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>
When we retire the last request on the ring, before we ever access that
ring again we know it will be completely idle and so we can advance the
ring->head fully to the end (i.e. ring->tail) and not just to the start
of the breadcrumb. This allows us to skip re-emitting the breadcrumb
after resetting the GPU if the ring was entirely idle. This prevents us
from overwriting a seqno wraparound by re-executing a stale breadcrumb,
i.e.
submit_request(1)
intel_engine_init_global_seqno(0)
i915_reset()
would then leave 1 in the HWS, but the next request to execute would
also be with seqno 1. The sanity checks upon submission detect this as a
timewarp and explode. By setting the ring as empty, upon reset the HWS
is left as 0, leaving it consistent with the timeline.
v2: Fix check for deleting last element of list. We know that this
request is always the first element of the ring, so only if next
points back to the start will this be the only request in flight.
v3: Remove opencoding of list_is_last()
v4: Move the block to its own function for some clarity.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=100144
Testcase: igt/gem_exec_whisper/hang-*
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170406170028.26871-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
The plat_data->input_bus_format and plat_data->input_bus_encoding
are unsigned long and are always >=0, but the value 0 was still
considered as RGB888 for input_bus_format and default color space
for input_bus_encoding in the reworked code.
This patch changes the if statement check for a non-zero value to
either use the default input bus_format and/or bus_encoding for a zero
value and the provided bus_format and/or bus_encoding for a
non zero value.
Thanks to Dan Carpenter for his bug report at [1].
Tested on Amlogic P230 (with CSC enabled for YUV444 to RGB) and Rockchip
RK3288 ACT8846 EVB Board (no CSC involved, direct RGB passthrough).
[1] http://lkml.kernel.org/r/20170406052120.GA26578@mwanda
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Fixes: def23aa7e9 ("drm: bridge: dw-hdmi: Switch to V4L bus format and encodings")
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Reviewed-by: Archit Taneja <architt@codeaurora.org>
[narmstrong@baylibre.com: reworded commit message and added Fixes tag]
Link: http://patchwork.freedesktop.org/patch/msgid/1491471244-24989-1-git-send-email-narmstrong@baylibre.com
This adds support for the Winstar Display Co. WF35LTIACD 3.5" QVGA TFT
LCD panel, which can be supported by the simple panel driver.
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Richard Genoud <richard.genoud@gmail.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
The Sitronix ST7789v controller is used to drive 240x320 LCD panels through
various interfaces, including SPI and RGB/Parallel.
The current driver is configuring it for the latter. Support for tinyDRM
can always be added later.
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Add simple-panel support for the Ampire AM-480272H3TMQW-T01H, which is a
4.3" WQVGA panel.
Signed-off-by: Yannick Fertre <yannick.fertre@st.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
The OMAP driver has its own OF graph helpers that are similar to the
common helpers. This commit replaces most of the calls with the common
helpers. There's still a couple of custom helpers left, but the driver
needs more extensive changes to get rid of them.
In dss_init_ports, we invert the loop, looping through the known ports
and matching them to DT nodes rather than looping thru DT nodes and
matching them to the ports.
Signed-off-by: Rob Herring <robh@kernel.org>
Signed-off-by: Sean Paul <seanpaul@chromium.org>
Similar to the previous commit, convert drivers open coding OF graph
parsing to use drm_of_find_panel_or_bridge instead.
This changes some error messages to debug messages (in the graph core).
Graph connections are often "no connects" depending on the particular
board, so we want to avoid spurious messages. Plus the kernel is not a
DT validator.
Signed-off-by: Rob Herring <robh@kernel.org>
Reviewed-by: Archit Taneja <architt@codeaurora.org>
Tested-by: Philipp Zabel <p.zabel@pengutronix.de>
Acked-by: Maxime Ripard <maxime.ripard@free-electrons.com>
[seanpaul dropped rockchip changes since they're now obsolete]
Signed-off-by: Sean Paul <seanpaul@chromium.org>
Convert drivers to use the new of_graph_get_remote_node() helper
instead of parsing the endpoint node and then getting the remote device
node. Now drivers can just specify the device node and which
port/endpoint and get back the connected remote device node. The details
of the graph binding are nicely abstracted into the core OF graph code.
This changes some error messages to debug messages (in the graph core).
Graph connections are often "no connects" depending on the particular
board, so we want to avoid spurious messages. Plus the kernel is not a
DT validator.
Signed-off-by: Rob Herring <robh@kernel.org>
Acked-by: Neil Armstrong <narmstrong@baylibre.com>
Tested-by: Liviu Dudau <liviu.dudau@arm.com>
Tested-by: Eric Anholt <eric@anholt.net>
Tested-by: Jyri Sarha <jsarha@ti.com>
Tested by: Archit Taneja <architt@codeaurora.org>
Signed-off-by: Sean Paul <seanpaul@chromium.org>
Many drivers have a common pattern of searching the OF graph for either an
attached panel or bridge and then finding the DRM struct for the panel
or bridge. Also, most drivers need to handle deferred probing when the
DRM device is not yet instantiated. Create a common function,
drm_of_find_panel_or_bridge, to find the connected node and the
associated DRM panel or bridge device.
Signed-off-by: Rob Herring <robh@kernel.org>
Acked-by: Philipp Zabel <p.zabel@pengutronix.de>
[seanpaul dropped extern from drm_of.h]
Signed-off-by: Sean Paul <seanpaul@chromium.org>
Currently we use a flag to change behavior in atomic commit
whether a conflicting encoder should be enabled or disabled.
This is used for the legacy set_config helper, which disables
connectors that have a conflicting encoder but not part of the
active crtc list.
There's no need for this to be handled in atomic commit, it
could be done in the set_config helper instead. This will
let the atomic check function reject any conflicting encoders,
while set_config can disable conflicting crtc's. This makes it
possible to recalculate the changed flags in 1 loop.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Sean Paul <seanpaul@chromium.org>
Link: http://patchwork.freedesktop.org/patch/msgid/1491477543-31257-2-git-send-email-maarten.lankhorst@linux.intel.com
A few more things for 4.12:
- ttm and amdgpu support for non-contiguous vram CPU mappings
- lots of bug fixes and cleanups for vega10
- misc bug fixes and code cleanups
[airlied: fix do_div error on 32-bit arm, not sure it's 100% correct]
* 'drm-next-4.12' of git://people.freedesktop.org/~agd5f/linux: (58 commits)
drm/amdgpu: use uintptr_t instead of unsigned long to store pointer
drm/amdgpu: Avoid using signed integer to store pointer value
drm/amdgpu:invoke new implemented AI MB func
drm/amdgpu/vega10:timeout set to equal with VI
drm/amdgpu:implement the reset MB func for vega10
drm/amdgpu:fix typo for mxgpu_ai
drm/amdgpu:no need to involv HDP in KIQ
drm/amdgpu:add PSP block only load_type=PSP (v2)
drm/amdgpu/smu9: update to latest driver interface
drm/amd/amdgpu: cleanup gfx_v9_0_gpu_init()
drm/amd/amdgpu: cleanup gfx_v9_0_rlc_reset()
drm/amd/amdgpu: cleanup gfx_v9_0_rlc_start()
drm/amd/amdgpu: simplify gfx_v9_0_cp_gfx_enable()
drm/amd/amdgpu: cleanup gfx_v9_0_kiq_init_register()
drm/amd/amdgpu: Drop gfx_v9_0_print_status()
drm/amd/amdgpu: cleanup gfx_v9_0_set_gfx_eop_interrupt_state()
drm/amd/amdgpu: cleanup gfx_v9_0_set_priv_reg_fault_state()
drm/amd/amdgpu: cleanup gfx_v9_0_set_priv_inst_fault_state()
drm/amd/amdgpu: cleanup gfx_v9_0_init_queue()
drm/amdgpu: Move function amdgpu_has_atpx near other similar functions
...
Highlights:
- Cooling device support from Russell, to allow GPU throttling on system
thermal overload.
- Explicit fencing support from Philipp, implemented in a similar way to
drm/msm.
* 'etnaviv/next' of https://git.pengutronix.de/git/lst/linux:
drm/etnaviv: submit support for out-fences
drm/etnaviv: return GPU fence through the submit structure
drm/etnaviv: submit support for in-fences
drm/etnaviv: add etnaviv cooling device
drm/etnaviv: switch to postclose
drm/etnaviv: add lockdep assert to fence allocation
imx-drm module/dependency changes
- The PRE/PRG drivers added an unwanted DRM dependency to the ipu-v3 driver.
Remove the dependency by conditionally disabling PRE/PRG support depending
on CONFIG_DRM.
- Merge the imx-ipuv3-crtc module into the imxdrm module. There is no reason
anymore for a separation between core drm driver and crtc/plane drivers,
especially since commit eb8c88808c ("drm/imx: add deferred plane
disabling"), which added a dependency on imx-ipuv3-crtc to the imxdrm
module.
* tag 'imx-drm-next-2017-04-04' of git://git.pengutronix.de/git/pza/linux:
drm/imx: merge imx-drm-core and ipuv3-crtc in one module
gpu: ipu-v3: don't depend on DRM being enabled
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
1,KIQ won't touch VRAM so no need to involv HDP flush/invalidate at all.
2,According to CP hw designer KIQ better not use any PM4 package lead to wait behave.
Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
SRIOV currently only can load ucode directly, and PSP
block is not supported by VF temporarily.
will remove this restrict and use PSP load all ucode
even for SRIOV later
v2: squash in check against module parameter
Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
