Whilst cleaning up the patches for submission, I mis-classified non-dirty
objects as purgeable. This was causing the backing pages for those
objects to be evicted under memory-pressure, discarding valid and
unreplaceable texture data.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
As evict_something() is called by routines that do not repeatedly search
again, try harder in the initial search to find an object that matches
the request.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
First the routine attempted to unlock a mutex it did not own along the
error path.
Secondly the routine should never be called on any list but the inactive
one, since we attempt to unbind those objects, so fix the calling semantics.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
By adding tracepoint equivalents for WATCH_BUF/EXEC we are able to monitor
the lifetimes of objects, requests and significant events. These events can
then be probed using the tracing frameworks, such as systemtap and, in
particular, perf.
For example to record the stack trace for every GPU stall during a run, use
$ perf record -e i915:i915_gem_request_wait_begin -c 1 -g
And
$ perf report
to view the results.
[Updated to fix compilation issues caused.]
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Ben Gamari <bgamari@gmail.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
* 'drm-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6: (133 commits)
drm/vgaarb: add VGA arbitration support to the drm and kms.
drm/radeon: some r420s have a CP race with the DMA engine.
drm/radeon/r600/kms: rv670 is not DCE3
drm/radeon/kms: r420 idle after programming GA_ENHANCE
drm/radeon/kms: more fixes to rv770 suspend/resume path.
drm/radeon/kms: more alignment for rv770.c with r600.c
drm/radeon/kms: rv770 blit init called too late.
drm/radeon/kms: move around new init path code to avoid posting at init
drm/radeon/r600: fix some issues with suspend/resume.
drm/radeon/kms: disable VGA rendering engine before taking over VRAM
drm/radeon/kms: Move radeon_get_clock_info() call out of radeon_clocks_init().
drm/radeon/kms: add initial connector properties
drm/radeon/kms: Use surfaces for scanout / cursor byte swapping on big endian.
drm/radeon/kms: don't fail if we fail to init GPU acceleration
drm/r600/kms: fixup number of loops per blit calculation.
drm/radeon/kms: reprogram format in set base.
drm/radeon: avivo chips have no separate int bit for display
drm/radeon/r600: don't do interrupts
drm: fix _DRM_GEM addmap error message
drm: update crtc x/y when only fb changes
...
Fixed up trivial conflicts in firmware/Makefile due to network driver
(cxgb3) and drm (mga/r128/radeon) firmware being listed next to each
other.
If the presumed_offset as feed to userspace and returned to the kernel
from a previous execbuffer is still valid, then we do not need to rewrite
the relocation entry and may skip the offset sanity checks.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Eric noted a potential concern with the low bits not being strictly used
as part of the absolute offset (instead part of the command stream to the
GPU), but in practice that should not be an issue.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Tested-by: Andy Whitcroft <apw@canonical.com>
Cc: Eric Anholt <eric@anholt.net>
CC: stable@kernel.org
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Due to the necessity of having to take the struct_mutex, the i915
shrinker can not free the inactive lists if we fail to allocate memory
whilst processing a batch buffer, triggering an OOM and an ENOMEM that
is reported back to userspace. In order to fare better under such
circumstances we need to manually retry a failed allocation after
evicting inactive buffers.
To do so involves 3 steps:
1. Marking the backing shm pages as NORETRY.
2. Updating the get_pages() callers to evict something on failure and then
retry.
3. Revamping the evict something logic to be smarter about the required
buffer size and prefer to use volatile or clean inactive pages.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Similar to the madvise() concept, the application may wish to mark some
data as volatile. That is in the event of memory pressure the kernel is
free to discard such buffers safe in the knowledge that the application
can recreate them on demand, and is simply using these as a cache.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
This should help GEM handle memory pressure sitatuions more gracefully.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
There is no need to store the gtt_alignment as it is either explicitly
set according to the hardware requirements (e.g. scanout) or the
minimum alignment is computed on demand.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
If we failed to set the domain, the buffer was no longer being tracked
on any list.
