Alex Deucher [Wed, 19 Mar 2008 19:37:56 +0000 (15:37 -0400)]
RADEON: switch over to new production microcode
This needs to be tested thoroughly before pushing to the
kernel.
Alex Deucher [Wed, 19 Mar 2008 18:57:42 +0000 (14:57 -0400)]
RADEON: production microcode for all radeons, r1xx-r6xx
This updated microcode is not in use yet.
Dave Airlie [Mon, 17 Mar 2008 23:07:45 +0000 (09:07 +1000)]
drm: add new rs690 pci id
Thomas Hellstrom [Mon, 17 Mar 2008 10:37:10 +0000 (11:37 +0100)]
Evict cached_mapped relocatee before applying reloc.
Fix that got left out after the intel-post-reloc merge.
Dave Airlie [Sun, 16 Mar 2008 21:05:46 +0000 (07:05 +1000)]
drm/pcigart: fix the pci gart to use the drm_pci wrapper.
This is the correct fix for the RS690 and hopefully the dma coherent work.
For now we limit everybody to a 32-bit DMA mask but it is possible for
RS690 to use a 40-bit DMA mask for the GART table itself,
and the PCIE cards can use 40-bits for the table entries.
Signed-off-by: Dave Airlie <airlied@redhat.com>
Thomas Hellstrom [Sun, 16 Mar 2008 19:07:14 +0000 (20:07 +0100)]
Avoid unnecessary waits for command regulator pause.
Thomas Hellstrom [Sun, 16 Mar 2008 10:44:35 +0000 (11:44 +0100)]
[via] Remove some leftover vars.
Thomas Hellstrom [Sun, 16 Mar 2008 10:39:18 +0000 (11:39 +0100)]
[via] Allow a little larger stride for SG DMA DownloadFromScreen.
Thomas Hellstrom [Sun, 16 Mar 2008 10:37:17 +0000 (11:37 +0100)]
[via] The millionth fixup for the millionth-1 attempt to stabilize the AGP
DMA command submission. It's worth remembering that all new bright ideas on how
to make this command reader work properly and according to docs
will probably fail :( Bring in some old code.
Thomas Hellstrom [Sun, 16 Mar 2008 10:29:57 +0000 (11:29 +0100)]
[via] Fix driver after vblank-rework merge.
Dave Airlie [Sun, 16 Mar 2008 05:01:27 +0000 (15:01 +1000)]
ati: fix rs690 igp gart by allocating the page table in 32-bit memory
Dave Airlie [Sun, 16 Mar 2008 04:00:16 +0000 (14:00 +1000)]
drm/rs690: set AGP_BASE_2 to 0
Dave Airlie [Sun, 16 Mar 2008 02:58:07 +0000 (12:58 +1000)]
drm: set rs690 gart base completly.
The docs state bits 4-11 represent bits 32-39 of a 40-bit address
Dave Airlie [Sun, 16 Mar 2008 02:56:11 +0000 (12:56 +1000)]
drm: this u32 should be a dma_addr_t
doesn't fix anything but just making it consistent
Alan Hourihane [Fri, 14 Mar 2008 00:25:42 +0000 (00:25 +0000)]
fix build problems
Mike Isely [Thu, 13 Mar 2008 23:53:05 +0000 (09:53 +1000)]
drm: Fix race that can lockup the kernel
The i915_vblank_swap() function schedules an automatic buffer swap
upon receipt of the vertical sync interrupt. Such an operation is
lengthy so it can't be allowed to happen in normal interrupt context,
thus the DRM implements this by scheduling the work in a kernel
softirq-scheduled tasklet. In order for the buffer swap to work
safely, the DRM's central lock must be taken, via a call to
drm_lock_take() located in drivers/char/drm/drm_irq.c within the
function drm_locked_tasklet_func(). The lock-taking logic uses a
non-interrupt-blocking spinlock to implement the manipulations needed
to take the lock. This semantic would be safe if all attempts to use
the spinlock only happen from process context. However this buffer
swap happens from softirq context which is really a form of interrupt
context. Thus we have an unsafe situation, in that
drm_locked_tasklet_func() can block on a spinlock already taken by a
thread in process context which will never get scheduled again because
of the blocked softirq tasklet. This wedges the kernel hard.
