Commit graph

53 commits

Author SHA1 Message Date
Francisco Jerez
4e03b4af6d drm/nouveau: Fix pushbufs over the 4GB mark.
Signed-off-by: Francisco Jerez <currojerez@riseup.net>
Tested-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2011-12-21 19:01:46 +10:00
Marcin Slusarz
5e60ee780e drm/nouveau: initialize chan->fence.lock before use
Fence lock needs to be initialized before any call to nouveau_channel_put
because it calls nouveau_channel_idle->nouveau_fence_update which uses
fence lock.

BUG: spinlock bad magic on CPU#0, test/24134
 lock: ffff88019f90dba8, .magic: 00000000, .owner: <none>/-1, .owner_cpu: 0
Pid: 24134, comm: test Not tainted 3.0.0-nv+ #800
Call Trace:
 spin_bug+0x9c/0xa3
 do_raw_spin_lock+0x29/0x13c
 _raw_spin_lock+0x1e/0x22
 nouveau_fence_update+0x2d/0xf1
 nouveau_channel_idle+0x22/0xa0
 nouveau_channel_put_unlocked+0x84/0x1bd
 nouveau_channel_put+0x20/0x24
 nouveau_channel_alloc+0x4ec/0x585
 nouveau_ioctl_fifo_alloc+0x50/0x130
 drm_ioctl+0x289/0x361
 do_vfs_ioctl+0x4dd/0x52c
 sys_ioctl+0x42/0x65
 system_call_fastpath+0x16/0x1b

It's easily triggerable from userspace.

Additionally remove double initialization of chan->fence.pending.

Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Cc: stable@kernel.org
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2011-11-10 08:58:53 +10:00
Ben Skeggs
8c06e60ed4 drm/nouveau: if requested, try harder at disabling sysmem pushbufs
On >=nv50, userspace would still end up allocating pushbufs in GART.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2011-09-20 16:11:41 +10:00
Ben Skeggs
7375c95b34 drm/nouveau: remove 'chan' argument from nouveau_bo_new
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2011-06-23 16:00:37 +10:00
Ben Skeggs
ce163f6967 drm/nv50-nvc0: explicitly map pushbuf bo into channel vm
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2011-06-23 16:00:09 +10:00
Ben Skeggs
180cc30637 drm/nouveau: convert bo.mem.start usage to bo.offset
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2011-06-23 15:59:44 +10:00
Ben Skeggs
07533ea549 drm/nouveau: convert some bo.offset use to vma.offset
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2011-06-23 15:59:39 +10:00
Ben Skeggs
dd6a46cc92 drm/nouveau: initialise any vm for a channel before pushbuf/ntfy
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2011-06-23 15:59:12 +10:00
Ben Skeggs
b7cb6c01ee drm/nouveau: modify gpuobj/ntfy takedown ordering
gpuobj really needs splitting into channel/gpuobj code instead...

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2011-06-23 15:59:07 +10:00
Ben Skeggs
e8a863c10f drm/nouveau: store a per-client channel list
Removes the need to disable IRQs to lookup channel struct on every pushbuf
ioctl, among others.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2011-06-23 15:58:25 +10:00
Ben Skeggs
a82dd49f14 drm/nouveau: remove remnants of nouveau_pgraph_engine
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2011-05-16 10:48:45 +10:00
Ben Skeggs
39a654d5b9 drm/nouveau: remove remnants of nouveau_pgraph_engine from nouveau_channel
The nouveau_wait_for_idle() call should hopefully not have been actually
necessary, we *do* wait for the channel to go idle already.  If it's
an issue somehow, the chipset-specific hooks can wait for idle themselves
before taking the lock.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2011-05-16 10:48:37 +10:00
Ben Skeggs
2703c21a82 drm/nv50/gr: move to exec engine interfaces
This needs a massive cleanup, but to catch bugs from the interface changes
vs the engine code cleanup, this will be done later.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2011-05-16 10:48:06 +10:00
Ben Skeggs
6dfdd7a61e drm/nouveau: working towards a common way to represent engines
There's lots of more-or-less independant engines present on NVIDIA GPUs
these days, and we generally want to perform the same operations on them.
Implementing new ones requires hooking into lots of different places,
the aim of this work is to make this simpler and cleaner.

