If the user requires patching of their batch or auxiliary buffers, we
currently make the alterations on the cpu. If they are active on the GPU
at the time, we wait under the struct_mutex for them to finish executing
before we rewrite the contents. This happens if shared relocation trees
are used between different contexts with separate address space (and the
buffers then have different addresses in each), the 3D state will need
to be adjusted between execution on each context. However, we don't need
to use the CPU to do the relocation patching, as we could queue commands
to the GPU to perform it and use fences to serialise the operation with
the current activity and future - so the operation on the GPU appears
just as atomic as performing it immediately. Performing the relocation
rewrites on the GPU is not free, in terms of pure throughput, the number
of relocations/s is about halved - but more importantly so is the time
under the struct_mutex.
v2: Break out the request/batch allocation for clearer error flow.
v3: A few asserts to ensure rq ordering is maintained
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Currently, the last object in the execlist is the always the batch.
However, when building the batch buffer we often know the batch object
first and if we can use the first slot in the execlist we can emit
relocation instructions relative to it immediately and avoid a separate
pass to adjust the relocations to point to the last execlist slot.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
This simply hides the EAGAIN caused by userptr when userspace causes
resource contention. However, it is quite beneficial with highly
contended userptr users as we avoid repeating the setup costs and
kernel-user context switches.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>
When choosing a slot for an execbuffer, we ideally want to use the same
address as last time (so that we don't have to rebind it) and the same
address as expected by the user (so that we don't have to fixup any
relocations pointing to it). If we first try to bind the incoming
execbuffer->offset from the user, or the currently bound offset that
should hopefully achieve the goal of avoiding the rebind cost and the
relocation penalty. However, if the object is not currently bound there
we don't want to arbitrarily unbind an object in our chosen position and
so choose to rebind/relocate the incoming object instead. After we
report the new position back to the user, on the next pass the
relocations should have settled down.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtien@linux.intel.com>
If we take a reference to the object/vma when it is first used in an
execbuf, we can keep that reference until the object's file-local handle
is closed. Thereby saving a frequent ref/unref pair.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
The major scaling bottleneck in execbuffer is the processing of the
execobjects. Creating an auxiliary list is inefficient when compared to
using the execobject array we already have allocated.
Reservation is then split into phases. As we lookup up the VMA, we
try and bind it back into active location. Only if that fails, do we add
it to the unbound list for phase 2. In phase 2, we try and add all those
objects that could not fit into their previous location, with fallback
to retrying all objects and evicting the VM in case of severe
fragmentation. (This is the same as before, except that phase 1 is now
done inline with looking up the VMA to avoid an iteration over the
execobject array. In the ideal case, we eliminate the separate reservation
phase). During the reservation phase, we only evict from the VM between
passes (rather than currently as we try to fit every new VMA). In
testing with Unreal Engine's Atlantis demo which stresses the eviction
logic on gen7 class hardware, this speed up the framerate by a factor of
2.
The second loop amalgamation is between move_to_gpu and move_to_active.
As we always submit the request, even if incomplete, we can use the
current request to track active VMA as we perform the flushes and
synchronisation required.
The next big advancement is to avoid copying back to the user any
execobjects and relocations that are not changed.
v2: Add a Theory of Operation spiel.
v3: Fall back to slow relocations in preparation for flushing userptrs.
v4: Document struct members, factor out eb_validate_vma(), add a few
more comments to explain some magic and hide other magic behind macros.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
If we write a relocation into the buffer, we require our own implicit
synchronisation added after the start of the execbuf, outside of the
user's control. As we may end up clflushing, or doing the patch itself
on the GPU, asynchronously we need to look at the implicit serialisation
on obj->resv and hence need to disable EXEC_OBJECT_ASYNC for this
object.
If the user does trigger a stall for relocations, we make sure the stall
is complete enough so that the batch is not submitted before we complete
those relocations.
