[RFC PATCH 00/14] IOMMU: irq-remapping and dmar hotplug support

[RFC PATCH 00/14] IOMMU: irq-remapping and dmar hotplug support

[Yinghai Lu]

During pci root bus hot-add, ioapic and iommu need to be parsed at first.
Using acpi_pci_driver->add to do that.

The patches need to apply to for-pci-root-bus-hotplug and for-x86-irq
and those two patches have been rebased to 3.4-rc1.

could get from
        git://git.kernel.org/pub/scm/linux/kernel/git/yinghai/linux-yinghai.git for-iommu

Thanks

Yinghai

Yinghai Lu (14):
  ACPI, PCI: Stop pci devices before acpi_pci_driver remove calling
  PCI, x86: Move pci_enable_bridges() down
  ACPI, PCI: Skip extra pci_enable_bridges for non hot-add root
  PCI: set dev_node early for pci_dev
  IOMMU: Update dmar units devices list during hotplug
  IOMMU: Fix tboot force iommu logic
  IOMMU: Don't clean handler data before free_irq()
  IOMMU: iommu_unique_seq_id()
  ACPI: Add acpi_run_dsm()
  IOMMU: Separate free_dmar_iommu from free_iommu
  IOMMU: Add init_dmar_one()
  IOMMU: Add intel_enable_irq_remapping_one()
  IOMMU: Add dmar_parse_one_drhd()
  IOMMU: Add intel iommu irq-remapping and dmar hotplug support

 arch/x86/pci/legacy.c               |    1 +
 drivers/acpi/bus.c                  |   77 +++++++
 drivers/acpi/pci_root.c             |   15 +-
 drivers/acpi/pci_root_hp.c          |    1 +
 drivers/iommu/dmar.c                |  420 +++++++++++++++++++++++++++++++++--
 drivers/iommu/intel-iommu.c         |  167 ++++++++++++--
 drivers/iommu/intel_irq_remapping.c |  172 ++++++++++++++-
 drivers/pci/probe.c                 |    2 +
 drivers/pci/remove.c                |    2 +-
 drivers/pci/setup-bus.c             |    2 -
 include/acpi/acpi_bus.h             |    1 +
 include/linux/acpi.h                |   10 +
 include/linux/dmar.h                |   59 +++++-
 include/linux/pci.h                 |    1 +
 14 files changed, 867 insertions(+), 63 deletions(-)

-- 
1.7.7

[RFC PATCH 01/14] ACPI, PCI: Stop pci devices before acpi_pci_driver remove calling

[Yinghai Lu]

During stop pci drivers, it still need to access ioapic and iommu.
So need to make sure those drivers need to be stop at first.

Also change the acpi_pci_drivers remove calling sequence to reverse.

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
---
 drivers/acpi/pci_root.c |   11 +++++++----
 drivers/pci/remove.c    |    2 +-
 include/linux/pci.h     |    1 +
 3 files changed, 9 insertions(+), 5 deletions(-)

diff --git a/drivers/acpi/pci_root.c b/drivers/acpi/pci_root.c
index 4a7d575..ce662ae 100644
--- a/drivers/acpi/pci_root.c
+++ b/drivers/acpi/pci_root.c
@@ -657,10 +657,6 @@ static int acpi_pci_root_remove(struct acpi_device *device, int type)
 
 	mutex_lock(&acpi_pci_root_lock);
 
-	list_for_each_entry(driver, &acpi_pci_drivers, node)
-		if (driver->remove)
-			driver->remove(root->device->handle);
-
 	/* that root bus could be removed already */
 	if (!pci_find_bus(root->segment, root->secondary.start)) {
 		dev_printk(KERN_DEBUG, &device->dev,
@@ -668,6 +664,13 @@ static int acpi_pci_root_remove(struct acpi_device *device, int type)
 		goto out;
 	}
 
+	/* stop normal pci drivers before we stop ioapic and dmar etc */
+	pci_stop_bus_devices(root->bus);
+
+	list_for_each_entry_reverse(driver, &acpi_pci_drivers, node)
+		if (driver->remove)
+			driver->remove(root->device->handle);
+
 	device_set_run_wake(root->bus->bridge, false);
 	pci_acpi_remove_bus_pm_notifier(device);
 
diff --git a/drivers/pci/remove.c b/drivers/pci/remove.c
index 18efb31..f672731 100644
--- a/drivers/pci/remove.c
+++ b/drivers/pci/remove.c
@@ -118,7 +118,7 @@ static void __pci_remove_bus_devices(struct pci_bus *bus)
 		__pci_remove_bus_device(dev);
 }
 
-static void pci_stop_bus_devices(struct pci_bus *bus)
+void pci_stop_bus_devices(struct pci_bus *bus)
 {
 	struct pci_dev *dev, *tmp;
 
diff --git a/include/linux/pci.h b/include/linux/pci.h
index 8199563..944f49e 100644
--- a/include/linux/pci.h
+++ b/include/linux/pci.h
@@ -706,6 +706,7 @@ extern void pci_dev_put(struct pci_dev *dev);
 extern void pci_remove_bus(struct pci_bus *b);
 extern void __pci_remove_bus_device(struct pci_dev *dev);
 extern void pci_stop_and_remove_bus_device(struct pci_dev *dev);
+void pci_stop_bus_devices(struct pci_bus *bus);
 void pci_stop_and_remove_bus(struct pci_bus *bus);
 extern void pci_stop_bus_device(struct pci_dev *dev);
 void pci_setup_cardbus(struct pci_bus *bus);
-- 
1.7.7

[RFC PATCH 02/14] PCI, x86: Move pci_enable_bridges() down

[Yinghai Lu]

After we get hot-added ioapic registered.

pci_enable_bridges will try to enable ioapic irq for pci bridge.

So need to move it down.

Or We can move out pcibios_enable_irq() out of pci_enable_device()
and call pcibios_enable_irq in pci_bus_add_devices ?
also will need to move ...
        pcibios_resource_survey_bus(root->bus);
        pci_assign_unassigned_bus_resources(root->bus);
to the start add ....

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
---
 arch/x86/pci/legacy.c   |    1 +
 drivers/acpi/pci_root.c |    3 +++
 drivers/pci/probe.c     |    1 +
 drivers/pci/setup-bus.c |    2 --
 4 files changed, 5 insertions(+), 2 deletions(-)

diff --git a/arch/x86/pci/legacy.c b/arch/x86/pci/legacy.c
index 8b6048a..5c63bb6 100644
--- a/arch/x86/pci/legacy.c
+++ b/arch/x86/pci/legacy.c
@@ -101,6 +101,7 @@ void __ref pcibios_root_rescan(void)
 
 		pcibios_resource_survey_bus(bus);
 		pci_assign_unassigned_bus_resources(bus);
+		pci_enable_bridges(bus);
 
 		pci_bus_add_devices(bus);
 	}
diff --git a/drivers/acpi/pci_root.c b/drivers/acpi/pci_root.c
index ce662ae..95b8f22 100644
--- a/drivers/acpi/pci_root.c
+++ b/drivers/acpi/pci_root.c
@@ -644,6 +644,9 @@ static int acpi_pci_root_start(struct acpi_device *device)
 		if (driver->add)
 			driver->add(device->handle);
 
