When we write watermark values to the hardware, those values are stored
in dev_priv->wm.skl_hw. However with recent watermark changes, the
results structure we're copying from only contains valid watermark and
DDB values for the pipes that are actually changing; the values for
other pipes remain 0. Thus a blind copy of the entire skl_wm_values
structure will clobber the values for unchanged pipes...we need to be
more selective and only copy over the values for the changing pipes.
This mistake was hidden until recently due to another bug that caused us
to erroneously re-calculate watermarks for all active pipes rather than
changing pipes. Only when that bug was fixed was the impact of this bug
discovered (e.g., modesets failing with "Requested display configuration
exceeds system watermark limitations" messages and leaving watermarks
non-functional, even ones initiated by intel_fbdev_restore_mode).
Changes since v1:
- Add a function for copying a pipe's wm values
(skl_copy_wm_for_pipe()) so we can reuse this later
Fixes: 734fa01f3a ("drm/i915/gen9: Calculate watermarks during atomic 'check' (v2)")
Fixes: 9b61302274 ("drm/i915/gen9: Re-allocate DDB only for changed pipes")
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Lyude <cpaul@redhat.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Cc: stable@vger.kernel.org
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Daniel Vetter <daniel.vetter@intel.com>
Cc: Radhakrishna Sripada <radhakrishna.sripada@intel.com>
Cc: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1471463761-26796-4-git-send-email-cpaul@redhat.com
Since the watermark calculations for Skylake are still broken, we're apt
to hitting underruns very easily under multi-monitor configurations.
While it would be lovely if this was fixed, it's not. Another problem
that's been coming from this however, is the mysterious issue of
underruns causing full system hangs. An easy way to reproduce this with
a skylake system:
- Get a laptop with a skylake GPU, and hook up two external monitors to
it
- Move the cursor from the built-in LCD to one of the external displays
as quickly as you can
- You'll get a few pipe underruns, and eventually the entire system will
just freeze.
After doing a lot of investigation and reading through the bspec, I
found the existence of the SAGV, which is responsible for adjusting the
system agent voltage and clock frequencies depending on how much power
we need. According to the bspec:
"The display engine access to system memory is blocked during the
adjustment time. SAGV defaults to enabled. Software must use the
GT-driver pcode mailbox to disable SAGV when the display engine is not
able to tolerate the blocking time."
The rest of the bspec goes on to explain that software can simply leave
the SAGV enabled, and disable it when we use interlaced pipes/have more
then one pipe active.
Sure enough, with this patchset the system hangs resulting from pipe
underruns on Skylake have completely vanished on my T460s. Additionally,
the bspec mentions turning off the SAGV with more then one pipe enabled
as a workaround for display underruns. While this patch doesn't entirely
fix that, it looks like it does improve the situation a little bit so
it's likely this is going to be required to make watermarks on Skylake
fully functional.
This will still need additional work in the future: we shouldn't be
enabling the SAGV if any of the currently enabled planes can't enable WM
levels that introduce latencies >= 30 µs.
Changes since v11:
- Add skl_can_enable_sagv()
- Make sure we don't enable SAGV when not all planes can enable
watermarks >= the SAGV engine block time. I was originally going to
save this for later, but I recently managed to run into a machine
that was having problems with a single pipe configuration + SAGV.
- Make comparisons to I915_SKL_SAGV_NOT_CONTROLLED explicit
- Change I915_SAGV_DYNAMIC_FREQ to I915_SAGV_ENABLE
- Move printks outside of mutexes
- Don't print error messages twice
Changes since v10:
- Apparently sandybridge_pcode_read actually writes values and reads
them back, despite it's misleading function name. This means we've
been doing this mostly wrong and have been writing garbage to the
SAGV control. Because of this, we no longer attempt to read the SAGV
status during initialization (since there are no helpers for this).
- mlankhorst noticed that this patch was breaking on some very early
pre-release Skylake machines, which apparently don't allow you to
disable the SAGV. To prevent machines from failing tests due to SAGV
errors, if the first time we try to control the SAGV results in the
mailbox indicating an invalid command, we just disable future attempts
to control the SAGV state by setting dev_priv->skl_sagv_status to
I915_SKL_SAGV_NOT_CONTROLLED and make a note of it in dmesg.
- Move mutex_unlock() a little higher in skl_enable_sagv(). This
doesn't actually fix anything, but lets us release the lock a little
sooner since we're finished with it.