To trigger this bug, run a dual-head cloned mode configuration which
uses the i915 drm, then execute an opengl application which
synchronizes buffer swaps against the vertical sync interrupt. In my
testing, a lockup always results after running anywhere from 5 minutes
to an hour and a half. I believe dual-head is needed to really
trigger the problem because then the vertical sync interrupt handling
is no longer predictable (due to being interrupt-sourced from two
different heads running at different speeds). This raises the
probability of the tasklet trying to run while the userspace DRI is
doing things to the GPU (and manipulating the DRM lock).
The fix is to change the relevant spinlock semantics to be the
interrupt-blocking form. After this change I am no longer able to
trigger the lockup; the longest test run so far was 20 hours (test
stopped after that point).
Note: I have examined the places where this spinlock is being
employed; all are reasonably short bounded sequences and should be
suitable for interrupts being blocked without impacting overall kernel
interrupt response latency.
Signed-off-by: Mike Isely <isely@pobox.com>
Alex Deucher [Wed, 12 Mar 2008 15:16:12 +0000 (11:16 -0400)]
Fix chip family for RV550
Ben Skeggs [Wed, 12 Mar 2008 12:37:29 +0000 (23:37 +1100)]
nv50: force channel vram access through vm
If we ever want to be able to use the 3D engine we have no choice. It
appears that the tiling setup (required for 3D on G8x) is in the page tables.
The immediate benefit of this change however is that it's now not possible
for a client to use the GPU to render over the top of important engine setup
tables, which also live in VRAM.
G8x VRAM size is limited to 512MiB at the moment, as we use a 1-1 mapping
of real vram pages to their offset within the start of a channel's VRAM
DMA object and only populate a single PDE for VRAM use.
Thomas Hellstrom [Wed, 12 Mar 2008 10:34:29 +0000 (11:34 +0100)]
Merge branch 'intel-post-reloc'
Conflicts:
linux-core/drm_compat.c
linux-core/drm_compat.h
linux-core/drm_ttm.c
shared-core/i915_dma.c
Bump driver minor to 13 due to introduction of new
relocation type.
Thomas Hellstrom [Wed, 12 Mar 2008 09:36:40 +0000 (10:36 +0100)]
Bug # 14712
Disable page saving for GPU read-only TTMs.
Thomas Hellstrom [Wed, 12 Mar 2008 09:19:36 +0000 (10:19 +0100)]
Dont allow !sysadmin clients to alter the memory type of
NO_EVICT buffers.
Alan Hourihane [Thu, 17 Jan 2008 13:04:42 +0000 (13:04 +0000)]
Fix for debug memory routines
Alan Hourihane [Thu, 17 Jan 2008 13:04:23 +0000 (13:04 +0000)]
Add error message
Thomas Hellstrom [Thu, 17 Jan 2008 12:10:36 +0000 (13:10 +0100)]
Add an emergency pinnable memory quota for root-only processes.
Thomas Hellstrom [Wed, 12 Mar 2008 09:07:56 +0000 (10:07 +0100)]
Fix kernel crash when we hit OOM conditions.
(Alan Hourihane)
Thomas Hellstrom [Wed, 12 Mar 2008 09:02:09 +0000 (10:02 +0100)]
Avoid duplicate calls to drm_ttm_bind in some cases.
Thomas Hellstrom [Wed, 12 Mar 2008 08:56:06 +0000 (09:56 +0100)]
Make sure other TTM memory types than TT is really unbound when evicted.
Thomas Hellstrom [Thu, 28 Feb 2008 08:08:52 +0000 (09:08 +0100)]
Avoid large kmallocs.
Stuart Bennett [Tue, 11 Mar 2008 00:33:58 +0000 (00:33 +0000)]
nouveau: move AGP reset to mem_init_agp
Also, power cycle PGRAPH when resetting AGP -- it seems to fix problems encountered by p0g on nv25
Dave Airlie [Mon, 10 Mar 2008 08:30:17 +0000 (18:30 +1000)]
remove unneeded load call
Keith Packard [Thu, 24 Jan 2008 19:46:45 +0000 (11:46 -0800)]
Switch from PIPE_VBLANK to PIPE_EVENT interrupts.
My 965GM gets interrupts stuck when using the old PIPE_VBLANK interrupt.