+	/* need to after hot-added ioapic is registered */
+	pci_enable_bridges(root->bus);
+
 	pci_bus_add_devices(root->bus);
 	mutex_unlock(&acpi_pci_root_lock);
 
diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c
index bfa111b..320f38e 100644
--- a/drivers/pci/probe.c
+++ b/drivers/pci/probe.c
@@ -2065,6 +2065,7 @@ unsigned int __ref pci_rescan_bus(struct pci_bus *bus)
 
 	max = pci_scan_child_bus(bus);
 	pci_assign_unassigned_bus_resources(bus);
+	pci_enable_bridges(bus);
 	pci_bus_add_devices(bus);
 
 	return max;
diff --git a/drivers/pci/setup-bus.c b/drivers/pci/setup-bus.c
index d9a5500..20ccd89 100644
--- a/drivers/pci/setup-bus.c
+++ b/drivers/pci/setup-bus.c
@@ -1504,6 +1504,4 @@ void pci_assign_unassigned_bus_resources(struct pci_bus *bus)
 	up_read(&pci_bus_sem);
 	__pci_bus_assign_resources(bus, &add_list, NULL);
 	BUG_ON(!list_empty(&add_list));
-
-	pci_enable_bridges(bus);
 }
-- 
1.7.7

[RFC PATCH 03/14] ACPI, PCI: Skip extra pci_enable_bridges for non hot-add root

[Yinghai Lu]

Pre-installed will be handled later, need to skip the one only for hot-added
roots.

Otherwise will get anonying failing message about can not reserve, irq ...

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
---
 drivers/acpi/pci_root.c    |    3 ++-
 drivers/acpi/pci_root_hp.c |    1 +
 include/acpi/acpi_bus.h    |    1 +
 3 files changed, 4 insertions(+), 1 deletions(-)

diff --git a/drivers/acpi/pci_root.c b/drivers/acpi/pci_root.c
index 95b8f22..b313125 100644
--- a/drivers/acpi/pci_root.c
+++ b/drivers/acpi/pci_root.c
@@ -645,7 +645,8 @@ static int acpi_pci_root_start(struct acpi_device *device)
 			driver->add(device->handle);
 
 	/* need to after hot-added ioapic is registered */
-	pci_enable_bridges(root->bus);
+	if (root->hot_added)
+		pci_enable_bridges(root->bus);
 
 	pci_bus_add_devices(root->bus);
 	mutex_unlock(&acpi_pci_root_lock);
diff --git a/drivers/acpi/pci_root_hp.c b/drivers/acpi/pci_root_hp.c
index e16b53d..7d7f809 100644
--- a/drivers/acpi/pci_root_hp.c
+++ b/drivers/acpi/pci_root_hp.c
@@ -86,6 +86,7 @@ static void acpi_root_configure_bridge(acpi_handle handle)
 
 	pcibios_resource_survey_bus(root->bus);
 	pci_assign_unassigned_bus_resources(root->bus);
+	root->hot_added = true;
 }
 
 static void handle_root_bridge_insertion(acpi_handle handle)
diff --git a/include/acpi/acpi_bus.h b/include/acpi/acpi_bus.h
index f1c8ca6..333d38c 100644
--- a/include/acpi/acpi_bus.h
+++ b/include/acpi/acpi_bus.h
@@ -370,6 +370,7 @@ struct acpi_pci_root {
 
 	u32 osc_support_set;	/* _OSC state of support bits */
 	u32 osc_control_set;	/* _OSC state of control bits */
+	bool hot_added;
 };
 
 /* helper */
-- 
1.7.7

[RFC PATCH 04/14] PCI: set dev_node early for pci_dev

[Yinghai Lu]

otherwise irq_desc for pci bridge with hot-added ioapic can not be
on local node.

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
---
 drivers/pci/probe.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c
index 320f38e..1eed2fd 100644
--- a/drivers/pci/probe.c
+++ b/drivers/pci/probe.c
@@ -1447,6 +1447,7 @@ void pci_device_add(struct pci_dev *dev, struct pci_bus *bus)
 	dev->dev.release = pci_release_dev;
 	pci_dev_get(dev);
 
+	set_dev_node(&dev->dev, pcibus_to_node(bus));
 	dev->dev.dma_mask = &dev->dma_mask;
 	dev->dev.dma_parms = &dev->dma_parms;
 	dev->dev.coherent_dma_mask = 0xffffffffull;
-- 
1.7.7

[RFC PATCH 05/14] IOMMU: Update dmar units devices list during hotplug

[Yinghai Lu]

When do pci remove/rescan on system that have more iommus, got

[  894.089745] Set context mapping for c4:00.0
[  894.110890] mpt2sas3: Allocated physical memory: size(4293 kB)
[  894.112556] mpt2sas3: Current Controller Queue Depth(1883), Max Controller Queue Depth(2144)
[  894.127278] mpt2sas3: Scatter Gather Elements per IO(128)
[  894.361295] DRHD: handling fault status reg 2
[  894.364053] DMAR:[DMA Read] Request device [c4:00.0] fault addr fffbe000
[  894.364056] DMAR:[fault reason 02] Present bit in context entry is cl

It turns out when remove/rescan, pci dev will be freed and will get another
new dev. But drhd units still keep old one... so dmar_find_matched_drhd_unit
will return wrong drhd and iommu for the device that is not on first iommu.

So need to update devices in drhd_units during pci remove/rescan.

Could save domain/bus/device/function aside in the list and according that
info restore new dev to drhd_units later.
Then dmar_find_matched_drdh_unit and device_to_iommu could return right drhd
and iommu.

Add remove_dev_from_drhd/restore_dev_to_drhd functions to do the real work.
call them in device ADD_DEVICE and UNBOUND_DRIVER

Need to do the samething to atsr. (expose dmar_atsr_units and add
atsru->segment)

After patch, will have right iommu for the new dev and will not get DMAR
error anymore.

-v2: add pci_dev_put/pci_dev_get to make refcount consistent.
-v3: fix one left over CONFIG_DMAR
-v4: pass pci_dev *dev in save_dev_dmaru()/get_dev_dmaru() according to Bjorn.
-v5: fix case only have intr-remap enabled.

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Vinod Koul <vinod.koul@intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: iommu@lists.linux-foundation.org
---
 drivers/iommu/dmar.c                |  129 +++++++++++++++++++++++++++++++++++
 drivers/iommu/intel-iommu.c         |   36 +++++++---
 drivers/iommu/intel_irq_remapping.c |   36 ++++++++++-
 include/linux/dmar.h                |   17 +++++
 4 files changed, 206 insertions(+), 12 deletions(-)

diff --git a/drivers/iommu/dmar.c b/drivers/iommu/dmar.c
index ec01bc7..fde4991 100644
--- a/drivers/iommu/dmar.c
+++ b/drivers/iommu/dmar.c
@@ -455,6 +455,135 @@ fail:
 	return ret;
 }
 
