Changes PRIME to use double buffering and synchronization if all required
driver functions are available.
rrcrtc.c:
Changes rrSetupPixmapSharing() to use double buffering and
synchronization in the case that all required driver functions are
available. Otherwise, falls back to unsynchronized single buffer.
Changes RRCrtcDetachScanoutPixmap() to properly clean up in the case of
double buffering.
Moves StopPixmapTracking() from rrDestroySharedPixmap() to
RRCrtcDetachScanoutPixmap().
Changes RRReplaceScanoutPixmap() to fail if we are using double buffering,
as it would need a second ppix parameter to function with double buffering,
and AFAICT no driver I've implemented double buffered source support in uses
RRReplaceScanoutPixmap().
randrstr.h:
Adds scanout_pixmap_back to struct _rrCrtc to facilitate PRIME
double buffering.
xf86Crtc.h:
Adds current_scanout_back to _xf86Crtc to facilitate detection
of changes to it in xf86RandR12CrtcSet().
xf86RandR12.c:
Changes xf86RandR12CrtcSet() to detect changes in
scanout_pixmap_back.
Adds scanout_pixmap_back to struct _rrCrtc to facilitate PRIME double
buffering.
v1: Initial commit
v2: Rename PresentTrackedFlippingPixmap to PresentSharedPixmap
v3: Refactor to accomodate moving (rr)StartFlippingPixmapTracking and
(rr)(Enable/Disable)SharedPixmapFlipping to rrScrPrivRec from ScreenRec
Add fallback if flipping funcs fail
v4: Detach scanout pixmap when destroying scanout_pixmap_back, to avoid
dangling pointers in some drivers
v5: Disable RRReplaceScanoutPixmap for double-buffered PRIME, it would need an
ABI change with support for 2 pixmaps if it were to be supported, but AFAICT
no driver that actually supports double-buffered PRIME uses it.
Refactor to use rrEnableSharedPixmapFlipping() as a substitute for
rrCrtcSetScanoutPixmap() in the flipping case.
Remove extraneous pSlaveScrPriv from DetachScanoutPixmap()
Remove extraneous protopix and pScrPriv from rrSetupPixmapSharing()
v6: Rebase onto ToT
v7: Unchanged
Reviewed-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Alex Goins <agoins@nvidia.com>
A single provider can be both a offload and source slave at the same time,
the use of seperate lists breaks in this case e.g. :
xrandr --listproviders
Providers: number : 2
Provider 0: id: 0x7b cap: 0xf, Source Output, Sink Output, Source Offload, Sink Offload crtcs: 3 outputs: 2 associated providers: 0 name:modesetting
Provider 1: id: 0x46 cap: 0xf, Source Output, Sink Output, Source Offload, Sink Offload crtcs: 2 outputs: 5 associated providers: 0 name:modesetting
xrandr --setprovideroutputsource 1 0x7b
xrandr --listproviders
Providers: number : 2
Provider 0: id: 0x7b cap: 0xf, Source Output, Sink Output, Source Offload, Sink Offload crtcs: 3 outputs: 2 associated providers: 1 name:modesetting
Provider 1: id: 0x46 cap: 0xf, Source Output, Sink Output, Source Offload, Sink Offload crtcs: 2 outputs: 5 associated providers: 1 name:modesetting
xrandr --setprovideroffloadsink 1 0x7b
xrandr --listproviders
Providers: number : 3
Provider 0: id: 0x7b cap: 0xf, Source Output, Sink Output, Source Offload, Sink Offload crtcs: 3 outputs: 2 associated providers: 2 name:modesetting
Provider 1: id: 0x46 cap: 0xf, Source Output, Sink Output, Source Offload, Sink Offload crtcs: 2 outputs: 5 associated providers: 2 name:modesetting
Provider 2: id: 0x46 cap: 0xf, Source Output, Sink Output, Source Offload, Sink Offload crtcs: 2 outputs: 5 associated providers: 2 name:modesetting
Not good. The problem is that the provider with id 0x46 now is on both
the output_slave_list and the offload_slave_list of the master screen.
