On Mon, 31 Oct 2011 14:04:00 +0300
Pavel Shilovsky <piastry@xxxxxxxxxxx> wrote:
> 2011/10/31 Pavel Shilovsky <piastry@xxxxxxxxxxx>:
> > 2011/10/31 Jeff Layton <jlayton@xxxxxxxxxxxxxxx>:
> >> On Sat, 29 Oct 2011 17:17:58 +0400
> >> Pavel Shilovsky <piastry@xxxxxxxxxxx> wrote:
> >>
> >>> Now we allocate a lock structure at first, then we request to the server
> >>> and save the lock if server returned OK though void function - it prevents
> >>> the situation when we locked a file on the server and then return -ENOMEM
> >>> from setlk.
> >>>
> >>> Signed-off-by: Pavel Shilovsky <piastry@xxxxxxxxxxx>
> >>> ---
> >>> fs/cifs/file.c | 64 ++++++++++++++++++++++++++++----------------------------
> >>> 1 files changed, 32 insertions(+), 32 deletions(-)
> >>>
> >>> diff --git a/fs/cifs/file.c b/fs/cifs/file.c
> >>> index c1f063c..d9cc07f 100644
> >>> --- a/fs/cifs/file.c
> >>> +++ b/fs/cifs/file.c
> >>> @@ -672,7 +672,7 @@ cifs_del_lock_waiters(struct cifsLockInfo *lock)
> >>> }
> >>>
> >>> static bool
> >>> -cifs_find_lock_conflict(struct cifsInodeInfo *cinode, __u64 offset,
> >>> +__cifs_find_lock_conflict(struct cifsInodeInfo *cinode, __u64 offset,
> >>> __u64 length, __u8 type, __u16 netfid,
> >>> struct cifsLockInfo **conf_lock)
> >>> {
> >>> @@ -694,6 +694,14 @@ cifs_find_lock_conflict(struct cifsInodeInfo *cinode, __u64 offset,
> >>> return false;
> >>> }
> >>>
> >>> +static bool
> >>> +cifs_find_lock_conflict(struct cifsInodeInfo *cinode, struct cifsLockInfo *lock,
> >>> + struct cifsLockInfo **conf_lock)
> >>> +{
> >>> + return __cifs_find_lock_conflict(cinode, lock->offset, lock->length,
> >>> + lock->type, lock->netfid, conf_lock);
> >>> +}
> >>> +
> >>> static int
> >>> cifs_lock_test(struct cifsInodeInfo *cinode, __u64 offset, __u64 length,
> >>> __u8 type, __u16 netfid, struct file_lock *flock)
> >>> @@ -704,8 +712,8 @@ cifs_lock_test(struct cifsInodeInfo *cinode, __u64 offset, __u64 length,
> >>>
> >>> mutex_lock(&cinode->lock_mutex);
> >>>
> >>> - exist = cifs_find_lock_conflict(cinode, offset, length, type, netfid,
> >>> - &conf_lock);
> >>> + exist = __cifs_find_lock_conflict(cinode, offset, length, type, netfid,
> >>> + &conf_lock);
> >>> if (exist) {
> >>> flock->fl_start = conf_lock->offset;
> >>> flock->fl_end = conf_lock->offset + conf_lock->length - 1;
> >>> @@ -723,40 +731,27 @@ cifs_lock_test(struct cifsInodeInfo *cinode, __u64 offset, __u64 length,
> >>> return rc;
> >>> }
> >>>
> >>> -static int
> >>> -cifs_lock_add(struct cifsInodeInfo *cinode, __u64 len, __u64 offset,
> >>> - __u8 type, __u16 netfid)
> >>> +static void
> >>> +cifs_lock_add(struct cifsInodeInfo *cinode, struct cifsLockInfo *lock)
> >>> {
> >>> - struct cifsLockInfo *li;
> >>> -
> >>> - li = cifs_lock_init(len, offset, type, netfid);
> >>> - if (!li)
> >>> - return -ENOMEM;
> >>> -
> >>> mutex_lock(&cinode->lock_mutex);
> >>> - list_add_tail(&li->llist, &cinode->llist);
> >>> + list_add_tail(&lock->llist, &cinode->llist);
> >>> mutex_unlock(&cinode->lock_mutex);
> >>> - return 0;
> >>> }
> >>>
> >>> static int
> >>> -cifs_lock_add_if(struct cifsInodeInfo *cinode, __u64 offset, __u64 length,
> >>> - __u8 type, __u16 netfid, bool wait)