+#ifdef CONFIG_HOTPLUG
+struct dev_dmaru {
+	struct list_head list;
+	void *dmaru;
+	int index;
+	int segment;
+	unsigned char bus;
+	unsigned int devfn;
+};
+
+static int save_dev_dmaru(struct pci_dev *dev, void *dmaru,
+			  int index, struct list_head *lh)
+{
+	struct dev_dmaru *m;
+
+	m = kzalloc(sizeof(*m), GFP_KERNEL);
+	if (!m)
+		return -ENOMEM;
+
+	m->segment = pci_domain_nr(dev->bus);
+	m->bus     = dev->bus->number;
+	m->devfn   = dev->devfn;
+	m->dmaru   = dmaru;
+	m->index   = index;
+
+	list_add(&m->list, lh);
+
+	return 0;
+}
+static void *get_dev_dmaru(struct pci_dev *dev, int *index,
+				struct list_head *lh)
+{
+	int segment = pci_domain_nr(dev->bus);
+	unsigned char bus = dev->bus->number;
+	unsigned int devfn = dev->devfn;
+	struct dev_dmaru *m;
+	void *dmaru = NULL;
+
+	list_for_each_entry(m, lh, list) {
+		if (m->segment == segment &&
+		    m->bus == bus && m->devfn == devfn) {
+			*index = m->index;
+			dmaru  = m->dmaru;
+			list_del(&m->list);
+			kfree(m);
+			break;
+		}
+	}
+
+	return dmaru;
+}
+
+static LIST_HEAD(saved_dev_drhd_list);
+void remove_dev_from_drhd(struct pci_dev *dev)
+{
+	struct dmar_drhd_unit *drhd = NULL;
+	int segment = pci_domain_nr(dev->bus);
+	int i;
+
+	for_each_drhd_unit(drhd) {
+		if (drhd->ignored)
+			continue;
+		if (segment != drhd->segment)
+			continue;
+
+		for (i = 0; i < drhd->devices_cnt; i++) {
+			if (drhd->devices[i] == dev) {
+				/* save it at first if it is in drhd */
+				save_dev_dmaru(dev, drhd, i,
+						&saved_dev_drhd_list);
+				/* always remove it */
+				pci_dev_put(dev);
+				drhd->devices[i] = NULL;
+				return;
+			}
+		}
+	}
+}
+void restore_dev_to_drhd(struct pci_dev *dev)
+{
+	struct dmar_drhd_unit *drhd = NULL;
+	int i;
+
+	/* find the stored drhd */
+	drhd = get_dev_dmaru(dev, &i, &saved_dev_drhd_list);
+	/* restore that into drhd */
+	if (drhd)
+		drhd->devices[i] = pci_dev_get(dev);
+}
+
+#ifdef CONFIG_INTEL_IOMMU
+static LIST_HEAD(saved_dev_atsr_list);
+void remove_dev_from_atsr(struct pci_dev *dev)
+{
+	struct dmar_atsr_unit *atsr = NULL;
+	int segment = pci_domain_nr(dev->bus);
+	int i;
+
+	for_each_atsr_unit(atsr) {
+		if (segment != atsr->segment)
+			continue;
+
+		for (i = 0; i < atsr->devices_cnt; i++) {
+			if (atsr->devices[i] == dev) {
+				/* save it at first if it is in drhd */
+				save_dev_dmaru(dev, atsr, i,
+						&saved_dev_atsr_list);
+				/* always remove it */
+				pci_dev_put(dev);
+				atsr->devices[i] = NULL;
+				return;
+			}
+		}
+	}
+}
+void restore_dev_to_atsr(struct pci_dev *dev)
+{
+	struct dmar_atsr_unit *atsr = NULL;
+	int i;
+
+	/* find the stored atsr */
+	atsr = get_dev_dmaru(dev, &i, &saved_dev_atsr_list);
+	/* restore that into atsr */
+	if (atsr)
+		atsr->devices[i] = pci_dev_get(dev);
+}
+#endif  /* CONFIG_INTEL_IOMMU */
+
+#endif  /* CONFIG_HOTPLUG */
 
 int __init dmar_table_init(void)
 {
diff --git a/drivers/iommu/intel-iommu.c b/drivers/iommu/intel-iommu.c
index 430224b..be867e6 100644
--- a/drivers/iommu/intel-iommu.c
+++ b/drivers/iommu/intel-iommu.c
@@ -3472,7 +3472,7 @@ rmrr_parse_dev(struct dmar_rmrr_unit *rmrru)
 	return ret;
 }
 
-static LIST_HEAD(dmar_atsr_units);
+LIST_HEAD(dmar_atsr_units);
 
 int __init dmar_parse_one_atsr(struct acpi_dmar_header *hdr)
 {
@@ -3486,6 +3486,7 @@ int __init dmar_parse_one_atsr(struct acpi_dmar_header *hdr)
 
 	atsru->hdr = hdr;
 	atsru->include_all = atsr->flags & 0x1;
+	atsru->segment = atsr->segment;
 
 	list_add(&atsru->list, &dmar_atsr_units);
 
@@ -3517,16 +3518,13 @@ int dmar_find_matched_atsr_unit(struct pci_dev *dev)
 {
 	int i;
 	struct pci_bus *bus;
-	struct acpi_dmar_atsr *atsr;
 	struct dmar_atsr_unit *atsru;
 
 	dev = pci_physfn(dev);
 
-	list_for_each_entry(atsru, &dmar_atsr_units, list) {
-		atsr = container_of(atsru->hdr, struct acpi_dmar_atsr, header);
-		if (atsr->segment == pci_domain_nr(dev->bus))
+	list_for_each_entry(atsru, &dmar_atsr_units, list)
+		if (atsru->segment == pci_domain_nr(dev->bus))
 			goto found;
-	}
 
 	return 0;
 
@@ -3586,20 +3584,36 @@ static int device_notifier(struct notifier_block *nb,
 	struct pci_dev *pdev = to_pci_dev(dev);
 	struct dmar_domain *domain;
 
-	if (iommu_no_mapping(dev))
+	if (unlikely(dev->bus != &pci_bus_type))
 		return 0;
 
-	domain = find_domain(pdev);
-	if (!domain)
-		return 0;
+	switch (action) {
+	case BUS_NOTIFY_UNBOUND_DRIVER:
+		if (iommu_no_mapping(dev))
+			goto out;
+
+		if (iommu_pass_through)
+			goto out;
+
+		domain = find_domain(pdev);
+		if (!domain)
+			goto out;
 
-	if (action == BUS_NOTIFY_UNBOUND_DRIVER && !iommu_pass_through) {
 		domain_remove_one_dev_info(domain, pdev);
 
 		if (!(domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE) &&
 		    !(domain->flags & DOMAIN_FLAG_STATIC_IDENTITY) &&
 		    list_empty(&domain->devices))
 			domain_exit(domain);
+out:
+		remove_dev_from_drhd(pdev);
+		remove_dev_from_atsr(pdev);
+
+		break;
+	case BUS_NOTIFY_ADD_DEVICE:
+		restore_dev_to_drhd(pdev);
+		restore_dev_to_atsr(pdev);
+		break;
 	}
 
 	return 0;
diff --git a/drivers/iommu/intel_irq_remapping.c b/drivers/iommu/intel_irq_remapping.c
index b4d3950..05ffe20 100644
--- a/drivers/iommu/intel_irq_remapping.c
+++ b/drivers/iommu/intel_irq_remapping.c
@@ -757,12 +757,46 @@ int __init parse_ioapics_under_ir(void)
 	return ir_supported;
 }
 
