Re: cciss update [2/2] fixes for Steeleye Lifekeeper

[mikem <mikem@beardog.cca.cpqcorp.net>]

On Fri, Oct 15, 2004 at 10:05:09AM -0500, James Bottomley wrote:
> On Thu, 2004-10-14 at 13:39, Christoph Hellwig wrote:
> > Such a volume has been configured and set up, and although it's still
> > ugly I'd say it's okay.  But the patch also adds one gendisk per controller
> > even if no volume is set up.
> 
> That's this bit of code:
> 
> @@ -2762,7 +2810,9 @@ static int __devinit cciss_init_one(stru
>                 disk->fops = &cciss_fops;
>                 disk->queue = hba[i]->queue;
>                 disk->private_data = drv;
> -               if( !(drv->nr_blocks))
> +               /* we must register the controller even if no disks
> exist */
> +               /* this is for the online array utilities */
> +               if(!drv->heads && j)
>                         continue;
>                 blk_queue_hardsect_size(hba[i]->queue, drv->block_size);
>                 set_capacity(disk, drv->nr_blocks);
> 
> Mike, is there a way we can only allocate a gendisk when we know there's
> actually a device there (if owned by another controller currently)?
> 
> James

This patch only registers the controller if no logical drives are configured. It will not result in all possible logical drives being added. I added printk's to the driver to show me what I'm registering.
What I see is the controller registers every time, and only drives that are phsically configured are registered. That is true for reserved drives, also.

Thanks,
mikem