Fixes: 77ae995789 ("drm/i915: Enable userspace to opt-out of implicit fencing")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
We can simplify our tracking of pending writes in an execbuf to the
single bit in the vma->exec_entry->flags, but that requires the
relocation function knowing the object's vma. Pass it along.
Note we have only been using a single bit to track flushing since
commit cc889e0f6c
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date: Wed Jun 13 20:45:19 2012 +0200
drm/i915: disable flushing_list/gpu_write_list
unconditionally flushed all render caches before the breadcrumb and
commit 6ac42f4148
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date: Sat Jul 21 12:25:01 2012 +0200
drm/i915: Replace the complex flushing logic with simple invalidate/flush all
did away with the explicit GPU domain tracking. This was then codified
into the ABI with NO_RELOC in
commit ed5982e6ce
Author: Daniel Vetter <daniel.vetter@ffwll.ch> # Oi! Patch stealer!
Date: Thu Jan 17 22:23:36 2013 +0100
drm/i915: Allow userspace to hint that the relocations were known
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
The advent of full-ppgtt lead to an extra indirection between the object
and its binding. That extra indirection has a noticeable impact on how
fast we can convert from the user handles to our internal vma for
execbuffer. In order to bypass the extra indirection, we use a
resizable hashtable to jump from the object to the per-ctx vma.
rhashtable was considered but we don't need the online resizing feature
and the extra complexity proved to undermine its usefulness. Instead, we
simply reallocate the hastable on demand in a background task and
serialize it before iterating.
In non-full-ppgtt modes, multiple files and multiple contexts can share
the same vma. This leads to having multiple possible handle->vma links,
so we only use the first to establish the fast path. The majority of
buffers are not shared and so we should still be able to realise
speedups with multiple clients.
v2: Prettier names, more magic.
v3: Many style tweaks, most notably hiding the misuse of execobj[].rsvd2
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
For ease of use (i.e. avoiding a few checks and function calls), store
the object's cache coherency next to the cache is dirty bit.
Specifically this patch aims to reduce the frequency of no-op calls to
i915_gem_object_clflush() to counter-act the increase of such calls for
GPU only objects in the previous patch.
v2: Replace cache_dirty & ~cache_coherent with cache_dirty &&
!cache_coherent as gcc generates much better code for the latter
(Tvrtko)
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Dongwon Kim <dongwon.kim@intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Tested-by: Dongwon Kim <dongwon.kim@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170616105455.16977-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Currently, we only mark the CPU cache as dirty if we skip a clflush.
This leads to some confusion where we have to ask if the object is in
the write domain or missed a clflush. If we always mark the cache as
dirty, this becomes a much simply question to answer.
The goal remains to do as few clflushes as required and to do them as
late as possible, in the hope of deferring the work to a kthread and not
block the caller (e.g. execbuf, flips).
v2: Always call clflush before GPU execution when the cache_dirty flag
is set. This may cause some extra work on llc systems that migrate dirty
buffers back and forth - but we do try to limit that by only setting
cache_dirty at the end of the gpu sequence.
v3: Always mark the cache as dirty upon a level change, as we need to
invalidate any stale cachelines due to external writes.
Reported-by: Dongwon Kim <dongwon.kim@intel.com>
Fixes: a6a7cc4b7d ("drm/i915: Always flush the dirty CPU cache when pinning the scanout")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Dongwon Kim <dongwon.kim@intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Tested-by: Dongwon Kim <dongwon.kim@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170615123850.26843-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
i915_vma_destroy() is now not used outside of i915_vma.c so we can
remove the export and make the function static.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170616123508.12673-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Attach the tv_format property to the SDVO connector instead of passing
a '0' in place of the pointer to the property. This got broken when
the SDVO connector properties were converted to atomic.
We can thank sparse for catching this:
drivers/gpu/drm/i915/intel_sdvo.c:2742:75: warning: Using plain integer as NULL pointer
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Fixes: 630d30a4ee ("drm/i915: Convert intel_sdvo connector properties to atomic.")
