Before this patch, we used the intel_display_power_{get,put} functions
to make sure the PW1 and Misc I/O power wells were enabled all the
time while LCPLL was enabled. We called a get() at
intel_ddi_pll_init() when we discovered that LCPLL was enabled, then
we would call put/get at skl_{un,}init_cdclk().
The problem is that skl_uninit_cdclk() is indirectly called by
intel_runtime_suspend(). So it will only release its power well
_after_ we already decided to runtime suspend. But since we only
decide to runtime suspend after all power wells and refcounts are
released, that basically means we will never decide to runtime
suspend.
So what this patch does to fix that problem is move the PW1 + Misc I/O
power well handling out of the runtime PM mechanism: instead of
calling intel_display_power_{get_put} - functions that touch the
refcount -, we'll call the low level intel_power_well_{en,dis}able,
which don't change the refcount. This way, it is now possible for the
refcount to actually reach zero, and we'll now start runtime
suspending/resuming.
v2 (from Paulo):
- Write a commit message since the original patch left it empty.
- Rebase after the intel_power_well_{en,dis}able rename.
- Use lookup_power_well() instead of hardcoded indexes.
Testcase: igt/pm_rpm/rte (and every other rpm test)
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Patrik Jakobsson <patrik.jakobsson@linux.intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=92211
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=92605
Signed-off-by: Imre Deak <imre.deak@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1446657859-9598-4-git-send-email-imre.deak@intel.com
When accessing through the GTT from one CPU whilst concurrently updating
the GGTT PTEs in another thread, the hardware likes to return random
data. As we have strong serialisation prevent us from modifying the PTE
of an active GTT mmapping, we have to conclude that it whilst modifying
other PTE's that error occurs. (I have not looked for any pattern such
as modifying PTE within the same page or cacheline as active PTE -
though checking whether revoking neighbouring objects should be enough
to test that theory.) The corruption also seems restricted to Braswell
and disappears with maxcpus=0. This patch stops all access through the
GTT by other CPUs when we update any PTE by stopping the machine around
the GGTT update.
Note that splitting up the 64 bit write into two 32 bit writes was
tried and found to fail too.
Testcase: igt/gem_concurrent_blit
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=89079
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
[danvet: Add note about 2x 32bits failing too.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Automated test data that is updated when a test is requested is not cleared
till next automated test request is recevied which can cause various
problems. This patch fixes this by clearing this during the next
short pulse and on hot unplug.
For example, when TEST_LINK_TRAINING is requested it is updated
to appropriate variable inside intel_dp_handle_test_request
but is also cleared only inside the same function. if the next
short pulse does not have the AUTOMATED_TEST_REQUEST bits set
the variable will not be cleared resulting in carrying incorrect
test status in local variables.
v2: Added comments and moved nack and defer variables before set_edid
(Sonika)
Signed-off-by: Sivakumar Thulasimani <sivakumar.thulasimani@intel.com>
Signed-off-by: Shubhangi Shrivastava <shubhangi.shrivastava@intel.com>
Reviewed-by: Sonika Jindal <sonika.jindal@intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
If ddb allocation for planes in current CRTC is changed, that doesn't
lead to ddb allocation change for other CRTCs, because our DDB allocation
is not dynamic according to plane parameters, ddb is allocated according
to number of CRTC enabled, & divided equally among CTRC's.
In current condition check during Watermark calculation, if number of
plane/ddb allocation changes for current CRTC, Watermark for other pipes
are recalculated. But there is no change in DDB allocation of other pipe
so watermark is also not changed, This leads to warning messages.
WARN_ON(!wm_changed)
This patch corrects this and check if DDB allocation for pipes is changed,
then only recalculate watermarks.
v2 (by Matt): Rebased to latest -nightly and fixed a typo
Signed-off-by: Kumar, Mahesh <mahesh1.kumar@intel.com>
Reviewed-by(v1): Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Currently we determine the location of the AUX registers in a confusing
way. First we assume the PCH registers are used always, but then we
override it for everything but HSW/BDW to use DP+0x10. Very confusing.
Let's just make it straightforward and simply add a few functions to
pick the right AUX_CTL based on the DP port.
To deal with VLV/CHV we'll include the display_mmio_offset into the
AUX register defines.
v2: Reorder patches (Chris)
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> (v1)
Link: http://patchwork.freedesktop.org/patch/msgid/1447266856-30249-5-git-send-email-ville.syrjala@linux.intel.com
Use kasprintf() to generate the "DPDDC-<port>" name for the aux helper.
To deal with errors properly make intel_dp_aux_init() return something,
and adjust the caller to match. It seems we were also missing a
intel_dp_mst_encoder_cleanup() call on edp (non-port A) init failures,
so add that too.
The whole error/cleanup ordering doesn't feel entirely sane to me, but
I'll leave that part alone for now.
v2: Use kasprintf() instead of a table, reorder patches (Chis)
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1447266856-30249-3-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
This removes two anti-patterns:
- Locking shouldn't be used to synchronize with async work (of any
form, whether callbacks, workers or other threads). This is what the
mutex_lock/unlock seems to have been for in intel_csr_load_program.
Instead ordering should be ensured with the generic
wait_for_completion()/complete(). Or more specific functions
provided by the core kernel like e.g.
flush_work()/cancel_work_sync() in the case of synchronizing with a
work item.
- Don't invent own completion like the following code did with the
(already removed) wait_for(csr_load_status_get()) pattern - it's
really hard to get these right when you want them to be _really_
correct (and be fast) in all cases. Furthermore it's easier to read
code using the well-known primitives than new ones using
non-standard names.
Before enabling/disabling DC6 check if the firmware is loaded
successfully. This is guaranteed during runtime s/r, since otherwise we
don't enable RPM, but not during system s/r.
Note that it's still unclear whether we need to enable/disable DC6
during system s/r, until that's clarified, keep the current behavior and
enable/disable DC6.
Also after this patch there is a race during system s/r where the
firmware may not be loaded yet, that's addressed in an upcoming patch.
v2-v3:
- unchanged
v4:
- rebased on latest drm-intel-nightly
Cc: Damien Lespiau <damien.lespiau@intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Cc: Sunil Kamath <sunil.kamath@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Signed-off-by: Animesh Manna <animesh.manna@intel.com>
[imre: added code and note about checking if the firmware loaded ok,
before enabling/disabling it]
Reviewed-by: Animesh Manna <animesh.manna@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Tested-by: Daniel Stone <daniels@collabora.com> # SKL
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1447341037-2623-1-git-send-email-imre.deak@intel.com
Reading the driver load/unload code leaves one confused as there's
an async_schedule() in the load, but not async_synchronize_full()
in sight. In fact it's hidden inside intel_fbdev.c. So let's move the
async_schedule() into intel_fbdev.c as well so that it's next to the
async_synchronize_full(), which should make the relationship easier
to see.
Plus this way we won't schedule a nop function call when fbdev is
disabled. And we were passing a pointer to a static inline
function to async_schedule(), which seems rather dubious to me.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1446815313-9490-4-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
We set up fbdev last during load, so doing the fbdev cleanup should be
first.
We weren't supposed to drop the init power during driver unload, but since
the fbdev teardown happened after intel_power_domains_fini() that could
have happened due in one of two ways. First it could have happened
during the modeset caused by normal fbdev cleanup. But in addition it
could have happened already via the intel_fbdev_initial_config() since
that is executed asynhronously, and the async_synchronize_full() was
done during fbdev cleanup, after intel_power_domains_fini(). All of
that got eliminated by
commit 292b990e86 ("drm/i915: Update power domains on readout.")
since we now drop the init power synchronously during driver load.