Changes since v9:
- Only enable/disable sagv on Skylake
Changes since v8:
- Add intel_state->modeset guard to the conditional for
skl_enable_sagv()
Changes since v7:
- Remove GEN9_SAGV_LOW_FREQ, replace with GEN9_SAGV_IS_ENABLED (that's
all we use it for anyway)
- Use GEN9_SAGV_IS_ENABLED instead of 0x1 for clarification
- Fix a styling error that snuck past me
Changes since v6:
- Protect skl_enable_sagv() with intel_state->modeset conditional in
intel_atomic_commit_tail()
Changes since v5:
- Don't use is_power_of_2. Makes things confusing
- Don't use the old state to figure out whether or not to
enable/disable the sagv, use the new one
- Split the loop in skl_disable_sagv into it's own function
- Move skl_sagv_enable/disable() calls into intel_atomic_commit_tail()
Changes since v4:
- Use is_power_of_2 against active_crtcs to check whether we have > 1
pipe enabled
- Fix skl_sagv_get_hw_state(): (temp & 0x1) indicates disabled, 0x0
enabled
- Call skl_sagv_enable/disable() from pre/post-plane updates
Changes since v3:
- Use time_before() to compare timeout to jiffies
Changes since v2:
- Really apply minor style nitpicks to patch this time
Changes since v1:
- Added comments about this probably being one of the requirements to
fixing Skylake's watermark issues
- Minor style nitpicks from Matt Roper
- Disable these functions on Broxton, since it doesn't have an SAGV
Signed-off-by: Lyude <cpaul@redhat.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: stable@vger.kernel.org
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1471463761-26796-3-git-send-email-cpaul@redhat.com
[mlankhorst: ENOSYS -> ENXIO, whitespace fixes]
This issue here is (I think) purely theoretical, since a compiler
would need to be especially foolish to recompute the value of
i915_gem_request_completed right after it was already used. Hence the
additional barrier() is also not really a restriction.
But I believe this to be at least permissible, and since our rcu
trickery is a beast it's worth to annotate all the corner cases.
Chris proposed to instead just wrap a READ_ONCE around
request->fence.seqno in i915_gem_request_completed. But that has a
measurable impact on code size, and everywhere we hold a full
reference to the underlying request it's also not needed. And
personally I'd like to have just enough barriers and locking needed
for correctness, but not more - it makes it much easier in the future
to understand what's going on.
Since the busy ioctl has now fully embraced it's races there's no
point annotating it there too. We really only need it in
active_get_rcu, since that function _must_ deliver a correct snapshot
of the active fences (and not chase something else).
v2: Polish the comment a bit more (Chris).
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>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1471856122-466-1-git-send-email-daniel.vetter@ffwll.ch
In the recent patch
bc3d674 drm/i915: Allow userspace to request no-error-capture upon ...
the final version moved the flags and the associated #defines around
so they were adjacent; unfortunately, they ended up between a comment
and the thing (hw_id) to which the comment applies :(
So this patch reshuffles the comment and subject back together.
Also, as we're touching 'hw_id', let's change it from just 'unsigned'
to a fully-specified 'unsigned int', because some code checking tools
(including checkpatch) object to plain 'unsigned'.
Fixes: bc3d674462 ("drm/i915: Allow userspace to request no-error-capture...")
Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1471616622-6919-1-git-send-email-david.s.gordon@intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
On the blitter (and in test code), we see long sequences of repeated
commands, e.g. XY_PIXEL_BLT, XY_SCANLINE_BLT, or XY_SRC_COPY. For these,
we can skip the hashtable lookup by remembering the previous command
descriptor and doing a straightforward compare of the command header.
The corollary is that we need to do one extra comparison before lookup
up new commands.
v2: Less magic mask (ok, it is still magic, but now you cannot see!)
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/20160818161718.27187-36-chris@chris-wilson.co.uk
The existing code's hashfunction is very suboptimal (most 3D commands
use the same bucket degrading the hash to a long list). The code even
acknowledge that the issue was known and the fix simple:
/*
* If we attempt to generate a perfect hash, we should be able to look at bits
* 31:29 of a command from a batch buffer and use the full mask for that
* client. The existing INSTR_CLIENT_MASK/SHIFT defines can be used for this.
*/
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/20160818161718.27187-35-chris@chris-wilson.co.uk
For simplicity, we want to continue using a contiguous mapping of the
command buffer, but we can reduce the number of vmappings we hold by
switching over to a page-by-page copy from the user batch buffer to the
shadow. The cost for saving one linear mapping is about 5% in trivial
workloads - which is more or less the overhead in calling kmap_atomic().
