Re: [PATCH 05/16] nfs: don't open in ->d_revalidate

[Miklos Szeredi]

On Wed, Mar 28, 2012 at 11:45 PM, Myklebust, Trond
<Trond.Myklebust@xxxxxxxxxx> wrote:
> On Wed, 2012-03-28 at 22:24 +0200, Miklos Szeredi wrote:
>> From: Miklos Szeredi <mszeredi@xxxxxxx>
>>
>> NFSv4 can't do reliable opens in d_revalidate, since it cannot know whether a
>> mount needs to be followed or not. ÂIt does check d_mountpoint() on the dentry,
>> which can result in a weird error if the VFS found that the mount does not in
>> fact need to be followed, e.g.:
>>
>> Â # mount --bind /mnt/nfs /mnt/nfs-clone
>> Â # echo something > /mnt/nfs/tmp/bar
>> Â # echo x > /tmp/file
>> Â # mount --bind /tmp/file /mnt/nfs-clone/tmp/bar
>> Â # cat Â/mnt/nfs/tmp/bar
>> Â cat: /mnt/nfs/tmp/bar: Not a directory
>>
>> Which should, by any sane filesystem, result in "something" being printed.
>>
>> So instead do the open in f_op->open() and in the unlikely case that the cached
>> dentry turned out to be invalid, drop the dentry and return ESTALE to let the
>> VFS retry.
>
>
> Just one comment. Would it now make sense for NFSv4 to just skip
> ->d_revalidate() if LOOKUP_OPEN is set, and LOOKUP_EXCL is not set? We
> will in any case be doing a revalidation in nfs4_file_open.

And dentry is positive and regular.  Which is basically what
nfs4_lookup_revalidate() does check at the moment.

One question is whether this can be done without dropping out of RCU
mode, which might be a real performance win. I'm not sure about
dereferencing inode->i_mode.  AFAICS it should be fine, considering
that destruction of the inode will leave the mode bits untouched,
but...

Thanks,
Miklos
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/