So there is no real bug wrt. the init power anymore, but still it seems
better to do the fbdev cleanup first, before we've potentially cleaned
up something else important.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1446815313-9490-3-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
intel_runtime_pm_disable() takes an extra rpm reference which combined
with the one we leak from intel_display_set_init_power() leaves the
usage count at <original>+1 after the driver has been unloaded.
The original ref is dropped explicitly in intel_runtime_pm_enable().
So the next time we load the driver we can no longer do runtime PM ever.
This used to work, but
commit 292b990e86 ("drm/i915: Update power domains on readout.")
broke things by not dropping the init power domain during fbdev
teardown. Based on the comment in intel_power_domains_fini(), the
way it used to to work wasn't intentional. As in we weren't supposed
to drop the init power during driver unload. And since we no longer
do, we now leak an extra rpm reference.
So fix things by throwing intel_runtime_pm_disable() to the bin, so
that the only leaked reference comes from the init power domain.
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Daniel Stone <daniels@collabora.com>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Fixes: 292b990e86 ("drm/i915: Update power domains on readout.")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1446815313-9490-2-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Use intel_dp->DP in the eDP PLL setup, instead of doing RMWs.
To do this we need to move DP_AUDIO_OUTPUT_ENABLE setup to happen later,
so that we don't enable audio accidentally while configuring the PLL.
Note that actually we already enabled audio before the port due to
the double port register write magic required by VLV/CHV from
7b713f50d7 ("drm/i915: Fix eDP link training when switching pipes on VLV/CHV")
So that gets changed now to keep audio off as long as the port is off.
Also intel_dp_link_down() must be made to update intel_dp->DP so that we
don't re-enable the port by accident when turning off the PLL. This is
safe now that we don't call intel_dp_link_down() during link retraining.
v2: Add a note about the audio vs. port enable (Daniel)
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1447164977-32315-1-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
We get underruns on the other pipe when enabling the CPU eDP PLL and
port on ILK.
Bspec knows about the PLL issue, and recommends doing a vblank wait just
prior to enabling the PLL. That does seem to help, but unfortunately we
get another underrun when actually enabling the CPU eDP port. Bspec
doesn't mention that at all, and the same vblank wait trick doesn't
appear to be effective there.
Since I have no better clue how to deal with this, just hide the errors.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1446146763-31821-10-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Doing the IBX transcoder B workaround causes underruns on
pipe/transcoder A. Just hide them by disabling underrun reporting for
pipe A around the workaround.
It might be possible to avoid the underruns by moving the workaround
to be applied only when enabling pipe A. But I was too lazy to try it
right now, and the current method has been proven to work, so didn't
want to change it too hastily.
Note that this can re-enable underrun reporting on pipe A if was
already disabled due to a previous actual underrun. But that's OK, we
may just get a second underrun report if another real underron occurrs
on pipe A.
v2: Note that pipe A underruns can get re-enabled due to this (Jani)
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org> (v1)
Link: http://patchwork.freedesktop.org/patch/msgid/1446225802-11180-1-git-send-email-ville.syrjala@linux.intel.com
Some hardware (IVB/HSW and CPT/PPT) have a shared error interrupt for
all the relevant underrun bits, so in order to keep the error interrupt
enabled, we need to have underrun reporting enabled on all PCH
transocders. Currently we leave the underrun reporting disabled when
the pipe is off, which means we won't get any underrun interrupts
when only a subset of the pipes are active.
Fix the problem by re-enabling the underrun reporting after the pipe has
been disabled. And to avoid the spurious underruns during pipe enable,
disable the underrun reporting before embarking on the pipe enable
sequence. So this way we have the error reporting disabled while
running through the modeset sequence.
v2: Re-enable PCH FIFO underrun reporting unconditionally on pre-HSW
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org> (v1)
Link: http://patchwork.freedesktop.org/patch/msgid/1446225691-10928-1-git-send-email-ville.syrjala@linux.intel.com
