Chris Wilson [Tue, 23 Aug 2016 15:50:24 +0000 (16:50 +0100)]
io-mapping.h: s/PAGE_KERNEL_IO/PAGE_KERNEL/
PAGE_KERNEL_IO is an x86-ism. Though it is used to define the pgprot_t
used for the iomapped region, it itself is just PAGE_KERNEL. On all
other arches, PAGE_KERNEL_IO is undefined so in a general header we must
refrain from using it.
v2: include pgtable for pgprot_combine()
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au> Fixes: cafaf14a5d8f ("io-mapping: Always create a struct to hold metadata about the io-mapping") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: linux-mm@kvack.org Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: http://patchwork.freedesktop.org/patch/msgid/20160823155024.22379-1-chris@chris-wilson.co.uk
Lyude [Mon, 22 Aug 2016 16:50:08 +0000 (12:50 -0400)]
drm/i915/skl: Update plane watermarks atomically during plane updates
Thanks to Ville for suggesting this as a potential solution to pipe
underruns on Skylake.
On Skylake all of the registers for configuring planes, including the
registers for configuring their watermarks, are double buffered. New
values written to them won't take effect until said registers are
"armed", which is done by writing to the PLANE_SURF (or in the case of
cursor planes, the CURBASE register) register.
With this in mind, up until now we've been updating watermarks on skl
like this:
So this patch moves all of the watermark writes into the right place;
inside of the vblank evasion where we update all of the registers for
each plane. While this patch doesn't fix everything, it does allow us to
update the watermark values in the way the hardware expects us to.
Changes since original patch series:
- Remove mutex_lock/mutex_unlock since they don't do anything and we're
not touching global state
- Move skl_write_cursor_wm/skl_write_plane_wm functions into
intel_pm.c, make externally visible
- Add skl_write_plane_wm calls to skl_update_plane
- Fix conditional for for loop in skl_write_plane_wm (level < max_level
should be level <= max_level)
- Make diagram in commit more accurate to what's actually happening
- Add Fixes:
Changes since v1:
- Use IS_GEN9() instead of IS_SKYLAKE() since these fixes apply to more
then just Skylake
- Update description to make it clear this patch doesn't fix everything
- Check if pipes were actually changed before writing watermarks
Changes since v2:
- Write PIPE_WM_LINETIME during vblank evasion
Changes since v3:
- Rebase against new SAGV patch changes
Changes since v4:
- Add a parameter to choose what skl_wm_values struct to use when
writing new plane watermarks
Changes since v5:
- Remove cursor ddb entry write in skl_write_cursor_wm(), defer until
patch 6
- Write WM_LINETIME in intel_begin_crtc_commit()
Changes since v6:
- Remove redundant dirty_pipes check in skl_write_plane_wm (we check
this in all places where we call this function, and it was supposed
to have been removed earlier anyway)
- In i9xx_update_cursor(), use dev_priv->info.gen >= 9 instead of
IS_GEN9(dev_priv). We do this everywhere else and I'd imagine this
needs to be done for gen10 as well
Changes since v7:
- Fix rebase fail (unused variable obj)
- Make struct skl_wm_values *wm const
- Fix indenting
- Use INTEL_GEN() instead of dev_priv->info.gen
Changes since v8:
- Don't forget calls to skl_write_plane_wm() when disabling planes
- Use INTEL_GEN(), not INTEL_INFO()->gen in intel_begin_crtc_commit()
Slightly less straightforward. Some of the drrs calls are done from
workers or from intel_ddi.c, pass along crtc_state when we can,
or crtc->config when we can't.
drm/i915: Remove unused loop from intel_dp_mst_compute_config
Now that conn_state is passed in as argument to compute_config, it's
guaranteed that there is a connector for the argument. The code that
looks for the connector is now dead, and completely unused. Delete it.
drm/i915: Pass crtc_state and connector_state to encoder functions
This is mostly code churn, with exception of a few places:
- intel_display.c has changes in intel_sanitize_encoder
- intel_ddi.c has intel_ddi_fdi_disable calling intel_ddi_post_disable,
and required a function change. Also affects intel_display.c
- intel_dp_mst.c passes a NULL crtc_state and conn_state to
intel_ddi_post_disable for shutting down the real encoder.
If we would pass conn_state, then conn_state->connector !=
intel_dig_port->connector and conn_state->best_encoder !=
to_intel_encoder(intel_dig_port).
