IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
pm_runtime_get_sync() increments the runtime PM usage counter even
when it returns an error code. Thus a pairing decrement is needed on
the error handling path to keep the counter balanced. For other error
paths after this call, things are the same.
Fix this by adding pm_runtime_put_noidle() after 'err_runtime_disable'
label. But in this case, the error path after pm_runtime_put_sync()
will decrease PM usage counter twice. Thus add an extra
pm_runtime_get_noresume() in this path to balance PM counter.
Signed-off-by: Dinghao Liu <dinghao.liu@zju.edu.cn>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
camss_probe() does not free camss on error handling paths. The patch
introduces an additional error label for this purpose. Besides, it
removes call of v4l2_async_notifier_cleanup() from
camss_of_parse_ports() since its caller, camss_probe(), cleans up all
its resources itself.
Found by Linux Driver Verification project (linuxtesting.org).
Signed-off-by: Evgeny Novikov <novikov@ispras.ru>
Co-developed-by: Anton Vasilyev <vasilyev@ispras.ru>
Signed-off-by: Anton Vasilyev <vasilyev@ispras.ru>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
pm_runtime_get_sync() increments the runtime PM usage counter even
when it returns an error code. Thus a pairing decrement is needed on
the error handling path to keep the counter balanced.
Signed-off-by: Dinghao Liu <dinghao.liu@zju.edu.cn>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Currently we are considering the instances which are available
in core->inst list for load calculation in min_loaded_core()
function, but this is incorrect because by the time we call
decide_core() for second instance, the third instance not
filled yet codec_freq_data pointer.
Solve this by considering the instances whose session has started.
Cc: stable@vger.kernel.org # v5.7+
Fixes: 4ebf969375bc ("media: venus: introduce core selection")
Tested-by: Douglas Anderson <dianders@chromium.org>
Signed-off-by: Mansur Alisha Shaik <mansur@codeaurora.org>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Currently for msm_csiphy, msm_csid, and msm_ispif subdevices the media
entity function field is set to MEDIA_ENT_F_IO_V4L. This is incorrect
as MEDIA_ENT_F_IO_V4L implies V4L2 video node.
Change it to MEDIA_ENT_F_PROC_VIDEO_PIXEL_FORMATTER as this is the best
fit from the functions defined in include/uapi/linux/media.h.
Signed-off-by: Andrey Konovalov <andrey.konovalov@linaro.org>
Tested-by: Robert Foss <robert.foss@linaro.org>
Reviewed-by: Robert Foss <robert.foss@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
pm_runtime_get_sync() increments the runtime PM usage counter even
when it returns an error code. Thus a pairing decrement is needed on
the error handling path to keep the counter balanced.
Signed-off-by: Dinghao Liu <dinghao.liu@zju.edu.cn>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
pm_runtime_get_sync() increments the runtime PM usage counter even
when it returns an error code. Thus a pairing decrement is needed on
the error handling path to keep the counter balanced.
Signed-off-by: Dinghao Liu <dinghao.liu@zju.edu.cn>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
pm_runtime_get_sync() increments the runtime PM usage counter even
when it returns an error code. Thus a pairing decrement is needed on
the error handling path to keep the counter balanced.
Signed-off-by: Dinghao Liu <dinghao.liu@zju.edu.cn>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
According to the v4l spec s_fmt must return EBUSY while the
particular queue is streaming. Add such check in encoder and
decoder s_fmt methods.
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
According to stateful Codec API the decoder will process all
remaining buffers from before the source change event in
dynamic-resolution-change state and mark the last buffer with
V4L2_BUF_FLAG_LAST.
In Venus case the firmware doesn't mark that last buffer and
some mechanism have to be created in v4l decoder driver.
Fortunately the firmware interface (HFI) claims that the
decoder output buffers will be returned to v4l decoder
driver before it send the insufficient event.
In order to do that we save last queued in the driver capture
buffer in the event_notify and issue flush on output firmware
buffers queue. Once the saved buffer is returned (as a result of
flush command) we mark it as LAST. For all that possible we
extend HFI flush command with one more argument and one more
flush_done HFI driver callback.
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Presently the list initialization is done only in
dynamic-resolution-change state, which leads to list corruptions
and use-after-free. Init list_head unconditionally in
vdec_stop_capture called by vb2 stop_streaming without takeing
into account current codec state.
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Once the hfi_session_flush is issued by the vdec all queued
buffers to firmware should be returned to the v4l driver. Some
of those buffers are not processed at the time of flush command,
those buffers has filled len zero (no data). Catch that in
buffer_done callback and mark not filled capture buffers with
error state so that client can discard them.
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Currently calling venus_helper_buffers_done() will return buffers to
user for both capture and output queues in the same call. This is
wrong because both queues are really separate and calling
stop_streaming on one queue shouldn't return buffers for the other.