Switch to the PIPE_EVENT interrupt mechanism, and set the PIPE*STAT
registers to use START_VBLANK on 965 and VBLANK on previous chips.
Dave Airlie [Fri, 7 Mar 2008 22:30:03 +0000 (08:30 +1000)]
drm/radeon: check sarea_priv exists
Ben Skeggs [Fri, 7 Mar 2008 04:08:59 +0000 (15:08 +1100)]
nouveau: redo channel idle detection
Will hopefully work a bit better than previous code, which depended on
knowing the channel's most recent PUT value. Some chips always return
0 on reading these regs, and currently userspace is the only other entity
which knows the value.
Ben Skeggs [Fri, 7 Mar 2008 03:38:05 +0000 (14:38 +1100)]
nouveau: don't touch NV_USER regs on channel destroy.
Not only was this entirely pointless, it actually causes my NV30GL to
die randomly when channels are destroyed.
Dave Airlie [Thu, 6 Mar 2008 23:29:35 +0000 (09:29 +1000)]
flush_agp_mappings commit
Dave Airlie [Thu, 6 Mar 2008 04:43:23 +0000 (14:43 +1000)]
drm/bo: allow non-suser priv to add kernel BOs.
modprobe can be run with dropped capabilities we still want the kernel bos
to work.
Benjamin Herrenschmidt [Wed, 5 Mar 2008 09:25:03 +0000 (19:25 +1000)]
drm: Fix for non-coherent DMA PowerPC
This patch fixes bits of the DRM so to make the radeon DRI work on
non-cache coherent PCI DMA variants of the PowerPC processors.
It moves the few places that needs change to wrappers to that
other architectures with similar issues can easily add their
own changes to those wrappers, at least until we have more useful
generic kernel API.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Dave Airlie [Mon, 18 Feb 2008 00:39:21 +0000 (10:39 +1000)]
ttm: make sure userspace can't destroy kernel create memory managers
this adds something to say the kernel initialised the memory region not
the userspace. and blocks userspace from deallocating kernel areas
Dave Airlie [Wed, 20 Feb 2008 03:27:10 +0000 (13:27 +1000)]
drm/ttm: add ioctl to get back memory managed area sized
taken from modesetting branch but could be useful outside it.
Dave Airlie [Wed, 5 Mar 2008 19:21:50 +0000 (05:21 +1000)]
drm: reorganise minor number handling using code from modesetting branch
Rip out the whole head thing and replace it with an idr and drm_minor
structure.
Xiang, Haihao [Wed, 5 Mar 2008 07:08:46 +0000 (15:08 +0800)]
i915: Evict if relocatee buffer is CACHED_MAPPED before
writting relocations, otherwise the GPU probably sees some
inconsistent data. Fix fd.o bug#14656
Eric Anholt [Tue, 4 Mar 2008 21:45:41 +0000 (13:45 -0800)]
Clarify when WAIT_LAZY is relevant to users.
Eric Anholt [Thu, 31 Jan 2008 03:02:56 +0000 (19:02 -0800)]
Remove unused DRM_FENCE_FLAG_WAIT_IGNORE_SIGNALS.
Eric Anholt [Tue, 4 Mar 2008 21:35:23 +0000 (13:35 -0800)]
Clarify through the names what drm_ttm_alloc_pages() and friend actually did.
These are all about the page directory (pointers to pages) rather than the
actual pages backing the allocation.
Eric Anholt [Tue, 4 Mar 2008 20:16:51 +0000 (12:16 -0800)]
Don't shortcut the info syscall for drmBOBusy on nonshareable objects.
This broke the results when you're trying to check if a buffer you dispatched
some time ago is done being rendered from.
Zou Nan hai [Mon, 3 Mar 2008 06:49:49 +0000 (14:49 +0800)]
[i915] 2D driver may reset Frame count value, this may lead driver
to leap it's vblank count a huge value.
This will stall some applications that switch video mode if vblank_mode is set to a non zero value in drirc.
Patrice Mandin [Fri, 29 Feb 2008 20:57:40 +0000 (21:57 +0100)]
FIX_KMAP_BEGIN requires CONFIG_HIMEM (see include/asm-i386.h/fixmap.h)
Thomas Hellstrom [Fri, 29 Feb 2008 14:38:55 +0000 (15:38 +0100)]
Working revision.
