Create out-fence using eglCreateSyncKHR or OUT_FENCE_PTR DRM
property when page-flipping.

Signed-off-by: Louis-Francis Ratté-Boulianne <lfrb@collabora.com>

Allow the modesetting present implementation to wait on
in-fences using ANDROID_native_fence_sync if the fence
was created with DRI3 FenceFromDMAFenceFD.

Signed-off-by: Louis-Francis Ratté-Boulianne <lfrb@collabora.com>

Modesetting/glamor can handle DMA fences if extensions EGL_KHR_wait_sync and
EGL_ANDROID_native_fence_sync are available and if IN_FENCE_FD
and OUT_FENCE_PTR DRM properties are implemented.

Signed-off-by: Louis-Francis Ratté-Boulianne <lfrb@collabora.com>

Drivers need to implement flip_with_fence and flush_with_fence
to create out-fences.

Signed-off-by: Louis-Francis Ratté-Boulianne <lfrb@collabora.com>

Instead of relying on the present module to wait on in-fences,
we added some hooks to allow the drivers/hardware to do the work.
Drivers need to implement can_wait_fence, wait_fence and
flip_with_fence to handle it.

v2: Rename hooks to can_wait_fence and wait_fence

Signed-off-by: Louis-Francis Ratté-Boulianne <lfrb@collabora.com>

Enable support for DMA fences in Glamor backend.

Signed-off-by: Louis-Francis Ratté-Boulianne <lfrb@collabora.com>

Bump DRI3 version to v1.2 that adds requests:

    * FenceFromDMAFenceFD is used to create a sync fence
      from a sync file fence.
    * DMAFenceFDFromFence is used to retrieve the sync
      file fd from a fence.

Drivers need to check the SyncFence type to make use of it
when possible.

Signed-off-by: Louis-Francis Ratté-Boulianne <lfrb@collabora.com>

Functions are available to create a SyncFence from a DMA fence
fd and to retrieve the file descriptor from a SyncFence.

DMA fences can't be triggered/reset externally.

v2: Make SetTriggered and Reset return a status and remove fence type

Signed-off-by: Louis-Francis Ratté-Boulianne <lfrb@collabora.com>

Needed so multiple types of sync fences can exist (SHM and DMA).

v2: Remove unused miSyncGetFenceType

Signed-off-by: Louis-Francis Ratté-Boulianne <lfrb@collabora.com>

Enable DRI3 v1.1 now that all functions have been implemented and
that there is at least one backend implementing the driver hooks
(modesetting/glamor).

Signed-off-by: Louis-Francis Ratté-Boulianne <lfrb@collabora.com>

Using modifier might allow the driver to use a more optimal format
(e.g. tiled/compressed). Let's try to use those if possible.

Signed-off-by: Louis-Francis Ratté-Boulianne <lfrb@collabora.com>

Implement function added in DRI3 v1.1.

A newest version of libepoxy (>= 1.4.4) is required as earlier
versions use a problematic version of Khronos
EXT_image_dma_buf_import_modifiers spec.

v4: Only send scanout-supported modifiers if flipping is possible

Signed-off-by: Louis-Francis Ratté-Boulianne <lfrb@collabora.com>

It relies on GBM >= 17.1.0 where we can import BO with multiple
planes and a format modifier (GBM_BO_IMPORT_FD_MODIFIER).

Signed-off-by: Louis-Francis Ratté-Boulianne <lfrb@collabora.com>

Add support for 'check_flip2' so that the present core can know
why it is impossible to flip in that scenario. The core can then
let know the client that the buffer format/modifier is suboptimal.

Signed-off-by: Louis-Francis Ratté-Boulianne <lfrb@collabora.com>

Use most optimal buffer format (e.g. tiled/compressed) available
for scanout.

v2: Don't use multi-plane modifier to create scanout buffer

v3: Add flag to retrieve modifiers set from enabled CRTCs only

v4: Fix uses when GBM/EGL driver doesn't support modifiers

Signed-off-by: Louis-Francis Ratté-Boulianne <lfrb@collabora.com>

Retrieve IN_FORMATS property from the plane. It gives the
allowed formats and modifiers for BO allocation.

