It's deprecated and only should be used by drivers which still use
drm_platform_init, not by anyone else.

And indeed it's entirely unused and can be nuked.

This required a bit more fudging, but I guess kirin_dc_ops really
wants to operate on the platform_device, not something else. Also
bonus points for implementing abstraction, and then storing the vfunc
in a global variable.

v2: Don't break the build soooo badly :(
Note that the cleanup function is a bit confused: ade_data was never
set as drvdata, and calling drm_crtc_cleanup directly is a bug - this
is called indirectly through drm_mode_config_cleanup, which calls into
crtc->destroy, which already has the call to drm_crtc_cleanup. Which
means we can just nuke it.

Note this is the 2nd attempt after the first one failed and had to be
reverted again in

commit 9cd2e854
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Wed Aug 17 13:59:40 2016 +0200

    Revert "drm/hisilicon: Don't set drm_device->platformdev"

Cc: Xinliang Liu <xinliang.liu@linaro.org>
Cc: Xinwei Kong <kong.kongxinwei@hisilicon.com>
Cc: Archit Taneja <architt@codeaurora.org>
Cc: Sean Paul <seanpaul@chromium.org>
Reviewed-by: NSean Paul <seanpaul@chromium.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161209141944.22121-1-daniel.vetter@ffwll.ch

With the last round of changes all ioctls called by modern drivers now
have their own locking. Everything else is only allowed for legacy
drivers and hence the lack of locking doesn't matter.

One exception is nouveau, due to the DRIVER_KMS_LEGACY_CONTEXT flag.
But that only works its magic on the context and bufs ioctls. And
drm_bufs.c is protected with dev->struct_mutex, and drm_context.c by
the same and dev->ctxlist_mutex. That should be all safe, and we can
finally mandata drm-bkl-less ioctls for everyone!
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDaniel Vetter <daniel.vetter@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161210215255.7765-3-daniel.vetter@ffwll.ch

It only updates per-file feature flags. And all the ioctl which change
behaviour depending upon these flags (they're all kms features) do
_not_ hold the BKL. Therefor this is pure cargo-cult and can be
removed.

Note that there's a risk that the ioctl will behave inconsistently
when userspace is racing with itself, but that's ok. The only thing
it's not allowed to do is oops the kernel, and from an audit all
places are safe.

v2: Clarify that the inconsistency is only when userspace races
(Chris).
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDaniel Vetter <daniel.vetter@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161210215255.7765-2-daniel.vetter@ffwll.ch

No one looks at the major/minor versions except the unique/busid
stuff. If we protect that with the master_mutex (since it also affects
the unique of each master, oh well) we can mark these two IOCTL with
DRM_UNLOCKED.

While doing this I realized that the comment for the magic_map is
outdated, I've forgotten to update it in:

commit d2b34ee6
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Fri Jun 17 09:33:21 2016 +0200

    drm: Protect authmagic with master_mutex

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: NEmil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161210215255.7765-1-daniel.vetter@ffwll.ch

Not setting the fb modifiers flag is something different from setting
the fb modifiers to 0 (which means explicitly linear). We kinda failed
to document that properly. Spotted by Kristian.

Cc: hoegsberg@google.com
Reviewed-by: NKristian H. Kristensen <hoegsberg@google.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1478694996-4200-1-git-send-email-daniel.vetter@ffwll.ch

Looping twice when we can do it once is silly. Also use a consistent
style. Note that there's a good race with the connector list walking,
since that is no longer protected by mode_config.mutex. But that's for
a later patch to fix.

v2: Actually try to not blow up, somehow I lost the hunk that checks
we don't copy too much. Noticed by Chris.

v3:
- squash all drm_mode_getresources cleanups into one
- use consistent style for walking objects (Chris)

v4:
- Use u64_to_user_ptr (Chris)
- Don't forget to copy the last connector (Chris)

v5: Chris was right ...