+static int device_notifier(struct notifier_block *nb,
+				  unsigned long action, void *data)
+{
+	struct device *dev = data;
+	struct pci_dev *pdev = to_pci_dev(dev);
+
+	if (unlikely(dev->bus != &pci_bus_type))
+		return 0;
+
+	switch (action) {
+	case BUS_NOTIFY_UNBOUND_DRIVER:
+		remove_dev_from_drhd(pdev);
+		break;
+	case BUS_NOTIFY_ADD_DEVICE:
+		restore_dev_to_drhd(pdev);
+		break;
+	}
+
+	return 0;
+}
+
+static struct notifier_block device_nb = {
+	.notifier_call = device_notifier,
+};
+
 int __init ir_dev_scope_init(void)
 {
+	int ret;
+
 	if (!irq_remapping_enabled)
 		return 0;
 
-	return dmar_dev_scope_init();
+	ret = dmar_dev_scope_init();
+	if (ret < 0)
+		return ret;
+
+	if (intel_iommu_enabled != 1)
+		bus_register_notifier(&pci_bus_type, &device_nb);
+
+	return ret;
 }
 rootfs_initcall(ir_dev_scope_init);
 
diff --git a/include/linux/dmar.h b/include/linux/dmar.h
index b029d1a..305db55 100644
--- a/include/linux/dmar.h
+++ b/include/linux/dmar.h
@@ -131,6 +131,18 @@ extern int dmar_set_interrupt(struct intel_iommu *iommu);
 extern irqreturn_t dmar_fault(int irq, void *dev_id);
 extern int arch_setup_dmar_msi(unsigned int irq);
 
+#ifdef CONFIG_HOTPLUG
+void remove_dev_from_drhd(struct pci_dev *dev);
+void restore_dev_to_drhd(struct pci_dev *dev);
+void remove_dev_from_atsr(struct pci_dev *dev);
+void restore_dev_to_atsr(struct pci_dev *dev);
+#else
+static inline void remove_dev_from_drhd(struct pci_dev *dev) { }
+static inline void restore_dev_to_drhd(struct pci_dev *dev) { }
+static inline void remove_dev_from_atsr(struct pci_dev *dev) { }
+static inline void restore_dev_to_atsr(struct pci_dev *dev) { }
+#endif
+
 #ifdef CONFIG_INTEL_IOMMU
 extern int iommu_detected, no_iommu;
 extern struct list_head dmar_rmrr_units;
@@ -146,14 +158,19 @@ struct dmar_rmrr_unit {
 #define for_each_rmrr_units(rmrr) \
 	list_for_each_entry(rmrr, &dmar_rmrr_units, list)
 
+extern struct list_head dmar_atsr_units;
 struct dmar_atsr_unit {
 	struct list_head list;		/* list of ATSR units */
 	struct acpi_dmar_header *hdr;	/* ACPI header */
 	struct pci_dev **devices;	/* target devices */
 	int devices_cnt;		/* target device count */
+	u16 segment;			/* PCI domain */
 	u8 include_all:1;		/* include all ports */
 };
 
+#define for_each_atsr_unit(atsr) \
+	list_for_each_entry(atsr, &dmar_atsr_units, list)
+
 int dmar_parse_rmrr_atsr_dev(void);
 extern int dmar_parse_one_rmrr(struct acpi_dmar_header *header);
 extern int dmar_parse_one_atsr(struct acpi_dmar_header *header);
-- 
1.7.7

[RFC PATCH 06/14] IOMMU: Fix tboot force iommu logic

[Yinghai Lu]

Should check dmar_disabled just after tboot_force_iommu.

Otherwise when tboot is not used, and intel_iommu=off, and nointrmap
still get dmar_table_init() called. that will cause some get_device
calling, and it will have some device refcount leaking.

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: iommu@lists.linux-foundation.org
---
 drivers/iommu/intel-iommu.c |    6 +++---
 1 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/iommu/intel-iommu.c b/drivers/iommu/intel-iommu.c
index be867e6..b7e5a7e 100644
--- a/drivers/iommu/intel-iommu.c
+++ b/drivers/iommu/intel-iommu.c
@@ -3630,6 +3630,9 @@ int __init intel_iommu_init(void)
 	/* VT-d is required for a TXT/tboot launch, so enforce that */
 	force_on = tboot_force_iommu();
 
+	if (no_iommu || dmar_disabled)
+		return -ENODEV;
+
 	if (dmar_table_init()) {
 		if (force_on)
 			panic("tboot: Failed to initialize DMAR table\n");
@@ -3642,9 +3645,6 @@ int __init intel_iommu_init(void)
 		return 	-ENODEV;
 	}
 
-	if (no_iommu || dmar_disabled)
-		return -ENODEV;
-
 	if (iommu_init_mempool()) {
 		if (force_on)
 			panic("tboot: Failed to initialize iommu memory\n");
-- 
1.7.7

[RFC PATCH 07/14] IOMMU: Don't clean handler data before free_irq()

[Yinghai Lu]

need that in __free_irq.

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
---
 drivers/iommu/intel-iommu.c |    1 -
 1 files changed, 0 insertions(+), 1 deletions(-)

diff --git a/drivers/iommu/intel-iommu.c b/drivers/iommu/intel-iommu.c
index b7e5a7e..4116944 100644
--- a/drivers/iommu/intel-iommu.c
+++ b/drivers/iommu/intel-iommu.c
@@ -1308,7 +1308,6 @@ void free_dmar_iommu(struct intel_iommu *iommu)
 		iommu_disable_translation(iommu);
 
 	if (iommu->irq) {
-		irq_set_handler_data(iommu->irq, NULL);
 		/* This will mask the irq */
 		free_irq(iommu->irq, iommu);
 		destroy_irq(iommu->irq);
-- 
1.7.7

[RFC PATCH 08/14] IOMMU: iommu_unique_seq_id()

[Yinghai Lu]

So for hot-remove/hot-add will reuse seq_id.

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
---
 drivers/iommu/dmar.c |   16 ++++++++++++++--
 1 files changed, 14 insertions(+), 2 deletions(-)

diff --git a/drivers/iommu/dmar.c b/drivers/iommu/dmar.c
index fde4991..043192e 100644
--- a/drivers/iommu/dmar.c
+++ b/drivers/iommu/dmar.c
@@ -707,13 +707,23 @@ int __init detect_intel_iommu(void)
 	return ret ? 1 : -ENODEV;
 }
 
+static DECLARE_BITMAP(iommu_allocated, 1024);
+
+static int iommu_unique_seq_id(void)
+{
+	int id;
+
+	id = find_first_zero_bit(iommu_allocated, 1024);
+	__set_bit(id, iommu_allocated);
+
+	return id;
+}
 
 int alloc_iommu(struct dmar_drhd_unit *drhd)
 {
 	struct intel_iommu *iommu;
 	int map_size;
 	u32 ver;
-	static int iommu_allocated = 0;
 	int agaw = 0;
 	int msagaw = 0;
 
@@ -726,7 +736,7 @@ int alloc_iommu(struct dmar_drhd_unit *drhd)
 	if (!iommu)
 		return -ENOMEM;
 
-	iommu->seq_id = iommu_allocated++;
+	iommu->seq_id = iommu_unique_seq_id();
 	sprintf (iommu->name, "dmar%d", iommu->seq_id);
 
 	iommu->reg = ioremap(drhd->reg_base_addr, VTD_PAGE_SIZE);
@@ -790,6 +800,7 @@ int alloc_iommu(struct dmar_drhd_unit *drhd)
  err_unmap:
 	iounmap(iommu->reg);
  error:
+	__clear_bit(iommu->seq_id, iommu_allocated);
 	kfree(iommu);
 	return -1;
 }
@@ -803,6 +814,7 @@ void free_iommu(struct intel_iommu *iommu)
 
 	if (iommu->reg)
 		iounmap(iommu->reg);
+	__clear_bit(iommu->seq_id, iommu_allocated);
 	kfree(iommu);
 }
 
