Random ESTALE due to lookup "/" in mount point?

Random ESTALE due to lookup "/" in mount point?

[Frank van Maarseveen]

NFSv3, udp, client=2.6.17.8, server=tru64

On the client I started to get some random ESTALE errors which disappeared
quickly afterwards without a remount. The errors only occurred when typing

	stat ..

in a subdirectory of a mountpoint. Doing ls -al in that subdirectory
gave an ESTALE only on "..". I could even change directory to / and back
without an ESTALE but the "stat .." command still gave an ESTALE.

A "tcpdump -i eth0 udp port 2049 -n -s 1500" showed something which is
unusual to my knowledge. The kernel tried to lookup "/" in the mount
point as it seems:

   16:42:03.644621 IP 172.17.1.64.3759242568 > 172.17.1.1.2049: 120 lookup fh 2956,259248/28524 "stat"
   16:42:03.644967 IP 172.17.1.1.2049 > 172.17.1.64.3759242568: reply ok 116 lookup ERROR: No such file or directory
   16:42:03.647345 IP 172.17.1.64.3776019784 > 172.17.1.1.2049: 112 getattr fh 2956,259248/2
   16:42:03.647496 IP 172.17.1.1.2049 > 172.17.1.64.3776019784: reply ok 112 getattr DIR 1777 ids 0/0 sz 0x2000
=> 16:42:03.648002 IP 172.17.1.64.3792797000 > 172.17.1.1.2049: 120 lookup fh 2956,259248/2 "/"
   16:42:03.648244 IP 172.17.1.1.2049 > 172.17.1.64.3792797000: reply ok 116 lookup ERROR: No such file or directory
   16:42:03.648440 IP 172.17.1.64.3809574216 > 172.17.1.1.2049: 112 getattr fh 2956,259248/2
   16:42:03.648680 IP 172.17.1.1.2049 > 172.17.1.64.3809574216: reply ok 112 getattr DIR 1777 ids 0/0 sz 0x2000
=> 16:42:03.648877 IP 172.17.1.64.3826351432 > 172.17.1.1.2049: 120 lookup fh 2956,259248/2 "/"
   16:42:03.649110 IP 172.17.1.1.2049 > 172.17.1.64.3826351432: reply ok 116 lookup ERROR: No such file or directory
   16:42:09.605452 IP 172.17.1.64.3879531864 > 172.17.1.1.2049: 116 access fh 3806,805074/16337 001f
   16:42:09.605758 IP 172.17.1.1.2049 > 172.17.1.64.3879531864: reply ok 120 access c 001f

-- 
Frank

-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: Random ESTALE due to lookup "/" in mount point?

[Trond Myklebust]

On Wed, 2006-08-23 at 17:09 +0200, Frank van Maarseveen wrote:
> NFSv3, udp, client=2.6.17.8, server=tru64
> 
> On the client I started to get some random ESTALE errors which disappeared
> quickly afterwards without a remount. The errors only occurred when typing
> 
> 	stat ..
> 
> in a subdirectory of a mountpoint. Doing ls -al in that subdirectory
> gave an ESTALE only on "..". I could even change directory to / and back
> without an ESTALE but the "stat .." command still gave an ESTALE.
> 
> A "tcpdump -i eth0 udp port 2049 -n -s 1500" showed something which is
> unusual to my knowledge. The kernel tried to lookup "/" in the mount
> point as it seems:

I've never seen that, and my own kernel certainly appears to do no such
thing when I stat '..'.

Any chance of a 'strace' with that?

Cheers,
  Trond


-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs

Re: Random ESTALE due to lookup "/" in mount point?

[Frank van Maarseveen]

On Wed, Aug 23, 2006 at 12:20:25PM -0400, Trond Myklebust wrote:
> On Wed, 2006-08-23 at 17:09 +0200, Frank van Maarseveen wrote:
> > NFSv3, udp, client=2.6.17.8, server=tru64
> > 
> > On the client I started to get some random ESTALE errors which disappeared
> > quickly afterwards without a remount. The errors only occurred when typing
> > 
> > 	stat ..
> > 
> > in a subdirectory of a mountpoint. Doing ls -al in that subdirectory
> > gave an ESTALE only on "..". I could even change directory to / and back
> > without an ESTALE but the "stat .." command still gave an ESTALE.
> > 
> > A "tcpdump -i eth0 udp port 2049 -n -s 1500" showed something which is
> > unusual to my knowledge. The kernel tried to lookup "/" in the mount
> > point as it seems:
> 
> I've never seen that, and my own kernel certainly appears to do no such
> thing when I stat '..'.
> 
> Any chance of a 'strace' with that?

It's reproduceable now but it appears that there's a nonstandard kernel
patch from me involved to do a selective "noac" (www.frankvm.com/nfs-noac,
it's a bit hackish...).  I think this results in an incorrect revalidation
of the mount point and I guess "/" is coming from the dcache: the fake
name of any root inode IIRC.

In short, I have to sort it out but it's probably only my patch at
fault. Sorry for the disturbance.

-- 
Frank

-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
NFS maillist  -  NFS@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs