At shutdown, the Xserver will free all its resources which includes the
RRCrtc and RROutput created.
Xwayland would do the same in its xwl_output_destroy() called from
xwl_close_screen(), leading to a double free of existing RRCrtc
RROutput:
Invalid read of size 4
at 0x4CDA10: RRCrtcDestroy (rrcrtc.c:689)
by 0x426E75: xwl_output_destroy (xwayland-output.c:301)
by 0x424144: xwl_close_screen (xwayland.c:117)
by 0x460E17: CursorCloseScreen (cursor.c:187)
by 0x4EB5A3: AnimCurCloseScreen (animcur.c:106)
by 0x4EF431: present_close_screen (present_screen.c:64)
by 0x556D40: dix_main (main.c:354)
by 0x6F0D290: (below main) (in /usr/lib/libc-2.24.so)
Address 0xbb1fc30 is 0 bytes inside a block of size 728 free'd
at 0x4C2BDB0: free (in
/usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
by 0x4CCE5F: RRCrtcDestroyResource (rrcrtc.c:719)
by 0x577541: doFreeResource (resource.c:895)
by 0x5787B5: FreeClientResources (resource.c:1161)
by 0x578862: FreeAllResources (resource.c:1176)
by 0x556C54: dix_main (main.c:323)
by 0x6F0D290: (below main) (in /usr/lib/libc-2.24.so)
Block was alloc'd at
at 0x4C2CA6A: calloc (in
/usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
by 0x4CC6DB: RRCrtcCreate (rrcrtc.c:76)
by 0x426D1C: xwl_output_create (xwayland-output.c:264)
by 0x4232EC: registry_global (xwayland.c:431)
by 0x76CB1C7: ffi_call_unix64 (in /usr/lib/libffi.so.6.0.4)
by 0x76CAC29: ffi_call (in /usr/lib/libffi.so.6.0.4)
by 0x556CEFD: wl_closure_invoke (connection.c:935)
by 0x5569CBF: dispatch_event.isra.4 (wayland-client.c:1310)
by 0x556AF13: dispatch_queue (wayland-client.c:1456)
by 0x556AF13: wl_display_dispatch_queue_pending
(wayland-client.c:1698)
by 0x556B33A: wl_display_roundtrip_queue (wayland-client.c:1121)
by 0x42371C: xwl_screen_init (xwayland.c:631)
by 0x552F60: AddScreen (dispatch.c:3864)
And:
Invalid read of size 4
at 0x522890: RROutputDestroy (rroutput.c:348)
by 0x42684E: xwl_output_destroy (xwayland-output.c:302)
by 0x423CF4: xwl_close_screen (xwayland.c:118)
by 0x4B6377: CursorCloseScreen (cursor.c:187)
by 0x539503: AnimCurCloseScreen (animcur.c:106)
by 0x53D081: present_close_screen (present_screen.c:64)
by 0x43DBF0: dix_main (main.c:354)
by 0x7068730: (below main) (libc-start.c:289)
Address 0xc403190 is 0 bytes inside a block of size 154 free'd
at 0x4C2CD5A: free (vg_replace_malloc.c:530)
by 0x521DF3: RROutputDestroyResource (rroutput.c:389)
by 0x45DA61: doFreeResource (resource.c:895)
by 0x45ECFD: FreeClientResources (resource.c:1161)
by 0x45EDC2: FreeAllResources (resource.c:1176)
by 0x43DB04: dix_main (main.c:323)
by 0x7068730: (below main) (libc-start.c:289)
Block was alloc'd at
at 0x4C2BBAD: malloc (vg_replace_malloc.c:299)
by 0x52206B: RROutputCreate (rroutput.c:84)
by 0x426763: xwl_output_create (xwayland-output.c:270)
by 0x422EDC: registry_global (xwayland.c:432)
by 0x740FC57: ffi_call_unix64 (unix64.S:76)
by 0x740F6B9: ffi_call (ffi64.c:525)
by 0x5495A9D: wl_closure_invoke (connection.c:949)
by 0x549283F: dispatch_event.isra.4 (wayland-client.c:1274)
by 0x5493A13: dispatch_queue (wayland-client.c:1420)
by 0x5493A13: wl_display_dispatch_queue_pending
(wayland-client.c:1662)
by 0x5493D2E: wl_display_roundtrip_queue (wayland-client.c:1085)
by 0x4232EC: xwl_screen_init (xwayland.c:632)
by 0x439F50: AddScreen (dispatch.c:3864)
Split xwl_output_destroy() into xwl_output_destroy() which frees the
wl_output and the xwl_output structure, and xwl_output_remove() which
does the RRCrtcDestroy() and RROutputDestroy() and call the latter only
when an output is effectively removed.
