Skip to contentSkip to conversation

32 bit stat() on a 64 bit kernel failing on a large NFS filesystem?

DISCUSSION James Pearson 2 participants 3 messages Jun 20, 2006 – Jun 22, 2006

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

I recently found a problem running a 32bit (i686) binary on a 64bit (x86_64) kernel that fails when stat'ing any file on a multi-TB NFS clustered file system (a non-Linux file system).

stat() returns EOVERFLOW (Value too large for defined data type)

The same 32bit binary running on a 32bit kernel stat'ing the same files works OK.

In both cases, I'm using RHEL4 based distros on the clients (I also tried a generic 2.6.16.x kernel on the 64 bit client, but that made no difference).

We can work round the problem by either making sure the binary is compiled with _FILE_OFFSET_BITS set to 64, or configuring the NFS file server to use 32 bit 'file IDs'.

My question are:

Is this a 'bug' or 'feature' with using a 32 bit app on a 64 bit kernel?

If it's a 'feature', can anyone explain why it works in this way?

Thanks

James Pearson

_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs

#1 lore raw
Quoted 17 lines · James Pearson

I recently found a problem running a 32bit (i686) binary on a 64bit (x86_64) kernel that fails when stat'ing any file on a multi-TB NFS clustered file system (a non-Linux file system).

stat() returns EOVERFLOW (Value too large for defined data type)

The same 32bit binary running on a 32bit kernel stat'ing the same files works OK.

In both cases, I'm using RHEL4 based distros on the clients (I also tried a generic 2.6.16.x kernel on the 64 bit client, but that made no difference).

We can work round the problem by either making sure the binary is compiled with _FILE_OFFSET_BITS set to 64, or configuring the NFS file server to use 32 bit 'file IDs'.

Quoted 5 lines

My question are:

Is this a 'bug' or 'feature' with using a 32 bit app on a 64 bit kernel?

If it's a 'feature', can anyone explain why it works in this way?

I already had the same problem, just see this thread:

http://lkml.org/lkml/2005/2/28/172

-- =

Bernd Schubert
PCI / Theoretische Chemie
Universit=E4t Heidelberg
INF 229
69120 Heidelberg

_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs

#2 lore raw
Quoted 21 lines · Bernd Schubert
Quoted 19 lines · James Pearson

I recently found a problem running a 32bit (i686) binary on a 64bit (x86_64) kernel that fails when stat'ing any file on a multi-TB NFS clustered file system (a non-Linux file system).

stat() returns EOVERFLOW (Value too large for defined data type)

The same 32bit binary running on a 32bit kernel stat'ing the same files works OK.

In both cases, I'm using RHEL4 based distros on the clients (I also tried a generic 2.6.16.x kernel on the 64 bit client, but that made no difference).

We can work round the problem by either making sure the binary is compiled with _FILE_OFFSET_BITS set to 64, or configuring the NFS file server to use 32 bit 'file IDs'.

My question are:

Quoted 8 lines
Quoted 3 lines

Is this a 'bug' or 'feature' with using a 32 bit app on a 64 bit kernel?

If it's a 'feature', can anyone explain why it works in this way?

I already had the same problem, just see this thread:

http://lkml.org/lkml/2005/2/28/172

Thanks - that thread sort of explains what is going on ...

James Pearson

All the advantages of Linux Managed Hosting--Without the Cost and Risk! Fully trained technicians. The highest number of Red Hat certifications in the hosting industry. Fanatical Support. Click to learn more
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=107521&bid=248729&dat=121642
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs