We try to avoid writing the relocations through the uncached GTT, if the
buffer is currently in the CPU write domain and so will be flushed out to
main memory afterwards anyway. Also on SandyBridge we can safely write
to the pages in cacheable memory, so long as the buffer is LLC mapped.
In either of these cases, we therefore do not need to force the
reallocation of the buffer into the mappable region of the GTT, reducing
the aperture pressure.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Originally the code tried to allocate a large enough array to perform
the copy using vmalloc, performance wasn't great and throughput was
improved by processing each individual relocation entry separately.
This too is not as efficient as one would desire. A compromise would be
to allocate a single page, or to allocate a few entries on the stack,
and process the copy in batches. The latter gives simpler code and more
consistent performance due to a lack of heuristic.

x11perf -copywinwin10:	n450/pnv	i3-330m		i5-2520m (cpu)
               before: 	  249000	 785000		 1280000 (80%)
                 page:	  264000	 896000		 1280000 (65%)
             on-stack:	  264000	 902000		 1280000 (67%)

v2: Use 512-bytes of stack for batching rather than allocate a page.
v3: Tidy the code slightly with more descriptive variable names
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

PIPE_CONTROL on snb needs global gtt mappings in place to workaround a
hw gotcha. No other commands need such a workaround. Luckily we can
detect a PIPE_CONTROL commands easily because they have a write_domain
= I915_GEM_DOMAIN_INSTRUCTION (and nothing else has that).

v2: Binding the target of such a reloc into the global gtt actually
works instead of binding the source, which is rather pointless ...

v3: Kill a superflous has_global_gtt_mapping assignement noticed by
Chris Wilson.
Reviewed-and-tested-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This adds support to bind/unbind objects and wires it up. Objects are
only put into the ppgtt when necessary, i.e. at execbuf time.

Objects are still unconditionally put into the global gtt.

v2: Kill the quick hack and explicitly pass cache_level to ppgtt_bind
like for the global gtt function. Noticed by Chris Wilson.
Reviewed-by: NBen Widawsky <ben@bwidawsk.net>
Tested-by: NChris Wilson <chris@chris-wilson.co.uk>
Tested-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Reviewed-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

These are all user-trigerable, so tune down their loudness a notch.
For some of these we have i-g-t tests (because they prevent
newly-discovered bugs), without this patches running the test suite
leaves behind a dirty dmesg.
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This confuses our domain tracking and can (for gtt write domains) lead
to a subsequent oops.

Tested by tests/gem_exec_bad_domains from i-g-t.
Reviewed-by: NEric Anholt <eric@anholt.net>
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

In order to correctly account for reserving space in the GTT and fences
for a batch buffer, we need to independently track whether the fence is
pinned due to a fenced GPU access in the batch or whether the buffer is
pinned in the aperture. Currently we count the fenced as pinned if the
buffer has already been seen in the execbuffer. This leads to a false
accounting of available fence registers, causing frequent mass evictions.
Worse, if coupled with the change to make i915_gem_object_get_fence()
report EDADLK upon fence starvation, the batchbuffer can fail with only
one fence required...

Fixes intel-gpu-tools/tests/gem_fenced_exec_thrash

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=38735Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Tested-by: NPaul Neumann <paul104x@yahoo.de>
[danvet: Resolve the functional conflict with Jesse Barnes sprite
patches, acked by Chris Wilson on irc.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

... and add a helpr function for the places where we want a flag.

This way we can use ring->id to index into arrays.

v2: Resurrect the missing beautification-space Chris Wilson noted.
I'm moving this space around because I'll reuse ring_str in the next
patch.
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: NBen Widawsky <ben@bwidawsk.net>
Reviewed-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Sometimes it may be the case when we idle the gpu or wait on something
we don't actually want to process the retiring list. This patch allows
callers to choose the behavior.
Reviewed-by: NKeith Packard <keithp@keithp.com>
Reviewed-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

These registers are automatically incremented by the hardware during
transform feedback to track where the next streamed vertex output
should go.  Unlike the previous generation, which had a packet for
setting the corresponding registers to a defined value, gen7 only has
MI_LOAD_REGISTER_IMM to do so.  That's a secure packet (since it loads
an arbitrary register), so we need to do it from the kernel, and it
needs to be settable atomically with the batchbuffer execution so that
two clients doing transform feedback don't stomp on each others'
state.

Instead of building a more complicated interface involcing setting the
registers to a specific value, just set them to 0 when asked and
userland can tweak its pointers accordingly.
Signed-off-by: NEric Anholt <eric@anholt.net>
Reviewed-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Reviewed-by: NKenneth Graunke <kenneth@whitecape.org>
Signed-off-by: NKeith Packard <keithp@keithp.com>

The docs say this is required for Gen7, and since the bit was added for
Gen6, we are also setting it there pit pf paranoia. Particularly as
Chris points out, if PIPE_CONTROL counts as a 3d state packet.

