#linux-media 2021-07-23,Fri

↑back Search ←Prev date Next date→ Show only urls(Click on time to select a line by its url)

WhoWhatWhen
***camus1 has joined #linux-media
camus has quit IRC (Read error: Connection reset by peer)
[00:08]
................................... (idle for 2h51mn)
camus has joined #linux-media [03:00]
camus1 has quit IRC (Ping timeout: 480 seconds) [03:05]
camus1 has joined #linux-media
camus has quit IRC (Remote host closed the connection)
[03:17]
.......... (idle for 49mn)
NiksDev has quit IRC (Ping timeout: 480 seconds)
uajain_ is now known as uajain
[04:06]
......... (idle for 44mn)
camus1 has quit IRC (Remote host closed the connection)
camus has joined #linux-media
[04:52]
..... (idle for 24mn)
miqztee has joined #linux-media [05:16]
........... (idle for 53mn)
camus has quit IRC (Remote host closed the connection)
camus has joined #linux-media
[06:09]
............ (idle for 56mn)
ao2 has joined #linux-media [07:06]
..... (idle for 21mn)
camus1 has joined #linux-media
camus has quit IRC (Ping timeout: 480 seconds)
ao2 has quit IRC (Quit: Leaving)
[07:27]
............... (idle for 1h11mn)
miqztee has quit IRC (Ping timeout: 480 seconds) [08:41]
..... (idle for 20mn)
djrscally has joined #linux-media [09:01]
......................... (idle for 2h0mn)
BrianG61UK_ has quit IRC (Read error: Connection reset by peer) [11:01]
BrianG61UK_ has joined #linux-media [11:10]
.......... (idle for 49mn)
camus has joined #linux-media
camus1 has quit IRC (Ping timeout: 480 seconds)
[11:59]
mort_Hey, so my topology is like this: https://p.mort.coffee/PO4 (graph: https://p.mort.coffee/yWK.png). Everything which is connected to /dev/video3 and has a resolution is set to 1280x960. Specifically, both "ov5645 4-003b" and "msm_vfe0_pix" is 1280x960.
However, vidioc_g_fmt returns 1920x1080; https://github.com/torvalds/linux/blob/master/drivers/media/platform/qcom/camss/camss-video.c#L678 - active_fmt.fmt.pix_mp.width/height is 1920 and 1080
any clue why that might be?
[12:16]
pinchartlhave you called VIDIOC_S_FMT(1280x960) on the video node ? [12:22]
mort_No. But wouldn't it default to its configured format?
And where's the 1080p coming from in the first place?
fwiw, `v4l2-compliance -d /dev/video3 --streaming` doesn't VIDIOC_S_FMT either and hence it fails
[12:29]
welp, if I `v4l2-ctl --set-fmt-video width=blah,height=blah` first, v4l2-compliance works. I can always make that run at boot right after the media-ctl command. But I feel like this is a symptom of a misconfiguration on my end which should be fixed. I may be wrong though, maybe setting the format with v4l2-ctl is the correct way to do it [12:36]
.... (idle for 18mn)
pinchartl1080p is the driver's default. you need to set a format if you want to override it [12:54]
mort_ok, so I've looked into it a bit more
there's one application which does VIDIOC_S_FMT to 540x351 and then starts streaming with that format; that used to work in an older version of the driver, but now it doesn't because, for whatever reason, the VIDIOC_S_FMT doesn't affect the resolution returned by video_get_subdev_format, even though it affects the active_fmt
[12:55]
pinchartlformats on subdevs and video nodes are set independently [12:58]
mort_quite possible that the code was always wrong and the new driver just checks more stuff, but it's weird that active_fmt changes in apparently illegal ways [12:58]
pinchartlthey need to match, but it's userspace's responsibility to set them [12:58]
mort_interesting
so the VIDIOC_S_FMT should actually be accompanied by some media-ctl command?
(or whatever ioctl media-ctl does under the hood)
[12:58]
pinchartlVIDIOC_SUBDEV_S_FMT on the subdev nodes, yes [12:59]
mort_sounds like the safest approach might be to just set the format of the device and the subdev once at the same time during boot, and then write code which andles whichever resolution VIDIOC_G_FMT gives [13:03]
..... (idle for 23mn)
pinchartlI'd recommend writing an application that will set all formats, to achieve the desired output resolution
or, better, using libcamera :-)
[13:26]
mort_libcamera would probably be a good idea [13:30]
pinchartlthere's support for the camss driver, but I haven't tested it personally [13:32]
mort_we end up using the raw video4linux says all API a lot anyways though due to using hardware encoders/decoders which are exposed as v4l devices
(IIRC ffmpeg supports hardware decoders exposed as v4l devices, but doesn't support hardware encoders, and the decoder was buggy and produced green frames; otherwise we would've used ffmpeg)
[13:33]
***miqztee has joined #linux-media
miqztee has quit IRC ()
[13:46]
.... (idle for 19mn)
camus1 has joined #linux-media [14:06]
camus has quit IRC (Ping timeout: 480 seconds) [14:11]
........... (idle for 50mn)
camus1 has quit IRC () [15:01]
......................... (idle for 2h3mn)
gouchi has joined #linux-media [17:04]
............. (idle for 1h2mn)
ndufresnemort_: msm venus driver was buggy yes (racy in many ways), hence why it failed sometimes with gst and ffmpeg, I haven't rested recently
but it should be way better
we support both encoder and decoders in gst, ffmpeg v4l2 encoder support shouldn't be hard to add really, it's just a very small amount of v4l2 code there btw
[18:06]
......................... (idle for 2h2mn)
***ao2 has joined #linux-media [20:08]
.............. (idle for 1h9mn)
gouchi has quit IRC (Remote host closed the connection) [21:17]
..... (idle for 21mn)
ao2 has quit IRC (Remote host closed the connection) [21:38]
...................... (idle for 1h48mn)
djrscally has quit IRC (Ping timeout: 480 seconds) [23:26]

↑back Search ←Prev date Next date→ Show only urls(Click on time to select a line by its url)