The size has been increased to 2MB starting from Gen11. GuC and HuC FWs
fit in 1MB so we were fine even with the legacy define, but let's still
move to the correct one before the blobs grow to avoid being caught off
guard in the future.
v2: return early if the platform doesn't have GuC, fix nits (Michal)
Bspec: 12690 Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com> Cc: Michal Wajdeczko <michal.wajdeczko@intel.com> Reviewed-by: Michal Wajdeczko <michal.wajdeczko@intel.com> Reviewed-by: Stuart Summers <stuart.summers@intel.com>
[ickle: use SZ consistently] Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Link: https://patchwork.freedesktop.org/patch/msgid/20190606224225.14287-2-daniele.ceraolospurio@intel.com
drm/i915/guc: always use Command Transport Buffers
Now that we've moved the Gen9 GuC blobs to version 32 we have CTB
support on all gens, so no need to restrict the usage to Gen11+.
Note that MMIO communication is still required for CTB initialization.
Hans de Goede [Wed, 5 Jun 2019 18:17:35 +0000 (20:17 +0200)]
drm/i915/dsi: Read back pclk set by GOP and use that as pclk (v3)
The GOP sometimes initializes the pclk at a (slightly) different frequency
then the pclk which we've calculated.
This commit makes the DSI code read-back the pclk set by the GOP and
if that is within a reasonable margin of the calculated pclk, uses
that instead.
This fixes the first modeset being a full modeset instead of a
fast modeset on systems where the GOP pclk is different.
Changes in v2:
-Use intel_encoder_current_mode() to get the pclk setup by the GOP
Changes in v3:
-Back to the readback approach, skipping the dsi_pll.ctrl / .dev checks
in intel_pipe_config_compare() when adjust is set leads to:
[drm:pipe_config_err [i915]] *ERROR* mismatch in dsi_pll.ctrl (...)
[drm:pipe_config_err [i915]] *ERROR* mismatch in dsi_pll.div (...)
-Do the readback and pclk overriding from vlv_dsi_init(), rather then from
intel_dsi_vbt_init() as the vbt code should not be touching the hw
Hans de Goede [Wed, 5 Jun 2019 18:17:34 +0000 (20:17 +0200)]
drm/i915/dsi: Move vlv/icl_dphy_param_init call out of intel_dsi_vbt_init (v2)
The vlv/icl_dphy_param_init calls do various calculations to set dphy
parameters based on the pclk.
Move the calling of vlv/icl_dphy_param_init to vlv_dsi_init to give
vlv_dsi_init a chance to tweak the pclk before these calculations are done.
Changes in v2:
-Also moves the icl and vlv specific dphy_param_init functions from the
generic intel_dsi_vbt.c file into the icl_ and vlv_dsi.c specific files.
Note icl_dphy_param_init() and vlv_dphy_param_init() are only moved,
otherwise they are completely unchanged.
Ville Syrjälä [Wed, 10 Apr 2019 17:08:35 +0000 (20:08 +0300)]
drm/i915/sdvo: Check that we have space for the infoframe
Before we go writing the infoframe let's make sure we have
the space for it. Not that it really matters since the write
loop would just terminate early in that case.
Ville Syrjälä [Tue, 9 Apr 2019 14:40:50 +0000 (17:40 +0300)]
drm/i915: Rename SDVO_AUDIO_ENABLE to HDMI_AUDIO_ENABLE
The "audio enable" bit on the SDVO/HDMI control register is only meant
for HDMI. Audio is never delivered over the SDVO bus. Rename the define
to reflect this fact.
Ville Syrjälä [Tue, 9 Apr 2019 14:40:49 +0000 (17:40 +0300)]
drm/i915/sdvo: Implement proper HDMI audio support for SDVO
Our SDVO audio support is pretty bogus. We can't push audio over the
SDVO bus, so trying to enable audio in the SDVO control register doesn't
do anything. In fact it looks like the SDVO encoder will always mix in
the audio coming over HDA, and there's no (at least documented) way to
disable that from our side. So HDMI audio does work currently on gen4
but only by luck really. On gen3 it got broken by the referenced commit.
And what has always been missing on every platform is the ELD.
