Chris Wilson [Mon, 11 Nov 2019 13:32:05 +0000 (13:32 +0000)]
drm/i915/execlists: Move reset_active() from schedule-out to schedule-in
The gem_ctx_persistence/smoketest was detecting an odd coherency issue
inside the LRC context image; that the address of the ring buffer did
not match our associated struct intel_ring. As we set the address into
the context image when we pin the ring buffer into place before the
context is active, that leaves the question of where did it get
overwritten. Either the HW context save occurred after our pin which
would imply that our idle barriers are broken, or we overwrote the
context image ourselves. It is only in reset_active() where we dabble
inside the context image outside of a serialised path from schedule-out;
but we could equally perform the operation inside schedule-in which is
then fully serialised with the context pin -- and remains serialised by
the engine pulse with kill_context(). (The only downside, aside from
doing more work inside the engine->active.lock, was the plan to merge
all the reset paths into doing their context scrubbing on schedule-out
needs more thought.)
Fixes: d12acee84ffb ("drm/i915/execlists: Cancel banned contexts on schedule-out")
Testcase: igt/gem_ctx_persistence/smoketest Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191111133205.11590-3-chris@chris-wilson.co.uk
Chris Wilson [Mon, 11 Nov 2019 13:32:04 +0000 (13:32 +0000)]
drm/i915/userptr: Handle unlocked gup retries
Enable gup to retry and fault the pages outside of the mmap_sem lock in
our worker. As we are inside our worker, outside of any critical path,
we can allow the mmap_sem lock to be dropped in order to service a page
fault; this in turn allows the mm to populate the page using a slow
fault handler.
Chris Wilson [Mon, 11 Nov 2019 13:32:03 +0000 (13:32 +0000)]
drm/i915/userptr: Try to acquire the page lock around set_page_dirty()
set_page_dirty says:
For pages with a mapping this should be done under the page lock
for the benefit of asynchronous memory errors who prefer a
consistent dirty state. This rule can be broken in some special
cases, but should be better not to.
Under those rules, it is only safe for us to use the plain set_page_dirty
calls for shmemfs/anonymous memory. Userptr may be used with real
mappings and so needs to use the locked version (set_page_dirty_lock).
However, following a try_to_unmap() we may want to remove the userptr and
so call put_pages(). However, try_to_unmap() acquires the page lock and
so we must avoid recursively locking the pages ourselves -- which means
that we cannot safely acquire the lock around set_page_dirty(). Since we
can't be sure of the lock, we have to risk skip dirtying the page, or
else risk calling set_page_dirty() without a lock and so risk fs
corruption.
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=203317
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=112012 Fixes: 5cc9ed4b9a7a ("drm/i915: Introduce mapping of user pages into video memory (userptr) ioctl")
References: cb6d7c7dc7ff ("drm/i915/userptr: Acquire the page lock around set_page_dirty()")
References: 505a8ec7e11a ("Revert "drm/i915/userptr: Acquire the page lock around set_page_dirty()"")
References: 6dcc693bc57f ("ext4: warn when page is dirtied without buffers") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Lionel Landwerlin <lionel.g.landwerlin@intel.com> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: stable@vger.kernel.org Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191111133205.11590-1-chris@chris-wilson.co.uk
Gwan-gyeong Mun [Wed, 6 Nov 2019 21:26:36 +0000 (23:26 +0200)]
drm/i915: Split a setting of MSA to MST and SST
The setting of MSA is done by the DDI .pre_enable() hook. And when we are
using MST, the MSA is only set to first mst stream by calling of
DDI .pre_eanble() hook. It raies issues to non-first mst streams.
Wrong MSA or missed MSA packets might show scrambled screen or wrong
screen.
This splits a setting of MSA to MST and SST cases. And In the MST case it
will call a setting of MSA after an allocating of Virtual Channel from
MST encoder pre_enable callback.
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=112212 Fixes: 0c06fa156006 ("drm/i915/dp: Add support of BT.2020 Colorimetry to DP MSA") Fixes: d4a415dcda35 ("drm/i915: Fix MST oops due to MSA changes") Signed-off-by: Gwan-gyeong Mun <gwan-gyeong.mun@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191106212636.502471-1-gwan-gyeong.mun@intel.com Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
[vsyrjala: nuke spurious newline] Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Chris Wilson [Sun, 10 Nov 2019 18:57:51 +0000 (18:57 +0000)]
drm/i915/execlists: Reduce barrier on context switch to a wmb()
Having been forced to reduce Braswell back to using the aliasing ppgtt,
the coherency issue we previously observed cannot impact us. Reduce the
performance penalty imposed on all platforms from using the mfence to a
mere sfence.
Chris Wilson [Sun, 10 Nov 2019 18:57:54 +0000 (18:57 +0000)]
drm/i915: Taint the kernel on dumping the GEM ftrace buffer
As the ftrace buffer is single shot, once dumped it will not update. As
such, it only provides information for the first bug and all subsequent
bugs are noise. The goal of CI is to have zero bugs, so taint the kernel
causing CI to reboot the machine; fix the bug and move on.
Chris Wilson [Mon, 11 Nov 2019 12:27:06 +0000 (12:27 +0000)]
drm/i915/selftests: Fill all the drm_vma_manager holes
To test mmap_offset_exhaustion, we first have to fill the entire vma
manager leaving a single page. Don't assume that the vma manager is not
already fragment, and fill all the holes.
Chris Wilson [Mon, 11 Nov 2019 12:27:05 +0000 (12:27 +0000)]
drm/i915/selftests: Exercise parallel blit operations on a single ctx
Make sure that our code is robust enough to handle multiple threads
trying to clear objects for a single client context. This brings the joy
of a shared GGTT to all!
Chris Wilson [Sat, 9 Nov 2019 10:53:56 +0000 (10:53 +0000)]
drm/i915/pmu: "Frequency" is reported as accumulated cycles
We report "frequencies" (actual-frequency, requested-frequency) as the
number of accumulated cycles so that the average frequency over that
period may be determined by the user. This means the units we report to
the user are Mcycles (or just M), not MHz.
Chris Wilson [Mon, 11 Nov 2019 11:43:23 +0000 (11:43 +0000)]
drm/i915: Cancel context if it hangs after it is closed
If we detect a hang in a closed context, just flush all of its requests
and cancel any remaining execution along the context. Note that after
closing the context, the last reference to the context may be dropped,
leaving it only valid under RCU.
Chris Wilson [Mon, 11 Nov 2019 11:43:22 +0000 (11:43 +0000)]
drm/i915: Show guilty context name on GPU reset
We mention that we are resetting the GPU, and dump the device state for
post mortem debugging. However, while that dump contains the active
processes and the one flagged as causing the error, we do not always
include that information in dmesg. Include the name of the guilty
process in dmesg for reference.
Chris Wilson [Mon, 11 Nov 2019 11:43:20 +0000 (11:43 +0000)]
drm/i915/gem: Embed context/timeline name inside the GEM context
Use a small char buffer inside the i915_gem_context to store the user
friendly name so that ctx->name has the same lifetime as the RCU
protected GEM context. That is, e.g. when using print_request() that
prints the timeline name (ctx->name), the name will not be prematurely
freed upon the context being closed and the last reference dropped.
Chris Wilson [Mon, 11 Nov 2019 11:43:19 +0000 (11:43 +0000)]
drm/i915: Protect context while grabbing its name for the request
Inside print_request(), we query the context/timeline name. Nothing
immediately protects the context from being freed if the request is
complete -- we rely on serialisation by the caller to keep the name
valid until they finish using it. Inside intel_engine_dump(), we
generally only print the requests in the execution queue protected by the
engine->active.lock, but we also show the pending execlists ports which
are not protected and so require a rcu_read_lock to keep the pointer
valid.
Fixes: c36eebd9ba5d ("drm/i915/gt: execlists->active is serialised by the tasklet") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191111114323.5833-1-chris@chris-wilson.co.uk
Chris Wilson [Sun, 10 Nov 2019 18:57:50 +0000 (18:57 +0000)]
drm/i915/icl: Refine PG_HYSTERESIS
After doing some measuring, Icelake behaves on a par with Broadwell, and
without having to compromise for low power cores with long latencies, we
can reduce the powergating hysteresis so that the powersaving is enabled
faster. No impact observed on client side throughput measures (so
negligible increase in extra switching), and inspection from high
frequency polling using igt/gem_exec_nop/sequential, provided an estimate
for the upper bound before we can measure a substantial impact on
latency.
Ville Syrjälä [Wed, 6 Nov 2019 17:23:49 +0000 (19:23 +0200)]
drm/i915: Don't oops in dumb_create ioctl if we have no crtcs
Make sure we have a crtc before probing its primary plane's
max stride. Initially I thought we can't get this far without
crtcs, but looks like we can via the dumb_create ioctl.