Solve this by add a new queue type argument and fix the clients of
the helper function.
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
The array is not changed in the code, so make it const.
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
The hfi_destroy function is called too early in remove method. It
destroys a mutex which is used later in the .remove from pmruntime.
Solve the issue by moving hfi_destroy after last usage of the mutex.
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
The Venus driver is voting Configuration NoC during .probe but not clear
voting in .suspend. Because of this NoC is up during shutdown also. As a
consequence the whole device could leak energy while in .suspend.
So correct this by moving voting in .resume and unvoting
in .suspend
Signed-off-by: Mansur Alisha Shaik <mansur@codeaurora.org>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:
struct foo {
int stuff;
struct boo array[];
};
By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.
Also, notice that, dynamic memory allocations won't be affected by
this change:
"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]
This issue was found with the help of Coccinelle.
[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 76497732932f ("cxgb3/l2t: Fix undefined behaviour")
Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:
struct foo {
int stuff;
struct boo array[];
};
By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.
Also, notice that, dynamic memory allocations won't be affected by
this change:
"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]
This issue was found with the help of Coccinelle.
[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 76497732932f ("cxgb3/l2t: Fix undefined behaviour")
Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Implement pmruntime autosuspend in video decoder. This will
allow to save power while the userspace is inactive for some
reasonable period of time. Here we power-off venus core clocks
and power domain and don't touch vcodec because it is under
hardware control. The later decision is made to simplify the
code and avoid a mess in the power management code.
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
With the latest cleanup in qcom scm driver the secure monitor
call for setting the remote processor state returns EINVAL when
it is called for the first time and after another scm call
auth_and_reset. The error returned from scm call could be ignored
because the state transition is already done in auth_and_reset.
Acked-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Cc: stable@vger.kernel.org
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Add encoder control for enabling/disabling frame rate control via
V4L2_CID_MPEG_VIDEO_FRAME_RC_ENABLE. It is enabled by default.
Signed-off-by: Jeffrey Kardatzke <jkardatzke@google.com>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
This fixes the following smatch warning in the error path:
drivers/media/platform/qcom/venus/vdec.c:968 vdec_start_streaming()
warn: inconsistent returns 'mutex:&inst->lock'.
Locked on: line 952
Unlocked on: line 963
line 968
by goto mutex unlock.
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Currently, v4l2_pipeline_pm_use() prototype is:
int v4l2_pipeline_pm_use(struct media_entity *entity, int use)
Where the 'use' argument shall only be set to '1' for enable/power-on,
or to '0' for disable/power-off. The integer return is specified
as only meaningful when 'use' is set to '1'.
Let's enforce this semantic by splitting the function in two:
v4l2_pipeline_pm_get and v4l2_pipeline_pm_put. This is done
for several reasons.
It makes the API easier to use (or harder to misuse).
It removes the constraint on the values the 'use' argument
shall take. Also, it removes the need to constraint
the return value, by making v4l2_pipeline_pm_put void return.
And last, it's more consistent with other kernel APIs, such
as the runtime pm APIs, which makes the code more symmetric.
Signed-off-by: Ezequiel Garcia <ezequiel@collabora.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
In order to use dynamic video codec assignment add a new sdm845-v2
DT compatible and new venus_resource structure to cover the binding
where all pmdomains and clocks are under the venus_core control.
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Handle 10bit video streams in the decoder by using dithering, i.e
the decoder output buffers will be in 8bit format.
The runtime handling is implemented by sending v4l2 event to
userspace application, then the application should stop the
streaming on capture queue and initiate format negotiation, and
start streaming again.
Signed-off-by: Aniket Masule <amasule@codeaurora.org>
Co-developed-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Presently the core (vcodec pipelines) assignment is static. Here we
introduce dynamic load balancing across the cores depending on the
current session load. The load on earch core is calculated and core
with minimum load is assigned to given instance. This will be
applicable on Venus v4 with more than one vcodec cores.
Signed-off-by: Aniket Masule <amasule@codeaurora.org>
Co-developed-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Redesign core (vcodec) clock control to give the venus core more
freedom to control them in order to make possible to use core
selection feature on Venus IP v4.
Move all clock and pmdomain functions in separate file and abstract
power control with common operations per Venus IP version.
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Those two properties are not implemented for Venus v4 (sdm845),
thus don't set them to firmware.
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Replace explicit use of members of struct resource by %pR to print
the resource.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Some older MSM8916 Venus firmware versions also seem to indicate
support for encoding HEVC, even though they really can't.
This will lead to errors later because hfi_session_init() fails
in this case.
HEVC is already ignored for "dec_codecs", so add the same for
"enc_codecs" to make these old firmware versions work correctly.
Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
'GRABBER' is a weird name, all other types map to the /dev
device names. Rename to 'VIDEO' to be consistent with the
other types.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
v4l2_compat_ioctl32() is the function that calls into
v4l2_file_operations->compat_ioctl32(), so setting that back to the same
function leads to a trivial endless loop, followed by a kernel
stack overrun.
Remove the incorrect assignment.
Cc: stable@vger.kernel.org
Fixes: 7472c1c69138 ("[media] media: venus: vdec: add video decoder files")
Fixes: aaaa93eda64b ("[media] media: venus: venc: add video encoder files")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
As reported by jenkins@linuxtv.org, the build with i386 fails
with:
ld: drivers/media/platform/qcom/venus/helpers.o: in function `venus_helper_load_scale_clocks':
(.text+0x1d77): undefined reference to `__udivdi3'
ld: (.text+0x1dce): undefined reference to `__udivdi3'
make: *** [Makefile:1094: vmlinux] Error 1
That's because it divides an u32 bit integer by a u64 one.
Reviewed-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Current clock scaling calculations are same for vpu4 and
previous versions. For vpu4, Clock scaling calculations
are updated with cycles/mb and bitrate. This helps in
getting precise clock required.
Signed-off-by: Aniket Masule <amasule@codeaurora.org>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Add vpp cycles for different types of codec.
It indicates the cycles required by video hardware
to process each macroblock. Add vsp cycles, cycles
required by stream processor. Initialize the codec
data with core resources.
Signed-off-by: Aniket Masule <amasule@codeaurora.org>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Failure to suspend (venus_suspend_3xx) happens when the system
is fresh booted and loading venus driver. This happens once and
after reload the venus driver modules the problem disrepair.
Fix the failure by skipping the check for WFI and IDLE bits if
PC_READY is on in control status register.
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
This aims to add a requests for bandwidth scaling depending
on the resolution and framerate (macroblocks per second). The
exact value of the requested bandwidth is get from a
pre-calculated tables for encoder and decoder.
Acked-by: Georgi Djakov <georgi.djakov@linaro.org>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
In downstream driver, there are two frequency tables defined,
one for the encoder and one for the decoder:
/* Encoders /
<972000 490000000 0x55555555>, / 4k UHD @ 30 /
<489600 320000000 0x55555555>, / 1080p @ 60 /
<244800 150000000 0x55555555>, / 1080p @ 30 /
<108000 75000000 0x55555555>, / 720p @ 30 */
/* Decoders /
<1944000 490000000 0xffffffff>, / 4k UHD @ 60 /
< 972000 320000000 0xffffffff>, / 4k UHD @ 30 /
< 489600 150000000 0xffffffff>, / 1080p @ 60 /
< 244800 75000000 0xffffffff>; / 1080p @ 30 */
It shows that encoder always needs a higher clock than decoder.
In current venus driver, the unified frequency table is aligned
with the downstream decoder table which causes performance issues
in encoding scenarios. Fix that by aligning frequency table on
worst case (encoding).
Signed-off-by: Loic Poulain <loic.poulain@linaro.org>
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
On 32-bits archs, a signed integer has 31 bits plus on extra
bit for signal. Due to that, touching the 32th bit with something
like:
int bar = 1 << 31;
has an undefined behavior in C on 32 bit architectures, as it
touches the signal bit. This is warned by cppcheck.
Instead, force the numbers to be unsigned, in order to solve this
issue.
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Tag all the coded formats where the venus vdec supports dynamic
resolution switching.
Signed-off-by: Maxime Jourdan <mjourdan@baylibre.com>
Acked-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Reviewed-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Cache flags, timestamps and timecode structure of OUTPUT buffers
in per-instance structure array and fill correctly the same when
the CAPTURE buffers are done.
This will make v4l2-compliance decoder streaming test happy.
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Correct handling of OUTPUT buffers field and make v4l2-compliance
happy.
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
This refactors code for start/stop streaming vb2 operations and
adds a state machine handling similar to the one in stateful codec
API documentation. One major change is that now the HFI session is
started on STREAMON(OUTPUT) and stopped on REQBUF(OUTPUT,count=0),
during that time STREAMOFF(CAP,OUT) just flush buffers but doesn't
stop the session. The other major change is that now the capture
and output queues are completely separated.
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
In most of the cases the client will know better what could be
the maximum size for compressed data buffers. Change the driver
to permit the user to set bigger size for the compressed buffer
but make reasonable sanitation.
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Until now we returned num_output_bufs set during reqbuf but
that could be wrong when we implement stateful Codec API. So
get the minimum buffers for capture from HFI. This is supposed
to be called after stream header parsing, i.e. after dequeue
v4l2 event for change resolution.
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
This adds three more helper functions:
* for internal buffers reallocation, applicable when we are doing
dynamic resolution change
* for initial buffer processing of capture and output queue buffer
types
All of them will be needed for stateful Codec API support.
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
This makes hfi_session_init to return an error when it is
already called without a call to hfi_session_deinit.
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
