Finally all the core gem and a lot of drivers are entirely free of
dev->struct_mutex depencies, and we can start to have an entirely
lockless unref path.
To make sure that no one who touches the core code accidentally breaks
existing drivers which still require dev->struct_mutex I've made the
might_lock check unconditional.
While at it de-inline the ref/unref functions, they've become a bit
too big.
v2: Make it not leak like a sieve.
v3: Review from Lucas:
- drop != NULL in pointer checks.
- fixup copypasted kerneldoc to actually match the functions.
v4:
Add __drm_gem_object_unreference as a fastpath helper for drivers who
abolished dev->struct_mutex, requested by Chris.
v5: Fix silly mistake in drm_gem_object_unreference_unlocked caught by
intel-gfx CI - I checked for gem_free_object instead of
gem_free_object_unlocked ...
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Alex Deucher <alexdeucher@gmail.com>
Cc: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de> (v3)
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> (v4)
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1462178451-1765-1-git-send-email-daniel.vetter@ffwll.ch
Here are a few amdkfd patches for 4.7, all of them fixes according to
the Coccinelle tool.
* tag 'drm-amdkfd-next-2016-05-04' of git://people.freedesktop.org/~gabbayo/linux:
amdkfd: Trim unnescessary intermediate err var in kfd_chardev.c
amdkfd: Trim off unnescessary semicolon from kfd_packet_manager.c
amdkfd: Use the canonical form in branch predicates
Here are some little fixes for rockchip drm, looks good for me, and seems there is no doubt on them, So I'd like you can land them.
* 'drm-rockchip-next-fixes-05-03' of https://github.com/markyzq/kernel-drm-rockchip:
drm/rockchip: vop: Initialize vskiplines to zero
drm/rockchip: vop: fix iommu crash with async atomic
drm/rockchip: support non-iommu buffer path
drm/rockchip: get rid of rockchip_drm_crtc_mode_config
drm/rockchip: inno_hdmi: fix an error code
drm/rockchip: don't leak iommu mapping
drm/rockchip: remove redundant statement
Summary:
- Support for pipeline clock between KMS drivers.
. Exynos SoC is required to control clocks across KMS drivers
according to Exynos SoC version. So this patch refactos
some relevant codes and provides generic solution for it.
- Add Exynos5433 SoC support to HDMI parts - HDMI and DECON-TV.
- Add HW trigger mode support to CRTC drivers.
. In case of using i80 Panel, some Exynos SoC supports HW trigger
mode so this patch makes trigger mode - HW or SW trigger - to be
set according to SoC version properly.
- And some cleanups and regression fixups.
* 'exynos-drm-next' of git://git.kernel.org/pub/scm/linux/kernel/git/daeinki/drm-exynos: (39 commits)
drm/exynos: clean up register definions for fimd and decon
drm/exynos: decon: clean up interface type
drm/exynos: fimd: add HW trigger support
drm/exynos: clean up wait_for_vblank
drm/exynos: mixer: use generic of_device_get_match_data helper
drm/exynos: mixer: remove support for non-dt platforms
drm/exynos: hdmi: use generic of_device_get_match_data helper
drm/exynos: rotator: use generic of_device_get_match_data helper
drm/exynos: fimd: use generic of_device_get_match_data helper
drm/exynos: dsi: use generic of_device_get_match_data helper
drm/exynos: exynos5433_decon: use generic of_device_get_match_data helper
drm/exynos: convert clock_enable crtc callback to pipeline clock
drm/exynos/mixer: enable HDMI-PHY before configuring MIXER
drm/exynos/decon5433: enable HDMI-PHY before configuring DECON
drm/exynos: add support for pipeline clock to the framework
drm/exynos: add helper to get crtc from pipe
drm/exynos/decon5433: do not protect window in plane disable
drm/exynos/decon5433: reset decon on start
drm/exynos/decon5433: fix DECON standalone update
drm/exynos/hdmi: remove registry dump
...
This pull request brings in DPI panel support, gamma ramp support, and
render nodes for vc4.
* tag 'drm-vc4-next-2016-05-02' of https://github.com/anholt/linux:
drm/vc4: Add missing render node support
drm/vc4: Add support for gamma ramps.
drm/vc4: Fix NULL deref in HDMI init error path
drm/vc4: Add DPI driver
drm: Add an encoder and connector type enum for DPI.
- prep work for struct_mutex-less gem_free_object
- more invasive/tricky mst fixes from Lyude for broken hw. I discussed
this with Ville/Jani and we all agreed more soaking in -next would be
real good this late in the -rc cycle. They're cc: stable too to make
sure they're not getting lost. Feel free to cherry-pick those four if
you disagree.