An additional benefit, on top of avoiding a double free, is to avoid
updating the screen size at shutdown.
Signed-off-by: Olivier Fourdan <ofourdan@redhat.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
The frame callback set up via wl_surface_frame() needs to be freed with
wl_callback_destroy() or we'll leak memory.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=97065
Signed-off-by: Olivier Fourdan <ofourdan@redhat.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
This loop was written in a buggy style, causing a NULL driver ptr to be
passed to copyScreen(). copyScreen() only uses that to generate an
identifier string, so this is mostly harmless on systems that accept
NULL for asprintf() "%s" format. (the generated identifiers are off
by one wrt the driver names and the last one contains NULL.
For systems that don't accept NULL for '%s' this would cause a
segmentation fault when this code is used (no xorg.conf, but partial
config in xorg.conf.d for instance).
Signed-off-by: Matthieu Herrb <matthieu@herrb.eu>
Reviewed-by: Keith Packard <keithp@keithp.com>
In function xf86VGAarbiterScrnInit when the "pEnt->bus.type" is
BUS_PLATFORM, the "pScrn->vgaDev" won't be set, so the "pScrn->vgaDev" is
equal to zero.
The variable "rsrc_decodes" in function "xf86VGAarbiterAllowDRI" is not
initialized. So it will occur error when "pScrn->vgaDev == 0", and
"vga_count > 1". For this case, as "pScrn->vgaDev == 0", the function
"pci_device_vgaarb_get_info" will only set the value of "vga_count",
but won't set the value of "rsrc_decodes", so it will has two different
return values for function "xf86VGAarbiterAllowDRI" in different
platforms. One platform will return TRUE, as the "rsrc_decodes" 's
default value is 0, but another platform will return FALSE, as the
"rsrc_decodes" 's default value is "32767", this will cause disable
direct rendering.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96937
Signed-off-by: Emily Deng <Emily.Deng@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Any code called from the driver ScreenInit may want to refer to
pScrn->pScreen. As the function passed to AddScreen is the first place
the DDX sees a new screen, the generic code needs to make sure that
value is set before passing control to the video driver's
initialization code.
This was found by running a driver which didn't bother to set this
value when the initial colormap was installed; xf86RandR12LoadPalette
tried to use pScrn->pScreen and crashed.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=97124
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-and-Tested-by: Michel Dänzer <michel.daenzer@amd.com>
This is a problem for the libinput driver that uses the same context across
multiple devices. The driver may be halfway through setting up an input device
(and the only way to do so is to add it to libinput) when the input thread
comes in an reads events. This then causes mayhem when data is dereferenced
that hasn't been set up yet.
In my case the cause was the call to libinput_path_remove_device() inside
preinit racing with evdev_dispatch_device() handling of ENODEV. The sequence
was:
- thread 2 gets an event and calls evdev_dispatch_device()
- thread 1 calls libinput_path_remove_device() which sets the device->source
to NULL
- thread 2 reads from the fd, gets ENODEV and now removes the device->source,
dereferencing the null-pointer
This is the one I could reproduce the most, but there are other potential
pitfalls that affect any driver that uses the same fd for multiple devices.
Avoid all this and wrap PreInit into the lock.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Keith Packard <keithp@keithp.com>
If a device couldn't be enabled we left the lock hanging.
This patch also removes the leftover OsReleaseSignals() call, now unnecessary.
Note that input_unlock() is later than previously OsReleaseSignals().
RemoveDevice() manipulates the input device and its file descriptors, it's
safer to put the input_unlock() call after RemoveDevice() to avoid events
coming in while the device is being removed.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Keith Packard <keithp@keithp.com>
Just use the RandR gamma ramp directly.
Fixes random on-monitor colours with drivers which don't call
xf86HandleColormaps, e.g. modesetting.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=97154
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Embarassingly, it looks like I introduced this dead function in
commit 13c7d53df8 a year ago.
