Only call synchronize_rcu_expedited after unlocking struct_mutex to
avoid deadlock because the workqueues depend on struct_mutex.
>From original patch by Andrea:
synchronize_rcu/synchronize_sched/synchronize_rcu_expedited() will
hang until its own workqueues are run. The i915 gem workqueues will
wait on the struct_mutex to be released. So we cannot wait for a
quiescent state using those rcu primitives while holding the
struct_mutex or it creates a circular lock dependency resulting in
kernel hangs (which is reproducible but goes undetected by lockdep).
kswapd0 D 0 700 2 0x00000000
Call Trace:
? __schedule+0x1a5/0x660
? schedule+0x36/0x80
? _synchronize_rcu_expedited.constprop.65+0x2ef/0x300
? wake_up_bit+0x20/0x20
? rcu_stall_kick_kthreads.part.54+0xc0/0xc0
? rcu_exp_wait_wake+0x530/0x530
? i915_gem_shrink+0x34b/0x4b0
? i915_gem_shrinker_scan+0x7c/0x90
? i915_gem_shrinker_scan+0x7c/0x90
? shrink_slab.part.61.constprop.72+0x1c1/0x3a0
? shrink_zone+0x154/0x160
? kswapd+0x40a/0x720
? kthread+0xf4/0x130
? try_to_free_pages+0x450/0x450
? kthread_create_on_node+0x40/0x40
? ret_from_fork+0x23/0x30
plasmashell D 0 4657 4614 0x00000000
Call Trace:
? __schedule+0x1a5/0x660
? schedule+0x36/0x80
? schedule_preempt_disabled+0xe/0x10
? __mutex_lock.isra.4+0x1c9/0x790
? i915_gem_close_object+0x26/0xc0
? i915_gem_close_object+0x26/0xc0
? drm_gem_object_release_handle+0x48/0x90
? drm_gem_handle_delete+0x50/0x80
? drm_ioctl+0x1fa/0x420
? drm_gem_handle_create+0x40/0x40
? pipe_write+0x391/0x410
? __vfs_write+0xc6/0x120
? do_vfs_ioctl+0x8b/0x5d0
? SyS_ioctl+0x3b/0x70
? entry_SYSCALL_64_fastpath+0x13/0x94
kworker/0:0 D 0 29186 2 0x00000000
Workqueue: events __i915_gem_free_work
Call Trace:
? __schedule+0x1a5/0x660
? schedule+0x36/0x80
? schedule_preempt_disabled+0xe/0x10
? __mutex_lock.isra.4+0x1c9/0x790
? del_timer_sync+0x44/0x50
? update_curr+0x57/0x110
? __i915_gem_free_objects+0x31/0x300
? __i915_gem_free_objects+0x31/0x300
? __i915_gem_free_work+0x2d/0x40
? process_one_work+0x13a/0x3b0
? worker_thread+0x4a/0x460
? kthread+0xf4/0x130
? process_one_work+0x3b0/0x3b0
? kthread_create_on_node+0x40/0x40
? ret_from_fork+0x23/0x30
Fixes: 3d3d18f086 ("drm/i915: Avoid rcu_barrier() from reclaim paths (shrinker)")
Reported-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
(cherry picked from commit 8f612d0551)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Last drm-misc-next pull req for 4.12
Core changes:
- fb_helper checkpatch cleanup and simplified _add_one_connector() (Thierry)
- drm_ioctl and drm_sysfs improved/gained documentation (Daniel)
- [ABI] Repurpose reserved field in drm_event_vblank for crtc_id (Ander)
- Plumb acquire ctx through legacy paths to avoid lock_all and legacy_backoff
(Daniel)
- Add connector_atomic_check to check conn constraints on modeset (Maarten)
- Add drm_of_find_panel_or_bridge to remove boilerplate in drivers (Rob)
Driver changes:
- meson moved to drm-misc (Neil)
- Added support for Amlogic GX SoCs in dw-hdmi (Neil)
- Rockchip unbind actually cleans up the things bind initializes (Jeffy)
- A couple misc fixes in virtio, dw-hdmi
NOTE: this also includes a backmerge of drm-next as well rc5 (we needed vmwgfx
as well as the new synopsys media formats)