NV84:NV98 PCRYPT moved over as a test.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2011-05-16 10:48:01 +10:00
Lucas De Marchi
25985edced Fix common misspellings
Fixes generated by 'codespell' and manually reviewed.

Signed-off-by: Lucas De Marchi <lucas.demarchi@profusion.mobi>
2011-03-31 11:26:23 -03:00
Ben Skeggs
d550c41e4f drm/nouveau: remove no_vm/mappable flags from nouveau_bo
'mappable' isn't really used at all, nor is it necessary anymore as the
bo code is capable of moving buffers to mappable vram as required.

'no_vm' isn't necessary anymore either, any places that don't want to be
mapped into a GPU address space should allocate the VRAM directly instead.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2011-02-25 06:45:34 +10:00
Ben Skeggs
ea5f2786a0 drm/nouveau: silence some compiler warnings
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2011-02-25 06:44:22 +10:00
Ben Skeggs
e457acaed4 drm/nouveau: create grctx on the fly on all chipsets
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2011-01-07 14:03:58 +10:00
Ben Skeggs
2a55c9a7ff drm/nvc0: reserve only subc 0 for kernel use
Current 3D driver expects this behaviour.  While this could be changed,
there's no compelling reason to reserve more than one subchannel for the
DRM.  If we ever need to use an object other then M2MF, we can just
re-bind subchannel 0 as required.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-12-30 11:55:07 +10:00
Michel Hermier
8e91182bbd drm/nouveau: Validate channel indices passed from userspace.
When hacking the libdrm for improvements, I triggered a kernel crash
related to the fact that the NOUVEAU_NOTIFIEROBJ_ALLOC ioctl calls
nouveau_channel_get with an unchecked channel index.
The patch ensures that the channel index is an unsigned and validates
its value in nouveau_channel_get.

Signed-off-by: Michel Hermier <hermier@frugalware.org>
Signed-off-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-12-27 10:41:01 +10:00
Ben Skeggs
96545299d7 drm/nvc0: fix channel dma init paths
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-12-21 17:17:34 +10:00
Ben Skeggs
4c13614298 drm/nv50: implement global channel address space on new VM code
As of this commit, it's guaranteed that if an object is in VRAM that its
GPU virtual address will be constant.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-12-08 13:48:13 +10:00
Ben Skeggs
d908175cca drm/nouveau: make fifo.create_context() responsible for mapping control regs
The regs belong to PFIFO, they're different for pretty much the same
generations we need different PFIFO control for, and NVC0 is going
to be even more different than the rest.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-12-08 03:00:34 +01:00
Francisco Jerez
6dccd311dd drm/nouveau: Synchronize with the user channel before GPU object destruction.
There have been reports of PFIFO cache errors during context take down
(fdo bug 31637). They are caused by some GPU objects being taken out
while the channel is still potentially processing commands. Make sure
that all the previous rendering has landed before releasing a GPU
object.

Reported-by: Grzesiek Sójka <pld@pfu.pl>
Reported-by: Patrice Mandin <patmandin@gmail.com>
Signed-off-by: Francisco Jerez <currojerez@riseup.net>
Acked-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-12-08 03:00:23 +01:00
Ben Skeggs
7f4a195fcb drm/nouveau: tidy up and extend dma object creation interfaces
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-12-03 15:11:59 +10:00
Francisco Jerez
332b242f47 drm/nouveau: Implement the pageflip ioctl.
nv0x-nv4x should be mostly fine, nv50 doesn't work yet.