Thomas Hellstrom [Fri, 29 Feb 2008 12:31:14 +0000 (13:31 +0100)]
Fix compilation breakage on x86-64.
Thomas Hellstrom [Fri, 29 Feb 2008 12:25:55 +0000 (13:25 +0100)]
More post-ioctl work.
Dave Airlie [Fri, 29 Feb 2008 00:16:24 +0000 (10:16 +1000)]
agp: export the correct symbol
Thomas Hellstrom [Thu, 28 Feb 2008 12:47:15 +0000 (13:47 +0100)]
Add a compat kmap_atomic_prot_pfn to do quick kernel map / unmaps of
PCI- or high memory.
This is substantially more efficient than drm_bo_kmap,
since the mapping only lives on a single processor.
Unmapping is done use kunmap_atomic(). Flushes only a single tlb() entry.
Add a support utility int drm_bo_pfn_prot() that returns the
pfn and desired page protection for a given bo offset.
This is all intended for relocations in bound TTMS or vram.
Mapping-accessing-unmapping must be atomic, either using preempt_xx() macros
or a spinlock.
Thomas Hellstrom [Thu, 28 Feb 2008 13:05:53 +0000 (14:05 +0100)]
Initial commit.
Thomas Hellstrom [Thu, 28 Feb 2008 12:47:15 +0000 (13:47 +0100)]
Add a compat kmap_atomic_prot_pfn to do quick kernel map / unmaps of
PCI- or high memory.
This is substantially more efficient than drm_bo_kmap,
since the mapping only lives on a single processor.
Unmapping is done use kunmap_atomic(). Flushes only a single tlb() entry.
Add a support utility int drm_bo_pfn_prot() that returns the
pfn and desired page protection for a given bo offset.
This is all intended for relocations in bound TTMS or vram.
Mapping-accessing-unmapping must be atomic, either using preempt_xx() macros
or a spinlock.
Thomas Hellstrom [Wed, 27 Feb 2008 20:44:40 +0000 (21:44 +0100)]
Reinstate buffer idle before applying relocations.
Thomas Hellstrom [Wed, 27 Feb 2008 18:46:28 +0000 (19:46 +0100)]
Don't wait for buffer idle before applying relocations.
Thomas Hellstrom [Tue, 26 Feb 2008 09:47:05 +0000 (10:47 +0100)]
[i915] Relocation fixes.
Thomas Hellstrom [Mon, 25 Feb 2008 23:01:09 +0000 (00:01 +0100)]
Make the execbuffer code reasonably safe against errors.
In particular -EAGAINs, which should be common during Xserver operation.
Also handle the fence creation failure case.
Roland Scheidegger [Sat, 23 Feb 2008 10:01:36 +0000 (11:01 +0100)]
fix texture uploads with large 3d textures (bug 13980)
Texture uploads could hit the blitter coordinate limit, adjust the texture
offset when uploading the pieces. Make sure to check the end address of the
upload too.
Jesse Barnes [Fri, 22 Feb 2008 21:57:38 +0000 (13:57 -0800)]
i915: put ARX back into index mode before doing restore
Fixes resume from hibernate in some configurations.
Maarten Maathuis [Fri, 22 Feb 2008 12:41:05 +0000 (13:41 +0100)]
nouveau: Remove some random (french) comment.
Maarten Maathuis [Fri, 22 Feb 2008 12:35:13 +0000 (13:35 +0100)]
nouveau: A single define of dma skips is more than enough.
Kristian Høgsberg [Fri, 22 Feb 2008 16:22:52 +0000 (11:22 -0500)]
Fix one last occurance of struct _drm_i915_batchbuffer.
Thanks to Todd Merrill for pointing it out.
Kristian Høgsberg [Fri, 22 Feb 2008 05:12:39 +0000 (00:12 -0500)]
i915: Remove leading underscore from struct tags.
This matches the changes in mesa to use the system drm includes
for the definitions of the drm ioctl structs.
Kristian Høgsberg [Thu, 21 Feb 2008 20:31:48 +0000 (15:31 -0500)]
Don't free irq resources until after we've unregistered the handler.
Michel Dänzer [Thu, 21 Feb 2008 10:49:38 +0000 (11:49 +0100)]
linux: Clean up vblank related resources from drm_irq_uninstall().
