"tail /dev/zero" hangs 2.4.19-pre10aa2 but not -pre10 and -pre10ac2

From: harisri (harisri@telstra.com)
Date: Mon Jun 17 2002 - 01:48:29 EST


Hello,

tail /dev/zero (ran as a normal user) hangs 2.4.19-pre10aa2 but not
-pre10 and -pre10ac2.

In case of -pre10aa2:

The vmstat output just before it started killing processes, and
later hang (ie, the vmstat itself hang):
   procs memory swap io system
cpu
 r b w swpd free buff cache si so bi bo in cs us
sy id
 0 0 0 0 92428 5420 14752 0 0 331 118 146 116 16
13 71
 0 0 0 0 92276 5568 14752 0 0 0 53 109 6 0
0 100
 0 0 0 0 91264 5652 14752 0 0 0 23 114 19 1
0 99
 0 0 0 0 90760 5864 14808 0 0 8 58 113 16 0
1 99
 2 1 1 40836 1528 1512 5016 78 5789 83 5802 241 250 19
12 69
 1 0 1 101168 1912 1512 5012 95 7930 95 7935 245 267 10
13 77
 2 0 2 152560 1500 1512 5012 62 8098 62 8098 243 200 10
21 70

With greater difficulty I was able to login (it killed the telnet
session many times, at various stages).

Some sample messages from dmesg:
__alloc_pages: 0-order allocation failed (gfp=0x1d2/0)
VM: killing process tail
__alloc_pages: 0-order allocation failed (gfp=0x1d2/0)

Multiply the "__alloc_pages" error message by 100 times or so.

And,
VM: killing process smbd

In the same way it killed run-crons, login, in.telnetd, tcpd, inetd and
X.

Though it killed the root cause - "tail" command, commands like
ps/top/reboot etc hang, vmstat got killed initially and hung later, su
got killed, sync works fine. I am unable to switch between virtual
terminals at the console using ctrl alt F1/F2.

In case of -pre10 and -pre10ac2:
It thrashes initially for about 10 secs (approx) and then everything is
back to normal. There is a single line of "Out of Memory: killed process
403 (tail)" (of course the process id of tail is different between
-pre10 and -pre10ac2). And the system comes back to normal state.

This is a:
Pentium - II @ 333 MHz
128 MB RAM,
150 MB swap,
gcc-2.95.2,
suse 7.1

I am not subscribed to lkml, please cc me if you can, else I will refer
the web archive at marctheaimsgroup.com

Thanks
Hari
harisri@bigpond.com

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Sun Jun 23 2002 - 22:00:12 EST