Re: [PATCH] btrfs: delete unused member nobarriers

[Anand Jain <anand.jain@oracle.com>]

On 04/03/2017 08:06 PM, David Sterba wrote:
> Please update the changelog to say why it's ok to remove it, eg. the
> commit that removed the last user.
>
> commit b25de9d6da49b1a8760a89672283128aa8c78345
> Author: Christoph Hellwig <hch@lst.de>
> Date:   Fri Apr 24 21:41:01 2015 +0200
>
>     block: remove BIO_EOPNOTSUPP

  Ah. I should have search the commit log. sorry about that.


IMO, there is a bug in generic_make_request_checks() in which
it should rather return EOPNOTSUPP, instead of EIO if QUEUE_FLAG_WC
is not supported.

------------------------------------------------
1853 static noinline_for_stack bool
1854 generic_make_request_checks(struct bio *bio)
1855 {
::

1892         /*
1893          * Filter flush bio's early so that make_request based
1894          * drivers without flush support don't have to worry
1895          * about them.
1896          */
1897         if ((bio->bi_opf & (REQ_PREFLUSH | REQ_FUA)) &&
1898             !test_bit(QUEUE_FLAG_WC, &q->queue_flags)) {
1899                 bio->bi_opf &= ~(REQ_PREFLUSH | REQ_FUA);
1900                 if (!nr_sectors) {
1901                         err = 0;
1902                         goto end_io; <- this should goto  not_supported
1903                 }
1904         }

::

1946 not_supported:
1947         err = -EOPNOTSUPP;
------------------------------------------------


  Pls ignore this patch.

  I have submitted
[PATCH] btrfs: check if the device is flush capable

  which will remain unaffected by the above bug/not-a-bug
  in the blk core code.

Thanks, Anand


> Otherwise the patch is ok.