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDaniel Vetter <daniel.vetter@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161211192019.29603-1-daniel.vetter@ffwll.ch

Joonas complained that writing ww_mutex_lock(&resv->lock, ctx) was too
intrusive compared to reservation_object_lock(resv, ctx);
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Cc: Sumit Semwal <sumit.semwal@linaro.org>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20161115154642.31850-1-chris@chris-wilson.co.uk

When DPMS was introduced to atomic, vblanks only worked when the crtc
was enabled and active. wait_for_vblanks were not converted to check for
crtc_state->active, which may cause an attempt for vblank_get to fail.

This is probably harmless, but convert from enable to active anyway.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1481204729-9058-2-git-send-email-maarten.lankhorst@linux.intel.com

It's supported now! Spotted while reviewing Chris' patch to add a
release hook.

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDaniel Vetter <daniel.vetter@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161208102847.3063-1-daniel.vetter@ffwll.ch

Currently the reference for the dmabuf->obj is incremented for the
dmabuf in drm_gem_prime_handle_to_fd() (at the high level userspace
interface), but is released in drm_gem_dmabuf_release() (the lowlevel
handler). Improve the symmetry of the dmabuf->obj ownership by acquiring
the reference in drm_gem_dmabuf_export(). This makes it easier to use
the prime functions directly.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
[danvet: Update kerneldoc.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20161207214527.22533-1-chris@chris-wilson.co.uk

vgem (and our igt tests using vgem) need this. I suspect etnaviv will
fare similarly.

v2. Make it build. Oops.

Fixes: d5264ed3 ("drm: Return -ENOTSUPP when called for KMS cap with a non-KMS driver")
Cc: Michel Dänzer <michel.daenzer@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: NMichel Dänzer <michel.daenzer@amd.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161207144939.22756-1-daniel.vetter@ffwll.ch

Instead of dealing with crtc details inside drm_atomic.c we should
just export a function that creates a new crtc fence for us and
use that.
Suggested-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1481046437-18778-1-git-send-email-gustavo@padovan.org

We will ignored PSR setting if panel not support it. So, in this case,
we should return from analogix_dp_enable/disable_psr() without any
error code. Let's retrun 0 instead of -EINVAL when panel not support PSR
in analogix_dp_enable/disable_psr().
Signed-off-by: Nzain wang <wzz@rock-chips.com>
Signed-off-by: NArchit Taneja <architt@codeaurora.org>
Link: http://patchwork.freedesktop.org/patch/msgid/1481072253-8917-1-git-send-email-wzz@rock-chips.com

We now support nonblocking commits on drm_atomic_helper_commit()
so the comment is not valid anymore.
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1480946626-30917-1-git-send-email-gustavo@padovan.org

If we cannot acquire the mode_config.mutex immediately, just back off and
queue a new attempt after the poll interval. This is mostly to stop the
hung task spam when the system is deadlocked, but it will also lessen
the load (in such extreme cases).
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: NEric Engestrom <eric.engestrom@imgtec.com>
[danvet:s/lock/mutex/ per Eric's comment.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20161206113715.30382-1-chris@chris-wilson.co.uk

This is an attempt to make the previous fix a bit more robust going
forward.

v2:
* Only allow DRM_CAP_TIMESTAMP_MONOTONIC with UMS drivers (Daniel
  Vetter, Alex Deucher)
* Different logic to keep DRM_CAP_TIMESTAMP_MONOTONIC separate from
  the other caps (Daniel Vetter)
Signed-off-by: NMichel Dänzer <michel.daenzer@amd.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20161201073731.5716-1-michel@daenzer.net

Backmerge v4.9-rc8 to get at

commit e94bd173
Author: Michel Dänzer <michel.daenzer@amd.com>
Date:   Wed Nov 30 17:30:01 2016 +0900

drm: Don't call drm_for_each_crtc with a non-KMS driver

so I can apply Michel's follow-up patch.
Signed-off-by: NDaniel Vetter <daniel.vetter@intel.com>

Since commit 8a357d10 ("drm: Nerf DRM_CONTROL nodes"), a
struct drm_device's ->control member is always NULL.

In the case of CONFIG_DEBUG_FS=y, amdgpu_debugfs_add_files() accesses
->control->debugfs_root though. This results in a NULL pointer
dereference.

Fix this by omitting the drm_debugfs_create_files() call for the
control minor debugfs directory which is now non-existent anyway.

Fixes: 8a357d10 ("drm: Nerf DRM_CONTROL nodes")
Signed-off-by: NNicolai Stange <nicstange@gmail.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=98915Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>
Tested-by: NMike Lothian <mike@fireburn.co.uk>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20161205203022.11671-1-nicstange@gmail.com

The drm_atomic_helper_connector_dpms
will set the connector back the old DPMS state
before returning. This makes it impossible to change
DPMS state of the connector.

Fixes: 0853695c
v2: edit of commit message
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Eric Engestrom <eric.engestrom@imgtec.com>
Cc: Sean Paul <seanpaul@chromium.org>
Cc: dri-devel@lists.freedesktop.org
Cc: <drm-intel-fixes@lists.freedesktop.org>
Signed-off-by: NMarta Lofstedt <marta.lofstedt@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20161205120408.13056-1-marta.lofstedt@intel.com

Merge branch 'exynos-drm-next' of git://git.kernel.org/pub/scm/linux/kernel/git/daeinki/drm-exynos into drm-next

Just refactoring HDMI driver by using infoframe helper
function, fixing GSC Kconfig dependency issue and including trivial
cleanups.

* 'exynos-drm-next' of git://git.kernel.org/pub/scm/linux/kernel/git/daeinki/drm-exynos:
  drm/exynos: Use VIDEO_SAMSUNG_EXYNOS_GSC=n as GSC Kconfig dependency
  drm/exynos: gsc: fix spelling mistakes
  exynos-drm: Fix error messages to print flags and size
  drm/exynos/hdmi: refactor infoframe code

- lots of code cleanup
- lots of bug fixes
- expose rpm based fan info via hwmon
- lots of clock and powergating fixes
- SI register header cleanup and conversion to common format used by newer asics

* 'drm-next-4.10' of git://people.freedesktop.org/~agd5f/linux: (54 commits)
  drm/amdgpu: drop is_display_hung from display funcs
  drm/amdgpu/uvd: reduce IB parsing overhead on UVD5+ (v2)
  drm/amdgpu/uvd: consolidate code for fetching addr from ctx
  drm/amdgpu: Disable DPM in virtualization
  drm/amdgpu: use AMDGPU_GEM_CREATE_VRAM_CLEARED for VM PD/PTs (v2)
  drm/amdgpu: improve AMDGPU_GEM_CREATE_VRAM_CLEARED handling (v2)
  drm/amdgpu: fix error handling in amdgpu_bo_create_restricted
  drm/amdgpu: fix amdgpu_fill_buffer (v2)
  drm/amdgpu: remove amdgpu_irq_get_delayed
  amdgpu: Wrap dev_err() calls on vm faults with printk_ratelimit()
  amdgpu: Use dev_err() over vanilla printk() in vm_decode_fault()
  drm/amd/amdgpu: port of DCE v6 to new headers (v3)
  drm/amdgpu: cleanup unused iterator members for sdma v2.4
  drm/amdgpu: cleanup unused iterator members for sdma v3
  drm/amdgpu:impl vgt_flush for VI(V5)
  drm/amdgpu: enable uvd mgcg for Fiji.
  drm/amdgpu: refine cz uvd clock gate logic.
  drm/amdgpu: change log level to KERN_INFO in ci_dpm.c
  drm/amdgpu: always un-gate UVD REGS path.
  drm/amdgpu/sdma: fix typo in packet setup
  ...

- fix dma-buf export path to return correct SG table
- trivially implement direct dma-buf mapping
- allow DRAW_INSTANCED commands in validator
- make the driver work on i.MX6SX, yielding a working 2D/3D stack
together with Mareks MXS DRM driver

* 'drm-etnaviv-next' of git://git.pengutronix.de/lst/linux:
  MAINTAINERS: add etnaviv mailinglist
  drm/etnaviv: move linear window on MC1.0 parts if necessary
  drm/etnaviv: don't invoke OOM killer from dump code
  drm/etnaviv: fix gem_prime_get_sg_table to return new SG table
  drm/etnaviv: Allow DRAW_INSTANCED commands
  drm/etnaviv: implement dma-buf mmap

- BIT_PERF_PTRS uses 32-bit pointers to its subtables, we were parsing
them as 16-bit, causing various issues on newer boards.
- Support for MXM on GM20x and up.
- More display-related fixes.

* 'linux-4.10' of git://github.com/skeggsb/linux:
  drm/nouveau/mxm: warn more loudly on unsupported DCB version
  drm/nouveau/mxm: handle DCB 4.1 modification
  drm/nouveau/bios/mxm: handle digital connector table 1.1
  drm/nouveau: Queue hpd_work on (runtime) resume
  drm/nouveau: Rename acpi_work to hpd_work
  drm/nouveau/kms/nv50: Fix atomic pageflip events.
  drm/nouveau/fb/ram/gp100-: fix memory detection where FBP_NUM != FBPA_NUM
  drm/nouveau/bios/volt: pointers are 32-bit
  drm/nouveau/bios/vmap: pointers are 32-bit
  drm/nouveau/bios/timing: pointers are 32-bit
  drm/nouveau/bios/therm: pointers are 32-bit
  drm/nouveau/bios/perf: pointers are 32-bit
  drm/nouveau/bios/iccsense: pointers are 32-bit
  drm/nouveau/bios/fan: pointers are 32-bit
  drm/nouveau/bios/cstep: pointers are 32-bit
  drm/nouveau/bios/boost: pointers are 32-bit

Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

Allows MXM DCB modification to be handled on GM20x and newer boards.
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

I suspect the version bump is just to signify that the table now specifies
pad macro/links instead of SOR/sublinks.

For our usage of the table, just recognising the new version is enough.
Signed-off-by: NBen Skeggs <bskeggs@redhat.com>

In function qxl_release_alloc(), when kmalloc() returns a NULL pointer,
it returns value 0 and parameter *ret is uninitialized. 0 means no error
to the callers of qxl_release_alloc(). The callers keep going and will
try to reference the uninitialized variable. This patch fixes the bug,
returning "-ENOMEM" when kmalloc() fails.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=188911Signed-off-by: NPan Bian <bianpan2016@163.com>
[seanpaul fixed up subject prefix]
Signed-off-by: NSean Paul <seanpaul@chromium.org>
Link: http://patchwork.freedesktop.org/patch/msgid/1480777902-7648-1-git-send-email-bianpan2016@163.com

Commit aeefb368 ("drm/exynos: gsc: add device tree support and remove
usage of static mappings") made the DRM_EXYNOS_GSC Kconfig symbol to only
be selectable if the exynos-gsc V4L2 driver isn't enabled, since both use
the same HW IP block.

But added the dependency as depends on !VIDEO_SAMSUNG_EXYNOS_GSC which is
not correct since Kconfig expressions are not boolean but tristate. So it
will only evaluate to 'n' if VIDEO_SAMSUNG_EXYNOS_GSC=y but will evaluate
to 'm' if VIDEO_SAMSUNG_EXYNOS_GSC=m.