-- 
1.7.7

[RFC PATCH 09/14] ACPI: Add acpi_run_dsm()

[Yinghai Lu]

Will need to use that get dmar entries for hotplug

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
---
 drivers/acpi/bus.c   |   77 ++++++++++++++++++++++++++++++++++++++++++++++++++
 include/linux/acpi.h |   10 ++++++
 2 files changed, 87 insertions(+), 0 deletions(-)

diff --git a/drivers/acpi/bus.c b/drivers/acpi/bus.c
index e0960c7..9025dbb 100644
--- a/drivers/acpi/bus.c
+++ b/drivers/acpi/bus.c
@@ -520,6 +520,83 @@ out_kfree:
 }
 EXPORT_SYMBOL(acpi_run_osc);
 
+static void acpi_print_dsm_error(acpi_handle handle,
+	struct acpi_dsm_context *context, char *error)
+{
+	struct acpi_buffer buffer = {ACPI_ALLOCATE_BUFFER};
+
+	if (ACPI_FAILURE(acpi_get_name(handle, ACPI_FULL_PATHNAME, &buffer)))
+		printk(KERN_DEBUG "%s for idx %x\n", error, context->func_idx);
+	else {
+		printk(KERN_DEBUG "%s:%s for idx %x\n",
+			(char *)buffer.pointer, error, context->func_idx);
+		kfree(buffer.pointer);
+	}
+}
+
+acpi_status acpi_run_dsm(acpi_handle handle, struct acpi_dsm_context *context)
+{
+	acpi_status status;
+	struct acpi_object_list input;
+	union acpi_object in_params[4];
+	union acpi_object *out_obj;
+	u8 uuid[16];
+	struct acpi_buffer output = {ACPI_ALLOCATE_BUFFER, NULL};
+
+	if (!context)
+		return AE_ERROR;
+	if (ACPI_FAILURE(acpi_str_to_uuid(context->uuid_str, uuid)))
+		return AE_ERROR;
+	context->ret.length = ACPI_ALLOCATE_BUFFER;
+	context->ret.pointer = NULL;
+
+	/* Setting up input parameters */
+	input.count = 4;
+	input.pointer = in_params;
+	in_params[0].type		= ACPI_TYPE_BUFFER;
+	in_params[0].buffer.length	= 16;
+	in_params[0].buffer.pointer	= uuid;
+	in_params[1].type		= ACPI_TYPE_INTEGER;
+	in_params[1].integer.value	= context->rev;
+	in_params[2].type		= ACPI_TYPE_INTEGER;
+	in_params[2].integer.value	= context->func_idx;
+	in_params[3].type		= ACPI_TYPE_PACKAGE;
+	in_params[3].package.count	= context->func_argc;
+	in_params[3].package.elements	= context->func_argv;
+
+	status = acpi_evaluate_object(handle, "_DSM", &input, &output);
+	if (ACPI_FAILURE(status))
+		return status;
+
+	if (!output.length)
+		return AE_NULL_OBJECT;
+
+	out_obj = output.pointer;
+	if (out_obj->type != ACPI_TYPE_BUFFER) {
+		acpi_print_dsm_error(handle, context,
+			"_DSM evaluation returned wrong type");
+		status = AE_TYPE;
+		goto out_kfree;
+	}
+
+	context->ret.length = out_obj->buffer.length;
+	context->ret.pointer = kmalloc(context->ret.length, GFP_KERNEL);
+	if (!context->ret.pointer) {
+		status =  AE_NO_MEMORY;
+		goto out_kfree;
+	}
+	memcpy(context->ret.pointer, out_obj->buffer.pointer,
+		context->ret.length);
+	status =  AE_OK;
+
+out_kfree:
+	kfree(output.pointer);
+	if (status != AE_OK)
+		context->ret.pointer = NULL;
+	return status;
+}
+EXPORT_SYMBOL(acpi_run_dsm);
+
 bool osc_sb_apei_support_acked;
 static u8 sb_uuid_str[] = "0811B06E-4A27-44F9-8D60-3CBBC22E7B48";
 static void acpi_bus_osc_support(void)
diff --git a/include/linux/acpi.h b/include/linux/acpi.h
index d78baa7..14ad7d7 100644
--- a/include/linux/acpi.h
+++ b/include/linux/acpi.h
@@ -273,6 +273,16 @@ struct acpi_osc_context {
 
 acpi_status acpi_run_osc(acpi_handle handle, struct acpi_osc_context *context);
 
+struct acpi_dsm_context {
+	char *uuid_str; /* uuid string */
+	int rev;
+	int func_idx;	/* function index */
+	int func_argc;
+	union acpi_object *func_argv;
+	struct acpi_buffer ret; /* free by caller if success */
+};
+acpi_status acpi_run_dsm(acpi_handle handle, struct acpi_dsm_context *context);
+
 /* platform-wide _OSC bits */
 #define OSC_SB_PAD_SUPPORT		1
 #define OSC_SB_PPC_OST_SUPPORT		2
-- 
1.7.7

[RFC PATCH 10/14] IOMMU: Separate free_dmar_iommu from free_iommu

[Yinghai Lu]

We will need separate two calling later for iommu_remove.

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
---
 drivers/iommu/dmar.c        |    2 --
 drivers/iommu/intel-iommu.c |    1 +
 2 files changed, 1 insertions(+), 2 deletions(-)

diff --git a/drivers/iommu/dmar.c b/drivers/iommu/dmar.c
index 043192e..ea9d210 100644
--- a/drivers/iommu/dmar.c
+++ b/drivers/iommu/dmar.c
@@ -810,8 +810,6 @@ void free_iommu(struct intel_iommu *iommu)
 	if (!iommu)
 		return;
 
-	free_dmar_iommu(iommu);
-
 	if (iommu->reg)
 		iounmap(iommu->reg);
 	__clear_bit(iommu->seq_id, iommu_allocated);
diff --git a/drivers/iommu/intel-iommu.c b/drivers/iommu/intel-iommu.c
index 4116944..662932f 100644
--- a/drivers/iommu/intel-iommu.c
+++ b/drivers/iommu/intel-iommu.c
@@ -2613,6 +2613,7 @@ error:
 		if (drhd->ignored)
 			continue;
 		iommu = drhd->iommu;
+		free_dmar_iommu(iommu);
 		free_iommu(iommu);
 	}
 	kfree(g_iommus);
-- 
1.7.7

[RFC PATCH 11/14] IOMMU: Add init_dmar_one()

[Yinghai Lu]

Will need that for hot added intel iommu

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
---
 drivers/iommu/dmar.c        |    4 +-
 drivers/iommu/intel-iommu.c |  123 +++++++++++++++++++++++++++++++++++++++----
 2 files changed, 115 insertions(+), 12 deletions(-)

diff --git a/drivers/iommu/dmar.c b/drivers/iommu/dmar.c
index ea9d210..cd5eee3 100644
--- a/drivers/iommu/dmar.c
+++ b/drivers/iommu/dmar.c
@@ -62,7 +62,7 @@ static void __init dmar_register_drhd_unit(struct dmar_drhd_unit *drhd)
 		list_add(&drhd->list, &dmar_drhd_units);
 }
 