To pass the ELD to the audio driver we need to write it to magic buffer
in the SDVO encoder hardware which then gets pulled out via HDA in the
other end. Ie. pretty much the same thing we had for native HDMI before
we started to just pass the ELD between the drivers. This sort of
explains why we even have that silly hardware buffer with native HDMI.
$ cat /proc/asound/card0/eld#1.0
-monitor_present 0
-eld_valid 0
+monitor_present 1
+eld_valid 1
+monitor_name LG TV
+connection_type HDMI
+...
This also fixes our state readout since we can now query the SDVO
encoder about the state of the "ELD valid" and "presence detect"
bits. As mentioned those don't actually control whether audio
gets sent over the HDMI cable, but it's the best we can do. And with
the state checker appeased we can re-enable HDMI audio for gen3.
Cc: stable@vger.kernel.org Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: zardam@gmail.com Tested-by: zardam@gmail.com
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108976 Fixes: de44e256b92c ("drm/i915/sdvo: Shut up state checker with hdmi cards on gen3") Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20190409144054.24561-3-ville.syrjala@linux.intel.com Reviewed-by: Imre Deak <imre.deak@intel.com>
Ville Syrjälä [Tue, 9 Apr 2019 14:40:48 +0000 (17:40 +0300)]
drm/i915/sdvo: Fix AVI infoframe TX rate readout
The AVI infoframe readout code currently issues a
SDVO_CMD_GET_HBUF_TXRATE before SDVO_CMD_SET_HBUF_INDEX, which is
not the correct order for these two operations. So far this wasn't
a problem since we left the index pointing at the AVI infoframe
buffer at the end of the modeset. However once we start to write
to other buffers (namely ELD) that is no longer going to be true.
Fix up the order so that we always read out the TX rate for the
correct buffer.
Ville Syrjälä [Mon, 3 Jun 2019 14:25:00 +0000 (17:25 +0300)]
drm/i915: Fix per-pixel alpha with CCS
We forgot to set .has_alpha=true for the A+CCS formats when the code
started to consult .has_alpha. This manifests as A+CCS being treated
as X+CCS which means no per-pixel alpha blending. Fix the format
list appropriately.
Cc: stable@vger.kernel.org Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Cc: Matt Roper <matthew.d.roper@intel.com> Cc: Heinrich Fink <heinrich.fink@daqri.com> Reported-by: Heinrich Fink <heinrich.fink@daqri.com> Tested-by: Heinrich Fink <heinrich.fink@daqri.com> Fixes: b20815255693 ("drm/i915: Add plane alpha blending support, v2.") Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20190603142500.25680-1-ville.syrjala@linux.intel.com Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Ville Syrjälä [Tue, 4 Jun 2019 14:02:14 +0000 (17:02 +0300)]
drm/i915: Drop pointless WARN_ON
intel_dp_link_down() is static and it's only called from the pre-ddi
DP functions, so having a WARN_ON(HAS_DDI) in there is quite pointless.
Remove it.
Ville Syrjälä [Tue, 4 Jun 2019 14:02:13 +0000 (17:02 +0300)]
drm/i915: Move intel_dp->prepare_link_train assignment into ddi code
It's a bit silly to go through intel_dp.c to assign the
prepare_link_train vfunc for ddi platforms when we can just
assign it directly from intel_ddi.c.
Tvrtko Ursulin [Fri, 7 Jun 2019 10:15:35 +0000 (11:15 +0100)]
drm/i915: Make Gen6/7 RING_FAULT_REG access engine centric
Similar to earlier conversions, eliminate the implicit dev_priv by
introducing some helpers which take the engine parameter (since the
register itself is per engine).
Lucas De Marchi [Wed, 5 Jun 2019 23:55:35 +0000 (16:55 -0700)]
drm/i915/dmc: protect against reading random memory
While loading the DMC firmware we were double checking the headers made
sense, but in no place we checked that we were actually reading memory
we were supposed to. This could be wrong in case the firmware file is
truncated or malformed.