This commit fixes this by unifying all 3 lists into a single slaves list.
Note that this does change the struct _Screen definition, so this is an ABI
break. I do not expect any of the drivers to actually use the removed / changed
fields so a recompile should suffice.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
This is an ABI break, in that we now pass NULL to a function that hasn't
accepted it before.
Alex Goins had a different patch for this but it wasn't symmetrical, it
freed something in a very different place than it allocated it, this
attempts to retain symmetry in the releasing of the backing bo.
v2: use a new toplevel API, though it still passes NULL to something
that wasn't expecting it.
v3: pass -1 instead of 0.
Signed-off-by: Dave Airlie <airlied@redhat.com>
Tested-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Alex Goins <agoins at nvidia.com>
Otherwise, we leave a dangling reference to the destroyed pixmap in the
master screen's pixmap_dirty_list.
Fixes regression from commit cf5d6414 ("randr: Factor out shared pixmap
destruction").
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
protopix is completely redundant with mscreenpix. Get rid of it.
We don't need rrScrPriv, so remove it.
[ajax: also squash an unused variable in RRCrtcDetachScanoutPixmap,
though it'll come back when the rest of this series lands]
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Alex Goins <agoins@nvidia.com>
The old version of rrCreateSharedPixmap(), in addition to actually creating
a shared pixmap with scanout, also set up pixmap tracking on the source
driver.
I will be needing to create multiple shared pixmaps for PRIME double
buffering, so factor the part that does shared pixmap creation into its own
function, the new rrCreateSharedPixmap(). Rename the old
rrCreateSharedPixmap() to rrSetupPixmapSharing(), a function that
replicates the old functionality of rrCreateSharedPixmap() using the new
rrCreateSharedPixmap().
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Alex Goins <agoins@nvidia.com>
Shared pixmap destruction is done by unrefing the master pixmap twice: once
for the original reference, and once for the reference implicitly added by
PixmapShareToSlave. Then, unrefing the slave pixmap once.
When I add PRIME double buffering and synchronization, I will need to do
this in multiple places. To avoid duplication of code and comments
explaining it everywhere, factor it out into its own function and use that
in place of where it was before.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Alex Goins <agoins@nvidia.com>
As we define sizeFits based on whether a CRTC is active, and skip trying
to redirect the scanout on a disable pipe, we then attempt to undo it
later and fail because crtc->scanout_pixmap != DRI2_Pixmap and
!sizeFits. Paper over this failure by skipping unredirected CRTC when
disabling.
v2: Unwind upon failure
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=84653
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Dave Airlie <airlied@redhat.com>
Reported-by: Christoph Haag <haagch@frickel.club>
Tested-by: Christoph Haag <haagch@frickel.club>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
One of the lacking features with output offloading was
that screen rotation didn't work at all.
This patch makes 0/90/180/270 rotation work with USB output
and GPU outputs.
When it allocates the shared pixmap it allocates it rotated,
and any updates to the shared pixmap are done using a composite
path that does the rotation. The slave GPU then doesn't need
to know about the rotation and just displays the pixmap.
v2:
rewrite the sync dirty helper to use the dst pixmap, and
avoid any strange hobbits and rotations.
This breaks ABI in two places.
Signed-off-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
In most of xserver code-base we define new functions at column 0, with
their return type provided on the previous line. Two functions did not
follow this rule so update them, and get them wrapped up to 80 as an
added bonus.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
This will allow us to make the code more readable, and the lines will
fit within 80 columns.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
To provide some information about the origin of the message.
Cc: Dave Airlie <airlied@redhat.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
All of these seem like left over from developments stage. Remove them as
they can cause excessive flood in the logs.
Cc: Dave Airlie <airlied@redhat.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Slaving two outputs on a secondary GPU to a primary GPU testing
picked this up, in that we'd try to resize to the totally the
wrong thing, then as usual segfault in the rotation code.
