Date: Thu, 24 May 2007 11:07:56 -0500
From: "Mike Miller (OS Dev)" <mikem@beardog.cca.cpqcorp.net>
To: linux-kernel@vger.kernel.org, tom.l.nguyen@intel.com
Cc: iss_storagedev@hp.com, akpm@linux-foundation.org, jens.axboe@oracle.com
Subject: msi_free_irqs #2
Message-ID: <20070524160756.GA14083@beardog.cca.cpqcorp.net>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
User-Agent: Mutt/1.5.9i
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1459
Lines: 29

So I guess I found the answer to my own question. msi_free_irqs was apparently added
in 2.6.22-something. I don't find it in 2.6.21.2 or anywhere else. So somebody broke a
couple of things.
The most noticable is cciss hangs after turning on interrupts. The reason for that is
the kernel now looks at my array of MSI-X vectors in reverse order. We have 4 ways of
generating an interrupt on Smart Array hw. They are:

#       define DOORBELL_INT     0
#       define PERF_MODE_INT    1
#       define SIMPLE_MODE_INT  2
#       define MEMQ_MODE_INT    3

For INTx these four lines are OR'ed together and run to one interrupt pin. MSI-X
breaks this hardware OR'ing so we must register either all 4 or at the least the
correct interrupt. When I first submitted the MSI/X support I was registering all 4.
Someone changed that to only register a single MSI-X vector. That worked fine until
2.6.22-something. 
Now it appears that the kernel is looking at the array in reverse order. IOW, I must
register PERF_MODE_INT in order for cciss to work. That's messed up. Anybody want to
`fess up to making these changes? :)
I'll keep working this, but I'm going to undo someones change when I figure out where
it's broke.

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