- few small things all over
* tag 'topic/drm-misc-2016-04-29' of git://anongit.freedesktop.org/drm-intel:
drm/atomic: Add missing drm_crtc_internal.h include
drm/dp: Allow signals to interrupt drm_aux-dev reads/writes
drm: Quiet down drm_mode_getresources
drm: Quiet down drm_mode_getconnector
drm: Protect dev->filelist with its own mutex
drm: Make drm_vm_open/close_locked private to drm_vm.c
drm: Hide master MAP cleanup in drm_bufs.c
drm: Forbid legacy MAP functions for DRIVER_MODESET
drm: Push struct_mutex into ->master_destroy
drm: Move drm_getmap into drm_bufs.c and give it a legacy prefix
drm: Put legacy lastclose work into drm_legacy_dev_reinit
drm: Give drm_agp_clear drm_legacy_ prefix
drm/sysfs: Annote lockless show functions with READ_ONCE
MAINTAINERS: Update the files list for the GMA500 DRM driver
drm: rcar-du: Fix compilation warning
drm/i915: Get rid of intel_dp_dpcd_read_wake()
drm/dp_helper: Perform throw-away read before actual read in drm_dp_dpcd_read()
drm/dp_helper: Retry aux transactions on all errors
drm/dp_helper: Always wait before retrying native aux transactions
- more userptr cornercase fixes from Chris
- clean up and tune forcewake handling (Tvrtko)
- more underrun fixes from Ville, mostly for ilk to appeas CI
- fix unclaimed register warnings on vlv/chv and enable the debug code to catch
them by default (Ville)
- skl gpu hang fixes for gt3/4 (Mika Kuoppala)
- edram improvements for gen9+ (Mika again)
- clean up gpu reset corner cases (Chris)
- fix ctx/ring machine deaths on snb/ilk (Chris)
- MOCS programming for all engines (Peter Antoine)
- robustify/clean up vlv/chv irq handler (Ville)
- split gen8+ irq handlers into ack/handle phase (Ville)
- tons of bxt rpm fixes (mostly around firmware interactions), from Imre
- hook up panel fitting for dsi panels (Ville)
- more runtime PM fixes all over from Imre
- shrinker polish (Chris)
- more guc fixes from Alex Dai and Dave Gordon
- tons of bugfixes and small polish all over (but with a big focus on bxt)
* tag 'drm-intel-next-2016-04-25' of git://anongit.freedesktop.org/drm-intel: (142 commits)
drm/i915: Update DRIVER_DATE to 20160425
drm/i915/bxt: Explicitly clear the Turbo control register
drm/i915: Correct the i915_frequency_info debugfs output
drm/i915: Macros to convert PM time interval values to microseconds
drm/i915: Make RPS EI/thresholds multiple of 25 on SNB-BDW
drm/i915: Fake HDMI live status
drm/i915/bxt: Force reprogramming a PHY with invalid HW state
drm/i915/bxt: Wait for PHY1 GRC done if PHY0 was already enabled
drm/i915/bxt: Use PHY0 GRC value for HW state verification
drm/i915: use dev_priv directly in gen8_ppgtt_notify_vgt
drm/i915/bxt: Enable DC5 during runtime resume
drm/i915/bxt: Sanitize DC state tracking during system resume
drm/i915/bxt: Don't uninit/init display core twice during system suspend/resume
drm/i915: Inline intel_suspend_complete
drm/i915/kbl: Don't WARN for expected secondary MISC IO power well request
drm/i915: Fix eDP low vswing for Broadwell
drm/i915: check for ERR_PTR from i915_gem_object_pin_map()
drm/i915/guc: local optimisations and updating comments
drm/i915/guc: drop cached copy of 'wq_head'
drm/i915/guc: keep GuC doorbell & process descriptor mapped in kernel
...
Update CDCLK_FREQ on BDW after changing the cdclk frequency. Not sure
if this is a late addition to the spec, or if I simply overlooked this
step when writing the original code.
This is what Bspec has to say about CDCLK_FREQ:
"Program this field to the CD clock frequency minus one. This is used to
generate a divided down clock for miscellaneous timers in display."
And the "Broadwell Sequences for Changing CD Clock Frequency" section
clarifies this further:
"For CD clock 337.5 MHz, program 337 decimal.
For CD clock 450 MHz, program 449 decimal.
For CD clock 540 MHz, program 539 decimal.
For CD clock 675 MHz, program 674 decimal."