This fixes at least two problems:
* The vblank_disable_fn timer callback could get called after the DRM was
de-initialized, e.g. on X server shutdown.
* Leak of vblank related resources when disabling and re-enabling the IRQ, e.g.
on an X server reset.
Alan Hourihane [Wed, 20 Feb 2008 22:22:49 +0000 (22:22 +0000)]
fix SAREA
Dave Airlie [Wed, 20 Feb 2008 00:02:20 +0000 (10:02 +1000)]
drm: add support for passing state into the suspend hooks.
fix i915 driver to use state for hibernate save avoidance.
Signed-off-by: Dave Airlie <airlied@redhat.com>
Keith Packard [Sun, 17 Feb 2008 03:19:29 +0000 (19:19 -0800)]
[915]: more registers for S3 (DSPCLK_GATE_D, CACHE_MODE_0, MI_ARB_STATE)
Failing to preserve the MI_ARB_STATE register was causing FIFO underruns on
the VGA output on my HP 2510p after resume.
Stephane Marchesin [Sat, 16 Feb 2008 02:50:10 +0000 (03:50 +0100)]
nouveau: no GART on ia64 either.
Ben Skeggs [Fri, 15 Feb 2008 17:33:27 +0000 (04:33 +1100)]
nv40: actually init all tile regs.
Kristian Høgsberg [Tue, 5 Feb 2008 18:27:16 +0000 (13:27 -0500)]
i915: Add a dri2 init path that gets the lock from the dri2 sarea.
Kristian Høgsberg [Tue, 12 Feb 2008 21:08:18 +0000 (16:08 -0500)]
i915: Only look up dev_priv->mmio_map if it's not already set up
Kristian Høgsberg [Tue, 5 Feb 2008 17:27:48 +0000 (12:27 -0500)]
i915: Add I915_PARAM_CHIPSET_ID param to get chipset ID.
Kristian Høgsberg [Tue, 5 Feb 2008 17:25:22 +0000 (12:25 -0500)]
i915: Make sarea_priv setup optional.
Jesse Barnes [Fri, 8 Feb 2008 01:33:28 +0000 (17:33 -0800)]
Restore pipeconf regs unconditionally
On many chipsets, the checks for DPLL enable or VGA mode will prevent the
pipeconf regs from being restored, which could result in a blank display or X
failing to come back after resume. So restore them unconditionally along with
actually restoring pipe B's palette correctly.
Jesse Barnes [Thu, 7 Feb 2008 19:21:09 +0000 (11:21 -0800)]
Fix saveGR array size
Make sure we have enough room for all the GR registers or we'll end up
clobbering the AR index register (which should actually be harmless
unless the BIOS is making an assumption about it).
Jesse Barnes [Thu, 7 Feb 2008 18:48:08 +0000 (10:48 -0800)]
i915: save/restore interrupt state
On resume, if the interrupt state isn't restored correctly, we may end
up with a flood of unexpected or ill-timed interrupts, which could cause
the kernel to disable the interrupt or vblank events to happen at the
wrong time. So save/restore them properly.
Jesse Barnes [Thu, 7 Feb 2008 18:40:06 +0000 (10:40 -0800)]
Fix vblank enable/disable callbacks
There were two problems with the existing callback code: the vblank
enable callback happened multiple times per disable, making drivers more
complex than they had to be, and there was a race between the final
decrement of the vblank usage counter and the next enable call, which
could have resulted in a put->schedule disable->get->enable->disable
sequence, which would be bad.
So add a new vblank_enabled array to track vblank enable on per-pipe
basis, and add a lock to protect it along with the refcount +
enable/disable calls to fix the race.
Thomas Hellstrom [Tue, 5 Feb 2008 09:35:56 +0000 (10:35 +0100)]
i915: Re-report breadcrumbs on poll to the fence manager,
since a breadcrumb may actually turn up before a corresponding fence object
has been placed on the fence ring.
Stuart Bennett [Mon, 28 Jan 2008 22:59:26 +0000 (22:59 +0000)]
nouveau: make nv34 work every time, not just every 2nd time
And make nv30_graph_init a bit more like mmio-traces
Maarten Maathuis [Sat, 2 Feb 2008 11:46:31 +0000 (12:46 +0100)]
nouveau: NV40 can/should now be able to run after the blob.
