[08:10] <lyakh> hverkuil: morning, you weren't answering while I was away? [08:32] <hverkuil> lyakh: No, you left the room, so there was no point :-) [08:33] <hverkuil> Topics would depend on who would be able to join the conference. First I need to see if there are enough devs to make it worthwhile. [08:34] <lyakh> I might be able to join, if topics are relevant :-) E.g. see my mail yesterday [08:34] <lyakh> but we did already discuss that [08:34] <lyakh> so, not sure it would be appropriate to discuss that again [08:35] <lyakh> point is I'm under some pressure to push that forward relatively quickly [08:35] <hverkuil> metadata, right? [08:36] <lyakh> yep [08:36] <lyakh> which might have some cross-references with the request API [08:44] *** stdint has quit IRC (Quit: Leaving) [08:50] <pinchartl> hverkuil: it also depends on how much time we'll have I suppose [08:50] <pinchartl> the main topic is to agree on a way forward for MC lifetime management [08:50] <pinchartl> if we have time left, I'd really like to discuss metadata [08:50] <pinchartl> but I fear we won't have time left [08:51] <hverkuil> Note that if there aren't enough devs to make a summit worthwhile, then Mauro probably won't attend the ELC. At least, that's what he told me. [08:52] <pinchartl> ok [08:52] <hverkuil> Right now we don't have enough people. [08:53] <pinchartl> how many people do we need to make it worth it ? :-) [09:57] <hverkuil> pinchartl: sailus: preparing for tomorrow and looking into the kref issue: [09:57] <hverkuil> so if you unbind a driver, then remove() is called immediately, and after remove() exits any devm_alloc'ed memory is released, right? [09:58] <hverkuil> And that happens even if an application is in the middle of an ioctl? [10:02] <pinchartl> hverkuil: we're preparing for the meeting too :-) but heading for lunch now. I'll reply right after [10:02] <pinchartl> or actually I can reply now [10:02] <pinchartl> yes, that's correct [10:02] <pinchartl> now I'm off to lunch :-) [10:02] <hverkuil> bon appetit! [10:14] <kbingham> Can't the ioctls take a reference to the devm objects to persist their lifetime? [10:17] <hverkuil> kbingham: no, it's an unconditional release of resources. [10:17] <hverkuil> (just checked the code) [10:18] <hverkuil> insane IMHO [10:18] <kbingham> hverkuil: Wheres the relevant place? (OOI) [10:18] <hverkuil> drivers/base/devres.c, devres_release_all() [10:18] <hverkuil> Called from drivers/base/dd.c, __device_release_driver() [10:23] <kbingham> hverkuil: thanks. So refcounted drivers then :D hehe [10:23] <kbingham> On the whole (excluding this flaw) devm_ seems to be quite useful for keeping resource allocations clean... [10:24] <kbingham> anyway - have fun in your meeting tomorrow. I don't think I can make it :D It's a long walk [10:30] <hverkuil> lyakh: will you be at the ELC, or does that depend on whether there is a media summit or not? [10:30] <lyakh> hverkuil: the latter, yes [11:26] * pinchartl is back [16:55] *** benjiG has left [21:42] *** awalls1 has left [22:20] *** neg has quit IRC (Read error: Connection reset by peer)