On Thu, May 26, 2022 at 10:35:11PM -0700, Christoph Hellwig wrote: > On Thu, May 26, 2022 at 08:29:02PM +0100, Matthew Wilcox (Oracle) wrote: > > Al's patch. > > > > Signed-off-by: Matthew Wilcox (Oracle) <willy@xxxxxxxxxxxxx> > > Hmm, that is a bit of a weird changelog.. I took parts of https://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs.git/commit/?h=new.iov_iter&id=447262443a4dbf88ae5b21de5c77197f084cdca4 and fixed it up a bit but dropped the btrfs parts. > > /* for data sync or sync, we need sync completion processing */ > > - if (iocb->ki_flags & IOCB_DSYNC) > > + if (iocb->ki_flags & IOCB_DSYNC && > > + !(dio_flags & IOMAP_DIO_NOSYNC)) > > dio->flags |= IOMAP_DIO_NEED_SYNC; > > > > /* > > I think we also need to skip the setting of IOMAP_DIO_WRITE_FUA below > (or find a way to communicate back to the that FUA was used, which > seems way more complicated) Probably ... I was just looking to avoid the deadlock. Patch not for upstream in this form; I'm expecting Al to submit a fixed version before the JFS code is ready for upstream.
