| ↑back Search ←Prev date Next date→ Show only urls | (Click on time to select a line by its url) |
| Who | What | When | |
|---|---|---|---|
| sailus | neg: Could you check my set "Unified fwnode endpoint parser" as well, please? | [08:10] | |
| ..... (idle for 20mn) | |||
| neg | sailus: will do | [08:30] | |
| ................... (idle for 1h34mn) | |||
| brisquit | Hi, I'm having an issue with an application I'm developing where it stops working after I've run some other application using v4l2. So I was thinking that I'm probably forgetting to set/clear some state/format... I ran "media-ctl -p" and compared, but the output was identical before and after... Does anybody have any pointers on how to debug further?
VIDIOC_STREAMON fails with EINVAL after I've run the other application | [10:04] | |
| sailus | Can you enable driver and media-entity.c debug logs if you have dynamic debug enabled in the kernel, and recheck?
It usually tells exactly where things go wrong. | [10:08] | |
| brisquit | sailus: I haven't used dynamic debug before... There's no file named /sys/kernel/debug/dynamic_debug/control which means it's probably not enabled? | [10:15] | |
| sailus | Probably not. You can also check if you can find CONFIG_DYNAMIC_DEBUG in your kernel configuration which is typically under /boot.
You can also try fishing this from the format configuration on the video nodes. That's commonly checked against the media bus formats but not printed by media-ctl -p. E.g. yavta can print format information. | [10:20] | |
| hverkuil | echo 1 >/sys/module/videobuf2_v4l2/parameters/debug
ditto for videobuf2_core Should give the reason for the einval in the kernel log. (videobuf2_v4l2 may not exist if you have an old kernel) | [10:24] | |
| sailus | hverkuil: If it's because of the IOCTL parameters. That doesn't seem like the case to me here, as it's another application changing something that breaks things. | [10:27] | |
| hverkuil | streamon fails, so it should report why it fails. Should go a long way towards figuring out what the other app changes. | [10:27] | |
| neg | hverkuil: I know this have been asked before and that due to lack of VIDIOC_QUERYCAP its not currently supported but is there any plans to add subdev operations to v4l-utils? Right now this prevents querying/setting std paramaters and programing edid of subdevices which are part of a media pipeline without hacks :-) | [10:30] | |
| hverkuil | neg: yes, there are plans. I hope to know more in 1-2 weeks. | [10:34] | |
| neg | hverkuil: OK thanks, then I won't spend a lot of time to make my hacks pretty untill I know more :-) | [10:36] | |
| .......... (idle for 45mn) | |||
| brisquit | hverkuil: https://pastebin.com/raw/H6d1iPvd | [11:21] | |
| hverkuil | brisquit: which kernel version? | [11:27] | |
| brisquit | hverkuil: Linux m3ulcb 4.6.0-yocto-standard #1 SMP PREEMPT Wed Feb 8 13:39:31 CET 2017 aarch64 GNU/Linux | [11:30] | |
| *** | pocek has quit IRC (Ping timeout: 255 seconds) | [11:32] | |
| hverkuil | brisquit: which driver are you using?
I don't see any message from the vb2 core, so I wonder if it is in the driver that -EINVAL is returned. | [11:35] | |
| brisquit | hverkuil: vsp2 | [11:43] | |
| hverkuil | if the vsp2 is anything like the vsp1, then there is a vsp2_video_streamon function that does lots of things and is probably returning the EINVAL.
You'll have to debug in there. probably a format mismatch in the video pipeline. | [11:45] | |
| brisquit | hverkuil: maybe a stupid question, but what's a good way to debug the driver? | [11:58] | |
| kbingham | brisquit: I'm quite sure the VSP2 is using the vsp1 driver. So you'll be looking at drivers/media/platform/vsp1/* | [12:12] | |
| pinchartl | kbingham: more than sure indeed
brisquit: unless you're using the driver from the BSP ? | [12:12] | |
| kbingham | pinchartl: ^ That was why I was only 'quite' sure rather than certain :)
on an m3ulcb ... brisquit: vsp1_video_streamon() has a few error paths that could return -EINVAL - in particular -vsp1_video_verify_format() would be a likely cause. Add some printk statements in there to see if those error paths are being taken. | [12:13] | |
| brisquit | well, there is a vsp2 kernel module loaded. | [12:17] | |
| pinchartl | brisquit: all my condolences
that's a BSP driver I wouldn't touch if with a 10-feet pole | [12:18] | |
| brisquit | pinchartl: probably this one then? https://github.com/renesas-rcar/vsp2driver/blob/76a87c8c9e8e14f4ae8a2733a4f3408ba7c8db4b/vsp2driver/vsp2_video.c#L933
could I use the vsp1 driver instead? | [12:22] | |
| pinchartl | brisquit: most likely. I'd recommend using the vsp1 driver, yes. you will likely need to backport the latest version, as the v4.6.0 driver didn't support Gen3
(at least not properly) then it also depends on how you use it from userspace. I don't know if the BSP driver contains features not available in the mainline driver | [12:24] | |
| kbingham | pinchartl: Yes, quick glances show that it does. They have added some basic support for fences for example.
Otherwise, they do a lot of 'cherry-picking' from our commits, and adapting to the vsp2 naming :) | [12:36] | |
| pinchartl | if I remember correctly, they also have some kind of "vsp manager" to multiplex access to the vsp | [12:38] | |
| kbingham | yes - I see mention of a vspm as well :) | [12:38] | |
| ............. (idle for 1h4mn) | |||
| mort | Hey. I'm using the qualcomm venus h264 encoder, and need to know its output's h264 profile and level.
I'm not entirely sure what that means, but webrtc expects me to specify a profile and level. apparently ffprobe can tell you what profile/level a h264 stream is, this encoder produces video with profile=High and level=50 | [13:42] | |
| pH5 | mort: the venus driver implements the V4L2_CID_MPEG_VIDEO_H264_PROFILE and V4L2_CID_MPEG_VIDEO_H264_LEVEL controls
those should tell you what profile and level the encoder is currently producing | [13:48] | |
| mort | alright, thanks | [13:50] | |
| ....................... (idle for 1h50mn) | |||
| Mason385 | Hello, I'm trying to use an IR RC driver, and it seems I am missing something
for example, ir-keytable complains "Couldn't find any node at /sys/class/rc/rc0/input0/event*."
| [15:40] | |
| ...... (idle for 26mn) | |||
| syoung: indeed I am missing +CONFIG_INPUT_EVDEV=y ... retesting now :-) | [16:08] | ||
| ...... (idle for 28mn) | |||
| how can the driver support concurrently NEC, RC5 and RC6? When the IR sensor receives a signal, the protocol can be deduced from the frequencies used? | [16:36] | ||
| .................. (idle for 1h29mn) | |||
| syoung | Mason385: IR is just infrared light which is turned on and off at different intervals. See http://www.sbprojects.com/knowledge/ir/nec.php and http://www.sbprojects.com/knowledge/ir/rc5.php for example | [18:05] | |
| Mason385: so different decoders can try to decode the IR at the same time | [18:15] | ||
| ↑back Search ←Prev date Next date→ Show only urls | (Click on time to select a line by its url) |