Cc: stable@kernel.org
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
There is a very real possibility that multiple CPUs will notice that the
GPU is wedged. This introduces all sorts of potential race conditions.
Make the wedged flag atomic to mitigate this risk.
Signed-off-by: Ben Gamari <bgamari.foss@gmail.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
We set a periodic timer to check on the GPU, resetting it every time a
batch is completed. If the timer elapses, we check acthd. If acthd
hasn't changed in two timer periods, we assume the chip is wedged.
This is implemented in such a way that it leaves the option open to
employ adaptive timer intervals in the future. One could wait until
several timer periods have elapsed before declaring the chip dead. If
the chip comes back after several periods but before the "dead"
threshold, the timer interval or dead threshold could be raised.
It is important to note that while checking for active requests, we need
to account for the fact that requests are removed from the list (i.e.
retired) in a deferred work queue handler. This means that merely
checking for an empty request_list is insufficient; the list could be
non-empty yet the GPU still idle, causing the hangcheck timer to
incorrectly mark the GPU as wedged (it took me a while to figure that
out---sigh...)
Signed-off-by: Ben Gamari <bgamari.foss@gmail.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
We'll need it in i915_irq.c for checking whether there are outstanding
requests. Also, the function really ought to return a bool, not an int.
Signed-off-by: Ben Gamari <bgamari.foss@gmail.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
i915_wait_request() only checks mm.wedged after it interacts with the
hardware, generally causing the driver to lock up waiting for a wedged
chip. Make sure we check mm.wedged as the first thing we do.
Reported-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ben Gamari <bgamari.foss@gmail.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
drm_ht_remove_item() does not handle removing an absent item and the hlist
in particular is incorrectly initialised. The easy remedy is simply skip
calling i915_gem_free_mmap_offset() unless we have actually created the
offset and associated ht entry.
This also fixes the mishandling of a partially constructed offset which
leaves pointers initialized after freeing them along the
i915_gem_create_mmap_offset() error paths.
In particular this should fix the oops found here:
https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/415357/comments/8
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
Cc: stable@kernel.org
Ever since we enabled GEM, the pre-9xx chipsets (particularly 865) have had
serious stability issues. Back in May a wbinvd was added to the DRM to
work around much of the problem. Some failure remained -- easily visible
by dragging a window around on an X -retro desktop, or by looking at bugzilla.
The chipset flush was on the right track -- hitting the right amount of
memory, and it appears to be the only way to flush on these chipsets, but the
flush page was mapped uncached. As a result, the writes trying to clear the
writeback cache ended up bypassing the cache, and not flushing anything! The
wbinvd would flush out other writeback data and often cause the data we wanted
to get flushed, but not always. By removing the setting of the page to UC
and instead just clflushing the data we write to try to flush it, we get the
desired behavior with no wbinvd.
This exports clflush_cache_range(), which was laying around and happened to
basically match the code I was otherwise going to copy from the DRM.
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Brice Goglin <Brice.Goglin@ens-lyon.org>
Cc: stable@kernel.org
Otherwise, some other userland writing into its buffer may race to land
writes either after the CPU thinks it's got a coherent view, or after its
GTT entries have been redirected to point at the scratch page. Either
result is unpleasant.
Signed-off-by: Eric Anholt <eric@anholt.net>
Reinette Chatre reports a frozen system (with blinking keyboard LEDs)
when switching from graphics mode to the text console, or when
suspending (which does the same thing). With netconsole, the oops
turned out to be
BUG: unable to handle kernel NULL pointer dereference at 0000000000000084
IP: [<ffffffffa03ecaab>] i915_driver_irq_handler+0x26b/0xd20 [i915]
and it's due to the i915_gem.c code doing drm_irq_uninstall() after
having done i915_gem_idle(). And the i915_gem_idle() path will do
i915_gem_idle() ->
i915_gem_cleanup_ringbuffer() ->
i915_gem_cleanup_hws() ->
dev_priv->hw_status_page = NULL;
but if an i915 interrupt comes in after this stage, it may want to
access that hw_status_page, and gets the above NULL pointer dereference.
