[dm-devel] [PATCH] scsi_dh_rdac: Add dynamic match functionality to rdac handler

[Moger, Babu]

Chandra/Mike, Thanks for your comments.

> -----Original Message-----
> From: Chandra Seetharaman [mailto:sekharan at us.ibm.com]
> Sent: Friday, April 20, 2012 1:06 PM
> To: Mike Christie
> Cc: Moger, Babu; linux-scsi at vger.kernel.org; device-mapper development;
> Krishnasamy, Somasundaram
> Subject: Re: [PATCH] scsi_dh_rdac: Add dynamic match functionality to rdac
> handler
> 
> On Fri, 2012-04-20 at 11:48 -0500, Mike Christie wrote:
> > On 04/20/2012 11:27 AM, Chandra Seetharaman wrote:
> > > Babu,
> > >
> > > So, in effect, any SCSI devices with INQ C8 page with "edid" would be
> > > treated as a rdac device, right 
> > > Is that enough to identify _any_ SCSI device as a rdac device (even in
> > > future) ?

These checks are only to pass the match function. Actual checks in attach function.
So, it should not cause any issues.

> > >
> > > Also, instead of repeating the code, can you use submit_inquiry() ?
> > >
> >
> > Yeah you are right, I think we should modify submit_inquiry to take in
> > the sense and data buffers instead of taking in the rdac_dh_data struct.
> > At the time the match function is run rdac_dh_data might not be allocated.
> 
> oh yeah,
> 
> I did not not realize that...
> >

Changing the submit_inquiry involves some more changes. Will try to see if
I can make those changes in my next set of patches.

Actually the bigger problem here is match is called with spin_lock held.
We are making the blocking calls  with spin lock held. Thanks to Hannes  for
pointing that out. Will replace the spin_lock with mutex lock.