diff mbox

[05/21] hpsa: fix hpsa_adjust_hpsa_scsi_table

Message ID 20151024195256.17127.93962.stgit@brunhilda (mailing list archive)
State Accepted, archived
Headers show

Commit Message

Don Brace Oct. 24, 2015, 7:52 p.m. UTC 
Fix a NULL pointer issue in the driver when devices are removed
during a reset.

Signed-off-by: Don Brace <don.brace@pmcs.com>
---
 drivers/block/cciss.h |    1 +
 drivers/scsi/hpsa.c   |   16 ++++++++++++++++
 drivers/scsi/hpsa.h   |    1 +
 3 files changed, 18 insertions(+)


--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Comments

Tomas Henzl Oct. 26, 2015, 2:55 p.m. UTC | #1 
On 24.10.2015 21:52, Don Brace wrote:
> Fix a NULL pointer issue in the driver when devices are removed
> during a reset.
>
> Signed-off-by: Don Brace <don.brace@pmcs.com>
> ---
>  drivers/block/cciss.h |    1 +
>  drivers/scsi/hpsa.c   |   16 ++++++++++++++++
>  drivers/scsi/hpsa.h   |    1 +
>  3 files changed, 18 insertions(+)
>
> diff --git a/drivers/block/cciss.h b/drivers/block/cciss.h
> index 7fda30e..036fb0f 100644
> --- a/drivers/block/cciss.h
> +++ b/drivers/block/cciss.h
> @@ -155,6 +155,7 @@ struct ctlr_info
>  	size_t reply_pool_size;
>  	unsigned char reply_pool_wraparound;
>  	u32 *blockFetchTable;
> +	atomic_t reset_in_progress;
>  };
>  
>  /*  Defining the diffent access_methods
> diff --git a/drivers/scsi/hpsa.c b/drivers/scsi/hpsa.c
> index 3520d75..714bb76 100644
> --- a/drivers/scsi/hpsa.c
> +++ b/drivers/scsi/hpsa.c
> @@ -1656,6 +1656,11 @@ static void adjust_hpsa_scsi_table(struct ctlr_info *h, int hostno,
>  	int nadded, nremoved;
>  	struct Scsi_Host *sh = NULL;
>  
> +	if (atomic_read(&h->reset_in_progress)) {
> +		h->drv_req_rescan = 1;
> +		return;
> +	}

Is it safe, what happens when a reset is invoked at this point?
Is it possible to somehow reuse the scan_lock or scan_finished
and change the logic and wait with the reset?
(maybe break and reschedule the scan)

Btw. why is an atomic variable needed here?