And since the NULL pointer dereference happens from within an interrupt,
and with the screen still in graphics mode, the common end result is
simply a silently hung machine.
Fix it by simply uninstalling the irq handler before idling rather than
after. Fixes
http://bugzilla.kernel.org/show_bug.cgi?id=13819
Reported-and-tested-by: Reinette Chatre <reinette.chatre@intel.com>
Acked-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
According to the docs, the ringbuffer is not allowed to wrap in the middle
of an instruction.
G45 PRM, Vol 1b, p101:
While the “free space” wrap may allow commands to be wrapped around the
end of the Ring Buffer, the wrap should only occur between commands.
Padding (with NOP) may be required to follow this restriction.
Do as commanded.
[Having seen bug reports where there is evidence of split commands, but
apparently the GPU has continued on merrily before a bizarre and untimely
death, this may or may not fix a few random hangs.]
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
CC: Eric Anholt <eric@anholt.net>
Signed-off-by: Eric Anholt <eric@anholt.net>
There are several sources of unnecessary power consumption on Intel
graphics systems. The first is the LVDS clock. TFTs don't suffer from
persistence issues like CRTs, and so we can reduce the LVDS refresh rate
when the screen is idle. It will be automatically upclocked when
userspace triggers graphical activity. Beyond that, we can enable memory
self refresh. This allows the memory to go into a lower power state when
the graphics are idle. Finally, we can drop some clocks on the gpu
itself. All of these things can be reenabled between frames when GPU
activity is triggered, and so there should be no user visible graphical
changes.
Signed-off-by: Jesse Barnes <jesse.barnes@intel.com>
Signed-off-by: Matthew Garrett <mjg@redhat.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
Remember to release the local reference if we fail to wait on
the rendering.
(Also whilst in the vicinity add some whitespace so that the phasing of
the operations is clearer.)
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
Several functions in the GEM kernel API used int as handle type, but
user API has it __u32 which is also the intended type.
Replace int with u32.
Signed-off-by: Pekka Paalanen <pq@iki.fi>
Signed-off-by: Dave Airlie <airlied@redhat.com>
As of 52dc7d32b8, we could leave an old
linear GTT mapping in place, so that apps trying to GTT-mapped write in
tiled data wouldn't get the fence added, and garbage would get displayed.
Signed-off-by: Eric Anholt <eric@anholt.net>
As we call unmap_mapping_range() twice in identical fashion, refactor
and attempt to explain why we need to call unmap_mapping_range().
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
Unitialized fence register could leads to corrupted display. Problem
encountered on MacBooks (revision 1 and 2), directly booting from EFI
or through BIOS emulation.
(bug #21710 at freedestop.org)
Signed-off-by: Grégoire Henry <henry@pps.jussieu.fr>
Signed-off-by: Eric Anholt <eric@anholt.net>
It hasn't been used in ages, and having the user tell your how much
memory is being freed at free time is a recipe for disaster even if it
was ever used.
Signed-off-by: Eric Anholt <eric@anholt.net>
The fence register value also depends upon the stride of the object, so we
need to clear the fence if that is changed as well.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
[anholt: Added 8xx and 965 paths, and renamed the confusing
i915_gem_object_tiling_ok function to i915_gem_object_fence_offset_ok]
Signed-off-by: Eric Anholt <eric@anholt.net>
With the work by Jesse Barnes to eliminate allocation of fences during
execbuffer, it becomes possible to write to the scan-out buffer with it
never acquiring a fence (simply by only ever writing to the object using
tiled GPU commands and never writing to it via the GTT). So for pre-i965
chipsets which require fenced access for tiled scan-out buffers, we need
to obtain a fence register.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
After performing an operation over the page list for a buffer retrieved by
i915_gem_object_get_pages() the pages need to be returned with
i915_gem_object_put_pages(). This was not being observed for the phys
objects which were thus leaking references to their backing pages.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
CC: Dave Airlie <airlied@gmail.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
To differentiate between encountering an out-of-memory error with running
out of space in the aperture, use ENOSPC for the later.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
The batch buffer may be shared with another read buffer, so we should not
ignore any previously set domains, but just or in the command domain (and
check that the buffer is not writable).