* tag 'drm-misc-next-2017-04-07' of git://anongit.freedesktop.org/git/drm-misc: (77 commits)
Revert "drm: Don't allow interruptions when opening debugfs/crc"
drm: Only take cursor locks when the cursor plane exists
drm/vmwgfx: Fix fbdev emulation using legacy functions
drm/rockchip: Shutdown all crtcs when unbinding drm
drm/rockchip: Reorder drm bind/unbind sequence
drm/rockchip: analogix_dp: Disable clock when unbinding
drm/rockchip: vop: Unprepare clocks when unbinding
drm/rockchip: vop: Enable pm domain before vop_initial
drm/rockchip: cdn-dp: Don't unregister audio dev when unbinding
drm/rockchip: cdn-dp: Don't try to release firmware when not loaded
drm: bridge: analogix: Destroy connector & encoder when unbinding
drm: bridge: analogix: Disable clock when unbinding
drm: bridge: analogix: Unregister dp aux when unbinding
drm: bridge: analogix: Detach panel when unbinding analogix dp
drm: Don't allow interruptions when opening debugfs/crc
drm/virtio: don't leak bo on drm_gem_object_init failure
drm: bridge: dw-hdmi: fix input format/encoding from plat_data
drm: omap: use common OF graph helpers
drm: convert drivers to use drm_of_find_panel_or_bridge
drm: convert drivers to use of_graph_get_remote_node
...
Last 4.12 feature pile:
GVT updates:
- Add mdev attribute group for per-vgpu info
- Time slice based vGPU scheduling QoS support (Gao Ping)
- Initial KBL support for E3 server (Han Xu)
- other misc.
i915:
- lots and lots of small fixes and improvements all over
- refactor fw_domain code (Chris Wilson)
- improve guc code (Oscar Mateo)
- refactor cursor/sprite code, precompute more for less overhead in
the critical path (Ville)
- refactor guc/huc fw loading code a bit (Michal Wajdeczko)
* tag 'drm-intel-testing-2017-04-03' of git://anongit.freedesktop.org/git/drm-intel: (121 commits)
drm/i915: Update DRIVER_DATE to 20170403
drm/i915: Clear gt.active_requests before checking idle status
drm/i915/uc: Drop use of MISSING_CASE on trivial enums
drm/i915: make a few DDI functions static
drm/i915: Combine reset_all_global_seqno() loops into one
drm/i915: Remove redudant wait for each engine to idle from seqno wrap
drm/i915: Wait for all engines to be idle as part of i915_gem_wait_for_idle()
drm/i915: Move retire-requests into i915_gem_wait_for_idle()
drm/i915/uc: Move fw path check to fetch_uc_fw()
drm/i915/huc: Remove unused intel_huc_fini()
drm/i915/uc: Add intel_uc_fw_fini()
drm/i915/uc: Add intel_uc_fw_type_repr()
drm/i915/uc: Move intel_uc_fw_status_repr() to intel_uc.h
drivers: gpu: drm: i915L intel_lpe_audio: Fix kerneldoc comments
drm/i915: Suppress busy status for engines if wedged
drm/i915: Do request retirement before marking engines as wedged
drm/i915: Drop verbose and archaic "ring" from our internal engine names
drm/i915: Use a dummy timeline name for a signaled fence
drm/i915: Ironlake do_idle_maps w/a may be called w/o struct_mutex
drm/i915/guc: Take enable_guc_loading check out of GEM core code
...
mode_valid() called from drm_helper_probe_single_connector_modes()
may need to look at connector->state because what a valid mode is may
depend on connector properties being set. For example some HDMI modes
might be rejected when a connector property forces the connector
into DVI mode.