Not sure if we shouldn't disable dumb buffer support entirely
when we have no crtcs, but that would require some amount of work
as the only thing currently being checked is dev->driver->dumb_create
which we'd have to convert to some device specific dynamic thing.
Cc: stable@vger.kernel.org Reported-by: Mika Kuoppala <mika.kuoppala@linux.intel.com> Fixes: aa5ca8b7421c ("drm/i915: Align dumb buffer stride to 4k to allow for gtt remapping") Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191106172349.11987-1-ville.syrjala@linux.intel.com Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Chris Wilson [Fri, 8 Nov 2019 10:35:11 +0000 (10:35 +0000)]
drm/i915/pmu: Only use exclusive mmio access for gen7
On gen7, we have to avoid concurrent access to the same mmio cacheline,
and so coordinate all mmio access with the uncore->lock. However, for
pmu, we want to avoid perturbing the system and disabling interrupts
unnecessarily, so restrict the w/a to gen7 where it is requied to
prevent machine hangs.
Chris Wilson [Fri, 8 Nov 2019 10:35:10 +0000 (10:35 +0000)]
drm/i915/pmu: Cheat when reading the actual frequency to avoid fw
We want to avoid taking forcewake when querying the performance stats,
as we wish to avoid perturbing the system under observation. (And with
the forcewake being kept alive for 1ms after use, sampling the frequency
from a 200Hz timer keeps forcewake 40% active.)
Chris Wilson [Thu, 7 Nov 2019 22:12:01 +0000 (22:12 +0000)]
drm/i915/selftests: Mark up sole accessor to ctx->vm as being protected
In the selftests, where we are accessing a private ctx from within the
confines of a single test, we know that the ctx->vm pointer is static
and bounded by the lifetime of the test. We can use a simple helper to
provide the RCU annotations to keep sparse happy.
Chris Wilson [Thu, 7 Nov 2019 21:39:29 +0000 (21:39 +0000)]
drm/i915/selftests: Complete transition to a real struct file mock
Since drm provided us with a real struct file we can use for our
anonymous internal clients (mock_file), complete our transition to using
that as the primary interface (and not the mocked up struct drm_file we
previous were using).
Masahiro Yamada [Fri, 8 Nov 2019 09:41:42 +0000 (18:41 +0900)]
drm/i915: make more headers self-contained
The headers in the gem/selftests/, gt/selftests, gvt/, selftests/
directories have never been compile-tested, but it would be possible
to make them self-contained.
This commit only addresses missing <linux/types.h> and forward
struct declarations.
Matt Roper [Thu, 7 Nov 2019 17:45:27 +0000 (09:45 -0800)]
drm/i915: Expand documentation for gen12 DP pre-enable sequence
Rather than just specifying the bullet numbers from the bspec (e.g.,
"4.b") actually include the description of what the bspec wants us to
do. Steps can be renumbered or moved so including the description will
help us match the code up to the spec. Plus if we add support for new
platforms, some of the steps may be added/removed so more descriptive
comments will be useful for ensuring all of the bspec requirements are
met.
Chris Wilson [Thu, 7 Nov 2019 18:06:01 +0000 (18:06 +0000)]
drm/i915/selftests: Verify mmap_gtt revocation on unbinding
Whenever, we unbind (or change fence registers) on an object, we must
revoke any and all mmap_gtt using the previous bindings. Those user PTEs
point at the GGTT which know points into a new object, the wrong object.
Ergo, those PTEs must be cleared so that any user access provokes a new
page fault.
Chris Wilson [Thu, 7 Nov 2019 18:06:00 +0000 (18:06 +0000)]
drm/i915/selftests: Wrap vm_mmap() around GEM objects
Provide a utility function to create a vma corresponding to an mmap() of
our device. And use it to exercise the equivalent of userspace
performing a GTT mmap of our objects.
Chris Wilson [Thu, 7 Nov 2019 18:05:59 +0000 (18:05 +0000)]
drm/i915/selftests: Replace mock_file hackery with drm's true fake
As drm now exports a method to create an anonymous struct file around a
drm_device for internal use, make use of it to avoid our horrible hacks.
Danial suggested that the mock_file_put() wrapper was suitable for
drm-core, along with the mock_drm_getfile() [and that the vestigal
mock_drm_file() in this patch should perhaps be the drm interface
itself]. However, the eventual goal is to remove the mock_drm_file() and
use the struct file and fput() directly, in this patch we take a simple
transition in that direction.
Chris Wilson [Thu, 7 Nov 2019 18:05:58 +0000 (18:05 +0000)]
drm: Expose a method for creating anonymous struct file around drm_minor
Sometimes we need to create a struct file to wrap a drm_device, as it
the user were to have opened /dev/dri/card0 but to do so anonymously
(i.e. for internal use). Provide a utility method to create a struct
file with the drm_device->driver.fops, that wrap the drm_device.
Chris Wilson [Thu, 7 Nov 2019 18:05:57 +0000 (18:05 +0000)]
drm: Move EXPORT_SYMBOL_FOR_TESTS_ONLY under a separate Kconfig
Currently, we only export symbols for drm-selftests which are either
compiled as modules or into the main drm builtin. However, if we want to
export symbols from drm.ko for the drivers' selftests, we require a
means of controlling that export separately. So we add a new Kconfig to
determine whether or not the EXPORT_SYMBOL_FOR_TESTS_ONLY() takes
effect.
Chris Wilson [Wed, 6 Nov 2019 09:13:12 +0000 (09:13 +0000)]
drm/i915/gem: Safely acquire the ctx->vm when copying
As we read the ctx->vm unlocked before cloning/exporting, we should
validate our reference is correct before returning it. We already do for
clone_vm() but were not so strict around get_ppgtt().
Ville Syrjälä [Wed, 30 Oct 2019 19:08:15 +0000 (21:08 +0200)]
drm/i915: Preload LUTs if the hw isn't currently using them
The LUTs are single buffered so in order to program them without
tearing we'd have to do it during vblank (actually to be 100%
effective it has to happen between start of vblank and frame start).
We have no proper mechanism for that at the moment so we just
defer loading them after the vblank waits have happened. That
is not quite sufficient (especially when committing multiple pipes
whose vblanks don't line up) so the LUT load will often leak into
the following frame causing tearing.
However in case the hardware wasn't previously using the LUT we
can preload it before setting the enable bit (which is double
buffered so won't tear). Let's determine if we can do such
preloading and make it happen. Slight variation between the
hardware requires some platforms specifics in the checks.
Hans is seeing ugly colored flash on VLV/CHV macchines (GPD win
and Asus T100HA) when the gamma LUT gets loaded for the first
time as the BIOS has left some junk in the LUT memory.
v2: Deal with uapi vs. hw crtc state split
s/GCM/CGM/ typo fix
Ramalingam C [Tue, 5 Nov 2019 14:44:14 +0000 (20:14 +0530)]
drm/i915: FB backing gem obj should reside in LMEM
If Local memory is supported by hardware, we want framebuffer backing
gem objects from local memory.
if the backing obj is not from LMEM, pin_to_display is failed.
v2:
memory regions are correctly assigned to obj->memory_regions [tvrtko]
migration failure is reported as debug log [Tvrtko]
v3:
Migration is dropped. only error is reported [Daniel]
mem region check is move to pin_to_display [Chris]
v4:
s/dev_priv/i915 [chris]
v5:
i915_gem_object_is_lmem is used for detecting the obj mem type. [Matt]
Chris Wilson [Wed, 6 Nov 2019 22:12:23 +0000 (22:12 +0000)]
drm/i915: Leave the aliasing-ppgtt size alone
The hidden aliasing-ppgtt's size is never revealed, as we only inspect
the front GTT when engaged. However, we were "fixing" the hidden ppgtt
to match, with the net result that we ended up leaking the unused
portion on Braswell were we preallocated the entire set of top level
PDP, see gen8_preallocate_top_level_pdp().
Reported-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Fixes: 1eda701eace2 ("drm/i915/gtt: Recursive cleanup for gen8")
References: c082afac86cb ("drm/i915: Move aliasing_ppgtt underneath its i915_ggtt") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191106221223.7437-1-chris@chris-wilson.co.uk
(the shrinker cannot get at an object while we're in get_pages, hence
this is safe). But it's confusing, so try to take the right subclass
of the lock.
This does a bit reduce our lockdep based checking, but then it's also
less fragile, in case we ever change the nesting around.
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Will Deacon <will@kernel.org> Cc: linux-kernel@vger.kernel.org Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191104173720.2696-3-daniel.vetter@ffwll.ch
Daniel Vetter [Tue, 5 Nov 2019 09:01:48 +0000 (10:01 +0100)]
drm/i915: Switch obj->mm.lock lockdep annotations on its head
The trouble with having a plain nesting flag for locks which do not
naturally nest (unlike block devices and their partitions, which is
the original motivation for nesting levels) is that lockdep will
never spot a true deadlock if you screw up.
