Commit Graph

3409 Commits (3e03d781f7c41a88d5d5f895be9c443bf3592ef0)

Author SHA1 Message Date
Eric Anholt 3e03d781f7 intel: Avoid pthread mutex recursion in bufmgr_fake.
Bug #18035. Fixes deadlock in glean texCube testcase.
2008-10-13 13:41:10 -07:00
Dave Airlie c6109df93b libdrm: don't depend or link to libdrm_intel 2008-10-13 07:16:33 +10:00
Robert Noland f5327aca0c [FreeBSD] Plug memory leak in drm_rmdraw() and drm_drawable_free_all() 2008-10-10 18:23:11 -04:00
Robert Noland cdd3e9fc56 [FreeBSD] Rework all of the memory allocations
Allocate memory from different pools.  This allows the OS to track memory
allocations for us, much like the linux memory debugging.  This will ease
tracking down memory leaks since the OS can track the number of allocations
from each pool and help to point us in the right direction.  Also replace
drm_alloc and friends with static __inline__ versions while we are here.
2008-10-10 13:06:22 -04:00
Robert Noland 1150a42d43 [FreeBSD] Fix linux list compat list_for_each_safe()
linux_for_each_safe would not handle lists with a single entry.
2008-10-09 22:13:26 -04:00
Robert Noland a8f73c214d i915: Cleanup interrupt handling 2008-10-09 22:11:55 -04:00
Keith Packard 94c88c151c Manage fences in user-mode bufmgr_fake to clean buffers
When using bufmgr_fake without DRM, the X server idles the ring whenever it
wants to wait for something to complete (brutal, but effective). In this
case, bufmgr_fake must treat the pending fence as having passed. However, it
wasn't recording the fences as it emitted them, nor cleaning buffers as they
passed.

Signed-off-by: Keith Packard <keithp@keithp.com>
2008-10-09 18:25:58 -07:00
Eric Anholt 3169d9639b intel: Protect bufmgr objects with a pthread mutex.
We want to be able to use the bufmgr from multiple threads for GL, and thus
we need to protect the internal structures.

The pthread-stubs package is used so that programs not linked against
pthreads get weak symbols to stubs and don't eat most of the cost.
2008-10-09 12:57:09 -07:00
Xiang, Haihao 604759d4a7 intel: fix for write_domain and static BOs.
http://bugs.freedesktop.org/show_bug.cgi?id=17705
2008-10-09 11:59:29 +08:00
Alex Deucher 728d8e226f radeon: add comment to clarify bus mastering on PCIE chips 2008-10-06 12:12:49 -04:00
Alex Deucher 6f9dfa098f radeon: fix duplicate define in my last commit
That's what I get for committing at 3 AM.
2008-10-06 12:01:11 -04:00
Dave Airlie 8e5f5ed189 radeon: PCIE cards don't appear to have explicit bus master 2008-10-07 04:47:54 +10:00
Dave Airlie d0aff12052 drm: add create gpu tree script 2008-10-07 04:07:40 +10:00
Alex Deucher 4b98f6d74f radeon: fix bus master enabled bits on newer asics 2008-10-06 03:08:27 -04:00
Mihail Zenkov 5a36cce349 radeon: fix missing bit from rs740 patch
See bug 17908
2008-10-06 00:49:15 -04:00
Alex Deucher ce40261012 radeon: Add support for HD2100 IGP (RS740) 2008-10-04 20:43:21 -04:00
Robert Noland 60cf3a4db4 [FreeBSD] Don't explicitly bzero driver softc.
This is already handled for us.

Suggested by John Baldwin
2008-10-03 14:11:20 -04:00
Robert Noland 9c0ce38df3 [FreeBSD] Use M_WAITOK when allocating driver memory.
We don't explicitly check for error here and M_WAITOK will just put the
process to sleep waiting on resources to become available.

Suggested by John Baldwin
2008-10-03 14:05:45 -04:00
Robert Noland 4c92abfa8d [FreeBSD] Do a bit of optimization on drm_order() 2008-10-03 13:56:50 -04:00
Robert Noland 81952c7dd1 Use devfs_get_cdevpriv in mmap as well.
d_mmap gets called twice and we are only able to associate the file_priv
during the first call.  The second call will return EBADF and we need to
assume that the call was succesful.  d_mmap will not tolerate having an
error returned for the second call.
2008-10-01 20:49:03 -04:00
Xiang, Haihao 073cb5ee1d intel: Copy data from card memory back to backing store when mapping.
Fixes http://bugs.freedesktop.org/show_bug.cgi?id=17705
2008-09-27 11:01:24 +08:00
Eric Anholt 2db8e0c8ef intel: Allow up to 15 seconds chewing on one buffer before acknowledging -EBUSY.
The gltestperf demo in some cases took over seven seconds to make it through
one batchbuffer on a GM965.

Bug #17004.
2008-09-23 17:10:04 -07:00
Eric Anholt 0dccf017ab intel: Replace wraparound test logic in bufmgr_fake. Again.
I'd swapped the operands, so if we weren't in lockstep with the hardware we
said the sequence was always passed.  Additionally, a race was available that
we might have failed at recovering from.  Instead, I've replaced the logic
with new stuff that should be more robust and not rely on all the parties in
userland following the same IRQ_EMIT() == 1 protocol.  Also, in a radical
departure from past efforts, include a long comment describing the failure
modes and how we're working around them.