Some implementations of detect() already lock all state,
so we have to pass an acquire_ctx to them to prevent a deadlock.
This means changing the function signature of detect() slightly,
and passing the acquire_ctx for locking multiple crtc's.
For the callbacks, it will always be non-zero. To allow callers
not to worry about this, drm_helper_probe_detect_ctx is added
which might handle -EDEADLK for you.
Changes since v1:
- Always set ctx parameter.
Changes since v2:
- Always take connection_mutex when probing.
Changes since v3:
- Remove the ctx from intel_dp_long_pulse, and add
WARN_ON(!connection_mutex) (danvet)
- Update docs to clarify the locking situation. (danvet)
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Boris Brezillon <boris.brezillon@free-electrons.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1491504920-4017-1-git-send-email-maarten.lankhorst@linux.intel.com
It's too chatty to have three places to tell us which one
is next vgpu for schedule. My log file was bloated to eat
all disk space..
Cc: Ping Gao <ping.a.gao@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
It just doesn't work. It probably stopped working way, way before that
(e.g. i915 grabbed random mutexes all over in modeset code at least
since gen6), but with atomic and all the ww_mutex stuff it's indeed
hopeless.
Remove ->mode_set_base_atomic from the 2 atomic drivers (i915 and
nouveau) that still had one (both had dummy implementations already
anyway), and shunt atomic drivers in the helpers debug_enter/leave
functions.
I'll leave the code in for radeon and amdgpu, but I think as soon as
amdgpu is atomic we should think about just ripping it out. Only
having it around for radeon and pre-nv50 is rather pointless. This
would also allow us to nuke all that code from fbdev.
Funny part is that _all_ kms drivers set this hook, despite that no
one else provides the required ->mode_set_base_atomic implementation.
The reason I'm jumping on this is that I want to wire up a full
acquire ctx for the benefit of atomic drivers, everywhere. And the
debug_enter/leave implementations call ->gamma_set. And there's just
no way ever we can create an acquire_ctx in the nmi context of kgdb.
Cc: Ben Skeggs <bskeggs@redhat.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170403083304.9083-11-daniel.vetter@ffwll.ch
Fix wrong initial csb read pointer value. This fixes the random
engine timeout issue in guest when guest boots up.
Fixes: 8453d674ae ("drm/i915/gvt: vGPU execlist virtualization")
Cc: stable@vger.kernel.org # v4.10+
Signed-off-by: Min He <min.he@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
drm-misc for 4.12:
Core:
- Removed some fb subsampling dimension checks from core (Ville)
- Some MST slot cleanup (Dhinakaran)
- Extracted drm_debugfs.h & drm_ioctl.h from drmP.h (Daniel)
- Added drm_atomic_helper_shutdown() to compliment suspend/resume counterparts
(Daniel)
- Pipe context through legacy modeset to remove legacy_backoff nasties (Daniel)
- Cleanups around vblank as well as allowing lockless counter reads (Chris W.)
- VGA Switcheroo added to MAINTAINERS with Lukas Wunner as reviewer (Lukas)
Drivers:
- Enhancements to rockchip driver probe (Jeffy) and dsi (Chris Z.)
- Thunderbolt external GPU awareness added (Lukas)
* tag 'drm-misc-next-2017-03-31' of git://anongit.freedesktop.org/git/drm-misc: (63 commits)
apple-gmux: Don't switch external DP port on 2011+ MacBook Pros
drm/nouveau: Don't register Thunderbolt eGPU with vga_switcheroo
drm/amdgpu: Don't register Thunderbolt eGPU with vga_switcheroo
drm/radeon: Don't register Thunderbolt eGPU with vga_switcheroo
PCI: Recognize Thunderbolt devices
MAINTAINERS: Add Lukas Wunner as reviewer for vga_switcheroo
drm: Fix locking gotcha in page_flip ioctl
drm: Clarify the role of plane_state argument to drm_simple update().