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
By sending a broken execbuffer (its length was not suitably aligned) I
triggered an operation upon a freed object. The invalid alignment was
discovered after updating the write_domain on the object but before the
object was placed on the active queue. So during the unwind process
following the error, the now freed object attempts to flush its
non-existent, but outstanding, GPU writes causing this use-after-free.
[drm:i915_dispatch_gem_execbuffer] *ERROR* alignment
[drm:i915_gem_execbuffer] *ERROR* dispatch failed -22
WARNING: at lib/kref.c:43 warn_slowpath_null+0x10/0x15()
Modules linked in:
Pid: 4552, comm: lt-csi-drm Not tainted 2.6.30-rc6 #423
Call Trace:
[<c0119ef3>] warn_slowpath_fmt+0x57/0x6d
[<c014de24>] ? get_pageblock_migratetype+0x18/0x1e
[<c014e8fd>] ? free_hot_page+0xa/0xc
[<c014e915>] ? __free_pages+0x16/0x1f
[<c0153ebf>] ? shmem_truncate_range+0x63e/0x656
[<c015fb2f>] ? slob_page_alloc+0x146/0x1c8
[<c0119f19>] warn_slowpath_null+0x10/0x15
[<c01f55f2>] kref_get+0x1b/0x21
[<c02605db>] i915_gem_object_move_to_active+0x1f/0x56
[<c0261302>] i915_add_request+0x156/0x19a
[<c026136e>] i915_gem_object_flush_gpu_write_domain+0x28/0x3f
[<c0261eca>] i915_gem_object_unbind+0x4a/0x124
[<c0261fd7>] i915_gem_free_object+0x33/0x9b
[<c0250d6b>] drm_gem_object_free+0x28/0x4a
[<c0250d43>] ? drm_gem_object_free+0x0/0x4a
[<c01f55ce>] kref_put+0x38/0x41
[<c0250cbf>] drm_gem_object_unreference+0x11/0x13
[<c0250d06>] drm_gem_object_handle_unreference+0x1e/0x21
[<c0250d13>] drm_gem_object_release_handle+0xa/0xe
[<c01f3e6b>] idr_for_each+0x5f/0x98
[<c0250d09>] ? drm_gem_object_release_handle+0x0/0xe
[<c0250daf>] drm_gem_release+0x22/0x34
[<c025046f>] drm_release+0x1e8/0x3c4
[<c0162d25>] __fput+0xaf/0x146
[<c0162dce>] fput+0x12/0x14
[<c01605ef>] filp_close+0x48/0x52
[<c011b182>] put_files_struct+0x57/0x9b
[<c011b1e4>] exit_files+0x1e/0x20
[<c011c6b6>] do_exit+0x16d/0x511
[<c03704ab>] ? __schedule+0x3d4/0x3e5
[<c0103f0d>] ? handle_irq+0xd/0x69
[<c011caa7>] do_group_exit+0x4d/0x73
[<c011cae0>] sys_exit_group+0x13/0x17
[<c010268c>] sysenter_do_call+0x12/0x2b
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
Update interrupt handling methods for IGDNG with new registers
for display and graphics interrupt functions. As we won't use
irq-based vblank sync in dri2, so display interrupt on new chip
will be used for hotplug only in future.
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
keithp didn't like the original 20ms plan because a cooperative client could
be starved by an uncooperative client. There may even have been problems
with cooperative clients versus cooperative clients. So keithp changed
throttle to just wait for the second to last seqno emitted by that client.
It worked well, until we started getting more round-trips to the server
due to DRI2 -- the server throttles in BlockHandler, and so if you did more
than one round trip after finishing your frame, you'd end up unintentionally
syncing to the swap.
