Re: [dm-devel] scsi_dh_rdac: Add empty set_params function to scsi_dh_rdac

[Mike Snitzer <snitzer redhat com>]

On Wed, Aug 08 2012 at 12:10pm -0400,
Moger, Babu <Babu Moger netapp com> wrote:

> This patch adds empty set_params function to scsi_dh_rdac.
> 
> This patch is required for the following features to work properly.
> 1. add  retain_attached_hw_handler feature
> 
> http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commit;h=a58a935d5a1b2ad267017a68c3a1bca26226cc76
> 
> 2. add scsi_dh_attached_handler_name
> 
> http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commit;h=7e8a74b177f17d100916b6ad415450f7c9508691
> 
> 
> DM layer detaches the handler if the set_params is not implemented or fails.
> 
> For example if we pass following parameters from multipath:
> 
> 	features "1 retain_attached_hw_handler" (consider rdac as attached_hw_handler)
> 	hardware_handler "2 alua 1"
> 
> If the attached_hw_handler is rdac, then DM anyway tries to call set_params on rdac. Because rdac does not 
> implement set_params, scsi_dh_set_params fails and DM detaches the handler.
> 
> This patch fixes this problem. 
> 
> Signed-off-by: Babu Moger <babu moger netapp com>

I think it'd be best to fix scsi_dh_set_params() so that it returns 0 if
scsi_dh->set_params is NULL.  That'd avoid all scsi_dh from having to
stub out a similar set_params to return 0.

But taking a step back, have you actually seen the problem you're saying
this patch fixes?

This problem shouldn't exist:
drivers/md/dm-mpath.c:parse_path will free m->hw_handler_params (and
reset it to NULL) if a scsi_dh is already attached.  Have a look at the
if (attached_handler_name) {} case.

Mike