-static int __init dmar_parse_one_dev_scope(struct acpi_dmar_device_scope *scope,
+static int dmar_parse_one_dev_scope(struct acpi_dmar_device_scope *scope,
 					   struct pci_dev **dev, u16 segment)
 {
 	struct pci_bus *bus;
@@ -119,7 +119,7 @@ static int __init dmar_parse_one_dev_scope(struct acpi_dmar_device_scope *scope,
 	return 0;
 }
 
-int __init dmar_parse_dev_scope(void *start, void *end, int *cnt,
+int dmar_parse_dev_scope(void *start, void *end, int *cnt,
 				struct pci_dev ***devices, u16 segment)
 {
 	struct acpi_dmar_device_scope *scope;
diff --git a/drivers/iommu/intel-iommu.c b/drivers/iommu/intel-iommu.c
index 662932f..56b886f 100644
--- a/drivers/iommu/intel-iommu.c
+++ b/drivers/iommu/intel-iommu.c
@@ -2407,19 +2407,12 @@ static int __init init_dmars(void)
 	 *    initialize and program root entry to not present
 	 * endfor
 	 */
-	for_each_drhd_unit(drhd) {
 		/*
 		 * lock not needed as this is only incremented in the single
 		 * threaded kernel __init code path all other access are read
 		 * only
 		 */
-		if (g_num_of_iommus < IOMMU_UNITS_SUPPORTED) {
-			g_num_of_iommus++;
-			continue;
-		}
-		printk_once(KERN_ERR "intel-iommu: exceeded %d IOMMUs\n",
-			  IOMMU_UNITS_SUPPORTED);
-	}
+	g_num_of_iommus = IOMMU_UNITS_SUPPORTED;
 
 	g_iommus = kcalloc(g_num_of_iommus, sizeof(struct intel_iommu *),
 			GFP_KERNEL);
@@ -2620,6 +2613,109 @@ error:
 	return ret;
 }
 
+int init_dmar_one(struct dmar_drhd_unit *drhd)
+{
+	struct intel_iommu *iommu;
+	int ret;
+
+	/*
+	 * for each drhd
+	 *    allocate root
+	 *    initialize and program root entry to not present
+	 * endfor
+	 */
+
+	if (drhd->ignored)
+		return 0;
+
+	iommu = drhd->iommu;
+	g_iommus[iommu->seq_id] = iommu;
+
+	ret = iommu_init_domains(iommu);
+	if (ret)
+		goto error;
+
+	/*
+	 * TBD:
+	 * we could share the same root & context tables
+	 * among all IOMMU's. Need to Split it later.
+	 */
+	ret = iommu_alloc_root_entry(iommu);
+	if (ret) {
+		printk(KERN_ERR "IOMMU: allocate root entry failed\n");
+		goto error;
+	}
+
+	/*
+	 * Start from the sane iommu hardware state.
+	 */
+	/*
+	 * If the queued invalidation is already initialized by us
+	 * (for example, while enabling interrupt-remapping) then
+	 * we got the things already rolling from a sane state.
+	 */
+	if (!iommu->qi) {
+		/*
+		 * Clear any previous faults.
+		 */
+		dmar_fault(-1, iommu);
+		/*
+		 * Disable queued invalidation if supported and already enabled
+		 * before OS handover.
+		 */
+		dmar_disable_qi(iommu);
+	}
+
+	if (dmar_enable_qi(iommu)) {
+		/*
+		 * Queued Invalidate not enabled, use Register Based
+		 * Invalidate
+		 */
+		iommu->flush.flush_context = __iommu_flush_context;
+		iommu->flush.flush_iotlb = __iommu_flush_iotlb;
+		printk(KERN_INFO
+		       "IOMMU %d 0x%Lx: using Register based invalidation\n",
+		       iommu->seq_id, (unsigned long long)drhd->reg_base_addr);
+	} else {
+		iommu->flush.flush_context = qi_flush_context;
+		iommu->flush.flush_iotlb = qi_flush_iotlb;
+		printk(KERN_INFO "IOMMU %d 0x%Lx: using Queued invalidation\n",
+		       iommu->seq_id, (unsigned long long)drhd->reg_base_addr);
+	}
+
+	/*
+	 * for each drhd
+	 *   enable fault log
+	 *   global invalidate context cache
+	 *   global invalidate iotlb
+	 *   enable translation
+	 */
+	iommu_flush_write_buffer(iommu);
+
+	ret = dmar_set_interrupt(iommu);
+	if (ret)
+		goto error;
+
+	iommu_set_root_entry(iommu);
+
+	iommu->flush.flush_context(iommu, 0, 0, 0, DMA_CCMD_GLOBAL_INVL);
+	iommu->flush.flush_iotlb(iommu, 0, 0, 0, DMA_TLB_GLOBAL_FLUSH);
+
+	ret = iommu_enable_translation(iommu);
+	if (ret)
+		goto error;
+
+	iommu_disable_protect_mem_regions(iommu);
+
+	return 0;
+error:
+	free_dmar_iommu(iommu);
+	free_iommu(iommu);
+	drhd->iommu = NULL;
+	return ret;
+}
+
+
 /* This takes a number of _MM_ pages, not VTD pages */
 static struct iova *intel_alloc_iova(struct device *dev,
 				     struct dmar_domain *domain,
@@ -3474,7 +3570,8 @@ rmrr_parse_dev(struct dmar_rmrr_unit *rmrru)
 
 LIST_HEAD(dmar_atsr_units);
 
-int __init dmar_parse_one_atsr(struct acpi_dmar_header *hdr)
+int __dmar_parse_one_atsr(struct acpi_dmar_header *hdr,
+			  struct dmar_atsr_unit **patsru)
 {
 	struct acpi_dmar_atsr *atsr;
 	struct dmar_atsr_unit *atsru;
@@ -3489,11 +3586,17 @@ int __init dmar_parse_one_atsr(struct acpi_dmar_header *hdr)
 	atsru->segment = atsr->segment;
 
 	list_add(&atsru->list, &dmar_atsr_units);
+	if (patsru)
+		*patsru = atsru;
 
 	return 0;
 }
+int __init dmar_parse_one_atsr(struct acpi_dmar_header *hdr)
+{
+	return __dmar_parse_one_atsr(hdr, NULL);
+}
 
-static int __init atsr_parse_dev(struct dmar_atsr_unit *atsru)
+int atsr_parse_dev(struct dmar_atsr_unit *atsru)
 {
 	int rc;
 	struct acpi_dmar_atsr *atsr;
-- 
1.7.7

[RFC PATCH 12/14] IOMMU: Add intel_enable_irq_remapping_one()

[Yinghai Lu]

Will need that for hot-added iommu interrupt remapping suppor.

