This might be a way to decrease complexity of locking in vfs.
Basic idea: for local filesystems, i_sem gets taken on several objects
only to protect i_nlink.
These can be removed if i_nlink is atomic.
For network filesystems, no amount of locking would ensure atomic
operations anyway, so
probably no loss there.
inode operations would then lock:
link: both parents
rename: both parents
A new per-superblock semaphore vfs_link_sem would be created, to be taken
first on both link and rename, dropped as soon as all other locks are taken.
This prevents deadlocks in pathelogical cases. vfs_rename_sem is still needed
(taken after vfs_link_sem) to prevent cycles from being created.
Grabbing the target on link doesn't do much for most filesystems
because the operation
is (or should be) syncronized by the page lock for each directory
pages. If this change
is made, it is possible to remove this locking from all the filesystems.
Note that locking of source or target is no longer necessary. Except
for i_nlink, nothing
can change the source or target anyway.
Comment as you will. I am competent to write this if it is wanted. I
that I will be flamed to ashes for this suggestion.
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
[Index of Archives]
[Video 4 Linux]
[Linux for the blind]