mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: "Nicholas A. Bellinger" <nab@linux-iscsi.org>
To: Christoph Hellwig <hch@infradead.org>
Cc: "Nicholas A. Bellinger" <nab@daterainc.com>,
	target-devel <target-devel@vger.kernel.org>,
	lkml <linux-kernel@vger.kernel.org>,
	linux-scsi <linux-scsi@vger.kernel.org>,
	Christoph Hellwig <hch@lst.de>, Hannes Reinecke <hare@suse.de>,
	Martin Petersen <martin.petersen@oracle.com>,
	Chris Mason <chris.mason@fusionio.com>,
	James Bottomley <JBottomley@Parallels.com>
Subject: Re: [PATCH 8/9] target: Add support for COMPARE_AND_WRITE emulation
Date: Wed, 21 Aug 2013 10:47:29 -0700	[thread overview]
Message-ID: <1377107249.32763.54.camel@haakon3.risingtidesystems.com> (raw)
In-Reply-To: <20130821161408.GB30848@infradead.org>

On Wed, 2013-08-21 at 09:14 -0700, Christoph Hellwig wrote:
> I don't like the layering here.  The re-execution of the same command
> for both reading and writing the data from/to the backend device already
> looks sketchy here due to doubling work of task attribute handling, the
> various state bits, etc.  And it will only get more complicated when
> the required locking is added.  In addition we have all that confusion
> about overloading the data direction.
> 
> I think the way this should be handled is:
> 
> 
>  - cmd->execute_cmd gets set to a new sbc_emulate_compare_and_write
>  - sbc_emulate_compare_and_write does all the setup for the locking,
>    sets up the read buffer, then calls ops->execute_rw to do the
>    read.  The complete callback does the comparism, then calls
>    ops->execute_rw to do the write, and the second time we hit
>    the complete callback we teard down the read buffer, stop the
>    locking, etc.
> 

I do like this approach better, however calling ops->execute_rw() the
second time around does require at least the TRANSPORT_PROCESSING and
other transport_state bits from target_execute_cmd() to be set for
things to work correctly.  Bypassing the aborted + CMD_*STOP checks
should be OK for the write submission, and will kick (if necessary)
during the final completion.

Setting up the read buffer from sbc_emulate_compare_and_write() would
require two extra COMPARE_AND_WRITE specific se_cmd elements, so I'm
tempted to continue to use the bidi fields for this (because they
already exist) with transport_generic_get_mem_bidi(), and use a
SCF_COMPARE_AND_WRITE flag to avoid any CDB specific checks in
transport_complete_ok(), and reverse dma direction mapping case.

> This also avoids bloating the command with another function pointer
> or having to change all execute_cmd prototypes.

Indeed.

--nab


  reply	other threads:[~2013-08-21 17:40 UTC|newest]

Thread overview: 28+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2013-08-20 20:07 [PATCH 0/9] target: Add support for COMPARE_AND_WRITE (VAAI) emulation Nicholas A. Bellinger
2013-08-20 20:07 ` [PATCH 1/9] scsi: Add CDB definition for COMPARE_AND_WRITE Nicholas A. Bellinger
2013-08-21  6:30   ` Christoph Hellwig
2013-08-20 20:07 ` [PATCH 2/9] target: Add return for se_cmd->transport_complete_callback Nicholas A. Bellinger
2013-08-20 20:07 ` [PATCH 3/9] target: Add memory allocation for bidirectional commands Nicholas A. Bellinger
2013-08-20 20:07 ` [PATCH 4/9] target: Add TCM_MISCOMPARE_VERIFY sense handling Nicholas A. Bellinger
2013-08-20 20:07 ` [PATCH 5/9] target: Skip ->queue_data_in() callbacks for COMPARE_AND_WRITE Nicholas A. Bellinger
2013-08-21  6:32   ` Christoph Hellwig
2013-08-21  7:20     ` Nicholas A. Bellinger
2013-08-20 20:07 ` [PATCH 6/9] target: Allow sbc_ops->execute_rw() to accept SGLs + data_direction Nicholas A. Bellinger
2013-08-21  6:35   ` Christoph Hellwig
2013-08-21  7:26     ` Nicholas A. Bellinger
2013-08-20 20:07 ` [PATCH 7/9] target: Add transport_reset_sgl_orig() for COMPARE_AND_WRITE Nicholas A. Bellinger
2013-08-20 20:07 ` [PATCH 8/9] target: Add support for COMPARE_AND_WRITE emulation Nicholas A. Bellinger
2013-08-21 16:14   ` Christoph Hellwig
2013-08-21 17:47     ` Nicholas A. Bellinger [this message]
2013-08-20 20:08 ` [PATCH 9/9] tcm_qla2xxx: Add special case for COMPARE_AND_WRITE data_direction Nicholas A. Bellinger
2013-08-21  6:37   ` Christoph Hellwig
2013-08-21  7:31     ` Nicholas A. Bellinger
2013-08-21 16:04       ` Christoph Hellwig
2013-08-21 14:38   ` Roland Dreier
2013-08-21 15:53     ` Christoph Hellwig
2013-08-21 17:52       ` Nicholas A. Bellinger
2013-08-21 16:47     ` Roland Dreier
2013-08-20 21:29 ` [PATCH 0/9] target: Add support for COMPARE_AND_WRITE (VAAI) emulation Christoph Hellwig
2013-08-20 21:53   ` Nicholas A. Bellinger
2013-08-20 22:01     ` Douglas Gilbert
2013-08-20 22:19       ` Nicholas A. Bellinger

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=1377107249.32763.54.camel@haakon3.risingtidesystems.com \
    --to=nab@linux-iscsi.org \
    --cc=JBottomley@Parallels.com \
    --cc=chris.mason@fusionio.com \
    --cc=hare@suse.de \
    --cc=hch@infradead.org \
    --cc=hch@lst.de \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-scsi@vger.kernel.org \
    --cc=martin.petersen@oracle.com \
    --cc=nab@daterainc.com \
    --cc=target-devel@vger.kernel.org \
    /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

all inboxes | Powered by JetHome®