| Message ID | 20180814001607.27026-1-peter@lekensteyn.nl (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [xf86-video-intel] SNA: fix PRIME output support since xserver 1.20 | expand |
Ping. This patch was independently verified working (see the linked bug report) and is essential for Xorg 1.20 using this driver. On Tue, Aug 14, 2018 at 02:16:07AM +0200, Peter Wu wrote: > Since xorg-server 1.20, an external monitor would remain blank when used > in a PRIME output slave setup. Only a cursor was visible. The cause is > "Make PixmapDirtyUpdateRec::src a DrawablePtr" in xserver, the "src" > pointer might point to the root window (created by the server) instead > of a pixmap (as created by xf86-video-intel). Use get_drawable_pixmap to > handle both cases. > > When built with -fsanitize=address, the following test will trigger a > heap-buffer-overflow error due to to_sna_from_pixmap receiving a window > instead of a pixmap. > > Test on a hybrid graphics laptop (Intel + modesetting/nouveau): > > xrandr --setprovideroutputsource modesetting Intel > xrandr --output DP-1-1 --mode 2560x1440 # should not crash > glxgears # should display gears on both screens > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=100086 > Signed-off-by: Peter Wu <peter@lekensteyn.nl> > --- > Tested with xserver 1.20.1 with ASAN enabled. Survives multiple > resolution changes, works with a Plasma desktop session, it seems > stable. Something like this patch is required to make multi-monitor > setups usable in a hybrid graphics setting with Xorg 1.20. > --- > src/sna/sna_accel.c | 18 ++++++++++++++++++ > 1 file changed, 18 insertions(+) > > diff --git a/src/sna/sna_accel.c b/src/sna/sna_accel.c > index 2f669bcf..80b116a3 100644 > --- a/src/sna/sna_accel.c > +++ b/src/sna/sna_accel.c > @@ -17510,7 +17510,11 @@ static bool has_offload_slaves(struct sna *sna) > PixmapDirtyUpdatePtr dirty; > > xorg_list_for_each_entry(dirty, &screen->pixmap_dirty_list, ent) { > +#ifdef HAS_DIRTYTRACKING_DRAWABLE_SRC > + assert(dirty->src == &sna->front->drawable); > +#else > assert(dirty->src == sna->front); > +#endif > if (RegionNotEmpty(DamageRegion(dirty->damage))) > return true; > } > @@ -17671,7 +17675,11 @@ static void sna_accel_post_damage(struct sna *sna) > if (RegionNil(damage)) > continue; > > +#ifdef HAS_DIRTYTRACKING_DRAWABLE_SRC > + src = get_drawable_pixmap(dirty->src); > +#else > src = dirty->src; > +#endif > dst = dirty->slave_dst->master_pixmap; > > region.extents.x1 = dirty->x; > @@ -17922,9 +17930,15 @@ migrate_dirty_tracking(PixmapPtr old_front, PixmapPtr new_front) > PixmapDirtyUpdatePtr dirty, safe; > > xorg_list_for_each_entry_safe(dirty, safe, &screen->pixmap_dirty_list, ent) { > +#ifdef HAS_DIRTYTRACKING_DRAWABLE_SRC > + assert(dirty->src == &old_front->drawable); > + if (dirty->src != &old_front->drawable) > + continue; > +#else > assert(dirty->src == old_front); > if (dirty->src != old_front) > continue; > +#endif > > DamageUnregister(&dirty->src->drawable, dirty->damage); > DamageDestroy(dirty->damage); > @@ -17939,7 +17953,11 @@ migrate_dirty_tracking(PixmapPtr old_front, PixmapPtr new_front) > } > > DamageRegister(&new_front->drawable, dirty->damage); > +#ifdef HAS_DIRTYTRACKING_DRAWABLE_SRC > + dirty->src = &new_front->drawable; > +#else > dirty->src = new_front; > +#endif > } > #endif > } > -- > 2.18.0
In meantime this patch has been picked up by the Arch Linux packages: https://bugs.archlinux.org/task/58895 Is there any chance that this patch can be reviewed/applied? The diff is quite small and should be easy to review. On Wed, Aug 29, 2018 at 12:22:21PM +0200, Peter Wu wrote: > Ping. > > This patch was independently verified working (see the linked bug > report) and is essential for Xorg 1.20 using this driver. > > On Tue, Aug 14, 2018 at 02:16:07AM +0200, Peter Wu wrote: > > Since xorg-server 1.20, an external monitor would remain blank when used > > in a PRIME output slave setup. Only a cursor was visible. The cause is > > "Make PixmapDirtyUpdateRec::src a DrawablePtr" in xserver, the "src" > > pointer might point to the root window (created by the server) instead > > of a pixmap (as created by xf86-video-intel). Use get_drawable_pixmap to > > handle both cases. > > > > When built with -fsanitize=address, the following test will trigger a > > heap-buffer-overflow error due to to_sna_from_pixmap receiving a window > > instead of a pixmap. > > > > Test on a hybrid graphics laptop (Intel + modesetting/nouveau): > > > > xrandr --setprovideroutputsource modesetting Intel > > xrandr --output DP-1-1 --mode 2560x1440 # should not crash > > glxgears # should display gears on both screens > > > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=100086 > > Signed-off-by: Peter Wu <peter@lekensteyn.nl> > > --- > > Tested with xserver 1.20.1 with ASAN enabled. Survives multiple > > resolution changes, works with