Nothing ever used it, not even then.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
Instead of breaking the former when the driver supports the latter,
hook them up so that the hardware LUTs reflect the combination of the
current colourmap and gamma states. I.e. combine the colourmap, the
global gamma value/ramp and the RandR 1.2 per-CRTC gamma ramps into one
combined LUT per CRTC.
Fixes e.g. gamma sliders not working in games.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=27222
v2:
* Initialize palette_size and palette struct members, fixes crash on
server startup.
v3:
* Free randrp->palette in xf86RandR12CloseScreen, fixes memory leak.
v4:
* Call CMapUnwrapScreen if xf86RandR12InitGamma fails (Emil Velikov).
* Still allow xf86HandleColormaps to be called with a NULL loadPalette
parameter in the xf86_crtc_supports_gamma case.
v5:
* Clean up inner loops in xf86RandR12CrtcComputeGamma (Keith Packard)
* Move palette update out of per-CRTC loop in xf86RandR12LoadPalette
(Keith Packard)
v6:
* Handle reallocarray failure in xf86RandR12LoadPalette (Keith Packard)
Reviewed-by: Keith Packard <keithp@keithp.com>
This would normally return the same values the core RandR code passed to
xf86RandR12CrtcSetGamma before, which is rather pointless. The only
possible exception would be if a driver tried initializing
crtc->gamma_red/green/blue to reflect the hardware LUT state on startup,
but that can't work correctly if whatever set the LUT before the server
started was running at a different depth.
Even the pointless round-trip case will no longer work with the
following change.
Reviewed-by: Keith Packard <keithp@keithp.com>
RRCrtcGammaSetSize cannot be used yet in xf86InitialConfiguration,
because randr_crtc isn't allocated yet at that point, but a following
change will require RRCrtcGammaSetSize to be called from
xf86RandR12CrtcInitGamma.
v2:
* Bail from xf86RandR12CrtcInitGamma if !crtc->funcs->gamma_set (Keith
Packard)
Reviewed-by: Keith Packard <keithp@keithp.com>
This uses the wrapper in case we need to emulate poll with select
as we do on Windows.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
Update for removal of fdset from Block/Wakeup handler API in 9d15912a
Signed-off-by: Jon Turney <jon.turney@dronecode.org.uk>
Reviewed-by: Keith Packard <keithp@keithp.com>
Update for removal of AddEnabledDevice in be5a513f. Use SetNotifyFd instead.
Signed-off-by: Jon Turney <jon.turney@dronecode.org.uk>
Reviewed-by: Keith Packard <keithp@keithp.com>
ATTR_KEY maps to ID_INPUT_KEY which is set for any device with keys.
ID_INPUT_KEYBOARD and thus ATTR_KEYBOARD is set for devices that are actual
keyboards (and have a set of expected keys).
Hand-written match rules may only apply ID_INPUT_KEYBOARD, so make sure we
match on that too.
Arguably we should've been matching on ATTR_KEYBOARD only all along but
changing that likely introduces regressions.
Reported-by: Marty Plummer <netz.kernel@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This error code can mean we're submitting more rects at once than the
driver can handle. If that happens, resubmit one at a time.
v2: Make the rect submit loop more error-proof (Walter Harms)
Signed-off-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Michael Thayer <michael.thayer@oracle.com>
This removes the last uses of fd_set from the server interfaces
outside of the OS layer itself.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
With no users of the interface needing the readmask anymore, we can
remove it from the argument passed to these functions.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Use SetNotifyFd instead, with the hope that someday someone will come
fix this to be more efficient -- right now, the wakeup handler is
doing the event reading, instead of the notify callback.
v2: no need to patch dmxsigio.c as it has been removed.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
This is a cleanup, proposed by Adam Jackson, but wasn't merged with
the original NotifyFD changes.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
This is a cleanup, proposed by Adam Jackson, but wasn't merged with
the original NotifyFD changes.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Remove code in xf86Wakeup for dealing with other input and switch to
using the new NotifyFd interface.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
This new libXfont API eliminates exposing internal X server symbols to
the font library, replacing those with a struct full of the entire API
needed to use that library.
v2: Use libXfont2 instead of libXfont_2
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
The intent here was that fallback drivers would be at the end of the
list in order, but if a fallback driver happened to be at the end of the
list already that's not what would happen. Rather than open-code
something smarter, just use qsort.
