On Mon, Oct 31, 2016 at 02:07:54PM +0100, Christoph Hellwig wrote: > On Mon, Oct 31, 2016 at 10:23:31AM +1100, Dave Chinner wrote: > > This doesn't belong in this patchset. > > It does. I can't fix up the calling conventions for a methods that > was never implemented.
That sounds like a problem with your fix - it should work regardless of whether a valid/implemented AIO function is called or not, right? There's no difference between an invalid command, IOCB_CMD_FSYNC where ->aio_fsync() is null, or some supported command that immediately returns -EIO, the end result should be the same... > > Regardless, can we just implement the damned thing rather than > > removing it? Plenty of people have asked for it and they still want > > this functionality. I've sent a couple of different prototypes that > > worked but got bikeshedded to death, and IIRC Ben also tried to get > > it implemented but that went nowhere because other parts of his > > patchset got bikeshedded to death. > > > > If nothing else, just let me implement it in XFS like I did the > > first time so when all the bikshedding stops we can convert it to > > the One True AIO Interface that is decided on. > > I'm not going to complain about a proper implementation, but right now > we don't have any, and I'm not even sure the method signature is > all that suitable. E.g. for the in-kernel users we'd really want a > ranged fsync like the normal fsync anyway. You mean like this version I posted a year ago: https://lkml.org/lkml/2015/10/29/517 Cheers, Dave. -- Dave Chinner da...@fromorbit.com
