2.4.21-rc1-ac2 NFS close-to-open question

2.4.21-rc1-ac2 NFS close-to-open question

[Shantanu Goel]

Hi,

In fs/nfs/inode.c:nfs_open() shouldn't the following

if (NFS_SERVER(inode)->flags & NFS_MOUNT_NOCTO) {
  err =
_nfs_revalidate_inode(NFS_SERVER(inode),inode);
  if (err)
    goto out;
}

be

if (!(NFS_SERVER(inode)->flags & NFS_MOUNT_NOCTO)) {
  err =
_nfs_revalidate_inode(NFS_SERVER(inode),inode);
  if (err)
    goto out;
}

If we desire close-to-open consistency, and assuming
my reading of the code is correct, is this a typo?

Thanks.
Shantanu



__________________________________
Do you Yahoo!?
The New Yahoo! Search - Faster. Easier. Bingo.
http://search.yahoo.com

Re: 2.4.21-rc1-ac2 NFS close-to-open question

[Trond Myklebust]

>>>>> " " == Shantanu Goel <sgoel01@yahoo.com> writes:

     > if (!(NFS_SERVER(inode)->flags & NFS_MOUNT_NOCTO)) {
     >   err =
     > _nfs_revalidate_inode(NFS_SERVER(inode),inode);
     >   if (err)
     >     goto out;
     > }

     > If we desire close-to-open consistency, and assuming my reading
     > of the code is correct, is this a typo?

Duh... Now *that* is downright embarassing...
Yup. Damn right...

Cheers,
  Trond

Re: 2.4.21-rc1-ac2 NFS close-to-open question

[Mike Fedyk]

On Sun, Apr 27, 2003 at 05:56:07PM +0200, Trond Myklebust wrote:
> >>>>> " " == Shantanu Goel <sgoel01@yahoo.com> writes:
> 
>      > if (!(NFS_SERVER(inode)->flags & NFS_MOUNT_NOCTO)) {
>      >   err =
>      > _nfs_revalidate_inode(NFS_SERVER(inode),inode);
>      >   if (err)
>      >     goto out;
>      > }
> 
>      > If we desire close-to-open consistency, and assuming my reading
>      > of the code is correct, is this a typo?
> 
> Duh... Now *that* is downright embarassing...
> Yup. Damn right...
> 
> Cheers,
>   Trond

Might that cause this too:

May  5 15:32:10 fileserver kernel: lockd: can't encode arguments: 5
May  5 15:33:08 fileserver last message repeated 18 times
May  5 15:34:07 fileserver last message repeated 27 times
May  5 15:35:07 fileserver last message repeated 23 times
May  5 15:36:08 fileserver last message repeated 10 times
May  5 15:37:09 fileserver last message repeated 22 times
May  5 15:37:34 fileserver last message repeated 9 times
May  5 17:24:47 fileserver kernel: lockd: can't encode arguments: 5
May  5 17:25:45 fileserver last message repeated 21 times

I have one system exporting a 112GB reiserfs partition and I've tested it
with two kernels:

2.4.20-rmap15e
2.4.21-rc1-rmap15g

2.4.21-rc1-rmap15g is giving the above error, and 2.4.20-rmap15e is not.
I'll leave it on 2.4.20-rmap15e for now and let you know if there are any
errors on that one too.

00:00.0 Host bridge: Intel Corp. 440FX - 82441FX PMC [Natoma] (rev 02)
00:07.0 ISA bridge: Intel Corp. 82371SB PIIX3 ISA [Natoma/Triton II] (rev 01)
00:07.1 IDE interface: Intel Corp. 82371SB PIIX3 IDE [Natoma/Triton II]
00:07.2 USB Controller: Intel Corp. 82371SB PIIX3 USB [Natoma/Triton II] (rev 01)
00:08.0 VGA compatible controller: S3 Inc. ViRGE/DX or /GX (rev 01)
00:0c.0 Ethernet controller: 3Com Corporation 3c905B 100BaseTX [Cyclone] (rev 24)

I can include .config files too if needed.

Thanks,

Mike

Re: 2.4.21-rc1-ac2 NFS close-to-open question

[Trond Myklebust]

>>>>> " " == Mike Fedyk <mfedyk@matchmail.com> writes:

     > Might that cause this too:

     > May 5 15:32:10 fileserver kernel: lockd: can't encode
     > arguments: 5

No.

     > 2.4.21-rc1-rmap15g is giving the above error, and
     > 2.4.20-rmap15e is not.  I'll leave it on 2.4.20-rmap15e for now
     > and let you know if there are any errors on that one too.

I'm confused. Are the rmap patches making changes to lockd?
I certainly don't see the above errors in standard 2.4.21-rc1.

Cheers,
  Trond

Re: 2.4.21-rc1-ac2 NFS close-to-open question

[Mike Fedyk]

On Tue, May 06, 2003 at 02:50:53PM +0200, Trond Myklebust wrote:
> I'm confused. Are the rmap patches making changes to lockd?
> I certainly don't see the above errors in standard 2.4.21-rc1.

Ok, I'll test with standard rc1...

Oh and one more thing, both have the debian freeswan kernel patch applied.

I'll try again without any patches.

Re: 2.4.21-rc1-ac2 NFS close-to-open question

[Mike Fedyk]

On Tue, May 06, 2003 at 02:50:53PM +0200, Trond Myklebust wrote:
> >>>>> " " == Mike Fedyk <mfedyk@matchmail.com> writes:
> 
>      > Might that cause this too:
> 
>      > May 5 15:32:10 fileserver kernel: lockd: can't encode
>      > arguments: 5
> 
> No.
> 
>      > 2.4.21-rc1-rmap15g is giving the above error, and
>      > 2.4.20-rmap15e is not.  I'll leave it on 2.4.20-rmap15e for now
>      > and let you know if there are any errors on that one too.
> 
> I'm confused. Are the rmap patches making changes to lockd?
> I certainly don't see the above errors in standard 2.4.21-rc1.

Oh, one more thing.

This is output from a server running 2.4.19-freeswan mounting an nfs export
from a reiserfs based 112GB export using 2.4.21-rc1-freeswan-rmap.  When the
112GBs are exported from 2.4.20-freeswan-rmap, there is no error.

I will retest, but I hope this isn't a bug making it through the pre & rc
process.