Signed-off-by: Yinghai <yinghai@kernel.org>
---
 drivers/iommu/intel_irq_remapping.c |  136 +++++++++++++++++++++++++++++++++--
 1 files changed, 129 insertions(+), 7 deletions(-)

diff --git a/drivers/iommu/intel_irq_remapping.c b/drivers/iommu/intel_irq_remapping.c
index 05ffe20..e20503d 100644
--- a/drivers/iommu/intel_irq_remapping.c
+++ b/drivers/iommu/intel_irq_remapping.c
@@ -214,7 +214,7 @@ static struct intel_iommu *map_ioapic_to_ir(int apic)
 {
 	int i;
 
-	for (i = 0; i < MAX_IO_APICS; i++)
+	for (i = 0; i < ir_ioapic_num; i++)
 		if (ir_ioapic[i].id == apic)
 			return ir_ioapic[i].iommu;
 	return NULL;
@@ -320,7 +320,7 @@ static int set_ioapic_sid(struct irte *irte, int apic)
 	if (!irte)
 		return -1;
 
-	for (i = 0; i < MAX_IO_APICS; i++) {
+	for (i = 0; i < ir_ioapic_num; i++) {
 		if (ir_ioapic[i].id == apic) {
 			sid = (ir_ioapic[i].bus << 8) | ir_ioapic[i].devfn;
 			break;
@@ -662,6 +662,7 @@ static void ir_parse_one_ioapic_scope(struct acpi_dmar_device_scope *scope,
 	struct acpi_dmar_pci_path *path;
 	u8 bus;
 	int count;
+	int i;
 
 	bus = scope->bus;
 	path = (struct acpi_dmar_pci_path *)(scope + 1);
@@ -678,11 +679,16 @@ static void ir_parse_one_ioapic_scope(struct acpi_dmar_device_scope *scope,
 		path++;
 	}
 
-	ir_ioapic[ir_ioapic_num].bus   = bus;
-	ir_ioapic[ir_ioapic_num].devfn = PCI_DEVFN(path->dev, path->fn);
-	ir_ioapic[ir_ioapic_num].iommu = iommu;
-	ir_ioapic[ir_ioapic_num].id    = scope->enumeration_id;
-	ir_ioapic_num++;
+	for (i = 0; i < ir_ioapic_num; i++)
+		if (!ir_ioapic[i].iommu)
+			break;
+
+	ir_ioapic[i].bus   = bus;
+	ir_ioapic[i].devfn = PCI_DEVFN(path->dev, path->fn);
+	ir_ioapic[i].iommu = iommu;
+	ir_ioapic[i].id    = scope->enumeration_id;
+	if (i == ir_ioapic_num)
+		ir_ioapic_num++;
 }
 
 static int ir_parse_ioapic_hpet_scope(struct acpi_dmar_header *header,
@@ -1082,6 +1088,122 @@ static int intel_setup_hpet_msi(unsigned int irq, unsigned int id)
 	return 0;
 }
 
+void disable_irq_remapping_one(struct dmar_drhd_unit *drhd)
+{
+	struct intel_iommu *iommu = drhd->iommu;
+	int i;
+
+	/*
+	 * Disable Interrupt-remapping for the DRHD's now.
+	 */
+	if (!ecap_ir_support(iommu->ecap))
+		return;
+
+	iommu_disable_irq_remapping(iommu);
+
+	/* remove it from ir_ioapic */
+	for (i = 0; i < ir_ioapic_num; i++)
+		if (ir_ioapic[i].iommu == iommu) {
+			ir_ioapic[i].iommu = NULL;
+			ir_ioapic[i].id	= 0xff;
+		}
+}
+
+static int parse_ioapics_under_ir_one(struct dmar_drhd_unit *drhd)
+{
+	int ir_supported = 0;
+
+	struct intel_iommu *iommu = drhd->iommu;
+
+	if (ecap_ir_support(iommu->ecap)) {
+		if (ir_parse_ioapic_hpet_scope(drhd->hdr, iommu))
+			return -1;
+
+		ir_supported = 1;
+	}
+
+	return ir_supported;
+}
+
+int intel_enable_irq_remapping_one(struct dmar_drhd_unit *drhd)
+{
+	int setup = 0;
+	int eim = 1;
+	int ret;
+	struct intel_iommu *iommu = drhd->iommu;
+
+	if (parse_ioapics_under_ir_one(drhd) != 1) {
+		printk(KERN_INFO "Not enable interrupt remapping\n");
+		return -1;
+	}
+
+	/*
+	 * If the queued invalidation is already initialized,
+	 * shouldn't disable it.
+	 */
+	if (!iommu->qi) {
+		/*
+		 * Clear previous faults.
+		 */
+		dmar_fault(-1, iommu);
+
+		/*
+		 * Disable intr remapping and queued invalidation, if already
+		 * enabled prior to OS handover.
+		 */
+		iommu_disable_irq_remapping(iommu);
+
+		dmar_disable_qi(iommu);
+	}
+
+	/*
+	 * check for the Interrupt-remapping support
+	 */
+
+	if (ecap_ir_support(iommu->ecap))
+		if (eim && !ecap_eim_support(iommu->ecap)) {
+			printk(KERN_INFO
+			"DRHD %Lx: EIM not supported by DRHD, ecap %Lx\n",
+				drhd->reg_base_addr, iommu->ecap);
+			return -1;
+		}
+
+	/*
+	 * Enable queued invalidation for the DRHD's.
+	 */
+	ret = dmar_enable_qi(iommu);
+
+	if (ret) {
+		printk(KERN_ERR
+			"DRHD %Lx: failed to enable queued, invalidation, ecap %Lx, ret %d\n",
+		       drhd->reg_base_addr, iommu->ecap, ret);
+		return -1;
+	}
+
+	/*
+	 * Setup Interrupt-remapping for the DRHD's now.
+	 */
+	if (ecap_ir_support(iommu->ecap)) {
+		if (intel_setup_irq_remapping(iommu, eim))
+			goto error;
+
+		setup = 1;
+	}
+
+	if (!setup)
+		goto error;
+
+	pr_info("Enabled IRQ remapping in %s mode\n", eim ? "x2apic" : "xapic");
+
+	return eim ? IRQ_REMAP_X2APIC_MODE : IRQ_REMAP_XAPIC_MODE;
+
+error:
+	/*
+	 * handle error condition gracefully here!
+	 */
+	return -1;
+}
+
 struct irq_remap_ops intel_irq_remap_ops = {
 	.supported		= intel_irq_remapping_supported,
 	.prepare		= dmar_table_init,
-- 
1.7.7

[RFC PATCH 13/14] IOMMU: Add dmar_parse_one_drhd()

[Yinghai Lu]

To parse new hot-added iommu dmar entry table

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
---
 drivers/iommu/dmar.c |   51 ++++++++++++++++++++++++++++++++++++++++---------
 include/linux/dmar.h |   46 ++++++++++++++++++++++++++++++++++++--------
 2 files changed, 78 insertions(+), 19 deletions(-)

diff --git a/drivers/iommu/dmar.c b/drivers/iommu/dmar.c
index cd5eee3..f96b8ce 100644
--- a/drivers/iommu/dmar.c
+++ b/drivers/iommu/dmar.c
@@ -171,8 +171,9 @@ int dmar_parse_dev_scope(void *start, void *end, int *cnt,
  * structure which uniquely represent one DMA remapping hardware unit
  * present in the platform
  */
-static int __init
-dmar_parse_one_drhd(struct acpi_dmar_header *header)
+static int
+__dmar_parse_one_drhd(struct acpi_dmar_header *header,
+		      struct dmar_drhd_unit **pdmaru)
 {
 	struct acpi_dmar_hardware_unit *drhd;
 	struct dmar_drhd_unit *dmaru;
@@ -194,10 +195,18 @@ dmar_parse_one_drhd(struct acpi_dmar_header *header)
 		return ret;
 	}
 	dmar_register_drhd_unit(dmaru);
+	if (pdmaru)
+		*pdmaru = dmaru;
 	return 0;
 }
 
