Re: [PATCH 3/3] multipath: implement controller framework for hardware handlers (fwd)

Re: [PATCH 3/3] multipath: implement controller framework for hardware handlers (fwd)

[Chip Coldwell]

From: Hannes Reinecke <hare@suse.de>
diff --git a/drivers/md/dm-hw-handler.h b/drivers/md/dm-hw-handler.h
> index 9216682..6f91b55 100644
> --- a/drivers/md/dm-hw-handler.h
> +++ b/drivers/md/dm-hw-handler.h
> @@ -15,10 +15,31 @@
>  
>  struct hw_handler_type;
>  
> +#define CTLR_ID_LEN 256
> +
> +struct hw_controller {
> +	struct list_head node;
> +	struct hw_controller_type *type;
> +	unsigned char ctlr_id[CTLR_ID_LEN];
> +	struct kref kref;
> +	spinlock_t lock;
> +	struct list_head cmd_list;
> +	int submitted;
> +	void *context;
> +};
> +
> +struct hw_controller_type {
> +	char *name;
> +
> +	int (*create) (struct hw_controller *ctlr);
> +	void (*destroy) (struct hw_controller *ctlr);
> +};
> +

Would it make sense to make hw_controller_type a kobj_type?  That way,
the controller topology could be reflected in sysfs, where it can be
reached by tools such as lsscsi.

Chip

-- 
Charles M. "Chip" Coldwell
Senior Software Engineer
Red Hat, Inc
978-392-2426

Re: [PATCH 3/3] multipath: implement controller framework for hardware handlers (fwd)

[Hannes Reinecke]

Chip Coldwell wrote:
> From: Hannes Reinecke <hare@suse.de>
> diff --git a/drivers/md/dm-hw-handler.h b/drivers/md/dm-hw-handler.h
>> index 9216682..6f91b55 100644
>> --- a/drivers/md/dm-hw-handler.h
>> +++ b/drivers/md/dm-hw-handler.h
>> @@ -15,10 +15,31 @@
>>  
>>  struct hw_handler_type;
>>  
>> +#define CTLR_ID_LEN 256
>> +
>> +struct hw_controller {
>> +	struct list_head node;
>> +	struct hw_controller_type *type;
>> +	unsigned char ctlr_id[CTLR_ID_LEN];
>> +	struct kref kref;
>> +	spinlock_t lock;
>> +	struct list_head cmd_list;
>> +	int submitted;
>> +	void *context;
>> +};
>> +
>> +struct hw_controller_type {
>> +	char *name;
>> +
>> +	int (*create) (struct hw_controller *ctlr);
>> +	void (*destroy) (struct hw_controller *ctlr);
>> +};
>> +
> 
> Would it make sense to make hw_controller_type a kobj_type?  That way,
> the controller topology could be reflected in sysfs, where it can be
> reached by tools such as lsscsi.
> 
Eventually, maybe.

Plan is to add userland support to manually declare and assign
controller to each priority group.
But I haven't really sorted out the details there (how to pass
information etc) and I really would like to have the basic
infrastructure agreed upon first.
And implementing a kobj_type at this point would be a bit
over the top as we are in the very preliminary stages here.

Once the controller framework is accepted we can think about it.

Cheers,

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