All of lore.kernel.org
 help / color / mirror / Atom feed
From: Wei Fang <fangwei1@huawei.com>
To: James Bottomley <jejb@linux.vnet.ibm.com>, emilne@redhat.com
Cc: Bart Van Assche <bart.vanassche@sandisk.com>,
	"martin.petersen@oracle.com" <martin.petersen@oracle.com>,
	"linux-scsi@vger.kernel.org" <linux-scsi@vger.kernel.org>,
	chenzengxi@huawei.com
Subject: Re: [PATCH] scsi: avoid a permanent stop of the scsi device's request queue
Date: Thu, 8 Dec 2016 14:38:16 +0800	[thread overview]
Message-ID: <5848FFD8.5070606@huawei.com> (raw)
In-Reply-To: <5848D205.8060007@huawei.com>

Hi, James, Ewan, Bart,

On 2016/12/8 11:22, Wei Fang wrote:
> I looked through those code and found that if we fix this bug
> by removing setting the state in scsi_sysfs_add_sdev(), it
> can't be fixed completely:
> 
> scsi_device_set_state(sdev, SDEV_RUNNING) in scsi_add_lun() and
> scsi_device_set_state(sdev, SDEV_CREATED_BLOCK) in scsi_internal_device_block()
> can be called simultaneously. Because there is no synchronization
> between scsi_device_set_state(), those calls may both return
> success, and the state may be SDEV_RUNNING after that, and the
> device queue is stopped.

Can we fix it in this way:

Add a state lock to make sure the result of simultaneously calling
of scsi_device_set_state() is not unpredictable.

diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c
index 253ee74..80cb493 100644
--- a/drivers/scsi/scsi_lib.c
+++ b/drivers/scsi/scsi_lib.c
@@ -2457,10 +2457,16 @@ EXPORT_SYMBOL(scsi_test_unit_ready);
 int
 scsi_device_set_state(struct scsi_device *sdev, enum scsi_device_state state)
 {
-       enum scsi_device_state oldstate = sdev->sdev_state;
+       enum scsi_device_state oldstate;
+       unsigned long flags;
+
+       spin_lock_irqsave(&sdev->state_lock, flags);
+       oldstate = sdev->sdev_state;

-       if (state == oldstate)
+       if (state == oldstate) {
+               spin_unlock_irqrestore(&sdev->state_lock, flags);
                return 0;
+       }

        switch (state) {
        case SDEV_CREATED:
@@ -2558,9 +2564,11 @@ scsi_device_set_state(struct scsi_device *sdev, enum scsi_device_state state)

        }
        sdev->sdev_state = state;
+       spin_unlock_irqrestore(&sdev->state_lock, flags);
        return 0;

  illegal:
+       spin_unlock_irqrestore(&sdev->state_lock, flags);
        SCSI_LOG_ERROR_RECOVERY(1,
                                sdev_printk(KERN_ERR, sdev,
                                            "Illegal state transition %s->%s",
diff --git a/drivers/scsi/scsi_scan.c b/drivers/scsi/scsi_scan.c
index 6f7128f..ba2f38f 100644
--- a/drivers/scsi/scsi_scan.c
+++ b/drivers/scsi/scsi_scan.c
@@ -238,6 +238,7 @@ static struct scsi_device *scsi_alloc_sdev(struct scsi_target *starget,
        INIT_LIST_HEAD(&sdev->starved_entry);
        INIT_LIST_HEAD(&sdev->event_list);
        spin_lock_init(&sdev->list_lock);
+       spin_lock_init(&sdev->state_lock);
        mutex_init(&sdev->inquiry_mutex);
        INIT_WORK(&sdev->event_work, scsi_evt_thread);
        INIT_WORK(&sdev->requeue_work, scsi_requeue_run_queue);
diff --git a/drivers/scsi/scsi_sysfs.c b/drivers/scsi/scsi_sysfs.c
index 0734927..82dfe07 100644
--- a/drivers/scsi/scsi_sysfs.c
+++ b/drivers/scsi/scsi_sysfs.c
@@ -1204,10 +1204,6 @@ int scsi_sysfs_add_sdev(struct scsi_device *sdev)
        struct request_queue *rq = sdev->request_queue;
        struct scsi_target *starget = sdev->sdev_target;

-       error = scsi_device_set_state(sdev, SDEV_RUNNING);
-       if (error)
-               return error;
-
        error = scsi_target_add(starget);
        if (error)
                return error;
diff --git a/include/scsi/scsi_device.h b/include/scsi/scsi_device.h
index 8990e58..e00764e 100644
--- a/include/scsi/scsi_device.h
+++ b/include/scsi/scsi_device.h
@@ -31,7 +31,7 @@ struct scsi_mode_data {
 enum scsi_device_state {
        SDEV_CREATED = 1,       /* device created but not added to sysfs
                                 * Only internal commands allowed (for inq) */
-       SDEV_RUNNING,           /* device properly configured
+       SDEV_RUNNING,           /* device properly initialized
                                 * All commands allowed */
        SDEV_CANCEL,            /* beginning to delete device
                                 * Only error handler commands allowed */
@@ -207,6 +207,7 @@ struct scsi_device {
        void                    *handler_data;

        unsigned char           access_state;
+       spinlock_t              state_lock;
        enum scsi_device_state sdev_state;
        unsigned long           sdev_data[0];
 } __attribute__((aligned(sizeof(unsigned long))));

Haven't tested yet. Sending this for your opinion.

Thanks,
Wei


  reply	other threads:[~2016-12-08  6:45 UTC|newest]

Thread overview: 22+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-12-06  9:12 [PATCH] scsi: avoid a permanent stop of the scsi device's request queue Wei Fang
2016-12-06 15:51 ` Bart Van Assche
2016-12-07  1:20   ` Wei Fang
2016-12-07  2:45     ` Bart Van Assche
2016-12-07  3:41       ` Wei Fang
2016-12-07  4:40         ` Bart Van Assche
2016-12-07  6:59           ` Wei Fang
2016-12-07 16:48             ` Bart Van Assche
2016-12-07 16:55               ` Bart Van Assche
2016-12-07 17:40                 ` Ewan D. Milne
2016-12-07 18:16                   ` James Bottomley
2016-12-07 19:24                     ` Ewan D. Milne
2016-12-07 20:09                       ` James Bottomley
2016-12-07 20:30                         ` Ewan D. Milne
2016-12-07 23:43                           ` James Bottomley
2016-12-08  2:28                             ` Wei Fang
2016-12-08  2:33                               ` James Bottomley
2016-12-08  3:22                                 ` Wei Fang
2016-12-08  6:38                                   ` Wei Fang [this message]
2016-12-08 14:04                                     ` Ewan D. Milne
2016-12-08 15:39                                   ` James Bottomley
2016-12-09  1:08                                     ` Wei Fang

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=5848FFD8.5070606@huawei.com \
    --to=fangwei1@huawei.com \
    --cc=bart.vanassche@sandisk.com \
    --cc=chenzengxi@huawei.com \
    --cc=emilne@redhat.com \
    --cc=jejb@linux.vnet.ibm.com \
    --cc=linux-scsi@vger.kernel.org \
    --cc=martin.petersen@oracle.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.