This patch is an attempt at trying better, by highlighting a bit more
of the actual nature of the nesting that's going on. Essentially we
have two kinds of objects:
- objects without pages allocated, which cannot be on any lru and are
hence inaccessible to the shrinker.
- objects which have pages allocated, which are on an lru, and which
the shrinker can decide to throw out.
For the former type of object, memory allocations while holding
obj->mm.lock are permissible. For the latter they are not. And
get/put_pages transitions between the two types of objects.
This is still not entirely fool-proof since the rules might change.
But as long as we run such a code ever at runtime lockdep should be
able to observe the inconsistency and complain (like with any other
lockdep class that we've split up in multiple classes). But there are
a few clear benefits:
- We can drop the nesting flag parameter from
__i915_gem_object_put_pages, because that function by definition is
never going allocate memory, and calling it on an object which
doesn't have its pages allocated would be a bug.
- We strictly catch more bugs, since there's not only one place in the
entire tree which is annotated with the special class. All the
other places that had explicit lockdep nesting annotations we're now
going to leave up to lockdep again.
- Specifically this catches stuff like calling get_pages from
put_pages (which isn't really a good idea, if we can call get_pages
so could the shrinker). I've seen patches do exactly that.
Of course I fully expect CI will show me for the fool I am with this
one here :-)
v2: There can only be one (lockdep only has a cache for the first
subclass, not for deeper ones, and we don't want to make these locks
even slower). Still separate enums for better documentation.
Real fix: don't forget about phys objs and pin_map(), and fix the
shrinker to have the right annotations ... silly me.
v3: Forgot usertptr too ...
v4: Improve comment for pages_pin_count, drop the IMPORTANT comment
and instead prime lockdep (Chris).
v5: Appease checkpatch, no double empty lines (Chris)
v6: More rebasing over selftest changes. Also somehow I forgot to
push this patch :-/
Also format comments consistently while at it.
v7: Fix typo in commit message (Joonas)
Also drop the priming, with the lmem merge we now have allocations
while holding the lmem lock, which wreaks the generic priming I've
done in earlier patches. Should probably be resurrected when lmem is
fixed. See
Chris Wilson [Wed, 6 Nov 2019 22:34:10 +0000 (22:34 +0000)]
drm/i915/gt: Cleanup heartbeat systole first
Before we grab the engine wakeref, tidy up the previous heartbeat
request. If we then abort because the engine powerwell is off, we ensure
the request is freed as we know we will not have freed it when
cancelling the work (as the work is running!).
Fixes: 841e86728615 ("drm/i915/gt: Only drop heartbeat.systole if the sole owner")
References: 058179e72e09 ("drm/i915/gt: Replace hangcheck by heartbeats") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191106223410.30334-1-chris@chris-wilson.co.uk
James Ausmus [Wed, 6 Nov 2019 00:55:26 +0000 (16:55 -0800)]
drm/i915/tgl: Add second TGL PCH ID
Another TGP ID has shown up, so let's add it to avoid South Display
breakage on systems that have this ID.
Cc: Lucas De Marchi <lucas.demarchi@intel.com> Cc: José Roberto de Souza <jose.souza@intel.com> Signed-off-by: James Ausmus <james.ausmus@intel.com> Reviewed-by: José Roberto de Souza <jose.souza@intel.com> Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191106005526.1500-1-james.ausmus@intel.com
Chris Wilson [Wed, 6 Nov 2019 13:31:29 +0000 (13:31 +0000)]
drm/i915/gt: Only drop heartbeat.systole if the sole owner
Mika spotted that only using cancel_delayed_work() could mean that we
attempted to clear the heartbeat.systole while the worker was still
running. Rectify the situation by only touching the systole from outside
the worker if we suceeded in cancelling the worker before it could run.
The worker is expected to clean up by itself upon idling.
Reported-by: Mika Kuoppala <mika.kuoppala@linux.intel.com> Fixes: 058179e72e09 ("drm/i915/gt: Replace hangcheck by heartbeats") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191106133129.17732-1-chris@chris-wilson.co.uk
Ville Syrjälä [Tue, 5 Nov 2019 17:14:46 +0000 (19:14 +0200)]
drm/i915: Frob the correct crtc state in intel_crtc_disable_noatomic()
The uapi vs. hw state split introduced a bug in
intel_crtc_disable_noatomic() where it's now frobbing an already
freed temp crtc state instead of adjusting the crtc state we
are really left with. Fix that by making a cleaner separation
beteen the two.