Thanks to haihao for catching the original issue.
2008-09-23 17:10:04 -07:00
Eric Anholt 1b3abe62b5 intel: Do strerror on errno, not on the -1 return value from ioctl. 2008-09-23 17:10:04 -07:00
Xiang, Haihao 3949f3c9ea intel: Fix driver-supplied argument to exec function (fd.o bug #17653). 2008-09-22 10:16:19 +08:00
Robert Noland 8ca06eb492 [FreeBSD] Convert to using cdevpriv for file_priv tracking 2008-09-17 23:15:08 -04:00
Ben Skeggs ee6bcabc50 nv50: add initial context for chipset 0xaa
This just doesn't look right..
2008-09-17 22:18:03 +10:00
Ben Skeggs d55e8090fa nv50: add initial context to match ctxprog for chipset 0x50 2008-09-17 22:03:38 +10:00
Ben Skeggs 4d2f1257fa nv50: add ctxprog for chipset 0x50 2008-09-17 15:13:27 +10:00
Ben Skeggs 301be1dc9b nv50: add ctxprog for chipset 0xaa 2008-09-17 15:02:54 +10:00
Ben Skeggs f152482bde nv50: add support for chipset 0x92 2008-09-17 14:52:22 +10:00
vehemens 973c634eaa Remove incomplete and obsolete free/net/open code.
Signed-off-by: Robert Noland <rnoland@2hip.net>
2008-09-10 23:35:10 -04:00
Eric Anholt b54d15d8fd Add missing \ to libdrm_la_SOURCES
Reported by jcristau.
2008-09-10 20:05:02 -07:00
Eric Anholt 368b392e6d intel: don't forget to include config.h in bufmgr code.
Thanks to airlied for catching this.
2008-09-10 14:07:19 -07:00
Eric Anholt f9d98beefc intel: move drm calls to exec buffers to libdrm_intel.
This avoids duplicating the effort in 3 places.  Also, added emit/wait fence
callbacks back in bufmgr_fake since we need it for non-drm 2d.  Sigh.
2008-09-10 14:07:18 -07:00
Eric Anholt 869d8bebed intel: Move IRQ emit/wait from callbacks into the bufmgr.
In the process, work around the glaring bugs of the kernel irq wait function.
2008-09-10 14:07:18 -07:00
Eric Anholt 738e36acbc Move intel libdrm stuff to libdrm_intel.so
dri_bufmgr.h is replaced by intel_bufmgr.h, and several functions are renamed,
though the structures and many functions remain dri_bufmgr_* and dri_bo_*
2008-09-10 14:07:18 -07:00
Eric Anholt 09cf0f0213 drm: Add tests for GEM_FLINK ioctl. 2008-09-10 14:07:18 -07:00
Robert Noland 828ae3f6b8 [FreeBSD] We need to call drm_detach before we free dev->driver.
The driver is in control of the show, so when you try and unload a module
the driver detach routine is called first.  It is what drives the whole
unload process and so lots of panics occur if dev->driver is already
free.
2008-09-08 16:40:52 -04:00
Robert Noland 2880c86eb2 [FreeBSD] Implement drm_ioremap_wc() to set a range of mem to write-combining 2008-09-07 12:44:02 -04:00
Robert Noland 740f09bffd [FreeBSD] IGP gart needs to be un-cached.
Airlied inadvertently discovered that the IGP gart needs to be un-cached
for radeon rs485 and rs690 to work.  Initial tests by placing a wbinvd()
after allocating the gart were successful.  This is an attempt at a more
appropriate method of achieving success.
2008-09-06 21:08:33 -04:00
vehemens be5fad45ee Free temp_pagelist on error. Free in reverse order. Noticed by open.
Signed-off-by: Robert Noland <rnoland@2hip.net>
2008-09-06 21:07:46 -04:00
vehemens 9ad5a6d0d7 Pass lock data like linux and open.
Signed-off-by: Robert Noland <rnoland@2hip.net>
2008-09-06 18:55:03 -04:00
vehemens b8a9cebddc Move order to end like linux.
Signed-off-by: Robert Noland <rnoland@2hip.net>
2008-09-06 18:55:03 -04:00
vehemens 0808cf923d Style white space cleanup part 2.
Signed-off-by: Robert Noland <rnoland@2hip.net>
2008-09-06 18:55:03 -04:00
Robert Noland 6f2479c674 [FreeBSD] Ensure that drm_pci_alloc is never called while locks are held. 2008-09-06 18:37:06 -04:00
vehemens 31709aa2be Reorder lock functions like linux.
Signed-off-by: Robert Noland <rnoland@2hip.net>
2008-09-05 12:42:41 -04:00
vehemens 76dd74c64e Style white space cleanup.
Signed-off-by: Robert Noland <rnoland@2hip.net>
2008-09-05 12:42:41 -04:00
vehemens ed6dd03818 Need M_NOWAIT for malloc.
Signed-off-by: Robert Noland <rnoland@2hip.net>
2008-09-05 12:38:41 -04:00
Robert Noland 3f915f68e8 i915: fix i915_ring_validate() 2008-09-05 12:38:41 -04:00