Signed-off-by: Louis-Francis Ratté-Boulianne <lfrb@collabora.com>

To make sure we also use the same primary plane and to avoid
mixing uses of two APIs, it is better to always use the atomic
modesetting API when possible.

v2: Don't use mode_output->connector_id

Signed-off-by: Louis-Francis Ratté-Boulianne <lfrb@collabora.com>

This allows the uses of CCS compressed or tiled pixmaps as BOs when
page-flipping.

Signed-off-by: Louis-Francis Ratté-Boulianne <lfrb@collabora.com>

In order to flip between compressed and uncompressed buffers -
something drmModePageFlip explicitly bans us from doing - we need
to port use the atomic modesetting API. It's only 'fake' atomic
though given we still commit for each CRTC separately and
CRTC and connector properties are not set with the atomic API.

The helper functions to retrieve DRM properties have been borrowed
from Weston.

Signed-off-by: Louis-Francis Ratté-Boulianne <lfrb@collabora.com>

Add 'check_flip2' hook for driver to let know the core
about why flipping is not possible ('reason').
If it is because of unsupported buffer format/modifier,
a PresentCompleteNotify event is sent to the client with
the PresentCompleteModeSuboptimalCopy mode.

v2: Check for PresentOptionSuboptimal and check driver version
    before using 'check_flip2'.

Signed-off-by: Louis-Francis Ratté-Boulianne <lfrb@collabora.com>

Initial implementation for DRI3 v1.1. Only the DRI3 implementation
is there, backends need to implement the proper hooks.

Version is still set to 1.0 so clients shouldn't use the new
requests yet.

v2: Use depth/bpp instead of DRM formats in requests

v3: Remove DMA fence requests from v1.1
    Add screen/drawable modifier sets

v4: Free array returned by 'get_drawable_modifiers()'

v5: Fix FD leak

Signed-off-by: Daniel Stone <daniels@collabora.com>
Signed-off-by: Louis-Francis Ratté-Boulianne <lfrb@collabora.com>

Your libc has ffs, I promise.

Signed-off-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>

This was always wide enough to work on an fd_mask ("mask" ffs
presumably). We don't operate on fd_masks anymore, so this can go.

Signed-off-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>

Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Signed-off-by: Giuseppe Bilotta <giuseppe.bilotta@gmail.com>

Running Xephyr under valgrind reveals that we're sending some
uninitialized memory over the wire (particularly, the leftover padding
that comes from rounding extraLen to the next 32-bit multiple).

Solve by calloc()ing the memory instead of malloc()ing (the alternative
would be to memset just the padding, but I'm not sure it's more
convenient.)

Signed-off-by: Giuseppe Bilotta <giuseppe.bilotta@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>

Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>

This fixes some “Conditional jump depends on uninitialized value(s)”
errors spotted by valgrind.

Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Signed-off-by: Giuseppe Bilotta <giuseppe.bilotta@gmail.com>

Reviewed-by: Adam Jackson <ajax@redhat.com>

Previously, before GLX_OML_swap_method was fixed, both the X server and
client ignored the swapMethod fbconfig value, which meant that, if the dri
driver thought it exposed more than one swapMethod, it actually just
exported a duplicated set of fbconfigs. When fixing GLX_OML_swap_method
and restricting the choice for built-in visuals to a single swap method
that meant we didn't have that many fbconfigs to choose from when pairing
the compositing visual with an fbconfig, resulting in the fbconfig paired
with the compositing visual becoming too restrictive for some applications,
(at least for kwin). This problem would also happen if the dri driver
only exposed a single swap method to begin with.

So, to make sure the compositing visual gets a good enough fbconfig,
duplicate fbconfigs that are suitable for compositing visuals and make
sure these duplicated fbconfigs can be used only by compositing visuals.
For duplicated fbconfigs not paired with a compositing visual, construct
new compositing visuals, making compositing clients able to choose visuals
/ fbconfig more adapted to their needs.