Cc: stable@vger.kernel.org
Cc: Mika Kahola <mika.kahola@intel.com>
Fixes: b432e5cfd5 ("drm/i915: BDW clock change support")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461689194-6079-2-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Mika Kahola <mika.kahola@intel.com>
(cherry picked from commit 7f1052a8fa)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Without this there was a double free of the metadata,
which ended up freeing the fd table for me here, and taking
out the machine more often than not.
I reproduced with X.org + modesetting DDX + latest llvm/mesa,
also required using dri3.
Cc: stable@vger.kernel.org
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
When I was writing an atomic wrapper for rmfb, I ran into the
following backtrace from lockdep:
=============================================
[ INFO: possible recursive locking detected ]
4.5.0-patser+ #4696 Tainted: G U
---------------------------------------------
kworker/2:2/2608 is trying to acquire lock:
(crtc_ww_class_mutex){+.+.+.}, at: [<ffffffffc00c9ddc>] drm_modeset_lock+0x7c/0x120 [drm]
but task is already holding lock:
(crtc_ww_class_mutex){+.+.+.}, at: [<ffffffffc00c98cd>] modeset_backoff+0x8d/0x220 [drm]
other info that might help us debug this:
Possible unsafe locking scenario:
CPU0
----
lock(crtc_ww_class_mutex);
lock(crtc_ww_class_mutex);
*** DEADLOCK ***
May be due to missing lock nesting notation
4 locks held by kworker/2:2/2608:
#0: ("events"){.+.+.+}, at: [<ffffffff810a5eea>] process_one_work+0x15a/0x6c0
#1: ((&arg.work)){+.+.+.}, at: [<ffffffff810a5eea>] process_one_work+0x15a/0x6c0
#2: (crtc_ww_class_acquire){+.+.+.}, at: [<ffffffffc004532a>] drm_atomic_helper_remove_fb+0x4a/0x1d0 [drm_kms_helper]
#3: (crtc_ww_class_mutex){+.+.+.}, at: [<ffffffffc00c98cd>] modeset_backoff+0x8d/0x220 [drm]
While lockdep probably catches this bug when it happens, it's better
to explicitly warn when state->acquire_ctx is not set.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1462266751-29123-1-git-send-email-maarten.lankhorst@linux.intel.com
The fsl dcu now uses the clk-provider interfaces, which are not available
when CONFIG_COMMON_CLK is disabled:
drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c: In function 'fsl_dcu_drm_probe':
drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.c:362:20: error: implicit declaration of function '__clk_get_name' [-Werror=implicit-function-declaration]
pix_clk_in_name = __clk_get_name(pix_clk_in);
This adds a Kconfig dependency to prevent the driver from being enabled
in this case.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: 2d701449bc ("drm/fsl-dcu: use common clock framework for pixel clock divider")
Acked-by: Stefan Agner <stefan@agner.ch>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1462186839-2224021-1-git-send-email-arnd@arndb.de
After async atomic_commit callback, drm_atomic_clean_old_fb will
clean all old fb, but because async, the old fb may be also on
the vop hardware, dma will access the old fb buffer, clean old
fb will cause iommu page fault.
Reference the fb and unreference it when the fb actuall swap out
from vop hardware.
Signed-off-by: Mark Yao <mark.yao@rock-chips.com>
Some rockchip vop not support iommu, need use non-iommu
buffer for it. And if we get iommu issues, we can compare
the issues with non-iommu path, that would help the debug.
Signed-off-by: Mark Yao <mark.yao@rock-chips.com>
We need to take care of the vop status when use
rockchip_drm_crtc_mode_config, if vop is disabled,
the function would failed, that is terrible.
Save output_type and output_mode into rockchip_crtc_state,
it's nice to make them into atomic.
Signed-off-by: Mark Yao <mark.yao@rock-chips.com>
Tested-by: John Keeping <john@metanate.com>
We were accidentally returning PTR_ERR(NULL) which means success when we
wanted to return a negative error code.
Fixes: 412d4ae6b7 ('drm/rockchip: hdmi: add Innosilicon HDMI support')
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Yakir Yang <ykk@rock-chips.com>
arm_iommu_attach_device() takes its own reference to the mapping we give
it. Since we do not keep a reference to the mapping ourselves, we must
release it before returning.
Also fix the error path, which fails to release the mapping if it has
called arm_iommu_detach_device() since that clears archdata.mapping.
Signed-off-by: John Keeping <john@metanate.com>
The call to arm_iommu_detach_device() on the previous line sets
dev->archdata.mapping to NULL so this call is always a no-op.