drm: Clear e after kfree in drm_mode_page_flip_ioctl
drm: Convert cmpxchg(bool) back to a two step operation
drm/bridge: ti-tfp410: support hpd via gpio
drm: use .hword to represent 16-bit numbers
Revert unrelated part of "drm: simplify the locking in the GETCRTC ioctl"
drm: Fixup failure paths in drm_atomic_helper_set_config
drm: Peek at the current counter/timestamp for vblank queries
drm: Refactor vblank sequence number comparison
drm: vblank cannot be enabled if dev->irq_enabled is false
drm: Mark up accesses of vblank->enabled outside of its spinlock
drm: Make the decision to keep vblank irq enabled earlier
drm/atomic-helper: Remove the backoff hack from set_config
...
Firmware loading interface for GVT-g golden HW state has been broken
before. This patch fixes GVT-g firmware loading interface. A user should
apply this patch if he wants to load GVT-g golden HW state from firmware
interface.
Fixes: 579cea5 ("drm/i915/gvt: golden virtual HW state management")
Cc: Zhenyu Wang <zhenyuw@linux.intel.com>
Cc: drm-intel-fixes@lists.freedesktop.org
Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Make i915_gem_wait_for_idle() be a little heavier in order to try and
guarantee that the GPU is indeed idle (by checking each engine
individually is idle, i.e. all writes are complete and the rings
stopped) after waiting for in-flight requests to be completed.
v2: And return the final error.
v3: Break the wait_for() out from under the WARN -- the macro expansion
is hideous and unreadable in the warning message
v4: If wait_for_engine() fails the result is catastrophic, mark the
device as wedged and wait for the repair team.
References: https://bugs.freedesktop.org/show_bug.cgi?id=98836
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170330145041.9005-4-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Michał Winiarski pointed out that the debugging infrastructure (such as
trace_dma_fence_release) likes to pretty print the timeline name, long
after we have freed the timeline. Our timelines currently live as part of
the GTT (due to the strict ordering we currently use through each) which
belong to the context. We aim to free the context and release its
hardware resources as soon as we able to (i.e. when the last
fence/request using it has been signaled and retired). As the
.get_timeline_name is purely a debug feature, rather than extending the
lifetime of the context, or splitting it into many different release
phases just to keep the name around, replace the timeline name with a
constant after the fence has been signaled. This avoids the potential
use-after-free.
Reported-by: Krzysztof Olinski <krzysztof.e.olinski@intel.com>
Fixes: 80b204bce8 ("drm/i915: Enable multiple timelines")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Michał Winiarski <michal.winiarski@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: <stable@vger.kernel.org> # v4.10+
Link: http://patchwork.freedesktop.org/patch/msgid/20170330111614.29757-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>
(cherry picked from commit 05506b5be0)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
The variable info is never NULL, which is checked by the caller. This
patch removes the redundant info NULL check logic.
Fixes: 695fbc08d8 ("drm/i915/gvt: replace the gvt_err with gvt_vgpu_err")
Signed-off-by: Tina Zhang <tina.zhang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
(cherry picked from commit 865f03d42e)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
From commit d1a513be1f ("drm/i915/gvt: add resolution definition for vGPU
type"), small type has been restricted to small resolution, so not
require larger high GM size any more. Change to smaller 384M for more
VM creation with vGPU enabled which still perform reasonable workload.
Fixes: d1a513be1f ("drm/i915/gvt: add resolution definition for vGPU type")
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
(cherry picked from commit bf39ec335e)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Unlocking is dangerous. In this case we combine an early update to the
out-of-queue request, because we know that it will be inserted into the
correct FIFO priority-ordered slot when it becomes ready in the future.
However, given sufficient enthusiasm, it may become ready as we are
continuing to reschedule, and so may gazump the FIFO if we have since
dropped its spinlock. The result is that it may be executed too early,
before its dependencies.
v2: Move all work into the second phase over the topological sort. This
removes the shortcut on the out-of-rbtree request to ensure that we only
adjust its priority after adjusting all of its dependencies.
