- 28 Jul, 2023 26 commits
-
-
Sakari Ailus authored
When an async sub-device is registered, it used to be that the first one of its connections were matched when found. Continue looking for matches until a notifier no longer has any. Signed-off-by:
Sakari Ailus <sakari.ailus@linux.intel.com> Tested-by: Philipp Zabel <p.zabel@pengutronix.de> # imx6qp Tested-by: Niklas Söderlund <niklas.soderlund@ragnatech.se> # rcar + adv746x Tested-by: Aishwarya Kothari <aishwarya.kothari@toradex.com> # Apalis i.MX6Q with TC358743 Tested-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com> # Renesas RZ/G2L SMARC Signed-off-by:
Mauro Carvalho Chehab <mchehab@kernel.org>
-
Sakari Ailus authored
The list entry is initialised as a head in v4l2_async_register_subdev() just before being added to the list. This isn't needed, drop the initialisation. Signed-off-by:
-
Sakari Ailus authored
When the v4l2-async framework was introduced, the use case for it was to connect a camera sensor with a parallel receiver. Both tended to be rather simple devices with a single connection between them. The framework has been since improved in multiple ways but there are limitations that have remained, for instance the assumption an async sub-device is connected towards a single notifier and via a single link only. This patch enables connecting a sub-device to one or more notifiers simultaneously, with one or more connections per notifier. The notifier information is moved from the sub-device to the connection and the connections in sub-device are no longer a pointer but a linked list. Signed-off-by:
-
Sakari Ailus authored
Add v4l2_async_connection_unique() function for obtaining a struct v4l2_async_connection, typically allocated by drivers together with their own information on an external sub-device. The relation between connections and sub-devices still remains 1:1 but this code becomes more complex when the relation soon changes. Signed-off-by:
-
Sakari Ailus authored
This patch re-arranges internal V4L2 async lists for preparation of supporting multiple connections per sub-device as well as cleaning up used lists. The list of unbound V4L2 sub-devices shall be maintained for the purpose of listing those sub-devices only, not for their bindin status. Also, the V4L2 async connections now have, instead of two list entries, a single list entry in the notifier's list, be that either waiting or done lists, while the notifier's asc_list is removed. The one-to-one relation between a sub-device and a connection is still maintained in this patch. Signed-off-by:
-
Sakari Ailus authored
The connections are checked for duplicates already when the notifier is registered. This is effectively a sanity check for driver (and possibly obscure firmware) bugs. Don't do this when adding the connection. Retain the int return type for now. It'll be needed very soon again. Signed-off-by:
-
Sakari Ailus authored
Add labels for error handling instead of doing it all in individual cases. Prepare for more functionality in this function. Signed-off-by:
Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com> Tested-by: Aishwarya Kothari <aishwarya.kothari@toradex.com> # Apalis i.MX6Q with TC358743 Tested-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com> # Renesas RZ/G2L SMARC Signed-off-by:
-
Sakari Ailus authored
Rename v4l2_async_subdev as v4l2_async_connection, in order to differentiate between the sub-devices and their connections: one sub-device can have many connections but the V4L2 async framework has so far allowed just a single one. Connections in this context will later translate into either MC ancillary or data links. This patch prepares changing that relation by changing existing users of v4l2_async_subdev to switch to v4l2_async_connection. Async sub-devices themselves will not be needed anymore Additionally, __v4l2_async_nf_add_subdev() has been renamed __v4l2_async_nf_add_connection(). Signed-off-by:
-
Sakari Ailus authored
