2010/9/28 Jeff Layton <jlayton@xxxxxxxxx>: > Ok, so suppose you make this change and don't change > cifs_file_aio_write at all, and at some point in the future, we make > cifs do async writes... > > I go and do a buffered write to a file, cifs_file_aio_write then calls > filemap_fdatawrite to flush out the data, but doesn't necessarily wait > for it to complete before returning. > > Now, I do I read from that file descriptor, which ends up being > satisfied before the write I just did got flushed out. I see old > data in the read. > > According to the write(2) manpage: > > POSIX Ârequires Âthat Âa Âread(2) Âwhich can be proved to occur after a > write() has returned returns the new data. ÂNote that not all file sysâ > tems are POSIX conforming. > > It's arguable that CIFS isn't POSIX conforming, but we do our best with > it. I think we ought to conform to POSIX in this regard. If you want > to do this, you probably also need to ensure that the dirty data gets > flushed to the server prior to returning from a write call. It's only read part and I am going to post write part as well. Of course, it will wait for complete. So, let's look at this only from the read side - at least, it doesn't break write related things in comparison with existing code. -- Best regards, Pavel Shilovsky. -- 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
