Otherwise we can encounter out of memory situations under extreme load.

v2: add documentation for the new function
Signed-off-by: NChristian König <deathsimple@vodafone.de>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

Otherwise the sa managers out of memory
handling doesn't work.
Signed-off-by: NChristian König <deathsimple@vodafone.de>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

DRM_IOCTL_MODESET_CTL must only be used for UMS drivers. Make it a no-op
for KMS drivers.
Signed-off-by: NLaurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: NMichel Dänzer <michel@daenzer.net>
Signed-off-by: NDave Airlie <airlied@gmail.com>

This merges Christian work that has been hanging around on the list.

Try to save whatever is on the rings when
we encounter an lockup.

v2: Fix spelling error. Free saved ring data if reset fails.
    Add documentation for the new functions.
v3: Some more spelling fixes
v4: It doesn't make sense to save anything if all fences
    are signaled
Signed-off-by: NChristian König <deathsimple@vodafone.de>
Reviewed-by: NMichel Dänzer <michel.daenzer@amd.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

Before emitting any indirect buffer, emit the offset of the next
valid ring content if any. This allow code that want to resume
ring to resume ring right after ib that caused GPU lockup.

v2: use scratch registers instead of storing it into memory
v3: skip over the surface sync for ni and si as well
v4: use SET_CONFIG_REG instead of PACKET0
Signed-off-by: NJerome Glisse <jglisse@redhat.com>
Signed-off-by: NChristian König <deathsimple@vodafone.de>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

Making it easier to control when it is executed.
Signed-off-by: NChristian König <deathsimple@vodafone.de>
Reviewed-by: NMichel Dänzer <michel.daenzer@amd.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

Just restore the page table instead. Addressing three
problem with this change:

1. Calling vm_manager_suspend in the suspend path is
   problematic cause it wants to wait for the VM use
   to end, which in case of a lockup never happens.

2. In case of a locked up memory controller
   unbinding the VM seems to make it even more
   unstable, creating an unrecoverable lockup
   in the end.

3. If we want to backup/restore the leftover ring
   content we must not unbind VMs in between.
Signed-off-by: NChristian König <deathsimple@vodafone.de>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

Just reinitialize the shader content on resume instead.
Signed-off-by: NChristian König <deathsimple@vodafone.de>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

The IB pool is in gart memory, so it is completely
superfluous to unpin / repin it on suspend / resume.
Signed-off-by: NChristian König <deathsimple@vodafone.de>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

It's not critical, but the current code isn't
100% correct.
Signed-off-by: NChristian König <deathsimple@vodafone.de>
Reviewed-by: NJerome Glisse <jglisse@redhat.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

For a normal suspend/resume we allready wait for
the rings to be empty, and for a suspend/reasume
in case of a lockup we REALLY don't want to wait
for anything.
Signed-off-by: NChristian König <deathsimple@vodafone.de>
Reviewed-by: NJerome Glisse <jglisse@redhat.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

Start with last signaled fence number instead
of last emitted one.
Signed-off-by: NChristian König <deathsimple@vodafone.de>
Reviewed-by: NJerome Glisse <jglisse@redhat.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

It is possible that radeon_fence_process is called
after writeback is disabled for suspend, leading
to an invalid read of register 0x0.

This fixes a problem for me where the fence value
is temporary incremented by 0x100000000 on
suspend/resume.
Signed-off-by: NChristian König <deathsimple@vodafone.de>
Reviewed-by: NJerome Glisse <jglisse@redhat.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

We don't need to pad anything if the number of dwords
written to the ring already matches the requirements.

Fixes some "writting more dword to ring than expected"
warnings.
Signed-off-by: NChristian König <deathsimple@vodafone.de>
Reviewed-by: NJerome Glisse <jglisse@redhat.com>
Reviewed-by: NMichel Dänzer <michel.daenzer@amd.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

GPU reset need to be exclusive, one happening at a time. For this
add a rw semaphore so that any path that trigger GPU activities
have to take the semaphore as a reader thus allowing concurency.

