Re: [PATCH 5/8] iomap: optionally use ioends for direct I/O

[Christoph Hellwig <hch@lst.de>]

On Thu, Dec 12, 2024 at 11:56:24AM -0800, Darrick J. Wong wrote:
> >  		return 0;
> > +	if (ioend->io_flags & IOMAP_IOEND_DIRECT)
> > +		return iomap_finish_ioend_direct(ioend);
> >  	return iomap_finish_ioend_buffered(ioend);
> >  }
> 
> I'm a little surprised that more of the iomap_ioend* functions didn't
> end up in ioend.c.

See the cover letter.  For development I wanted to avoid churn.  Once
we have general approval for the concept I'd like to move more code.

> > +			WRITE_ONCE(dio->submit.waiter, NULL);
> > +			blk_wake_io_task(waiter);
> > +		} else if (!inode->i_mapping->nrpages) {
> > +			WRITE_ONCE(iocb->private, NULL);
> > +
> > +			/*
> > +			 * We must never invalidate pages from this thread to
> > +			 * avoid deadlocks with buffered I/O completions.
> > +			 * Tough luck if you hit the tiny race with someone
> > +			 * dirtying the range now.
> 
> What happens, exactly?  Does that mean that the dirty pagecache always
> survives?

Yes.