We also shouldn't pass crtc_state, because in that case the
disabling sequence may potentially be different depending on
which crtc is disabled last. Nice way to introduce bugs.
No other functional changes are done, diff stat is already huge.
Each encoder type will need to be fixed to use the atomic states
separately.
drm/i915: Pass atomic state to crtc enable/disable functions
This is required for supporting nonblocking modesets. Iterating over
the connector list will no longer be allowed when we don't hold
connection_mutex, so we have to use the atomic state.
Fix disable_noatomic by populating the minimal state required to
disable a connector.
Chris Wilson [Sat, 20 Aug 2016 14:54:08 +0000 (15:54 +0100)]
drm/i915: Ignore stuck requests when considering hangs
If the engine isn't being retired (worker starvation?) then it is
possible for us to repeatedly observe that between consecutive
hangchecks the seqno on the ring to be the same and there remain
unretired requests. Ignore these completely and only regard the engine
as busy for the purpose of hang detection (not stall detection) if there
are outstanding breadcrumbs.
In recent history we have looked at using both the request and seqno as
indication of activity on the engine, but that was reduced to just
inspecting seqno in commit cffa781e5907 ("drm/i915: Simplify check for
idleness in hangcheck"). However, in commit dcff85c8443e ("drm/i915:
Enable i915_gem_wait_for_idle() without holding struct_mutex"), I made
the decision to use the new common lockless function, under the
assumption that request retirement was more frequent than hangcheck and
so we would not have a stuck busy check. The flaw there was in
forgetting that we accumulate the hang score, and so successive checks
seeing a stuck request, albeit with the GPU advancing elsewhere and so
not necessary the same stuck request, would eventually trigger the hang.
Chris Wilson [Mon, 22 Aug 2016 07:44:31 +0000 (08:44 +0100)]
drm/i915: Allow DMA pagetables to use highmem
As we never need to directly access the pages we allocate for scratch and
the pagetables, and always remap them into the GTT through the dma
remapper, we do not need to limit the allocations to lowmem i.e. we can
pass in the __GFP_HIGHMEM flag to the page allocation.
For backwards compatibility, e.g. certain old GPUs not liking highmem
for certain functions that may be accidentally mapped to the scratch
page by userspace, keep the GMCH probe as only allocating from DMA32.
Chris Wilson [Mon, 22 Aug 2016 07:44:30 +0000 (08:44 +0100)]
drm/i915: Embed the scratch page struct into each VM
As the scratch page is no longer shared between all VM, and each has
their own, forgo the small allocation and simply embed the scratch page
struct into the i915_address_space.
David Weinehall [Mon, 22 Aug 2016 10:59:31 +0000 (13:59 +0300)]
drm/i915: debugfs spring cleaning
Just like with sysfs, we do some major overhaul.
Pass dev_priv instead of dev to all feature macros (IS_, HAS_,
INTEL_, etc.). This has the side effect that a bunch of functions
now get dev_priv passed instead of dev.
All calls to INTEL_INFO()->gen have been replaced with
INTEL_GEN().
We want access to to_i915(node->minor->dev) in a lot of places,
so add the node_to_i915() helper to accommodate for this.
Finally, we have quite a few cases where we get a void * pointer,
and need to cast it to drm_device *, only to run to_i915() on it.
Add cast_to_i915() to do this.
v2: Don't introduce extra dev (Chris)
v3: Make pipe_crc_info have a pointer to drm_i915_private instead of
drm_device. This saves a bit of space, since we never use
drm_device anywhere in these functions.
Also some minor fixup that I missed in the previous version.
v4: Changed the code a bit so that dev_priv is passed directly
to various functions, thus removing the need for the
cast_to_i915() helper. Also did some additional cleanup.
v5: Additional cleanup of newly introduced changes.
David Weinehall [Mon, 22 Aug 2016 10:32:44 +0000 (13:32 +0300)]
drm/i915: pdev cleanup
In an effort to simplify things for a future push of dev_priv instead
of dev wherever possible, always take pdev via dev_priv where
feasible, eliminating the direct access from dev. Right now this
only eliminates a few cases of dev, but it also obviates that we pass
dev into a lot of functions where dev_priv would be the more obvious
choice.
v2: Fixed one more place missing in the previous patch set
David Weinehall [Mon, 22 Aug 2016 10:32:43 +0000 (13:32 +0300)]
drm/i915: i915_sysfs.c cleanup
Various cleanup for i915_sysfs.c; we now use dev_priv whenever
possible. The kdev_to_drm_minor() helper function has been
replaced by one that converts from struct device *
to struct drm_i915_private *.
We already have a seemingly identical helper (kdev_to_i915())
in i915_drv.h. But that one cannot be used here.
Unlike the version in i915_drv.h, this helper
reaches i915 through drm_minor.
v2: Rename kdev_to_i915_dm() to kdev_minor_to_i915() (Chris)
David Weinehall [Mon, 22 Aug 2016 10:32:42 +0000 (13:32 +0300)]
drm/i915: consistent struct device naming
We currently have a mix of struct device *device, struct device *kdev,
and struct device *dev (the latter forcing us to refer to
struct drm_device as something else than the normal dev).
To simplify things, always use kdev when referring to struct device.
v2: Replace the dev_to_drm_minor() macro with the inline function
kdev_to_drm_minor().
drm/i915: Fix botched merge that downgrades CSR versions.
Merge commit 5e580523d9128a4d8 reverts the version bumping parts of
commit 4aa7fb9c3c4fa0. Bump the versions again and request the specific
firmware version.
The currently recommended versions are: SKL 1.26, KBL 1.01 and BXT 1.07.
Cc: Patrik Jakobsson <patrik.jakobsson@linux.intel.com> Cc: Imre Deak <imre.deak@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=97242 Cc: drm-intel-fixes@lists.freedesktop.org Fixes: 5e580523d912 ("Backmerge tag 'v4.7' into drm-next") Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1471266567-22443-1-git-send-email-maarten.lankhorst@linux.intel.com Reviewed-by: Imre Deak <imre.deak@intel.com>
Lyude [Wed, 17 Aug 2016 19:55:57 +0000 (15:55 -0400)]
drm/i915/skl: Ensure pipes with changed wms get added to the state
If we're enabling a pipe, we'll need to modify the watermarks on all
active planes. Since those planes won't be added to the state on
their own, we need to add them ourselves.
Signed-off-by: Lyude <cpaul@redhat.com> Reviewed-by: Matt Roper <matthew.d.roper@intel.com> Cc: stable@vger.kernel.org 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/1471463761-26796-6-git-send-email-cpaul@redhat.com
Matt Roper [Wed, 17 Aug 2016 19:55:55 +0000 (15:55 -0400)]
drm/i915/gen9: Only copy WM results for changed pipes to skl_hw
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: 734fa01f3a17 ("drm/i915/gen9: Calculate watermarks during atomic 'check' (v2)") Fixes: 9b6130227495 ("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
Lyude [Wed, 17 Aug 2016 19:55:54 +0000 (15:55 -0400)]
drm/i915/skl: Add support for the SAGV, fix underrun hangs
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
Lyude [Wed, 17 Aug 2016 19:55:53 +0000 (15:55 -0400)]
drm/i915/gen6+: Interpret mailbox error flags
In order to add proper support for the SAGV, we need to be able to know
what the cause of a failure to change the SAGV through the pcode mailbox
was. The reasoning for this is that some very early pre-release Skylake
machines don't actually allow you to control the SAGV on them, and
indicate an invalid mailbox command was sent.
This also might come in handy in the future for debugging.
Changes since v1:
- Add functions for interpreting gen6 mailbox error codes along with
gen7+ error codes, and actually interpret those codes properly
- Renamed patch to reflect new behavior
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/1471463761-26796-2-git-send-email-cpaul@redhat.com
[mlankhorst: -ENOSYS -> -ENXIO for checkpatch]
Paulo Zanoni [Wed, 17 Aug 2016 19:41:44 +0000 (16:41 -0300)]
drm/i915: Call intel_fbc_pre_update() after pinning the new pageflip
intel_fbc_pre_update() depends upon the new state being already pinned
in place in the Global GTT (primarily for both fencing which wants both
an offset and a fence register, if assigned). This requires the call to
intel_fbc_pre_update() be after intel_pin_and_fence_fb() - but commit e8216e502aca ("drm/i915/fbc: call intel_fbc_pre_update earlier during
page flips") moved the code way too much up in its attempt to call it
before the page flip.
v2 (from Paulo):
- Point the original bad commit.
- Add a comment to maybe prevent further regressions.
