When executing in a PCI passthrough based virtuzliation environment, the
hypervisor will usually attempt to send a PCIe bus reset signal to the
ASIC when the VM reboots. In this scenario, the card is not correctly
initialized, but we still consider it to be posted. Therefore, in a
passthrough based environemnt we should always post the card to guarantee
it is in a good state for driver initialization.

Ported from amdgpu commit:
amdgpu: fix asic initialization for virtualized environments

Cc: Andres Rodriguez <andres.rodriguez@amd.com>
Cc: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

Seems r600/r700 does not like hard reset while freezing for hibernation
(regression due to 274ad65c which itself
is a fix for hibernation on some GPU families). Until i can debug further
issue with r600, let just disable this for r600/r700 as they are very
similar family and bug affecting one likely affect the other.
Reviewed-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NJérôme Glisse <jglisse@redhat.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Some GPU block like UVD and VCE require hard reset to be properly
resume if there is no real powerdown of the asic like during various
hibernation step. This patch perform such hard reset.
Reviewed-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NJérôme Glisse <jglisse@redhat.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

This reverts commit e64c952e.

ATPX is the ACPI method for controlling AMD PowerXpress laptops.
There are flags to indicate which methods are supported.  If
the dGPU power down flag is not supported, the driver needs to
implement the dGPU power down manually.  We had previously
always forced the driver to assume the ATPX dGPU power down
was present, but this causes problems on boards where it is
not, leading to GPU hangs when attempting to power down the
dGPU.  Manual dGPU power down is not currently supported in
the Linux driver.  Some laptops indicate that the ATPX
dGPU power down method is not present, but it actually
apparently is.  I'm not sure if this is a bios bug and it should
be set or if there is a reason it was unset and the method should
not be used.  This is not an issue on other OSes since both the
ATPX and the manual driver power down methods are supported.

This is apparently fairly widespread, so just revert for now.

bugs:
https://bugzilla.kernel.org/show_bug.cgi?id=115321
https://bugzilla.kernel.org/show_bug.cgi?id=116581
https://bugzilla.kernel.org/show_bug.cgi?id=116251

Cc: stable@vger.kernel.org

The pci revision id is useful in debugging certain things as
it's part of how SKUs are defined on newer asics.
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

I hate doing this but it hurts my eyes to go over code that does not
comply with indentation rules. Only thing that is not only space change
is in atom.c all other files are space indentation issues.
Acked-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NJérôme Glisse <jglisse@redhat.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Some PX laptops don't provide an ACPI method to control dGPU power.  On
those systems, the driver is responsible for handling the dGPU power
state.  Disable runtime PM on them until support for this is implemented.
Reviewed-by: NMichel Dänzer <michel.daenzer@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

This reverts commit dbb17a21.

It turns out that commit can cause problems for systems with multiple
GPUs, and causes X to hang on at least a HP Pavilion dv7 with hybrid
graphics.

This got noticed originally in 4.4.4, where this patch had already
gotten back-ported, but 4.5-rc7 was verified to have the same problem.

Alexander Deucher says:
 "It looks like you have a muxed system so I suspect what's happening is
  that one of the display is being reported as connected for both the
  IGP and the dGPU and then the desktop environment gets confused or
  there some sort problem in the detect functions since the mux is not
  switched to the dGPU.  I don't see an easy fix unless Dave has any
  ideas.  I'd say just revert for now"
Reported-by: NJörg-Volker Peetz <jvpeetz@web.de>
Acked-by: NAlexander Deucher <Alexander.Deucher@amd.com>
Cc: Dave Airlie <airlied@gmail.com>
Cc: stable@kernel.org  # wherever dbb17a21 got back-ported
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Reviewed-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlexandre Demers <alexandre.f.demers@gmail.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Need to call this on resume if displays changes during
suspend in order to properly be notified of changes.
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

Currently everyone and their dog has their own favourite spelling
for vga_switcheroo. This makes it hard to grep dmesg for log entries
relating to vga_switcheroo. It also makes it hard to find related
source files in the tree.

vga_switcheroo.c uses pr_fmt "vga_switcheroo". Use that everywhere.
Signed-off-by: NLukas Wunner <lukas@wunner.de>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

In

commit 7a3f3d66
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Thu Jul 9 23:44:28 2015 +0200

    drm: Check locking in drm_for_each_connector

I added locking checks to drm_for_each_connector but failed that
through drm_helper_connector_dpms -> drm_helper_choose_encoder_dpms
it's used in a few more places in the radeon resume/suspend code.

