On Wed, May 06, 2009 at 11:19:12PM +0100, Al Viro wrote: > On Wed, May 06, 2009 at 10:16:05PM +0200, Christoph Hellwig wrote: > > Add a central MS_RDONLY to sync_super instead of having it in the ->write_super > > methods (and even that not consistently). The other two callers are already > > protected: sync_filesystems has the same s_umount protected MS_RDONLY check > > and file_fsync can only be called on a writeable file descriptor. > > > > Also make sure to clear s_dirt if it was set on a read-only superblock to > > avoid calling into these filesystems again and again. > > > Index: vfs-2.6/fs/affs/super.c > > Index: vfs-2.6/fs/bfs/inode.c > > Index: vfs-2.6/fs/ext2/super.c > > Index: vfs-2.6/fs/fat/inode.c > > Index: vfs-2.6/fs/hfs/super.c > > Index: vfs-2.6/fs/hfsplus/super.c > > Index: vfs-2.6/fs/jffs2/fs.c > > Index: vfs-2.6/fs/jffs2/super.c > > Index: vfs-2.6/fs/nilfs2/super.c > > Index: vfs-2.6/fs/super.c > > Index: vfs-2.6/fs/sysv/inode.c > > Index: vfs-2.6/fs/ufs/super.c > > You've missed xfs. Added and applied (reiserfs, ext4 and exofs are also > not touched, but they don't need a change here). Eh... Actually, this is wrong as described. ->fsync() *can* be called for file opened r/o. Moreover, fs might decide to go r/o on its own. file_fsync() is not widely used, but users include affs, bfs, fat, hfs, hfsplus and ufs. No go, at least as long as file_fsync() is used for those. Patch dropped until we decide what to do with those. It shouldn't be all that hard to push file_fsync() out of existence, so it might be doable, but... -- 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
