Enabling rps (turbo setup) was put in a work queue because it may
take quite awhile. This change flushes the work queue to initialize
rps values before use by sysfs or debugfs. Specifically,
rps.delayed_resume_work is flushed before using rps.hw_max,
rps.max_delay, rps.min_delay, or rps.cur_delay.
This change fixes a problem in sysfs where show functions using
uninitialized values show incorrect values and store functions
using uninitialized values in range checks incorrectly fail to
store valid input values. This change also addresses similar use
before initialized problems in debugfs.
Signed-off-by: Tom O'Rourke <Tom.O'Rourke@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We lost the ability to capture the first error for a stuck ring in the
recent hangcheck robustification. Whilst both error states are
interesting (why does the GPU not recover is also essential to debug),
our primary goal is to fix the initial hang and so we need to capture
the first error state upon taking hangcheck action.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Let's try to avoid these confusing negated booleans.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Regression introduced by:
commit 311a20949f
Author: Paulo Zanoni <paulo.r.zanoni@intel.com>
drm/i915: don't init DP or HDMI when not supported by DDI port
Since the commit above it is possible to have a DDI encoder that has
the HDMI connector but not the DP connector (in case the port doesn't
support DP). In this case, we must properly free the DP connector.
We just leak this once, so it's not a big deal.
Reported by kmemleak.
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The DP spec allows this, and requires it when full link training is
started with non-minimum voltage swing and/or non-zero pre-emphasis.
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
For future platforms we'll need to initialize our MMIO function pointers
even earlier. Specifically, we'll need to be able to have register
reads/writes at GTT initialization (in i915_gem_gtt_init). Similarly,
these platforms also have MMIO differences based on the PCH id, so
while moving stuff around, also move the PCH initialization.
CC: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
[danvet: Mention the function where we need register access.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
At the end of haswell_crtc_enable we have an intel_wait_for_vblank
with a big comment, and the message suggests it's a workaround for
something we don't really understand. So I removed that wait and
started getting HW state readout error messages saying that the IPS
state is not what we expected.
I investigated and concluded that after you write IPS_ENABLE to
IPS_CTL, the bit will only actually become 1 on the next vblank. So
add code to wait for the IPS_ENABLE bit. We don't really need this
wait right now due to the wait I already mentioned, but at least this
one has a reason to be there, while the other one is just to
workaround some problem: we may remove it in the future.
The wait also acts as a POSTING_READ which we missed.
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Something already got misplaced (although it's from a patch from
before Paulo's cleanup). Move it to the right spot.
v2: Remove the line to keep a neat block, requested by Paulo.
Reported-by: Paulo Zanoni <przanoni@gmail.com>
Cc: Paulo Zanoni <przanoni@gmail.com>
Reviewed-by: Paulo Zanoni <przanoni@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Now that MMIO has been split up into gen specific functions it is
obvious when HAS_FPGA_DBG_UNCLAIMED, HAS_FORCE_WAKE are needed. As such,
we can remove this extraneous condition.
As a result of this, as well as previously existing function pointers
for forcewake, we no longer need the has_force_wake member in the device
specific data structure.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Similar to the previous patch which implemented GEN specific reads; this
patch does the same for writes. Writes have a bit of adding complexity
due to the FPGA_DBG feature of HSW plus:
gen[2-4]: nothing special
gen5: ILK dummy write
gen[6-7]: forcewake shenanigans
gen[HSW}: forcewake shenanigans + FPGA_DBG
I was a bit torn about whether or not to combine 6-HSW as one function,
since the FPGA_DBG is cleanly separated, and it wouldn't make the 6-7
MMIO too messy. In the end, I chose the clearest possible solution which
splits out HSW.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Extracting the MMIO read functionality makes per gen handling a bit
simpler, and the overall function a lot easier to read. The increasing
complexity of reads doesn't get too bad as the generation number
increases:
gen[2-4]: Nothing special
gen5: ILK dummy write workaround
gen6+: forcewake shenanigans
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Just to make the churn and code duplication in upcoming patches a bit
less, turn code which is common to all GEN MMIO functions into a macro.
v2: Fix typo in subject
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
In preparation for having per GEN MMIO functions, create, and start
using MMIO functions in our uncore data structure. This simply makes the
transition easier by allowing us to just plug in the per GEN stuff
later.
For simplicity, I moved the intel_uncore_init() function down since
those rely on static functions defined lower in the file. This is most
of the churn in this patch.
