Re: cciss_scsi.c: Fix me

From: scameron
Date: Wed Jul 23 2014 - 10:43:23 EST


On Wed, Jul 23, 2014 at 02:15:29PM +0000, Benesh, Scott wrote:
> From: Nick Krause [mailto:xerofoify@xxxxxxxxx]
> Sent: Saturday, July 19, 2014 11:51 PM
> To: mike.miller@xxxxxx
> Cc: ISS StorageDev; linux-kernel@xxxxxxxxxxxxxxx
> Subject: cciss_scsi.c: Fix me
>
> Hey Mike,
> I seem to be hitting a fix me message in this file in function,cciss_scsi_queue_command_lck.
> I am wondering what you want to do when C is Null?
> Cheers Nick

Hi Nick,

Mike's moved on from HP to Canonical now.

It looks like you're running out of commands for tape drives,
which shouldn't ever happen, since we set

sh->can_queue = cciss_tape_cmds;

and we allocate that many commands + 2....

scsi_cmd_stack_setup(ctlr_info_t *h, struct cciss_scsi_adapter_data_t *sa)
{
int i;
struct cciss_scsi_cmd_stack_t *stk;
size_t size;

stk = &sa->cmd_stack;
stk->nelems = cciss_tape_cmds + 2;

You're apparently hitting this:

spin_lock_irqsave(&h->lock, flags);
c = scsi_cmd_alloc(h);
spin_unlock_irqrestore(&h->lock, flags);
if (c == NULL) { /* trouble... */
dev_warn(&h->pdev->dev, "scsi_cmd_alloc returned NULL!\n");
/* FIXME: next 3 lines are -> BAD! <- */
cmd->result = DID_NO_CONNECT << 16;
done(cmd);
return 0;
}

which means that scsi_cmd_alloc returned NULL, which only happens
if the thing has run out of commands.

It's not obvious to me how it can be that it runs out of commands.
Maybe we're losing them somehow, but this has not previously been
a problem that I'm aware of.

Are you able to reproduce the problem?

What's going on on the system when it happens?

putting in a dump_stack(); near that FIXME might give a clue.

Which kernel are you running?

-- steve


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/