[PATCH 0/2] 'default' hardware handler for multipath

[PATCH 0/2] 'default' hardware handler for multipath

[Hannes Reinecke]

This patchset introduces a 'default' hardware handler for
dm-multipath.
Modern storage arrays typically support two failover methods,
the original proprietary and the modern ALUA-based one.
The device_handler implementation will currently select the
ALUA handler, and falling back to the proprietary one if
ALUA isn't supported.
However, in the built-in hardware table for multipath one
can specify only one hardware handler, causing the original
hardware handler to be overwritten.
By specifying a 'default' hardware handler multipath will not
try to attach a specific hardware handler, but rather using
the currently attached on.

Hannes Reinecke (2):
  scsi_dh: Allow NULL hardware handler name in scsi_dh_attach()
  dm-mpath: Allow 'default' hardware handler

 drivers/md/dm-mpath.c                 |    8 ++++++--
 drivers/scsi/device_handler/scsi_dh.c |    8 ++++++--
 2 files changed, 12 insertions(+), 4 deletions(-)

-- 
1.7.7

[PATCH 1/2] scsi_dh: Allow NULL hardware handler name in scsi_dh_attach()

[Hannes Reinecke]

This patch allows to pass in a NULL hardware handler to
scsi_dh_attach(), causing the reference count of the existing
hardware handler to be increased.
An error will be returned if no hardware handler is attached.

Signed-off-by: Hannes Reinecke <hare@suse.de>
---
 drivers/scsi/device_handler/scsi_dh.c |    8 ++++++--
 1 files changed, 6 insertions(+), 2 deletions(-)

diff --git a/drivers/scsi/device_handler/scsi_dh.c b/drivers/scsi/device_handler/scsi_dh.c
index 48e46f5..77fa67e 100644
--- a/drivers/scsi/device_handler/scsi_dh.c
+++ b/drivers/scsi/device_handler/scsi_dh.c
@@ -475,10 +475,14 @@ int scsi_dh_attach(struct request_queue *q, const char *name)
 {
 	unsigned long flags;
 	struct scsi_device *sdev;
-	struct scsi_device_handler *scsi_dh;
+	struct scsi_device_handler *scsi_dh = NULL;
 	int err = 0;
 
-	scsi_dh = get_device_handler(name);
+	if (name) {
+		scsi_dh = get_device_handler(name);
+	} else if (sdev->scsi_dh_data) {
+		scsi_dh = sdev->scsi_dh_data->scsi_dh;
+	}
 	if (!scsi_dh)
 		return -EINVAL;
 
-- 
1.7.7

[PATCH 2/2] dm-mpath: Allow 'default' hardware handler

[Hannes Reinecke]

This patch introduces a 'default' hardware handler for dm-mpath.
When specifying '1 default' dm-multipath will be using the currently
attached hardware handler.

Signed-off-by: Hannes Reinecke <hare@suse.de>
---
 drivers/md/dm-mpath.c |    8 ++++++--
 1 files changed, 6 insertions(+), 2 deletions(-)

diff --git a/drivers/md/dm-mpath.c b/drivers/md/dm-mpath.c
index 922a338..bbf950f 100644
--- a/drivers/md/dm-mpath.c
+++ b/drivers/md/dm-mpath.c
@@ -586,15 +586,19 @@ static struct pgpath *parse_path(struct dm_arg_set *as, struct path_selector *ps
 
 	if (m->hw_handler_name) {
 		struct request_queue *q = bdev_get_queue(p->path.dev->bdev);
+		char *hw_handler = NULL;
 
-		r = scsi_dh_attach(q, m->hw_handler_name);
+		if (strncmp(m->hw_handler_name, "default", 6))
+			hw_handler = m->hw_handler_name;
+		
+		r = scsi_dh_attach(q, hw_handler);
 		if (r == -EBUSY) {
 			/*
 			 * Already attached to different hw_handler,
 			 * try to reattach with correct one.
 			 */
 			scsi_dh_detach(q);
-			r = scsi_dh_attach(q, m->hw_handler_name);
+			r = scsi_dh_attach(q, hw_handler);
 		}
 
 		if (r < 0) {
-- 
1.7.7

Re: [PATCH 2/2] dm-mpath: Allow 'default' hardware handler

[Mike Snitzer]