Fix this by keeping track of the client's requests, so the client can wait
when it has an outstanding request over 20ms old. This should have
non-starving behavior, good behavior in the presence of restarts, and less
waiting. Improves high-settings openarena performance on my GM45 by 50%.
Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
This could be triggered by a gtt mapping fault on 965 that decides to
remove the fence from another object that happens to be active currently.
Since the other object doesn't rely on the fence reg for its execution, we
don't wait for it to finish. We'll soon be not waiting on 915 most of the
time as well, so just drop the BUG_ON.
Signed-off-by: Eric Anholt <eric@anholt.net>
When a GEM object is evicted from the GTT we set it to the CPU domain,
as it might get swapped in and out or ever mmapped regularly. If the
object is mmapped through the GTT it can still get evicted in this way
by other objects requiring GTT space. When the GTT mapping is touched
again we fault it back into the GTT, but fail to set it back to the
GTT domain. This means we fail to flush any cached CPU writes to the
pages backing the object which will then happen "eventually", typically
after we write to the page through the uncached GTT mapping.
[anholt: Note that userland does do a set_domain(GTT, GTT) when starting
to access the GTT mapping. That covers getting the existing mapping of the
object synchronized if it's bound to the GTT. But set_domain(GTT, GTT)
doesn't do anything if the object is currently unbound. This fix covers the
transition to being bound for GTT mapping.]
Fixes glyph and other pixmap corruption during swapping. fd.o bug #21790
Signed-off-by: Kristian Høgsberg <krh@redhat.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
On the 865, but not the 855, the clflush we do appears to not actually make
it out to the hardware all the time. An easy way to safely reproduce was
X -retro, which would show that some of the blits involved in drawing the
lovely root weave didn't make it out to the hardware. Those blits are 32
bytes each, and 1-2 would be missing at various points around the screen.
Other experimentation (doing more clflush, doing more AGP chipset flush,
poking at some more device registers to maybe trigger more flushing) didn't
help. krh came up with the wbinvd as a way to successfully get all those
blits to appear.
Signed-off-by: Eric Anholt <eric@anholt.net>
The pitch field is an exponent on pre-965, so we were rejecting buffers
on 8xx that we shouldn't have. 915 got lucky in that the largest legal
value happened to match (8KB / 512 = 0x10), but 8xx has a smaller tile width.
Additionally, we programmed that bad value into the register on 8xx, so the
only pitch that would work correctly was 4096 (512-1023 pixels), while others
would probably give bad rendering or hangs.
Signed-off-by: Eric Anholt <eric@anholt.net>
fd.o bug #20473.
For some reason we never added 8xx desktop cursor support to the
kernel. This patch fixes that.
[krh: Also set the size on pre-i915 hw.]
Tested-by: Kristian Høgsberg <krh@redhat.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Eric Anholt <eric@anholt.net>
For awhile now, many of the GEM code paths have allocated page or
object arrays with the slab allocator. This is nice and fast, but
won't work well if memory is fragmented, since the slab allocator works
with physically contiguous memory (i.e. order > 2 allocations are
likely to fail fairly early after booting and doing some work).
This patch works around the issue by falling back to vmalloc for
>PAGE_SIZE allocations. This is ugly, but much less work than chaining
a bunch of pages together by hand (suprisingly there's not a bunch of
generic kernel helpers for this yet afaik). vmalloc space is somewhat
precious on 32 bit kernels, but our allocations shouldn't be big enough
to cause problems, though they're routinely more than a page.
Note that this patch doesn't address the unchecked
alloc-based-on-ioctl-args in GEM; that needs to be fixed in a separate
patch.
Also, I've deliberately ignored the DRM's "area" junk. I don't think
anyone actually uses it anymore and I'm hoping it gets ripped out soon.
[Updated: removed size arg to new free function. We could unify the
free functions as well once the DRM mem tracking is ripped out.]
fd.o bug #20152 (part 1/3)
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Eric Anholt <eric@anholt.net>
We might sleep here anyway so I hope an extra uncached read is ok to
add.
