Re: [PATCH v7 0/3] FDP and per-io hints
Keith Busch <[email protected]>
| Newsgroups | gmane.linux.kernel.aio.general,gmane.linux.file-systems,gmane.linux.kernel.io-uring,gmane.linux.block |
|---|---|
| Message-ID | <[email protected]> |
On Fri, Oct 11, 2024 at 02:21:02PM +0200, Javier Gonzalez wrote: > > That's a lot of marketing babble :) What exact thing is missing > > from the passthrough interface when using say spdx over io_uring? > > The block layer provides a lot of functionality that passthru cannot > provide. A simple example would be splits. You know this :) > > I am sure Jens and Keith can give you more specifics on their particular > reasons. Splitting, merging, cgroups, iostats, error retries, failover. We have a recent change staged in 6.13 to try to count iostats on passthrough, but it doesn't do discards. Passthrough to partitions requires root access and the slow CAP_SYSADMIN check on every IO. Page cache. Though that may not readily work with this io_uring proposal as-is either. The generic IO path is safe to changing formats; passthrough isn't. And it's just generally more work to port existing applications to use the passthrough interface. It causes duplicating solutions to scenarios the kernel already handles. -- To unsubscribe, send a message with 'unsubscribe linux-aio' in the body to [email protected]. For more info on Linux AIO, see: http://www.kvack.org/aio/ Don't email: <a href=mailto:"[email protected]">[email protected]</a>