From: Jeff Dike <jdike@ka...> - 2001-06-15 15:00:52
> i seem to be able to get process to hang quite alot in the D state
> #4 0x1001c9bd in ___wait_on_page (page=0x50025740) at filemap.c:641
> #5 0x1001d286 in do_generic_file_read (filp=0x50d4c674, ppos=0x50d4c694,
The last time this happened, it was only fixed after I started being able to
> (gdb) print io_count
> $12 = 30674
> and there have been 30674 ubd irqs
At least it's different this time. Last time, the io count would have been 1
less than the irq count.
From: James Stevenson <mistral@st...> - 2001-06-15 16:25:36
i have had it a few more times since.
what sort of stuff would be usefull to look for ?
> OK, this just happened to me, so it's getting some attention now :-)
Mobile: +44 07779080838
5:20pm up 4:19, 4 users, load average: 1.36, 1.33, 1.30
From: Jeff Dike <jdike@ka...> - 2001-06-15 18:10:33
> i have had it a few more times since. what sort of stuff would be
> usefull to look for ?
I don't know yet. It only happened once, and I didn't learn much.
The IO apparently did happen, but somehow the process that was waiting for it
never woke up.
Also, it seems that any further IO to the block device never happens. This
seems to be the biggest clue so far.