On Mon, Jun 04, 2007 at 10:05:35AM +0200, Jens Axboe wrote: > On Mon, Jun 04 2007, Fengguang Wu wrote: > > Hi Jens, > > > > This is another try, still not in a comfortable state though. > > //Busy waiting is possible for interleaved reads. > > A few random comments... > > Adding an internal flag is fine, but please put it at the upper end of > the spectrum. So, use (1 << 31) for that flag.
OK. > And please work on the #splice branch of the block repo, not -mm. There > are quite a few things pending for inclusion in there, and I'm sure your > patch as-is wont apply. I'm afraid this patch cannot be moved over to your branch trivially. The core of the algorithm reuses f_ra.prev_index to record the state. It is OK for the on-demand readahead in the -mm tree. But the current readahead code in 2.6.22-rc3 is sensible to the change. And it also does not reliably tell if readahead I/O has been submitted. We can either try other ways of doing non-blocking I/O, or just wait until the merge of on-demand readahead? The current patch should work perfect with single splice reader. In the case of multiple readers on the same fd, we might simply err on the side of I/O waiting, since busy EAGAIN looping is not acceptable. Fengguang Wu - 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/
