2.6.0-test3-bk10-final regression getting out of hand

Voluspa

author

2003-10-31 20:46 UTC #0 lore raw

As reported in:
http://marc.theaimsgroup.com/?l=linux-kernel&m=106195494828814&w=2
I'm experiencing a serious NFS regression since 2.6.0-test3-bk10-final (bk-set determined now). Since that, along the road to 2.6.0-test9, something has gotten completely out of hand.

cp -a /mnt/oden/etc .

_2.6.0-test3-bk10_
2 minutes

_2.6.0-test3-bk10-final_
3 minutes 47 seconds

_2.6.0-test9_
29 minutes 54 seconds

Nothing has changed in hardware or setup, and the cp tests were done today. Not being a programmer I tried to delete individual patches in the bk-set, but couldn't find the exact breakage point.

Erasing changes to:
linux-2.6.0-test3-bk10/fs/nfs/dir.c
No change.

Erasing changes to:
linux-2.6.0-test3-bk10/fs/nfs/nfsroot.c
No change.

Erasing changes to:
linux-2.6.0-test3-bk10/include/linux/sunrpc/timer.h linux-2.6.0-test3-bk10/net/sunrpc/timer.c No change.

Erasing changes to:
linux-2.6.0-test3-bk10/net/sunrpc/clnt.c
Doesn't compile.

Erasing changes to:
linux-2.6.0-test3-bk10/include/linux/sunrpc/xprt.h linux-2.6.0-test3-bk10/net/sunrpc/xprt.c Doesn't compile.

Using TCP or DIRECTIO in -test9 makes no difference. Here's the relevant .config-section:

#
# Network File Systems
#
CONFIG_NFS_FS=y
CONFIG_NFS_V3=y
# CONFIG_NFS_V4 is not set
# CONFIG_NFS_DIRECTIO is not set
CONFIG_NFSD=y
CONFIG_NFSD_V3=y
# CONFIG_NFSD_V4 is not set
# CONFIG_NFSD_TCP is not set
CONFIG_LOCKD=y
CONFIG_LOCKD_V4=y
CONFIG_EXPORTFS=y
CONFIG_SUNRPC=y
# CONFIG_SUNRPC_GSS is not set
# CONFIG_SMB_FS is not set
# CONFIG_CIFS is not set
# CONFIG_NCP_FS is not set
# CONFIG_CODA_FS is not set
# CONFIG_INTERMEZZO_FS is not set
# CONFIG_AFS_FS is not set

Mvh
Mats Johannesson

-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it
help you create better code? SHARE THE LOVE, and help us help
YOU! Click Here: http://sourceforge.net/donate/
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs

Martin Spott

2003-11-01 16:30 UTC #1 lore raw

Quoted 3 lines · Voluspa

I'm experiencing a serious NFS regression since 2.6.0-test3-bk10-final (bk-set determined now). Since that, along the road to 2.6.0-test9, something has gotten completely out of hand.

I can second the experiences on heavy trouble with 2.6.0-test[8,9] as NFS _client_ against 2.4.2[1,2] as a server. I didn't have different servers for testing against but I pretty much _believe_ it's a client issue.

I only sticked to 'official' 2.6.0-test patches and I'd like to express, that beginning with 2.6.0-test8 the NFS client is absolutely unsuable for me. Trying to copy (with 'cp') a file from a local filesystem to the server over NFS.V2 or .V3 completely freezes _any_ NFS traffic with the same server.
Are there any significant changes to the NFS client I should back out for the purpose of finding the 'bug' ?

Martin.

Signature · 12 lines

 Unix _IS_ user friendly - it's just selective about who its friends are !
--------------------------------------------------------------------------


-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
Does SourceForge.net help you be more productive?  Does it
help you create better code?   SHARE THE LOVE, and help us help
YOU!  Click Here: http://sourceforge.net/donate/
_______________________________________________
NFS maillist  -  [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs

Trond Myklebust

2003-11-04 06:35 UTC #2 lore raw

Quoted 1 line

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
" " == Martin Spott <[email protected]> writes:

> completely freezes _any_ NFS traffic with the same server. Are > there any significant changes to the NFS client I should back > out for the purpose of finding the 'bug' ?

Err... If you believe it is the client, they I suggest backing out _all_ client related patches between test7 and test8 for a start...

Personally, I'm starting to suspect some of the IPv4 changes. I'm seeing wierd crap with the TCP code which suggests some form of memory corruption. The corruption is occurring on both the client *and* server side.
The fact that I'm seeing signs of memory poisoning suggests it is probably an skb being freed too early or something like that.

Cheers,
Trond

-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it
help you create better code? SHARE THE LOVE, and help us help
YOU! Click Here: http://sourceforge.net/donate/
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs

Robert Williamson

2003-11-04 16:34 UTC #3 lore raw

In reply to Martin Spott

I've been seeing NFS client problems as well....for quite sometime. I too thought it was the NFS code, however further testing took me to possible problems with kernel memory code.

For more info on my problems checkout
http://bugme.osdl.org/show_bug.cgi?id=1097

-Robbie

Robert V. Williamson <[email protected]>
Linux Test Project
IBM Linux Technology Center
Web: http://ltp.sourceforge.net
IRC: #ltp on freenode.irc.net
====================
"Only two things are infinite, the universe and human stupidity, and I'm not sure about the former." -Albert Einstein

Martin Spott
<Martin.Spott@uni-du To: [email protected]
isburg.de> cc:
Sent by: Subject: Re: [NFS] 2.6.0-test3-bk10-final regression getting out of hand
[email protected]
ceforge.net

11/01/2003 10:29 AM

Quoted 3 lines · Voluspa

I'm experiencing a serious NFS regression since 2.6.0-test3-bk10-final (bk-set determined now). Since that, along the road to 2.6.0-test9, something has gotten completely out of hand.

I can second the experiences on heavy trouble with 2.6.0-test[8,9] as NFS _client_ against 2.4.2[1,2] as a server. I didn't have different servers for testing against but I pretty much _believe_ it's a client issue.

I only sticked to 'official' 2.6.0-test patches and I'd like to express, that beginning with 2.6.0-test8 the NFS client is absolutely unsuable for me. Trying to copy (with 'cp') a file from a local filesystem to the server over NFS.V2 or .V3 completely freezes _any_ NFS traffic with the same server.
Are there any significant changes to the NFS client I should back out for the purpose of finding the 'bug' ?

Martin.
--
Unix _IS_ user friendly - it's just selective about who its friends are !
--------------------------------------------------------------------------

-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it
help you create better code? SHARE THE LOVE, and help us help
YOU! Click Here: http://sourceforge.net/donate/
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs