linux-block.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Martin Wilck <mwilck@suse.com>
To: Bart Van Assche <bvanassche@acm.org>,
	Mike Snitzer <snitzer@redhat.com>,
	Alasdair G Kergon <agk@redhat.com>,
	dm-devel@redhat.com, Hannes Reinecke <hare@suse.de>
Cc: Daniel Wagner <dwagner@suse.de>,
	linux-block@vger.kernel.org, Paolo Bonzini <pbonzini@redhat.com>,
	Christoph Hellwig <hch@lst.de>
Subject: Re: [dm-devel] [RFC PATCH v2 2/2] dm: add CONFIG_DM_MULTIPATH_SG_IO - failover for SG_IO on dm-multipath
Date: Thu, 29 Apr 2021 23:08:23 +0200	[thread overview]
Message-ID: <413c998fd75f6bc9c7f70d8a1a2b107b443a75a0.camel@suse.com> (raw)
In-Reply-To: <5d1967f2-8017-c711-dec0-3ffe751974de@acm.org>

On Thu, 2021-04-29 at 09:32 -0700, Bart Van Assche wrote:
> On 4/29/21 8:50 AM, mwilck@suse.com wrote:
> > +       if (hdr.dxfer_len > (queue_max_hw_sectors(bdev->bd_disk-
> > >queue) << 9))
> > +               return -EIO;
> 
> How about using SECTOR_SHIFT instead of the number 9?

no problem. That line was just copied from the scsi_ioctl code.

> 
> > +               /*
> > +                * Errors resulting from invalid parameters
> > shouldn't be retried
> > +                * on another path.
> > +                */
> > +               switch (rc) {
> > +               case -ENOIOCTLCMD:
> > +               case -EFAULT:
> > +               case -EINVAL:
> > +               case -EPERM:
> > +                       goto out;
> > +               default:
> > +                       break;
> > +               }
> 
> Will -ENOMEM result in an immediate retry? Is that what's desired?

No, I overlooked that case. Thanks for pointing this out. 

> 
> > +               if (rhdr.info & SG_INFO_CHECK) {
> > +                       int result;
> > +                       blk_status_t sts;
> > +
> > +                       __set_status_byte(&result, rhdr.status);
> > +                       __set_msg_byte(&result, rhdr.msg_status);
> > +                       __set_host_byte(&result, rhdr.host_status);
> > +                       __set_driver_byte(&result,
> > rhdr.driver_status);
> > +
> > +                       sts = __scsi_result_to_blk_status(&result,
> > result);
> > +                       rhdr.host_status = host_byte(result);
> > +
> > +                       /* See if this is a target or path error.
> > */
> > +                       if (sts == BLK_STS_OK)
> > +                               rc = 0;
> > +                       else if (blk_path_error(sts))
> > +                               rc = -EIO;
> > +                       else {
> > +                               rc = blk_status_to_errno(sts);
> > +                               goto out;
> > +                       }
> > +               }
> 
> Will SAM_STAT_CHECK_CONDITION be treated as an I/O error? Is that
> what's
> desired? If not, does that mean that scsi_result_to_blk_status()
> shouldn't be used but instead that a custom SCSI result conversion is
> needed?

This mimics the logic for regular SCSI block I/O. By default, CHECK
CONDITION indeed maps to a BLK_STS_IO_ERR, and will be treated as a
path error. As you probably know, there are some exceptions that are
handled in the SCSI mid-layer beforehand. For example, check_sense()
sets DID_TARGET_FAILURE or DID_MEDIUM_ERROR for certain sense codes,
which map to target errors. So yes, I think this is correct.

> If __scsi_result_to_blk_status() is the right function to call, how
> about making that function accept the driver status, host status, msg
> and SAM status as four separate arguments such that the
> __set_*_byte()
> calls can be left out?
> 
> > +                       char *argv[2] = { "fail_path", bdbuf };
> 
> Can the above array be declared static?

How would that work? The function needs to be reentrant, and bdbuf is
on the stack. I don't see an issue here, it's really just two pointers
on the stack.

Regards,
Martin



      reply	other threads:[~2021-04-29 21:08 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-04-29 15:50 [RFC PATCH v2 0/2] dm: dm_blk_ioctl(): implement failover for SG_IO on dm-multipath mwilck
2021-04-29 15:50 ` [RFC PATCH v2 1/2] scsi: convert scsi_result_to_blk_status() to inline mwilck
2021-04-29 16:20   ` [dm-devel] " Bart Van Assche
2021-04-29 20:33     ` Martin Wilck
2021-04-30 21:12       ` Bart Van Assche
2021-04-29 15:50 ` [RFC PATCH v2 2/2] dm: add CONFIG_DM_MULTIPATH_SG_IO - failover for SG_IO on dm-multipath mwilck
2021-04-29 16:32   ` [dm-devel] " Bart Van Assche
2021-04-29 21:08     ` Martin Wilck [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=413c998fd75f6bc9c7f70d8a1a2b107b443a75a0.camel@suse.com \
    --to=mwilck@suse.com \
    --cc=agk@redhat.com \
    --cc=bvanassche@acm.org \
    --cc=dm-devel@redhat.com \
    --cc=dwagner@suse.de \
    --cc=hare@suse.de \
    --cc=hch@lst.de \
    --cc=linux-block@vger.kernel.org \
    --cc=pbonzini@redhat.com \
    --cc=snitzer@redhat.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).