In #20896 we found that vbetool clobbers the IER. In KMS mode this is
particularly bad since we don't set the interrupt regs late (in
EnterVT), so we'd fail to get *any* interrupts at all after X started
(since some distros have scripts that call vbetool at X startup
apparently).
So this patch checks IER at wait_request time, and re-enables
interrupts if it's been clobbered. In a proper config this check
should never be triggered.
This is really a distro issue, but having a sanity check is nice, as
long as it doesn't have a real performance hit.
Tested-by: Mateusz Kaduk <mateusz.kaduk@gmail.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
[anholt: Moved the check inside of the sleeping case to avoid perf cost]
Signed-off-by: Eric Anholt <eric@anholt.net>
regression caused by commit 5e118f4139:
i915_gem_object_move_to_inactive() should be called in task context,
as it calls fput();
Signed-off-by: Shaohua Li<shaohua.li@intel.com>
[anholt: Add more detail to the comment about the lock break that's added]
Signed-off-by: Eric Anholt <eric@anholt.net>
Save the bit 17 state of the pages when freeing the page list, and
reswizzle them if necessary when rebinding the pages (in case they were
swapped out). Since we have userland with expectations that the swizzle
enums let it pread and pwrite contents accurately, we can't expose a new
swizzle enum for bit 17 (which it would have to GTT map to handle), so we
handle it down in pread and pwrite by swizzling the copy when bit 17 of the
page address is set.
Signed-off-by: Eric Anholt <eric@anholt.net>
Otherwise, the results of our read didn't show up when we were faulting in
the page being read into (as happened with a testcase reading into a big
stack area). Likely accounts for some conformance test failures.
Signed-off-by: Eric Anholt <eric@anholt.net>
i915_gem_put_relocs_to_user returned an uninitialized value which
got returned to userspace. This caused libdrm in my setup to never
get out of a do{}while() loop retrying i915_gem_execbuffer.
result was hanging X, overheating of cpu and 2-3gb of logfile-spam.
This patch adresses the issue by
1. initializing vars in this file where necessary
2. correcting wrongly interpreted return values of copy_[from/to]_user
Signed-off-by: Florian Mickler <florian@mickler.org>
[anholt: cleanups of unnecessary changes, consistency in APIs]
Signed-off-by: Eric Anholt <eric@anholt.net>
We create a debugfs node (i915_ringbuffer_data) to expose a hex dump
of the ring buffer itself. We also expose another debugfs node
(i915_ringbuffer_info) with information on the state (i.e. head, tail
addresses) of the ringbuffer.
For batchbuffer dumping, we look at the device's active_list, dumping
each object which has I915_GEM_DOMAIN_COMMAND in its read
domains. This is all exposed through the dri/i915_batchbuffers debugfs
file with a header for each object (giving the objects gtt_offset so
that it can be matched against the offset given in the
BATCH_BUFFER_START command.
Signed-off-by: Ben Gamari <bgamari@gmail.com>
Signed-off-by: Carl Worth <cworth@cworth.org>
Signed-off-by: Eric Anholt <eric@anholt.net>
This is a baby-step in the direction of having finer-grained
locking than the struct_mutex. Specifically, this will enable
new debugging code to read the active list for printing out
GPU state when the GPU is wedged, (while the struct_mutex is
held, of course).
Signed-off-by: Carl Worth <cworth@cworth.org>
[anholt: indentation fix]
Signed-off-by: Eric Anholt <eric@anholt.net>
Indicates something is wrong with the mapping; and apparently triggers
in current kernels.
Signed-off-by: Jesse Barnes <jbarnes@virtuosugeek.org>
Signed-off-by: Eric Anholt <eric@anholt.net>
This fixes all the tiling problems with the 2d ddx. glxgears still doesn't work.
Changes:
- fix a copy&paste error in i8xx fence reg setup. It resulted in an at most a
512KB offset of the fence reg window, so was only visible sometimes.
- add tests for stride and object size constrains (also for i915 and 1965 class
hw). Userspace seems to have an of-by-one bug there, which changes the fence
size by at most 512KB due to an overflow.
