In the future, we need to call allocate_va_range on the aliasing-ppgtt
which means moving the call down from the vma into the vm (which is
more appropriate for calling the vm function).
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170215084357.19977-8-chris@chris-wilson.co.uk
As these are now both plain and simple kmap_atomic/kunmap_atomic pairs,
we can remove the wrappers for a small gain of clarity (in particular,
not hiding the atomic critical sections!).
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170215084357.19977-7-chris@chris-wilson.co.uk
We flush the entire page every time we update a few bytes, making the
update of a page table many, many times slower than is required. If we
create a WC map of the page for our updates, we can avoid the clflush
but incur additional cost for creating the pagetable. We amoritize that
cost by reusing page vmappings, and only changing the page protection in
batches.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Improve the sg iteration and in hte process eliminate a bug in
miscomputing the pml4 length as orig_nents<<PAGE_SHIFT is no longer the
full length of the sg table.
v2: Check for the end of the fourth level page table (the final pdpe)
and move onto the next.
v3: Assert that 3lvl insert_pte_entries doesn't overflow its smaller set
of PDP.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170215084357.19977-3-chris@chris-wilson.co.uk
Inline the address computation to avoid the vfunc call for every page.
We still have to pay the high overhead of sg_page_iter_next(), but now
at least GCC can optimise the inner most loop, giving a significant
boost to some thrashing Unreal Engine workloads.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170215084357.19977-2-chris@chris-wilson.co.uk
If there is no OPREGION_ASLE_EXT then a VBT stored in mailbox #4 may
use the ASLE_EXT parts of the opregion. Adjust the vbt_size calculation
for a vbt in mailbox #4 for this.
This fixes the driver not finding the VBT on a jumper ezpad mini3
cherrytrail tablet and on a ACER SW5_017 machine.
Cc: stable@vger.kernel.org
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1487088758-30050-1-git-send-email-jani.nikula@intel.com
Currently we apply the jump to rpe if we are below it and the GPU needs
more power. For some GPUs, the rpe is 75% of the maximum range causing
us to dramatically overshoot low power applications *and* unable to
reach the low frequency that can most efficiently deliver their
workload.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170210150348.22146-3-chris@chris-wilson.co.uk
Reviewed-by: Radoslaw Szwichtenberg <radoslaw.szwichtenberg@intel.com>
If we receive a DOWN_TIMEOUT rps interrupt, we respond by reducing the
GPU clocks significantly. Before we do, double check that the frequency
we pick is actually a decrease.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170210150348.22146-2-chris@chris-wilson.co.uk
Reviewed-by: Radoslaw Szwichtenberg <radoslaw.szwichtenberg@intel.com>
When the RPS tuning was applied to Baytrail, in commit 8fb55197e6
("drm/i915: Agressive downclocking on Baytrail"), concern was given that
it might cause Cherryview excess wakeups of the common power well.
However, the static thresholds perform poorly for Kodi, and the GPU is
unable to deliver the video frames on time. Enabling the dynamic, finer
thresholds used on all other platforms (including Skylake and Broxton
that also have the same multiple powerwell concerns) allows the GPU to
pick a more appropriate frequency and not drop frames.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170210150348.22146-1-chris@chris-wilson.co.uk
Reviewed-by: Radoslaw Szwichtenberg <radoslaw.szwichtenberg@intel.com>
Once upon a time before we had automated GPU state capture upon hangs,
we had intel_gpu_dump. Now we come almost full circle and reinstate that
view of the current GPU queues and registers by using the error capture
facility to snapshot the GPU state when debugfs/.../i915_gpu_info is
opened - which should provided useful debugging to both the error
capture routines (without having to cause a hang and avoid the error
state being eaten by igt) and generally.
v2: Rename drm_i915_error_state to i915_gpu_state to alleviate some name
collisions between the error state dump and inspecting the gpu state.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170214164611.11381-1-chris@chris-wilson.co.uk
We no longer need to take the struct_mutex for freeing objects, and on
the finalisation paths here the mutex is not been used for serialisation
of the pointer access, so remove the BKL wart.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170214133420.7977-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
In general, the compiler should not be able to detect if we do any
passes through the test loops:
In file included from drivers/gpu/drm/i915/i915_gem.c:5029:
drivers/gpu/drm/i915/selftests/i915_gem_coherency.c: In function 'igt_gem_coherency':
drivers/gpu/drm/i915/selftests/i915_gem_coherency.c:274: error: 'err' may be used uninitialized in this function
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170214143509.15719-1-chris@chris-wilson.co.uk
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
gcc-4.7 spotted that
In file included from drivers/gpu/drm/i915/i915_gem_gtt.c:3791:0:
drivers/gpu/drm/i915/selftests/i915_gem_gtt.c: In function ‘pot_hole’:
drivers/gpu/drm/i915/selftests/i915_gem_gtt.c:594:6: error: ‘err’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
So set it to 0 should we ever skip over a hole smaller than a few pages.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170214113756.27834-1-chris@chris-wilson.co.uk
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
When using the mock_ppgtt selftest, the GTT is large enough to cause an
overflow in pot_hole() when adding 2 pages to the address. Avoid the
overflow by computing the final valid address and iterating up to that
address.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170214092344.12330-1-chris@chris-wilson.co.uk
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
With latest Punit FW, vgg input voltag drop falling to minimum is fixed.
So reverting the WA patch & moving to turbo freq opreation range to [RPn -> RP0]
This is not a 1:1 revert of the commit 5b7c91b78b.
You can refer to commit 5b5929cbe3 ("drm/i915/chv: remove
pre-production hardware workarounds") as the reason for the discrepancy
commit 5b7c91b78b
Author: Deepak S <deepak.s@linux.intel.com>
Date: Sat May 9 18:15:46 2015 +0530
drm/i915/chv: Set min freq to efficient frequency on chv
v2: Fix inconsistent return type. (Chris)
v3: drop pre-production hw case (Ville)
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1471007801-86075-1-git-send-email-deepak.s@linux.intel.com
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Dump out more of the DSI configuration details during init.
This includes pclk, burst_mode_ratio, lane_count, pixel_overlap,
video_mode_format and reset_timer_val.
v2: Dump more info (Chris)
v3: Use the VIDEO_MODE_ defines for consistency (Chris)
Dump dphy_reg too (Chris)
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20161221143114.23530-1-ville.syrjala@linux.intel.com
This workaround for BDW was incomplete as it also requires EUTC clock
gating to be disabled via UCGCTL1.
v2: read modify write UCGTL1 in broadwell_init_clock_gating (Ville)
Signed-off-by: Robert Bragg <robert@sixbynine.org>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170212133252.20990-1-robert@sixbynine.org
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
For some reason my compiler (and CI as well) failed to spot the
uninitialized ret in mi_set_context.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Fixes: 73dec95e6b ("drm/i915: Emit to ringbuffer directly")
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170214152901.20361-1-tvrtko.ursulin@linux.intel.com
This removes the usage of intel_ring_emit in favour of
directly writing to the ring buffer.
intel_ring_emit was preventing the compiler for optimising
fetch and increment of the current ring buffer pointer and
therefore generating very verbose code for every write.
It had no useful purpose since all ringbuffer operations
are started and ended with intel_ring_begin and
intel_ring_advance respectively, with no bail out in the
middle possible, so it is fine to increment the tail in
intel_ring_begin and let the code manage the pointer
itself.
Useless instruction removal amounts to approximately
two and half kilobytes of saved text on my build.
Not sure if this has any measurable performance
implications but executing a ton of useless instructions
on fast paths cannot be good.
v2:
* Change return from intel_ring_begin to error pointer by
popular demand.
* Move tail increment to intel_ring_advance to enable some
error checking.
v3:
* Move tail advance back into intel_ring_begin.
* Rebase and tidy.
v4:
* Complete rebase after a few months since v3.
v5:
* Remove unecessary cast and fix !debug compile. (Chris Wilson)
v6:
* Make intel_ring_offset take request as well.
* Fix recording of request postfix plus a sprinkle of asserts.
(Chris Wilson)
v7:
* Use intel_ring_offset to get the postfix. (Chris Wilson)
* Convert GVT code as well.
v8:
* Rename *out++ to *cs++.
v9:
* Fix GVT out to cs conversion in GVT.
v10:
* Rebase for new intel_ring_begin in selftests.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Zhi Wang <zhi.a.wang@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
I screwed up the rebase of commit d8fc70b736 ("drm/i915: Make power
domain masks 64 bit long") before sending v2, causing a couple of
conversions from 32 to 64 bit masks to be lost.
Fixes: d8fc70b736 ("drm/i915: Make power domain masks 64 bit long")
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Cc: Daniel Vetter <daniel.vetter@intel.com>
Cc: intel-gfx@lists.freedesktop.org
Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213145733.8779-1-ander.conselvan.de.oliveira@intel.com
The i915_gem_object_wait_fence() uses an incoming timeout=0 to query
whether the current fence is busy or idle, without waiting. This can be
used by the wait-ioctl to implement a busy query.
Fixes: e95433c73a ("drm/i915: Rearrange i915_wait_request() accounting with callers")
Testcase: igt/gem_wait/basic-busy-write-all
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.william.auld@gmail.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: <drm-intel-fixes@lists.freedesktop.org> # v4.10-rc1+
Cc: stable@vger.kernel.org
Link: http://patchwork.freedesktop.org/patch/msgid/20170212215344.16600-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Explicitly disable stolen memory when running as a guest in a virtual
machine, since the memory is not mediated between clients and reserved
entirely for the host. The actual size should be reported as zero, but
like every other quirk we want to tell the user what is happening.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=99028
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Zhenyu Wang <zhenyuw@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161109103905.17860-1-chris@chris-wilson.co.uk
Reviewed-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Cc: stable@vger.kernel.org
As the page-table trees within the GTT are naturally aligned to
power-of-two boundaries, by inserting an object that crosses a
power-of-two (and the power-of-two intervals) we can quickly check the
code for errors in switching between levels in the tree.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-46-chris@chris-wilson.co.uk
Check we can create and execution within a context.
v2: Write one set of dwords through each context/engine to exercise more
contexts within the same time period.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-38-chris@chris-wilson.co.uk
It is possible whilst allocating the page-directory tree for a ppgtt
bind that the shrinker may run and reap unused parts of the tree. If the
shrinker happens to remove a chunk of the tree that the
allocate_va_range has already processed, we may then try to insert into
the dangling tree. This test uses the fault-injection framework to force
the shrinker to be invoked before we allocate new pages, i.e. new chunks
of the PD tree.
References: https://bugs.freedesktop.org/show_bug.cgi?id=99295
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Allocate objects with varying number of pages (which should hopefully
consist of a mixture of contiguous page chunks and so coalesced sg
lists) and check that the sg walkers in insert_pages cope.
v2: Check both small <-> large
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-28-chris@chris-wilson.co.uk
Add a late selftest that walks over all forcewake registers (those below
0x40000) and uses the mmio debug register to check to see if any are
unclaimed. This is possible if we fail to wake the appropriate
powerwells for the register.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-24-chris@chris-wilson.co.uk