This causes explosions on any machine that boots up with pipes
already running but not hooked up to any encoder (typical
behaviour for gen2-4 VBIOS).
Chris Wilson [Tue, 5 Nov 2019 14:53:05 +0000 (14:53 +0000)]
drm/i915/gem: Early rejection of no-aperture map_ggtt
If the device does not have an aperture through which we can indirectly
access and detile the buffers, simply reject the ioctl. Later we can
extend the ioctl to support different modes, but as an extension the
user must opt in and explicitly control the mmap type (viz
MMAP_OFFSET_IOCTL).
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Abdiel Janulgue <abdiel.janulgue@linux.intel.com> Cc: Matthew Auld <matthew.auld@intel.com> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Reviewed-by: Abdiel Janulgue <abdiel.janulgue@linux.intel.com> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191105145305.14314-1-chris@chris-wilson.co.uk
Jani Nikula [Mon, 4 Nov 2019 14:14:38 +0000 (16:14 +0200)]
drm/i915/dsc: split out encoder specific parts from DSC compute params
Split out the DP specific parts, making it easier to add DSI specific
configuration. Also move the encoder specific parts towards the end, to
allow overriding generic configuration if needed. This also improves
clarity by making it clear the encoder independent configuration does
not depend on the encoder specific parts.
Jani Nikula [Fri, 1 Nov 2019 14:20:24 +0000 (16:20 +0200)]
drm/i915: update rawclk also on resume
Since CNP it's possible for rawclk to have two different values, 19.2
and 24 MHz. If the value indicated by SFUSE_STRAP register is different
from the power on default for PCH_RAWCLK_FREQ, we'll end up having a
mismatch between the rawclk hardware and software states after
suspend/resume. On previous platforms this used to work by accident,
because the power on defaults worked just fine.
Update the rawclk also on resume. The natural place to do this would be
intel_modeset_init_hw(), however VLV/CHV need it done before
intel_power_domains_init_hw(). Thus put it there even if it feels
slightly out of place.
v2: Call intel_update_rawclck() in intel_power_domains_init_hw() for all
platforms (Ville).
Reported-by: Shawn Lee <shawn.c.lee@intel.com> Cc: Shawn Lee <shawn.c.lee@intel.com> Cc: Ville Syrjala <ville.syrjala@linux.intel.com> Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Tested-by: Shawn Lee <shawn.c.lee@intel.com> Signed-off-by: Jani Nikula <jani.nikula@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191101142024.13877-1-jani.nikula@intel.com
Lucas De Marchi [Wed, 30 Oct 2019 01:24:46 +0000 (18:24 -0700)]
drm/i915/tgl: do not enable transcoder clock twice on MST
For MST on Tiger Lake there are different moments when we need to
configure the transcoder clock select. For the first link this is in step
7.a of the spec, before training the link. For additional streams this
should be done as part of step 8.b after programming receiver VC Payload
ID.
Ville Syrjälä [Thu, 31 Oct 2019 16:56:48 +0000 (18:56 +0200)]
drm/i915: Add missing 10bpc formats for pipe B sprites on CHV
CHV pipe B sprites gained support for the 10bpc X/ARGB pixel formats.
On VLV and CHV pipe A/C these are only supported by the primary
plane. Add the require bits to expose the new formats.
After the reset, we do another clflush before checking the CSB to be
sure we see whatever was left in the CSB prior to the reset. So it is
unlikely to be an incoherent view of the CSB, and more likely that
Icelake didn't reset its pointers.
Chris Wilson [Sun, 3 Nov 2019 16:23:05 +0000 (16:23 +0000)]
drm/i915: Protect request peeking with RCU
Since the execlists_active() is no longer protected by the
engine->active.lock, we need to protect the request pointer with RCU to
prevent it being freed as we evaluate whether or not we need to preempt.
Fixes: df403069029d ("drm/i915/execlists: Lift process_csb() out of the irq-off spinlock") Fixes: 13ed13a4dcbf ("drm/i915: Don't set queue_priority_hint if we don't kick the submission") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191104090158.2959-2-chris@chris-wilson.co.uk
Chris Wilson [Sun, 3 Nov 2019 09:50:38 +0000 (09:50 +0000)]
drm/i915/gt: Drop false assertion on user_forcewake
The counter is removed from the pm wakeref count, but it remains intact
so that we can restore it upon resume. Ergo inside suspend, it may have
a value.