This was found through doc inspection by Ken and applies to Gen6+;
Reported-by: NKenneth Graunke <kenneth@whitecape.org>
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: NEric Anholt <eric@anholt.net>
Signed-off-by: NKeith Packard <keithp@keithp.com>

dev_priv keeps track of the current addressing mode that gets set at
execbuffer time. Unfortunately the existing code was doing this before
acquiring struct_mutex which leaves a race with another thread also
doing an execbuffer. If that wasn't bad enough, relocate_slow drops
struct_mutex which opens a much more likely error where another thread
comes in and modifies the state while relocate_slow is being slow.

The solution here is to just defer setting this state until we
absolutely need it, and we know we'll have struct_mutex for the
remainder of our code path.

v2: Keith noticed a bug in the original patch.
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: NKeith Packard <keithp@keithp.com>

Semaphores still cause problems on some machines:

> From Udo Steinberg:
>
> With Linux-3.2-rc6 I'm frequently seeing GPU hangs when large amounts of
> text scroll in an xterm, such as when extracting a tar archive. Such as this
> one (note the timestamps):
>
>  I can reproduce it fairly easily with something
>  as simple as:
>
>	  while true; do dmesg; done

This patch turns them off on SNB while leaving them on for IVB.
Reported-by: NUdo Steinberg <udo@hypervisor.org>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: Eugeni Dodonov <eugeni@dodonov.net>
Signed-off-by: NKeith Packard <keithp@keithp.com>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

This adds a default setting for semaphores parameter, and enables
semaphores by default on IVB.

For now, as semaphores interaction with VTd causes random issues on
SNB, we do not enable them by default. But they can still be enabled
via the semaphores=1 kernel parameter.

v2: enables semaphores on SNB when IO remapping is disabled, with base
on Keith Packard patch.

CC: Daniel Vetter <daniel.vetter@ffwll.ch>
CC: Ben Widawsky <ben@bwidawsk.net>
CC: Keith Packard <keithp@keithp.com>
CC: Jesse Barnes <jbarnes@virtuousgeek.org>
CC: Chris Wilson <chris@chris-wilson.co.uk>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=42696
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=40564
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=41353
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=38862Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Signed-off-by: NKeith Packard <keithp@keithp.com>

While I think the previous code is correct, it was hard to follow and
hard to debug. Since we already have a ring abstraction, might as well
use it to handle the semaphore updates and compares.

I don't expect this code to make semaphores better or worse, but you
never know...

v2:
Remove magic per Keith's suggestions.
Ran Daniel's gem_ring_sync_loop test on this.

v3:
Ignored one of Keith's suggestions.

v4:
Removed some bloat per Daniel's recommendation.

Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Keith Packard <keithp@keithp.com>
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NKeith Packard <keithp@keithp.com>

This reverts commit 4a684a41.
Userland has always been required to set the object's domain to GTT
before using it through a GTT mapping, it's not something that the
kernel is supposed to enforce.  (The pagefault support is so that we
can handle multiple mappings without userland having to pin across
them, not so that userland can use GTT after GPU domains without
telling the kernel).

Fixes 19.2% +/- 0.8% (n=6) performance regression in cairo-gl
firefox-talos-gfx on my T420 latop.
Signed-off-by: NKeith Packard <keithp@keithp.com>

Along the fast path for relocation handling, we attempt to copy directly
from the user data structures whilst holding our mutex. This causes
lockdep to warn about circular lock dependencies if we need to pagefault
the user pages. [Since when handling a page fault on a mmapped bo, we
need to acquire the struct mutex whilst already holding the mm
semaphore, it is then verboten to acquire the mm semaphore when already
holding the struct mutex. The likelihood of the user passing in the
relocations contained in a GTT mmaped bo is low, but conceivable for
extreme pathology.] In order to force the mm to return EFAULT rather
than handle the pagefault, we therefore need to disable pagefaults
across the relocation fast path.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Cc: stable@kernel.org
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

... as if we are only reading from it, we can do that concurrently with
the queue flip.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>

Andi Kleen narrowed his GPU hangs on his Sugar Bay (SNB desktop) rev 09
down to the use of GPU semaphores, and we already know that they appear
broken up to Huron River (mobile) rev 08. (I'm optimistic that disabling
GPU semaphores is simply hiding another bug by the latency and
side-effects of the additional device interaction it introduces...)

However, use of semaphores is a massive performance improvement... Only
as long as the system remains stable. Enable at your peril.
Reported-by: NAndi Kleen <andi-fd@firstfloor.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=33921Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>

This seems to be running stably on my test laptop, so hopefully the
reported hangs where just symptoms of other bugs.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>

Userspace has a legitimate requirement to use a delta that points to
outside of the target bo, and so we need to enable this. (As this is an
abi break, albeit a relaxation of the current restrictions, mark the change
with a new flag.)
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>

The code paths for modesetting are growing in complexity as we may need
to move the buffers around in order to fit the scanout in the aperture.
Therefore we face a choice as to whether to thread the interruptible status
through the entire pinning and unbinding code paths or to add a flag to
the device when we may not be interrupted by a signal. This does the
latter and so fixes a few instances of modesetting failures under stress.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>

As we just need a temporary array whilst performing the relocations for
the execbuffer, first attempt to allocate using kmalloc even if it is
not of order page-0. This avoids the overhead of remapping the
discontiguous array and so gives a moderate boost to execution
throughput.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>

Dave Airlie spotted that we had a potential bug should we ever rearrange
the drm_i915_gem_object so not the base drm_gem_object was not its first
member. He noticed that we often convert the return of
drm_gem_object_lookup() immediately into drm_i915_gem_object and then
check the result for nullity. This is only valid when the base object is
the first member and so the superobject has the same address. Play safe
instead and use the compiler to convert back to the original return
address for sanity testing.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>

A lot of minor tweaks to fix the tracepoints, improve the outputting for
ftrace, and to generally make the tracepoints useful again. It is a start
and enough to begin identifying performance issues and gaps in our
coverage.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>

There are I915_NUM_RINGS-1 inter-ring synchronisation counters, but we
were clearing I915_NUM_RINGS of them. Oops.
Reported-by: NJiri Slaby <jirislaby@gmail.com>
Tested-by: NJiri Slaby <jirislaby@gmail.com>
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>

Move code around and invoke iomem annotation in a few more places in
order to silence sparse. Still a few more iomem annotations to go...
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>

Hopefully, this is a temporary measure whilst the root cause is
understood. At the moment, we experience a hard hang whilst looping
urbanterror that has been identified as a result of the use of
semaphores, but so far only on SNB mobile.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=32752
Tested-by: mengmeng.meng@intel.com
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>

After reordering the sequence of relocating objects, commit 6fe4f140,
we can no longer rely on seeing all reloc targets prior to performing
the relocation. As a result we were ignoring the need to flush objects
from the render cache and invalidate the sampler caches, resulting in
rendering glitches. So we need to clear the relocation domains earlier.
Reported-by: NLinus Torvalds <torvalds@linux-foundation.org>
Tested-by: NLinus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>

On the fault path, commit 6fe4f140 introduction a regression whereby it
changed the sequence of the objects but continued to use the original
ordering of relocation entries. The result was that incorrect GTT offsets
were being fed into the execbuffer causing lots of misrendering and
potential hangs.
Reported-by: NLinus Torvalds <torvalds@linux-foundation.org>
Tested-by: NLinus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>

As the mappable portion of the aperture is always a small subset at the
start of the GTT, it is allocated preferentially by drm_mm. This is
useful in case we ever need to map an object later. However, if you have
a large object that can consume the entire mappable region of the
GTT this prevents the batchbuffer from fitting and so causing an error.
Instead allocate all those that require a mapping up front in order to
improve the likelihood of finding sufficient space to bind them.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>

Before releasing the lock in order to copy the relocation list from user
pages, we need to drop all the object references as another thread may
usurp and execute another batchbuffer before we reacquire the lock.
However, the code was buggy and failed to clear the list...
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Cc: stable@kernel.org

... in order to avoid a BUG() and potential unbounded waits.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>

Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>

We need to ensure that writes through the GTT land before any
modification to the MMIO registers and so must impose a mandatory write
barrier when flushing the GTT domain. This was revealed by relaxing the
write ordering by experimentally mapping the registers and the GATT as
write-combining.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>

The relative-to-general state default is useless as it means having to
rewrite the streaming kernels for each batch. Relative-to-surface is
more useful, as that stream usually needs to be rewritten for each
batch. And absolute addressing mode, vital if you start streaming
state, is also only available by adjusting the register...
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>

Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>

As we provide a list of all objects that will be accessed from the
batchbuffer, we can build a lut of the handles associated with those
objects for this invocation and use that to avoid the overhead of
looking up those objects again for every relocation.

The cost of building and searching a small hash table is much less than
that of acquiring a spinlock, searching a radix tree and manipulating an
atomic refcnt per relocation.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>

Userspace should not have been declaring that it needed fenced GPU
access with gen4+ as those GPUs have no fenced commands, but to be on
the safe side it is easier to ignore userspace in case they did.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>

The bulk of the change is to convert the growing list of rings into an
array so that the relationship between the rings and the semaphore sync
registers can be easily computed.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>