V4L2 async sub-device matching originally used the device nodes only. Endpoint nodes were taken into use instead as using the device nodes was problematic for it was in some cases ambiguous which link might have been in question. There is however no need to use endpoint nodes on both sides, as the async sub-device's fwnode can always be trivially obtained using fwnode_graph_get_remote_endpoint() when needed while what counts is whether or not the link is between two device nodes, i.e. the device nodes match. This will briefly break the adv748x driver but it will be fixed later in the set, by patch "media: adv748x: Return to endpoint matching". Signed-off-by:
-
Sakari Ailus authored
The naming of list heads and list entries is confusing as they're named similarly. Use _list for list head and _entry for list entries. Signed-off-by:
-
Sakari Ailus authored
Pass only information required for sub-device matching to functions checking whether the async sub-device already exists. Do the same for debug message printing. This makes further changes to other aspects of async sub-devices easier. Accordingly, also perform further renames: asd_equal as v4l2_async_match_equal, v4l2_async_nf_has_async_subdev as v4l2_async_nf_has_async_match, __v4l2_async_nf_has_async_subdev as v4l2_async_nf_has_async_subdev_entry and v4l2_async_nf_asd_valid as v4l2_async_nf_match_valid. Signed-off-by:
-
Sakari Ailus authored
The async match type is a struct field now, rename V4L2_ASYNC_MATCH_* macros as V4L2_ASYNC_MATCH_TYPE_* instead. This patch has been produced by: git grep -l V4L2_ASYNC_MATCH_ -- drivers/media/ drivers/staging/media/ \ include/ Documentation/|xargs perl -i -pe \ 's/V4L2_ASYNC_MATCH_\K/TYPE_/g' so it must be correct. Signed-off-by:
-
Sakari Ailus authored
Make V4L2 async match information a struct, making it easier to use it elsewhere outside the scope of struct v4l2_async_subdev. Also remove an obsolete comment --- none of these fields are supposed to be touched by drivers. Signed-off-by:
-
Sakari Ailus authored
The callers do pass a non-NULL asd to v4l2_async_nf_asd_valid() already. There's no need for the NULL check here. Signed-off-by:
-
Sakari Ailus authored
There's a need to verify that a single async sub-device isn't being added multiple times, this would be an error. This takes place at the time of adding the async sub-device to the notifier's list as well as when the notifier is added to the global notifier's list. Use the pointer to the sub-device for testing this instead of an index to an array that is long gone. (There was an array of async sub-devices in the notifier before it was converted to a linked list by commit 66beb323 ("media: v4l2: async: Remove notifier subdevs array").) Signed-off-by:
-
Sakari Ailus authored
Just add some debug prints for V4L2 async sub-device matching process. These might come useful in figuring out why things don't work as expected. Signed-off-by:
-
Sakari Ailus authored
Document v4l2_async_nf_cleanup() which must be called before releasing an unregistered notifier's memory. Also remove the sentence regarding v4l2_async_nf_init() arguments --- those are documented in kerneldoc which is referred here. Signed-off-by:
-
Sakari Ailus authored
Also v4l2_async_nf_add_fwnode() may be used to add an async sub-device descriptor to a notifier. Document this. Also remove a redundant sentence. Signed-off-by:
-
Sakari Ailus authored
Document that the notifier of an async sub-device is, besider unregistered, also cleaned up using v4l2_async_unregister_subdev(). Signed-off-by:
-
Sakari Ailus authored
Add section titles for async documentation. While the documentation is mostly fine as-is, it has grown from its original state but remains without internal structure. Add it now. Also remove an extra newline. Signed-off-by:
-
Sakari Ailus authored
An example adding an async sub-device to a V4L2 async notifier. The name of the variable in error handling was wrong (asd vs. my_asd). Signed-off-by:
-
Sakari Ailus authored
Move the creation of the links between external sub-devices and ISP sub-devices to the bound callback. This way we can also remove the need to access the sub-device's notifier field that will soon be removed. Signed-off-by:
-
Sakari Ailus authored
The async notifier bound callback does a lot of checks that have probably been always unnecessary. Remove the lookup of the async subev that we already have, as well as the debug print that is already printed by the framework. Signed-off-by:
-
Sakari Ailus authored
Remove an unneeded forward declaration for struct v4l2_async_subdev. Signed-off-by:
-
Sakari Ailus authored
Remove an unneeded declaration for struct fwnode_handle. Signed-off-by:
-
Jacopo Mondi authored
The v4l2_async_nf_parse_fwnode_endpoints() function, part of v4l2-fwnode.c, was a helper meant to register one async sub-dev for each fwnode endpoint of a device. The function is marked as deprecated in the documentation and is actually not used anywhere anymore. Drop it and remove the helper function v4l2_async_nf_fwnode_parse_endpoint() from v4l2-fwnode.c. This change allows to make the helper function __v4l2_async_nf_add_connection() visibility private to v4l2-async.c so that there is no risk drivers can mistakenly use it. [Sakari Ailus: Small fixups on top.] Signed-off-by:
Jacopo Mondi <jacopo.mondi@ideasonboard.com> Signed-off-by:
-
- 27 Jul, 2023 1 commit
-
-
Nícolas F. R. A. Prado authored
Commit fe8a3397 ("media: mediatek: vcodec: Read HW active status from syscon") allowed the driver to read the VDEC_SYS io space from a syscon instead of from the reg property when reg-names are supplied. However as part of that change, a smatch warning was introduced: drivers/media/platform/mediatek/vcodec/mtk_vcodec_dec_drv.c:142 mtk_vcodec_get_reg_bases() error: buffer overflow 'mtk_dec_reg_names' 11 <= 11 With a correct Devicetree, that is, one that follows the dt-binding, it wouldn't be possible to trigger such a buffer overflow. Even so, update the range validation of the reg property, so that the smatch warning is fixed and if an incorrect Devicetree is ever supplied the code errors out instead of causing memory corruption. Reported-by:
Hans Verkuil <hverkuil-cisco@xs4all.nl> Closes: https://lore.kernel.org/all/b5fd2dff-14a5-3ad8-9698-d1a50f4516fa@xs4all.nl Fixes: fe8a3397 ("media: mediatek: vcodec: Read HW active status from syscon") Reviewed-by:
AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com> Signed-off-by:
Nícolas F. R. A. Prado <nfraprado@collabora.com> Signed-off-by:
-
- 26 Jul, 2023 1 commit
-
-
Hans Verkuil authored
Document missing fields. This resolves two warnings: drivers/media/platform/qcom/venus/core.h:226: warning: Function parameter or member 'venus_ver' not described in 'venus_core' drivers/media/platform/qcom/venus/core.h:501: warning: Function parameter or member 'enc_state' not described in 'venus_inst' Signed-off-by:
Dikshita Agarwal <quic_dikshita@quicinc.com>
-
- 25 Jul, 2023 12 commits
-
-
Dan Carpenter authored
If we encounter any error in the vdec_msg_queue_init() then we need to set "msg_queue->wdma_addr.size = 0;". Normally, this is done inside the vdec_msg_queue_deinit() function. However, if the first call to allocate &msg_queue->wdma_addr fails, then the vdec_msg_queue_deinit() function is a no-op. For that situation, just set the size to zero explicitly and return. There were two other error paths which did not clean up before returning. Change those error paths to goto mem_alloc_err. Fixes: b199fe46 ("media: mtk-vcodec: Add msg queue feature for lat and core architecture") Fixes: 2f5d0aef ("media: mediatek: vcodec: support stateless AV1 decoder") Signed-off-by:
Dan Carpenter <dan.carpenter@linaro.org> Reviewed-by:
Nicolas Dufresne <nicolas.dufresne@collabora.com> Signed-off-by:
-
Dan Carpenter authored
The "lat_buf->private_data" needs to be set to NULL to prevent a double free. How this would happen is if vdec_msg_queue_init() failed twice in a row and on the second time it failed earlier than on the first time. The vdec_msg_queue_init() function has a loop which does: for (i = 0; i < NUM_BUFFER_COUNT; i++) { Each iteration initializes one element in the msg_queue->lat_buf[] array and then the clean up function vdec_msg_queue_deinit() frees each element of the msg_queue->lat_buf[] array. This clean up code relies on the assumption that every element is either initialized or zeroed. Leaving a freed pointer which is non-zero breaks the assumption. Fixes: b199fe46 ("media: mtk-vcodec: Add msg queue feature for lat and core architecture") Signed-off-by: -