This is in some sense equivalent to adding a new "TRUECOLOR_COMPOSITING"
GLX visualtype.

Fixes: 4486d199 ("glx: Fix visual fbconfig matching with respect to
                      swap method")
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=102806


Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Tested-By: Nick Sarnie <commendsarnex@gmail.com>
Tested-by: Fredrik Höglund <fredrik@kde.org>
Reviewed-by: Adam Jackson <ajax@redhat.com>

I don't know how many times I've had a broken server due to a bad
directory to xkbcomp, and only finding the whole path has shown me
where I went wrong.

Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>

The failing struct comes from the python test written by Michal Srb
<msrb@suse.com>.

v2: Use a drawable (root window) and gc, so that PolyLines hopefully
    actually tries processing things.  However, the request seems to
    process successfully so the poll() just stalls out.  However, this
    does let us distinguish between detecting the bigrequests error
    and not, at least.
v3: Clean up the description of what we expect the poll() call to do.
v4: Use XI2 instead of PolyLine to trigger a predictable error. We know the
    server replies with BadValue for a zero num_masks argument. So if we send
    a bigreq with a num_masks 0 and a length 0, we can just check whether we
    get killed (good) or a BadValue (bad). It doesn't test for specific memory
    overflows or crashes, but based on the assumption that we shouldn't look
    at *any* BigReq of size 0, this seems to be sufficient.

Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

Historically we didn't need to care about this case but more devices are
having invalid types set and they cannot be unset with a hwdb entry (which
doesn't handle the empty string). Allow for "0" to mean "unset" because
anything else would be crazy anyway.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

Roundhouse kick replacing the various (sizeof(foo)/sizeof(foo[0])) with
the ARRAY_SIZE macro from dix.h when possible. A semantic patch for
coccinelle has been used first. Additionally, a few macros have been
inlined as they had only one or two users.

Signed-off-by: Daniel Martin <consume.noise@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>

unsigned_tests[] was used to compute the amount of signed numbers to
test.

Signed-off-by: Daniel Martin <consume.noise@gmail.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>

input.c: In function ‘dix_event_to_core’:
../include/inputstr.h:61:55: warning: ‘*((void *)&ev+80)’ is used uninitialized in this function [-Wuninitialized]
 #define SetBit(ptr, bit)  (((BYTE *) (ptr))[(bit)>>3] |= (1 << ((bit) & 7)))
                                                       ^~

Signed-off-by: Daniel Martin <consume.noise@gmail.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>

If one misconfigures a ZaphodHeads value (more than 20 characters
without a delimiter), we get an overflow of our buffer.  Use
xstrtokenize() instead of writing/fixing our own tokenizer.

Signed-off-by: Daniel Martin <consume.noise@gmail.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>

Signed-off-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Eric Anholt <eric@anholt.net>

Signed-off-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Eric Anholt <eric@anholt.net>

Effectively always true anyway.

Signed-off-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Eric Anholt <eric@anholt.net>

Change 677c32bc refactored all usages of drmWaitVBlank() into a helper function,
ms_queue_vblank().

ms_queue_vblank() takes in an MS_QUEUE_RELATIVE flag to indicate that the
sequence number is relative rather than absolute, but still treats the actual
sequence number as absolute, passing it through ms_crtc_msc_to_kernel_msc()
unconditionally before calling drmWaitVBlank().

ms_crtc_msc_to_kernel_msc() works by subtracting a vblank offset from the
provided sequence number, which only makes sense for absolute sequence numbers.
In the case of PRIME Sync, drmmode_SharedPixmapPrsentOnVBlank() passes in 1,
which results in a large negative vblank offset. After subtracting, we're left
with a relative sequence number of 100,000+, i.e. wait for 100,000+ vblanks...

In the relative case we want to pass in the sequence number unmodified. Simply
add a check to do this.

Signed-off-by: Alex Goins <agoins@nvidia.com>
Reviewed-by: Keith Packard <keithp@keithp.com>