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/20160818161718.27187-34-chris@chris-wilson.co.uk
The single largest factor in the overhead of parsing the commands is the
setup of the virtual mapping to provide a continuous block for the batch
buffer. If we keep those vmappings around (against the better judgement
of mm/vmalloc.c, which we offset by handwaving and looking suggestively
at the shrinker) we can dramatically improve the performance of the
parser for small batches (such as media workloads). Furthermore, we can
use the prepare shmem read/write functions to determine how best we
need to clflush the range (rather than every page of the object).
The impact of caching both src/dst vmaps is +80% on ivb and +140% on byt
for the throughput on small batches. (Caching just the dst vmap and
iterating over the src, doing a page by page copy is roughly 5% slower
on both platforms. That may be an acceptable trade-off to eliminate one
cached vmapping, and we may be able to reduce the per-page copying overhead
further.) For *this* simple test case, the cmdparser is now within a
factor of 2 of ideal performance.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.william.auld@gmail.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20160818161718.27187-33-chris@chris-wilson.co.uk
We track the LRU access for eviction and bump the last access for the
user GGTT on set-to-gtt. When we do so we need to not only bump the
primary GGTT VMA but all partials as well. Similarly we want to
bump the last access tracking for when unpinning an object from the
scanout so that they do not get promptly evicted and hopefully remain
available for reuse on the next frame.
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/20160818161718.27187-29-chris@chris-wilson.co.uk
The existing ABI says that scanouts are pinned into the mappable region
so that legacy clients (e.g. old Xorg or plymouthd) can write directly
into the scanout through a GTT mapping. However if the surface does not
fit into the mappable region, we are better off just trying to fit it
anywhere and hoping for the best. (Any userspace that is capable of
using ginormous scanouts is also likely not to rely on pure GTT
updates.) With the partial vma fault support, we are no longer
restricted to only using scanouts that we can pin (though it is still
preferred for performance reasons and for powersaving features like
FBC).
v2: Skip fence pinning when not mappable.
v3: Add a comment to explain the possible ramifications of not being
able to use fences for unmappable scanouts.
v4: Rebase to skip over some local patches
v5: Rebase to defer until after we have unmappable GTT fault support
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Deepak S <deepak.s@linux.intel.com>
Cc: Damien Lespiau <damien.lespiau@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20160818161718.27187-27-chris@chris-wilson.co.uk
With the introduction of the reloc page cache, we are just one step away
from refactoring the relocation write functions into one. Not only does
it tidy the code (slightly), but it greatly simplifies the control logic
much to gcc's satisfaction.
v2: Add selftests to document the relationship between the clflush
flags, the KMAP bit and packing into the page-aligned pointer.
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/20160818161718.27187-13-chris@chris-wilson.co.uk
There is an improbable, but not impossible, case that if we leave the
pages unpin as we operate on the object, then somebody via the shrinker
may steal the lock (which lock? right now, it is struct_mutex, THE lock)
and change the cache domains after we have already inspected them.
(Whilst here, avail ourselves of the opportunity to take a couple of
steps to make the two functions look more similar.)
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/20160818161718.27187-11-chris@chris-wilson.co.uk
If we quickly switch from writing through the GTT to a read of the
physical page directly with the CPU (e.g. performing relocations through
the GTT and then running the command parser), we can observe that the
writes are not visible to the CPU. It is not a coherency problem, as
extensive investigations with clflush have demonstrated, but a mere
timing issue - we have to wait for the GTT to complete it's write before
we start our read from the CPU.
The issue can be illustrated in userspace with:
gtt = gem_mmap__gtt(fd, handle, 0, OBJECT_SIZE, PROT_READ | PROT_WRITE);
cpu = gem_mmap__cpu(fd, handle, 0, OBJECT_SIZE, PROT_READ | PROT_WRITE);
gem_set_domain(fd, handle, I915_GEM_DOMAIN_GTT, I915_GEM_DOMAIN_GTT);
for (i = 0; i < OBJECT_SIZE / 64; i++) {
int x = 16*i + (i%16);
gtt[x] = i;
clflush(&cpu[x], sizeof(cpu[x]));
assert(cpu[x] == i);
}
Experimenting with that shows that this behaviour is indeed limited to
recent Atom-class hardware.
Testcase: igt/gem_exec_flush/basic-batch-default-cmd #byt
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/20160818161718.27187-10-chris@chris-wilson.co.uk
