On 06/23/13 23:13, Mike Christie wrote:
> On 06/12/2013 08:28 AM, Bart Van Assche wrote:
>> + /*
>> + * It can occur that after fast_io_fail_tmo expired and before
>> + * dev_loss_tmo expired that the SCSI error handler has
>> + * offlined one or more devices. scsi_target_unblock() doesn't
>> + * change the state of these devices into running, so do that
>> + * explicitly.
>> + */
>> + spin_lock_irq(shost->host_lock);
>> + __shost_for_each_device(sdev, shost)
>> + if (sdev->sdev_state == SDEV_OFFLINE)
>> + sdev->sdev_state = SDEV_RUNNING;
>> + spin_unlock_irq(shost->host_lock);
>
> Is it possible for this to race with scsi_eh_offline_sdevs? Can it be
> looping over cmds offlining devices while this is looping over devices
> onlining them?
>
> It seems this can also happen for all transports/drivers. Maybe a a scsi
> eh/lib helper function that syncrhonizes with the scsi eh completion
> would be better.
I'm not sure it's possible to avoid such a race without introducing
a new mutex. How about something like the (untested) SCSI core patch
below, and invoking scsi_block_eh() and scsi_unblock_eh() around any
reconnect activity not initiated from the SCSI EH thread ?
[PATCH] Add scsi_block_eh() and scsi_unblock_eh()
---
drivers/scsi/hosts.c | 1 +
drivers/scsi/scsi_error.c | 10 ++++++++++
include/scsi/scsi_host.h | 1 +
3 files changed, 12 insertions(+)
diff --git a/drivers/scsi/hosts.c b/drivers/scsi/hosts.c
index 17e2ccb..0df3ec8 100644
--- a/drivers/scsi/hosts.c
+++ b/drivers/scsi/hosts.c
@@ -360,6 +360,7 @@ struct Scsi_Host *scsi_host_alloc(struct scsi_host_template
*sht, int privsize)
init_waitqueue_head(&shost->host_wait);
mutex_init(&shost->scan_mutex);
+ mutex_init(&shost->block_eh_mutex);
/*
* subtract one because we increment first then return, but we need to
diff --git a/drivers/scsi/scsi_error.c b/drivers/scsi/scsi_error.c
index ab16930..566daaa 100644
--- a/drivers/scsi/scsi_error.c
+++ b/drivers/scsi/scsi_error.c
@@ -551,6 +551,10 @@ static int scsi_begin_eh(struct Scsi_Host *host)
{
int res;
+ res = mutex_lock_interruptible(&host->block_eh_mutex);
+ if (res)
+ goto out;
+
spin_lock_irq(host->host_lock);
switch (host->shost_state) {
case SHOST_DEL:
@@ -565,6 +569,10 @@ static int scsi_begin_eh(struct Scsi_Host *host)
}
spin_unlock_irq(host->host_lock);
+ if (res)
+ mutex_unlock(&host->block_eh_mutex);
+
+out:
return res;
}
@@ -579,6 +587,8 @@ static void scsi_end_eh(struct Scsi_Host *host)
if (host->eh_active == 0)
wake_up(&host->host_wait);
spin_unlock_irq(host->host_lock);
+
+ mutex_unlock(&host->block_eh_mutex);
}
/**
diff --git a/include/scsi/scsi_host.h b/include/scsi/scsi_host.h
index 9785e51..d7ce065 100644
--- a/include/scsi/scsi_host.h
+++ b/include/scsi/scsi_host.h
@@ -573,6 +573,7 @@ struct Scsi_Host {
spinlock_t *host_lock;
struct mutex scan_mutex;/* serialize scanning activity */
+ struct mutex block_eh_mutex; /* block ML LLD EH calls */
struct list_head eh_cmd_q;
struct task_struct * ehandler; /* Error recovery thread. */
--
1.7.10.4
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