- because i8xx hw is quite old (and therefore not as well-tested) I left 2 debug
WARN_ONs in the i8xx fence reg setup code to hopefully catch any further
overflows in the bit-fields. Lastly there's one small change to make the
alignment checks more consistent.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=20289
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Eric Anholt <eric@anholt.net>
This produced a warning on my build, not sure why super-warning-man didn't
notice this one, its much worse than the %z one.
Signed-off-by: Dave Airlie <airlied@redhat.com>
agp_chipset_flush() is for flushing the intel GMCH write cache via the
IFP, these two uses are for when we're getting the object into the cpu
READ domain, and thus should not be needed. This confused me when I was
getting my head around the code.
With thanks to airlied for helping me check my mental picture of how the
flushes and clflushes are supposed to be used.
Signed-off-by: Owain G. Ainsworth <oga@openbsd.org>
Signed-off-by: Eric Anholt <eric@anholt.net>
This introduces allocation in the batch submission path that wasn't there
previously, but these are compatibility paths so we care about simplicity
more than performance.
kernel.org bug #12419.
Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Keith Packard <keithp@keithp.com>
Acked-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Like the GTT pwrite path fix, this uses an optimistic path and a
fallback to get_user_pages. Note that this means we have to stop using
vfs_write and roll it ourselves.
Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
We've wanted this for a few consumers that touch the pages directly (such as
the following commit), which have been doing the refcounting outside of
get/put pages.
Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Since the pagefault path determines that the lock order we use has to be
mmap_sem -> struct_mutex, we can't allow page faults to occur while the
struct_mutex is held. To fix this in pwrite, we first try optimistically to
see if we can copy from user without faulting. If it fails, fall back to
using get_user_pages to pin the user's memory, and map those pages
atomically when copying it to the GPU.
Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
agp_chipset_flush() is for flushing the intel GMCH write cache via the
IFP, these two uses are for when we're getting the object into the cpu
READ domain, and thus should not be needed. This confused me when I was
getting my head around the code.
With thanks to airlied for helping me check my mental picture of how the
flushes and clflushes are supposed to be used.
Signed-off-by: Owain G. Ainsworth <oga@openbsd.org>
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Once upon a time, the DRM made the distinction between the drm_map
data structure exchanged with user space and the drm_local_map used
in the kernel.
For some reasons, while the BSD port still has that "feature", the
linux part abused drm_map for kernel internal usage as the local
map only existed as a typedef of the struct drm_map.
This patch fixes it by declaring struct drm_local_map separately
(though its content is currently identical to the userspace variant),
and changing the kernel code to only use that, except when it's a
user<->kernel interface (ie. ioctl).
This allows subsequent changes to the in-kernel format
I've also replaced the use of drm_local_map_t with struct drm_local_map
in a couple of places. Mostly by accident but they are the same (the
former is a typedef of the later) and I have some remote plans and
half finished patch to completely kill the drm_local_map_t typedef
so I left those bits in.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@linux.ie>
The last 8 fence registers sit at a different offset, so when we went to set
fence number 8 in the lower offset, we instead set PGETBL_CTL, and the GPU
got all sorts of angry at us.
fd.o bug #20567. Easily reproducible by running glxgears and killing it about
6 times.
Signed-off-by: Eric Anholt <eric@anholt.net>
The i915 also uses the fence registers for GPU access to tiled buffers so
we cannot reallocate one whilst it is on the active list. By performing a
LRU scan of the fenced buffers we also avoid waiting the possibility of
waiting on a pinned, or otherwise unusable, buffer.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
We need to check and report if there are no available fences - or else we
spin endlessly waiting for a buffer to magically unpin itself.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
As we may steal the fence register of an unpinned buffer for another,
every time we repin the buffer we need to recheck whether it needs to be
allocated a fence.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
If we wait upon a request and successfully unbind a buffer occupying a
fence register, then that slot will be freed and cause a NULL derefrence
upon rescanning.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
If userspace passes an object list with the same object appearing more
than once, we end up hitting the BUG_ON() in
i915_gem_object_set_to_gpu_domain() as it gets called a second time
for the same object.
