This is my version of the AIO O_SYNC series, which is a lot simpler than the previous attempts. The first patch adds generic support for deferring AIO completions to a workqueue by taking the XFS code that alter got copied to ext4 to the core. At least the XFS code could be further simplified after this and a memory allocation could be dropped, but due to Dave having other changes in that area pending I'm not going to add those bits yet. The second patch simply adds a call to generic_write_sync from the I/O completion handler and makes sure it is called from user context if we have to. This works fine for the "normal" blkdev_direct_IO callers and XFS, but it currently breaks ext4 due to the mess that ext4 I/O completions are. We could work around this by requiring the filesystem to call generic_write_sync from its own end_io handler if we we absolutely need to, but I'd be much happier if someone who understands ext4 better than I could fix it. More details in the second patch. -- 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
