Hanging processes on disk I/O, kernel 2.1.35 etc

Stephen Coffin (scoffin@netcom.com)
Wed, 23 Apr 1997 07:45:50 -0700

>> I didn't have quite as much luck as you had. It patched cleanly into the
>> kernel for me, but this morning after it had been running all night I
>> typed sync, it synced but didn't ever return. I switched to another tty
>> and killall sync just hung also, I switched to another tty kill -9 the
>> killall and that worked, but I couldn't kill -9 the sync process. I
>> changed to root and started killing process to reboot and a good many of
>> them would turn into zombie processes and a few would would exit
>> correctly.

I am having a similar problem with all kernels 2.1.29 - 2.1.35. 2.1.29 was
the oldest 2.1 kernel I have played with. On my machine, the symptom
is hanging processes that are trying to do disk I/O (just a feeling, but
it seems that output to disk causes the hang, not input from disk....
(no proof for this, just a hunch) The hanging processes cannot be killed
with "kill -9", but the machine as a whole seems to stay up....

Another curious fact is that the "Load Average" seems to get very large
when one or more of these hung processes is around....

Of course, 2.0.x kernels do not have this problem (I usually use 2.0.26)....

>> Tyan Tomcat III, k5, ncr53c810, 64megs memory, (gcc driver version
>> Objective-C snapshot 960906 executing gcc version

P6/200 with Aurora motherboard, 2940 PCI SCSI controller with my only
disk on the 2940, Diamond 2001 video board.... I did not select any
modules in my kernel config, everything is either IN or OUT :-)

Version Info:
Linux yawnxx 2.1.35 #11 Mon Apr 21 14:36:38 MDT 1997 i686 unknown
Kernel modules 2.1.34
Gnu C 2.7.2p
Linux C Library 5.4.23
Dynamic Linker (ld.so) 1.8.9
Linux C++ Library 27.1.4
Procps 1.01
Mount 2.6g
Net-tools 1.1.79
Kbd 0.89
Sh-utils 1.16

If any of the kernel or driver team needs help to debug this problem, I can
demonstrate it well on my machine :-) Send me email if you wish
to use my machine or if I can help in any way....

ps= Is it possible I am doing something simple and wrong??