Before this patch:
# ls -l /lib/firmware/i915/icl_dmc_ver1_07.bin
-rw-r--r-- 1 root root 25716 Feb 1 12:26 icl_dmc_ver1_07.bin
# truncate -s 25700 /lib/firmware/i915/icl_dmc_ver1_07.bin
# modprobe i915
# dmesg| grep -i dmc
[drm:intel_csr_ucode_init [i915]] Loading i915/icl_dmc_ver1_07.bin
[drm] Finished loading DMC firmware i915/icl_dmc_ver1_07.bin (v1.7)
i.e. it loads random data. Now it fails like below:
[drm:intel_csr_ucode_init [i915]] Loading i915/icl_dmc_ver1_07.bin
[drm:csr_load_work_fn [i915]] *ERROR* Truncated DMC firmware, rejecting.
i915 0000:00:02.0: Failed to load DMC firmware i915/icl_dmc_ver1_07.bin. Disabling runtime power management.
i915 0000:00:02.0: DMC firmware homepage: https://git.kernel.org/pub/scm/linux/kernel/git/firmware/linux-firmware.git/tree/i915
Before reading any part of the firmware file, validate the input first.
Chris Wilson [Fri, 31 May 2019 11:32:45 +0000 (12:32 +0100)]
drm/i915: Report an earlier wedged event when suspending the engines
On i915_gem_load_power_context() we do care whether or not we succeed in
completing the switch back to the kernel context (via idling the
engines). Currently, we detect if an error occurs while we wait, but we
do not report one if it occurred beforehand (and the status of the
switch is undefined). Check the current terminally wedged status on
entering the wait, and report it after flushing the requests, as if it
had occurred during our own wait.
Chris Wilson [Tue, 4 Jun 2019 15:24:08 +0000 (16:24 +0100)]
drm/i915: Skip context_barrier emission for unused contexts
The intent was to skip unused HW contexts by checking ce->state.
However, this only works for execlists where the ppGTT pointers is
stored inside the HW context. For gen7, the ppGTT is alongside the
logical state and must be updated on all active engines but, crucially,
only on active engines. As we need different checks, and to keep
context_barrier_task() agnostic, pass in the predicate.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110836 Fixes: 62c8e423450d ("drm/i915: Skip unused contexts for context_barrier_task()") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20190604152408.24468-1-chris@chris-wilson.co.uk
Fixes: 10be98a77c55 ("drm/i915: Move more GEM objects under gem/") Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com> Signed-off-by: Jani Nikula <jani.nikula@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20190605095657.23601-1-jani.nikula@intel.com
Matt Roper [Wed, 5 Jun 2019 21:18:32 +0000 (14:18 -0700)]
drm/i915/ehl: Support HBR3 on EHL combo PHY
Unlike ICL, EHL's combo PHYs can support HBR3 data rates. Note that
this just extends the upper limit; we will continue to honor the max
data rate specified in the VBT in cases where it is lower than HBR3.
Hans de Goede [Fri, 24 May 2019 17:40:27 +0000 (19:40 +0200)]
drm/i915/dsi: Use a fuzzy check for burst mode clock check
Prior to this commit we fail to init the DSI panel on the GPD MicroPC:
https://www.indiegogo.com/projects/gpd-micropc-6-inch-handheld-industry-laptop#/
The problem is intel_dsi_vbt_init() failing with the following error:
*ERROR* Burst mode freq is less than computed
The pclk in the VBT panel modeline is 70000, together with 24 bpp and
4 lines this results in a bitrate value of 70000 * 24 / 4 = 420000.
But the target_burst_mode_freq in the VBT is 418000.
This commit works around this problem by adding an intel_fuzzy_clock_check
when target_burst_mode_freq < bitrate and setting target_burst_mode_freq to
bitrate when that checks succeeds, fixing the panel not working.
Jani Nikula [Fri, 31 May 2019 13:14:55 +0000 (16:14 +0300)]
drm/i915/bios: use port info child pointer to determine LSPCON presence
Avoid iterating the child devices. This should be a non-functional
change, but theoretically this might enable LSPCON on some extra ports
with buggy VBTs.
Jani Nikula [Fri, 31 May 2019 13:14:52 +0000 (16:14 +0300)]
drm/i915/bios: store child device pointer in DDI port info
This allows us to avoid iterating the child devices in some cases.
Also replace the presence bit with child device being non-NULL, and set
the child device pointer last to allow us to take advantage of it in
follow-up work.
Swati Sharma [Wed, 29 May 2019 09:50:52 +0000 (15:20 +0530)]
drm/i915: Enable intel_color_get_config()
In this patch, intel_color_get_config() is enabled and support
for read_luts() will be added platform by platform incrementally
in the follow-up patches.
v4: -Renamed intel_get_color_config to intel_color_get_config [Jani]
-Added the user early on such that support for get_color_config()
can be added platform by platform incrementally [Jani]
v5: -Incorrect place for calling intel_color_get_config() in
haswell_get_pipe_config() [Ville]
v6: -Renamed intel_color_read_luts() to intel_color_get_config()
[Jani and Ville]
Swati Sharma [Wed, 29 May 2019 09:50:51 +0000 (15:20 +0530)]
drm/i915: Introduce vfunc read_luts() to create hw lut
In this patch, a vfunc read_luts() is introduced to create a hw lut
i.e. lut having values read from gamma/degamma registers which will
later be used to compare with sw lut to validate gamma/degamma lut values.
v3: -Rebase
v4: -Renamed intel_get_color_config to intel_color_get_config [Jani]
-Wrapped get_color_config() [Jani]
v5: -Renamed intel_color_get_config() to intel_color_read_luts()
-Renamed get_color_config to read_luts
v6: -Renamed intel_color_read_luts() back to intel_color_get_config()
[Jani and Ville]
Chris Wilson [Tue, 4 Jun 2019 15:38:30 +0000 (16:38 +0100)]
drm/i915/gtt: Replace struct_mutex serialisation for allocation
Instead of relying on the caller holding struct_mutex across the
allocation, push the allocation under a tree of spinlocks stored inside
the page tables. Not only should this allow us to avoid struct_mutex
here, but it will allow multiple users to lock independent ranges for
concurrent allocations, and operate independently. This is vital for
pushing the GTT manipulation into a background thread where dependency
on struct_mutex is verboten, and for allowing other callers to avoid
struct_mutex altogether.
v2: Restore lost GEM_BUG_ON for removing too many PTE from
gen6_ppgtt_clear_range.
Chris Wilson [Tue, 4 Jun 2019 12:00:21 +0000 (13:00 +0100)]
drm/i915: Use unchecked writes for setting up the fences
As the fence registers are not part of the engine powerwells, we do not
need to fiddle with forcewake in order to update a fence. Avoid using
the heavyweight debug checking normal mmio writes as the checking
dominates the selftest runtime and is superfluous!
In the process, retire the I915_WRITE() implicit macro with the new
intel_uncore_write interface.
Ville Syrjälä [Fri, 17 May 2019 19:31:32 +0000 (22:31 +0300)]
drm/i915: Fix plane state dumps
Stop dumping plane->state for planes. That is the old state most of the
time and dumping stale information only serves to confuse people.
Instead dump the new state just for the planes included in the
operation. For now we'll include only the planes for the modeset/fastset
pipes in the dumps. But probably we want to dump them all eventually,
just not quite sure how to present that information nicely to the user.
And while at it let's dump a few more interesting bits from the state.
Ville Syrjälä [Fri, 17 May 2019 19:31:31 +0000 (22:31 +0300)]
drm/i915: Make state dumpers take a const state
Constify a bunch of the arguments of various state dumping
functions. Makes it clear they don't mutate the states.
And fix up some indent fails while at it.
Ville Syrjälä [Fri, 17 May 2019 19:31:30 +0000 (22:31 +0300)]
drm/i915: Dump failed crtc states during atomic check
Currently we're only dumping the failed crtc state if
intel_modeset_pipe_config() fails. Let's do the state
dump if anything else fails afterwards. The downside
is that we lose the immediate knowledge which crtc caused
the failure (unless a lower level function indicates it
with an additional debug print) but having the full state
dumped seems like something that could be beneficial.
Ville Syrjälä [Fri, 17 May 2019 19:31:29 +0000 (22:31 +0300)]
drm/i915: Include crtc_state.active in crtc state dumps
Currently we're not dumping out whether the crtc is actually
active or in dpms off state. Let's include that in the dumps.
And while at it compress out a few lines from the state dump.
Ville Syrjälä [Fri, 17 May 2019 19:31:28 +0000 (22:31 +0300)]
drm/i915: Move state dump to the end of atomic_check()
Currently we're dumping the crtc states before they have
been fully calculated. Move the dumping to the end of
.atomic_check() so we get a fully up to date dump.
Let's also do the dump for fully disabled pipes, but we'll
limit that to just saying that the pipe is disabled since
the rest of the state is going to be nonsense in that case.
v2:
- Moved symlink remove to unregister (Ville)
- Clarified commit message (Jani)
- Changed WARN to DRM_ERROR (Jani)
- Minor codestyle changes proposed by Jani
v3: added blank line
Cc: Arkadiusz Hiler <arkadiusz.hiler@intel.com> Cc: Imre Deak <imre.deak@intel.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Jani Nikula <jani.nikula@intel.com> Signed-off-by: Oleg Vasilev <oleg.vasilev@intel.com> Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20190520150642.3477-1-oleg.vasilev@intel.com
drm/i915: extract intel_display_power.h/c from intel_runtime_pm.h/c
Keep all the device-level PM management in intel_runtime_pm.h/c and move
all the display specific bits into their own file. Also add the new
header to Makefile.header-test.
Apart from the giant code move, the only difference is with the
intel_runtime_<get/put>_raw() functions, which are now exposed in the
header. The _put() version is also not conditionally compiled anymore
since it is ok to always pass the wakeref taken from the _get() to
__intel_runtime_pm_put (it is -1 if tracking is disabled).
Suggested-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Imre Deak <imre.deak@intel.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Jani Nikula <jani.nikula@intel.com> 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/20190531222409.9177-2-daniele.ceraolospurio@intel.com
Imre Deak [Fri, 31 May 2019 08:26:26 +0000 (11:26 +0300)]
drm/i915/icl: Ensure port A combo PHY HW state is correct
Make sure the HW state of the port A combo PHY is correct wrt. the
IREFGEN setting. This will force a reprogramming during init or a WARN
during uninit if the setting is incorrect.
On my ICL RVP I haven't seen this check failing and leading to a forced
reinit/WARN, but let's add it still for consistency.
Chris Wilson [Thu, 30 May 2019 20:35:00 +0000 (21:35 +0100)]
drm/i915: Report all objects with allocated pages to the shrinker
Currently, we try to report to the shrinker the precise number of
objects (pages) that are available to be reaped at this moment. This
requires searching all objects with allocated pages to see if they
fulfill the search criteria, and this count is performed quite
frequently. (The shrinker tries to free ~128 pages on each invocation,
before which we count all the objects; counting takes longer than
unbinding the objects!) If we take the pragmatic view that with
sufficient desire, all objects are eventually reapable (they become
inactive, or no longer used as framebuffer etc), we can simply return
the count of pinned pages maintained during get_pages/put_pages rather
than walk the lists every time.
The downside is that we may (slightly) over-report the number of
objects/pages we could shrink and so penalize ourselves by shrinking
more than required. This is mitigated by keeping the order in which we
shrink objects such that we avoid penalizing active and frequently used
objects, and if memory is so tight that we need to free them we would
need to anyway.
v2: Only expose shrinkable objects to the shrinker; a small reduction in
not considering stolen and foreign objects.
v3: Restore the tracking from a "backup" copy from before the gem/ split
Chris Wilson [Thu, 30 May 2019 20:34:59 +0000 (21:34 +0100)]
drm/i915: Track the purgeable objects on a separate eviction list
Currently the purgeable objects, I915_MADV_DONTNEED, are mixed in the
normal bound/unbound lists. Every shrinker pass starts with an attempt
to purge from this set of unneeded objects, which entails us doing a
walk over both lists looking for any candidates. If there are none, and
since we are shrinking we can reasonably assume that the lists are
full!, this becomes a very slow futile walk.
If we separate out the purgeable objects into own list, this search then
becomes its own phase that is preferentially handled during shrinking.
Instead the cost becomes that we then need to filter the purgeable list
if we want to distinguish between bound and unbound objects.
Jani Nikula [Mon, 6 May 2019 13:48:01 +0000 (16:48 +0300)]
drm/i915: add force_probe module parameter to replace alpha_support
The i915.alpha_support module parameter has caused some confusion along
the way. Add new i915.force_probe parameter to specify PCI IDs of
devices to probe, when the devices are recognized but not automatically
probed by the driver. The name is intended to reflect what the parameter
effectively does, avoiding any overloaded semantics of "alpha" and
"support".