The GPU reset path take the semaphore as a writer ensuring that
no concurrent reset take place.

v2: init rw semaphore
Signed-off-by: NJerome Glisse <jglisse@redhat.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

Don't return success if scheduling the IB fails, otherwise
we end up with an oops in ttm_eu_fence_buffer_objects.
Signed-off-by: NChristian König <deathsimple@vodafone.de>
Reviewed-by: NJerome Glisse <jglisse@redhat.com>
Reviewed-by: NMichel Dänzer <michel.daenzer@amd.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

Waiting for a fence can fail for different reasons,
the most common is a deadlock.
Signed-off-by: NChristian König <deathsimple@vodafone.de>
Reviewed-by: NMichel Dänzer <michel.daenzer@amd.com>
Reviewed-by: NJerome Glisse <jglisse@redhat.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

Instead of returning the error handle it directly
and while at it fix the comments about the ring lock.
Signed-off-by: NChristian König <deathsimple@vodafone.de>
Reviewed-by: NMichel Dänzer <michel.daenzer@amd.com>
Reviewed-by: NJerome Glisse <jglisse@redhat.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

In order to support snoopable memory on non-LLC architectures (so that
we can bind vgem objects into the i915 GATT for example), we have to
avoid the prefetcher on the GPU from crossing memory domains and so
prevent allocation of a snoopable PTE immediately following an uncached
PTE. To do that, we need to extend the range allocator with support for
tracking and segregating different node colours.

This will be used by i915 to segregate memory domains within the GTT.

v2: Now with more drm_mm helpers and less driver interference.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Cc: Dave Airlie <airlied@redhat.com
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Ben Skeggs <bskeggs@redhat.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: NDave Airlie <airlied@gmail.com>

If drm can't find proc it should fail more gracefully, than just
oopsing, this tests drm_class is NULL, and sets it to NULL in the
fail paths.
Reported-by: NFengguang Wu <fengguang.wu@intel.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Merge tag 'drm-intel-next-2012-07-06' of git://people.freedesktop.org/~danvet/drm-intel into drm-next

Daniel writes:
New pull for -next. Highlights:
- rc6/turbo support for hsw (Eugeni)
- improve corner-case of the reset handling code - gpu reset handling
  should be rock-solid now
- support for fb offset > 4096 pixels on gen4+ (yeah, you need some fairly
  big screens to hit that)
- the "Flush Me Harder" patch to fix the gen6+ fallout from disabling the
  flushing_list
- no more /dev/agpgart on gen6+!
- HAS_PCH_xxx improvements from Paulo
- a few minor bits&pieces all over, most of it in thew hsw code

* tag 'drm-intel-next-2012-07-06' of git://people.freedesktop.org/~danvet/drm-intel: (40 commits)
  drm/i915: program FDI_RX TP and FDI delays
  drm/i915: introduce for_each_encoder_on_crtc
  drm/i915: adjust framebuffer base address on gen4+
  drm/i915: introduce crtc->dspaddr_offset
  drm/i915: Reject page flips with changed format/offset/pitch
  drm/i915: Zero initialize mode_cmd
  drm/i915: don't return a spurious -EIO from intel_ring_begin
  drm/i915: properly SIGBUS on I/O errors
  drm/i915: don't hang userspace when the gpu reset is stuck
  drm/i915: non-interruptible sleeps can't handle -EAGAIN
  drm/i915: don't trylock in the gpu reset code
  drm/i915: fix PIPE_DDI_PORT_MASK
  drm/i915: prevent bogus intel_update_fbc notifications
  drm/i915: re-initialize DDI buffer translations after resume
  drm/i915: don't ironlake_init_pch_refclk() on LPT
  drm/i915: get rid of dev_priv->info->has_pch_split
  drm/i915: add PCH_NONE to enum intel_pch
  drm/i915: prefer wide & slow to fast & narrow in DP configs
  drm/i915: fix up ilk rc6 disabling confusion
  drm/i915: move force wake support into intel_pm
  ...

This is required for a stable FDI connection.

v2: fix and simplify the FDI_RX_MISC bits as noticed by Paulo Zanoni.

CC: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Reviewed-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

We already have this pattern at quite a few places, and moving part of
the modeset helper stuff into the driver will add more.

v2: Don't clobber the crtc struct name with the macro parameter ...

v3: Convert two more places noticed by Paulo Zanoni.
Reviewed-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The tileoffset register only supports a limited offset in x/y of 4096,
so for giant screen configuration with a shared fb we wrap around.

Fix this by computing a linear offset in tiles (pages) and only use
the tileoffset register to offset within the tile.
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

To avoid recomputing the display framebuffer offset on gen2/3
pageflips. This is also prep work to do similar trickery on gen4+

Also:
- kill "Start", such upper-case remnants from the ddx must surely die.
- rename "Offset" to linear_offset, to make it clearer that on gen4+
  this is only used by the hw for linear buffers, for tiled buffers it
  uses the TILEOFF register.
- call DSAPADDR DSPLINOFF on gen4+ for the same reason (and because
  the documentation really renamed the register).
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

MI display flips can't handle some changes in the framebuffer
format or layout. Return an error in such cases.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Zero initialize the mode_cmd structure when creating the kernel
framebuffer. Avoids having uninitialized data in offsets[0] for
instance.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The issue with this check is that it results in userspace receiving an
-EIO while the gpu reset hasn't completed, resulting in fallback to sw
rendering or worse.

Now there's also a stern comment in intel_ring_wait_seqno saying that
intel_ring_begin should not return -EAGAIN, ever, because some callers
can't handle that. But after an audit of the callsites I don't see any
issues. I guess the last problematic spot disappeared with the removal
of the pipelined fencing code.

So do the right thing and call check_wedge, which should properly
decide whether an -EAGAIN or -EIO is appropriate if wedged is set.

Note that the early check for a wedged gpu before touching the ring is
rather important (and it took me quite some time of acting like the
densest doofus to figure that out): If we don't do that and the gpu
died for good, not having been resurrect by the reset code, userspace
can merrily fill up the entire ring until it notices that something is
amiss.

Allowing userspace to emit more render, despite that we know that it
will fail can't lead to anything good (and by experience can lead to
all sorts of havoc, including angering the OOM gods and hard-hanging
the hw for good).

v2: Fix EAGAIN mispell, noticed by Chris Wilson.
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Tested-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

... instead of looping endless with no hope of ever serving that
page-fault. We only need to break out of this loop when the gpu died,
to run the reset work (and hopefully resurrect it).

To clarify questions Chris raised on irc: This is about handling I/O
errors not from our own code, but e.g. when the disk died when trying
to swap in a gem bo. So this patch remidies the issue that the current
handling only handles gpu-death-induced cases of -EIO. Admittedly,
dying disks are much rarer than hanging gpus ...To clarify questions
Chris raised on irc: This is about handling I/O errors not from our
own code, but e.g. when the disk died when trying to swap in a gem bo.
So this patch remidies the issue that the current handling only
handles gpu-death-induced cases of -EIO. Admittedly, dying disks are
much rarer than hanging gpus ...

This seems to have been lost in:

commit d9bc7e9f
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Mon Feb 7 13:09:31 2011 +0000

    drm/i915: Fix infinite loop regression from 21dd3734Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Tested-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

With the gpu reset no longer using a trylock we've increased the
chances of userspace getting stuck quite a bit. To make that
(hopefully) rare case more paletable time out when waiting for the gpu
reset code to complete and signal this little issue to the caller by
returning -EIO.

This should help userspace to somewhat gracefully fall back and
hopefully allow the user to grab some logs and reboot the machine
(instead of staring at a frozen X screen in agony).

Suggested by Chris Wilson because I've been stubborn about allowing
the gpu reset code no to fail, ever (by removing the trylock).
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Tested-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

So don't return -EAGAIN, even in the case of a gpu hang. Remap it to
-EIO instead. Note that this isn't really an issue with
interruptability, but more that we have quite a few codepaths (mostly
around kms stuff) that simply can't handle any errors and hence not
even -EAGAIN. Instead of adding proper failure paths so that we could
restart these ioctls we've opted for the cheap way out of sleeping
non-interruptibly.  Which works everywhere but when the gpu dies,
which this patch fixes.

So essentially interruptible == false means 'wait for the gpu or die
trying'.'

This patch is a bit ugly because intel_ring_begin is all non-interruptible
and hence only returns -EIO. But as the comment in there says,
auditing all the callsites would be a pain.

To avoid duplicating code, reuse i915_gem_check_wedge in __wait_seqno
and intel_wait_ring_buffer. Also use the opportunity to clarify the
different cases in i915_gem_check_wedge a bit with comments.

v2: Don't access dev_priv->mm.interruptible from check_wedge - we
might not hold dev->struct_mutex, making this racy. Instead pass
interruptible in as a parameter. I've noticed this because I've hit a
BUG_ON(!mutex_is_locked) at the top of check_wedge. This has been
added in

commit b4aca010
Author: Ben Widawsky <ben@bwidawsk.net>
Date:   Wed Apr 25 20:50:12 2012 -0700

    drm/i915: extract some common olr+wedge code

although that commit is missing any justification for this. I guess
it's just copy&paste, because the same commit add the same BUG_ON
check to check_olr, where it indeed makes sense.

But in check_wedge everything we access is protected by other means,
so this is superflous. And because it now gets in the way (we add a
new caller in __wait_seqno, which can be called without
dev->struct_mutext) let's just remove it.

v3: Group all the i915_gem_check_wedge refactoring into this patch, so
that this patch here is all about not returning -EAGAIN to callsites
that can't handle syscall restarting.

v4: Add clarification what interuptible == fales means in our code,
requested by Ben Widawsky.

v5: Fix EAGAIN mispell noticed by Chris Wilson.
Reviewed-by: NBen Widawsky <ben@bwidawsk.net>
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Tested-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Simply failing to reset the gpu because someone else might still hold
the mutex isn't a great idea - I see reliable silent reset failures.
And gpu reset simply needs to be reliable and Just Work.

"But ... the deadlocks!"

We already kick all processes waiting for the gpu before launching the
reset work item. New waiters need to check the wedging state anyway
and then bail out. If we have places that can deadlock, we simply need
to fix them.

"But ... testing!"

We have the gpu hangman, and if the current gpu load gem_exec_nop
isn't good enough to hit a specific case, we can add a new one.

"But ...  don't we return -EAGAIN for non-interruptible calls to
wait_seqno now?"

Yep, but this problem already exists in the current code. A follow up
patch will remedy this by returning -EIO for non-interruptible sleeps
if the gpu died and the low-level wait bails out with -EAGAIN.
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Tested-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Only bits 30:28, bit 31 is PIPE_DDI_FUNC_ENABLE.
Signed-off-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This pollutes dmesg output even if we do not have FBC for the device, so
move the DRM_DEBUG_KMS statement lower.

v2: just kill the message as suggested by Daniel.
Signed-off-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This is necessary for the modesetting to work correctly after a
suspend-resume cycle. Without this, the pipes and clocks got the correct
configuration, but the underlying DDI buffers configuration was lost.
Signed-off-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Reviewed-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This function is used to set the PCH_DREF_CONTROL register, which does
not exist on LPT anymore.
Signed-off-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Previously we had has_pch_split to tell us whether we had a PCH or not
and we also had dev_priv->pch_type to tell us which kind of PCH it
was, but it could only be used if we were 100% sure we did have a PCH.
Now that PCH_NONE was added to dev_priv->pch_type we don't need
has_pch_split anymore: we can just check for pch_type != PCH_NONE.

The HAS_PCH_{IBX,CPT,LPT} macros use dev_priv->pch_type, so they can
only be called after intel_detect_pch. The HAS_PCH_SPLIT macro looks
at dev_priv->info->has_pch_split, which is available earlier.

Since the goal is to implement HAS_PCH_SPLIT using dev_priv->pch_type
instead of dev_priv->info->has_pch_split, we need to make sure that
intel_detect_pch is called before any calls to HAS_PCH_SPLIT are made.
So we moved the intel_detect_pch call to an earlier stage.
Signed-off-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

And rely on the fact that it's 0 to assume that machines without a PCH
will have PCH_NONE as dev_priv->pch_type.

Just today I finally realized that HAS_PCH_IBX is true for machines
without a PCH. IMHO this is totally counter-intuitive and I don't
think it's a good idea to assume that we're going to check for
HAS_PCH_IBX only after we check for HAS_PCH_SPLIT.

I believe that in the future we'll have more PCH types and checks
like:

    if (HAS_PCH_IBX(dev) || HAS_PCH_CPT(dev))

will become more and more common. There's a good chance that we may
break non-PCH machines by adding these checks in code that runs on all
machines. I also believe that the HAS_PCH_SPLIT check will become less
common as we add more and more different PCH types. We'll probably
start replacing checks like:

    if (HAS_PCH_SPLIT(dev))
        foo();
    else
        bar();

with:

    if (HAS_PCH_NEW(dev))
        baz();
    else if (HAS_PCH_OLD(dev) || HAS_PCH_IBX(dev))
        foo();
    else
        bar();

and this may break gen 2/3/4.

As far as we have investigated, this patch will affect the behavior of
intel_hdmi_dpms and intel_dp_link_down on gen 4. In both functions the
code inside the HAS_PCH_IBX check is for IBX-specific workarounds, so
we should be safe. If we start bisecting gen 2/3/4 bugs to this commit
we should consider replacing the HAS_PCH_IBX checks with something
else.

V2: Improve commit message, list possible side effects and solution.
Signed-off-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

High frequency link configurations have the potential to cause trouble
with long and/or cheap cables, so prefer slow and wide configurations
instead.  This patch has the potential to cause trouble for eDP
configurations that lie about available lanes, so if we run into that we
can make it conditional on eDP.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=45801
Tested-by: peter@colberg.org
Signed-off-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>