Track the list of in-flight submits. If the gpu hangs, retire up to an
including the offending submit, and then re-submit the remainder. This
way, for concurrently running piglit tests (for example), one failing
test doesn't cause unrelated tests to fail simply because it's submit
was queued up after one that triggered a hang.
Signed-off-by: Rob Clark <robdclark@gmail.com>
The 'timeout' value comes from userspace (CLOCK_MONOTONIC), but
converting this directly to jiffies doesn't take into account the
initial jiffies count at boot, which may differ from the base time
of CLOCK_MONOTONIC.
TODO: add ktime_delta_jiffies() when rebasing on 4.1 and use that
instead of ktime_sub/ktime_to_timespec/timespec_to_jiffies combo (as
suggested by Arnd)
v2: switch over from 'struct timespec' to ktime_t throughout, since
'struct timespec' will be deprecated (as suggested by Arnd)
v3: minor cosmetic tweaks
Cc: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Rob Clark <robdclark@gmail.com>
Some targets (eg: msm8994) use the pinctrl framework to configure
interface pins. This change adds support for initialization and
pinctrl active/sleep state control for the HDMI driver.
Signed-off-by: Stephane Viau <sviau@codeaurora.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
DRM device's dev (hdmi->dev->dev) points to the mdss_mdp device
handle. Instead, we should get a reference to the mdss_hdmi
handle.
Signed-off-by: Stephane Viau <sviau@codeaurora.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
That will complete the lists of Alpha + RGB formats.
Signed-off-by: Stephane Viau <sviau@codeaurora.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
When the .atomic_commit() handler fails, clean up planes previoulsy
prepared by drm_atomic_helper_prepare_planes() with a call to
drm_atomic_helper_cleanup_planes().
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Rob Clark <robdclark@gmail.com>
DSI video mode engine can only take active-high sync signals. This
change prevents MDP5 sending active-low sync signals to DSI in any
case.
Signed-off-by: Hai Li <hali@codeaurora.org>
Tested-by: Archit Taneja <architt@codeaurora.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
I'm not sure where, exactly, but somewhere in here we must be relying on
an implicit include.
drivers/gpu/drm/msm/dsi/dsi_host.c: In function ‘dsi_host_init_panel_gpios’:
drivers/gpu/drm/msm/dsi/dsi_host.c:1356:2: error: implicit declaration of function ‘devm_gpiod_get’ [-Werror=implicit-function-declaration]
msm_host->disp_en_gpio = devm_gpiod_get(panel_device,
^
drivers/gpu/drm/msm/dsi/dsi_host.c:1356:25: warning: assignment makes pointer from integer without a cast [enabled by default]
msm_host->disp_en_gpio = devm_gpiod_get(panel_device,
^
drivers/gpu/drm/msm/dsi/dsi_host.c:1364:3: error: implicit declaration of function ‘gpiod_direction_output’ [-Werror=implicit-function-declaration]
ret = gpiod_direction_output(msm_host->disp_en_gpio, 0);
^
drivers/gpu/drm/msm/dsi/dsi_host.c:1371:20: warning: assignment makes pointer from integer without a cast [enabled by default]
msm_host->te_gpio = devm_gpiod_get(panel_device, "disp-te");
^
drivers/gpu/drm/msm/dsi/dsi_host.c:1378:3: error: implicit declaration of function ‘gpiod_direction_input’ [-Werror=implicit-function-declaration]
ret = gpiod_direction_input(msm_host->te_gpio);
^
drivers/gpu/drm/msm/dsi/dsi_host.c: In function ‘msm_dsi_host_power_on’:
drivers/gpu/drm/msm/dsi/dsi_host.c:1918:3: error: implicit declaration of function ‘gpiod_set_value’ [-Werror=implicit-function-declaration]
gpiod_set_value(msm_host->disp_en_gpio, 1);
^
Signed-off-by: Brian Norris <computersforpeace@gmail.com>
Cc: David Airlie <airlied@linux.ie>
Cc: Rob Clark <robdclark@gmail.com>
Cc: Hai Li <hali@codeaurora.org>
Cc: dri-devel@lists.freedesktop.org
Signed-off-by: Rob Clark <robdclark@gmail.com>
Since 39b2bbe3d7 (gpio: add flags argument to gpiod_get*() functions)
which appeared in v3.17-rc1, the gpiod_get* functions take an additional
parameter that allows to specify direction and initial value for output.
Also there is a variant to find optional gpios that returns NULL if
there is no gpio instead of -ENOENT.
Make use of both features to simplify the driver.
This makes error checking more strict because errors like -ENOSYS ("no
gpio support compiled in") or -EPROBE_DEFER ("gpio not ready yet") are
handled correctly now.
Furthermore this is one caller less that stops us making the flags
argument to gpiod_get*() mandatory.
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Rob Clark <robdclark@gmail.com>
There are different types of PHY from one chipset to another, while
the DSI host controller is relatively consistent across platforms.
Also, the PLL inside PHY is providing the source of DSI byte and
pixel clocks, which are used by DSI host controller. Separated devices
for clock provider and clock consumer make DSI driver better fit into
common clock framework.
Signed-off-by: Hai Li <hali@codeaurora.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
This change activates PLL driver for DSI to work with
common clock framework.
Signed-off-by: Hai Li <hali@codeaurora.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
DSI byte clock and pixel clocks are sourced from DSI PLL.
This change adds the DSI PLL source clock driver under
common clock framework.
This change handles DSI 28nm PLL only.
Signed-off-by: Hai Li <hali@codeaurora.org>
Signed-off-by: Archit Taneja <architt@codeaurora.org>
Signed-off-by: Stephane Viau <sviau@codeaurora.org>
Signed-off-by: Wentao Xu <wentaox@codeaurora.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
regulator_get() never returns NULL. There's no need for IS_ERR_OR_NULL()
Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: Rob Clark <robdclark@gmail.com>
msm_ioremap() never returns NULL. There's no need for IS_ERR_OR_NULL()
Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: Rob Clark <robdclark@gmail.com>
CRTCs in DSI command mode data path should wait for pingpong done,
instead of vblank, to finish atomic commit.
This change is to enable PP_DONE irq on command mode CRTCs and wait for
this irq happens before atomic commit completion.
Signed-off-by: Hai Li <hali@codeaurora.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
MDP FLUSH registers could indicate if the previous flush updates
has taken effect at vsync boundary. Making use of this H/W feature
can catch the vsync that happened between CRTC atomic_flush and
*_wait_for_vblanks, to avoid unnecessary wait.
This change allows kms CRTCs to use their own *_wait_for_commit_done
functions to wait for FLUSH register cleared at vsync, before commit
completion.
Signed-off-by: Hai Li <hali@codeaurora.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
wait_for_completion_timeout returns 0 in case of timeout and never
return < 0 so there is no additional information in printing the
value of time_left here as it will always be 0, thus it can be dropped.
Signed-off-by: Nicholas Mc Guire <hofrat@osadl.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
wait_for_completion_timeout returns >= 0 but never
negative - so the error check should be against equality
to 0 not <= 0.
Signed-off-by: Nicholas Mc Guire <hofrat@osadl.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
return type of wait_for_completion_timeout is unsigned long not int, this
patch assigns the return value of wait_for_completion_timeout to an
appropriately typed and named variable.
Signed-off-by: Nicholas Mc Guire <hofrat@osadl.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
Some DSI peripherals rely on the HS clock on DSI clock lane as their clock
source. If the clock lane transitions between HS and LP states, it
can disrupt the functioning of such peripherals.
The mipi dsi mode flag MIPI_DSI_CLOCK_NON_CONTINUOUS already exists for
such peripheral drivers. Use it to configure the bit CLKLN_HS_FORCE_REQUEST
in DSI_LANE_CTRL.
Signed-off-by: Archit Taneja <architt@codeaurora.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
If the GEM object is imported, drm_prime_gem_destroy needs to be
called to clean up dma buffer related information.
Signed-off-by: Jilai Wang <jilaiw@codeaurora.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
Fix build warning when building edp/edp_aux.o due to missing
prototype for edp_aux_transfer.
This function is only used in edp_aux.c so just make it static.
Reported-by: Nicholas Mc Guire <hofrat@osadl.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
wait_for_completion_timeout returns 0 in case of timeout so printing the
return value here will always yield 0 and is therefor redundant - dropped.
Signed-off-by: Nicholas Mc Guire <hofrat@osadl.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
The timeout is passed as a constant which makes it HZ dependent because
jiffies are expected so it should be converted to jiffies. The actual
value is not clear from the code - my best guess is that this should be
300 milliseconds given that other timeouts are in milliseconds based on
looking at other drm drivers (e.g. exynos_drm_dsi.c:356 300ms,
tegra/dpaux.c:188 250ms) - this needs to be confirmed by someone who
knows the details of the driver.
Signed-off-by: Nicholas Mc Guire <hofrat@osadl.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
wait_for_completion_timeout return >= 0 but never negative so the check
logic looks inconsistent. Further the return value of
wait_for_completion_timeout was being passed up the call chain but the
x call sites as drm_dp_i2c_do_msg()/drm_dp_dpcd_access() check for < 0
thus timeout was being treated as success case.
<snip> drivers/gpu/drm/drm_dp_helper.c:drm_dp_i2c_do_msg()
mutex_lock(&aux->hw_mutex);
ret = aux->transfer(aux, msg);
mutex_unlock(&aux->hw_mutex);
if (ret < 0) {
<snip>
logic in edp_aux_transfer() seems incorrect as it could return 0 (timeout)
but checks of <= 0 to indicate error so the return probably should be
-ETIMEDOUT in case wait_for_completion_timeout returns 0 (timeout
occurred).
Signed-off-by: Nicholas Mc Guire <hofrat@osadl.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
Seems like disable can race with complete_flip() in process of disabling
a crtc, leading to:
[ 49.065364] Call trace:
[ 49.071441] [<ffffffc00041d5a0>] mdp5_ctl_blend+0x20/0x1c0
[ 49.073788] [<ffffffc00041ebcc>] mdp5_crtc_disable+0x3c/0xa8
[ 49.079348] [<ffffffc0003e7854>] disable_outputs.isra.4+0x11c/0x220
[ 49.085164] [<ffffffc0003e7afc>] drm_atomic_helper_commit_modeset_disables+0x14/0x38
[ 49.091155] [<ffffffc000425c80>] complete_commit+0x40/0xb8
[ 49.099136] [<ffffffc0004260ac>] msm_atomic_commit+0x364/0x398
[ 49.104430] [<ffffffc00040a614>] drm_atomic_commit+0x3c/0x70
[ 49.110249] [<ffffffc0003e67b8>] drm_atomic_helper_set_config+0x1b0/0x3e0
[ 49.116065] [<ffffffc0003f99bc>] drm_mode_set_config_internal+0x64/0xf8
[ 49.122746] [<ffffffc0003fa624>] drm_framebuffer_remove+0xe4/0x128
[ 49.129171] [<ffffffc0003feaf8>] drm_mode_rmfb+0xc0/0x100
[ 49.135420] [<ffffffc0003efba8>] drm_ioctl+0x258/0x4d0
[ 49.140889] [<ffffffc0001b0388>] do_vfs_ioctl+0x338/0x5d0
[ 49.145921] [<ffffffc0001b06a8>] SyS_ioctl+0x88/0xa0
It makes no sense to free the ctl without disabling all stages, so lets
just move them together to avoid the crash.
Signed-off-by: Rob Clark <robdclark@gmail.com>
As found in apq8016 (used in DragonBoard 410c) and msm8916.
Note that numerically a306 is actually 307 (since a305c already claimed
306). Nice and confusing.
Signed-off-by: Rob Clark <robdclark@gmail.com>
A few spots in the driver have support for downstream android
CONFIG_MSM_BUS_SCALING. This is mainly to simplify backporting the
driver for various devices which do not have sufficient upstream
kernel support. But the intentionally dead code seems to cause
some confusion. Rename the #define to make this more clear.
Signed-off-by: Rob Clark <robdclark@gmail.com>
Using fb modifier flag, support NV12MT format in MDP4.
v2:
- rework the modifier's description [Daniel Vetter's comment]
- drop .set_mode_config() callback [Rob Clark's comment]
v3:
- change VENDOR's name and restrict usage to NV12 [pointed by Daniel]
Signed-off-by: Rob Clark <robdclark@gmail.com>
Dump a bit more info when the GPU hangs, without having hang_debug
enabled (which dumps a *lot* of registers). Also dump the scratch
registers, as they are useful for determining where in the cmdstream
the GPU hung (and they seem always safe to read when GPU has hung).
Note that the freedreno gallium driver emits increasing counter values
to SCRATCH6 (to identify tile #) and SCRATCH7 (to identify draw #), so
these two in particular can be used to "triangulate" where in the
cmdstream the GPU hung.
Signed-off-by: Rob Clark <robdclark@gmail.com>
The goto is correct, and we never reach the return statement so just
delete the dead code.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
virtio_gpu_alloc_object() returns an error pointer, it never returns
NULL.
Fixes: dc5698e80c ('Add virtio gpu driver.')
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
We were doing it in the common code and in the IP specific code.
Remove the IP specific code. The common code handles the
ordering properly.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Enable VCE dpm and powergating. VCE dpm dynamically scales the VCE clocks on
demand.
Signed-off-by: Sonny Jiang <sonny.jiang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
smu_init allocates buffers and initializes them. It does not
touch the hw. There is no need to do it again on resume. It
should really be part of sw_init (and smu_fini should be part
of sw_fini), but we need the firmware sizes from the other IPs
for firmware loading so we have to wait until sw init is done
for all other IPs.
Reviewed-by: Sonny Jiang <Sonny.Jiang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
smu_init allocates buffers and initializes them. It does not
touch the hw. There is no need to do it again on resume. It
should really be part of sw_init (and smu_fini should be part
of sw_fini), but we need the firmware sizes from the other IPs
for firmware loading so we have to wait until sw init is done
for all other IPs.
Reviewed-by: Sonny Jiang <Sonny.Jiang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
smu_init allocates buffers and initializes them. It does not
touch the hw. There is no need to do it again on resume. It
should really be part of sw_init (and smu_fini should be part
of sw_fini), but we need the firmware sizes from the other IPs
for firmware loading so we have to wait until sw init is done
for all other IPs.
Reviewed-by: Sonny Jiang <Sonny.Jiang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Some leftover copy and pastes from radeon that never
got updated.
Reviewed-by: Christian König <christian.koenig@amd.com>
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Need to adjust the number of CUs and RBs.
v2: get proper values
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Samuel Li <samuel.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The driver makes use of this information so print if to aid in
debugging.
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Samuel Li <samuel.li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Remove the unnecessary returned status and make the IOCTL write only.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