--tm


> +
>  	added = kzalloc(sizeof(*added) * HPSA_MAX_DEVICES, GFP_KERNEL);
>  	removed = kzalloc(sizeof(*removed) * HPSA_MAX_DEVICES, GFP_KERNEL);
>  
> @@ -1764,8 +1769,14 @@ static void adjust_hpsa_scsi_table(struct ctlr_info *h, int hostno,
>  		goto free_and_out;
>  
>  	sh = h->scsi_host;
> +	if (sh == NULL) {
> +		dev_warn(&h->pdev->dev, "%s: scsi_host is null\n", __func__);
> +		return;
> +	}
>  	/* Notify scsi mid layer of any removed devices */
>  	for (i = 0; i < nremoved; i++) {
> +		if (!removed[i])
> +			continue;
>  		if (removed[i]->expose_state & HPSA_SCSI_ADD) {
>  			struct scsi_device *sdev =
>  				scsi_device_lookup(sh, removed[i]->bus,
> @@ -1790,6 +1801,8 @@ static void adjust_hpsa_scsi_table(struct ctlr_info *h, int hostno,
>  
>  	/* Notify scsi mid layer of any added devices */
>  	for (i = 0; i < nadded; i++) {
> +		if (!added[i])
> +			continue;
>  		if (!(added[i]->expose_state & HPSA_SCSI_ADD))
>  			continue;
>  		if (scsi_add_device(sh, added[i]->bus,
> @@ -5227,12 +5240,15 @@ static int hpsa_eh_device_reset_handler(struct scsi_cmnd *scsicmd)
>  
>  	hpsa_show_dev_msg(__stringify(KERN_WARNING), h, dev, "resetting");
>  
> +	atomic_set(&h->reset_in_progress, 1);
> +
>  	/* send a reset to the SCSI LUN which the command was sent to */
>  	rc = hpsa_do_reset(h, dev, dev->scsi3addr, HPSA_RESET_TYPE_LUN,
>  			   DEFAULT_REPLY_QUEUE);
>  	snprintf(msg, sizeof(msg), "reset %s",
>  		 rc == 0 ? "completed successfully" : "failed");
>  	hpsa_show_dev_msg(__stringify(KERN_WARNING), h, dev, msg);
> +	atomic_set(&h->reset_in_progress, 0);
>  	return rc == 0 ? SUCCESS : FAILED;
>  }
>  
> diff --git a/drivers/scsi/hpsa.h b/drivers/scsi/hpsa.h
> index 642c8ce..c7c1697 100644
> --- a/drivers/scsi/hpsa.h
> +++ b/drivers/scsi/hpsa.h
> @@ -271,6 +271,7 @@ struct ctlr_info {
>  	wait_queue_head_t abort_cmd_wait_queue;
>  	wait_queue_head_t event_sync_wait_queue;
>  	struct mutex reset_mutex;
> +	atomic_t reset_in_progress;
>  };
>  
>  struct offline_device_entry {
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Don Brace Oct. 26, 2015, 3:22 p.m. UTC | #2 
On 10/26/2015 09:55 AM, Tomas Henzl wrote:
> On 24.10.2015 21:52, Don Brace wrote:
>> Fix a NULL pointer issue in the driver when devices are removed
>> during a reset.
>>
>> Signed-off-by: Don Brace <don.brace@pmcs.com>
>> ---
>>   drivers/block/cciss.h |    1 +
>>   drivers/scsi/hpsa.c   |   16 ++++++++++++++++
>>   drivers/scsi/hpsa.h   |    1 +
>>   3 files changed, 18 insertions(+)
>>
>> diff --git a/drivers/block/cciss.h b/drivers/block/cciss.h
>> index 7fda30e..036fb0f 100644
>> --- a/drivers/block/cciss.h
>> +++ b/drivers/block/cciss.h
>> @@ -155,6 +155,7 @@ struct ctlr_info
>>   	size_t reply_pool_size;
>>   	unsigned char reply_pool_wraparound;
>>   	u32 *blockFetchTable;
>> +	atomic_t reset_in_progress;
>>   };
>>   
>>   /*  Defining the diffent access_methods
>> diff --git a/drivers/scsi/hpsa.c b/drivers/scsi/hpsa.c
>> index 3520d75..714bb76 100644
>> --- a/drivers/scsi/hpsa.c
>> +++ b/drivers/scsi/hpsa.c
>> @@ -1656,6 +1656,11 @@ static void adjust_hpsa_scsi_table(struct ctlr_info *h, int hostno,
>>   	int nadded, nremoved;
>>   	struct Scsi_Host *sh = NULL;
>>   
>> +	if (atomic_read(&h->reset_in_progress)) {
>> +		h->drv_req_rescan = 1;
>> +		return;
>> +	}
> Is it safe, what happens when a reset is invoked at this point?
> Is it possible to somehow reuse the scan_lock or scan_finished
> and change the logic and wait with the reset?
> (maybe break and reschedule the scan)
I wanted to reschedule the scan to see what happened after
a reset.

There also seems to be some raciness between the reset
handler and the re-scan procedure.

> Btw. why is an atomic variable needed here?
Overkill. I can change reset_in_progress to an int.

> --tm
>
>
>> +
>>   	added = kzalloc(sizeof(*added) * HPSA_MAX_DEVICES, GFP_KERNEL);
>>   	removed = kzalloc(sizeof(*removed) * HPSA_MAX_DEVICES, GFP_KERNEL);
>>   
>> @@ -1764,8 +1769,14 @@ static void adjust_hpsa_scsi_table(struct ctlr_info *h, int hostno,
>>   		goto free_and_out;
>>   
>>   	sh = h->scsi_host;
>> +	if (sh == NULL) {
>> +		dev_warn(&h->pdev->dev, "%s: scsi_host is null\n", __func__);
>> +		return;
>> +	}
>>   	/* Notify scsi mid layer of any removed devices */
>>   	for (i = 0; i < nremoved; i++) {
>> +		if (!removed[i])
>> +			continue;
>>   		if (removed[i]->expose_state & HPSA_SCSI_ADD) {
>>   			struct scsi_device *sdev =
>>   				scsi_device_lookup(sh, removed[i]->bus,
>> @@ -1790,6 +1801,8 @@ static void adjust_hpsa_scsi_table(struct ctlr_info *h, int hostno,
>>   
>>   	/* Notify scsi mid layer of any added devices */
>>   	for (i = 0; i < nadded; i++) {
>> +		if (!added[i])
>> +			continue;
>>   		if (!(added[i]->expose_state & HPSA_SCSI_ADD))
>>   			continue;
>>   		if (scsi_add_device(sh, added[i]->bus,
>> @@ -5227,12 +5240,15 @@ static int hpsa_eh_device_reset_handler(struct scsi_cmnd *scsicmd)
>>   
>>   	hpsa_show_dev_msg(__stringify(KERN_WARNING), h, dev, "resetting");
>>   
>> +	atomic_set(&h->reset_in_progress, 1);
>> +
>>   	/* send a reset to the SCSI LUN which the command was sent to */
>>   	rc = hpsa_do_reset(h, dev, dev->scsi3addr, HPSA_RESET_TYPE_LUN,
>>   			   DEFAULT_REPLY_QUEUE);
>>   	snprintf(msg, sizeof(msg), "reset %s",
>>   		 rc == 0 ? "completed successfully" : "failed");
>>   	hpsa_show_dev_msg(__stringify(KERN_WARNING), h, dev, msg);
>> +	atomic_set(&h->reset_in_progress, 0);
>>   	return rc == 0 ? SUCCESS : FAILED;
>>   }
>>   
>> diff --git a/drivers/scsi/hpsa.h b/drivers/scsi/hpsa.h
>> index 642c8ce..c7c1697 100644
>> --- a/drivers/scsi/hpsa.h
>> +++ b/drivers/scsi/hpsa.h
>> @@ -271,6 +271,7 @@ struct ctlr_info {
>>   	wait_queue_head_t abort_cmd_wait_queue;
>>   	wait_queue_head_t event_sync_wait_queue;
>>   	struct mutex reset_mutex;
>> +	atomic_t reset_in_progress;
>>   };
>>   
>>   struct offline_device_entry {
>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> --
> To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
diff mbox