> >>> +cifs_lock_add_if(struct cifsInodeInfo *cinode, struct cifsLockInfo *lock,
> >>> + bool wait)
> >>> {
> >>> - struct cifsLockInfo *lock, *conf_lock;
> >>> + struct cifsLockInfo *conf_lock;
> >>> bool exist;
> >>> int rc = 0;
> >>>
> >>> - lock = cifs_lock_init(length, offset, type, netfid);
> >>> - if (!lock)
> >>> - return -ENOMEM;
> >>> -
> >>> try_again:
> >>> exist = false;
> >>> mutex_lock(&cinode->lock_mutex);
> >>>
> >>> - exist = cifs_find_lock_conflict(cinode, offset, length, type, netfid,
> >>> - &conf_lock);
> >>> + exist = cifs_find_lock_conflict(cinode, lock, &conf_lock);
> >>> if (!exist && cinode->can_cache_brlcks) {
> >>> list_add_tail(&lock->llist, &cinode->llist);
> >>> mutex_unlock(&cinode->lock_mutex);
> >>> @@ -781,7 +776,6 @@ try_again:
> >>> }
> >>> }
> >>>
> >>> - kfree(lock);
> >>> mutex_unlock(&cinode->lock_mutex);
> >>> return rc;
> >>> }
> >>> @@ -1254,20 +1248,26 @@ cifs_setlk(struct file *file, struct file_lock *flock, __u8 type,
> >>> }
> >>>
> >>> if (lock) {
> >>> - rc = cifs_lock_add_if(cinode, flock->fl_start, length,
> >>> - type, netfid, wait_flag);
> >>> + struct cifsLockInfo *lock;
> >>> +
> >>> + lock = cifs_lock_init(length, flock->fl_start, type, netfid);
> >>> + if (!lock)
> >>> + return -ENOMEM;
> >>> +
> >>> + rc = cifs_lock_add_if(cinode, lock, wait_flag);
> >>
> >> Here, you're adding "lock" to the list...
> >
> > If we added the lock to the list cifs_lock_add_if returns 0 and we
> > will jump to out label.
> >
> >>
> >>> if (rc < 0)
> >>> - return rc;
> >>> - else if (!rc)
> >>> + kfree(lock);
> >>> + if (rc <= 0)
> >>> goto out;
> >>>
> >>> rc = CIFSSMBLock(xid, tcon, netfid, current->tgid, length,
> >>> flock->fl_start, 0, 1, type, wait_flag, 0);
> >>> - if (rc == 0) {
> >>> - /* For Windows locks we must store them. */
> >>> - rc = cifs_lock_add(cinode, length, flock->fl_start,
> >>> - type, netfid);
> >>> + if (rc) {
> >>> + kfree(lock);
> >>
> >> ...and here you're freeing "lock" without removing it from the list.
> >> Isn't that like to cause a problem?
> >
> > So, if CIFSSMBLock returns a error we free the lock that hasn't been
> > added to the list. If CIFSSMBLock returns ok, we will add it to the
> > list with cifs_lock_add void function.
> >
> > Seems no problem with it.
> >
> > --
> > Best regards,
> > Pavel Shilovsky.
> >
>
> So, to be clear - brlock cache function cifs_lock_add_if returns:
>
> "0" if we store the lock in the cache and have an oplock for batching
> brlocks - no need to request to the server.
>
> "1" if there is no locks preventing us to set this lock but we haven't
> an oplock for batching brlocks - need to request to the server and
> then add the lock to the list if server accepts.
>
> "-EACCESS" if we have a lock on the client that prevent us and this
> call is non-blocked.
>
Ok, it looks like this patch is probably ok then, but it would be nice
to have the above explanation enshrined in a comment over
cifs_lock_add_if to ensure that other people working in this code
understand it.
Can you spin up a patch to add that when you get time? For this patch
though...
Acked-by: Jeff Layton <jlayton@xxxxxxxxx>
--
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
