Age | Commit message (Collapse) | Author | Files | Lines |
|
The function was an empty since 2008 at least. Remove it alongside the
unused GLX visuals code.
With this all the GL/GLX dependencies in the driver are gone.
Note: the driver currently does not build if XAA is missing
Cc: Matthias Hopf <mhopf@suse.de>
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
---
Analogous to the earlier VIA one - patch is for posterity.
Feel free to pick or ignore.
|
|
Found by Jens Lody.
|
|
Signed-off-by: Matthias Hopf <mhopf@suse.de>
|
|
Date: Tue, 06 Apr 2010 10:37:18 +0200
From: Michel Dänzer <michel@daenzer.net>
Subject: Re: xserver 1.8.0 radeonhd crashing
Most likely pSrcPicture->pDrawable is NULL (i.e. it's a source-only,
solid or gradient picture) in R300CheckComposite(), and that fails to
check the pointer before dereferencing it.
|
|
|
|
|
|
This doesn't completely conform with the notion that probe() should return
with all resources freed. However, resources are outside the Xserver in this
case.
X configuration is unknown at probe time. Thus this cannot be disabled even
with option nodri.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
This issue was pointed out to us by Peter Blum.
|
|
|
|
|
|
If kernel modesetting (KMS) is detected and active, we abort the
PciProbe and do not touch the GPU in any way. Then at least the KMS
console continues working, and there is an error message for the user
to read.
This is only temporary until KMS has been added to this driver.
|
|
Signed-off-by: Alan Coopersmith <alan.coopersmith@sun.com>
|
|
|
|
The radeonhd matching table has a incorrect LIST_END macro.
This causes xorg -configure to do silly things like configure a radeonhd entry
for every PCI function in the system.
Signed-off-by: Matthias Hopf <mhopf@suse.de>
|
|
Should fix assert in rhd_dig.c and removes RHDHdmiCommitAudioWorkaround,
from AllocFree becuase it's later called anyway.
Signed-off-by: Matthias Hopf <mhopf@suse.de>
|
|
This patch adds audio support for the RV710 and RV730 chipsets.
A special "Thank you" goes to Andrew Wasielewski, for finding the
0x7604 register dependency, submitting a bunch of logs and testing
really a lot of different patches until we figured this out completely.
Signed-off-by: Matthias Hopf <mhopf@suse.de>
|
|
|
|
|
|
Reported by Andreas Bittner.
|
|
|
|
In some AtomBIOSes voltage values are broken (e.g. 0xff01).
|
|
|
|
|
|
Reported by Dmitry Rybin.
|
|
Some VoltageInfo tables seem to be inconsistent. Check sizes
so that we don't accidentally read beyond the end of the BIOS.
|
|
Some Voltage Object tables contain a size of 0. If this isn't caught
the driver will hang in an endless loop.
|
|
|
|
Don't read out or set pipe parameters when engine is active.
Doing so might lock up the engine.
This fixes a long standing issue with 2D acceleration on R5xx.
|
|
two contexts.
|
|
The Radeon driver attempts to save/restore the hw scaler registers
by reading/writing the range of registers which address the video scaler.
While doing so it also touches the TAP Control index/data registers.
Uncontrolled touching of the data register seems to have an adverse
effect on the ability to run HW supported I2C read out. This running
RadeonHD after running the Radeon driver results in no DDC data.
The reason for this has not been understood however it has been found
experimentally that reading back the scaler data register is able to
'fix' this issue.
We put this hacky workaround in the code until we obtain a better
understanding of the issue.
|
|
AtomBIOS takes in a TV mode to determine the correct multi tab
values. Although TV mode support has not been completed this patch
adds support for this.
It also replaces scaler register values by their symbolic names.
|
|
|
|
This patch adds support for FreeBSD and DragonFlyBSD via acpi_video(4).
Signed-off-by: Egbert Eich <eich@freedesktop.org>
|
|
Mode line sanitization may change the HSync and VRefresh values.
Therefore it needs to be done before the Panel H/V ranges are
determined.
Also report if a mode had to be sanitized.
|
|
|
|
|
|
There are two methods of doing DAC load detection on DCE3.x chips: one
works 'on the fly' ie doesn't produce noticeable flickering but
doesn't work as reliably on some chips RS780 especially in connection
with a KVM switch where it may produce a false negative, the other may
produce some visible artefacts but also seems to work reliably with KVM
switches. It's the one used in most AtomBIOSes.
We now use both methods: if the 'on the fly' one produces a negative
result we fall back to the other one and let it decide wether something
is connected or not.
|
|
|
|
The correct value was not read out as 'Set' was called instead of 'Get'.
|
|
rs880 needs to be treated as pcie just like rs780.
Fixes DRI init.
|
|
Check was introduced by 69d4dabe27aba0ed6aff52d210b045876ed36fce,
supposedly to work around an old Ubuntu bug. We use GL/glxtokens.h and
GL/glxint.h but not GL/gl.h, which is not part of glproto anyway.
Removing the check. If this end up causing problems, we can add
explicit checks for the proper headers.
|
|
|
|
On pre-R6xx the DDC clock is derived from the engine clock, thus changing the
engine clock implies that the DDC clock is to be updated as well.
|