Note that qsort puts things in ascending order, so somewhat backwardsly
fallbacks are greater than native drivers, and vesa is greater than
modesetting.
v2: Use strcmp to compare non-fallback drivers so we get a predictable
result if your libc's qsort isn't stable (Keith Packard)
Reviewed-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Adam Jackson <ajax@redhat.com>
Otherwise if the geometry changes but the mode doesn't we end up with
the previous geometry from RR's point of view.
Fixes https://bugzilla.gnome.org/show_bug.cgi?id=768710
Reviewed-by: Jonas Ådahl <jadahl@gmail.com>
Signed-off-by: Rui Matos <tiagomatos@gmail.com>
/dev/vc/0 is a devfs thing which is long dead, so stop trying to open
/dev/vc/0, besides being a (small) code cleanup this will also fix the
"parse_vt_settings: Cannot open /dev/tty0 (%s)\n" error message to
display the actual error, rather then the -ENOENT from also trying
/dev/vc/0.
BugLink: https://patchwork.freedesktop.org/patch/8768/
Reported-by: Chad Versace <chad.versace@linux.intel.com>
Suggested-by: Julien Cristau <jcristau@debian.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Julien Cristau <jcristau@debian.org>
Reviewed-by: Chad Versace <chad.versace@intel.com>
Commit 80e64dae: "modesetting: Implement PRIME syncing as a sink" originally was
supposed to have this line, but it was dropped as part of the merge process.
Foregoing the NULL assignment causes a ton of problems with dereferencing
uninitialized memory.
Signed-off-by: Alex Goins <agoins@nvidia.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
In newer laptops with switchable graphics, the GPU may have 0 outputs,
in this case the modesetting driver should still load if the GPU is
SourceOffload capable, so that it can be used as an offload source provider.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
When a card has import capability it can be an offload _sink_, not
a source and vice versa for export capability.
This commit fixes the modesetting driver to properly set these
capabilities, this went unnoticed sofar because most gpus have both
import and export capability.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Server GPUs often have a VNC feature attached to allow remote console.
The controller implementing this feature is usually not very powerful,
and we can easily swamp it with work. This is made somewhat worse by
damage over-reporting the size of the dirty region, and a whole lot
worse by applications (or shells) that update the screen with identical
pixel content as was already there.
Fix this by double-buffering the shadow fb, using memcmp to identify
dirty tiles on each update pass. Since both shadows are in host memory
the memcmp is cheap, and worth it given the win in network bandwidth.
The tile size is somewhat arbitrarily chosen to be one cacheline wide at
32bpp on Intel Core.
By default we enable this behaviour for (a subset of) known server GPUs;
the heuristic could use work.
Signed-off-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
24bpp front buffers tend to be the least well tested path for client
rendering. On the qemu cirrus emulation, and on some Matrox G200 server
chips, the hardware can't do 32bpp at all. It's better to just allocate
a 32bpp shadow and downconvert in the upload hook than expose a funky
pixmap format to clients.
[ajax: Ported from RHEL and separate modesetting driver, lifted kbpp
into the drmmode struct, cleaned up commit message, fixed 16bpp]
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Dave Airlied <airlied@redhat.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
[hdegoede@redhat.com: rebase, also use kbpp for rotate shadow fb]
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
With the previous patch, the modesetting driver can now return whether
the driver supports hw cursor. However, it alone doesn't suffice,
unfortunately. drmmode_load_cursor_argb_check() is called in the
following chain:
xf86CursorSetCursor()
-> xf86SetCursor()
-> xf86DriverLoadCursorARGB()
-> xf86_load_cursor_argb()
-> xf86_crtc_load_cursor_argb()
-> drmmode_load_cursor_argb_check()
*but* at first with drmmode_crtc->cursor_up = FALSE. Then the
function doesn't actually set the cursor but returns TRUE
unconditionally. The actual call of drmmode_set_cursor() is done at
first via the show_cursor callback, and there is no check of sw cursor
fallback any longer at this place. Since it's called only once per
cursor setup, so the xserver still thinks as if the hw cursor is
supported.
This patch is an ad hoc fix to correct the behavior somehow: it does
call drmmode_set_cursor() at the very first time even if cursor_up is
FALSE, then quickly hides again. In that way, whether the hw cursor
is supported is evaluated in the right place at the right time.
Of course, it might be more elegant if we have a more proper mechanism
to fall back to sw cursor at any call path. But it'd need more
rework, so I leave this workaround as is for now.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
The modesetting driver still has an everlasting bug of invisible
cursor on cirrus and other KMS drivers where no hardware cursor is
supported. This patch is a part of an attempt to address it.
