Re: dump(8) deadlock (ULE bug?)

From: Doug White <dwhite_at_gumbysoft.com>
Date: Mon, 8 Mar 2004 18:18:26 -0800 (PST)
On Sat, 6 Mar 2004, Lukas Ertl wrote:

> Hi,
>
> I have this filesystem:
>
> Filesystem  1K-blocks    Used     Avail Capacity  Mounted on
> /dev/da0s1g  16244334  992310  13952478     7%
>
> and I'm trying to dump it like:
>
> dump -0uf - /dev/da0s1g > news.dump

$PWD isn't on /devad0s1g, is it?

> The problem is that dump(8) seems to run into a deadlock or race in Pass
> IV [regular files], because then the dump procs all go into state "paused"
> and don't do anything anymore.
>
>  1080 root      20    0  2116K  1664K pause  0   0:04  0.00%  0.00% dump
>  1082 root      20    0  2116K  1664K pause  3   0:04  0.00%  0.00% dump
>  1081 root      20    0  2116K  1664K pause  2   0:04  0.00%  0.00% dump
>  1079 root       4    0  2244K  1704K sbwait 3   0:03  0.00%  0.00% dump

pause means the process has called pause(3) and is waiting for a signal.
Considering the parent is in sbwait, which is a network send, it appears
to be waiting for network I/O.  Why it would be doing network I/O, I don't
know; I'd suggest tcpdump and/or ktrace.

I don't know if dump is SIGIO based or not.

-- 
Doug White                    |  FreeBSD: The Power to Serve
dwhite_at_gumbysoft.com          |  www.FreeBSD.org
Received on Mon Mar 08 2004 - 17:18:26 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:37:46 UTC