Fix them up.

Note that we could use the connector iterator macros in there too, but
that's for the future.
Reported-and-tested-by: NBorislav Petkov <bp@alien8.de>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@intel.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Everything is evicted from VRAM before suspend, so we need to make
sure all BOs are unpinned and re-pinned after resume. Fixes broken
mouse cursor after resume introduced by commit b9729b17.

[Michel Dänzer: Add pinning BOs on resume]

v2:
[Alex Deucher: merge cursor unpin into fb unpin loop]

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=100541
Cc: stable@vger.kernel.org
Reviewed-by: Christian König <christian.koenig@amd.com> (v1)
Signed-off-by: NGrigori Goronzy <greg@chown.ath.cx>
Signed-off-by: NMichel Dänzer <michel.daenzer@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Newer ASICs have more VRAM on average and allocating more GART as
well can have advantages. Also see commit edcd26e8.

Ideally, we should scale GART size based on actual VRAM size, but
that requires significant restructuring of initialization.

v2: extract small helper, apply to error paths
Reviewed-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NGrigori Goronzy <greg@chown.ath.cx>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Laptop with Turks/Thames GPU will freeze if dpm is enabled. It seems
the SMC engine is relying on some state inside the CP engine. CP needs
to chew at least one packet for it to get in good state for dynamic
power management.

This patch simply disabled and re-enable DPM after the ring test which
is enough to avoid the freeze.
Signed-off-by: NJérôme Glisse <jglisse@redhat.com>
Cc: stable@vger.kernel.org
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Userspace will be able to tell whether a GPU reset occured by comparing
an old referece value of the counter with a new value.
Reviewed-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NMarek Olšák <marek.olsak@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

This adds initial DP 1.2 MST support to radeon, on CAYMAN
and up in theory.

This is off by default.

v2: agd5f:
- add UNIPHY3 offsets
- move atom cmd table code into atombios_encoders.c
- whitespace cleanup
- replace some magic numbers with proper defines
Signed-off-by: NDave Airlie <airlied@redhat.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

get_page_entry calculates the GART page table entry, which is just written
to the GART page table by set_page_entry.

This is a prerequisite for the following fix.
Reviewed-by: NChristian König <christian.koenig@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: NMichel Dänzer <michel.daenzer@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

While developing MST support I noticed I often got the wrong data
back from a transaction, in a racy fashion. I noticed the scratch
space wasn't locked against concurrent users.

Based on a patch by Alex, but I've made it a bit more obvious when
things are locked.
Signed-off-by: NDave Airlie <airlied@redhat.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

The page offset is 12 bits.  For example if we have an
8 GB VM, we'd need 33 bits.  The number of bits needed
for PD + PT is 21 (33 - 12 or log2(8) + 18), not 20
(log2(8) + 17).

Noticed by Alexey during code review.
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

Adds an extra argument to radeon_bo_create, which is only used in radeon_prime.c.
Reviewed-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Seems to have problems turning the dGPU on/off.

bug:
https://bugzilla.kernel.org/show_bug.cgi?id=51381Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Use the new vga_switcheroo_fini_domain_pm_ops function
to unregister the pm ops.

Based on a patch from:
Pali Rohár <pali.rohar@gmail.com>

bug:
https://bugzilla.kernel.org/show_bug.cgi?id=84431Reviewed-by: NBen Skeggs <bskeggs@redhat.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: NPali Rohár <pali.rohar@gmail.com>
Cc: Ben Skeggs <bskeggs@redhat.com>

Changes since v1:
- Kill the sw interrupt dance, add and use
  radeon_irq_kms_sw_irq_get_delayed instead.
- Change custom wait function, lockdep complained about it.
  Holding exclusive_lock in the wait function might cause deadlocks.
  Instead do all the processing in .enable_signaling, and wait
  on the global fence_queue to pick up gpu resets.
- Process all fences in radeon_gpu_reset after reset to close a race
  with the trylock in enable_signaling.
Changes since v2:
- Small changes to work with the rewritten lockup recovery patches.
Changes since v3:
- Call radeon_fence_schedule_check when exclusive_lock cannot be
  acquired to always cause a wake up.
- Reset irqs from hangup check.
- Drop reading seqno in the callback, use cached value.
- Fix indentation in radeon_fence_default_wait
- Add a radeon_test_signaled function, drop a few test_bit calls.
- Make to_radeon_fence global.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>

