Subject: An extra datapoint re hangs
To: None <port-arm26@netbsd.org>
From: Gavan Fantom <gavan@coolfactor.org>
List: port-arm26
Date: 12/03/2001 03:35:49
I've just experimented with an extra file in /etc/rc.d which just does an
"ls /" and an "ls /dev"

"ls /" works fine, "ls /dev" hangs the machine. tcpdump shows the
following:

[...]
03:00:31.972324 dhcp-99.oblivion.net.2965251291 > tetra.oblivion.net.nfs:
112 lookup fh 0,6/1931 "malloc.conf"
03:00:31.973276 tetra.oblivion.net.nfs > dhcp-99.oblivion.net.2965251291:
reply ok 28 lookup ERROR: No such file or directory
03:00:31.973651 tetra.oblivion.net.nfs > dhcp-99.oblivion.net.2965251290:
reply ok 96 write
03:00:32.007493 dhcp-99.oblivion.net.2965251292 > tetra.oblivion.net.nfs:
96 getattr fh 0,6/1931
03:00:32.007841 tetra.oblivion.net.nfs > dhcp-99.oblivion.net.2965251292:
reply ok 96 getattr DIR 40755 ids 0/0 sz 5632
03:00:32.015509 dhcp-99.oblivion.net.2965251293 > tetra.oblivion.net.nfs:
96 getattr fh 0,6/1931
03:00:32.015848 tetra.oblivion.net.nfs > dhcp-99.oblivion.net.2965251293:
reply ok 96 getattr DIR 40755 ids 0/0 sz 512
03:00:32.027270 dhcp-99.oblivion.net.2965251294 > tetra.oblivion.net.nfs:
96 getattr fh 0,6/1931
03:00:32.027566 tetra.oblivion.net.nfs > dhcp-99.oblivion.net.2965251294:
reply ok 96 getattr DIR 40755 ids 0/0 sz 5632
03:00:32.034667 dhcp-99.oblivion.net.2965251295 > tetra.oblivion.net.nfs:
92 fsstat fh 0,6/1931
03:00:32.034984 tetra.oblivion.net.nfs > dhcp-99.oblivion.net.2965251295:
reply ok 48 fsstat [|nfs]
03:00:32.040815 dhcp-99.oblivion.net.2965251296 > tetra.oblivion.net.nfs:
104 readdir fh 0,6/1931 8192 bytes @ 0
03:00:32.055959 tetra.oblivion.net.nfs > dhcp-99.oblivion.net.2965251296:
reply ok 1472 readdir offset 1 size 5952  eof (frag 54754:1480@0+)
03:00:32.056607 tetra.oblivion.net > dhcp-99.oblivion.net: (frag
54754:1480@1480+)
03:00:32.057850 tetra.oblivion.net > dhcp-99.oblivion.net: (frag
54754:1480@2960+)
03:00:32.059095 tetra.oblivion.net > dhcp-99.oblivion.net: (frag
54754:1480@4440+)
03:00:32.060340 tetra.oblivion.net > dhcp-99.oblivion.net: (frag
54754:1480@5920+)
03:00:32.061584 tetra.oblivion.net > dhcp-99.oblivion.net: (frag
54754:720@7400)

The machine hangs at this point.

So it seems to me as if readdir causes problems on directories over a
certain size. I'm compiling a new kernel with DDB in the hope that there
is a key combination to drop into DDB on arm26, and also that putting in
DDB doesn't prevent a multi-user boot due to lack of memory.

-- 
Gillette - the best a man can forget