On Mon, Apr 02 2012 at 12:43pm -0400,
Hannes Reinecke <hare@suse.de> wrote:

> This patch introduces a 'default' hardware handler for dm-mpath.
> When specifying '1 default' dm-multipath will be using the currently
> attached hardware handler.
> 
> Signed-off-by: Hannes Reinecke <hare@suse.de>
> ---
>  drivers/md/dm-mpath.c |    8 ++++++--
>  1 files changed, 6 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/md/dm-mpath.c b/drivers/md/dm-mpath.c
> index 922a338..bbf950f 100644
> --- a/drivers/md/dm-mpath.c
> +++ b/drivers/md/dm-mpath.c
> @@ -586,15 +586,19 @@ static struct pgpath *parse_path(struct dm_arg_set *as, struct path_selector *ps
>  
>  	if (m->hw_handler_name) {
>  		struct request_queue *q = bdev_get_queue(p->path.dev->bdev);
> +		char *hw_handler = NULL;
>  
> -		r = scsi_dh_attach(q, m->hw_handler_name);
> +		if (strncmp(m->hw_handler_name, "default", 6))
> +			hw_handler = m->hw_handler_name;

Should be: if (strncmp(m->hw_handler_name, "default", 7))

Re: [PATCH 1/2] scsi_dh: Allow NULL hardware handler name in scsi_dh_attach()

[Mike Snitzer]

On Mon, Apr 02 2012 at 12:43pm -0400,
Hannes Reinecke <hare@suse.de> wrote:

> This patch allows to pass in a NULL hardware handler to
> scsi_dh_attach(), causing the reference count of the existing
> hardware handler to be increased.
> An error will be returned if no hardware handler is attached.
> 
> Signed-off-by: Hannes Reinecke <hare@suse.de>
> ---
>  drivers/scsi/device_handler/scsi_dh.c |    8 ++++++--
>  1 files changed, 6 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/scsi/device_handler/scsi_dh.c b/drivers/scsi/device_handler/scsi_dh.c
> index 48e46f5..77fa67e 100644
> --- a/drivers/scsi/device_handler/scsi_dh.c
> +++ b/drivers/scsi/device_handler/scsi_dh.c
> @@ -475,10 +475,14 @@ int scsi_dh_attach(struct request_queue *q, const char *name)
>  {
>  	unsigned long flags;
>  	struct scsi_device *sdev;
> -	struct scsi_device_handler *scsi_dh;
> +	struct scsi_device_handler *scsi_dh = NULL;
>  	int err = 0;
>  
> -	scsi_dh = get_device_handler(name);
> +	if (name) {
> +		scsi_dh = get_device_handler(name);
> +	} else if (sdev->scsi_dh_data) {
> +		scsi_dh = sdev->scsi_dh_data->scsi_dh;
> +	}

Extra brackets aren't needed.  But otherwise:

Acked-by: Mike Snitzer <snitzer@redhat.com>

Re: [PATCH 2/2] dm-mpath: Allow 'default' hardware handler

[Hannes Reinecke]

On 04/02/2012 07:04 PM, Mike Snitzer wrote:
> On Mon, Apr 02 2012 at 12:43pm -0400,
> Hannes Reinecke <hare@suse.de> wrote:
> 
>> This patch introduces a 'default' hardware handler for dm-mpath.
>> When specifying '1 default' dm-multipath will be using the currently
>> attached hardware handler.
>>
>> Signed-off-by: Hannes Reinecke <hare@suse.de>
>> ---
>>  drivers/md/dm-mpath.c |    8 ++++++--
>>  1 files changed, 6 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/md/dm-mpath.c b/drivers/md/dm-mpath.c
>> index 922a338..bbf950f 100644
>> --- a/drivers/md/dm-mpath.c
>> +++ b/drivers/md/dm-mpath.c
>> @@ -586,15 +586,19 @@ static struct pgpath *parse_path(struct dm_arg_set *as, struct path_selector *ps
>>  
>>  	if (m->hw_handler_name) {
>>  		struct request_queue *q = bdev_get_queue(p->path.dev->bdev);
>> +		char *hw_handler = NULL;
>>  
>> -		r = scsi_dh_attach(q, m->hw_handler_name);
>> +		if (strncmp(m->hw_handler_name, "default", 6))
>> +			hw_handler = m->hw_handler_name;
> 
> Should be: if (strncmp(m->hw_handler_name, "default", 7))
A-hem.

Counting by hand is at times tricky.
I'll be sending an updated patchset.

Cheers,

Hannes
-- 
Dr. Hannes Reinecke		      zSeries & Storage
hare@suse.de			      +49 911 74053 688
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: J. Hawn, J. Guild, F. Imendörffer, HRB 16746 (AG Nürnberg)
--
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

Re: [dm-devel] [PATCH 2/2] dm-mpath: Allow 'default' hardware handler

[Jun'ichi Nomura]

Hi Hannes,

On 04/03/12 02:07, Hannes Reinecke wrote:
> On 04/02/2012 07:04 PM, Mike Snitzer wrote:
>> On Mon, Apr 02 2012 at 12:43pm -0400,
>> Hannes Reinecke <hare@suse.de> wrote:
>>
>>> This patch introduces a 'default' hardware handler for dm-mpath.
>>> When specifying '1 default' dm-multipath will be using the currently
>>> attached hardware handler.
>>>
>>> Signed-off-by: Hannes Reinecke <hare@suse.de>
>>> ---
>>>  drivers/md/dm-mpath.c |    8 ++++++--
>>>  1 files changed, 6 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/drivers/md/dm-mpath.c b/drivers/md/dm-mpath.c
>>> index 922a338..bbf950f 100644
>>> --- a/drivers/md/dm-mpath.c
>>> +++ b/drivers/md/dm-mpath.c
>>> @@ -586,15 +586,19 @@ static struct pgpath *parse_path(struct dm_arg_set *as, struct path_selector *ps
>>>  
>>>  	if (m->hw_handler_name) {
>>>  		struct request_queue *q = bdev_get_queue(p->path.dev->bdev);
>>> +		char *hw_handler = NULL;
>>>  
>>> -		r = scsi_dh_attach(q, m->hw_handler_name);
>>> +		if (strncmp(m->hw_handler_name, "default", 6))
>>> +			hw_handler = m->hw_handler_name;
>>
>> Should be: if (strncmp(m->hw_handler_name, "default", 7))
> A-hem.
> 
> Counting by hand is at times tricky.
> I'll be sending an updated patchset.

It seems you will hit the following error in parse_hw_handler().
Or is it avoided somehow?

        request_module("scsi_dh_%s", m->hw_handler_name);
        if (scsi_dh_handler_exist(m->hw_handler_name) == 0) {
                ti->error = "unknown hardware handler type";
                ret = -EINVAL;
                goto fail;
        }

Thanks,
-- 
Jun'ichi Nomura, NEC Corporation

Re: [dm-devel] [PATCH 1/2] scsi_dh: Allow NULL hardware handler name in scsi_dh_attach()

[Jun'ichi Nomura]

Hi Hannes,

On 04/03/12 01:43, Hannes Reinecke wrote:
> This patch allows to pass in a NULL hardware handler to
> scsi_dh_attach(), causing the reference count of the existing
> hardware handler to be increased.
> An error will be returned if no hardware handler is attached.
> 
> Signed-off-by: Hannes Reinecke <hare@suse.de>
> ---
>  drivers/scsi/device_handler/scsi_dh.c |    8 ++++++--
>  1 files changed, 6 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/scsi/device_handler/scsi_dh.c b/drivers/scsi/device_handler/scsi_dh.c
> index 48e46f5..77fa67e 100644
> --- a/drivers/scsi/device_handler/scsi_dh.c
> +++ b/drivers/scsi/device_handler/scsi_dh.c
> @@ -475,10 +475,14 @@ int scsi_dh_attach(struct request_queue *q, const char *name)
>  {
>  	unsigned long flags;
>  	struct scsi_device *sdev;
> -	struct scsi_device_handler *scsi_dh;
> +	struct scsi_device_handler *scsi_dh = NULL;
>  	int err = 0;
>  
> -	scsi_dh = get_device_handler(name);
> +	if (name) {
> +		scsi_dh = get_device_handler(name);
> +	} else if (sdev->scsi_dh_data) {
> +		scsi_dh = sdev->scsi_dh_data->scsi_dh;
> +	}
>  	if (!scsi_dh)
>  		return -EINVAL;

It won't work. sdev is not initialized...