Signed-off-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Aaron Plattner <aplattner@nvidia.com>
The function RRCrtcSet call checks to see if the config being set is
already configured, but, doesn't check that the selected outputs are
connected to the crtc before skipping. This means that the following
sequence will omit the final CrtcSet call to the driver:
CRTC c1 connect to output o
CRTC c2 connect to output o
CRTC c1 connect to output o
This change adds the check to ensure that each of the calls are made to
the driver.
Signed-off-by: David Sodman <dsodman@chromium.org>
Reviewed-by: Stéphane Marchesin <marcheu@chromium.org>
Signed-off-by: Keith Packard <keithp@keithp.com>
Don't use rrScrPriv for nested screen private fetching.
Eliminate a duplicate fetch of the randr screen private in rrCheckPixmapBounding.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
This lets us stop using the 'pointer' typedef in Xdefs.h as 'pointer'
is used throughout the X server for other things, and having duplicate
names generates compiler warnings.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Send RRResourceChangeNotify event when provider, output or crtc was created or
destroyed. I.e. when the list of resources returned by RRGetScreenResources and
RRGetProviders changes.
Reviewed-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Michal Srb <msrb@suse.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
Introduce a wrapper interface so we can fix things up for multi-gpu
situations later.
This just introduces the API for now.
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>
rrcrtc.c: In function 'RRCrtcDetachScanoutPixmap':
rrcrtc.c:366:9: warning: variable 'ret' set but not used
[-Wunused-but-set-variable]
rrcrtc.c: In function 'rrCheckPixmapBounding':
rrcrtc.c:505:13: warning: variable 'ret' set but not used
[-Wunused-but-set-variable]
rrcrtc.c:445:9: warning: unused variable 'i' [-Wunused-variable]
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Mark Kettenis <kettenis@openbsd.org>
When an RandR shared pixmap is created in rrCreateSharedPixmap, it has a refcnt
of 1. Then, PixmapShareToSlave bumps the refcnt to 2. However, there's no
corresponding PixmapUnshareFromSlave where the refcnt can be decreased again,
and there's no convenient common place where the refcnt can be decremented when
the slave pixmap is destroyed.
Fix this by just unreffing the pixmap twice in RRCrtcDetachScanoutPixmap.
Signed-off-by: Aaron Plattner <aplattner@nvidia.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Tested-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Tested-by: Daniel Stone <daniel@fooishbar.org>
Ensures padding bytes are zero-filled
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Tested-by: Daniel Stone <daniel@fooishbar.org>
When passing variable pointers to functions or otherwise doing long
sequences to compute values for replies, create & use some new
temporary variables, to allow for simpler initialization of reply
structures in the following patches.
Move memsets & other initializations to group with the rest of the
filling in of the reply structure, now that they're not needed so
early in the code path.
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Tested-by: Daniel Stone <daniel@fooishbar.org>
Casting return to (void) was used to tell lint that you intended
to ignore the return value, so it didn't warn you about it.
Casting the third argument to (char *) was used as the most generic
pointer type in the days before compilers supported C89 (void *)
(except for a couple places it's used for byte-sized pointer math).
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Tested-by: Daniel Stone <daniel@fooishbar.org>
For DRI2 in some offload cases we need to set a new pixmap on the crtc,
this hook allows dri2 to call into randr to do the necessary work to set
a pixmap as the scanout pixmap for the crtc the drawable is currently on.
This is really only to be used for unredirected full screen apps in composited
environments.
Reviewed-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
The master contains the complete screen size bounds, so check
the width/height against it.
Reviewed-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Current code constrains the cursor to the crtcs on the master
device, for slave outputs to work we have to include their crtcs
in the constrain calculations.
Reviewed-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
When randr notices a crtc configuration request for a slave device,
it checks if the slave allocated pixmap exists and is suitable,
if not it allocates a new shared pixmap from the master, shares
it to the slave, and starts the master tracking damage to it,
to keep it updated from the current front pixmap.
If the resize means the crtc is no longer used it will destroy
the slave pixmap.