Patch

diff --git a/drivers/block/cciss.h b/drivers/block/cciss.h
index 7fda30e..036fb0f 100644
--- a/drivers/block/cciss.h
+++ b/drivers/block/cciss.h
@@ -155,6 +155,7 @@  struct ctlr_info
 	size_t reply_pool_size;
 	unsigned char reply_pool_wraparound;
 	u32 *blockFetchTable;
+	atomic_t reset_in_progress;
 };
 
 /*  Defining the diffent access_methods
diff --git a/drivers/scsi/hpsa.c b/drivers/scsi/hpsa.c
index 3520d75..714bb76 100644
--- a/drivers/scsi/hpsa.c
+++ b/drivers/scsi/hpsa.c
@@ -1656,6 +1656,11 @@  static void adjust_hpsa_scsi_table(struct ctlr_info *h, int hostno,
 	int nadded, nremoved;
 	struct Scsi_Host *sh = NULL;
 
+	if (atomic_read(&h->reset_in_progress)) {
+		h->drv_req_rescan = 1;
+		return;
+	}
+
 	added = kzalloc(sizeof(*added) * HPSA_MAX_DEVICES, GFP_KERNEL);
 	removed = kzalloc(sizeof(*removed) * HPSA_MAX_DEVICES, GFP_KERNEL);
 
@@ -1764,8 +1769,14 @@  static void adjust_hpsa_scsi_table(struct ctlr_info *h, int hostno,
 		goto free_and_out;
 
 	sh = h->scsi_host;
+	if (sh == NULL) {
+		dev_warn(&h->pdev->dev, "%s: scsi_host is null\n", __func__);
+		return;
+	}
 	/* Notify scsi mid layer of any removed devices */
 	for (i = 0; i < nremoved; i++) {
+		if (!removed[i])
+			continue;
 		if (removed[i]->expose_state & HPSA_SCSI_ADD) {
 			struct scsi_device *sdev =
 				scsi_device_lookup(sh, removed[i]->bus,
@@ -1790,6 +1801,8 @@  static void adjust_hpsa_scsi_table(struct ctlr_info *h, int hostno,
 
 	/* Notify scsi mid layer of any added devices */
 	for (i = 0; i < nadded; i++) {
+		if (!added[i])
+			continue;
 		if (!(added[i]->expose_state & HPSA_SCSI_ADD))
 			continue;
 		if (scsi_add_device(sh, added[i]->bus,
@@ -5227,12 +5240,15 @@  static int hpsa_eh_device_reset_handler(struct scsi_cmnd *scsicmd)
 
 	hpsa_show_dev_msg(__stringify(KERN_WARNING), h, dev, "resetting");
 
+	atomic_set(&h->reset_in_progress, 1);
+
 	/* send a reset to the SCSI LUN which the command was sent to */
 	rc = hpsa_do_reset(h, dev, dev->scsi3addr, HPSA_RESET_TYPE_LUN,
 			   DEFAULT_REPLY_QUEUE);
 	snprintf(msg, sizeof(msg), "reset %s",
 		 rc == 0 ? "completed successfully" : "failed");
 	hpsa_show_dev_msg(__stringify(KERN_WARNING), h, dev, msg);
+	atomic_set(&h->reset_in_progress, 0);
 	return rc == 0 ? SUCCESS : FAILED;
 }
 
diff --git a/drivers/scsi/hpsa.h b/drivers/scsi/hpsa.h
index 642c8ce..c7c1697 100644
--- a/drivers/scsi/hpsa.h
+++ b/drivers/scsi/hpsa.h
@@ -271,6 +271,7 @@  struct ctlr_info {
 	wait_queue_head_t abort_cmd_wait_queue;
 	wait_queue_head_t event_sync_wait_queue;
 	struct mutex reset_mutex;
+	atomic_t reset_in_progress;
 };
 
 struct offline_device_entry {