Thanks,
-- 
Jun'ichi Nomura, NEC Corporation

RE: [dm-devel] [PATCH 0/2] 'default' hardware handler for multipath

[Moger, Babu]

Thanks Hannes. We appreciate your work on this.

> -----Original Message-----
> From: dm-devel-bounces@redhat.com [mailto:dm-devel-
> bounces@redhat.com] On Behalf Of Hannes Reinecke
> Sent: Monday, April 02, 2012 11:44 AM
> To: linux-scsi@vger.kernel.org
> Cc: dm-devel@redhat.com; Mike Snitzer
> Subject: [dm-devel] [PATCH 0/2] 'default' hardware handler for multipath
> 
> This patchset introduces a 'default' hardware handler for dm-multipath.
> Modern storage arrays typically support two failover methods, the original
> proprietary and the modern ALUA-based one.
> The device_handler implementation will currently select the ALUA handler,
> and falling back to the proprietary one if ALUA isn't supported.
> However, in the built-in hardware table for multipath one can specify only
> one hardware handler, causing the original hardware handler to be
> overwritten.
> By specifying a 'default' hardware handler multipath will not try to attach a
> specific hardware handler, but rather using the currently attached on.

I think we still have some issues here. Right now we load the driver
either by adding it in initrd or by using request_module call from device mapper.
If the user passes, hardware_handler   "1 default"  from multipath then request_module will fail. 
How are we going to load the driver if these handlers are not loaded. 

> 
> Hannes Reinecke (2):
>   scsi_dh: Allow NULL hardware handler name in scsi_dh_attach()
>   dm-mpath: Allow 'default' hardware handler
> 
>  drivers/md/dm-mpath.c                 |    8 ++++++--
>  drivers/scsi/device_handler/scsi_dh.c |    8 ++++++--
>  2 files changed, 12 insertions(+), 4 deletions(-)
> 
> --
> 1.7.7
> 
> --
> dm-devel mailing list
> dm-devel@redhat.com
> https://www.redhat.com/mailman/listinfo/dm-devel

Re: [dm-devel] [PATCH 0/2] 'default' hardware handler for multipath

[Hannes Reinecke]

On 04/03/2012 11:12 PM, Moger, Babu wrote:
> Thanks Hannes. We appreciate your work on this.
> 
>> -----Original Message-----
>> From: dm-devel-bounces@redhat.com [mailto:dm-devel-
>> bounces@redhat.com] On Behalf Of Hannes Reinecke
>> Sent: Monday, April 02, 2012 11:44 AM
>> To: linux-scsi@vger.kernel.org
>> Cc: dm-devel@redhat.com; Mike Snitzer
>> Subject: [dm-devel] [PATCH 0/2] 'default' hardware handler for multipath
>>
>> This patchset introduces a 'default' hardware handler for dm-multipath.
>> Modern storage arrays typically support two failover methods, the original
>> proprietary and the modern ALUA-based one.
>> The device_handler implementation will currently select the ALUA handler,
>> and falling back to the proprietary one if ALUA isn't supported.
>> However, in the built-in hardware table for multipath one can specify only
>> one hardware handler, causing the original hardware handler to be
>> overwritten.
>> By specifying a 'default' hardware handler multipath will not try to attach a
>> specific hardware handler, but rather using the currently attached on.
> 
> I think we still have some issues here. Right now we load the driver
> either by adding it in initrd or by using request_module call from device mapper.
> If the user passes, hardware_handler   "1 default"  from multipath then request_module will fail. 
> How are we going to load the driver if these handlers are not loaded. 
> 
The idea of the 'default' hardware handler is to avoid multipath
overriding the kernel matchine algorithm. To make this work we
obviously have to load the modules prior to start multipathing.