Fixes: e8216e502aca ("drm/i915/fbc: call intel_fbc_pre_update earlier...") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Cc: Daniel Vetter <daniel.vetter@intel.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Cc: Patrik Jakobsson <patrik.jakobsson@linux.intel.com> Cc: drm-intel-fixes@lists.freedesktop.org Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Link: http://patchwork.freedesktop.org/patch/msgid/1471462904-842-1-git-send-email-paulo.r.zanoni@intel.com Cc: stable@vger.kernel.org
Daniel Vetter [Mon, 22 Aug 2016 08:55:22 +0000 (10:55 +0200)]
drm/i915: Ensure consistent control flow __i915_gem_active_get_rcu
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).
Chris Wilson [Mon, 22 Aug 2016 07:44:29 +0000 (08:44 +0100)]
drm/i915: Stop marking the unaccessible scratch page as UC
Since by design, if not entirely by practice, nothing is allowed to
access the scratch page we use to background fill the VM, then we do not
need to ensure that it is coherent between the CPU and GPU.
set_pages_uc() does a stop_machine() after changing the PAT, and that
significantly impacts upon context creation throughput.
Chris Wilson [Fri, 19 Aug 2016 15:54:28 +0000 (16:54 +0100)]
drm/i915: Use remap_io_mapping() to prefault all PTE in a single pass
Very old numbers indicate this is a 66% improvement when remapping the
entire object for fence contention - due to the elimination of
track_pfn_insert and its strcmp.
Chris Wilson [Fri, 19 Aug 2016 15:54:26 +0000 (16:54 +0100)]
io-mapping: Always create a struct to hold metadata about the io-mapping
Currently, we only allocate a structure to hold metadata if we need to
allocate an ioremap for every access, such as on x86-32. However, it
would be useful to store basic information about the io-mapping, such as
its page protection, on all platforms.
Chris Wilson [Fri, 19 Aug 2016 15:54:24 +0000 (16:54 +0100)]
drm/i915/fbc: Don't set an illegal fence if unfenced
If the frontbuffer doesn't have an associated fence, it will have a
fence reg of -1. If we attempt to OR in this register into the FBC
control register we end up setting all control bits, oops!
Chris Wilson [Fri, 19 Aug 2016 15:54:23 +0000 (16:54 +0100)]
drm/i915: Flush delayed fence releases after reset
What I never hit in testing, but Mika immediately did, was a GPU hang
with a pending fence release (where a tiled object has been changed by
the user to be untiled, and the update has not yet been committed to the
fence register). As the stride/tiling is 0, this causes a divide-by-zero
error when trying to write the new fence parameters:
Reported-by: Mika Kuoppala <mika.kuoppala@intel.com> Fixes: 49ef5294cda2 ("drm/i915: Move fence tracking from object to vma") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20160819155428.1670-1-chris@chris-wilson.co.uk
Dave Gordon [Fri, 19 Aug 2016 14:23:42 +0000 (15:23 +0100)]
drm/i915: Reattach comment, complete type specification
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'.
Chris Wilson [Thu, 18 Aug 2016 16:17:17 +0000 (17:17 +0100)]
drm/i915/cmdparser: Use binary search for faster register lookup
A significant proportion of the cmdparsing time for some batches is the
cost to find the register in the mmiotable. We ensure that those tables
are in ascending order such that we could do a binary search if it was
ever merited. It is.
Chris Wilson [Thu, 18 Aug 2016 16:17:15 +0000 (17:17 +0100)]
drm/i915/cmdparser: Compare against the previous command descriptor
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!)
Chris Wilson [Thu, 18 Aug 2016 16:17:14 +0000 (17:17 +0100)]
drm/i915/cmdparser: Improve hash function
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.
*/
Chris Wilson [Thu, 18 Aug 2016 16:17:13 +0000 (17:17 +0100)]
drm/i915/cmdparser: Only cache the dst vmap
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().
Chris Wilson [Thu, 18 Aug 2016 16:17:12 +0000 (17:17 +0100)]
drm/i915/cmdparser: Use cached vmappings
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.
Chris Wilson [Thu, 18 Aug 2016 16:17:11 +0000 (17:17 +0100)]
drm/i915/cmdparser: Add the TIMESTAMP register for the other engines
Since I have been using the BCS_TIMESTAMP to measure latency of
execution upon the blitter ring, allow regular userspace to also read
from that register. They are already allowed RCS_TIMESTAMP!
Chris Wilson [Thu, 18 Aug 2016 16:17:10 +0000 (17:17 +0100)]
drm/i915/cmdparser: Make initialisation failure non-fatal
If the developer adds a register in the wrong order, we BUG during boot.
That makes development and testing very difficult. Let's be a bit more
friendly and disable the command parser with a big warning if the tables
are invalid.
Chris Wilson [Thu, 18 Aug 2016 16:17:09 +0000 (17:17 +0100)]
drm/i915: Stop discarding GTT cache-domain on unbind vma
Since commit 43566dedde54 ("drm/i915: Broaden application of
set-domain(GTT)") we allowed objects to be in the GTT domain, but unbound.
Therefore removing the GTT cache domain when removing the GGTT vma is no
longer semantically correct.
An unfortunate side-effect is we lose the wondrously named
i915_gem_object_finish_gtt(), not to be confused with
i915_gem_gtt_finish_object()!
Chris Wilson [Thu, 18 Aug 2016 16:17:08 +0000 (17:17 +0100)]
drm/i915: Bump the inactive tracking for all VMA accessed
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.
Chris Wilson [Thu, 18 Aug 2016 16:17:07 +0000 (17:17 +0100)]
drm/i915: Track display alignment on VMA
When using the aliasing ppgtt and pageflipping with the shrinker/eviction
active, we note that we often have to rebind the backbuffer before
flipping onto the scanout because it has an invalid alignment. If we
store the worst-case alignment required for a VMA, we can avoid having
to rebind at critical junctures.
Chris Wilson [Thu, 18 Aug 2016 16:17:06 +0000 (17:17 +0100)]
drm/i915: Fallback to using unmappable memory for scanout
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
Chris Wilson [Thu, 18 Aug 2016 16:17:05 +0000 (17:17 +0100)]
drm/i915: Choose not to evict faultable objects from the GGTT
Often times we do not want to evict mapped objects from the GGTT as
these are quite expensive to teardown and frequently reused (causing an
equally, if not more so, expensive setup). In particular, when faulting
in a new object we want to avoid evicting an active object, or else we
may trigger a page-fault-of-doom as we ping-pong between evicting two
objects.
Chris Wilson [Thu, 18 Aug 2016 16:17:04 +0000 (17:17 +0100)]
drm/i915: Drop ORIGIN_GTT for untracked GTT writes
If FBC is set on a framebuffer that is unmapped, all GTT faults will be
from a partial mapping. Writes by the user through the partial VMA are
then untracked by the FBC and so we must use the ORIGIN_CPU when flushing
the I915_GEM_DOMAIN_GTT.
Chris Wilson [Thu, 18 Aug 2016 16:17:03 +0000 (17:17 +0100)]
drm/i915: Convert partial ggtt vma to full ggtt if it spans the entire object
If we want to create a partial vma from a chunk that is the same size as
the object, create a normal ggtt vma instead. The benefit is that it
will match future requests for the normal ggtt.
Chris Wilson [Thu, 18 Aug 2016 16:17:02 +0000 (17:17 +0100)]
drm/i915: Fix partial GGTT faulting
We want to always use the partial VMA as a fallback for a failure to
bind the object into the GGTT. This extends the support partial objects
in the GGTT to cover everything, not just objects too large.
Chris Wilson [Thu, 18 Aug 2016 16:17:01 +0000 (17:17 +0100)]
drm/i915: Choose partial chunksize based on tile row size
In order to support setting up fences for partial mappings of an object,
we have to align those mappings with the fence. The minimum chunksize we
choose is at least the size of a single tile row.
v2: Make minimum chunk size a define for later use
Chris Wilson [Thu, 18 Aug 2016 16:16:59 +0000 (17:16 +0100)]
drm/i915: Rename fence.lru_list to link
Our current practice is to only name the actual list (here
dev_priv->fence_list) using "list", and elements upon that list are
referred to as "link". Further, the lru nature is of the list and not of
the node and including in the name does not disambiguate the link from
anything else.
Chris Wilson [Thu, 18 Aug 2016 16:16:58 +0000 (17:16 +0100)]
drm/i915/userptr: Make gup errors stickier
Keep any error reported by the gup_worker until we are notified that the
arena has changed (via the mmu-notifier). This has the importance of
making two consecutive calls to i915_gem_object_get_pages() reporting
the same error, and curtailing a loop of detecting a fault and requeueing
a gup_worker.