- Moved the fix from the ddx to drm, because it seemed more appropriate.
- Don't be shy, report if it works for you or not.
Thomas Hellstrom [Thu, 31 Jan 2008 13:11:12 +0000 (14:11 +0100)]
Add an fence_class_manager::last_queued_sequence member, since a
sequence number may actually turn up before the corresponding fence
object has been queued on the ring.
Fence drivers can use this member to determine whether a
sequence number must be re-reported.
Thomas Hellstrom [Wed, 30 Jan 2008 21:14:02 +0000 (22:14 +0100)]
i915: Avoid calling drm_fence_flush_old excessively.
Thomas Hellstrom [Wed, 30 Jan 2008 21:06:02 +0000 (22:06 +0100)]
Simplify the fencing code and differentiate between flushes and
waiting types.
Add a "command_stream_barrier" method to the bo driver.
Ben Skeggs [Wed, 30 Jan 2008 00:40:13 +0000 (11:40 +1100)]
nv40: some more nv67 changes
With some luck the drm-side will be OK now for this chipset.
Mirko [Tue, 29 Jan 2008 15:11:27 +0000 (10:11 -0500)]
Add new RV380 pci id
bug 14289
Jesse Barnes [Tue, 29 Jan 2008 05:05:22 +0000 (21:05 -0800)]
Fix hibernate save/restore of VGA attribute regs
In hibernate, we may end up calling the VGA save regs function twice, so we
need to make sure it's idempotent. That means leaving ARX in index mode after
the first save operation. Fixes hibernate on 965.
Maciej Cencora [Sun, 27 Jan 2008 02:50:31 +0000 (12:50 +1000)]
drm: add initial rs690 support for drm.
This adds support for configuring the RS690 GART.
George Sapountzis [Fri, 25 Jan 2008 14:54:29 +0000 (16:54 +0200)]
mach64: fix after vblank-rework
don't disable vblank interrupts (similar to r128)
Jesse Barnes [Fri, 25 Jan 2008 04:59:51 +0000 (20:59 -0800)]
Fixup modeset ioctl number & typedef usage
Should be 0x08 rather than 0xa0, and shouldn't use typedefs.
Eric Anholt [Thu, 24 Jan 2008 20:32:08 +0000 (12:32 -0800)]
Merge commit 'airlied/i915-ttm-cfu'
This requires updated Mesa to handle the new relocation format.
Jesse Barnes [Thu, 24 Jan 2008 16:57:04 +0000 (08:57 -0800)]
Remove broken 'in vblank' accounting
We need to return an accurate vblank count to the callers of
->get_vblank_counter, and in the Intel case the actual frame count
register isn't udpated until the next active line is displayed, so we
need to return one more than the frame count register if we're currently
in a vblank period.
However, none of the various ways of doing this is working yet, so
disable the logic for now. This may result in a few missed events, but
should fix the hangs some people have seen due to the current code
tripping the wraparound logic in drm_update_vblank_count.
Dave Airlie [Thu, 24 Jan 2008 05:18:09 +0000 (15:18 +1000)]
i915: fix missing header when copying data from userspace
Dave Airlie [Thu, 24 Jan 2008 04:37:40 +0000 (14:37 +1000)]
i915 make relocs use copy from user
Switch relocs to using copy from user and remove index and pass buffer
handles in instead.
Jesse Barnes [Wed, 23 Jan 2008 16:39:57 +0000 (08:39 -0800)]
Fix thinko in get_vblank_counter
Should use vtotal not htotal to figure out if we're in a vblank period.
Jesse Barnes [Wed, 23 Jan 2008 16:38:01 +0000 (08:38 -0800)]
Fix IS_I915G macro
One to many parantheses...
Maarten Maathuis [Wed, 23 Jan 2008 15:40:19 +0000 (16:40 +0100)]
nouveau: Fix warning in nouveau_mem.c
Dave Airlie [Wed, 23 Jan 2008 06:45:09 +0000 (16:45 +1000)]
i915/flush: get the ret the right way around
Dave Airlie [Wed, 23 Jan 2008 06:44:51 +0000 (16:44 +1000)]
drm/i915: add support for E7221