Blocking completely innocent processes with a GPU reset is
a pretty bad idea. Just set needs_reset and let the next
command submission or fence wait do the job.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

This is needed for the next commit, because the lockup detection
will need the read lock to run.

v4 (chk): split out forced fence completion, remove unrelated changes,
          add and handle in_reset flag
v5 (agd5f): rebase fix
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Instead of resetting all fence numbers, only reset the
number of the problematic ring. Split out from a patch
from Maarten Lankhorst <maarten.lankhorst@canonical.com>

v2 (agd5f): rebase build fix
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Otherwise we won't test if the fallback to PCIe GART really worked.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NMichel Dänzer <michel.daenzer@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

If the display hw was reset or a hard reset was used,
we need to re-init some of the common display hardware as well.
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

pm_suspend is handled in the radeon_suspend callbacks.
pm_resume has special handling depending on whether
dpm or legacy pm is enabled.  Change radeon_gpu_reset
to mirror the behavior in the suspend and resume
pathes.
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>
Cc: stable@vger.kernel.org

Whenever userspace mapping related to our userptr change
we wait for it to become idle and unmap it from GTT.

v2: rebased, fix mutex unlock in error path
v3: improve commit message
Signed-off-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

v2: rebase on vm_size scale change. Adjust vm_size default to 8,
    Better handle the default and smaller values.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Some PX laptops seems to have problems turning the dGPU on/off.
Add a quirk list to disable runpm by default on those systems.
Also convert the current PX d3 delay handling to a quirk.

bug:
https://bugzilla.kernel.org/show_bug.cgi?id=51381
https://bugzilla.kernel.org/show_bug.cgi?id=74551Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

v2: fix rebase onto drm-fixes
Signed-off-by: NMichel Dänzer <michel.daenzer@amd.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

VM sizes smaller than 1GB doesn't make much sense anyway.

v2: fix typo and grammer
Signed-off-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Implementing a lock for selecting and accessing shader engines and arrays.
This lock will make sure that radeon and amdkfd are not colliding when
accessing shader engines and arrays with GRBM_GFX_INDEX register.
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NOded Gabbay <oded.gabbay@amd.com>

Signed-off-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

v2: agd5f: simplify patch
Signed-off-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Setting the power state prior to restoring the display
hardware leads to blank screens on some systems.  Drop
the power state set from dpm resume.  The power state
will get set as part of the mode set sequence.  Also
add an explicit power state set after mode set resume
to cover PX and headless systems.

bug:
https://bugzilla.kernel.org/show_bug.cgi?id=76761Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: NDave Airlie <airlied@redhat.com>

So a few people complained that

commit 177cf92d
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Tue Apr 1 22:14:59 2014 +0200

    drm/crtc-helpers: fix dpms on logic

which was merged into 3.15-rc1, broke resume on radeons. Strangely git
bisect lead everyone to

commit 25f397a4
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Fri Jul 19 18:57:11 2013 +0200

    drm/crtc-helper: explicit DPMS on after modeset

which was merged long ago and actually part of 3.14.

Digging deeper I've noticed (again) that the call to
drm_helper_resume_force_mode in the radeon resume handlers was a no-op
previously because everything gets shut down on suspend. radeon does
this with explicit calls to drm_helper_connector_dpms with DPMS_OFF.
But with 177c we now force the dpms state to ON, so suddenly
resume_force_mode actually forced the crtcs back on.

This is the intention of the change after all, the problem is that
radeon resumes the fbdev console layer _before_ restoring the display,
through calling fb_set_suspend. And fbcon does an immediate ->set_par,
which in turn causes the same forced mode restore to happen.

Two concurrent modeset operations didn't lead to happiness. Fix this
by delaying the fbcon resume until the end of the readeon resum
functions.

v2: Fix up a bit of the spelling fail.

References: https://lkml.org/lkml/2014/5/29/1043
References: https://lkml.org/lkml/2014/5/2/388
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=74751Tested-by: NKen Moffat <zarniwhoop@ntlworld.com>
Cc: Alex Deucher <alexdeucher@gmail.com>
Cc: Ken Moffat <zarniwhoop@ntlworld.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: NDave Airlie <airlied@gmail.com>