Chris Wilson [Thu, 18 Aug 2016 16:16:57 +0000 (17:16 +0100)]
drm/i915: Allocate rings from stolen
If we have stolen available, make use of it for ringbuffer allocation.
Previously this was restricted to !llc platforms, as writing to stolen
requires a GGTT mapping - but now that we have partial mappable support,
the mappable aperture isn't quite so precious so we can use it more
freely and ringbuffers are a good user for the otherwise wasted stolen.
Chris Wilson [Thu, 18 Aug 2016 16:16:56 +0000 (17:16 +0100)]
drm/i915: Allow ringbuffers to be bound anywhere
Now that we have WC vmapping available, we can bind our rings anywhere
in the GGTT and do not need to restrict them to the mappable region.
Except for stolen objects, for which direct access is verbatim and we
must use the mappable aperture.
Chris Wilson [Thu, 18 Aug 2016 16:16:55 +0000 (17:16 +0100)]
drm/i915: Move map-and-fenceable tracking to the VMA
By moving map-and-fenceable tracking from the object to the VMA, we gain
fine-grained tracking and the ability to track individual fences on the VMA
(subsequent patch).
Chris Wilson [Thu, 18 Aug 2016 16:16:53 +0000 (17:16 +0100)]
drm/i915: Fallback to single page GTT mmappings for relocations
If we cannot pin the entire object into the mappable region of the GTT,
try to pin a single page instead. This is much more likely to succeed,
and prevents us falling back to the clflush slow path.
Chris Wilson [Thu, 18 Aug 2016 16:16:52 +0000 (17:16 +0100)]
drm/i915: Refactor execbuffer relocation writing
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.
Chris Wilson [Thu, 18 Aug 2016 16:16:50 +0000 (17:16 +0100)]
drm/i915: Pin the pages first in shmem prepare read/write
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.)
Chris Wilson [Thu, 18 Aug 2016 16:16:49 +0000 (17:16 +0100)]
drm/i915: Wait for writes through the GTT to land before reading back
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.
Chris Wilson [Thu, 18 Aug 2016 16:16:48 +0000 (17:16 +0100)]
drm/i915: Before accessing an object via the cpu, flush GTT writes
If we want to read the pages directly via the CPU, we have to be sure
that we have to flush the writes via the GTT (as the CPU can not see
the address aliasing).
This is a companion to i915_gem_obj_prepare_shmem_read() that prepares
the backing storage for direct writes. It first serialises with the GPU,
pins the backing storage and then indicates what clfushes are required in
order for the writes to be coherent.
Whilst here, fix support for ancient CPUs without clflush for which we
cannot do the GTT+clflush tricks.
v2: Add i915_gem_obj_finish_shmem_access() for symmetry
Chris Wilson [Thu, 18 Aug 2016 16:16:46 +0000 (17:16 +0100)]
drm/i915: Cache kmap between relocations
When doing relocations, we have to obtain a mapping to the page
containing the target address. This is either a kmap or iomap depending
on GPU and its cache coherency. Neighbouring relocation entries are
typically within the same page and so we can cache our kmapping between
them and avoid those pesky TLB flushes.
Note that there is some sleight-of-hand in how the slow relocate works
as the reloc_entry_cache implies pagefaults disabled (as we are inside a
kmap_atomic section). However, the slow relocate code is meant to be the
fallback from the atomic fast path failing. Fortunately it works as we
already have performed the copy_from_user for the relocation array (no
more pagefaults there) and the kmap_atomic cache is enabled after we
have waited upon an active buffer (so no more sleeping in atomic).
Magic!
Chris Wilson [Thu, 18 Aug 2016 16:16:45 +0000 (17:16 +0100)]
drm/i915: Fallback to single page pwrite/pread if unable to release fence
If we cannot release the fence (for example if someone is inexplicably
trying to write into a tiled framebuffer that is currently pinned to the
display! *cough* kms_frontbuffer_tracking *cough*) fallback to using the
page-by-page pwrite/pread interface, rather than fail the syscall
entirely.
Since this is triggerable by the user (along pwrite) we have to remove
the WARN_ON(fence->pin_count).