I made one unrelated change here by using off_t datatype for the offset
of the register to write. I like the clarity that this brings to the
code. If I did it as a separate patch, I am pretty certain it would get
bikeshedded to oblivion.
Requested-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
In order to be able to have virtual functions for the MMIO, we need to
use the raw access function. To keep things simple, just move this to
our early_sanitize code in uncore.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
For upcoming patches which will have GEN specific MMIO functions, we'll
need to initialize the uncore data structure earlier than we do today.
If we do not do this, the following will be problematic:
intel_uncore_sanitize
intel_disable_gt_powersave
gen6_disable_rps
I915_WRITE(GEN6_RC_CONTROL, 0); <--- MMIO
intel_uncore_init // initializes MMIO
By initializing the function pointers first, we should be safe.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
At least on my i830M here it reliably results in hard system hangs
nowadays. This is much worse than falling back to software rendering,
so I think we should simply rip this out.
After all we don't have any gpu reset for gen3 either, and there are a
lot more of those still around.
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
In truly crazy circumstances shmem might give us the wrong type of
page. So be a bit paranoid and double check this.
Reviewer: Damien Lespiau <damien.lespiau@intel.com>
Cc: Rob Clark <robdclark@gmail.com>
References: http://lkml.org/lkml/2011/7/11/238
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The PIPEA quirk is specifically for the issue with the PIPEB PLL on
830gm being slaved to the PIPEA PLL, and so to use PIPEB requires PIPEA
running. i845 doesn't even have the second PLL or pipe, and enabling
the quirk results in a blank DVO LVDS.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The policy's max frequency is not equal to the CPU's max frequency. The
ring frequency is derived from the CPU frequency, and not the policy
frequency.
One example of how this may differ through sysfs. If the sysfs max
frequency is modified, that will be used for the max ring frequency
calculation.
(/sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq). As far as I
know, no current governor uses anything but max as the default, but in
theory, they could. Similarly distributions might set policy as part of
their init process.
It's ideal to use the real frequency because when we're currently scaled
up on the GPU. In this case we likely want to race to idle, and using a
less than max ring frequency is non-optimal for this situation.
AFAIK, this patch should have no impact on a majority of people.
This behavior hasn't been changed since it was first introduced:
commit 23b2f8bb92
Author: Jesse Barnes <jbarnes@virtuousgeek.org>
Date: Tue Jun 28 13:04:16 2011 -0700
drm/i915: load a ring frequency scaling table v3
CC: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Flush the primary plane changes when enabling/disabling the primary
plane in response to sprite visibility.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The new names make it clearer which plane we're talking about.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
[danvet: Resolve small conflict with the haswell_crtc_disable_planes
extraction.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The intel_flush_primary_plane name actually tells us which plane
we're talking about.
Also reorganize the internals a bit and add a missing POSTING_READ()
to make sure the hardware has seen the changes by the time we
return from the function.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
IPS should be OK as long as one plane is enabled on the pipe, but
it does seem to cause problems when going between primary only and
sprite only.
This needs more investigations, but for now just disable IPS whenever
the primary plane is disabled.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Disable fbc before disabling the primary plane, and enable fbc after
the primary plane has been enabled again.
Also use intel_disable_fbc() to disable FBC to avoid the pointless
overhead of intel_update_fbc(), and especially avoid having to clean
up and set up the stolen mem compressed buffer again.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
If the setplane operation fails, we shouldn't save the user's requested
plane coordinates. Since we adjust the coordinates during the clipping
process, make a copy of the originals, and once the operation has
succeeded save them for later reuse when the plane gets re-enabled.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Move the variable initialization to where the variables are declared,
and kill a pointless to_intel_crtc() cast when we already have the
casted pointer.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Let's not use goto when a simple if suffices. This is not error handling
code or anything, so the goto looks out of place.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We used to call the entire intel specific update_plane hook while
holding struct_mutex. Actually we only need to hold struct_mutex while
pinning/unpinning the obj. The plane state itself is protected by the
kms locks, and as the object is pinned we can dig out the offset and
tiling information from it without fearing that it would change
underneath us.
So now we don't need to drop and reacquire the lock around the
wait_for_vblank. Also we will need another wait_for_vblank in the IVB
specific update_plane hook, and this way we don't need to worry about
struct_mutex there either.
Also move the intel_plane->obj=NULL assignment outside strut_mutex in
disable_plane to make it clear that it's not protected by struct_mutex.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We allow cursors to be set up when the pipe is disabled. Do the same for
sprites as well.
