summaryrefslogtreecommitdiff
AgeCommit message (Collapse)AuthorFilesLines
2010-01-08build and starts on F12 with these hacks.f12-hackupDave Airlie1-4/+8
2010-01-08qxl: make build against later serversDave Airlie1-0/+2
2009-12-23qxl: make default be 1024x768Dave Airlie1-0/+16
This change lets us validate a lot of modes and if the user doesn't supply a virtual line they'll get 1024x768, because we have no monitor, its hard to know what the best thing to do in this case. To get larger modes add a Virtual line to xorg.conf
2009-12-23qxl: add sanity check to make sure we aren't back in VGA mode.Dave Airlie1-0/+12
For some reason we still sometimes end up back in VGA mode, I suspect a kernel VGA access somewhere while under KD_GRAPHICS, on real HW it would ignore this, qxl seems to "unique" and not in a good way.
2009-12-22qxl: fix access to qxl ringDave Airlie1-8/+8
This really needs to be volatile as the hw will change it underneath us. Really we should be using proper MMIO accessors for this stuff not casting structs to mmio memory.
2009-12-22qxl: fix 2 trivial cosmetic issuesDave Airlie1-4/+4
1. use correct values for unmap vid mem BAR sizes, removes warning in logs. 2. print values from ROM not uninitialised values from qxl struct.
2009-11-07Bump versionSøren Sandmann Pedersen1-1/+1
2009-11-07Avoid submitting commands when device is in VGA mode.Søren Sandmann Pedersen2-8/+27
When someone runs "init 3", the device will be switched into VGA mode with no notification to the driver. However, if commands are submitted when the device is in VGA mode, they will be queued up, and then, the next time a mode set set, an assertion in the device will take down the entire virtual machine. So this patch checks whether the device is in VGA mode before submitting each command. This is racy because the switch to VGA could happen just after we check, but there doesn't seem to be a better way. (Apart from removing the assert() in the QXL device itself thereby making it robust against command submission in VGA mode. But the author of the device is opposed to that).
2009-11-06Do not wait for idle before mode setSøren Sandmann Pedersen1-2/+0
2009-07-15Reset before setting modeSøren Sandmann Pedersen1-0/+2
2009-07-01Eliminate a couple of warningsSøren Sandmann Pedersen2-2/+4
2009-07-01Add lookup3.h to Makefile.amSøren Sandmann Pedersen1-0/+1
2009-07-01Bump version number and include stdint.hSøren Sandmann Pedersen2-1/+3
2009-07-01Comment out some debugging spamSøren Sandmann Pedersen2-3/+15
2009-06-30Update vendor IDSøren Sandmann Pedersen2-3/+3
2009-04-27Remove mmtest.c test programSøren Sandmann Pedersen2-79/+0
2009-04-09Dump display tree on OOMSøren Sandmann Pedersen3-0/+48
2009-04-09Use a linked hash table implementations; use the Jenkins hash functionSøren Sandmann Pedersen5-97/+884
2009-04-08Move hash table back into qxl_image.cSøren Sandmann Pedersen2-115/+152
2009-04-08Beginning of new hash table implementationSøren Sandmann Pedersen1-10/+38
2009-03-31Bump versionSøren Sandmann Pedersen1-1/+1
2009-03-31Initial naive image caching implementation.Søren Sandmann Pedersen3-48/+149
It attempts to cache all image puts without exception; the id is produced by hashing and copying every time.
2009-03-29Compute hash codes for imagesSøren Sandmann Pedersen4-24/+57
2009-03-29Formatting qxl.h, add qxl_image_destroy()Søren Sandmann Pedersen3-48/+63
2009-03-29TODO; beginning of image hashing; move virtual/phsyical functions to qxl.hSøren Sandmann Pedersen6-89/+111
2009-03-19Merge branch 'master' of ↵Søren Sandmann Pedersen1-6/+22
git+ssh://sandmann@shell.devel.redhat.com/home/boston/sandmann/public_git/xf86-video-qxl
2009-03-19Cast-fest to get rid of void pointer arithmeticSøren Sandmann Pedersen1-7/+7
2009-03-19Add pointer arithmetic warnings back inSøren Sandmann Pedersen1-1/+1
2009-03-18TODOSøren Sandmann Pedersen1-6/+22
2009-03-12TODOSøren Sandmann Pedersen1-2/+46
2009-03-06Reenable hardware cursorsSøren Sandmann Pedersen1-2/+0
2009-03-06Copy bits in qxlCopyNtoN rather than relying on a separateSøren Sandmann Pedersen1-27/+23
fbCopyArea() to do it This ensures that miHandleExposures() which get called at the end of fbDoCopy(), will be called with the shadow framebuffer in a sane state. This fixes a number of issues with scrolling partially covered windows.
2009-03-06Fix qxl_sleep(); make CopyArea deal better with Paintwindow damageSøren Sandmann Pedersen1-11/+68
2009-03-04Add hardware cursor supportSøren Sandmann Pedersen4-47/+148
2009-03-03Remove #if 0'd codeSøren Sandmann Pedersen1-265/+2
2009-03-03Add cursor structs; delete some unused variablesSøren Sandmann Pedersen2-65/+30
2009-02-27Bump version numberSøren Sandmann Pedersen1-1/+1
2009-02-27Use window offsets rather than pixmap offsets, among other thingsSøren Sandmann Pedersen1-65/+234
2009-02-27Debug spamSøren Sandmann Pedersen1-3/+46
2009-02-26Add copy acceleration; use the right fill colorsSøren Sandmann Pedersen2-5/+142
2009-02-26Get the offsets closer to rightSøren Sandmann Pedersen2-20/+54
2009-02-26Accelerate window background paintingSøren Sandmann Pedersen2-0/+56
2009-02-26Wrap PolyFillRectsSøren Sandmann Pedersen2-17/+59
2009-02-26Wrap CreateGCSøren Sandmann Pedersen2-6/+43
2009-02-26TODOSøren Sandmann Pedersen1-0/+21
2009-02-25Don't use shadow framebuffer; use our own damage tracker instead.Søren Sandmann Pedersen2-41/+54
2009-02-25TODOSøren Sandmann Pedersen3-5/+43
2009-02-24Set version number to 0.0.1Søren Sandmann Pedersen1-1/+1
2009-02-24Wait for the command ring to go idle before setting the modeSøren Sandmann Pedersen3-12/+43
Set the correct stride in the ModifyPixmapHeader call, and onlhy call qxl_mem_free_all() if the memory system is initialized.
2009-02-24Squash compiler warningsSøren Sandmann Pedersen2-45/+11
As part of this, don't use -Wpointer-arith
generated by cgit v1.2.3 (git 2.39.1) at 2024-06-11 15:14:52 +0000