Skip to contentSkip to conversation

Different rsize/wsize for different NFS server software

DISCUSSION Shehjar Tikoo 2 participants 4 messages May 12, 2009 3 messages not in the archive

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

Hi All

I've been running some tests using the linux NFS3 client using the in-kernel NFS server and the unfs3 project.

I see that the client mounts the in-kernel server with default [rw]size values of 32k and the unfs3 server using [rw]size values of 512k. Where is this difference coming from, because I do not specify these different [rw]size value on the command line for mount.

The client kernel is Debian's 2.6.18-5-amd64 and the server kernel is Fedora's 2.6.18-92.1.6.el5.

Thanks
Shehjar

A message that is not in the archive (sent off-list or not yet received): 4A0959F9.1010107-iP+wyIN8Xge1Z/[email protected]

#2 lore raw
Quoted 10 lines · Shehjar Tikoo

Hi All

I've been running some tests using the linux NFS3 client using the in-kernel NFS server and the unfs3 project.

I see that the client mounts the in-kernel server with default [rw]size values of 32k and the unfs3 server using [rw]size values of 512k. Where is this difference coming from, because I do not specify these different [rw]size value on the command line for mount.

Through wireshark, I see that an FSINFO request is sent
by the client in order to get the wtpref and rtpref values and that is what determines the wsize and rsize values at the client.

The next strange behaviour is that, with unfs3 as the server, I try to mount using wsize set to 65536 with the command:

$ mount  192.168.1.201:/server/ -o wsize=65536 /test/mount

but the /proc/mounts entry still reports a 512k wsize value.

$ cat /proc/mounts
192.168.1.201:/server/ /test/mount nfs rw,vers=3,rsize=524288,wsize=524288,hard,intr,proto=tcp,timeo=600,retrans=2,sec=sys,addr=192.168.1.201 0 0

and wireshark also shows client sending 512k requests. mount.fs version is: mount.nfs (linux nfs-utils 1.1.4)

Is there any way to force the client to use a 64k wsize value?

Thanks
Shehjar

Quoted 10 lines

The client kernel is Debian's 2.6.18-5-amd64 and the server kernel is Fedora's 2.6.18-92.1.6.el5.

Thanks
Shehjar

Signature · 3 lines
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

A message that is not in the archive (sent off-list or not yet received): 4A097DE9.6030709-iP+wyIN8Xge1Z/[email protected]

#4 lore raw
Quoted 28 lines · Shehjar Tikoo
Quoted 10 lines · Shehjar Tikoo

Hi All

I've been running some tests using the linux NFS3 client using the in-kernel NFS server and the unfs3 project.

I see that the client mounts the in-kernel server with default [rw]size values of 32k and the unfs3 server using [rw]size values of 512k. Where is this difference coming from, because I do not specify these different [rw]size value on the command line for mount.

Through wireshark, I see that an FSINFO request is sent
by the client in order to get the wtpref and rtpref values and that is what determines the wsize and rsize values at the client.

The next strange behaviour is that, with unfs3 as the server, I try to mount using wsize set to 65536 with the command:

$ mount  192.168.1.201:/server/ -o wsize=65536 /test/mount

but the /proc/mounts entry still reports a 512k wsize value.

$ cat /proc/mounts
192.168.1.201:/server/ /test/mount nfs 
rw,vers=3,rsize=524288,wsize=524288,hard,intr,proto=tcp,timeo=600,retrans=2,sec=sys,addr=192.168.1.201 
0 0
Quoted 4 lines

and wireshark also shows client sending 512k requests. mount.fs version is: mount.nfs (linux nfs-utils 1.1.4)

Is there any way to force the client to use a 64k wsize value?

You probably have the exact same filesystem still mounted somewhere else on that client. A 2.6.18 kernel will, in those circumstances, share the same super block with both mount points, and will therefore not permit you to use different mount options.

This is something which has been fixed in more recent kernels.

Trond

A message that is not in the archive (sent off-list or not yet received): 1242137345.21939.6.camel-rJ7iovZKK19ZJLDQqaL3InhyD016LWXt@public.gmane.org

#6 lore raw
Quoted 2 lines · Trond Myklebust
Quoted 1 line · Shehjar Tikoo

Shehjar Tikoo wrote:

...
...
...

Quoted 6 lines
Quoted 1 line

Is there any way to force the client to use a 64k wsize value?

You probably have the exact same filesystem still mounted somewhere else on that client. A 2.6.18 kernel will, in those circumstances, share the same super block with both mount points, and will therefore not permit you to use different mount options.

That was, in fact, the case.

Thanks
Shehjar

Quoted 11 lines

This is something which has been fixed in more recent kernels.

Trond

--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html