<!-- Some styling for better description lists --><style type='text/css'>dt { font-weight: bold;float: left;display:inline;margin-right: 1em} dd { display:block; margin-left: 2em}</style>
***: ChanServ sets mode: +v mchehab
mkrufky: ``` (06:58:18 AM) Sigyn: Invitation denied, there are only 61 users in #linuxtv: contact staffers if needed. ```
<br> ?!?
***: ChanServ sets mode: +v snawrocki
mkrufky: meh, looks like im going to have to get staff involved
hverkuil: hi guys!
mkrufky: hi hans :-)
syoung: Hello
sailus: Hello!
<br> Just pinged pinchartl...
hverkuil: <u>mchehab</u>: ping
sailus: <u>hverkuil</u>: Wasn't mchehab travelling these days?
<br> Did he say whether he'd join?
hverkuil: good question
<br> He's spending two weeks at the brazil office.
<br> but should be able to join the meeting AFAICT.
syoung: During last week's meeting: 12:03 <+mchehab> not sure yet if I'll be able to join for the meetings but I'll likely find a way
<br> I'm guessing he did not find a way :/
sailus: It's ten past. I guess we could start.
hverkuil: I don't have much to report. I've been working on this: http://people.canonical.com/~ubuntu-security/cve/2017/CVE-2017-13166.html
<br> Expect a bunch of patches soon. Besides fixing that CVE I also found a ton of bugs in the compat-ioctl32 code.
sailus: Gosh.
<br> I should expect to see a flow of patches to stable kernels then?
hverkuil: We'll see. That's a lot of work.
sailus: That seems pretty bad, doesn't it?
hverkuil: I think it's a real security risk that is exploitable. The patch Daniel Mentz came up with has been there for almost half a year, but nobody could be bothered to inform upstream. That really pisses me off.
mchehab: sorry, I had some issues to connect today
sailus: <u>hverkuil</u>: Agreed.
mchehab: yep, they should have pinged us
<br> <u>hverkuil</u>: the better would be to apply those patchesets for 4.16, of course c/c stable
<br> from my side, I'm still stuck with the softirq issue
<br> trying to setup a way to test it at the office
hverkuil: right now they only apply to 4.15 and up.
mchehab: the previous patchset didn't work. A new one was submitted today
<br> <u>hverkuil</u>: well, let's solve 4.15+ first
<br> then work on backporting it
hverkuil: yup
mchehab: at least the CVE ones should be send to earlier Kernels
<br> we should try to simplify the compat32 on V4L
<br> i have the feeling that it is more complex than it should be
<br> as even compatible ioctls should be listed there
hverkuil: the problem is that v4l2-compat-ioctl32.c changes a lot, so for almost every kernel you'd need to do some manual work.
mchehab: that's due to that: all IOCTLS are listed there
hverkuil: No, only ioctls that need conversion are listed.
<br> It used to be that all ioctls were listed, but that was changed some time ago.
mchehab: ah, OK!
hverkuil: Anyway, because of this I didn't have time to do any patch handling.
mchehab: ok
<br> yeah, this has priority over normal patch handling
<br> sailus, mkrufky, pinchartl, snawrocki, syoung: anything from your side?
hverkuil: vivid + v4l2-compliance have been incredibly useful to test this. I get good coverage that way.
<br> (ioctl coverage)
mchehab: good!
sailus: <u>mchehab</u>: Thanks for pulling the patches. I think that was it. I'll postpone all non-fix patches until we have rc1.
mchehab: ok
<br> my guess is that the merge window will start next week
sailus: On media device refcounting --- I've done some more work on it but lately I've been stuck debugging a crash.
<br> I've converted some existing drivers for refcounting and I've done testing with a rtl28xxu stick.
mchehab: ok
sailus: That's media device refcounting only; the DVB framework appears to have similar issues but also makes an attept to work with them by waiting until all users are gone, it seems.
<br> I'll review Alexandre's request patches in the near future.
mchehab: dvb refcounting used to work fine, until 2016
<br> it is more complex than V4L, as, on DVB, there is a thread that starts running when frontend is opened
sailus: Some of that seems fine whereas elsewhere there's an apparent assumption of a use case such as a stateless video codec.
mchehab: the refcount logic there has to wait for all users to go
sailus: It's not refcounting, but waiting. :-)
<br> That was another approach considered for media device btw., and still needed in some cases, for example to shut down hardware.
hverkuil: <u>pinchartl</u>: can you comment on https://www.mail-archive.com/linux-media@vger.kernel.org/msg125078.html?
mchehab: <u>sailus</u>: v4l2 has a waiting logic too for some structs
hverkuil: omap_vout.c is badly broken. Perhaps we should retire that driver?
mchehab: anything else?
hverkuil: not from me
mchehab: (with regards to omap_vout, if it is broken, I'm OK to retire it)
<br> send a RFC
<br> if none complains, it is probably safe to retire it
hverkuil: will do
<br> I finished my last tests for the CVE fix. I'll post the patch series tomorrow morning.