Messages in this thread | | | Date | Sun, 21 Dec 2003 16:12:03 -0500 | From | Ben Slusky <> | Subject | Re: [PATCH] loop.c patches, take two |
| |
On Sun, 21 Dec 2003 22:49:47 +0200, Mika Penttil? wrote: > Yet another Big Loop Patch... :) > > It's not obvious which parts are bug fixes, and which performance > improvements. What exactly breaks loops on journalling filesystems, and > how do you solve it?
See <URL:http://www.ussg.iu.edu/hypermail/linux/kernel/0310.3/1151.html>... I'd submitted these patches a while back. Andrew observed that the problems they fixed did not manifest in file-backed loops, so his solution (which was merged into -mm but not mainstream) was to cut out the block-backed code path entirely. THAT is what breaks journaling filesystems on loops (note: not vice versa).
> What's the clone_bio() business? Why on reads only?
There's no need to allocate memory for a second copy of the data on a read. This is a performance improvenment but I'd say it's closely related to the main point of the patch (i.e. take what pages you can get and recycle them); I'm making the block-backed code path significantly more complex and at the same time having reads take a shortcut. But I could split that into a separate patch if desired.
-- Ben Slusky | Yakka foob mog. Grug pubbawup sluskyb@paranoiacs.org | zink wattoom gazork. Chumble sluskyb@stwing.org | spuzz. PGP keyID ADA44B3B | -Calvin - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |