| === maclin1 is now known as maclin | ||
| === maclin1 is now known as maclin | ||
| === JanC is now known as Guest5212 | ||
| === JanC_ is now known as JanC | ||
| === JanC_ is now known as JanC | ||
| ogra_ | slangasek, i wiped all calendar data from my phone and re-installed/re-created everything ... how does the TB meeting even end up in any way related to my calendar, i never added it, did someone in canonical do that ? | 14:21 |
|---|---|---|
| === asac` is now known as asac | ||
| zyga | mwhudson: hey | 19:45 |
| zyga | mwhudson: would you mind having a look at a patch for golang | 19:45 |
| mwhudson | zyga: um ok but i'm not really here (sunday and all) | 19:45 |
| zyga | mwhudson: haha, I didn't anticipate an instant reply ;) | 19:46 |
| zyga | :-) | 19:46 |
| mwhudson | heh | 19:46 |
| zyga | http://paste.ubuntu.com/24159510/ | 19:46 |
| mwhudson | zyga: seems like a good thing, although that really is a hacky workaround | 19:47 |
| mwhudson | zyga: you'll also need to cover the non-cgo case | 19:47 |
| zyga | mwhudson: where is that? | 19:47 |
| mwhudson | zyga: good question :) | 19:47 |
| zyga | mwhudson: I think that it is a hack but given the fact that we exec it cannot hurt | 19:47 |
| zyga | exec will replace all the threads | 19:48 |
| zyga | too bad the kernel doesn't say anything like that | 19:48 |
| zyga | EABOUTTOEXEC | 19:48 |
| zyga | I patched my kernel to see where this fails exactly | 19:48 |
| mwhudson | oh right yes | 19:48 |
| mwhudson | how does this manage to have bad effects other than the message? | 19:49 |
| zyga | and as some people anticipated it fails in kernel/fork.c:1031 | 19:49 |
| zyga | mwhudson: not clear because the thread just doens't get created | 19:49 |
| zyga | something doens't happen | 19:49 |
| zyga | I didn't run elaborate tests yet | 19:49 |
| mwhudson | the non-cgo case is src/runtime/os_linux.go:newosproc | 19:49 |
| zyga | thanks, checking | 19:49 |
| mwhudson | so yeah, add the non-cgo case and come up with a good explanation of why this is a problem :) | 19:50 |
| mwhudson | and it looks ok | 19:51 |
| zyga | mwhudson: does the execing sample attached in the commit message look sufficient? | 19:52 |
| mwhudson | zyga: so in the bad case, the exec does not happen? | 19:52 |
| zyga | mwhudson: and the non-cgo case, is that in os1_linux.go? | 19:52 |
| zyga | mwhudson: yes | 19:52 |
| zyga | mwhudson: run that in while ./bug; do :; done | 19:52 |
| zyga | and it returns quickly with | 19:53 |
| zyga | runtime/cgo: pthread_create failed: Resource temporarily unavailable | 19:53 |
| mwhudson | zyga: uh, maybe it's os1_linux.go in older go's, it's os_linux.go in tip | 19:53 |
| mwhudson | zyga: ah that does sound bad enough yeah | 19:53 |
| zyga | OK | 19:53 |
| zyga | will do | 19:53 |
| zyga | for you it's sunday right? | 19:53 |
| mwhudson | yeah | 19:53 |
| zyga | mwhudson: I'll try to patch that part too, never touched the go compiler | 19:56 |
| zyga | oh, nice sched_yield is already implemented under osyield | 20:01 |
| zyga | I noticed after I wrote the copy :) | 20:01 |
| zyga | I think I got it | 20:04 |
| jtaylor | hm what is the context? sched_yield should not be used | 20:38 |
| jtaylor | on linux | 20:38 |
| jtaylor | ah read up, you are only using it a couple times that is probably fine | 20:40 |
| jtaylor | but don't use it for busy looping | 20:40 |
| zyga | jtaylor: TBH I'm not yet sure what is the correct solution | 20:46 |
| zyga | jtaylor: the case is where one thread runs clone/pthread_create while another one runs exefc | 20:47 |
| zyga | exec* | 20:47 |
| zyga | mwhudson: it worked :) | 21:13 |
| zyga | mwhudson: but I think sched_yield is not a good idea indeed, the number of retries is unrelible as well | 21:13 |
| jtaylor | don't other programs run into this? or does go just produce a larger amount of threads than everything else? | 21:15 |
| zyga | jtaylor: go is special here, normally you don't just create threads in a program that just execs | 21:16 |
| zyga | jtaylor: but in go you do | 21:16 |
| zyga | http://paste.ubuntu.com/24160079/ | 21:16 |
| zyga | run this in a loopc | 21:16 |
| zyga | comment out the "C" import for another variant | 21:16 |
| zyga | while ./golang-bug; do :; done | 21:16 |
| zyga | I have two patches that retry pthread_create/clone up to 1000 times | 21:17 |
| zyga | I can still reproduce it but far far less frequently | 21:17 |
| zyga | while terrible I'm inclined to sleep :/ | 21:17 |
| zyga | or patch syscall | 21:18 |
| zyga | to detect exec and set a flag | 21:18 |
| jtaylor | might be something the kernel might be interested in, give userspace the ability to detect a real EAGAIN | 21:18 |
| zyga | jtaylor: it is a real EAGAIN | 21:18 |
| jtaylor | that one seems more like the typical EINTR case not one where retrying again will result in the same result | 21:18 |
| zyga | jtaylor: but ... we just care about one of those | 21:18 |
| zyga | jtaylor: not the resource contention EAGAINs | 21:19 |
| jtaylor | yes but EAGAIN can be a case where trying again won't help as you hit a limit | 21:19 |
| zyga | jtaylor: this EAGAIN is also undocumented | 21:19 |
| jtaylor | the others are not real EGAINs | 21:19 |
| zyga | jtaylor: well I think this is up to interpretation | 21:19 |
| zyga | jtaylor: but I worry that given how kernel development is done this will not change (userspace sees this) | 21:19 |
| jtaylor | yeah, but the kernels input on how this should be handled would be valuable | 21:19 |
| zyga | jtaylor: and even if, now you have to roll out a fixed kernel everywhere | 21:19 |
| zyga | jtaylor: yes, I agree | 21:19 |
| zyga | jtaylor: ideally golang would not do this but I think this is somewhat unavoidable | 21:20 |
| zyga | jtaylor: perhaps checkinng syscall.Exec is the way actually | 21:20 |
| zyga | jtaylor: just set a flag (retry on EAGAIN) | 21:20 |
| zyga | jtaylor: and use that flag to retry | 21:20 |
| jtaylor | zyga: maybe I don't now go to well to really help | 21:23 |
| jtaylor | my alarm clocks just rang on mention of sched_yield as that occurs in lots of old numerics software I work with, written back when sched_yield was more or less equivalent to pause and not what it is today | 21:24 |
| jtaylor | but used here is probably a reasonable workaround until a maybe better option if found | 21:25 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!