Link: http://patchwork.freedesktop.org/patch/msgid/20170615172308.10121-1-ville.syrjala@linux.intel.com
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
gvt-next-2017-06-08
First gvt-next pull for 4.13:
- optimization for per-VM mmio save/restore (Changbin)
- optimization for mmio hash table (Changbin)
- scheduler optimization with event (Ping)
- vGPU reset refinement (Fred)
- other misc refactor and cleanups, etc.
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170608093547.bjgs436e3iokrzdm@zhen-hp.sh.intel.com
This reverts commit bb9d85f6e9.
New ddb allocation algorithm is a show stopper on my SKL system.
Besides not be able to get external DP 4k@60 (through USB type C),
It fully hang my screen when unplugging the USB type C.
Bugzilla: https://patchwork.freedesktop.org/patch/161571/
Fixes: bb9d85f6e9 ("drm/i915/skl: New ddb allocation algorithm")
Cc: Mahesh Kumar <mahesh1.kumar@intel.com>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1497376350-3400-1-git-send-email-rodrigo.vivi@intel.com
As per BSEPC, if device ready bit is '0' in enable IO sequence
then its a cold boot/reset scenario eg: S3/S4 resume. If cold boot
scenario detected in enable IO, then prepare port immediately.
In normal boot scenario, prepare port after glk_dsi_device_ready().
Without cold boot sequence enabled, features like S3/S4 doesn't work.
Signed-off-by: Madhav Chauhan <madhav.chauhan@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1497340095-5877-2-git-send-email-madhav.chauhan@intel.com
This patch divides glk_dsi_device_ready() function into
two part. First part will program LP wake and MIPI DSI mode
to MIPI_CTRL reg using newly defined function glk_dsi_enable_io().
glk_dsi_enable_io() will be called from intel_dsi_pre_enable.
Second part will do remaining device ready activities using
the existing function glk_dsi_device_ready().
Signed-off-by: Madhav Chauhan <madhav.chauhan@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1497340095-5877-1-git-send-email-madhav.chauhan@intel.com
Maarten and Ville noticed that we are enabling backlight via DP aux very
early in the modeset_init path via the intel_dp_aux_setup_backlight()
function, since commit e7156c8339 ("drm/i915: Add Backlight Control using
DPCD for eDP connectors (v9)"). Looks like all we need to do during
_setup_backlight() is read the current brightness state instead of
modifying it.
v2: Rewrote commit message.
Cc: Ville Syrjala <ville.syrjala@linux.intel.com>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Jani Nikula <jani.nikula@intel.com>
Cc: Yetunde Adebisi <yetundex.adebisi@intel.com>
Signed-off-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Acked-by: Jani Nikula <jani.nikula@intel.com>
Tested-by: Puthikorn Voravootivat <puthik@chromium.org>
Fixes: e7156c8339 ("drm/i915: Add Backlight Control using DPCD for eDP connectors (v9)")
Link: http://patchwork.freedesktop.org/patch/msgid/1497384239-2965-1-git-send-email-dhinakaran.pandiyan@intel.com
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
The function cnl_ddi_dp_set_dpll_hw_state does not need to be in global
scope, so make it static.
Cleans up sparse warning:
"symbol 'cnl_ddi_dp_set_dpll_hw_state' was not declared. Should it
be static?"
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170613134751.29196-1-colin.king@canonical.com
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
With 830 the only thing needing pipe quirks, we can just drop the quirk
defines and replace the checks with IS_I830() checks.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170601143619.27840-8-ville.syrjala@linux.intel.com
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
The pipe A force quirk shouldn't needed except on 830. So let's nuke it
for the IBM Thinkpad T60 945 machines. This quirk pre-dates
KMS so it's usefulness is doubtful at best now.
