linux-scsi.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Mike Christie <michaelc@cs.wisc.edu>
To: Bart Van Assche <bvanassche@acm.org>
Cc: linux-scsi <linux-scsi@vger.kernel.org>,
	James Bottomley <jbottomley@parallels.com>,
	Jun'ichi Nomura <j-nomura@ce.jp.nec.com>,
	Stefan Richter <stefanr@s5r6.in-berlin.de>,
	Tomas Henzl <thenzl@redhat.com>,
	Mike Snitzer <snitzer@redhat.com>
Subject: Re: [PATCH 2/3] Stop accepting SCSI requests before removing a device
Date: Fri, 04 May 2012 15:30:42 -0500	[thread overview]
Message-ID: <4FA43C72.3000108@cs.wisc.edu> (raw)
In-Reply-To: <4FA43912.2060706@cs.wisc.edu>

[-- Attachment #1: Type: text/plain, Size: 1781 bytes --]

On 05/04/2012 03:16 PM, Mike Christie wrote:
> On 05/04/2012 10:06 AM, Bart Van Assche wrote:
>> diff --git a/drivers/scsi/scsi_sysfs.c b/drivers/scsi/scsi_sysfs.c
>> index 42c35ff..f8fc240 100644
>> --- a/drivers/scsi/scsi_sysfs.c
>> +++ b/drivers/scsi/scsi_sysfs.c
>> @@ -955,12 +955,20 @@ int scsi_sysfs_add_sdev(struct scsi_device *sdev)
>>  void __scsi_remove_device(struct scsi_device *sdev)
>>  {
>>  	struct device *dev = &sdev->sdev_gendev;
>> +	struct request_queue *q = sdev->request_queue;
>> +
>> +	/*
>> +	 * Stop accepting new requests before tearing down the
>> +	 * device. Note: the actual queue deallocation happens in
>> +	 * scsi_device_dev_release_usercontext().
>> +	 */
>> +	blk_cleanup_queue(q);
>>  
>>  	if (sdev->is_visible) {
>>  		if (scsi_device_set_state(sdev, SDEV_CANCEL) != 0)
>>  			return;
>>  
>> -		bsg_unregister_queue(sdev->request_queue);
>> +		bsg_unregister_queue(q);
>>  		device_unregister(&sdev->sdev_dev);
>>  		transport_remove_device(dev);
>>  		device_del(dev);
>> @@ -971,8 +979,6 @@ void __scsi_remove_device(struct scsi_device *sdev)
> 
> My fault. The blk_cleanup_queue call should be right before the
> scsi_device_set_state(sdev, SDEV_DEL); call in this function.
> 
> It needs to be after the device_del call, because that triggers the SCSI
> ULD removal which can send IO.
> 

I made the attached patch. It allows the scsi ULD removal IO to execute
ok. I tested with iscsi devices that have write caches.

With your patch I get:

Apr 27 19:09:51 noisyr6 kernel: sd 11:0:0:1: [sdc] Synchronizing SCSI cache
Apr 27 19:09:51 noisyr6 kernel: sd 11:0:0:1: [sdc] Result:
hostbyte=DID_TRANSPORT_FAILFAST driverbyte=DRIVER_OK


With my patch I just get:

May  4 16:29:27 noisyr6 kernel: sd 8:0:0:1: [sdc] Synchronizing SCSI cache

[-- Attachment #2: scsi-clean-queue-after-del.patch --]
[-- Type: text/x-patch, Size: 1165 bytes --]

diff --git a/drivers/scsi/scsi_sysfs.c b/drivers/scsi/scsi_sysfs.c
index 42c35ff..11dc3ff 100644
--- a/drivers/scsi/scsi_sysfs.c
+++ b/drivers/scsi/scsi_sysfs.c
@@ -955,24 +955,30 @@ int scsi_sysfs_add_sdev(struct scsi_device *sdev)
 void __scsi_remove_device(struct scsi_device *sdev)
 {
 	struct device *dev = &sdev->sdev_gendev;
+	struct request_queue *q = sdev->request_queue;
 
 	if (sdev->is_visible) {
 		if (scsi_device_set_state(sdev, SDEV_CANCEL) != 0)
 			return;
 
-		bsg_unregister_queue(sdev->request_queue);
+		bsg_unregister_queue(q);
 		device_unregister(&sdev->sdev_dev);
 		transport_remove_device(dev);
 		device_del(dev);
 	} else
 		put_device(&sdev->sdev_dev);
+	/*
+	 * Stop accepting new requests before tearing down the
+	 * queue. Note: the actual queue deallocation happens in
+	 * scsi_device_dev_release_usercontext().
+	 */
+	blk_cleanup_queue(q);
+
 	scsi_device_set_state(sdev, SDEV_DEL);
 	if (sdev->host->hostt->slave_destroy)
 		sdev->host->hostt->slave_destroy(sdev);
 	transport_destroy_device(dev);
 
-	/* Freeing the queue signals to block that we're done */
-	blk_cleanup_queue(sdev->request_queue);
 	put_device(dev);
 }
 

  reply	other threads:[~2012-05-04 20:31 UTC|newest]

Thread overview: 21+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2012-05-04 15:00 [PATCH 0/3 v6] Fixes for SCSI device removal Bart Van Assche
2012-05-04 15:03 ` [PATCH 1/3] sd: Fix device removal NULL pointer dereference Bart Van Assche
2012-05-04 15:06 ` [PATCH 2/3] Stop accepting SCSI requests before removing a device Bart Van Assche
2012-05-04 20:16   ` Mike Christie
2012-05-04 20:30     ` Mike Christie [this message]
2012-05-05 13:04       ` Bart Van Assche
2012-05-29 15:00         ` Bart Van Assche
2012-05-29 17:35           ` Mike Christie
2012-05-30  6:56             ` Bart Van Assche
2012-05-30 17:27               ` Mike Christie
2012-05-30 20:00                 ` Bart Van Assche
2012-06-01  3:13                   ` Mike Christie
2012-05-04 15:07 ` [PATCH 3/3] Make scsi_free_queue() abort pending requests Bart Van Assche
2012-05-04 20:25   ` Mike Christie
2012-05-04 20:32     ` Mike Christie
2012-05-05  6:07       ` Bart Van Assche
2012-05-07  0:44         ` Mike Christie
2012-05-07  1:15           ` Mike Christie
2012-05-14 18:43           ` Bart Van Assche
2012-05-29 14:56             ` Bart Van Assche
2012-05-05 13:41     ` Bart Van Assche

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=4FA43C72.3000108@cs.wisc.edu \
    --to=michaelc@cs.wisc.edu \
    --cc=bvanassche@acm.org \
    --cc=j-nomura@ce.jp.nec.com \
    --cc=jbottomley@parallels.com \
    --cc=linux-scsi@vger.kernel.org \
    --cc=snitzer@redhat.com \
    --cc=stefanr@s5r6.in-berlin.de \
    --cc=thenzl@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).