Chris Wilson [Thu, 18 Aug 2016 16:16:44 +0000 (17:16 +0100)]
drm/i915: Mark up the GTT flush following WC writes as ORIGIN_CPU
Similarly to invalidating beforehand, if the object is mmapped via
I915_MMAP_WC we cannot track writes through the I915_GEM_DOMAIN_GTT. At
the conclusion of the write, i915_gem_object_flush_gtt_writes() we also
need to treat the origin carefully in case it may have been untracked.
See also commit aeecc9696aa0 ("drm/i915: use ORIGIN_CPU for frontbuffer
invalidation on WC mmaps").
Chris Wilson [Thu, 18 Aug 2016 16:16:43 +0000 (17:16 +0100)]
drm/i915: Use ORIGIN_CPU for fb invalidation from pwrite
As pwrite does not use the fence for its GTT access, and may even go
through a secondary interface avoiding the main VMA, we cannot treat the
write as automatically invalidated by the hardware and so we require
ORIGIN_CPU frontbufer invalidate/flushes.
Chris Wilson [Thu, 18 Aug 2016 16:16:41 +0000 (17:16 +0100)]
agp/intel: Flush chipset writes after updating a single PTE
After we update one PTE for a page, the caller expects to be able to
immediately use that through a GGTT read/write. To comply with the
callers expectations we therefore need to flush the chipset buffers
before returning.
Reported-by: Matti Hämäläinen <ccr@tnsp.org> Fixes: d6473f566417 ("drm/i915: Add support for mapping an object page...") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Ankitprasad Sharma <ankitprasad.r.sharma@intel.com> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Tested-by: Matti Hämäläinen <ccr@tnsp.org> Cc: drm-intel-fixes@lists.freedesktop.org Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20160818161718.27187-2-chris@chris-wilson.co.uk
Chris Wilson [Thu, 18 Aug 2016 16:16:40 +0000 (17:16 +0100)]
drm/i915: Unconditionally flush any chipset buffers before execbuf
If userspace is asynchronously streaming into the batch or other
execobjects, we may not flush those writes along with a change in cache
domain (as there is no change). Therefore those writes may end up in
internal chipset buffers and not visible to the GPU upon execution. We
must issue a flush command or otherwise we encounter incoherency in the
batchbuffers and the GPU executing invalid commands (i.e. hanging) quite
regularly.
v2: Throw a paranoid wmb() into the general flush so that we remain
consistent with before.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=90841 Fixes: 1816f9236303 ("drm/i915: Support creation of unbound wc user...") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Akash Goel <akash.goel@intel.com> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> Tested-by: Matti Hämäläinen <ccr@tnsp.org> Cc: stable@vger.kernel.org Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20160818161718.27187-1-chris@chris-wilson.co.uk
Matt Roper [Fri, 17 Jun 2016 20:42:20 +0000 (13:42 -0700)]
drm/i915/gen9: Drop invalid WARN() during data rate calculation
It's possible to have a non-zero plane mask and still wind up with a
total data rate of zero. There are two cases where this can happen:
* planes are active (from the KMS point of view), but are
all fully clipped (positioned offscreen)
* the only active plane on a CRTC is the cursor (which is handled
independently and not counted into the general data rate computations
These are both valid display setups (although unusual), so we need to
drop the WARN().
Matt Roper [Fri, 17 Jun 2016 20:42:18 +0000 (13:42 -0700)]
drm/i915/gen9: Initialize intel_state->active_crtcs during WM sanitization (v2)
intel_state->active_crtcs is usually only initialized when doing a
modeset. During our first atomic commit after boot, we're effectively
faking a modeset to sanitize the DDB/wm setup, so ensure that this field
gets initialized before use.
v2:
- Don't clobber active_crtcs if our first commit really is a modeset
(Maarten)
- Grab connection_mutex when faking a modeset during sanitization
(Maarten)
Chris Wilson [Wed, 17 Aug 2016 10:33:33 +0000 (11:33 +0100)]
drm/i915: Remember to set vma->pages for the preallocated stolen object
In commit 247177ddd517 ("drm/i915: Always set the vma->pages"), as it
title implies, we always set vma->pages for bound objects. Even before
that, we would set vma->ggtt_view.pages, for globally bound objects.
This was forgotten for the fixup inside the preallocated stolen objects,
which has to recreate a global GTT binding outside of the usual VMA
insertion path
Chris Wilson [Wed, 17 Aug 2016 11:09:06 +0000 (12:09 +0100)]
drm/i915: Mark i915_hpd_poll_init_work as static
Local function with forgotten static declaration.
