From: David Milburn <dmilburn@redhat.com>
Date: Mon, 6 Oct 2008 12:00:51 -0500
Subject: [ata] libata: ata_piix sata/ide combined mode fix
Message-id: 20081006170051.GA15138@dhcp-210.hsv.redhat.com
O-Subject: [RHEL5.3 PATCH] libata: ata_piix sata/ide combined mode fix
Bugzilla: 463716
RH-Acked-by: Alan Cox <alan@redhat.com>

Intel has reported a crash during initialization of the ata_piix
driver after changing the BIOS setting for the sata controller
from ahci to ide mode. I was able to reproduce this problem, the
ata_piix driver is recognizing that a port is busy since the
ide driver is controlling the port. And RHEL5 sets the ata port
operations for this port to a set of dummy ops in ata_init_legacy_host.
But, during piix_init_one in between ata_pci_sff_prepare_host and
ata_pci_sff_activate_host the ata_piix driver is resetting the
ops for this port in piix_init_sidpr eventually causing a crash
in ata_sff_check_status since the registers have not been mapped
in for the busy port. This patch is specific to RHEL5 in order to
support combined mode where one port maybe driver by libata and
another port driven by drivers/ide. This patch has been verified
on -117.el5 internally and by Intel. This resolves BZ 463716
and an exception has been requested. Please review and ack.

Thanks,
David

diff --git a/drivers/ata/ata_piix.c b/drivers/ata/ata_piix.c
index f8ce185..5a8e371 100644
--- a/drivers/ata/ata_piix.c
+++ b/drivers/ata/ata_piix.c
@@ -1368,6 +1368,10 @@ static void __devinit piix_init_sidpr(struct ata_host *host)
 		if (hpriv->map[i] == IDE)
 			return;
 
+	if (ata_port_is_dummy(host->ports[0]) ||
+	    ata_port_is_dummy(host->ports[1]))
+		return;
+
 	if (!(host->ports[0]->flags & PIIX_FLAG_SIDPR))
 		return;