This patch particularly converts the current load_cursor_argb callback
of modesetting driver to load_cursor_argb_check so that it can return
whether the driver handles the hw cursor or falls back to the sw
cursor.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
[hdegoede@redhat.com: Add extra comment suggested by Kenneth]
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
The error value isn't always -EINVAL, e.g. the kernel drm core returns
-ENXIO when the corresponding ops doesn't exist. Without this fix,
DRM_IOCTL_MODE_CURSOR2 would be dealt as success even if it
shouldn't.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Implements (Start/Stop)FlippingPixmapTracking, PresentSharedPixmap, and
RequestSharedPixmapNotifyDamage, the source functions for PRIME
synchronization and double buffering. Allows modesetting driver to be used
as a source with PRIME synchronization.
v1: N/A
v2: N/A
v3: N/A
v4: Initial commit
v5: Move disabling of reverse PRIME on sink to sink commit
v6: Rebase onto ToT
v7: Unchanged
Reviewed-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Alex Goins <agoins@nvidia.com>
Reverse PRIME seems to be designed with discrete graphics as a sink in
mind, designed to do an extra copy from sysmem to vidmem to prevent a
discrete chip from needing to scan out from sysmem.
The criteria it used to detect this case is if we are a GPU screen and
Glamor accelerated. It's possible for i915 to fulfill these conditions,
despite the fact that the additional copy doesn't make sense for i915.
Normally, you could just set AccelMethod = none as an option for the device
and call it a day. However, when running with modesetting as both the sink
and the source, Glamor must be enabled.
Ideally, you would be able to set AccelMethod individually for devices
using the same driver, but there seems to be a bug in X option parsing that
makes all devices on a driver inherit the options from the first detected
device. Thus, glamor needs to be enabled for all or for none until that bug
(if it's even a bug) is fixed.
Nonetheless, it probably doesn't make sense to do the extra copy on i915
even if Glamor is enabled for the device, so this is more user friendly by
not requiring users to disable acceleration for i915.
v1: N/A
v2: N/A
v3: N/A
v4: Initial commit
v5: Unchanged
v6: Rebase onto ToT
v7: NULL check and free drmVersionPtr
Reviewed-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Alex Goins <agoins@nvidia.com>
UDL (USB 2.0 DisplayLink DRM driver) and other drivers for USB transport devices
have strange semantics when it comes to vblank events, due to their inability to
get the actual vblank info.
When doing a page flip, UDL instantly raises a vblank event without waiting for
vblank. It also has no support for DRM_IOCTL_WAIT_VBLANK, and has some strange
behavior with how it handles damage when page flipping.
It's possible to get something semi-working by hacking around these issues,
but even then there isn't much value-add vs single buffered PRIME, and it
reduces maintainability and adds additional risks to the modesetting driver
when running with more well-behaved DRM drivers.
Work needs to be done on UDL in order to properly support synchronized
PRIME. For now, just blacklist it, causing RandR to fall back to
unsynchronized PRIME.
This patch originally blacklisted UDL by name, but it was pointed out that there
are other USB transport device drivers with similar limitations, so it was
expanded to blacklist all USB transport devices.
v1: N/A
v2: N/A
v3: Initial commit
v4: Move check to driver.c for consistency/visibility
v5: Refactor to accomodate earlier changes
v6: Rebase onto ToT
v7: Expand to blacklist all USB transport devices, not just UDL
Signed-off-by: Alex Goins <agoins@nvidia.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
DPMS would prevent page flip / vblank events from being raised, freezing
the screen until PRIME flipping was reinitialized. To handle DPMS cleanly,
suspend PRIME page flipping when DPMS mode is not on, and resume it when
DPMS mode is on.
v1: Initial commit
v2: Moved flipping_active check from previous commit to here
v3: Unchanged
v4: Unchanged
v5: Move flipping_active check to sink support commit
v6: Rebase onto ToT
v7: Unchanged
Reviewed-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Alex Goins <agoins@nvidia.com>
Implements (Enable/Disable)SharedPixmapFlipping and
SharedPixmapNotifyDamage, the sink functions for PRIME synchronization and
double buffering. Allows modesetting driver to be used as a sink with PRIME
synchronization.
Changes dispatch_slave_dirty to flush damage from both scanout pixmaps.