The original bug report [1] describes the symptoms as "system hang on
closing T60 panel lid", and we already dropped a similar quirk for
another 945 machine in
commit 736a69ca8c ("drm/i915: Drop PIPE-A quirk for 945GSE HP Mini")
so I'm hopeful we can drop this one as well.
The quirk was added into xf86-video-intel in
commit 08903abe4dc0 ("Add pipe a force enable quirk for Lenovo T60")
[1] https://bugs.freedesktop.org/show_bug.cgi?id=16494
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170601143619.27840-7-ville.syrjala@linux.intel.com
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
The pipe A force quirk shouldn't needed except on 830. So let's nuke it
for the Toshiba Protege R-205/S-209 945 machines. This quirk pre-dates
KMS so it's usefulness is doubtful at best now.
Unfortunately the original bug report [1] isn't very helpful since it
doesn't describe the symptoms. And the commit message in xf86-video-intel
commit ecdb5963ef68 ("Add pipe A force enable quirk for Toshiba Portege R205-S209")
is not much help either.
However, if we assume the problem was the typical "closing the lid
hangs the box" type of thing, we already nuked the quirk for another
945 machine in
commit 736a69ca8c ("drm/i915: Drop PIPE-A quirk for 945GSE HP Mini")
and so I hope we can drop this one as well.
[1] https://bugs.freedesktop.org/show_bug.cgi?id=14944
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170601143619.27840-6-ville.syrjala@linux.intel.com
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
830 more or less requires both pipes and DPLLs to remain on as long
as either pipe is needed. However, when neither pipe is actually needed,
we can save a bit of power by turning everything off. To do that we add
a new "power well" that turns both pipes and DPLLs on and off in the
right order. Seems to save ~50mW on my Fujitsu-Siemens Lifebook S6010.
This also avoids having to abuse the load detection to force pipe A on
at init time. That was never very robust, and it only worked for one
pipe, whereas 830 really needs both pipes enabled. As a bonus the 830
pipe quirk is now a bit more isolated from the rest of the mode setting
infrastructure, which should mean that it's much less likely someone
will accidentally break it in the future. The extra cost is of course
slight code duplication, but that seems like a worthwile tradeoff here.
v2; s/BIT/BIT_ULL/
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170601143619.27840-5-ville.syrjala@linux.intel.com
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
The magic "enable the DPLL three times" sequence feels like it
deserves a loop.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170601143619.27840-4-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
If intel_crtc_disable_noatomic() were to ever get called during resume
we'd end up deadlocking since resume has its own acqcuire_ctx but
intel_crtc_disable_noatomic() still tries to use the
mode_config.acquire_ctx. Pass down the correct acquire ctx from the top.
Cc: stable@vger.kernel.org
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Fixes: e2c8b8701e ("drm/i915: Use atomic helpers for suspend, v2.")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170601143619.27840-3-ville.syrjala@linux.intel.com
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Pass down the correct acquire context to the pipe A quirk load detect
hack during display resume. Avoids deadlocking the entire thing.
Cc: stable@vger.kernel.org
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Fixes: e2c8b8701e ("drm/i915: Use atomic helpers for suspend, v2.")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170601143619.27840-2-ville.syrjala@linux.intel.com
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Currently the vma has one link member that is used for both holding its
place in the execbuf reservation list, and in any eviction list. This
dual property is quite tricky and error prone.
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/20170615081435.17699-3-chris@chris-wilson.co.uk
This has the benefit of not requiring us to manipulate the
vma->exec_link list when tearing down the execbuffer, and is a
marginally cheaper test to detect the user error.