Signed-off-by: Kristian Høgsberg <krh@redhat.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
dev_priv->hw_status_page can be NULL, if i915_gem_retire_requests()
is called from i915_gem_busy_ioctl().
Signed-off-by Karsten Wiese <fzu@wemgehoertderstaat.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The object is dereferenced before the NULL check. Oops.
Fixes http://bugs.freedesktop.org/show_bug.cgi?id=20235
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This ensures that the user gets the latest information from the hardware
on whether the buffer is busy, potentially reducing the working set of objects
that the user chooses.
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>
In the KMS case, we need to suspend/resume GEM as well. So on suspend, make
sure we idle GEM and stop any new rendering from coming in, and on resume,
re-init the framebuffer and clear the suspended flag.
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>
The problem was that object_set_to_gpu_domain would set the new write_domains
that are getting set by this batchbuffer, then the accumulated flushes required
for all the objects in preparation for this batchbuffer were posted, and the
brand new write domain would get cleared by the flush being posted. Instead,
hang on to the new (or old if we're not changing it) value and set it after
the flush is queued.
Results from this noticably included conformance test failures from reads
shortly after writes (where the new write domain had been lost and thus not
flushed and waited on), but is a suspected cause of hangs in some apps when
a write domain is lost on a buffer that gets reused for instruction or
commmand state.
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>
While not strictly required, it helped while thinking about the following
change. This change should be invariant.
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This fixes potential fault at fault time if the object was unreferenced
while the mapping still existed. Now, while the mmap_offset only lives
for the lifetime of the object, the object also stays alive while a vma
exists that needs it.
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>
If we fail to create the ringbuffer, then we need to cleanup the allocated
hws.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@linux.ie>
A missing unpin on the error path.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@linux.ie>
A missing unpin on the error path.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@linux.ie>
A missing unreference and unpin after rejecting the relocation for an
invalid memory domain.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@linux.ie>
We failed to unlock the mutex after failing to create the mmap offset.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@linux.ie>
Lockdep warns that i915_gem_execbuffer() can trigger a page fault (which
takes mmap_sem) while holding dev->struct_mutex, while drm_vm_open()
(which is called with mmap_sem already held) takes dev->struct_mutex.
So this is a potential AB-BA deadlock.
The way that i915_gem_execbuffer() triggers a page fault is by doing
copy_to_user() when returning new buffer offsets back to userspace;
however there is no reason to hold the struct_mutex when doing this
copy, since what is being copied is the contents of an array private to
i915_gem_execbuffer() anyway. So we can fix the potential deadlock (and
get rid of the lockdep warning) by simply moving the copy_to_user()
outside of where struct_mutex is held.
This fixes <http://bugzilla.kernel.org/show_bug.cgi?id=12491>.
Reported-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@linux.ie>
A missing unreference if the user calls pin() a second time on a pinned
buffer.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@linux.ie>
Also spotted by Owain Ainsworth.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@linux.ie>
The DRI people seem to have a hard time getting these right (see also
commit aeb565dfc3).
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
If we failed to allocate a new fence register we would return
VM_FAULT_SIGBUS without relinquishing the lock.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@linux.ie>
Adds code to set up fence registers at execbuf time on pre-965 chips as
necessary. Also fixes up a few bugs in the pre-965 tile register support
(get_order != ffs). The number of fences available to the kernel defaults
to the hw limit minus 3 (for legacy X front/back/depth), but a new parameter
allows userspace to override that as needed.
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@linux.ie>
Previously, the caller would continue along without knowing that the
function failed, resulting in potential mis-rendering. Right now vm_fault
just returns SIGBUS in that case, and we may need to disable signal handling
to avoid that happening.
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@linux.ie>
We'd love to just be using PAT, but even on chips with PAT it gets disabled
sometimes due to an errata. It would probably be better to have pat_enabled
exported and only bother with this when !pat_enabled.
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@linux.ie>