We need to be somewhat careful with the primary disable logic as we
don't want to accidentally enable the primary plane on a disabled pipe.
v2: Skip primary enable/disable and plane registers
writes on disabled pipe
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
If the primary gets marked as disabled while the pipe is off for
instance, we should still re-enable it when the pipe is turned on,
unless the sprite covers it fully also in that configuration.
Unfortunately we do the plane visibility checks only in the sprite code,
which is executed after the primary enabling when turning the pipe off.
Ideally we should compute the plane visibility before touching the
hardware at all, but for now just set the primary_disabld flag
in intel_{enable,disable}_plane.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
If channel equalization succeeds, there's no indication something went
wrong in clock recovery (unless debug is enabled). We should shout about
the failures and fix them instead of hiding them under the carpet.
This has allowed bugs like [1] stay dormant for a long time.
[1] https://bugs.freedesktop.org/show_bug.cgi?id=70117
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The VGACNTRL register contains a bunch of other stuff besides
the VGA_DISP_DISABLE bit. When we write the register we always set those
other bits to zero, so normally the current check would work.
However on HSW disabling and re-enabling the power well will reset the
VGACNTRL register to its default value, which has several of the other
bits set as well.
So only look at the VGA_DISP_DISABLE bit when checking whether the VGA
plane needs re-disabling.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Everyone else uses intel_PLL_is_valid(), so make VLV use it as well.
We don't have any special p and m limits on VLV, so skip those tests,
and we also need to skip the m1<=m2 test line PNV.
Reorganize the function a bit to move the n check alongside the rest of
the test for the non-derived dividers, and check the derived values
afterwards.
Note that this changes vlv_find_best_dpll() in two ways:
- The .vco comparison is now >max instead of >=max, and since we round
down when calculating that stuff, we may now allow frequencies slightly
above the max as we do on other platforms. The previous method
disallowed exactly max and anything above it.
- We now check the .dot frequency against the data rate limits, which we
didn't do before.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
If vlv_find_best_dpll() couldn't find suitable PLL settings,
just say so instead of lying to caller.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
After aligning the p1 divider limits, and removing the unused p and m
limits, intel_limits_vlv_dac and intel_limits_vlv_hdmi are identical.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We don't use .dot_limit for anything on VLV, so don't populate it.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We never check the p and m limits (which according to comments are
based on someone's guesswork), so just remove them.
VLV2_DPLL_mphy_hsdpll_frequency_table_ww6_rev1p1.xlsm has no p and m
limits listed.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
VLV2_DPLL_mphy_hsdpll_frequency_table_ww6_rev1p1.xlsm tells us that the
minimum p2 divider is 2. Use that limit on the code.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
According to VLV2_DPLL_mphy_hsdpll_frequency_table_ww6_rev1p1.xlsm p1
can be 2-3 always.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
For some reason there's a sort of off by one issue with the p1 divider.
The actual p1 limits according to
VLV2_DPLL_mphy_hsdpll_frequency_table_ww6_rev1p1.xlsm is 2-3, so we should
just say that instead of saying 1-3 and avoiding the 1 via the choice of
comparison operator.
I don't know why we're using different p1 limits for intel_limits_vlv_dac
and intel_limits_vlv_hdmi, but let's preserve that for now.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We limit the maximum n divider value in order to make sure the PLL's
reference inout is at least 19.2 MHz. I assume that is done to satisfy
some hardware requirement.
However we never check whether that calculated limit is below the
maximum supoorted N divider value (7). In practice that is always true
since we only support 100 MHz reference clock, but making the code
safe against higher reference clocks seems like a reasoanble thing to
do.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The p2 divider on VLV needs to be even when it's > 10. The current code
to make that happen is rather weird. Just make the step size adjustement
in the for loop decrement step.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Rewrite vlv_find_best_dpll() to use intel_clock_t rather than
an army of local variables.
Also extract the code to calculate the derived values into
vlv_clock().
v2: Split up the earlier fixes, extract vlv_clock()
v3: Initialize best_clock
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We do 'bestppm - 10' in vlv_find_best_dpll() but never check whether
that might underflow. Add such a check.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Use div_u64() to make the ppm calculation in vlv_find_best_dpll() safe
against interger overflows.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The interface uses an unsigned long, and we can use the unsigned counter
throughout our code, so do so. In the process, we notice one instance
where the shrink count is based on a heuristic rather than the result,
and another where we ask for too many pages to be purged.
v2: nr_to_scan needs to be promoted to a long as well, so just use
sc->nr_to_scan directly.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Tom O'Rourke