From: mchristi@redhat.com <mchristi@redhat.com>
Date: Fri, 24 Jul 2009 13:39:14 -0500
Subject: [scsi] mptsas: fix max_id initialization
Message-id: 1248460754-7139-1-git-send-email-mchristi@redhat.com
O-Subject: [PATCH 1/1] RHEL5.4: mptsas: fix max_id initialization
Bugzilla: 455678
RH-Acked-by: Rob Evers <revers@redhat.com>
RH-Acked-by: David Miller <davem@redhat.com>
RH-Acked-by: Stefan Assmann <sassmann@redhat.com>

From: Mike Christie <mchristi@redhat.com>

This is for BZ 455678.

The mptsas driver is setting the max_id based on a firmware
limit. The problem is that for sas in linux, the target id
is just used as a OS id value. It does not track the number
of targets in the system and does not match the hardware
target id. The sas class just increases target id after
a target is created, so if you pull a cable (which for sas
deletes the target after X seconds) and plug it in again enough
times the target id will hit the shost->max_id and the target struct
will not get scanned.

This fixes the problem by having mptsas set the max_id to
-1 like is done for other sas drivers.

The patch is not upstream yet. I will be sending it later today.
I have not tested the patch. LSI did some quick tests on it, but
did not stress it heavily. We are still waiting for confirmation
from Intel that it fixes their specific problem in the BZ, but
we are not sure when they will be done testing. This patch
will only affect mptsas, LSI thinks it is safe and it fixes a bug
we know about in that code path, so I wanted to still send it.

diff --git a/drivers/message/fusion/mptsas.c b/drivers/message/fusion/mptsas.c
index ce0788e..239de23 100644
--- a/drivers/message/fusion/mptsas.c
+++ b/drivers/message/fusion/mptsas.c
@@ -3066,7 +3066,7 @@ mptsas_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 	/* set 16 byte cdb's */
 	sh->max_cmd_len = 16;
 
-	sh->max_id = ioc->pfacts[0].PortSCSIID;
+	sh->max_id = -1;
 	sh->max_lun = max_lun;
 
 	sh->transportt = mptsas_transport_template;