We stash the pending flip mode into the closure, but failed to copy that
over to the right field when continuing with a chain of flips. The
impact was that we would fail to issue the required SwapComplete to
unblock the clients.
Fixes regression from
commit 96ce362ba4
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date: Tue May 13 09:46:45 2014 +0100
sna/dri2: Do not rate-limit vblank=0 swaps
Reported-by: francisbrwn
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
By using move-area-to-gpu and specifying that we overwrite the target
area, we can optimize away any needless damage migration.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
This is fortunately a no-op, as it gets initialized to zero already
(that is the pixmap is writeable). However, we may as well do the right
thing...
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Copied from kernel commit 7d87a7f709650bde4d7d63117f25ee1c095da5dd
Author: Ville Syrjälä <ville.syrjala@linux.intel.com>
Date: Wed Apr 9 18:19:04 2014 +0300
srm/i915/chv: Add Cherryview PCI IDs
and also includes non-functional changes from
commit fd3c269f8ff940cc0fbb3b7f7e84c0572f6f759a
Author: Zhao Yakui <yakui.zhao@intel.com>
Date: Thu Apr 17 10:37:35 2014 +0800
drm/i915: Split the BDW device definition to prepare for dual BSD rings on BDW GT3
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Avoid busy-spinning waiting for the cache to expire after a failure. Try
to free up any overdue expiration, then if we still fail, we free the
entire cache. Currently, it would spin over the expiration until all
items were overdue and then freed.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
OML_sync_control:
If there are multiple outstanding swaps for the same window, at most
one such swap can be satisfied per increment of MSC. The order of
satisfying outstanding swaps of a window must be the order they were
issued.
The only challenge is to keep both this behaviour and vblank_mode=0
benchmarking.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
If we complete a flip after resizing and recreating the TearFree shadow,
we may process the flip completion events whilst the output is only
partially reconfigured.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
We have to open-code kgem_bo_is_busy() in order to avoid prematurely
retiring the rq->bo. The unfortunate outcome here is that the bo may be
recycled for a new batch before it is finally released from the request.
Reported-by: Zdenek Kabelac <zkabelac@redhat.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
GetDrawableMSC is specified as returning the frame counter and timestamp
of the last flip event associated with the drawable, not the current
hardware timestamp.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Clients are not yet ready for outputs that disappear at runtime, often
unexpectedly dieing with asynchronous BadOutput errors. A simple
workaround is to not remove any output just yet - but allow users to opt
in via xorg.conf, i.e.
Section "Device"
Option "DeleteUnusedDP12Displays" "true"
EndSection
The name chosen is to be consistent with the nvidia driver, which did
MST first and encountered all of these issues first, and has also been
adopted for -modesetting. If this is combined with persistent output
naming (i.e. based on DisplayPort branch topology), the number of
outputs presented to the user should be static.
Instead of removing the outputs, we have to mark them detached instead
to avoid throwing errors from the kernel.
Based on the patch for -modesetting by Dave Airlie.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
If valgrind is running, keep it simple and avoid using unknown
instructions (causing valgrind to die with SIGILL).
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
An addition to DRI2 that was overlooked at the time, was the support
added for having multiple outstanding swap requests in the core. The
importance of this is that we can then send the SwapComplete reply after
the pageflip is completed and not before as we currently do to fake
triple buffering - in clear violation of OML_sync_control.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
When switching back to the VT, rerun the output discovery (if such an
event was pending) before we attempt to set the desired modes.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
If we keep the separate discovery uevent, we need to run it first in the
cases where we get combined discovery+hotplug events.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
We need to preserve possible_crtcs for the later check that we can use
the output with the particular ZaphodHead (which owns a CRTC).
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
In the future, we may see a setup where connectors are created without
being attached to a specific encoder, instead associated with multiple
generic encoders - such as DP MST sink devices. To handle this, we need
to recheck their active encoder at runtime after branch configuration
events (or possibly hotplug).
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
If we fail to disable the remote output during initialisation, copy the
current configuration in order to try and keep the bookkeeping in order.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
In order to prevent a subsequent BadAccess when we try to use it as a
ShmPixmap, we need to mark the segment as writeable.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
References: https://bugs.freedesktop.org/show_bug.cgi?id=78293
Creating a ShmPixmap may cause an asynchronous BadAccess error, so wrap
the construction with XSync and check for an error before proceeding.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
References: https://bugs.freedesktop.org/show_bug.cgi?id=78293
Only mark an output as disabled if we do successfully disable it. This
will require a little more work to make sure that such errors are
cleanly propagated back to the host...
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
If we remove a connector/output, we need to not advance the iterator or
else we skip over an output (as the array is reduces inside the loop).
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
If the output was ignored, we would leak the allocations. However, we
can check the output name after the first GETCONNECTOR/GETENCODER
request before any allocations.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
From OML_sync_control:
"The graphics MSC value is incremented once for each screen refresh.
For a non-interlaced display, this means that the graphics MSC value
is incremented for each frame."
When the pipe is disabled, return the last known MSC such that the
reported value is always monotonic and consistent with the pipe when it
is active again. That also is consistent with OML_sync_control which
says that the MSC is a vertical refresh counter, and by implication does
not increase whilst the pipe is off.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
If the drawable is even partially on the PrimaryOutput, use that pipe
for determining synchronisation counters and synchronised updates.
OML_sync_control:
"For a multi-monitor system, the monitor used to
determine MSC is screen 0 of <display>."
The PrimaryOutput is the nearest approximation we have to "screen 0". We
also deviate in handling what hapens when that screen is off, which is
not mentioned at all in OML_sync_control.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Ideally, we want to assert that we have sufficient stashed cursors to
allocate for the unused CRTCS, so something like
assert(num_stash + num_active_cursors >= 0);
Reported-by: Zdenek Kabelac <zkabelac@redhat.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
One of the side-effects of MST is that we need to support dynamic
attachment and removal of displays as the branch hierachy changes.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
In the near future we will be dealing with dynamic discovery of
connectors, and so we cannot assume that the mode resources will be
static for our lifetime. The first step is to not keep that struct
around, but pull out the useful information and discard it.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
The major change here is not do a full promotion if we are filling
through a clip region. The original intention was to catch doing the
initial clear of the background, but it was also triggering for drawing
rectangular outlines etc.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Chris Wilson