Chris Wilson [Fri, 1 Nov 2019 18:10:22 +0000 (18:10 +0000)]
drm/i915/selftests: Flush all active callbacks
Flushing the outer i915_active is not enough, as we need the barrier to
be applied across all the active dma_fence callbacks. So we must
serialise with each outstanding fence.
Chris Wilson [Fri, 1 Nov 2019 14:10:09 +0000 (14:10 +0000)]
drm/i915: Defer rc6 shutdown to suspend_late
Currently we shutdown rc6 during i915_gem_resume() but this is called
during the preparation phase (i915_drm_prepare) for all suspend paths,
but we only want to shutdown rc6 for S3+. Move the actual shutdown to
i915_gem_suspend_late().
We then need to differentiate between suspend targets, to distinguish S0
(s2idle) where the device is kept awake but needs to be in a low power
mode (the same as runtime suspend) from the device suspend levels where
we lose control of HW and so must disable any HW access to dangling
memory.
Chris Wilson [Fri, 1 Nov 2019 13:04:06 +0000 (13:04 +0000)]
drm/i915/gt: Pull timeline initialise to intel_gt_init_early
Our timelines are currently contained within an intel_gt, and we only
need to perform list/spinlock initialisation, so we can pull the
intel_timelines_init() into our intel_gt_init_early().
drm/i915: Remove special case slave handling during hw programming, v3.
Now that we split plane_state which I didn't want to do yet, we can
program the slave plane without requiring the master plane.
This is useful for programming bigjoiner slave planes as well. We
will no longer need the master's plane_state.
Changes since v1:
- set src/dst rectangles after copy_uapi_to_hw_state.
Changes since v2:
- Use the correct color_plane for pre-gen11 by using planar_linked_plane != NULL.
- Use drm_format_info_is_yuv_semiplanar in skl_plane_check() to fix gen11+.
Splitting plane state is easier than splitting crtc_state,
before plane check we copy the drm properties to hw so we can
do the same in bigjoiner later on.
We copy the state after we did all the modeset handling, but fortunately
i915 seems to be split correctly and nothing during modeset looks
at plane_state.
Changes since v1:
- Do not clear hw state on duplication.
drm/i915: Add aliases for uapi and hw to plane_state
Prepare to split up hw and uapi machinally, by adding a uapi and
hw alias. We will remove the base in a bit. This is a split from the
original uapi/hw patch, which did it all in one go.
Now that we separated everything into uapi and hw, it's
time to make the split definitive. Remove the union and
make a copy of the hw state on modeset and fastset.
Color blobs are copied in crtc atomic_check(), right
before color management is checked.
Changes since v1:
- Copy all blobs immediately after drm_atomic_helper_check_modeset().
- Clear crtc_state->hw on disable, instead of using clear_intel_crtc_state().
Changes since v2:
- Use intel_crtc_free_hw_state + clear in intel_crtc_disable_noatomic().
- Make a intel_crtc_prepare_state() function that clears the crtc_state
and copies hw members.
- Remove setting uapi.adjusted_mode, we now have a direct call to
drm_calc_timestamping_constants().
Changes since v3:
- Rename prefix copy_hw_to_uapi_state() with intel_crtc.
- Copy color blobs to uapi as well.
- Add a intel_crtc_copy_uapi_to_hw_state_nomodeset() function for clarity.
Changes since v4:
- Copy hw.adjusted_mode back to uapi.adjusted_mode, to shut up
the call to drm_calc_timestamping_constants() in
drm_atomic_helper_update_legacy_modeset_state().
- Use drm_property_replace_blob (Ville).
Changes since v5:
- Use hw->mode in intel_modeset_readout_hw_state(). (Ville)
- Copy to uapi.mode using drm_atomic_set_mode_for_crtc(). (Ville)
drm/i915: Add aliases for uapi and hw to crtc_state
Prepare to split up hw and uapi machinally, by adding a uapi and
hw alias. We will remove the base in a bit. This is a split from the
original uapi/hw patch, which did it all in one go.
Chris Wilson [Fri, 1 Nov 2019 10:15:28 +0000 (10:15 +0000)]
drm/i915/selftests: Spin on all engines simultaneously
Vanshidhar Konda asked for the simplest test "to verify that the kernel
can submit and hardware can execute batch buffers on all the command
streamers in parallel." We have a number of tests in userspace that
submit load to each engine and verify that it is present, but strictly
we have no selftest to prove that the kernel can _simultaneously_
execute on all known engines. (We have tests to demonstrate that we can
submit to HW in parallel, but we don't insist that they execute in
parallel.)
v2: Improve the igt_spinner support for older gen.
Chris Wilson [Fri, 1 Nov 2019 08:49:40 +0000 (08:49 +0000)]
drm/i915/selftests: Start kthreads before stopping
An interesting observation made with our parallel selftests was that on
our small/single cpu systems we would call kthread_stop() before the
kthreads were spawned. If this happens, the kthread is never run at all;
completely bypassing the test.
A simple yield() from the parent will ensure that all children have the
opportunity to start before we reap them.
Chris Wilson [Thu, 31 Oct 2019 09:01:04 +0000 (09:01 +0000)]
drm/i915/gem: Refine occupancy test in kill_context()
Don't just look at the very last request in a queue when deciding if we
need to evict the context from the GPU, as that request may still be in
the submission queue while the rest of the context is running!
Instead, walk back along the queued requests looking for the active
request and checking that.
Fixes: 2e0986a58cc4 ("drm/i915/gem: Cancel contexts when hangchecking is disabled")
Testcase: igt/gem_ctx_persistence/queued Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Cc: Matthew Auld <matthew.auld@intel.com> Reviewed-by: Matthew Auld <matthew.auld@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191031090104.22245-1-chris@chris-wilson.co.uk
Matthew Auld [Wed, 30 Oct 2019 17:33:20 +0000 (17:33 +0000)]
drm/i915/lmem: add the fake lmem region
Intended for upstream testing so that we can still exercise the LMEM
plumbing and !i915_ggtt_has_aperture paths. Smoke tested on Skull Canyon
device. This works by allocating an intel_memory_region for a reserved
portion of system memory, which we treat like LMEM. For the LMEMBAR we
steal the aperture and 1:1 it map to the stolen region.
To enable simply set the i915 modparam fake_lmem_start= on the kernel
cmdline with the start of reserved region(see memmap=). The size of the
region we can use is determined by the size of the mappable aperture, so
the size of reserved region should be >= mappable_end. For now we only
enable for the selftests. Depends on CONFIG_DRM_I915_UNSTABLE being
enabled.
Signed-off-by: Matthew Auld <matthew.auld@intel.com> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Abdiel Janulgue <abdiel.janulgue@linux.intel.com> Cc: Arkadiusz Hiler <arkadiusz.hiler@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Link: https://patchwork.freedesktop.org/patch/msgid/20191030173320.8850-1-matthew.auld@intel.com
drm/i915/dp: Do not switch aux to TBT mode for non-TC ports
Non-TC ports always have tc_mode == TC_PORT_TBT_ALT so it was
switching aux to TBT mode for all combo-phy ports, happily this did
not caused any issue but is better follow BSpec.
Also this is reserved bit before ICL.
Recent GuC doesn't require the shared area. We still have one user in
i915 (engine reset via guc) because we haven't updated the command to
match the current guc submission flow [1]. Since the flow in guc is
about to change again, just disable the command for now and add a note
that we'll implement it as part of the new flow.
Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com> Cc: Michal Wajdeczko <michal.wajdeczko@intel.com> Cc: John Harrison <John.C.Harrison@Intel.com> Cc: Matthew Brost <matthew.brost@intel.com> Cc: Fernando Pacheco <fernando.pacheco@intel.com> Acked-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Matthew Brost <matthew.brost@intel.com> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Link: https://patchwork.freedesktop.org/patch/msgid/20191031013040.25803-2-daniele.ceraolospurio@intel.com
Recent GuC binaries (including all the ones we're currently using)
don't require this shared area anymore, having moved the relevant
entries into the stage pool instead. i915 itself doesn't write
anything into it either, so we can safely drop it.
Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Michal Wajdeczko <michal.wajdeczko@intel.com> Cc: John Harrison <John.C.Harrison@Intel.com> Cc: Matthew Brost <matthew.brost@intel.com> Acked-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Matthew Brost <matthew.brost@intel.com> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Link: https://patchwork.freedesktop.org/patch/msgid/20191031013040.25803-1-daniele.ceraolospurio@intel.com