linux-stable.git/drivers/gpu/drm/amd/amdgpu, branch linux-4.8.y

drm/amdgpu: fix init save/restore list in gfx_v8.0

2017-01-09T07:21:52+00:00

commit 202e0b227b906cb80a2791f21216a55d9468d61b upstream.

set valid data to mmRLC_SRM_INDEX_CNTL_ADDRx/DATAx.

Signed-off-by: Rex Zhu 
Reviewed-by: Alex Deucher 
Signed-off-by: Alex Deucher 
Signed-off-by: Greg Kroah-Hartman

drm/amd/amdgpu: enable GUI idle INT after enabling CGCG

2017-01-09T07:21:49+00:00

commit dd31ae9ac933636c3712b7dd0f6152c1d71f81fe upstream.

GUI idle interrupts should be enabled only after we
have enabled coarse grain clock gating (CGCG). This
prevents GFX engine generating idle interrupt even
though CGCG is not completely enabled.

Most of the time this goes un-noticed, but on some
Stoney ASICs this results in GFX engine hang after
system resumes from suspend. The issue is not
particular to Stoney though and could have occured
on any ASIC. The patch fixes this issue.

Reviewed-by: Alex Deucher 
Reported-by: Sunil Uttarwar 
Signed-off-by: Arindam Nath 
Signed-off-by: Alex Deucher 
Signed-off-by: Greg Kroah-Hartman

drm/amdgpu: fix check for port PM availability

2016-12-08T06:16:19+00:00

commit 7ac33e47d5769632010e537964c7e45498f8dc26 upstream.

The ATPX method does not always exist on the dGPU, it may be located at
the iGPU. The parent device of the iGPU is the root port for which
bridge_d3 is false. This accidentally enables the legacy PM method which
conflicts with port PM and prevented the dGPU from powering on.

Fixes: 1db4496f167b ("drm/amdgpu: fix power state when port pm is unavailable")

Reported-and-tested-by: Mike Lothian 
Signed-off-by: Peter Wu 
Signed-off-by: Alex Deucher 
Signed-off-by: Greg Kroah-Hartman

drm/amdgpu: fix power state when port pm is unavailable

2016-12-08T06:16:19+00:00

commit 1db4496f167bcc7c6541d449355ade2e7d339d52 upstream.

When PCIe port PM is not enabled (system BIOS is pre-2015 or the
pcie_port_pm=off parameter is set), legacy ATPX PM should still be
marked as supported. Otherwise the GPU can fail to power on after
runtime suspend. This affected a Dell Inspiron 5548.

Ideally the BIOS date in the PCI core is lowered to 2013 (the first year
where hybrid graphics platforms using power resources was introduced),
but that seems more risky at this point and would not solve the
pcie_port_pm=off issue.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=98505
Reported-and-tested-by: Nayan Deshmukh 
Signed-off-by: Peter Wu 
Signed-off-by: Alex Deucher 
Acked-by: Christian König 
Reviewed-by: Alex Deucher 
Signed-off-by: Greg Kroah-Hartman

drm/amdgpu: Attach exclusive fence to prime exported bo's. (v5)

2016-11-26T08:56:55+00:00

commit 8e94a46c1770884166b31adc99eba7da65a446a7 upstream.

External clients which import our bo's wait only
for exclusive dmabuf-fences, not on shared ones,
ditto for bo's which we import from external
providers and write to.

Therefore attach exclusive fences on prime shared buffers
if our exported buffer gets imported by an external
client, or if we import a buffer from an external
exporter.

See discussion in thread:
https://lists.freedesktop.org/archives/dri-devel/2016-October/122370.html

Prime export tested on Intel iGPU + AMD Tonga dGPU as
DRI3/Present Prime render offload, and with the Tonga
standalone as primary gpu.

v2: Add a wait for all shared fences before prime export,
    as suggested by Christian Koenig.

v3: - Mark buffer prime_exported in amdgpu_gem_prime_pin,
    so we only use the exclusive fence when exporting a
    bo to external clients like a separate iGPU, but not
    when exporting/importing from/to ourselves as part of
    regular DRI3 fd passing.

