[PATCH 0/4] amd-iommu fixes for 2.6.31

[PATCH 0/4] amd-iommu fixes for 2.6.31

[Joerg Roedel]

This small patchset contain the fixes for the AMD IOMMU driver in the
Linux kernel I have queued up for 2.6.31. Please review.

diffstat:

 arch/x86/kernel/amd_iommu.c      |    6 +++++-
 arch/x86/kernel/amd_iommu_init.c |    6 ++++--
 2 files changed, 9 insertions(+), 3 deletions(-)

shortlog:

Chris Wright (1):
      amd iommu: properly detach from protection domain on ->remove

Joerg Roedel (1):
      amd-iommu: make sure only ivmd entries are parsed

Neil Turton (2):
      amd-iommu: fix an off-by-one error in the AMD IOMMU driver.
      amd-iommu: fix the handling of device aliases in the AMD IOMMU driver.

[PATCH 1/4] amd-iommu: fix an off-by-one error in the AMD IOMMU driver.

[Joerg Roedel]

From: Neil Turton <nturton@solarflare.com>

[ impact: bugfix ]

The variable amd_iommu_last_bdf holds the maximum bdf of any device
controlled by an IOMMU, so the number of device entries needed is
amd_iommu_last_bdf+1.  The function tbl_size used amd_iommu_last_bdf
instead.  This would be a problem if the last device were a large
enough power of 2.

Signed-off-by: Neil Turton <nturton@solarflare.com>
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu_init.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c
index 8c0be09..35fc965 100644
--- a/arch/x86/kernel/amd_iommu_init.c
+++ b/arch/x86/kernel/amd_iommu_init.c
@@ -175,7 +175,7 @@ static inline void update_last_devid(u16 devid)
 static inline unsigned long tbl_size(int entry_size)
 {
 	unsigned shift = PAGE_SHIFT +
-			 get_order(amd_iommu_last_bdf * entry_size);
+			 get_order(((int)amd_iommu_last_bdf + 1) * entry_size);
 
 	return 1UL << shift;
 }
-- 
1.6.3.1

[PATCH 2/4] amd-iommu: fix the handling of device aliases in the AMD IOMMU driver.

[Joerg Roedel]

From: Neil Turton <nturton@solarflare.com>

[ impact: set flags for the right req-id in the device table ]

The devid parameter to set_dev_entry_from_acpi is the requester ID
rather than the device ID since it is used to index the IOMMU device
table.  The handling of IVHD_DEV_ALIAS used to pass the device ID.
This patch fixes it to pass the requester ID.

Signed-off-by: Neil Turton <nturton@solarflare.com>
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu_init.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c
index 35fc965..53f93db 100644
--- a/arch/x86/kernel/amd_iommu_init.c
+++ b/arch/x86/kernel/amd_iommu_init.c
@@ -618,7 +618,7 @@ static void __init init_iommu_from_acpi(struct amd_iommu *iommu,
 		case IVHD_DEV_ALIAS:
 			devid = e->devid;
 			devid_to = e->ext >> 8;
-			set_dev_entry_from_acpi(iommu, devid, e->flags, 0);
+			set_dev_entry_from_acpi(iommu, devid_to, e->flags, 0);
 			amd_iommu_alias_table[devid] = devid_to;
 			break;
 		case IVHD_DEV_ALIAS_RANGE:
-- 
1.6.3.1

[PATCH 3/4] amd-iommu: make sure only ivmd entries are parsed

[Joerg Roedel]

[ impact: protect against broken ivrs acpi table ]

The bug never triggered. But it should be fixed to protect against
broken ACPI tables in the future.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu_init.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c
index 53f93db..a3a2b98 100644
--- a/arch/x86/kernel/amd_iommu_init.c
+++ b/arch/x86/kernel/amd_iommu_init.c
@@ -906,6 +906,8 @@ static int __init init_unity_map_range(struct ivmd_header *m)
 
 	switch (m->type) {
 	default:
+		kfree(e);
+		return 0;
 	case ACPI_IVMD_TYPE:
 		e->devid_start = e->devid_end = m->devid;
 		break;
-- 
1.6.3.1

[PATCH 4/4] amd iommu: properly detach from protection domain on ->remove

[Joerg Roedel]

From: Chris Wright <chrisw@sous-sol.org>

[ impact: bugfix ]

Some drivers may use the dma api during ->remove which will
cause a protection domain to get reattached to a device.  Delay the
detach until after the driver is completely unbound.

[ joro: added a little merge helper ]

Signed-off-by: Chris Wright <chrisw@redhat.com>
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu.c |    6 +++++-
 1 files changed, 5 insertions(+), 1 deletions(-)

diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index a97db99..d689883 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -57,6 +57,10 @@ static int dma_ops_unity_map(struct dma_ops_domain *dma_dom,
 static struct dma_ops_domain *find_protection_domain(u16 devid);
 
 
+#ifndef BUS_NOTIFY_UNBOUND_DRIVER
+#define BUS_NOTIFY_UNBOUND_DRIVER 0x0005
+#endif
+
 #ifdef CONFIG_AMD_IOMMU_STATS
 
 /*
@@ -1012,7 +1016,7 @@ static int device_change_notifier(struct notifier_block *nb,
 		printk(KERN_INFO "AMD IOMMU: Using protection domain %d for "
 		       "device %s\n", dma_domain->domain.id, dev_name(dev));
 		break;
-	case BUS_NOTIFY_UNBIND_DRIVER:
+	case BUS_NOTIFY_UNBOUND_DRIVER:
 		if (!domain)
 			goto out;
 		detach_device(domain, devid);
-- 
1.6.3.1