This adds the concept of a scanout_pixmap to the randr_crtc object,
and also adds a master pixmap pointer to the pixmap object, along
with defining some pixmap helper functions for getting pixmap box/regions.
v2: split out pixmap sharing to a separate function.
v3: update for void *
Reviewed-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
set but not used variables
shadowing a previous local
A hidden problem was that the VERIFY_RR_* macros define local 'rc'
variables, any other local definitions for those would be shadowed and
generate warnings from gcc. I've renamed the other locals 'ret'
instead of 'rc'.
Signed-off-by: Keith Packard <keithp@keithp.com>
This is strictly the application of the script 'x-indent-all.sh'
from util/modular. Compared to the patch that Daniel posted in
January, I've added a few indent flags:
-bap
-psl
-T PrivatePtr
-T pmWait
-T _XFUNCPROTOBEGIN
-T _XFUNCPROTOEND
-T _X_EXPORT
The typedefs were needed to make the output of sdksyms.sh match the
previous output, otherwise, the code is formatted badly enough that
sdksyms.sh generates incorrect output.
The generated code was compared with the previous version and found to
be essentially identical -- "assert" line numbers and BUILD_TIME were
the only differences found.
The comparison was done with this script:
dir1=$1
dir2=$2
for dir in $dir1 $dir2; do
(cd $dir && find . -name '*.o' | while read file; do
dir=`dirname $file`
base=`basename $file .o`
dump=$dir/$base.dump
objdump -d $file > $dump
done)
done
find $dir1 -name '*.dump' | while read dump; do
otherdump=`echo $dump | sed "s;$dir1;$dir2;"`
diff -u $dump $otherdump
done
Signed-off-by: Keith Packard <keithp@keithp.com>
Acked-by: Daniel Stone <daniel@fooishbar.org>
Acked-by: Alan Coopersmith <alan.coopersmith@oracle.com>
The constraining logic in RRConstrainCursorHarder allows the cursor to reach
crtc positions of x = width and y = height while the constraining code in
miPointerSetPosition only allows it to reach x = width - 1 and y = height - 1
for the analogous screen case.
This patch makes the former's logic equivalent to the latter's which allows
applications to benefit from Fitts's law. E.g. a maximized application
adjacent to a crtc border wouldn't get pointer events if the user moved the
pointer all the way until it's contained.
Signed-off-by: Rui Matos <tiagomatos@gmail.com>
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Keith Packard <keithp@keithp.com>
Swapping the wrong size was never caught because swap{l,s} are macros.
It's clear in the case of Xext/xres.c, that the author believed
client_major/minor to be CARD16 from looking at the code in the first
hunk.
v2: dmx.c fixes from Keith.
Reviewed-by: Peter Harris <pharris@opentext.com>
Signed-off-by: Matt Turner <mattst88@gmail.com>
Also, fix whitespace, mainly around
swaps(&rep.sequenceNumber)
Reviewed-by: Peter Harris <pharris@opentext.com>
Signed-off-by: Matt Turner <mattst88@gmail.com>
Providing an argument to return in a function with void return type
is not allowed by the C standard, and makes the Sun compilers unhappy.
Signed-off-by: Nicolas Kaiser <nikai@nikai.net>
Reviewed-by: Jeremy Huddleston <jeremyhu@apple.com>
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
Confine cursor motion to within the bounds of a single CRTC, iff all the
CRTCs within a ScreenRec are reachable from each other. If not you get
the same "cursor floats within the bounding rect" behaviour you get now.
v3:
- Incorporate review feedback from Christopher James Halse Rogers
v4:
- Add mode field.
Signed-off-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
If the client sends invalid data for this request, the server
will jump to 'sendReply' and call RRFreeCrtcConfigs, passing it the
uninitialized 'configs' and 'num_configs' values.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Aaron Plattner <aplattner@nvidia.com>
This provides for separate sizes for the screen scanout and rendering
buffer and the application-visible screen size.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Aaron Plattner <aplattner@nvidia.com>