The parameter supports "" to disable, "<pci-id>,[<pci-id>,...]" to
enable force probe for one or more devices, and "*" to enable force
probe for all known devices.
Also add new CONFIG_DRM_I915_FORCE_PROBE config option to replace the
DRM_I915_ALPHA_SUPPORT option. This defaults to "*" if
DRM_I915_ALPHA_SUPPORT=y.
Instead of replacing i915.alpha_support immediately, let the two coexist
for a while, with a deprecation message, for a transition period.
Colin Ian King [Fri, 31 May 2019 10:32:01 +0000 (11:32 +0100)]
drm/i915: fix use of uninitialized pointer vaddr
The assignment of err is using the incorrect pointer vaddr that has
not been initialized. Fix this by using the correct pointer obj instead.
Addresses-Coverity: ("Uninitialized pointer read") Fixes: 6501aa4e3a45 ("drm/i915: add in-kernel blitter client") Signed-off-by: Colin Ian King <colin.king@canonical.com> 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/20190531103201.10124-1-colin.king@canonical.com
Chris Wilson [Thu, 30 May 2019 12:13:11 +0000 (13:13 +0100)]
drm/i915: Make default value for i915.mmio_debug a compile time option
The normal behaviour is to periodically check for a mmio access error,
and once detected enable mmio access checking. However this is useless
if the error only occurs once during module load, and so we may miss
such errors in CI. To allow ourselves to catch them, allow CI to opt into
always enabling mmio debugging.
drm/i915: Split off pci_driver.remove() tail to drm_driver.release()
In order to support driver hot unbind, some cleanup operations, now
performed on PCI driver remove, must be called later, after all device
file descriptors are closed.
Split out those operations from the tail of pci_driver.remove()
callback and put them into drm_driver.release() which is called as soon
as all references to the driver are put. As a result, those cleanups
will be now run on last drm_dev_put(), either still called from
pci_driver.remove() if all device file descriptors are already closed,
or on last drm_release() file operation.
Matthew Auld [Wed, 29 May 2019 12:31:08 +0000 (13:31 +0100)]
drm/i915: add in-kernel blitter client
The plan is to use the blitter engine for async object clearing when
using local memory, but before we can move the worker to get_pages() we
have to first tame some more of our struct_mutex usage. With this in
mind we should be able to upstream the object clearing as some
selftests, which should serve as a guinea pig for the ongoing locking
rework and upcoming async get_pages() framework.
Matthew Auld [Wed, 29 May 2019 12:31:07 +0000 (13:31 +0100)]
drm/i915/gtt: grab wakeref in gen6_alloc_va_range
Some steps in gen6_alloc_va_range require the HW to be awake, so ideally
we should be grabbing the wakeref ourselves and not relying on the
caller already holding it for us.
Chris Wilson [Wed, 29 May 2019 09:34:07 +0000 (10:34 +0100)]
drm/i915/gtt: Avoid overflowing the WC stash
An interesting issue cropped with making the pagetables be allocated and
freed concurrently (i.e. removing their grandeous struct_mutex guard)
was that we would overflow the page stash. This happens when we have
multiple allocators grabbing WC pages such that we fill the vm's local
page stash and then when we free another page, the page stash is already
full and we overflow.
The fix is quite simple: to check for a full page stash before adding
another. This results in us keeping a vm local page stash around for
much longer, which is both a blessing and a curse.
Jani Nikula [Wed, 29 May 2019 08:21:50 +0000 (11:21 +0300)]
Revert "drm/i915: Expand subslice mask"
This reverts commit 1ac159e23c2c ("drm/i915: Expand subslice mask"),
which kills ICL due to GEM_BUG_ON() sanity checks before CI even gets a
chance to do anything.
The commit exposes an issue in commit 1e40d4aea57b ("drm/i915/cnl:
Implement WaProgramMgsrForCorrectSliceSpecificMmioReads"), which will
also need to be addressed.
There's a proposed fix [1], but considering the seeming uncertainty with
the fix as well as the size of the regressing commit (in this context,
the one that actually brings down ICL), this warrants a revert to get
ICL working, and gives us time to get all of this right without
rushing. Even if this means shooting the messenger.
