2003-02-11 22:33:40

by Steven Cole

Subject: 2.5.60 problems with dbench and unkillable processes.

With 2.5.60, dbench starts, prints "1 clients started", but never prints
any "..." and doesn't finish.

With 2.5.59-mm10, dbench runs normally.

I ran this:
ps -ewo user,pid,priority,%cpu,stat,command,wchan
with the 2.5.60 System.map copied to /boot, and got this:

root 1103 15 0.1 S /usr/sbin/sshd schedule_timeout
steven 1104 15 0.0 S -bash wait4
steven 1142 15 0.0 S ./dbench 1 wait4
steven 1143 16 0.0 S ./dbench 1 pause
steven 1153 16 0.0 S /bin/sh /home/st wait4
steven 1154 16 0.0 R ps -ewo user,pid -


I control-c'ed the dbench 1, and started ./dbench 4, and got this:

root 1103 15 0.0 S /usr/sbin/sshd schedule_timeout
steven 1104 15 0.0 S -bash wait4
steven 1143 16 0.0 S ./dbench 1 pause
steven 1160 15 0.0 S ./dbench 4 wait4
steven 1161 16 0.0 S ./dbench 4 pause
steven 1162 16 0.0 S ./dbench 4 pause
steven 1163 16 0.0 S ./dbench 4 pause
steven 1164 16 0.0 S ./dbench 4 pause
steven 1166 16 0.0 S /bin/sh /home/st wait4
steven 1167 17 0.0 R ps -ewo user,pid -

That process 1143 is now unkillable with kill -9.

After control-c-ing the ./dbench 4, I now have 4
dbench 4 processes which are unkillable with kill -9.

The system is 2-way PIII, kernel is SMP, PREEMPT.

Steven

2003-02-11 22:38:05

by Linus Torvalds

Subject: Re: 2.5.60 problems with dbench and unkillable processes.

On 11 Feb 2003, Steven Cole wrote:
>
> With 2.5.60, dbench starts, prints "1 clients started", but never prints
> any "..." and doesn't finish.

Yeah, it's fixed in the current BK tree, and I already posted a patch to
linux-kernel for non-BK users. Sorry,

Linus