This means that both the V4L2 and DRM drivers can be enabled if the former
is enabled as a module, which isn't what we want since otherwise 2 drivers
could attempt to use the hardware at the same time.
Signed-off-by: NJavier Martinez Canillas <javier@osg.samsung.com>
Signed-off-by: NInki Dae <inki.dae@samsung.com>

Trivial fixes to spelling mistakes "precalser" to "prescaler"
in dev_err messages
Signed-off-by: NColin Ian King <colin.king@canonical.com>
Reviewed-by: NJavier Martinez Canillas <javier@osg.samsung.com>
Signed-off-by: NInki Dae <inki.dae@samsung.com>

Fix exynos_drm_gem_create() error messages to include flags and size when
flags and size are invalid.
Signed-off-by: NShuah Khan <shuahkh@osg.samsung.com>
Signed-off-by: NInki Dae <inki.dae@samsung.com>

Use core helpers to generate infoframes and generate vendor frame if necessary.

Changelog:
- changed 'ret >= 0' checks to '!ret'
Signed-off-by: NAndrzej Hajda <a.hajda@samsung.com>
Signed-off-by: NInki Dae <inki.dae@samsung.com>

They're gone since 8a357d10 ("drm: Nerf DRM_CONTROL nodes").
Spotted while doing a full audit when revieng a similar patch from
Nicolai for radeon.

v2: Drink coffee first aka don't forget the unregister side.

Cc: Dave Airlie <airlied@redhat.com>
Fixes: 8a357d10 ("drm: Nerf DRM_CONTROL nodes")
Cc: Nicolai Stange <nicstange@gmail.com>
Acked-by: NDave Airlie <airlied@redhat.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161205072926.12546-1-daniel.vetter@ffwll.ch

Since commit 8a357d10 ("drm: Nerf DRM_CONTROL nodes"), a
struct drm_device's ->control member is always NULL.

In the case of CONFIG_DEBUG_FS=y, radeon_debugfs_add_files() accesses
->control->debugfs_root though. This results in the following Oops:

  BUG: unable to handle kernel NULL pointer dereference at 0000000000000018
  IP: radeon_debugfs_add_files+0x90/0x100 [radeon]
  PGD 0
  Oops: 0000 [#1] SMP
  [...]
  Call Trace:
   ? work_on_cpu+0xb0/0xb0
   radeon_fence_driver_init+0x120/0x150 [radeon]
   si_init+0x122/0xd50 [radeon]
   ? _raw_spin_unlock_irq+0x2c/0x40
   ? device_pm_check_callbacks+0xb3/0xc0
   radeon_device_init+0x958/0xda0 [radeon]
   radeon_driver_load_kms+0x9a/0x210 [radeon]
   drm_dev_register+0xa9/0xd0 [drm]
   drm_get_pci_dev+0x9c/0x1e0 [drm]
   radeon_pci_probe+0xb8/0xe0 [radeon]
  [...]

Fix this by omitting the drm_debugfs_create_files() call for the
control minor debugfs directory which is now non-existent anyway.

Fixes: 8a357d10 ("drm: Nerf DRM_CONTROL nodes")
Signed-off-by: NNicolai Stange <nicstange@gmail.com>
Acked-by: NDave Airlie <airlied@redhat.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20161203144700.2307-1-nicstange@gmail.com

Linux 4.9-rc8

Daniel requested this so we could apply some follow on fixes cleanly to -next.

Pull drm fixes from Dave Airlie:
 "A pretty small pull request: a couple of AMD powerxpress regression
  fixes and a power management fix, a couple of i915 fixes and one hdlcd
  fix, along with one core don't oops because of incorrect API usage fix"

* tag 'drm-fixes-for-v4.9-rc8' of git://people.freedesktop.org/~airlied/linux:
  drm/i915: drop the struct_mutex when wedged or trying to reset
  drm/i915: Don't touch NULL sg on i915_gem_object_get_pages_gtt() error
  drm: Don't call drm_for_each_crtc with a non-KMS driver
  drm/radeon: fix check for port PM availability
  drm/amdgpu: fix check for port PM availability
  drm/amd/powerplay: initialize the soft_regs offset in struct smu7_hwmgr
  drm: hdlcd: Fix cleanup order

Merge tag 'drm-intel-fixes-2016-12-01' of git://anongit.freedesktop.org/git/drm-intel into drm-fixes

2 intel fixes.

* tag 'drm-intel-fixes-2016-12-01' of git://anongit.freedesktop.org/git/drm-intel:
  drm/i915: drop the struct_mutex when wedged or trying to reset
  drm/i915: Don't touch NULL sg on i915_gem_object_get_pages_gtt() error

Merge more fixes from Andrew Morton:
 "2 fixes"

* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
  mm, vmscan: add cond_resched() into shrink_node_memcg()
  mm: workingset: fix NULL ptr in count_shadow_nodes

Boris Zhmurov has reported RCU stalls during the kswapd reclaim:

  INFO: rcu_sched detected stalls on CPUs/tasks:
   23-...: (22 ticks this GP) idle=92f/140000000000000/0 softirq=2638404/2638404 fqs=23
   (detected by 4, t=6389 jiffies, g=786259, c=786258, q=42115)
  Task dump for CPU 23:
  kswapd1         R  running task        0   148      2 0x00000008
  Call Trace:
    shrink_node+0xd2/0x2f0
    kswapd+0x2cb/0x6a0
    mem_cgroup_shrink_node+0x160/0x160
    kthread+0xbd/0xe0
    __switch_to+0x1fa/0x5c0
    ret_from_fork+0x1f/0x40
    kthread_create_on_node+0x180/0x180

a closer code inspection has shown that we might indeed miss all the
scheduling points in the reclaim path if no pages can be isolated from
the LRU list.  This is a pathological case but other reports from Donald
Buczek have shown that we might indeed hit such a path:

        clusterd-989   [009] .... 118023.654491: mm_vmscan_direct_reclaim_end: nr_reclaimed=193
         kswapd1-86    [001] dN.. 118023.987475: mm_vmscan_lru_isolate: isolate_mode=0 classzone=0 order=0 nr_requested=32 nr_scanned=4239830 nr_taken=0 file=1
         kswapd1-86    [001] dN.. 118024.320968: mm_vmscan_lru_isolate: isolate_mode=0 classzone=0 order=0 nr_requested=32 nr_scanned=4239844 nr_taken=0 file=1
         kswapd1-86    [001] dN.. 118024.654375: mm_vmscan_lru_isolate: isolate_mode=0 classzone=0 order=0 nr_requested=32 nr_scanned=4239858 nr_taken=0 file=1
         kswapd1-86    [001] dN.. 118024.987036: mm_vmscan_lru_isolate: isolate_mode=0 classzone=0 order=0 nr_requested=32 nr_scanned=4239872 nr_taken=0 file=1
         kswapd1-86    [001] dN.. 118025.319651: mm_vmscan_lru_isolate: isolate_mode=0 classzone=0 order=0 nr_requested=32 nr_scanned=4239886 nr_taken=0 file=1
         kswapd1-86    [001] dN.. 118025.652248: mm_vmscan_lru_isolate: isolate_mode=0 classzone=0 order=0 nr_requested=32 nr_scanned=4239900 nr_taken=0 file=1
         kswapd1-86    [001] dN.. 118025.984870: mm_vmscan_lru_isolate: isolate_mode=0 classzone=0 order=0 nr_requested=32 nr_scanned=4239914 nr_taken=0 file=1
  [...]
         kswapd1-86    [001] dN.. 118084.274403: mm_vmscan_lru_isolate: isolate_mode=0 classzone=0 order=0 nr_requested=32 nr_scanned=4241133 nr_taken=0 file=1

this is minute long snapshot which didn't take a single page from the
LRU.  It is not entirely clear why only 1303 pages have been scanned
during that time (maybe there was a heavy IRQ activity interfering).

In any case it looks like we can really hit long periods without
scheduling on non preemptive kernels so an explicit cond_resched() in
shrink_node_memcg which is independent on the reclaim operation is due.

