On Tue, Mar 20, 2012 at 11:56:25PM +0100, Jan Kara wrote: > Move clearing of I_SYNC into inode_sync_complete(). It is more logical to have > clearing of I_SYNC bit and waking of waiters in one place. Also later we will > have two places needing to clear I_SYNC and wake up waiters so this allows them > to use the common helper. Moving of I_SYNC clearing to a later stage of > writeback_single_inode() is safe since we hold i_lock all the time. > > Signed-off-by: Jan Kara <jack@xxxxxxx> The code changes look good, but should we really remove a comment that describes memory barrier? IMHO any undocumented barrier is a bug waiting to happen. Reviewed-by: Christoph Hellwig <hch@xxxxxx> -- 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