Signed-off-by: John Keeping <john@metanate.com>
There shouldn't be any other driver support necessary, since none of
the driver-specific ioctls ever required auth, and none of them deal
with modesetting.
Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We could possibly save a bit of power by not requesting gamma
conversion when the ramp happens to be 1:1, but at least if all the
CRTCs are off the SRAM will be disabled.
This should fix brightness sliders in a lot of fullscreen games.
Signed-off-by: Eric Anholt <eric@anholt.net>
This patch adds support for software clock gating to UVD 5
and UVD 6 blocks with a preliminary commented out hardware
gating routine.
Currently hardware gating does not work so it's not activated.
Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The values of all but the RADEON_HPD_NONE members of the radeon_hpd_id
enum transform 1:1 into bit positions within the 'enabled' bitset as
assembled by evergreen_hpd_init():
enabled |= 1 << radeon_connector->hpd.hpd;
However, if ->hpd.hpd happens to equal RADEON_HPD_NONE == 0xff, UBSAN
reports
UBSAN: Undefined behaviour in drivers/gpu/drm/radeon/evergreen.c:1867:16
shift exponent 255 is too large for 32-bit type 'int'
[...]
Call Trace:
[<ffffffff818c4d35>] dump_stack+0xbc/0x117
[<ffffffff818c4c79>] ? _atomic_dec_and_lock+0x169/0x169
[<ffffffff819411bb>] ubsan_epilogue+0xd/0x4e
[<ffffffff81941cbc>] __ubsan_handle_shift_out_of_bounds+0x1fb/0x254
[<ffffffffa0ba7f2e>] ? atom_execute_table+0x3e/0x50 [radeon]
[<ffffffff81941ac1>] ? __ubsan_handle_load_invalid_value+0x158/0x158
[<ffffffffa0b87700>] ? radeon_get_pll_use_mask+0x130/0x130 [radeon]
[<ffffffff81219930>] ? wake_up_klogd_work_func+0x60/0x60
[<ffffffff8121a35e>] ? vprintk_default+0x3e/0x60
[<ffffffffa0c603c4>] evergreen_hpd_init+0x274/0x2d0 [radeon]
[<ffffffffa0c603c4>] ? evergreen_hpd_init+0x274/0x2d0 [radeon]
[<ffffffffa0bd196e>] radeon_modeset_init+0x8ce/0x18d0 [radeon]
[<ffffffffa0b71d86>] radeon_driver_load_kms+0x186/0x350 [radeon]
[<ffffffffa03b6b16>] drm_dev_register+0xc6/0x100 [drm]
[<ffffffffa03bc8c4>] drm_get_pci_dev+0xe4/0x490 [drm]
[<ffffffff814b83f0>] ? kfree+0x220/0x370
[<ffffffffa0b687c2>] radeon_pci_probe+0x112/0x140 [radeon]
[...]
=====================================================================
radeon 0000:01:00.0: No connectors reported connected with modes
At least on x86, there should be no user-visible impact as there
1 << 0xff == 1 << (0xff & 31) == 1 << 31
holds and 31 > RADEON_MAX_HPD_PINS. Thus, this patch is a cosmetic one.
All of the above applies analogously to evergreen_hpd_fini(),
r100_hpd_init(), r100_hpd_fini(), r600_hpd_init(), r600_hpd_fini(),
rs600_hpd_init() and rs600_hpd_fini()
Silence UBSAN by checking ->hpd.hpd for RADEON_HPD_NONE before oring it
into the 'enabled' bitset in the *_init()- or the 'disabled' bitset in
the *_fini()-functions respectively.
Signed-off-by: Nicolai Stange <nicstange@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
this is to fix fatal page fault error that occured if:
job is signaled/released after its timeout work is already
put to the global queue (in this case the cancel_delayed_work
will return false), which will lead to NX-protection error
page fault during job_timeout_func.
Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add two callbacks to scheduler to maintain jobs, and invoked for
job timeout calculations. Now TDR measures time gap from
job is processed by hw.
v2:
fix typo
Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
original time out detect routine is incorrect, cuz it measures
the gap from job scheduled, but we should only measure the
gap from processed by hw.
Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
the mirror_list will be used for later time out detect
feature. This is needed to properly detect a GPU
timeout with the scheduler.
Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
for those jobs submitted through scheduler, do not
free it immediately after scheduled, instead free it
in global workqueue by its sched fence signaling
callback function.
v2:
call uf's bo_undef after job_run()
call job's sync free after job_run()
no static inline __amdgpu_job_free() anymore, just use
kfree(job) to replace it.
Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Consolidate job initialization in one place rather than
duplicating it in multiple places.
Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
