[11:56] <mchehab> hi all
[11:57] <hverkuil> Hi Mauro!
[11:59] <mchehab> hi Hans
[12:02] <mchehab> syoung: I'm ok on merging the fix from Akihiro
[12:02] <mchehab> sailus, mkrufky, pinchartl: ping
[12:03] <syoung> hi!
[12:03] <syoung> mchehab: thanks!
[12:03] <syoung> so I'll send a PR
[12:03] <mchehab> OK
[12:03] <mkrufky> hello
[12:04] <hverkuil> While waiting for the others I can give a quick update on Dafna's Outreachy project.
[12:04] <mchehab> I'll handle fixup stuff today - and add those typo fixes I sent after rebased (the documentation one will likely won't apply as-is, as there were some documentation file split)
[12:04] <hverkuil> The project ends on Monday, but she managed to finish the stateless decoder support (and is working on the stateless encoder, but that won't be finished)
[12:05] <hverkuil> All in all she did a great job, and I'm very pleased with it.
[12:05] <mchehab> great, good to know!
[12:05] <mchehab> yeah, I saw her patches floating around the ML - didn't review them, though
[12:06] <hverkuil> Can you take care of this one as well: https://patchwork.linuxtv.org/patch/54737/
[12:06] <hverkuil> Fixes annoying imx7 sparse/smatch warnings.
[12:06] <mchehab> ok
[12:06] <mchehab> send a PR, please
[12:06] <hverkuil> I will probably wait with making a pull request for the vicodec stuff until rc1 is released.
[12:07] * mchehab has been busy those days... still doing paperwork and needing to do some internal stuff that depends on vim2m
[12:07] <mchehab> for testing purposes
[12:08] <mchehab> I'll likely merge one of the vim2m patches I sent today... it fixes a buffer overflow condition
[12:08] <mchehab> s/today/this week/
[12:09] <mchehab> this one: https://patchwork.linuxtv.org/patch/54719/
[12:11] <hverkuil> I'll review that one. It's a bit weird.
[12:11] <mchehab> please do that...
[12:12] <mchehab> basically vim2m allows setting any size between 32x32 to 640x480 for capture or output
[12:12] <hverkuil> I'm still working on the media regression tests. Lots of little nasty issues that surface once you start testing corner cases :-(
[12:12] <mchehab> in thesis, a real driver would scale... vim2m just crops the data, but it does it on a bad way, allowing to write past the buffer
[12:12] <hverkuil> I was wondering why the plane_size check in vim2m_buf_prepare was removed.
[12:13] <mchehab> (it could be a regression I introduced, or it could be something that it is there since the beginning)
[12:13] <hverkuil> The log didn't mention anything about that.
[12:13] <mchehab> yeah, I avoided mentioning buffer overflow or other things that would rise security issues
[12:14] <mchehab> it is removed because it now checks lines and widths directly instead of buffer size
[12:14] <mchehab> it is possible to use, for example, out buffer 32x32 and cap buffer 640x480
[12:15] <mchehab> (it will show a very small image, and all black on the rest of the buffer)
[12:15] <mchehab> or 640x480 out buffer, 32x32 cap buffer
[12:15] <mchehab> (it will crop the image to 32x32)
[12:15] <hverkuil> But the check in vim2m_buf_prepare happens during VIDIOC_QBUF, so the applications gets an error directly. Doing it in the process step means it is too late and userspace gets no feedback that the buffer it queued was too small.
[12:15] <mchehab> there's no need to check... it will work
[12:16] <mchehab> (it is a "poor man" scaler, that doesn't scale)
[12:17] <hverkuil> If the format is set to e.g. 640x480, then the buffer you queue should be sized to handle 640x480, since that is what will be returned. Nothing to do with different formats between capture and output.
[12:19] <hverkuil> You remove that check, so you can pass a buffer that is smaller than sizeimage (using DMABUF or USERPTR streaming modes)
[12:19] <mchehab> hmm... doesn't VB2 already do such check?
[12:20] <mchehab> I mean: doesn't VB2 already checks it?
[12:20] <hverkuil> no. It doesn't know sizeimage, that's something only the driver knows.
[12:21] <mchehab> ah, this is VB2 buf_prepare... yeah, you're right... that code can't be removed. I'll send a version 3 without that
[12:21] <hverkuil> Let me review the patch first, I might have more concerns.
[12:21] <mchehab> ok, sure
[12:22] <mchehab> somehow, I mis-read that logic that it would be comparing the capture/output buffers - I guess there used to be such code on a past version
[12:23] <mchehab> any other thing that should be considered before the merge window?
[12:24] <mchehab> btw, I have another thing for discussions at #media-security
[12:24] <syoung> mchehab: I'm banned
[12:25] <sailus> Hello!
[12:25] <mchehab> let me unban
[12:25] <hverkuil> I don't have anything else at the moment.
[12:25] <sailus> Good timing. :-)
[12:25] <sailus> Not that it would have been intended. Apologies for being late.
[12:26] <sailus> I've reworked the ImgU warning fix set and Raj has done *some* testing on it, will send v2 soonish.
[12:27] <sailus> I'm also been thinking of how to improve V4L2 fwnode API for devices with multiple ports. Patches will follow...
[12:31] <mchehab> syoung: try again
[12:31] <mchehab> sailus: good
[12:37] <mchehab> sailus, syoung: did you try joining #media-security?
[12:38] <syoung> 12:38 -!- Cannot join to channel #media-security (You are banned)
[12:39] <sailus> ... Out of curiosity, what is #media-security?
[12:40] <mchehab> syoung: are you registered?
[12:43] * mchehab needs to see how can I disable autoban
[12:43] <mchehab> and why /mode -b is not working
[12:44] <mchehab> try again
[12:45] <mchehab> syoung: ^
[12:59] <hverkuil> mchehab: posted pull request for imx7 sparse/smatch fix.
[12:59] <mchehab> ok, thanks!
[14:09] <mchehab> hverkuil: see my reply to your comments on vim2m patch
[14:47] *** ChanServ sets mode: +v mchehab`