a Plasma desktop session, it seems > > stable. Something like this patch is required to make multi-monitor > > setups usable in a hybrid graphics setting with Xorg 1.20. > > --- > > src/sna/sna_accel.c | 18 ++++++++++++++++++ > > 1 file changed, 18 insertions(+) > > > > diff --git a/src/sna/sna_accel.c b/src/sna/sna_accel.c > > index 2f669bcf..80b116a3 100644 > > --- a/src/sna/sna_accel.c > > +++ b/src/sna/sna_accel.c > > @@ -17510,7 +17510,11 @@ static bool has_offload_slaves(struct sna *sna) > > PixmapDirtyUpdatePtr dirty; > > > > xorg_list_for_each_entry(dirty, &screen->pixmap_dirty_list, ent) { > > +#ifdef HAS_DIRTYTRACKING_DRAWABLE_SRC > > + assert(dirty->src == &sna->front->drawable); > > +#else > > assert(dirty->src == sna->front); > > +#endif > > if (RegionNotEmpty(DamageRegion(dirty->damage))) > > return true; > > } > > @@ -17671,7 +17675,11 @@ static void sna_accel_post_damage(struct sna *sna) > > if (RegionNil(damage)) > > continue; > > > > +#ifdef HAS_DIRTYTRACKING_DRAWABLE_SRC > > + src = get_drawable_pixmap(dirty->src); > > +#else > > src = dirty->src; > > +#endif > > dst = dirty->slave_dst->master_pixmap; > > > > region.extents.x1 = dirty->x; > > @@ -17922,9 +17930,15 @@ migrate_dirty_tracking(PixmapPtr old_front, PixmapPtr new_front) > > PixmapDirtyUpdatePtr dirty, safe; > > > > xorg_list_for_each_entry_safe(dirty, safe, &screen->pixmap_dirty_list, ent) { > > +#ifdef HAS_DIRTYTRACKING_DRAWABLE_SRC > > + assert(dirty->src == &old_front->drawable); > > + if (dirty->src != &old_front->drawable) > > + continue; > > +#else > > assert(dirty->src == old_front); > > if (dirty->src != old_front) > > continue; > > +#endif > > > > DamageUnregister(&dirty->src->drawable, dirty->damage); > > DamageDestroy(dirty->damage); > > @@ -17939,7 +17953,11 @@ migrate_dirty_tracking(PixmapPtr old_front, PixmapPtr new_front) > > } > > > > DamageRegister(&new_front->drawable, dirty->damage); > > +#ifdef HAS_DIRTYTRACKING_DRAWABLE_SRC > > + dirty->src = &new_front->drawable; > > +#else > > dirty->src = new_front; > > +#endif > > } > > #endif > > } > > -- > > 2.18.0
diff --git a/src/sna/sna_accel.c b/src/sna/sna_accel.c index 2f669bcf..80b116a3 100644 --- a/src/sna/sna_accel.c +++ b/src/sna/sna_accel.c @@ -17510,7 +17510,11 @@ static bool has_offload_slaves(struct sna *sna) PixmapDirtyUpdatePtr dirty; xorg_list_for_each_entry(dirty, &screen->pixmap_dirty_list, ent) { +#ifdef HAS_DIRTYTRACKING_DRAWABLE_SRC + assert(dirty->src == &sna->front->drawable); +#else assert(dirty->src == sna->front); +#endif if (RegionNotEmpty(DamageRegion(dirty->damage))) return true; } @@ -17671,7 +17675,11 @@ static void sna_accel_post_damage(struct sna *sna) if (RegionNil(damage)) continue; +#ifdef HAS_DIRTYTRACKING_DRAWABLE_SRC + src = get_drawable_pixmap(dirty->src); +#else src = dirty->src; +#endif dst = dirty->slave_dst->master_pixmap; region.extents.x1 = dirty->x; @@ -17922,9 +17930,15 @@ migrate_dirty_tracking(PixmapPtr old_front, PixmapPtr new_front) PixmapDirtyUpdatePtr dirty, safe; xorg_list_for_each_entry_safe(dirty, safe, &screen->pixmap_dirty_list, ent) { +#ifdef HAS_DIRTYTRACKING_DRAWABLE_SRC + assert(dirty->src == &old_front->drawable); + if (dirty->src != &old_front->drawable) + continue; +#else assert(dirty->src == old_front); if (dirty->src != old_front) continue; +#endif DamageUnregister(&dirty->src->drawable, dirty->damage); DamageDestroy(dirty->damage); @@ -17939,7 +17953,11 @@ migrate_dirty_tracking(PixmapPtr old_front, PixmapPtr new_front) } DamageRegister(&new_front->drawable, dirty->damage); +#ifdef HAS_DIRTYTRACKING_DRAWABLE_SRC + dirty->src = &new_front->drawable; +#else dirty->src = new_front; +#endif } #endif }
Since xorg-server 1.20, an external monitor would remain blank when used in a PRIME output slave setup. Only a cursor was visible. The cause is "Make PixmapDirtyUpdateRec::src a DrawablePtr" in xserver, the "src" pointer might point to the root window (created by the server) instead of a pixmap (as created by xf86-video-intel). Use get_drawable_pixmap to handle both cases. When built with -fsanitize=address, the following test will trigger a heap-buffer-overflow error due to to_sna_from_pixmap receiving a window instead of a pixmap. Test on a hybrid graphics laptop (Intel + modesetting/nouveau): xrandr --setprovideroutputsource modesetting Intel xrandr --output DP-1-1 --mode 2560x1440 # should not crash glxgears # should display gears on both screens Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=100086 Signed-off-by: Peter Wu <peter@lekensteyn.nl> --- Tested with xserver 1.20.1 with ASAN enabled. Survives multiple resolution changes, works with a Plasma desktop session, it seems stable. Something like this patch is required to make multi-monitor setups usable in a hybrid graphics setting with Xorg 1.20. --- src/sna/sna_accel.c | 18 ++++++++++++++++++ 1 file changed, 18 insertions(+)