| soren | Keybuk: I've got defunct process with ppid 1.. :( Anything you want me to do? | 12:06 |
|---|---|---|
| Keybuk | soren: ps output? | 12:14 |
| soren | All of it? Or just: soren 13177 0.1 0.0 0 0 ? Zl Nov23 5:31 [epiphany-browse] <defunct> | 12:15 |
| Keybuk | ps ajx for that pid | 12:16 |
| Keybuk | ps j -p 13177 | 12:18 |
| Keybuk | should do the trick | 12:18 |
| soren | 13177 7364 7364 ? Zl 5:31 [epiphany-browse] <defunct> | 12:19 |
| Keybuk | it's ppid isn't 1 then | 12:19 |
| Keybuk | ? | 12:19 |
| soren | That's what ps -efl says. | 12:19 |
| Keybuk | ps -j ! :) | 12:19 |
| soren | ps -j -p gives these headers: | 12:20 |
| Keybuk | no | 12:20 |
| Keybuk | ps j | 12:20 |
| soren | Gah. | 12:20 |
| Keybuk | ps j -p 13177 | 12:20 |
| Keybuk | ;) | 12:20 |
| soren | I'm an idiot. | 12:20 |
| soren | PPID PID PGID SID TTY TPGID STAT UID TIME COMMAND | 12:20 |
| soren | 1 13177 7364 7364 ? -1 Zl 1000 5:31 [epiphany-browse] <defunct> | 12:20 |
| Keybuk | and curse ps for having the most insanely bad command-line ever | 12:20 |
| Keybuk | hmm | 12:20 |
| Keybuk | if you kill -CHLD 1 does it go away? | 12:20 |
| soren | Wow.. No. | 12:21 |
| Keybuk | sudo initctl version | 12:21 |
| Keybuk | ? | 12:21 |
| Keybuk | (note: no --) | 12:21 |
| soren | Now init (upstart 0.3.9) | 12:21 |
| soren | Er.. -now :) | 12:22 |
| Keybuk | did it go away? | 12:22 |
| soren | No. | 12:22 |
| Keybuk | *shrug* | 12:22 |
| Keybuk | iz kernel boog | 12:22 |
| soren | What makes you say that? | 12:22 |
| soren | Surely, it's upstart's job to wait() for it? | 12:22 |
| Keybuk | yes | 12:22 |
| Keybuk | and you just did two things that guaranteed that upstart would have called wait() :P | 12:23 |
| soren | Eh? What was the other one? initctl version? | 12:23 |
| Keybuk | yeah | 12:23 |
| soren | *g* | 12:23 |
| soren | Why? | 12:23 |
| Keybuk | upstart always calls wait() in a loop inside its main loop | 12:23 |
| soren | Oh. | 12:23 |
| Keybuk | rather than in a signal handler | 12:23 |
| Keybuk | so you just did two things that would have definitely caused upstart to repeat its main loop | 12:24 |
| Keybuk | so definitely caused wait() to be called | 12:24 |
| Keybuk | which kernel version? | 12:24 |
| soren | Just once? Or until there aren't any more children? | 12:24 |
| Keybuk | until there aren't any more children | 12:24 |
| Keybuk | it calls waitid (..., WNOHANG) in a loop | 12:24 |
| soren | uname -a says: Linux butch 2.6.22-14-generic #1 SMP Sun Oct 14 21:45:15 GMT 2007 x86_64 GNU/Linux | 12:24 |
| soren | I may or may not have upgraded the kernel package since I booted, so that's the best I've got. | 12:25 |
| Keybuk | sudo stop tty1 | 12:25 |
| Keybuk | sudo start tty1 | 12:25 |
| soren | Same. | 12:25 |
| Keybuk | any getty processes now zombies? | 12:25 |
| soren | Nope | 12:25 |
| Keybuk | sudo status tty1 | 12:26 |
| Keybuk | then kill that pid | 12:26 |
| Keybuk | does it respawn? | 12:26 |
| soren | Yup | 12:26 |
| Keybuk | any zombie gettys? | 12:26 |
| soren | Nope. | 12:26 |
| Keybuk | epiphany still there? | 12:26 |
| soren | Yup | 12:27 |
| Keybuk | *shrug* | 12:27 |
| Keybuk | not my fault then ;) | 12:27 |
| soren | Gah.. | 12:27 |
| Keybuk | definitely a kernel issue | 12:27 |
| Keybuk | try sending SIGCONT to 13177 ? | 12:27 |
| soren | To a zombie? Er.. No. I can, though. | 12:28 |
| Keybuk | try ie | 12:28 |
| Keybuk | try it | 12:28 |
| soren | Same. | 12:28 |
| Keybuk | fair enough | 12:28 |
| soren | What effect could that possibly have? | 12:28 |
| Keybuk | -> #ubuntu-kernel :) | 12:28 |
| Keybuk | might have released an in-kernel lock | 12:28 |
| soren | Oh. | 12:28 |
| soren | The epiphany process has an npviewer.bin child process, but I doubt that can mess things up this badly. | 12:29 |
| Keybuk | well | 12:30 |
| Keybuk | that kinda proves the kernel is stuffed then | 12:30 |
| soren | That's a good point. | 12:30 |
| soren | :) | 12:30 |
| Keybuk | zombie processes can't, by definition, have children | 12:32 |
| soren | Precisely. | 12:32 |
| Keybuk | since when they exit, the children would have been reparented to init first | 12:33 |
| soren | I didn't think of that. | 12:33 |
| soren | Yeah. | 12:33 |
| soren | Ah... dmesg reveals trouble, too. | 12:33 |
| soren | E.g. "Fixing recursive fault but reboot is needed!" | 12:33 |
| soren | ...but I had the vmware server modules loaded, so I'm SOL. | 12:34 |
| soren | I'll just blame them. There. Problem solved. | 12:35 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!