[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Fri, 10 Sep 2021 07:42:05 +1000
From: Dave Chinner <david@...morbit.com>
To: Jens Axboe <axboe@...nel.dk>
Cc: Linus Torvalds <torvalds@...ux-foundation.org>,
Al Viro <viro@...iv.linux.org.uk>,
Pavel Begunkov <asml.silence@...il.com>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
linux-fsdevel <linux-fsdevel@...r.kernel.org>
Subject: Re: [git pull] iov_iter fixes
On Thu, Sep 09, 2021 at 03:19:56PM -0600, Jens Axboe wrote:
> On 9/9/21 1:37 PM, Linus Torvalds wrote:
> > I'd like the comments expanded too. In particular that
> >
> > /* some cases will consume bytes even on error returns */
>
> That comment is from me, and it goes back a few years. IIRC, it was the
> iomap or xfs code that I hit this with, but honestly I don't remember
> all the details at this point. I can try and play with it and see if it
> still reproduces.
You might well be thinking of the problem fixed by commit
883a790a8440 ("xfs: don't allow NOWAIT DIO across extent
boundaries").
This fix was indicative of a whole class of issues with IOCB_NOWAIT
being used for multi-IO operations at the filesystem level and being
applied to each sub-segment of the IO that was constructed, rather
than the IO as a whole. Hence a failure on the second or subsequent
segments could return -EAGAIN (and potentially other errors) to the
caller after the segments we successfully submitted consumed part of
the iov...
Cheers,
Dave.
--
Dave Chinner
david@...morbit.com
Powered by blists - more mailing lists