Defer the actual wait until the next use of the screen pixmap, and then
if needed replace the GPU bo with an alternative back buffer.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
We need to also take account of upload buffers which include an offset
to the base map. To do so, we also need to stop conflating the cpu hint
with the current mapping flag.
References: https://bugs.freedesktop.org/show_bug.cgi?id=70461
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
And make sure we consider such overflowing strings for correct clipping
against Windows.
To offset the cost of doing a full extents check (~10% on aa10text), we
introduce an approximate extents query (~1% on aa10text). The disparity
should be rare, and should be an overestimate to force redundant
clipping.
Reported-by: Clemens Eisserer <linuxhippy@gmail.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=70541
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
After some probing mechanisms, we may end up with a valid device without
knowing its PCI address a priori. Having a valid device, we can just
query it for the correct device id, and can safely abort any path that
requires PCI information that we don't have. (Those paths are not valid
under such hosting anyway - if it may be required, we could reconstruct
the address.)
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
We know that when we access either a CPU or GTT mmap we are vulernable
to receiving a SIGBUS. In fact, we can catch these and abort the
operation preventing X and all of its clients from randomly dieing.
This helps for instance if you try and use a 1GiB frontbuffer on a 2GiB
machine...
For complete protection, we also need to catch signals for all GTT maps,
such as VBO and staging buffers. (TBD)
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
The operation is in theory redundant, and in the case of Haswell where
we have multiple outputs aliasing to the same encoder, actually harmful.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=68030
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Too often our implementation of vsync or pageflip is buggy, or for some
other reason it is desired by the user to disable those code paths. Make
it possible once again by restoring the Options for the user to control.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
This problematic GPU still seems to like to fallover when faced with
Y-tiling. It was reserved only for use with glyphs, but even that
occasionally runs into trouble, so disable all selection of Y-tiling for
our own use.
Bugzilla: https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/1222203
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
When on-battery, we would prefer to use more power efficient operations.
For example, the BCS is far more economical to more data around with, but
it doesn't have quite the same throughput as the hungry RCS. (Not that
there is any reason why, the BCS is supposed to run at full memory
speed, unfortunately that is main memory speed and not the caches...)
Note: that X already listens to acpid for video switch notifications, it
would be useful if we could extend that interface to emit power
notifications as well.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
With lots of updates by Christopher James Halse Rogers as he updated the
XMir API - but now supposedly frozen!
"<RAOF> ickle: I think the xmir api should be pretty much stable now,
barring people coming up with more awesome ways of doing things."
Signed-off-by: Christopher James Halse Rogers <raof@ubuntu.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Like it's sibling sna_pixmap_move_to_gpu(), it helps to know the private
sna_pixmap after the operation rather than just a boolean success/fail
result, and make it more robust in the process.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Always maintain one spare so that we can reconfigure for any number of
desired outputs on the fly.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Prevent a NULL dereference for the small system pixmaps. Introduced with
commit f22d7f68b8
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date: Wed Aug 28 14:24:33 2013 +0100
sna/gen6+: Improve ring stickyness for BLT composite ops
Reported-by: Sami Farin <hvtaifwkbgefbaei@gmail.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=68728
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Previously, we instantiated a fake output in case we had a machine with
no output. (For certain server-class products.) The Bumblee project were
also doing something very similar in order to fake an extended desktop
on the Intel igfx and copy portions onto a discrete GPU. (The preferred
method for doing this upstream is through the use of PRIME). As the code
is very similar, we can support both use-cases simultaneously.
This adds the option:
Section "Device"
Driver "intel"
Option "VirtualHeads" "<count>"
EndSection
to allow the user to specify an additional set of fake outputs, which
can then be controlled using xrandr.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
If we are hosted, we do not own the CRTC configuration, and deferencing
the private data structures believing them to be ours, only leads to
disaster.
Based on patches by Christopher James Halse Rogers.
Reported-by: Christopher James Halse Rogers <raof@ubuntu.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
When you don't have many cycles to play with, every one counts.
Here we make sure we cache negative lookups for large glyphs.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Start adding the infrastructure to disable direct hardware access if X
is being run under a system compositor (aka "hosted").
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
We can shave a few instructions off the routine by incrementally
performing the "is-empty" check as soon as we compute the intersection
in each dimension.
Ideally, the method of swapping is something that the applications have
control over, along with how to synchronise to the vertical refresh.
Whilst triple buffering is good to reduce jitter for games (at the cost of
an extra frame of latency, usually considered a good tradeoff), it
prevents the applications from accurately controlling the presentation
of animations. One vocal critique is Owen Taylor, who demands accurate
swap control for smooth animations in gnome-shell. For example,
http://blog.fishsoup.net/2012/11/28/avoiding-jitter-in-composited-frame-display/
In lieu of application control, just apply a quirk for the compositor.
Everyone else will just have to wait for DRI3.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Atter a modeset or KMS takeover, we do a quick readback of the kernel
state in order to verify that it matches our expectations. If we find
that a foreign framebuffer is attached, or no mode if set on the output,
we then turn off that connection and release any resources associated
with that pipe. This patch tries to reduce the number of superfluous
requests to turn off a connection.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Avoid having to search the device tree once again in order to simply
recover the path we used to open the device.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
This allows us to more accurately track when the GTT is dirty. However,
the only danger is that we may prematurely flush the scanout and clear
the dirty bit and not preserve the flush timer.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
In the future, we may need to explicitly flush GTT writes to the
scanout, so add the infrastructure to do so now.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Otherwise we end up cloning the scanout only to leave it dangling if the
client copies the from the front-buffer and then writes to it.
Reported-by: Nick Bowler <nbowler@draconx.ca>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=64675
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Whilst we are not DRM master, not only is another server in control of
the outputs and responding to the udev event, we ourselves cannot
change modes and just cause contention upon the DRM device. Instead
inform userspace of the change as soon as we are DRM master again and
back in control.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Along the slow path, skip all processing of glyphs that are not visible.
This is important as the slow path handles the per-glyph redirection
case, which is much more expensive.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Chris Wilson
Raul Fernandes