Changes drmmode_set_scanout_pixmap*() functions to
drmmode_set_target_scanout_pixmap*() that take an additional parameter
PixmapPtr *target. Then, treat *target as it did prime_pixmap. This allows
me to use it to explicitly set both prime_pixmap and prime_pixmap_back
individually. drmmode_set_scanout_pixmap() without the extra parameter
remains to cover the single-buffered case, but only works if we aren't
already double buffered.
driver.c:
Add plumbing for rr(Enable/Disable)SharedPixmapFlipping and
SharedPixmapNotifyDamage.
Change dispatch_dirty_crtc to dispatch_dirty_pixmap, which functions the
same but flushes damage associated with a ppriv instead of the crtc, and
chanage dispatch_slave_dirty to use it on both scanout pixmaps if
applicable.
drmmode_display.h:
Add flip_seq field to msPixmapPrivRec to keep track of the event handler
associated with a given pixmap, if any.
Add wait_for_damage field to msPixmapPrivRec to keep track if we have
requested a damage notification from the source.
Add enable_flipping field to drmmode_crtc_private_rec to keep track if
flipping is enabled or disabled.
Add prime_pixmap_back to drmmode_crtc_private_rec to keep track of back
buffer internally.
Add declarations for drmmode_SetupPageFlipFence(),
drmmode_EnableSharedPixmapFlipping(),
drmmode_DisableSharedPixmapFlipping, drmmode_SharedPixmapFlip(), and
drmmode_SharedPixmapPresentOnVBlank().
Move slave damage from crtc to ppriv.
drmmode_display.c:
Change drmmode_set_scanout_pixmap*() functions to
drmmode_set_target_scanout_pixmap*() that take an additional parameter
PixmapPtr *target for explicitly setting different scanout pixmaps.
Add definitions for functions drmmode_SharedPixmapFlip(),
drmmode_SharedPixmapPresentOnVBlank(),
drmmode_SharedPixmapPresent(),
drmmode_SharedPixmapVBlankEventHandler(),
drmmode_SharedPixmapVBlankEventAbort(),
drmmode_EnableSharedPixmapFlipping(), and
drmmode_DisableSharedPixmapFlipping,
drmmode_InitSharedPixmapFlipping(), and
drmmode_FiniSharedPixmapFlipping, along with struct
vblank_event_args.
The control flow is as follows:
pScrPriv->rrEnableSharedPixmapFlipping() makes its way to
drmmode_EnableSharedPixmapFlipping(), which sets enable_flipping to
TRUE and sets both scanout pixmaps prime_pixmap and
prime_pixmap_back.
When setting a mode, if prime_pixmap is defined, modesetting
driver will call drmmode_InitSharedPixmapFlipping(), which if
flipping is enabled will call drmmode_SharedPixmapPresent() on
scanout_pixmap_back.
drmmode_SharedPixmapPresent() requests that for the source to
present on the given buffer using master->PresentSharedPixmap(). If
it succeeds, it will then attempt to flip to that buffer using
drmmode_SharedPixmapFlip(). Flipping shouldn't fail, but if it
does, it will raise a warning and try drmmode_SharedPixmapPresent()
again on the next vblank using
drmmode_SharedPixmapPresentOnVBlank().
master->PresentSharedPixmap() could fail, in most cases because
there is no outstanding damage on the mscreenpix tracked by the
shared pixmap. In this case, drmmode_SharedPixmapPresent() will
attempt to use master->RequestSharedPixmapNotifyDamage() to request
for the source driver to call slave->SharedPixmapNotifyDamage() in
response to damage on mscreenpix. This will ultimately call
into drmmode_SharedPixmapPresentOnVBlank() to retry
drmmode_SharedPixmapPresent() on the next vblank after
accumulating damage.
drmmode_SharedPixmapFlip() sets up page flip event handler by
packing struct vblank_event_args with the necessary parameters, and
registering drmmode_SharedPixmapVBlankEventHandler() and
drmmode_SharedPixmapVBlankEventAbort() with the modesetting DRM
event handler queue. Then, it uses the drmModePageFlip() to flip on
the next vblank and raise an event.
drmmode_SharedPixmapPresentOnVBlank() operates similarly to
drmmode_SharedPixmapFlip(), but uses drmWaitVBlank() instead of
drmModePageFlip() to raise the event without flipping.
