- 16 Nov, 2020 40 commits
-
-
Niklas Söderlund authored
In preparation of suspend/resume support cache the chsel value when written to the register so it can be restored on resume if needed. Signed-off-by:
Niklas Söderlund <niklas.soderlund+renesas@ragnatech.se> Reviewed-by:
Laurent Pinchart <laurent.pinchart@ideasonboard.com> Reviewed-by:
Jacopo Mondi <jacopo+renesas@jmondi.org> Signed-off-by:
Hans Verkuil <hverkuil-cisco@xs4all.nl> Signed-off-by:
Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
-
Niklas Söderlund authored
When returning buffers to user-space it's no longer needed to examine the buffers given to hardware as recent changes guarantees the only buffer present in the hardware registers when the driver is not in the running state is the scratch buffer. Remove the special case and rename the function to better describe it now only deals with buffers queued to the driver but not yet recorded in the hardware registers. Signed-off-by:
-
Niklas Söderlund authored
In its early stages the VIN driver did not use an internal scratch buffer. This leads to a unnecessary complex start and stop behavior, specially for TB/BT. The driver now always allocates a scratch buffer to deal with buffer under-runs, use the scratch buffer to also simplify starting and stopping. When capture is starting use the scratch buffer instead of a user-space buffers while syncing the driver with the hardware state. This allows the driver to know that no user-space buffer is given to the hardware before the running state is reached. When capture is stopping use the scratch buffer instead of leaving the user-space buffers in place and add a check that all user-space buffers are processed by the hardware before transitioning from the stopping to stopped state. This allows the driver to know all user-space buffers given to the hardware are fully processed. This change in itself does not improve the driver much but it paves the way for future simplifications and enhancements. One direct improvement of this change is that TB/BT buffers returned to user-space while stopping will always contain both fields, that was not guaranteed before. Signed-off-by:
-
Thomas Gleixner authored
The memory allocations in cx231xx_init_*() happen all in task context with GFP_KERNEL. Therefore a dev_err() trying to deduce whether this is called from task or interrupt context is pretty useless. Remove these historical leftovers. Signed-off-by:
Thomas Gleixner <tglx@linutronix.de> Signed-off-by:
-
Thomas Gleixner authored
The memory allocations in au0828_init_isoc() happen all in task context with GFP_KERNEL. Therefore a printk() trying to deduce whether this is called from task or interrupt context is pretty useless. Convert it to au0828_isocdbg() as the other one in that function and for completeness sake add one for the URB allocation as well. Signed-off-by:
-
Thomas Gleixner authored
None of these BUG_ON()'s is justified. BUG_ON() should only be used when there is really no way to survive. If at all these could be replaced by lockdep_assert_preemption_enabled() to cover all invalid caller context and not just those covered by in_interrupt(). But all functions which are invoked from those places contain already debug mechanisms to catch wrong context, so having these extra checks is not having any advantage. Signed-off-by:
-
Yu Kuai authored
mtk_vcodec_release_enc_pm() will be called in two places: a. mtk_vcodec_init_enc_pm() succeed while mtk_vcodec_probe() return error. b. mtk_vcodec_enc_remove(). In both cases put_device() call is needed, since of_find_device_by_node() was called in mtk_vcodec_init_enc_pm() previously. Thus add put_devices() call in mtk_vcodec_release_enc_pm() Signed-off-by:
Yu Kuai <yukuai3@huawei.com> Signed-off-by:
-
Yu Kuai authored
if of_find_device_by_node() succeed, mtk_vcodec_init_enc_pm() doesn't have a corresponding put_device(). Thus add jump target to fix the exception handling for this function implementation. Fixes: 4e855a6e ("[media] vcodec: mediatek: Add Mediatek V4L2 Video Encoder Driver") Signed-off-by:
-
Yu Kuai authored
mtk_vcodec_release_dec_pm() will be called in two places: a. mtk_vcodec_init_dec_pm() succeed while mtk_vcodec_probe() return error. b. mtk_vcodec_dec_remove(). In both cases put_device() call is needed, since of_find_device_by_node() was called in mtk_vcodec_init_dec_pm() previously. Thus add put_devices() call in mtk_vcodec_release_dec_pm() Fixes: 590577a4 ("[media] vcodec: mediatek: Add Mediatek V4L2 Video Decoder Driver") Signed-off-by:
-
Yu Kuai authored
if of_find_device_by_node() succeed, mtk_vcodec_init_dec_pm() doesn't have a corresponding put_device(). Thus add jump target to fix the exception handling for this function implementation. Fixes: 590577a4 ("[media] vcodec: mediatek: Add Mediatek V4L2 Video Decoder Driver") Signed-off-by:
-
Yu Kuai authored
if of_find_device_by_node() succeeds, mtk_jpeg_clk_init() doesn't have a corresponding put_device(). Thus add put_device() to fix the exception handling for this function implementation. Fixes: 648372a8 ("media: platform: Change the call functions of getting/enable/disable the jpeg's clock") Signed-off-by:
-
Hans Verkuil authored
If a CEC device node is unregistered, then it should be marked as unregistered before waking up any filehandles that are waiting for an event. This ensures that there is no race condition where an application can call CEC_DQEVENT and have the ioctl return 0 instead of ENODEV. Signed-off-by:
-
Lukas Bulwahn authored
Commit 754f0f1b ("media: MAINTAINERS: change maintainer of the zoran driver") added a new section in MAINTAINERS with an invalid file entry and at the wrong place for alphabetic ordering. Hence, ./scripts/get_maintainer.pl --self-test=patterns complains: warning: no file matches F: Documentation/media/v4l-drivers/zoran.rst Point the file entry to the right location and move the section to the right place in MAINTAINERS. Signed-off-by:
Lukas Bulwahn <lukas.bulwahn@gmail.com> Signed-off-by:
-
Gustavo A. R. Silva authored
In order to enable -Wimplicit-fallthrough for Clang[1], replace the existing /* fall through */ comments with the new pseudo-keyword macro fallthrough[2]. [1] https://git.kernel.org/linus/e2079e93f562c7f7a030eb7642017ee5eabaaa10 [2] https://www.kernel.org/doc/html/v5.7/process/deprecated.html?highlight=fallthrough#implicit-switch-case-fall-throughSigned-off-by:
Gustavo A. R. Silva <gustavoars@kernel.org> Signed-off-by:
-
Colin Ian King authored
The are two instances of sizeof() being used incorrectly. The sizeof(void *) is incorrect because urb_buffer is a char ** pointer, fix this by using sizeof(*dev->urb_buffer). The sizeof(dma_addr_t *) is incorrect, it should be sizeof(*dev->urb_dma), which is a dma_addr_t and not a dma_addr_t *. This errors did not cause any issues because it just so happens the sizes are the same. Addresses-Coverity: ("Sizeof not portable (SIZEOF_MISMATCH)") Fixes: 16427faf ("[media] tm6000: Add parameter to keep urb bufs allocated") Signed-off-by:Colin Ian King <colin.king@canonical.com> Signed-off-by:
-
Hans Verkuil authored
cec-core.rst was missing 'are', and cec-ioc-g-mode.rst was missing a comma. In both cases this made the description of the Monitor All functionality hard to read. Signed-off-by:
-
Ezequiel Garcia authored
Make interrupt naming more consistent by using a common pattern for video and jpeg interrupts. With this commit, interrupts are shown as: 29: 0 0 GPC 12 Level coda-video 30: 0 0 GPC 3 Level coda-jpeg Signed-off-by:
Ezequiel Garcia <ezequiel@collabora.com> Acked-by:
Philipp Zabel <p.zabel@pengutronix.de> Signed-off-by:
-
Ezequiel Garcia authored
Instead of printing just the device type, let's use the device name, which makes the message more useful. With this commit, the messages shown when the driver is registered are: coda 2040000.vpu: Firmware code revision: 570363 coda 2040000.vpu: Initialized CODA960. coda 2040000.vpu: Firmware version: 3.1.1 coda 2040000.vpu: coda-jpeg-encoder registered as video0 coda 2040000.vpu: coda-jpeg-decoder registered as video1 coda 2040000.vpu: coda-video-encoder registered as video2 coda 2040000.vpu: coda-video-decoder registered as video3 Signed-off-by:
-
Ezequiel Garcia authored
The H.264 small buffer padding is done under the (ctx->qsequence == 0 && payload < 512) condition. Given this is the exact same condition immediately above, we can move it right there, making the code slightly clearer. This change shouldn't affect functionality as it's just cosmetics. Signed-off-by:
-
Ezequiel Garcia authored
The ctx->initialized flag is set in __coda_decoder_seq_init, so it's redundant to set it in coda_dec_seq_init_work. Remove the redundant set, which allows to simplify the implementation quite a bit. This change shouldn't affect functionality as it's just cosmetics. Signed-off-by:
-
Evgeny Novikov authored
zr364xx_start_readpipe() can fail but callers do not care about that. This can result in various negative consequences. The patch adds missed error handling. Found by Linux Driver Verification project (linuxtesting.org). Signed-off-by:
Evgeny Novikov <novikov@ispras.ru> Signed-off-by:
-
Liu Shixin authored
Use DEFINE_SEQ_ATTRIBUTE macro to simplify the code. Signed-off-by:
Liu Shixin <liushixin2@huawei.com> Signed-off-by:
-
YueHaibing authored
platform_get_irq() will call dev_err() itself on failure, so there is no need for the driver to also do this. This is detected by coccinelle. Signed-off-by:
YueHaibing <yuehaibing@huawei.com> Signed-off-by:
-
Helen Koike authored
aclk_isp_wrap is a child of aclk_isp, and hclk_isp_wrap is a child of hclk_isp, thus we can remove parents from the list. Also, for the isp0, we only need the ISP clock, ACLK and HCLK. In the future we'll need a pixel clock for RK3288 and RK3399, and a JPEG clock for RK3288. So with the goal to cleanup the dt-bindings and remove it from staging, simplify clock names to isp, aclk and hclk. Assigned clocks are meant to refer to the full path in the clock tree, i.e. the leaf in the tree. For instance, in RK3399, the clock responsible for ACLK (ISP AXI CLOCK) is aclk_isp0_wrapper. For reference, this is the isp clock topology on RK3399: xin24m pll_npll npll clk_isp1 clk_isp0 pll_cpll cpll aclk_isp1 aclk_isp1_noc hclk_isp1 aclk_isp1_wrapper hclk_isp1_noc aclk_isp0 hclk_isp1_wrapper aclk_isp0_wrapper aclk_isp0_noc hclk_isp0 hclk_isp0_wrapper hclk_isp0_noc pclkin_isp1_wrapper Signed-off-by:Helen Koike <helen.koike@collabora.com> Reviewed-by:
Tomasz Figa <tfiga@chromium.org> Reviewed-by:
Rob Herring <robh@kernel.org> Signed-off-by:
-
Helen Koike authored
Fix the following error found with make ARCH=arm64 dt_binding_check: Documentation/devicetree/bindings/media/rockchip-isp1.example.dts:24.27-101.11: Warning (unit_address_vs_reg): /example-0/parent@0: node has a unit name, but no reg or ranges property Reported-by:
Johan Jonker <jbx6244@gmail.com> Signed-off-by:
-
Helen Koike authored
Organize properties order in dt-bindings to move it out of staging. On top: compatible, reg and interrupts. Then alphabetical order, then properties starting with '#'. Signed-off-by:
-
Helen Koike authored
Add missing required items in Rockchip ISP1 dt-bindings example for a complete i2c node. Drop unit address to Fix error: /example-0/parent/i2c@ff160000: node has a unit name, but no reg or ranges property Remove unecessary fields for the example. Signed-off-by:
-
Helen Koike authored
Add missing required nodes in json-schema yaml file for Rockchip ISP1 dt-bindings. Signed-off-by:
-
Helen Koike authored
Allow streaming from self picture path and main picture path at the same time. Take care for s_stream() callbacks to not be called twice. When starting a stream, s_stream(true) shouldn't be called for the isp and the sensor if the other stream is already enabled (since it was already called). When stopping a stream, s_stream(false) shouldn't be called for isp and the sensor if the other stream is still enabled. Remove the callback function scheme for navigating through the topology, simplifying the code, improving readability. Remove multistreaming item from the TODO list. Signed-off-by:
Dafna Hirschfeld <dafna.hirschfeld@collabora.com> Signed-off-by:
-
Dafna Hirschfeld authored
On 's/try_fmt' ioctl, set the quantization field to V4L2_QUANTIZATION_DEFAULT. This fixes the compliance errors: fail: v4l2-test-formats.cpp(357): quantization >= 0xff fail: v4l2-test-formats.cpp(473): testColorspace(!node->is_io_mc, pix_mp.pixelformat, pix_mp.colorspace, pix_mp.ycbcr_enc, pix_mp.quantization) test VIDIOC_TRY_FMT: FAIL fail: v4l2-test-formats.cpp(357): quantization >= 0xff fail: v4l2-test-formats.cpp(473): testColorspace(!node->is_io_mc, pix_mp.pixelformat, pix_mp.colorspace, pix_mp.ycbcr_enc, pix_mp.quantization) test VIDIOC_S_FMT: FAIL Signed-off-by:
-
Dafna Hirschfeld authored
Add "WITH Linux-syscall-note" to the uapi header SPDX comment. Signed-off-by:
-
Dafna Hirschfeld authored
There are several 'if' expression where double parentheses is used when one is enough. Remove the extra parentheses. Signed-off-by:
-
Dafna Hirschfeld authored
There are several 'if' conditions of the form: if (!!(module_ens & SOME_FLAG)) Those can be replaced with: if (module_ens & SOME_FLAG) Signed-off-by:
-
Dafna Hirschfeld authored
The params and stats entities have a field 'is_streaming'. This field is not needed since the entities can have available buffers only if they stream and therefore it is enough to check if there are buffers available. As a result, their start_stream callbacks can be removed since they only set the 'is_streaming' field. Signed-off-by:
-
Dafna Hirschfeld authored
The buffer lists of stats and params entities are initialized in queue_setup callback with 'INIT_LIST_HEAD'. It is enough to initialize the lists only upon registration. For the stats entity the list is already initialize upon registration. Signed-off-by:
-
Dafna Hirschfeld authored
Currently the code uses 'list_cut_position' to move the buffers to a temporary list. Replace it with 'list_splice_init'. This is nicer since we don't need to access params.prev. Also, replace INIT_LIST_HEAD with the simpler LIST_HEAD. Signed-off-by:
-
Dafna Hirschfeld authored
In function rkisp1_vb2_start_streaming, the call to media_pipeline_start should be the first thing in order to validate the links and prevents their state from being modified before power up and streaming. Adjust stop streaming to the same logic, call media_pipeline_stop after we disable streaming on the entities in the topology. Signed-off-by:
-
Hans Verkuil authored
The vivid driver is a bit flaky w.r.t. the kthread timing, esp. when running inside a virtual machine. This is caused by calling schedule_timeout_uninterruptible(1) which can actually take more than one jiffie. A while loop with schedule() turns out to be a lot more precise. Also, if mutex_trylock() fails, then just call schedule() instead of schedule_timeout_uninterruptible(1). There is no need to wait until the next jiffer, just schedule(), then try to get the lock again. This is still not precise enough, it is still relatively easy to get missed frames. This really should be converted to use a proper timer, but for now this solves the worst problems. Signed-off-by:
-
Colin Ian King authored
There is a spelling mistake in a pci_err error message. Fix this and make the error message a little more meaningful. Signed-off-by:
-
Hans Verkuil authored
The description of the V4L2_COLORSPACE_470_SYSTEM_BG stated that it was superseded by SMPTE 170M. That is incorrect. The probably root cause of this is that the HDMI standard does not support this colorspace and, unless otherwise signaled, will fall back to SMPTE 170M for SDTV. However, EBU Tech. 3321 states that sources should signal Rec. 709 as the colorimetry when using HDMI since the difference between Rec. 709 and Tech. 3213 are negligible. Update the text accordingly. Also drop a spurious " at the end of the Tech 3213 title in the bibliography. Signed-off-by:
-
