diff mbox

[1/6] ipr: Fix missed EH wakeup

Message ID 20170315215837.5146A13603C@b03ledav002.gho.boulder.ibm.com (mailing list archive)
State Accepted, archived
Headers show

Commit Message

Brian King March 15, 2017, 9:58 p.m. UTC
Following a command abort or device reset, ipr's EH handlers wait
for the commands getting aborted to get sent back from the adapter
prior to returning from the EH handler. This fixes up some cases
where the completion handler was not getting called, which would
have resulted in the EH thread waiting until it timed out, greatly
extending EH time.

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
---

 drivers/scsi/ipr.c |   16 ++++++++++++----
 1 file changed, 12 insertions(+), 4 deletions(-)

Comments

Martin K. Petersen March 23, 2017, 2:05 p.m. UTC | #1
Brian King <brking@linux.vnet.ibm.com> writes:

> Following a command abort or device reset, ipr's EH handlers wait
> for the commands getting aborted to get sent back from the adapter
> prior to returning from the EH handler. This fixes up some cases
> where the completion handler was not getting called, which would
> have resulted in the EH thread waiting until it timed out, greatly
> extending EH time.

Somebody please review this series so I can get it queued up.

Thanks!
wenxiong@linux.vnet.ibm.com March 23, 2017, 3:54 p.m. UTC | #2
On 2017-03-15 16:58, Brian King wrote:

I have reviewed this serial of patches and tested them on IBM systems 
successfully

Thanks for your help!
Wendy
> Following a command abort or device reset, ipr's EH handlers wait
> for the commands getting aborted to get sent back from the adapter
> prior to returning from the EH handler. This fixes up some cases
> where the completion handler was not getting called, which would
> have resulted in the EH thread waiting until it timed out, greatly
> extending EH time.
> 
> Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
> ---
> 
>  drivers/scsi/ipr.c |   16 ++++++++++++----
>  1 file changed, 12 insertions(+), 4 deletions(-)
> 
> diff -puN drivers/scsi/ipr.c~ipr_fix_missed_eh_wakeup 
> drivers/scsi/ipr.c
> ---
> linux-2.6.git/drivers/scsi/ipr.c~ipr_fix_missed_eh_wakeup	2017-03-13
> 14:52:17.974545318 -0500
> +++ linux-2.6.git-bjking1/drivers/scsi/ipr.c	2017-03-13 
> 17:03:13.635568644 -0500
> @@ -836,8 +836,10 @@ static void ipr_sata_eh_done(struct ipr_
> 
>  	qc->err_mask |= AC_ERR_OTHER;
>  	sata_port->ioasa.status |= ATA_BUSY;
> -	list_add_tail(&ipr_cmd->queue, &ipr_cmd->hrrq->hrrq_free_q);
>  	ata_qc_complete(qc);
> +	if (ipr_cmd->eh_comp)
> +		complete(ipr_cmd->eh_comp);
> +	list_add_tail(&ipr_cmd->queue, &ipr_cmd->hrrq->hrrq_free_q);
>  }
> 
>  /**
> @@ -5947,8 +5949,10 @@ static void ipr_erp_done(struct ipr_cmnd
>  		res->in_erp = 0;
>  	}
>  	scsi_dma_unmap(ipr_cmd->scsi_cmd);
> -	list_add_tail(&ipr_cmd->queue, &ipr_cmd->hrrq->hrrq_free_q);
>  	scsi_cmd->scsi_done(scsi_cmd);
> +	if (ipr_cmd->eh_comp)
> +		complete(ipr_cmd->eh_comp);
> +	list_add_tail(&ipr_cmd->queue, &ipr_cmd->hrrq->hrrq_free_q);
>  }
> 
>  /**
> @@ -6338,8 +6342,10 @@ static void ipr_erp_start(struct ipr_ioa
>  	}
> 
>  	scsi_dma_unmap(ipr_cmd->scsi_cmd);
> -	list_add_tail(&ipr_cmd->queue, &ipr_cmd->hrrq->hrrq_free_q);
>  	scsi_cmd->scsi_done(scsi_cmd);
> +	if (ipr_cmd->eh_comp)
> +		complete(ipr_cmd->eh_comp);
> +	list_add_tail(&ipr_cmd->queue, &ipr_cmd->hrrq->hrrq_free_q);
>  }
> 
>  /**
> @@ -6365,8 +6371,10 @@ static void ipr_scsi_done(struct ipr_cmn
>  		scsi_dma_unmap(scsi_cmd);
> 
>  		spin_lock_irqsave(ipr_cmd->hrrq->lock, lock_flags);
> -		list_add_tail(&ipr_cmd->queue, &ipr_cmd->hrrq->hrrq_free_q);
>  		scsi_cmd->scsi_done(scsi_cmd);
> +		if (ipr_cmd->eh_comp)
> +			complete(ipr_cmd->eh_comp);
> +		list_add_tail(&ipr_cmd->queue, &ipr_cmd->hrrq->hrrq_free_q);
>  		spin_unlock_irqrestore(ipr_cmd->hrrq->lock, lock_flags);
>  	} else {
>  		spin_lock_irqsave(ioa_cfg->host->host_lock, lock_flags);
> _
Martin K. Petersen March 23, 2017, 4:04 p.m. UTC | #3
Brian King <brking@linux.vnet.ibm.com> writes:

> Following a command abort or device reset, ipr's EH handlers wait
> for the commands getting aborted to get sent back from the adapter
> prior to returning from the EH handler. This fixes up some cases
> where the completion handler was not getting called, which would
> have resulted in the EH thread waiting until it timed out, greatly
> extending EH time.

Applied to 4.12/scsi-queue.
diff mbox

Patch

diff -puN drivers/scsi/ipr.c~ipr_fix_missed_eh_wakeup drivers/scsi/ipr.c
--- linux-2.6.git/drivers/scsi/ipr.c~ipr_fix_missed_eh_wakeup	2017-03-13 14:52:17.974545318 -0500
+++ linux-2.6.git-bjking1/drivers/scsi/ipr.c	2017-03-13 17:03:13.635568644 -0500
@@ -836,8 +836,10 @@  static void ipr_sata_eh_done(struct ipr_
 
 	qc->err_mask |= AC_ERR_OTHER;
 	sata_port->ioasa.status |= ATA_BUSY;
-	list_add_tail(&ipr_cmd->queue, &ipr_cmd->hrrq->hrrq_free_q);
 	ata_qc_complete(qc);
+	if (ipr_cmd->eh_comp)
+		complete(ipr_cmd->eh_comp);
+	list_add_tail(&ipr_cmd->queue, &ipr_cmd->hrrq->hrrq_free_q);
 }
 
 /**
@@ -5947,8 +5949,10 @@  static void ipr_erp_done(struct ipr_cmnd
 		res->in_erp = 0;
 	}
 	scsi_dma_unmap(ipr_cmd->scsi_cmd);
-	list_add_tail(&ipr_cmd->queue, &ipr_cmd->hrrq->hrrq_free_q);
 	scsi_cmd->scsi_done(scsi_cmd);
+	if (ipr_cmd->eh_comp)
+		complete(ipr_cmd->eh_comp);
+	list_add_tail(&ipr_cmd->queue, &ipr_cmd->hrrq->hrrq_free_q);
 }
 
 /**
@@ -6338,8 +6342,10 @@  static void ipr_erp_start(struct ipr_ioa
 	}
 
 	scsi_dma_unmap(ipr_cmd->scsi_cmd);
-	list_add_tail(&ipr_cmd->queue, &ipr_cmd->hrrq->hrrq_free_q);
 	scsi_cmd->scsi_done(scsi_cmd);
+	if (ipr_cmd->eh_comp)
+		complete(ipr_cmd->eh_comp);
+	list_add_tail(&ipr_cmd->queue, &ipr_cmd->hrrq->hrrq_free_q);
 }
 
 /**
@@ -6365,8 +6371,10 @@  static void ipr_scsi_done(struct ipr_cmn
 		scsi_dma_unmap(scsi_cmd);
 
 		spin_lock_irqsave(ipr_cmd->hrrq->lock, lock_flags);
-		list_add_tail(&ipr_cmd->queue, &ipr_cmd->hrrq->hrrq_free_q);
 		scsi_cmd->scsi_done(scsi_cmd);
+		if (ipr_cmd->eh_comp)
+			complete(ipr_cmd->eh_comp);
+		list_add_tail(&ipr_cmd->queue, &ipr_cmd->hrrq->hrrq_free_q);
 		spin_unlock_irqrestore(ipr_cmd->hrrq->lock, lock_flags);
 	} else {
 		spin_lock_irqsave(ioa_cfg->host->host_lock, lock_flags);