Fixes: 19625e85c6ec ("drm/i915: Enable polling when we don't have hpd") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: Lyude <cpaul@redhat.com> Link: http://patchwork.freedesktop.org/patch/msgid/1471432146-5196-2-git-send-email-chris@chris-wilson.co.uk Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Tvrtko Ursulin [Tue, 16 Aug 2016 16:04:21 +0000 (17:04 +0100)]
drm/i915: Initialize legacy semaphores from engine hw id indexed array
Build the legacy semaphore initialisation array using the engine
hardware ids instead of driver internal ones. This makes the
static array size dependent only on the number of gen6 semaphore
engines.
Also makes the per-engine semaphore wait and signal tables
hardware id indexed saving some more space.
v2: Refactor I915_GEN6_NUM_ENGINES to GEN6_SEMAPHORE_LAST. (Chris Wilson)
v3: More polish. (Chris Wilson)
Tvrtko Ursulin [Tue, 16 Aug 2016 16:04:20 +0000 (17:04 +0100)]
drm/i915: Add enum for hardware engine identifiers
Put the engine hardware id in the common header so they are
not only associated with the GuC since they are needed for
the legacy semaphores implementation.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Chris Wilson [Tue, 16 Aug 2016 08:50:40 +0000 (09:50 +0100)]
drm/i915: Embrace the race in busy-ioctl
Daniel Vetter proposed a new challenge to the serialisation inside the
busy-ioctl that exposed a flaw that could result in us reporting the
wrong engine as being busy. If the request is reallocated as we test
its busyness and then reassigned to this object by another thread, we
would not notice that the test itself was incorrect.
We are faced with a choice of using __i915_gem_active_get_request_rcu()
to first acquire a reference to the request preventing the race, or to
acknowledge the race and accept the limitations upon the accuracy of the
busy flags. Note that we guarantee that we never falsely report the
object as idle (providing userspace itself doesn't race), and so the
most important use of the busy-ioctl and its guarantees are fulfilled.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Daniel Vetter <daniel.vetter@intel.com> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Reviewed-by: Daniel Vetter <daniel.vetter@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1471337440-16777-1-git-send-email-chris@chris-wilson.co.uk
Chris Wilson [Mon, 15 Aug 2016 09:49:10 +0000 (10:49 +0100)]
drm/i915: Only record active and pending requests upon a GPU hang
There is no other state pertaining to the completed requests in the
hang, other than gleamed through the ringbuffer, so including the
expired requests in the list of outstanding requests simply adds noise.
Chris Wilson [Mon, 15 Aug 2016 09:49:09 +0000 (10:49 +0100)]
drm/i915: Print the batchbuffer offset next to BBADDR in error state
It is useful when looking at captured error states to check the recorded
BBADDR register (the address of the last batchbuffer instruction loaded)
against the expected offset of the batch buffer, and so do a quick check
that (a) the capture is true or (b) HEAD hasn't wandered off into the
badlands.
Chris Wilson [Mon, 15 Aug 2016 09:49:08 +0000 (10:49 +0100)]
drm/i915: Move debug only per-request pid tracking from request to ctx
Since contexts are not currently shared between userspace processes, we
have an exact correspondence between context creator and guilty batch
submitter. Therefore we can save some per-batch work by inspecting the
context->pid upon error instead. Note that we take the context's
creator's pid rather than the file's pid in order to better track fd
passed over sockets.
Chris Wilson [Mon, 15 Aug 2016 09:49:07 +0000 (10:49 +0100)]
drm/i915: Introduce i915_ggtt_offset()
This little helper only exists to safely discard the upper unused 32bits
of the general 64-bit VMA address - as we know that all Global GTT
currently are less than 4GiB in size and so that the upper bits must be
zero. In many places, we use a u32 for the global GTT offset and we want
to document where we are discarding the full VMA offset.
Chris Wilson [Mon, 15 Aug 2016 09:49:06 +0000 (10:49 +0100)]
drm/i915: Track pinned VMA
Treat the VMA as the primary struct responsible for tracking bindings
into the GPU's VM. That is we want to treat the VMA returned after we
pin an object into the VM as the cookie we hold and eventually release
when unpinning. Doing so eliminates the ambiguity in pinning the object
and then searching for the relevant pin later.
projects / karo-tx-linux.git / log