Irui Wang authored
"fb_use_list" is used to store used or referenced frame buffers for vp9 stateful decoder. "NULL" should be returned when getting target frame buffer failed from "fb_use_list", not a random unexpected one. Fixes: f77e8985 ("[media] vcodec: mediatek: Add Mediatek VP9 Video Decoder Driver") Signed-off-by:
Irui Wang <irui.wang@mediatek.com> Reviewed-by:
-
Xiaoyong Lu authored
Fix av1 decode fail when iova is 36bit. Decoder hardware will access incorrect iova address when tile buffer is 36bit, it will lead to iommu fault when hardware access dram data. Fixes: 2f5d0aef ("media: mediatek: vcodec: support stateless AV1 decoder") Signed-off-by: Xiaoyong Lu<xiaoyong.lu@mediatek.com> Signed-off-by:
-
Nícolas F. R. A. Prado authored
Remove the requirement of a VDEC_SYS reg iospace for both MT8173 and MT8183. To achieve that, rely on a vdecsys syscon to be passed through the DT, and use it to directly read the VDEC_HW_ACTIVE bit during IRQ handling to check whether the HW is active. Also update the VP8 stateful decoder to use the syscon, if present, for writes to VDEC_SYS. The old behavior is still present when reg-names aren't supplied, as to keep backward compatibility. Signed-off-by:
-
Nícolas F. R. A. Prado authored
The VDEC_HW_ACTIVE bit is located at offset 0, bit 4 of the VDECSYS iospace. Only the mask was previously defined, with the address being implicit. Explicitly define the address, and append a '_MASK' suffix to the mask, to make accesses to this bit clearer. This commit brings no functional change. Signed-off-by:
-
Nícolas F. R. A. Prado authored
The binding expects the first register space to be VDEC_SYS. However this register space is already assigned to a different node on both MT8173 and MT8183: a clock-controller node called 'vdecsys' which is also a syscon. In order to resolve the overlapping address ranges, remove the VDEC_SYS register space from the video decoder, and add a new property to hold the phandle to the syscon, so that iospace can still be handled. Also add reg-names to be able to tell that this new register schema is used, so the driver can keep backward compatibility. Signed-off-by:
Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> Signed-off-by:
-
Nícolas F. R. A. Prado authored
On MT8183 it's not necessary to configure the parent for the clocks. Remove the assigned-clocks and assigned-clock-parents from the required list. Signed-off-by:
Matthias Brugger <matthias.bgg@gmail.com> Reviewed-by:
-
Nícolas F. R. A. Prado authored
MT8173 and MT8183 have different clocks, and consequently clock-names. Relax the number of clocks and set clock-names based on compatible. Reviewed-by:
-
Dan Carpenter authored
This is intended to print the error code but there is a typo so it prints IS_ERR() instead of PTR_ERR(). Fixes: 77f3b023f452 ("media: mediatek: vcodec: Add debugfs interface to get debug information") Signed-off-by: -
Dan Carpenter authored
The list iterator "dbgfs_inst" is always non-NULL. This means that the test for NULL inside the loop is unnecessary and it also means that the test for NULL outside the loop will not work. If we do not find the item on the list with the correct the ctx_id then it will free invalid memory leading to a crash. Fixes: cd403a6a0419 ("media: mediatek: vcodec: Add a debugfs file to get different useful information") Signed-off-by: -
Ming Qian authored
the supported_instance_count determine the instance index range, it shouldn't exceed the bits number of instance_mask, otherwise the bitops of instance_mask may cross boundaries Fixes: 9f599f35 ("media: amphion: add vpu core driver") Reviewed-by:
Ming Qian <ming.qian@nxp.com> Signed-off-by:
-
