We can also take advantage of the new 'no retire' mode for seqno waiting
to avoid having to take a reference on the old fence object whilst
flushing an existing fence.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Now that we have a routine that is able to clear the fences as well as
setup up the register for a tiled object, remove the surplus routines to
clear the fences.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
One clarification that we make is to the existing semantics of
obj->tiling_changed to only mean that we need to update an associated
fence register (including the NO_FENCE when executing an untiled but
fenced GPU command). If we do not have a fence register or pending
fenced GPU access for the object (after put_fence() for example), then
we can clear the tiling_changed flag as any fence will necessarily be
rewritten upon acquisition.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Update the existing architecture specific fence writing routines to
either update the fence to point to a tiled object or to clear them in
preparation to remove the other fence writing routes.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
As i915_wait_request() will first check for an already passed seqno,
doing it also in the caller is a waste of space for a cold path.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
As the fences are stored in LRU order, we can simply reuse the oldest if
we do not have an unused register.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
As we now never pipeline a fence update, obj->last_fenced_ring is always
the same as the obj->ring whenever obj->last_fenced_seqno is active, so
remove it.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
As we now no longer track a pipelined fence change, we never use
ring->setup_seqno and can kill it.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Step 2 is then to replace the pipelined parameter with NULL and perform
constant folding to remove dead code.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We never succeeded in getting pipelined fencing to work (unresolved
spurious GPU hangs), so begin the process of dismantling and removal
the broken code.
Step 1 is the removal of the pipeline parameter to get_fence().
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
During modeset we have to disable the pipe to reconfigure its timings
and maybe its size. Userspace may have queued up command buffers that
depend upon the pipe running in a certain configuration and so the
commands may become confused across the modeset. At the moment, we use a
less than satisfactory kick-scanline-waits should the GPU hang during
the modeset. It should be more reliable to wait for the pending
operations to complete first, even though we still have a window for
userspace to submit a broken command buffer during the modeset.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
On gen2 MI_EXE_FLUSH is actually an AGP flush bit and on gen3 marked as
reserved. On both it is documented as being must-be-zero. So obey the
documentation, and separate the gen2 flush into its own little routine
and share with gen3.
This means that we can rename the existing render_ring_flush() to
reflect the generation from which it first applies and remove the code
for handling earlier generations from it.
v2: Applies to gen3 as well
v3: Make it compile and improve the commit message.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
As we need to manipulate our device structure and allocate queue a task,
it is no longer a simple atomic operation and cannot be performed along
the atomic modeset paths. Instead make sure that we disable FBC (which
must be therefore kept as a set of simple register writes) when
performing the atomic modeset and leave the heavy-weight
intel_update_fbc() for the normal modeset.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
This originally started as a patch from Bernard as a way of simply
setting the VS scheduler. After submitting the RFC patch, we decided to
also modify the DS scheduler. To be most explicit, I've made the patch
explicitly set all scheduler modes, and included the defines for other
modes (in case someone feels frisky later).
The rest of the story gets a bit weird. The first version of the patch
showed an almost unbelievable performance improvement. Since rebasing my
branch it appears the performance improvement has gone, unfortunately.
But setting these bits seem to be the right thing to do given that the
docs describe corruption that can occur with the default settings.
In summary, I am seeing no more perf improvements (or regressions) in my
limited testing, but we believe this should be set to prevent rendering
corruption, therefore cc stable.
v1: Clear bit 4 also (Ken + Eugeni)
Do a full clear + set of the bits we want (Me).
Cc: Bernard Kilarski <bernard.r.kilarski@intel.com>
Cc: stable <stable@vger.kernel.org>
Reviewed-by (RFC): Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
Reviewed-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The (2<<6) virtual memory space selector harks back to gen3 and is
mandatory given our use of GTT space for batchbuffers. On gen4+, use of
the GTT became mandatory and bit6 marked reserved. However the code must
now explicitly set (1<<7), which conveniently is also (2<<6).
To clarify the meaning for future readers, replace the open coded (2<<6)
with MI_BATCH_GTT.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
As we defer updating the fence register from set-tiling to the point of
use, we need to declare every access through the GTT as either fenced or
unfenced.
This patches fixes an old bug in the execbuffer relocation processing
which could conceivably be hit by a pathological userspace.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Sparse doesn't like:
"error: bad constant expression"
Signed-off-by: Ben Widawsky <benjamin.widawsky@intel.com>
[danvet: apply s/drm_malloc_ab/kcalloc bikeshed. If it's small enough
for the stack, it's small enough for kmalloc.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
When the PCH split occurred, hw dropped support for separate hsync and
vsync disable in the VGA DAC. So add a PCH specific DPMS function that
just uses the port enable bit for controlling DPMS states.
Before this fix, when anything other than a full DPMS off occurred,
the VGA port would be left enabled and scanning out while all the other
heads would turn off as expected.
v2: duplicate encoder helper vtable into pch and gmch versions (Daniel)
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=48491
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
[danvet: s/intel_crt_dpms/gmch_crt_dpms as suggested by Chris.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Not only do the pageflip work without it at non-native modes (i.e. with
the panel fitter enabled), it also causes normal (non-pageflipped)
modesets to fail.
Reported-by: Adam Jackson <ajax@redhat.com>
Tested-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Wanted-by-for-fixes: Dave Airlie <airlied@gmail.com>
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The purpose of this patch is to avoid zeroing the lower 12 reserved bits
of surface base address registers (framebuffer & sprite). There are bits
in that range that may occasionally be set by BIOS or by other components.
Signed-off-by: Armin Reese <armin.c.reese@intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