On the next vblank, DRM will raise an event that will ultimately be
handled by drmmode_SharedPixmapVBlankEventHandler(). If we flipped,
it will update prime_pixmap and prime_pixmap_back to reflect that
frontTarget is now being displayed, and use
drmmode_SharedPixmapPresent(backTarget) to start the process again
on the now-hidden shared pixmap. If we didn't flip, it will just
use drmmode_SharedPixmapPresent(frontTarget) to start the process
again on the still-hidden shared pixmap.
Note that presentation generally happens asynchronously, so with
these changes alone tearing is reduced, but we can't always
guarantee that the present will finish before the flip. These
changes are meant to be paired with changes to the sink DRM driver
that makes flips wait on fences attached to dmabuf backed buffers.
The source driver is responsible for attaching the fences and
signaling them when presentation is finished.
Note that because presentation is requested in response to a
vblank, PRIME sources will now conform to the sink's refresh rate.
At teardown, pScrPriv->rrDisableSharedPixmapFlipping() will be
called, making its way to drmmode_FiniSharedPixmapFlipping().
There, the event handlers for prime_pixmap and prime_pixmap_back
are aborted, freeing the left over parameter structure. Then,
prime_pixmap and prime_pixmap back are unset as scanout pixmaps.
Register and tear down slave damage per-scanout pixmap instead of
per-crtc.
v1: Initial commit
v2: Renamed PresentTrackedFlippingPixmap to PresentSharedPixmap
Renamed flipSeq to flip_seq
Warn if flip failed
Use SharedPixmapNotifyDamage to retry on next vblank after damage
v3: Refactor to accomodate moving (rr)StartFlippingPixmapTracking and
(rr)(Enable/Disable)SharedPixmapFlipping to rrScrPrivRec from ScreenRec
Do damage tracking on both scanout pixmaps
v4: Tweaks to commit message
v5: Revise for internal storage of prime pixmap ptrs
Move disabling for reverse PRIME from source commit to here
Use drmmode_set_target_scanout_pixmap*() to set scanout pixmaps
internally to EnableSharedPixmapFlipping().
Don't support flipping if ms->drmmode.pageflip == FALSE.
Move flipping_active check to this commit
v6: Rebase onto ToT
v7: Unchanged
Reviewed-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Alex Goins <agoins@nvidia.com>
ms->drmmode.pageflip was only loaded from options if ms->drmmode.glamor was
defined, otherwise it would always assume FALSE.
PRIME Synchronization requires ms->drmmode.pageflip even if we aren't using
glamor, so load it unconditionally.
v1: N/A
v2: N/A
v3: N/A
v4: N/A
v5: Initial commit
v6: Rebase onto ToT
v7: Unchanged
Reviewed-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Alex Goins <agoins@nvidia.com>
drmmode_set_scanout_pixmap_(cpu/gpu) would only do teardown if ppix ==
NULL. This meant that if there were consecutive calls to
SetScanoutPixmap(ppix != NULL) without calls to SetScanoutPixmap(ppix ==
NULL) in between, earlier calls would be leaked. RRReplaceScanoutPixmap()
does this today.
Instead, when setting a scanout pixmap, always do teardown of the existing
scanout pixmap before setting up the new one. Then, if there is no new one
to set up, stop there.
This maintains the previous behavior in all cases except those with
multiple consecutive calls to SetScanoutPixmap(ppix != NULL).
v1: N/A
v2: N/A
v3: N/A
v4: N/A
v5: Initial commit
v6: Rebase onto ToT
v7: Unchanged
Reviewed-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Alex Goins <agoins@nvidia.com>
modesetting relied on randr_crtc->scanout_pixmap being consistent with
calls to SetScanoutPixmap, which is very fragile and makes a lot of
assumptions about the caller's behavior.
For example, RRReplaceScanoutPixmap(), when dropping off with !size_fits,
will set randr_crtc->scanout_pixmap = NULL and then call SetScanoutPixmap.
Without this patch, drmmode_set_scanout_pixmap_(cpu/gpu) will think that
there is no scanout pixmap to tear down, because it's already been set to
NULL.
By keeping track of the scanout pixmap in its internal state, modesetting
can avoid these types of bugs and reduce constraints on calling
conventions.
v1: N/A
v2: N/A
v3: N/A
v4: N/A
v5: Initial commit
v6: Rebase onto ToT
v7: Unchanged
Reviewed-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Alex Goins <agoins@nvidia.com>