On Tue, Oct 23, 2012 at 9:41 AM, Jeff Layton <jlayton@xxxxxxxxxx> wrote:
> On Tue, 23 Oct 2012 08:24:51 -0500
> shirishpargaonkar@xxxxxxxxx wrote:
>
>> From: Shirish Pargaonkar <shirishpargaonkar@xxxxxxxxx>
>>
>>
>> Add support of Alternate Data Streams (ads).
>>
>> The generic access flags that cifs client currently employs are sufficient
>> for alternate data streams as well (MS-CIFS 2.2.4.64.1).
>>
>> The stream file and stream type are specified using : after the file name,
>> so that is used to differentiate between a regular file and its
>> alternate data streams and stream types.
>> Since they all have the same file id, each path name,
>> file name:stream name:stream type, has a separate inode with that same
>> file id but a distinct private data (path name) in that inode to
>> distinguish them.
>>
>> This scheme applies only to non-posix compliant servers such as Windows.
>>
>
> Question: are alternate data streams really part of the protocol
> per-se? Or are they an implementation detail of the exported
> filesystem? Is it possible that a (say non-windows) server could export
> a filesystem that did allow ':' in a filename without posix extensions?
>
> If so, are we at risk of treating such a file as an alternate data
> stream instead of a separate file altogether? Is there a way to
> distinguish such a situation?
>
Jeff, they are not part of the protocol, part of the implementation
details of the exported file system.
It is possible that a cifs/smb server such as a non-windows
server which does not support posix extensions could export a
file system that does allow ':' in a filename but it would not be a
issue because file ids / unique ids at the server would be different.
>> One operation that does not work is Rename (0x7).
>>
>
> I'm confused. What does this have to do with rename?
>
I was stating the observation that, for a file stream, against
a Windows server, cifs/smb operation of renaming that file stream
either to a regular file or to a another file stream fails.
(whereas create / delete / lookup etc. operations succeed).
>>
>> Signed-off-by: Shirish Pargaonkar <shirishpargaonkar@xxxxxxxxx>
>> ---
>> fs/cifs/cifsfs.c | 1 +
>> fs/cifs/cifsglob.h | 2 ++
>> fs/cifs/inode.c | 31 ++++++++++++++++++++++++++++++-
>> 3 files changed, 33 insertions(+), 1 deletions(-)
>>
>> diff --git a/fs/cifs/cifsfs.c b/fs/cifs/cifsfs.c
>> index a41044a..1a02c81 100644
>> --- a/fs/cifs/cifsfs.c
>> +++ b/fs/cifs/cifsfs.c
>> @@ -264,6 +264,7 @@ cifs_evict_inode(struct inode *inode)
>> {
>> truncate_inode_pages(&inode->i_data, 0);
>> clear_inode(inode);
>> + kfree(inode->i_private);
>> cifs_fscache_release_inode_cookie(inode);
>> }
>>
>> diff --git a/fs/cifs/cifsglob.h b/fs/cifs/cifsglob.h
>> index f5af252..26d65c7 100644
>> --- a/fs/cifs/cifsglob.h
>> +++ b/fs/cifs/cifsglob.h
>> @@ -1251,6 +1251,7 @@ struct dfs_info3_param {
>> #define CIFS_FATTR_DELETE_PENDING 0x2
>> #define CIFS_FATTR_NEED_REVAL 0x4
>> #define CIFS_FATTR_INO_COLLISION 0x8
>> +#define CIFS_FATTR_ALTDATASTR 0x10
>>
>> struct cifs_fattr {
>> u32 cf_flags;
>> @@ -1268,6 +1269,7 @@ struct cifs_fattr {
>> struct timespec cf_atime;
>> struct timespec cf_mtime;
>> struct timespec cf_ctime;
>> + char *cf_private;
>> };
>>
>> static inline void free_dfs_info_param(struct dfs_info3_param *param)
>> diff --git a/fs/cifs/inode.c b/fs/cifs/inode.c
>> index afdff79..b7c6dcf 100644
>> --- a/fs/cifs/inode.c
>> +++ b/fs/cifs/inode.c
>> @@ -615,10 +615,12 @@ cifs_get_inode_info(struct inode **inode, const char *full_path,
>> struct tcon_link *tlink;
>> struct cifs_sb_info *cifs_sb = CIFS_SB(sb);
>> char *buf = NULL;
>> + char *altstr = NULL;
>> bool adjust_tz = false;
>> struct cifs_fattr fattr;
>> struct cifs_search_info *srchinf = NULL;
>>
>> + fattr.cf_private = NULL;
>> tlink = cifs_sb_tlink(cifs_sb);
>> if (IS_ERR(tlink))
>> return PTR_ERR(tlink);
>> @@ -746,9 +748,26 @@ cifs_get_inode_info(struct inode **inode, const char *full_path,
>> }
>>
>> if (!*inode) {
>> + altstr = strchr(full_path, ':');
>> + if (altstr) {
>> + fattr.cf_private = kstrdup(altstr, GFP_KERNEL);
>> + if (!fattr.cf_private) {
>> + rc = -ENOMEM;
>> + goto cgii_exit;
>> + }
>> + fattr.cf_flags |= CIFS_FATTR_ALTDATASTR;
>> + }
>> +
>> *inode = cifs_iget(sb, &fattr);
>> - if (!*inode)
>> + if (*inode) {
>> + if (altstr && !((*inode)->i_private))
>> + (*inode)->i_private = fattr.cf_private;
>> + else
>> + kfree(fattr.cf_private);
>> + } else {
>> rc = -ENOMEM;
>> + kfree(fattr.cf_private);
>> + }
>> } else {
>> cifs_fattr_to_inode(*inode, &fattr);
>> }
>> @@ -784,6 +803,16 @@ cifs_find_inode(struct inode *inode, void *opaque)
>> if (S_ISDIR(inode->i_mode) && !hlist_empty(&inode->i_dentry))
>> fattr->cf_flags |= CIFS_FATTR_INO_COLLISION;
>>
>> + /* looking for an inode of a alternate data stream (full pathname) */
>> + if (fattr->cf_flags & CIFS_FATTR_ALTDATASTR) {
>> + if (!(inode->i_private)) {
>> + return 0;
>> + } else {
>> + if (strcmp(inode->i_private, fattr->cf_private))
>> + return 0;
>> + }
>> + }
>> +
>> return 1;
>> }
>>
>
>
> --
> Jeff Layton <jlayton@xxxxxxxxxx>
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