    - Propagate failure of reservation_object_wait_rcu back
    to caller.

v4: - Switch to a prime_shared_count counter instead of a
      flag, which gets in/decremented on prime_pin/unpin, so
      we can switch back to shared fences if all clients
      detach from our exported bo.

    - Also switch to exclusive fence for prime imported bo's.

v5: - Drop lret, instead use int ret -> long ret, as proposed
      by Christian.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=95472
Tested-by: Mike Lothian  (v1)
Signed-off-by: Mario Kleiner 
Reviewed-by: Christian König .
Cc: Christian König 
Cc: Michel Dänzer 
Signed-off-by: Alex Deucher 
Signed-off-by: Greg Kroah-Hartman

drm/amdgpu: fix a vm_flush fence leak

2016-11-18T09:51:53+00:00

commit 2d7c17be00e0dce3bc1a092a2c277a9f86c69ca9 upstream.

Looks like .last_flush reference is left at teardown.
Leak reported by CONFIG_SLUB_DEBUG.

Fixes: 41d9eb2c5a2a ("drm/amdgpu: add a fence after the VM flush")
Reviewed-by: Chunming Zhou 
Reviewed-by: Christian König 
Signed-off-by: Grazvydas Ignotas 
Signed-off-by: Alex Deucher 
Signed-off-by: Greg Kroah-Hartman

drm/amdgpu: fix fence slab teardown

2016-11-18T09:51:53+00:00

commit 0f10425e811355986907c54f7d1d06703e406092 upstream.

To free fences, call_rcu() is used, which calls amdgpu_fence_free()
after a grace period. During teardown, there is no guarantee all
callbacks have finished, so amdgpu_fence_slab may be destroyed before
all fences have been freed. If we are lucky, this results in some slab
warnings, if not, we get a crash in one of rcu threads because callback
is called after amdgpu has already been unloaded.

Fix it with a rcu_barrier().

Fixes: b44135351a3a ("drm/amdgpu: RCU protected amdgpu_fence_release")
Acked-by: Chunming Zhou 
Reviewed-by: Christian König 
Signed-off-by: Grazvydas Ignotas 
Signed-off-by: Alex Deucher 
Signed-off-by: Greg Kroah-Hartman

drm/amd: fix scheduler fence teardown order v2

2016-11-18T09:51:49+00:00

commit c24784f01549ecdf23fc00d0588423bcf8956714 upstream.

Some fences might be alive even after we have stopped the scheduler leading
to warnings about leaked objects from the SLUB allocator.

Fix this by allocating/freeing the SLUB allocator from the module
init/fini functions just like we do it for hw fences.

v2: make variable static, add link to bug

Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=97500

Reported-by: Grazvydas Ignotas 
Signed-off-by: Christian König 
Reviewed-by: Alex Deucher  (v1)
Signed-off-by: Alex Deucher 
Signed-off-by: Greg Kroah-Hartman

drm/amdgpu: fix crash in acp_hw_fini

2016-11-18T09:51:49+00:00

commit 757124d95c42bb579d67df51e51789849929ee31 upstream.

On CZ/ST systems with AZ rather than ACP audio, we need to bail
early in hw_fini since there is nothing to do.

bug: https://bugs.freedesktop.org/show_bug.cgi?id=98276

Signed-off-by: Alex Deucher 
Signed-off-by: Greg Kroah-Hartman

drm/amdgpu: disable runtime pm in certain cases

2016-11-18T09:51:49+00:00

commit 84b1528e8cef55274f0df20e93513b3060ce495a upstream.

If the platform does not support hybrid graphics or ATPX dGPU
power control.

Acked-by: Christian König 
Signed-off-by: Alex Deucher 
Signed-off-by: Greg Kroah-Hartman