[linux-cifs-client] Re: cthon test progress

[Steve French]

The bad news is that it looks like Windows 2008 returns ok on the
original rename (instead of EEXIST) so we have to move this UniqueId
detection up ... (which will slow things down in the main path)

On Wed, Sep 24, 2008 at 7:12 PM, Jeff Layton <jlayton at redhat.com> wrote:
> On Wed, 24 Sep 2008 18:44:41 -0500
> "Steve French" <smfrench at gmail.com> wrote:
>
>> With LinuxExtensionsDisabled and this patch applied, now make it farther.
>>
>> diff --git a/fs/cifs/inode.c b/fs/cifs/inode.c
>> index db091c5..ba0d9fa 100644
>> --- a/fs/cifs/inode.c
>> +++ b/fs/cifs/inode.c
>> @@ -19,7 +19,6 @@
>>   *   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
>>   */
>>  #include <linux/fs.h>
>> -#include <linux/stat.h>
>>  #include <linux/pagemap.h>
>>  #include <asm/div64.h>
>>  #include "cifsfs.h"
>> @@ -1309,8 +1308,33 @@ int cifs_rename(struct inode *source_inode,
>> struct dentry *source_direntry,
>>                                       info_buf_target->UniqueId))
>>                               /* same file, POSIX says that this is a noop */
>>                               goto cifs_rename_exit;
>> -             } /* else ... BB we could add the same check for Windows by
>> -                  checking the UniqueId via FILE_INTERNAL_INFO */
>> +             } else {
>> +                     __u64 src_inode_num;
>> +                     __u64 target_inode_num;
>> +                     cFYI(1, ("check hardlinks windows style on %s",
>> +                              fromName));
>> +                     rc = CIFSGetSrvInodeNumber(xid, pTcon,
>> +                                     fromName, &src_inode_num,
>> +                                     cifs_sb_source->local_nls,
>> +                                     cifs_sb_source->mnt_cifs_flags &
>> +                                                CIFS_MOUNT_MAP_SPECIAL_CHR);
>> +                     if (rc)
>> +                             goto unlink_target;
>> +
>> +                     rc = CIFSGetSrvInodeNumber(xid, pTcon,
>> +                                     toName, &target_inode_num,
>> +                                     cifs_sb_source->local_nls,
>> +                                     cifs_sb_source->mnt_cifs_flags &
>> +                                                CIFS_MOUNT_MAP_SPECIAL_CHR);
>> +                     if (rc)
>> +                             goto unlink_target;
>> +
>> +                     if ((src_inode_num != 0) &&
>> +                                     (src_inode_num == target_inode_num)) {
>> +                             cFYI(1, ("inode numbers match - rename noop"));
>> +                             goto cifs_rename_exit;
>> +                     }
>> +             }
>>  unlink_target:
>>               /*
>>                * we either can not tell the files are hardlinked (as with
>>
>> The "special" connectathon tests now fails at:
>>
>> second check for lost reply on non-idempotent requests
>> testing 50 idempotencies in directory "testdir"
>> stat 1: bad file type/size 0102767/30
>> special tests failed
>>
>>
>
> Nice work.
>
> It would also be nice to see the hardlink detection code code broken
> out into a separate function. cifs_rename is getting pretty large and
> unweildy and that seems like a good clean spot for an interface.
> --
> Jeff Layton <jlayton at redhat.com>
>



-- 
Thanks,

Steve