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/20170615081435.17699-2-chris@chris-wilson.co.uk
Add OA support for Kabylake (pretty much identical to Skylake), and
also add the associated OA configurations.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Add macros to detect GT2/GT3 skus so we can apply the proper OA
configuration later.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
In earlier iterations of the i915-perf driver we had a number of
callbacks/hooks from other parts of the i915 driver to e.g. notify us
when a legacy context was pinned and these could run asynchronously with
respect to the stream file operations and might also run in atomic
context.
dev_priv->perf.hook_lock had been for serialising access to state needed
within these callbacks, but as the code has evolved some of the hooks
have gone away or are implemented to avoid needing to lock any state.
The remaining use of this lock was actually redundant considering how
the gen7 oacontrol state used to be updated as part of a context pin
hook.
Signed-off-by: Robert Bragg <robert@sixbynine.org>
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
An oa_exponent_to_ns() utility and per-gen timebase constants where
recently removed when updating the tail pointer race condition WA, and
this restores those so we can update the _PROP_OA_EXPONENT validation
done in read_properties_unlocked() to not assume we have a 12.5MHz
timebase as we did for Haswell.
Accordingly the oa_sample_rate_hard_limit value that's referenced by
proc_dointvec_minmax defining the absolute limit for the OA sampling
frequency is now initialized to (timestamp_frequency / 2) instead of the
6.25MHz constant for Haswell.
v2:
Specify frequency of 19.2MHz for BXT (Ville)
Initialize oa_sample_rate_hard_limit per-gen too (Lionel)
Signed-off-by: Robert Bragg <robert@sixbynine.org>
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
These are auto generated from an XML description of metric sets,
currently maintained in gputop, ref:
https://github.com/rib/gputop
> gputop-data/oa-*.xml
> scripts/i915-perf-kernelgen.py
$ make -C gputop-data -f Makefile.xml
Signed-off-by: Robert Bragg <robert@sixbynine.org>
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Enables access to OA unit metrics for BDW, CHV, SKL and BXT which all
share (more-or-less) the same OA unit design.
Of particular note in comparison to Haswell: some OA unit HW config
state has become per-context state and as a consequence it is somewhat
more complicated to manage synchronous state changes from the cpu while
there's no guarantee of what context (if any) is currently actively
running on the gpu.
The periodic sampling frequency which can be particularly useful for
system-wide analysis (as opposed to command stream synchronised
MI_REPORT_PERF_COUNT commands) is perhaps the most surprising state to
have become per-context save and restored (while the OABUFFER
destination is still a shared, system-wide resource).
This support for gen8+ takes care to consider a number of timing
challenges involved in synchronously updating per-context state
primarily by programming all config state from the cpu and updating all
current and saved contexts synchronously while the OA unit is still
disabled.
The driver intentionally avoids depending on command streamer
programming to update OA state considering the lack of synchronization
between the automatic loading of OACTXCONTROL state (that includes the
periodic sampling state and enable state) on context restore and the
parsing of any general purpose BB the driver can control. I.e. this
implementation is careful to avoid the possibility of a context restore
temporarily enabling any out-of-date periodic sampling state. In
addition to the risk of transiently-out-of-date state being loaded
automatically; there are also internal HW latencies involved in the
loading of MUX configurations which would be difficult to account for
from the command streamer (and we only want to enable the unit when once
the MUX configuration is complete).
Since the Gen8+ OA unit design no longer supports clock gating the unit
off for a single given context (which effectively stopped any progress
of counters while any other context was running) and instead supports
tagging OA reports with a context ID for filtering on the CPU, it means
we can no longer hide the system-wide progress of counters from a
non-privileged application only interested in metrics for its own
context. Although we could theoretically try and subtract the progress
of other contexts before forwarding reports via read() we aren't in a
position to filter reports captured via MI_REPORT_PERF_COUNT commands.
As a result, for Gen8+, we always require the
dev.i915.perf_stream_paranoid to be unset for any access to OA metrics
if not root.
v5: Drain submitted requests when enabling metric set to ensure no
lite-restore erases the context image we just updated (Lionel)
v6: In addition to drain, switch to kernel context & update all
context in place (Chris)
v7: Add missing mutex_unlock() if switching to kernel context fails
(Matthew)
v8: Simplify OA period/flex-eu-counters programming by using the
batchbuffer instead of modifying ctx-image (Lionel)
v9: Back to updating the context image (due to erroneous testing,
batchbuffer programming the OA unit doesn't actually work)
(Lionel)
Pin context before updating context image (Chris)
Drop MMIO programming now that we switch to a kernel context with
right values in initial context image (Chris)
v10: Just pin_map the contexts we want to modify or let the
configuration happen on first use (Chris)
v11: Update kernel context OA config through the batchbuffer rather
than on the fly ctx-image update (Lionel)
v12: Rework OA context registers update again by swithing away from
user contexts and reconfiguring the kernel context through the
batchbuffer and updating all the other contexts' context image.
Also take care to lock slice/subslice configuration when OA is
on. (Lionel)
v13: Request rpcs updates on all engine when updating the OA config
(Lionel)
v14: Drop any kind of rpcs management now that we monitor sseu
configuration changes in a later patch (Lionel)
Remove usleep after programming the NOA configs on Gen8+, this
doesn't seem to be needed (Lionel)
v15: Respect coding style for block comments (Chris)
v16: Add missing i915_add_request() in case we fail to emit OA
configuration (Matthew)
Signed-off-by: Robert Bragg <robert@sixbynine.org>
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com> \o/
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Adds a static OA unit, MUX, B Counter + Flex EU configurations for basic
render metrics on Broadwell, Cherryview, Skylake and Broxton. These are
auto generated from an XML description of metric sets, currently
maintained in gputop, ref:
https://github.com/rib/gputop
> gputop-data/oa-*.xml
> scripts/i915-perf-kernelgen.py
$ make -C gputop-data -f Makefile.xml WHITELIST=RenderBasic
v2: add newlines to debug messages + fix comment (Matthew Auld)
Signed-off-by: Robert Bragg <robert@sixbynine.org>
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Gen8+ might have mux configurations per slices/subslices. Depending on
whether slices/subslices have been fused off, only part of the
configuration needs to be applied. This change reworks the mux
configurations query mechanism to allow more than one set of registers
to be programmed.
v2: s/n_mux_regs/n_mux_configs/ (Matthew)
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Assuming a uniform mask across all slices, this enables userspace to
determine the specific sub slices can be enabled. This information is
required, for example, to be able to analyse some OA counter reports
where the counter configuration depends on the HW sub slice
configuration.
Signed-off-by: Robert Bragg <robert@sixbynine.org>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Enables userspace to determine the maximum number of slices that can
be enabled on the device and also know what specific slices can be
enabled. This information is required, for example, to be able to
analyse some OA counter reports where the counter configuration
depends on the HW slice configuration.
Signed-off-by: Robert Bragg <robert@sixbynine.org>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
I removed the zapping of the reservation_object->fence array of shared
fences prematurely. We don't yet have the code to zap that array when
retiring the object, and so currently it remains possible to continually
grow the shared array trapping requests when reusing the batch_pool
object across many timelines.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Matthew Auld <matthew.auld@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170518094638.5469-4-chris@chris-wilson.co.uk
Having resolved whether or not we would deadlock upon a call to
mutex_lock(&dev->struct_mutex), we can then spin for the contended
struct_mutex if we are not the owner. We cannot afford to simply block
and wait for the mutex, as the owner may itself be waiting for the
allocator -- i.e. a cyclic deadlock. This should significantly improve
the chance of running the shrinker for other processes whilst the GPU is
busy.
A more balanced approach would be to optimistically spin whilst the
mutex owner was on the cpu and there was an opportunity to acquire the
mutex for ourselves quickly. However, that requires support from
kernel/locking/ and a new mutex_spin_trylock() primitive.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170609110350.1767-4-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
In our first pass, we do not want to use reclaim at all as we want to
solely reap the i915 buffer caches (its purgeable pages). But we don't
mind it initiates IO or pulls via the FS (but it shouldn't anyway as we
say no to reclaim!). Just drop the GFP_IO constraint for simplicity.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170609110350.1767-3-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
I tried __GFP_NORETRY in the belief that __GFP_RECLAIM was effective. It
struggles with handling reclaim of our dirty buffers and relies on
reclaim via kswapd. As a result, a single pass of direct reclaim is
unreliable when i915 occupies the majority of available memory, and the
only means of effectively waiting on kswapd to amke progress is by not
setting the __GFP_NORETRY flag and lopping. That leaves us with the
dilemma of invoking the oomkiller instead of propagating the allocation
failure back to userspace where it can be handled more gracefully (one
hopes). In the future we may have __GFP_MAYFAIL to allow repeats up until
we genuinely run out of memory and the oomkiller would have been invoked.
Until then, let the oomkiller wreck havoc.
v2: Stop playing with side-effects of gfp flags and await __GFP_MAYFAIL
v3: Update comments that direct reclaim only appears to be ignoring our
dirty buffers!
Fixes: 24f8e00a8a ("drm/i915: Prefer to report ENOMEM rather than incur the oom for gfx allocations")
Testcase: igt/gem_tiled_swapping
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Michal Hocko <mhocko@suse.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170609110350.1767-2-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Commit 24f8e00a8a ("drm/i915: Prefer to report ENOMEM rather than
incur the oom for gfx allocations") made the bold decision to try and
avoid the oomkiller by reporting -ENOMEM to userspace if our allocation
failed after attempting to free enough buffer objects. In short, it
appears we were giving up too easily (even before we start wondering if
one pass of reclaim is as strong as we would like). Part of the problem
is that if we only shrink just enough pages for our expected allocation,
the likelihood of those pages becoming available to us is less than 100%
To counter-act that we ask for twice the number of pages to be made
available. Furthermore, we allow the shrinker to pull pages from the
active list in later passes.
v2: Be a little more cautious in paging out gfx buffers, and leave that
to a more balanced approach from shrink_slab(). Important when combined
with "drm/i915: Start writeback from the shrinker" as anything shrunk is
immediately swapped out and so should be more conservative.
Fixes: 24f8e00a8a ("drm/i915: Prefer to report ENOMEM rather than incur the oom for gfx allocations")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170609110350.1767-1-chris@chris-wilson.co.uk
All here is pretty much like Kabylake.
Including CFL-U has to use same ddi translation table
as KBL-U for now.
v2: Include missed IS_COFFEELAKE on edp trans table. (DK)
Handle CFL-U with same translation table as KBL-U. (DK and
confirmed with HW engineers)
v3: Adding missed case for IS_CFL_ULT. (DK).
v4: Duh! Now with the real IS_CFL_ULT instead of KBL one. (DK)
Also use IS_GEN9_BC when possible. (DK)
Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1497045770-21302-1-git-send-email-rodrigo.vivi@intel.com
Enable wrpll computation for Cannonlake platform to support
pll's required for HDMI output. The patch contains the following features
- compute Cannonlake port clock programming
dividers P, Q, and K.
- compute PLL parameters for Cannonlake. These parameters
set the values on DPLL registers.
- find the register values to program wrpll for Cannonlake.
The reference clock can be either 19.2MHz or 24MHz.
v2: rebase
v3: squash wrpll patches into one (Rodrigo)
v4: switch order of getting even dividers (Paulo)
update divider register values for PDiv and KDiv (Paulo)
update wrpll computation algorithm (Paulo)
v5: Remove ref clock division by 1000. (Rodrigo)
v6: Rodrigo rebasing on top of latest code.
Signed-off-by: Kahola, Mika <mika.kahola@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Clint Taylor <Clinton.A.Taylor@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1497047175-27250-18-git-send-email-rodrigo.vivi@intel.com