Fixes: 20311bd350 ("drm/i915/scheduler: Execute requests in order of priorities")
Testcase: igt/gem_exec_whisper
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: <stable@vger.kernel.org> # v4.10+
Link: http://patchwork.freedesktop.org/patch/msgid/20170327202143.7972-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
(cherry picked from commit a79a524e92)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
We pretty print the name of an engine in several places, mostly for
debug, but also in the GPU hang report. Using "ring" in the name is
archaic (we call those engines now to differentiate them from the
multiple rings of commands we execute on each engine), quite verbose and
often tautological. We run out of room in our GPU hang report for
instance if we have more than a couple of engines hung simultaneously.
Bit the bullet and update the strings to reflect the common internal names.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170330134820.12273-1-chris@chris-wilson.co.uk
Michał Winiarski pointed out that the debugging infrastructure (such as
trace_dma_fence_release) likes to pretty print the timeline name, long
after we have freed the timeline. Our timelines currently live as part of
the GTT (due to the strict ordering we currently use through each) which
belong to the context. We aim to free the context and release its
hardware resources as soon as we able to (i.e. when the last
fence/request using it has been signaled and retired). As the
.get_timeline_name is purely a debug feature, rather than extending the
lifetime of the context, or splitting it into many different release
phases just to keep the name around, replace the timeline name with a
constant after the fence has been signaled. This avoids the potential
use-after-free.
Reported-by: Krzysztof Olinski <krzysztof.e.olinski@intel.com>
Fixes: 80b204bce8 ("drm/i915: Enable multiple timelines")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Michał Winiarski <michal.winiarski@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: <stable@vger.kernel.org> # v4.10+
Link: http://patchwork.freedesktop.org/patch/msgid/20170330111614.29757-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>
When test GVTg as below scenario:
VM boot --> failsafe --> kill qemu --> VM boot.
Qemu report error at the second boot:
ERROR: PCI region size must be pow2 type=0x0, size=0x1fa1000
Qemu need access PCI_ROM_ADDRESS reg to determine the size of expansion
PCI rom. The mechanism just like the BAR reg (write-read) and we should
return the size 0 since we have no rom. If we reject the write to
PCI_ROM_ADDRESS, Qemu cannot get the correct size of rom.
Essentially, GVTg failsafe mode should not break PCI function. So we
exclude cfg space from failsafe mode. This can fix above issue.
v2: add Fixes and Bugzilla link.
Fixes: fd64be6367 ("drm/i915/gvt: introduced failsafe mode into vgpu")
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=100296
Signed-off-by: Changbin Du <changbin.du@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
This patch introduces two functions for activating/de-activating vGPU in
mdev ops.
A racing condition was found between virtual vblank emulation and KVGMT
mdev release path. V-blank emulation will emulate and inject V-blank
interrupt for every active vGPU with holding gvt->lock, while in mdev
release path, it will directly release hypervisor handle without changing
vGPU status or taking gvt->lock, so a kernel oops is encountered when
vblank emulation is injecting a interrupt with a invalid hypervisor
handle. (Reported by Terrence)
To solve this problem, we factor out vGPU activation/de-activation from
vGPU creation/destruction path and let KVMGT mdev release ops de-activate
the vGPU before release hypervisor handle. Once a vGPU is de-activated,
GVT-g will not emulate v-blank for it or touch the hypervisor handle.
Fixes: 659643f ("drm/i915/gvt/kvmgt: add vfio/mdev support to KVMGT")
Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
The timeslice usage will determine vGPU whether has chance to
schedule or not at every vGPU switch checkpoint.
Signed-off-by: Ping Gao <ping.a.gao@intel.com>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
vGPU resource is allocated by scheduler. To account for non-allocated
free cycles, we create an idle vGPU as the placeholder similar to idle task
concept, which is useful to handle some corner cases in scheduling policy.
Signed-off-by: Ping Gao <ping.a.gao@intel.com>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