Chris Wilson [Tue, 28 May 2019 15:40:53 +0000 (16:40 +0100)]
drm/i915: Avoid refcount_inc on known zero count
In intel_wakeref_auto, we use refcount_inc_not_zero to detect the first
use and initialise the timer. On doing so, we have to avoid using
refcount_inc on that zero count as the debug code flags that as an
error:
refcount_t: increment on 0; use-after-free.
Rearrange the code so that if we know the count is 0 and we are
initialising, we explicitly set it to 1.
Dan Carpenter [Wed, 29 May 2019 11:03:55 +0000 (14:03 +0300)]
drm/i915: selftest_lrc: Check the correct variable
We should check "request[n]" instead of just "request".
Fixes: 78e41ddd2198 ("drm/i915: Apply an execution_mask to the virtual_engine") Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> 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/20190529110355.GA19119@mwanda
Chris Wilson [Tue, 28 May 2019 07:53:54 +0000 (08:53 +0100)]
drm/i915: Take a runtime pm wakeref for atomic commits
Before we start prepping the system for an atomic modeset, wake the
device up. We then keep track of this wakeref until we complete the
atomic commit, so we hold keep the device awake for all potential HW
access, and do not allow the device to sleep with a pending modeset.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110771 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20190528075354.22341-1-chris@chris-wilson.co.uk
Stuart Summers [Fri, 24 May 2019 15:40:22 +0000 (08:40 -0700)]
drm/i915: Expand subslice mask
Currently, the subslice_mask runtime parameter is stored as an
array of subslices per slice. Expand the subslice mask array to
better match what is presented to userspace through the
I915_QUERY_TOPOLOGY_INFO ioctl. The index into this array is
then calculated:
slice * subslice stride + subslice index / 8
v2: fix spacing in set_sseu_info args
use set_sseu_info to initialize sseu data when building
device status in debugfs
rename variables in intel_engine_types.h to avoid checkpatch
warnings
v3: update headers in intel_sseu.h
v4: add const to some sseu_dev_info variables
use sseu->eu_stride for EU stride calculations
v5: address review comments from Tvrtko and Daniele
v6: remove extra space in intel_sseu_get_subslices
return the correct subslice enable in for_each_instdone
add GEM_BUG_ON to ensure user doesn't pass invalid ss_mask size
use printk formatted string for subslice mask
v7: remove string.h header and rebase
Stuart Summers [Fri, 24 May 2019 15:40:21 +0000 (08:40 -0700)]
drm/i915: Refactor sseu helper functions
Move functions to intel_sseu.h and remove inline qualifier.
Additionally, ensure these are all prefixed with intel_sseu_*
to match the convention of other functions in i915.
v2: fix spacing from checkpatch warning
v3: squash helper function changes into a single patch
break 80 character line to fix checkpatch warning
move get/set_eus helpers to intel_device_info.c
v4: Remove intel_ prefix from static functions in
intel_device_info.c and correctly copy changes
to stride calculation in those functions.
Stuart Summers [Fri, 24 May 2019 15:40:19 +0000 (08:40 -0700)]
drm/i915: Add macro for SSEU stride calculation
Subslice stride and EU stride are calculated multiple times in
i915_query. Move this calculation to a macro to reduce code duplication.
v2: update headers in intel_sseu.h
v3: use GEN_SSEU_STRIDE for stride calculations in intel_sseu.h
apply s/bits/max_entries/ to GEN_SSEU_STRIDE parameter
Ville Syrjälä [Mon, 13 May 2019 13:39:03 +0000 (16:39 +0300)]
drm/i915: Update pipe gamma enable bits when C8 planes are getting enabled/disabled
When the first C8 plane gets enabled, or the last one gets disabled we
may need to enable/disable the pipe gamma for the other active planes.
Check for that and run through the normal intel_color_check() path.
Chris Wilson [Tue, 28 May 2019 09:29:56 +0000 (10:29 +0100)]
drm/i915: Drop the deferred active reference
An old optimisation to reduce the number of atomics per batch sadly
relies on struct_mutex for coordination. In order to remove struct_mutex
from serialising object/context closing, always taking and releasing an
active reference on first use / last use greatly simplifies the locking.