Bug 79078

Summary:	Machine lock up when using NFS
Product:	[Retired] Red Hat Linux	Reporter:	Mark <mark.piercewright>
Component:	nfs-utils	Assignee:	Dave Jones <davej>
Status:	CLOSED WONTFIX	QA Contact:	Ben Levenson <benl>
Severity:	high	Docs Contact:
Priority:	medium
Version:	7.3	CC:	pfrields
Target Milestone:	---
Target Release:	---
Hardware:	i586
OS:	Linux
Whiteboard:
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2004-11-27 23:15:29 UTC	Type:	---
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:

Description Mark 2002-12-05 13:42:25 UTC

From Bugzilla Helper:
User-Agent: Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 4.0; PANCREDITv5.0; 
PANCREDIT5.0; .NET CLR 1.0.3705)

Description of problem:
I have two Linux boxes.  A single processor p133 on RH7.1 and a dual processor 
p133 running RH7.3.  A job on the RH7.1 box mount a directory on the RH7.3 SMP 
box and copies any changed files, then unmounts.  At some point, the RH7.3 SMP 
box locks up, no console no ping no ftp or telnet.

Version-Release number of selected component (if applicable):


How reproducible:
Always

Steps to Reproduce:
1.mount a directory from RH71 to RH73 SMP
2.use a "cp -pu" to copy changed files
3.wait for the target machine to die
	

Actual Results:  After a period of time, could be 5 minutes could be 20, the 
SMP box ceases to respond to all external stimuli.

Expected Results:  The job should finish after an hour or so and all is well 
with the world.

Additional info:

There does not appear to be an error message, I think it is such a complete 
halt that Linux is not getting chance to write an error to the message log.
If the NFS services are disabled then the machine seems to stay up quite 
happily.  The copy job has worked without modification for many months when the 
target machine was a RH6.2 single processor box.

Comment 1 Arjan van de Ven 2002-12-05 13:50:35 UTC

question: what exact kernel is running on the hanging box?
also can you try adding "nmi_watchdog=1" to the kernel commandline of the
hanging box (that's the vmlinuz line in /boot/grub/grub.conf). Doing this will
make the kernel try to detect deadlocks and print a backtrace if it detects one.

Comment 2 Mark 2002-12-05 14:15:57 UTC

Thanks for such a rapid response.  The machines are at home, I will provide the 
requested details tomorrow.

Comment 3 Mark 2002-12-06 13:04:52 UTC

Here is my grub.conf
=========================================
default=0
timeout=10
splashimage=(hd0,0)/grub/splash.xpm.gz
password --md5 $1$cF\V1.BG$556CULbW1ITuEhGZ4.gLO1
title Red Hat Linux (2.4.18-3smp)
        root (hd0,0)
        kernel /vmlinuz-2.4.18-3smp ro root=/dev/hda6 noapic
        initrd /initrd-2.4.18-3smp.img
title Red Hat Linux-up (2.4.18-3)
        root (hd0,0)
        kernel /vmlinuz-2.4.18-3 ro root=/dev/hda6
        initrd /initrd-2.4.18-3.img
=========================================
I tried running with nmi-watch=1
Dec  5 18:22:27 puntus kernel: Default MP configuration #6
Dec  5 18:22:27 puntus kernel: Processor #0 Pentium(tm) APIC version 16
Dec  5 18:22:27 puntus kernel: Processor #1 Pentium(tm) APIC version 16
Dec  5 18:22:27 puntus kernel: I/O APIC #2 Version 16 at 0xFEC00000.
Dec  5 18:22:27 puntus kernel: Processors: 2
Dec  5 18:22:27 puntus kernel: Kernel command line: ro root=/dev/hda6 noapic
nmi_watchdog=1
Dec  5 18:22:27 puntus kernel: Initializing CPU#0
Dec  5 18:22:27 puntus kernel: Detected 133.610 MHz processor.
 ++++++++++++++++++++++
I was tailing the messages file when it crashed
gort.squirrelsoft:1005 for /backup (/backup)
Dec  5 18:28:32 puntus rpc.mountd: authenticated mount request from
gort.squirrelsoft:1008 for /backup (/backup)
Dec  5 18:28:51 puntus rpc.mountd: authenticated mount request from
gort.squirrelsoft:762 for /backup (/backup)
Dec  5 18:38:53 puntus su(pam_unix)[1702]: session opened for user news by
(uid=0)
Dec  5 18:38:54 puntus su(pam_unix)[1702]: session closed for user news

+++++++++++++++++++++++++++++++++++++++

No trace was found (where would it be?)

I also ran in single processor mode but it still hung.

Comment 4 Arjan van de Ven 2002-12-06 13:26:44 UTC

hmm
noapic and the nmi watchdog are exclusive ;(

Comment 5 Mark 2002-12-06 13:32:43 UTC

If i don't run noapic, i wold normally get lost interupt errors relating to the 
hard disks.  Will nmi watchdog still stop that happining or am I stucj now?

Comment 6 Mark 2002-12-12 11:34:10 UTC

I have given up on this one.  I have reverted to the old single processor box.  
Thanks for your time but let's not waste any more effort.  Cheers, Mark.