Link: http://lkml.kernel.org/r/20161202095841.16648-1-mhocko@kernel.orgSigned-off-by: NMichal Hocko <mhocko@suse.com>
Reported-by: NBoris Zhmurov <bb@kernelpanic.ru>
Tested-by: NBoris Zhmurov <bb@kernelpanic.ru>
Reported-by: NDonald Buczek <buczek@molgen.mpg.de>
Reported-by: N"Christopher S. Aker" <caker@theshore.net>
Reported-by: NPaul Menzel <pmenzel@molgen.mpg.de>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Commit 0a6b76dd ("mm: workingset: make shadow node shrinker memcg
aware") has made the workingset shadow nodes shrinker memcg aware.  The
implementation is not correct though because memcg_kmem_enabled() might
become true while we are doing a global reclaim when the sc->memcg might
be NULL which is exactly what Marek has seen:

  BUG: unable to handle kernel NULL pointer dereference at 0000000000000400
  IP: [<ffffffff8122d520>] mem_cgroup_node_nr_lru_pages+0x20/0x40
  PGD 0
  Oops: 0000 [#1] SMP
  CPU: 0 PID: 60 Comm: kswapd0 Tainted: G           O   4.8.10-12.pvops.qubes.x86_64 #1
  task: ffff880011863b00 task.stack: ffff880011868000
  RIP: mem_cgroup_node_nr_lru_pages+0x20/0x40
  RSP: e02b:ffff88001186bc70  EFLAGS: 00010293
  RAX: 0000000000000000 RBX: ffff88001186bd20 RCX: 0000000000000002
  RDX: 000000000000000c RSI: 0000000000000000 RDI: 0000000000000000
  RBP: ffff88001186bc70 R08: 28f5c28f5c28f5c3 R09: 0000000000000000
  R10: 0000000000006c34 R11: 0000000000000333 R12: 00000000000001f6
  R13: ffffffff81c6f6a0 R14: 0000000000000000 R15: 0000000000000000
  FS:  0000000000000000(0000) GS:ffff880013c00000(0000) knlGS:ffff880013d00000
  CS:  e033 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: 0000000000000400 CR3: 00000000122f2000 CR4: 0000000000042660
  Call Trace:
    count_shadow_nodes+0x9a/0xa0
    shrink_slab.part.42+0x119/0x3e0
    shrink_node+0x22c/0x320
    kswapd+0x32c/0x700
    kthread+0xd8/0xf0
    ret_from_fork+0x1f/0x40
  Code: 66 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 3b 35 dd eb b1 00 55 48 89 e5 73 2c 89 d2 31 c9 31 c0 4c 63 ce 48 0f a3 ca 73 13 <4a> 8b b4 cf 00 04 00 00 41 89 c8 4a 03 84 c6 80 00 00 00 83 c1
  RIP  mem_cgroup_node_nr_lru_pages+0x20/0x40
   RSP <ffff88001186bc70>
  CR2: 0000000000000400
  ---[ end trace 100494b9edbdfc4d ]---

This patch fixes the issue by checking sc->memcg rather than
memcg_kmem_enabled() which is sufficient because shrink_slab makes sure
that only memcg aware shrinkers will get non-NULL memcgs and only if
memcg_kmem_enabled is true.

Fixes: 0a6b76dd ("mm: workingset: make shadow node shrinker memcg aware")
Link: http://lkml.kernel.org/r/20161201132156.21450-1-mhocko@kernel.orgSigned-off-by: NMichal Hocko <mhocko@suse.com>
Reported-by: NMarek Marczykowski-Górecki <marmarek@mimuw.edu.pl>
Tested-by: NMarek Marczykowski-Górecki <marmarek@mimuw.edu.pl>
Acked-by: NVladimir Davydov <vdavydov.dev@gmail.com>
Acked-by: NJohannes Weiner <hannes@cmpxchg.org>
Acked-by: NBalbir Singh <bsingharora@gmail.com>
Cc: <stable@vger.kernel.org>	[4.6+]
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>