Given that we (ie RH and us) are already loading the device-handler
modules from the initrd independent from multipathing we should be fine.

But thinking a bit more about it, it might be better to handle this
via features.
Implementing a feature like 'default_hw_handler' would not override
the default hardware handler from the hardware table.
So the requested hardware handler would still be loaded, but the
actual initialisation would be controlled via the feature.

Yep, that sound better.
I'll be preparing a patch.

Cheers,

Hannes
-- 
Dr. Hannes Reinecke		      zSeries & Storage
hare@suse.de			      +49 911 74053 688
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: J. Hawn, J. Guild, F. Imendörffer, HRB 16746 (AG Nürnberg)
--
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

RE: [dm-devel] [PATCH 0/2] 'default' hardware handler for multipath

[Moger, Babu]

Hannes,

> -----Original Message-----
> From: Hannes Reinecke [mailto:hare@suse.de]
> Sent: Wednesday, April 18, 2012 9:08 AM
> To: Moger, Babu
> Cc: device-mapper development; linux-scsi@vger.kernel.org; Mike Snitzer
> Subject: Re: [dm-devel] [PATCH 0/2] 'default' hardware handler for multipath
> 
> On 04/03/2012 11:12 PM, Moger, Babu wrote:
> > Thanks Hannes. We appreciate your work on this.
> >
> >> -----Original Message-----
> >> From: dm-devel-bounces@redhat.com [mailto:dm-devel-
> >> bounces@redhat.com] On Behalf Of Hannes Reinecke
> >> Sent: Monday, April 02, 2012 11:44 AM
> >> To: linux-scsi@vger.kernel.org
> >> Cc: dm-devel@redhat.com; Mike Snitzer
> >> Subject: [dm-devel] [PATCH 0/2] 'default' hardware handler for multipath
> >>
> >> This patchset introduces a 'default' hardware handler for dm-multipath.
> >> Modern storage arrays typically support two failover methods, the original
> >> proprietary and the modern ALUA-based one.
> >> The device_handler implementation will currently select the ALUA
> handler,
> >> and falling back to the proprietary one if ALUA isn't supported.
> >> However, in the built-in hardware table for multipath one can specify only
> >> one hardware handler, causing the original hardware handler to be
> >> overwritten.
> >> By specifying a 'default' hardware handler multipath will not try to attach a
> >> specific hardware handler, but rather using the currently attached on.
> >
> > I think we still have some issues here. Right now we load the driver
> > either by adding it in initrd or by using request_module call from device
> mapper.
> > If the user passes, hardware_handler   "1 default"  from multipath then
> request_module will fail.
> > How are we going to load the driver if these handlers are not loaded.
> >
> The idea of the 'default' hardware handler is to avoid multipath
> overriding the kernel matchine algorithm. To make this work we
> obviously have to load the modules prior to start multipathing.
> 
> Given that we (ie RH and us) are already loading the device-handler
> modules from the initrd independent from multipathing we should be fine.
> 
> But thinking a bit more about it, it might be better to handle this
> via features.
> Implementing a feature like 'default_hw_handler' would not override
> the default hardware handler from the hardware table.
> So the requested hardware handler would still be loaded, but the
> actual initialisation would be controlled via the feature.

hardware_handler is also used to pass on the arguments to handler. Like.
	hardware_handler        "2 emc 1"
We have to make sure that this feature does not break here. Considering all this 
don't you think it may be better if we make this decision in multipath tools itself.
Common factor here is alua. It is always between alua or some other handler.
Does it make sense to check TPGS in tools? 

> Yep, that sound better.
> I'll be preparing a patch.
> 
> Cheers,
> 
> Hannes
> --
> Dr. Hannes Reinecke		      zSeries & Storage
> hare@suse.de			      +49 911 74053 688
> SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg
> GF: J. Hawn, J. Guild, F. Imendörffer, HRB 16746 (AG Nürnberg)
--
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