Signed-off-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-12-03 15:11:12 +10:00
Marcin Slusarz
382d62e524 drm/nouveau: fix annoying nouveau_fence type issue
nouveau_fence_* functions are not type safe, which could lead to bugs.
Additionally every use of nouveau_fence_unref had to cast struct
nouveau_fence to void **.
Fix it by renaming old functions and creating static inline functions with
new prototypes. We still need old functions, because we pass function
pointers to ttm.
As we are wrapping functions, drop unused "void *arg" parameter where possible.

Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Signed-off-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-12-03 15:11:07 +10:00
Ben Skeggs
f4512e6579 drm/nv50: create graph and crypt contexts on demand
This really needs cleaning up somehow, and probably investigate what's
needed to do this on earlier generations.  NVIDIA do something similar
there too.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-12-03 15:10:44 +10:00
Ben Skeggs
bd2e597de8 drm/nv84: add support for the PCRYPT engine
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-12-03 15:06:58 +10:00
Francisco Jerez
f091a3d403 drm/nouveau: Implement weak channel references.
nouveau_channel_ref() takes a "weak" channel reference that doesn't
prevent the hardware channel resources from being released, it just
keeps the channel data structure alive.

Signed-off-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-12-03 15:06:42 +10:00
Francisco Jerez
feeb0aecfb drm/nouveau: Add unlocked variants of nouveau_channel_get/put.
Signed-off-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-12-03 15:06:38 +10:00
Francisco Jerez
f175b745b5 drm/nouveau: Fix race condition in channel refcount handling.
nouveau_channel_put() can be executed after the 'refcount == 0' check
in nouveau_channel_get() and before the channel reference count is
incremented. In that case CPU0 will take the context down while CPU1
thinks it owns the channel and 'refcount == 1'.

Signed-off-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-12-03 15:06:36 +10:00
Francisco Jerez
3945e47543 drm/nouveau: Refactor context destruction to avoid a lock ordering issue.
The destroy_context() engine hooks call gpuobj management functions to
release the channel resources, these functions use HARDIRQ-unsafe locks
whereas destroy_context() is called with the HARDIRQ-safe
context_switch_lock held, that's a lock ordering violation.

Push the engine-specific channel destruction logic into destroy_context()
and let the hardware-specific code lock and unlock when it's actually
needed. Change the engine destruction order to avoid a race in the small
gap between pgraph and pfifo context uninitialization.

Reported-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Signed-off-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-12-03 15:06:35 +10:00
Ben Skeggs
b12120a58e drm/nouveau: switch to unlocked ioctls
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-12-03 15:05:20 +10:00
Ben Skeggs
cff5c13324 drm/nouveau: add more fine-grained locking to channel list + structures
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-12-03 15:05:18 +10:00
Ben Skeggs
6a6b73f254 drm/nouveau: add per-channel mutex, use to lock access to drm's channel
This fixes a race condition between fbcon acceleration and TTM buffer
moves.  To reproduce:

- start X
- switch to vt and "while (true); do dmesg; done"
- switch to another vt and "sleep 2 && cat /path/to/debugfs/dri/0/evict_vram"
- switch back to vt running dmesg

We don't make use of this on any other channel yet, they're currently
protected by drm_global_mutex.  This will change in the near future.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-12-03 15:05:10 +10:00
Ben Skeggs
d961db75ce drm/ttm: restructure to allow driver to plug in alternate memory manager
Nouveau will need this on GeForce 8 and up to account for the GPU
reordering physical VRAM for some memory types.

Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Acked-by: Thomas Hellström <thellstrom@vmware.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-10-05 10:01:20 +10:00
Francisco Jerez
2730723bbc drm/nouveau: Minor refactoring/cleanup of the fence code.
Mainly to make room for inter-channel sync.

Signed-off-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-10-05 09:58:59 +10:00
Ben Skeggs
a8eaebc6c5 drm/nouveau: remove nouveau_gpuobj_ref completely, replace with sanity
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-09-24 16:20:14 +10:00
Ben Skeggs
9d59e8a1fc drm/nouveau: require explicit unmap of kmapped bos
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-09-24 16:18:31 +10:00
Dave Airlie
1b2f148963 drm: block userspace under allocating buffer and having drivers overwrite it (v2)
With the current screwed but its ABI, ioctls for the drm, Linus pointed out that we could allow userspace to specify the allocation size, but we pass it to the driver which then uses it blindly to store a struct. Now if userspace specifies the allocation size as smaller than the driver needs, the driver can possibly overwrite memory.

This patch restructures the driver ioctls so we store the structure size we are expecting, and make sure we allocate at least that size. The copy from/to userspace are still restricted to the size the user specifies, this allows ioctl structs to grow on both sides of the equation.

Up until now we didn't really use the DRM_IOCTL defines in the kernel, so this cleans them up and adds them for nouveau.

v2:
fix nouveau pushbuf arg (thanks to Ben for pointing it out)

Reported-by: Linus Torvalds <torvalds@linuxfoundation.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-08-17 14:52:25 +10:00
Ben Skeggs
047d1d3cae drm/nouveau: reduce usage of fence spinlock to when absolutely necessary
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-07-13 10:14:04 +10:00
Ben Skeggs
b6d3d87178 drm/nouveau: remove dev_priv->init_state and friends
Nouveau will no longer load at all if card initialisation fails, so all
these checks are unnecessary.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-07-13 10:13:37 +10:00
Jordan Crouse
01d73a6967 drm: Remove drm_resource wrappers
Remove the drm_resource wrappers and directly use the
actual PCI and/or platform functions in their place.

[airlied: fixup nouveau properly to build]

Signed-off-by: Jordan Crouse <jcrouse@codeaurora.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-06-01 10:07:24 +10:00
Ben Skeggs
952eb63543 drm/nouveau: remove some unused members from drm_nouveau_private
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-04-09 10:11:57 +10:00
Ben Skeggs
a1606a9596 drm/nouveau: new gem pushbuf interface, bump to 0.0.16
This commit breaks the userspace interface, and requires a new libdrm for
nouveau to operate again.

The multiple GEM_PUSHBUF ioctls that were present in 0.0.15 for
compatibility purposes are now gone, and replaced with the new ioctl which
allows for multiple push buffers to be submitted (necessary for hw index
buffers in the nv50 3d driver) and relocations to be applied on any buffer.

A number of other ioctls (CARD_INIT, GEM_PIN, GEM_UNPIN) that were needed
for userspace modesetting have also been removed.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Francisco Jerez <currojerez@riseup.net>
2010-02-25 15:08:44 +10:00
Ben Skeggs
d87897d4c8 drm/nv50: make pushbuf dma object cover entire vm
This allows us to submit push buffers from any memtype to the hardware.
We'll need this ability for VRAM index buffers at some point.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-02-25 15:08:33 +10:00
Maarten Maathuis
ff9e5279b1 drm/nouveau: protect channel create/destroy and irq handler with a spinlock
The nv50 pgraph handler (for example) could reenable pgraph fifo access
and that would be bad when pgraph context is being unloaded (we need the
guarantee a ctxprog isn't running).

Signed-off-by: Maarten Maathuis <madman2003@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-02-25 15:07:53 +10:00
Maarten Maathuis
a51a3bf50d drm/nv50: avoid unloading pgraph context when ctxprog is running
- We need to disable pgraph fifo access before checking the current channel,
  otherwise we could still hit a running ctxprog.
- The writes to 0x400500 are already handled by pgraph->fifo_access and are
  therefore redundant, moreover pgraph fifo access should not be reenabled
  before current context is set as invalid. So remove them altogether.

Signed-off-by: Maarten Maathuis <madman2003@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-02-10 08:19:36 +10:00
Ben Skeggs
75c99da6cd drm/nouveau: initialise DMA tracking parameters earlier
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2010-01-11 09:06:45 +10:00