kNFS troubles (2.2.14pre15, Mandrake 7.0)

[Jens Benecke]

Hi,

since I updated my NFS server to the new Mandrake 7.0 distro I cannot run a
single KDE application any more. Strangely, only KDE apps are affected (so
far). I'll just forward the details I posted to the KDE mailing list.

The problem occurs whenever a KDE app tries to access its configuration or
starts a file open dialog. It hangs with CPU stuck at about 40% on both
server and client. strace and tcpdump logs are below.

Whose fault is it? knfsd, or KDE/Qt?

I'll happily give anyone a login to debug the problem.


Thanks for any help!


----------------------------------------------------------------------

Server:	2.2.14-15, Mandrake 7.0, nfs-utils-1.5.1-1mdk, glibc-2.1.2-9mdk
Linux pinguin.whm.tu-harburg.de 2.2.13-7mdk #1 Wed Sep 15 18:02:18 CEST 1999 i586 unknown

Client:	see above
Linux ds9.zuhause.de 2.2.14pre13 #1 Die Dez 14 19:48:25 CET 1999 i586 unknown
kdelibs-1.1.2-7mdk, kdebase-1.1.2-9mdk, qt-1.44-12mdk


The fstab line on the client looks like this:
192.168.1.100:/home     /home   nfs exec,nodev,nosuid,rw,rsize=16384,wsize=16384,hard,intr 1 0
 

This is the tcpdump.

SERVER:
23:33:17.208311 ds9.zuhause.de.2454775921 > server.zuhause.de.nfs: 160 lookup [|nfs]
23:33:17.208733 server.zuhause.de.nfs > ds9.zuhause.de.2454775921: reply ok 128 lookup [|nfs]
23:33:17.209006 ds9.zuhause.de.2471553137 > server.zuhause.de.nfs: 164 lookup [|nfs]
23:33:17.226905 server.zuhause.de.nfs > ds9.zuhause.de.2471553137: reply ok 128 lookup [|nfs]
23:33:17.227201 ds9.zuhause.de.2488330353 > server.zuhause.de.nfs: 164 lookup [|nfs]
23:33:17.227513 server.zuhause.de.nfs > ds9.zuhause.de.2488330353: reply ok 128 lookup [|nfs]
23:33:17.227888 ds9.zuhause.de.2505107569 > server.zuhause.de.nfs: 164 read [|nfs]
23:33:17.230519 server.zuhause.de.nfs > ds9.zuhause.de.2505107569: reply ok 112 read [|nfs]
23:33:23.622595 ds9.zuhause.de.2521884785 > server.zuhause.de.nfs: 164 lookup [|nfs]
23:33:23.622981 server.zuhause.de.nfs > ds9.zuhause.de.2521884785: reply ok 128 lookup [|nfs]

CLIENT:
23:33:21.160583 ds9.zuhause.de.2454775921 > earth.zuhause.de.nfs: 160 lookup [|nfs]
23:33:21.161161 earth.zuhause.de.nfs > ds9.zuhause.de.2454775921: reply ok 128 lookup [|nfs]
23:33:21.161278 ds9.zuhause.de.2471553137 > earth.zuhause.de.nfs: 164 lookup [|nfs]
23:33:21.179344 earth.zuhause.de.nfs > ds9.zuhause.de.2471553137: reply ok 128 lookup [|nfs]
23:33:21.179476 ds9.zuhause.de.2488330353 > earth.zuhause.de.nfs: 164 lookup [|nfs]
23:33:21.179951 earth.zuhause.de.nfs > ds9.zuhause.de.2488330353: reply ok 128 lookup [|nfs]
23:33:21.180161 ds9.zuhause.de.2505107569 > earth.zuhause.de.nfs: 164 read [|nfs]
23:33:21.182945 earth.zuhause.de.nfs > ds9.zuhause.de.2505107569: reply ok 112 read [|nfs]
23:33:27.575956 ds9.zuhause.de.2521884785 > earth.zuhause.de.nfs: 164 lookup [|nfs]
23:33:27.576488 earth.zuhause.de.nfs > ds9.zuhause.de.2521884785: reply ok 128 lookup [|nfs]


I chose to strace krdb because it's the first app that is started by the
startkde script (and it hangs).

krdb STRACE:
read(6, 0x40508000, 4096)               = -1 EPERM (Operation not permitted)
fstat(6, {st_mode=S_IFREG|S_ISGID|0640, st_size=76, ...}) = 0
read(6, 0x40508000, 4096)               = -1 EIO (Input/output error)
fstat(6, {st_mode=S_IFREG|S_ISGID|0640, st_size=76, ...}) = 0
read(6, 0x40508000, 4096)               = -1 EPERM (Operation not permitted)
fstat(6, {st_mode=S_IFREG|S_ISGID|0640, st_size=76, ...}) = 0
read(6, 0x40508000, 4096)               = -1 EIO (Input/output error)
fstat(6, {st_mode=S_IFREG|S_ISGID|0640, st_size=76, ...}) = 0
read(6, 0x40508000, 4096)               = -1 EPERM (Operation not permitted)
fstat(6, {st_mode=S_IFREG|S_ISGID|0640, st_size=76, ...}) = 0
read(6, 0x40508000, 4096)               = -1 EIO (Input/output error)
fstat(6, {st_mode=S_IFREG|S_ISGID|0640, st_size=76, ...}) = 0
read(6, 0x40508000, 4096)               = -1 EPERM (Operation not permitted)
fstat(6, {st_mode=S_IFREG|S_ISGID|0640, st_size=76, ...}) = 0
read(6, 0x40508000, 4096)               = -1 EIO (Input/output error)

Is it trying to open/read a suid file (which one) and getting IO errors? I can
do that with the shell perfectly...

> As an aside, I've gone back to userland nfs from knfsd because of some
> problems with submounts (using knfsd) that I'd consider fairly severe.

I have been experiencing crashes and instabilities and "D" state hung NFS
servers over five nfsd releases and at least twenty different kernel releases
(including special Alan Cox patches etc) with the user space daemon, so I
decided to switch. NFS is very important here and our server did not survive
longer than a week without several Oops messages, and that was simply not
acceptable.

I am using the Mandrake 6.1 distro now with kernel based NFS. It's Redhat
based if that helps you. :)
 

Thanks for your help!
 

-- 
_ciao, Jens_______________________________ http://www.pinguin.conetix.de
·
"Please don't tell my parents I work for Microsoft. They still think I'm a
bartender in a gay brothel."				-- Usenet



-- 
_ciao, Jens_______________________________ http://www.pinguin.conetix.de
·
"[Microsoft] ... guarantees 99.8% NT uptime for certain hard-/software.
 That's exactly the 3 minutes daily that my NT server needs to reboot."
							-- ZDnet editorial

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