c @@ -97,6 +97,8 @@ static int drm_syncobj_fence_get_or_add_callback(struct drm_syncobj *syncobj, { int ret; + WARN_ON(*fence); + *fence = drm_syncobj_fence_get(syncobj); if (*fence) return 1; @@ -743,6 +745,9 @@ static signed long drm_syncobj_array_wait_timeout(struct drm_syncobj **syncobjs, if (flags & DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT) { for (i = 0; i < count; ++i) { + if (entries[i].fence) + continue; + drm_syncobj_fence_get_or_add_callback(syncobjs[i], &entries[i].fence, &entries[i].syncobj_cb,
We attempt to get fences earlier in the hopes that everything will already have fences and no callbacks will be needed. If we do succeed in getting a fence, getting one a second time will result in a duplicate ref with no unref. This is causing memory leaks in Vulkan applications that create a lot of fences; playing for a few hours can, apparently, bring down the system. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107899 Signed-off-by: Jason Ekstrand <jason@jlekstrand.net> Cc: stable@vger.kernel.org --- drivers/gpu/drm/drm_syncobj.c | 5 +++++ 1 file changed, 5 insertions(+)