2012/2/10 Shirish Pargaonkar <shirishpargaonkar@xxxxxxxxx>:
> On Thu, Feb 9, 2012 at 12:08 PM, Pavel Shilovsky <piastry@xxxxxxxxxxx> wrote:
>> Currently we do inc_nlink/drop_nlink for a parent directory for every
>> mkdir and rmdir calls. That's wrong when POSIX extensions are disabled
>> because in this case a server doesn't do the same things and returns
>> the old value on the next QueryInfo request. As the result, we update
>> our value with the server one and then decrement it on every rmdir
>> call - go to negative nlink values.
>>
>> Fix this by doing inc_nlink/drop_nlink for parent directory in mkdir
>> and rmdir in POSIX case only. Also add cERROR when nlink value <= 2
>> and we still try to decrement it (possible broken servers).
>>
>> Signed-off-by: Pavel Shilovsky <piastry@xxxxxxxxxxx>
>> ---
>> fs/cifs/inode.c | 14 ++++++++++----
>> 1 files changed, 10 insertions(+), 4 deletions(-)
>>
>> diff --git a/fs/cifs/inode.c b/fs/cifs/inode.c
>> index a5f54b7..9af6ec7 100644
>> --- a/fs/cifs/inode.c
>> +++ b/fs/cifs/inode.c
>> @@ -1355,11 +1355,11 @@ mkdir_retry_old:
>> d_drop(direntry);
>> } else {
>> mkdir_get_info:
>> - inc_nlink(inode);
>> - if (pTcon->unix_ext)
>> + if (pTcon->unix_ext) {
>> + inc_nlink(inode);
>> rc = cifs_get_inode_info_unix(&newinode, full_path,
>> inode->i_sb, xid);
>> - else
>> + } else
>> rc = cifs_get_inode_info(&newinode, full_path, NULL,
>> inode->i_sb, xid, NULL);
>>
>> @@ -1475,7 +1475,13 @@ int cifs_rmdir(struct inode *inode, struct dentry *direntry)
>> cifs_put_tlink(tlink);
>>
>> if (!rc) {
>> - drop_nlink(inode);
>> + if (pTcon->unix_ext) {
>> + if (inode->i_nlink > 2)
>> + drop_nlink(inode);
>> + else
>> + cERROR(1, "%s: unexpected nlink number(%u)",
>> + __func__, inode->i_nlink);
>> + }
>> spin_lock(&direntry->d_inode->i_lock);
>> i_size_write(direntry->d_inode, 0);
>> clear_nlink(direntry->d_inode);
>> --
>> 1.7.1
>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-cifs" in
>> the body of a message to majordomo@xxxxxxxxxxxxxxx
>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
> Then we need to apply the same logic to set_nlink() calls and
> wherever we access i_nlink field also?
No, only mkdir/rmdir need this. Other places in cifs with i_nlink
access is either read only or inc_nlink/drop_nlink for
non-directories.
--
Best regards,
Pavel Shilovsky.
--
To unsubscribe from this list: send the line "unsubscribe linux-cifs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