-static int __init dmar_parse_dev(struct dmar_drhd_unit *dmaru)
+static int __init
+dmar_parse_one_drhd(struct acpi_dmar_header *header)
+{
+	return __dmar_parse_one_drhd(header, NULL);
+}
+
+static int dmar_parse_dev(struct dmar_drhd_unit *dmaru)
 {
 	struct acpi_dmar_hardware_unit *drhd;
 	int ret = 0;
@@ -218,10 +227,10 @@ static int __init dmar_parse_dev(struct dmar_drhd_unit *dmaru)
 	return ret;
 }
 
-#ifdef CONFIG_ACPI_NUMA
-static int __init
+static int
 dmar_parse_one_rhsa(struct acpi_dmar_header *header)
 {
+#ifdef CONFIG_ACPI_NUMA
 	struct acpi_dmar_rhsa *rhsa;
 	struct dmar_drhd_unit *drhd;
 
@@ -245,12 +254,11 @@ dmar_parse_one_rhsa(struct acpi_dmar_header *header)
 		dmi_get_system_info(DMI_BIOS_VERSION),
 		dmi_get_system_info(DMI_PRODUCT_VERSION));
 
+#endif
 	return 0;
 }
-#endif
 
-static void __init
-dmar_table_print_dmar_entry(struct acpi_dmar_header *header)
+static void dmar_table_print_dmar_entry(struct acpi_dmar_header *header)
 {
 	struct acpi_dmar_hardware_unit *drhd;
 	struct acpi_dmar_reserved_memory *rmrr;
@@ -364,9 +372,7 @@ parse_dmar_table(void)
 			ret = dmar_parse_one_atsr(entry_header);
 			break;
 		case ACPI_DMAR_HARDWARE_AFFINITY:
-#ifdef CONFIG_ACPI_NUMA
 			ret = dmar_parse_one_rhsa(entry_header);
-#endif
 			break;
 		default:
 			printk(KERN_WARNING PREFIX
@@ -545,6 +551,23 @@ void restore_dev_to_drhd(struct pci_dev *dev)
 		drhd->devices[i] = pci_dev_get(dev);
 }
 
+static void remove_dmaru_from_saved_list(void *drhd,
+				  struct list_head *lh)
+{
+	struct dev_dmaru *m, *n;
+
+	list_for_each_entry_safe(m, n, lh, list) {
+		if (m->dmaru == drhd) {
+			list_del(&m->list);
+			kfree(m);
+		}
+	}
+}
+static void remove_dmaru_from_saved_dev_drhd_list(void *drhd)
+{
+	remove_dmaru_from_saved_list(drhd, &saved_dev_drhd_list);
+}
+
 #ifdef CONFIG_INTEL_IOMMU
 static LIST_HEAD(saved_dev_atsr_list);
 void remove_dev_from_atsr(struct pci_dev *dev)
@@ -581,6 +604,14 @@ void restore_dev_to_atsr(struct pci_dev *dev)
 	if (atsr)
 		atsr->devices[i] = pci_dev_get(dev);
 }
+static void remove_atsru_from_saved_dev_atsru_list(void *atsr)
+{
+	remove_dmaru_from_saved_list(atsr, &saved_dev_atsr_list);
+}
+#else
+static void remove_atsru_from_saved_dev_atsru_list(struct dmar_atsr_unit *atsr)
+{
+}
 #endif  /* CONFIG_INTEL_IOMMU */
 
 #endif  /* CONFIG_HOTPLUG */
diff --git a/include/linux/dmar.h b/include/linux/dmar.h
index 305db55..79c3898 100644
--- a/include/linux/dmar.h
+++ b/include/linux/dmar.h
@@ -143,6 +143,15 @@ static inline void remove_dev_from_atsr(struct pci_dev *dev) { }
 static inline void restore_dev_to_atsr(struct pci_dev *dev) { }
 #endif
 
+struct dmar_atsr_unit {
+	struct list_head list;		/* list of ATSR units */
+	struct acpi_dmar_header *hdr;	/* ACPI header */
+	struct pci_dev **devices;	/* target devices */
+	int devices_cnt;		/* target device count */
+	u16 segment;			/* PCI domain */
+	u8 include_all:1;		/* include all ports */
+};
+
 #ifdef CONFIG_INTEL_IOMMU
 extern int iommu_detected, no_iommu;
 extern struct list_head dmar_rmrr_units;
@@ -159,24 +168,19 @@ struct dmar_rmrr_unit {
 	list_for_each_entry(rmrr, &dmar_rmrr_units, list)
 
 extern struct list_head dmar_atsr_units;
-struct dmar_atsr_unit {
-	struct list_head list;		/* list of ATSR units */
-	struct acpi_dmar_header *hdr;	/* ACPI header */
-	struct pci_dev **devices;	/* target devices */
-	int devices_cnt;		/* target device count */
-	u16 segment;			/* PCI domain */
-	u8 include_all:1;		/* include all ports */
-};
-
 #define for_each_atsr_unit(atsr) \
 	list_for_each_entry(atsr, &dmar_atsr_units, list)
 
 int dmar_parse_rmrr_atsr_dev(void);
 extern int dmar_parse_one_rmrr(struct acpi_dmar_header *header);
 extern int dmar_parse_one_atsr(struct acpi_dmar_header *header);
+int __dmar_parse_one_atsr(struct acpi_dmar_header *header,
+			  struct dmar_atsr_unit **patsru);
+int atsr_parse_dev(struct dmar_atsr_unit *atsru);
 extern int dmar_parse_dev_scope(void *start, void *end, int *cnt,
 				struct pci_dev ***devices, u16 segment);
 extern int intel_iommu_init(void);
+int init_dmar_one(struct dmar_drhd_unit *drhd);
 #else /* !CONFIG_INTEL_IOMMU: */
 static inline int intel_iommu_init(void) { return -ENODEV; }
 static inline int dmar_parse_one_rmrr(struct acpi_dmar_header *header)
@@ -187,10 +191,34 @@ static inline int dmar_parse_one_atsr(struct acpi_dmar_header *header)
 {
 	return 0;
 }
+static inline int __dmar_parse_one_atsr(struct acpi_dmar_header *header,
+			  struct dmar_atsr_unit **patsru)
+{
+	return 0;
+}
+static inline int atsr_parse_dev(struct dmar_atsr_unit *atsru)
+{
+	return 0;
+}
 static inline int dmar_parse_rmrr_atsr_dev(void)
 {
 	return 0;
 }
+static inline int init_dmar_one(struct dmar_drhd_unit *drhd)
+{
+	return 0;
+}
 #endif /* CONFIG_INTEL_IOMMU */
 
+#ifdef CONFIG_IRQ_REMAP
+void disable_irq_remapping_one(struct dmar_drhd_unit *drhd);
+int intel_enable_irq_remapping_one(struct dmar_drhd_unit *drhd);
+#else
+static inline void disable_irq_remapping_one(struct dmar_drhd_unit *drhd) { }
+static inline int intel_enable_irq_remapping_one(struct dmar_drhd_unit *drhd)
+{
+	return 0;
+}
+#endif
+
 #endif /* __DMAR_H__ */
-- 
1.7.7

[RFC PATCH 14/14] IOMMU: Add intel iommu irq-remapping and dmar hotplug support

[Yinghai Lu]

Will use acpi_pci_driver to parse and init them between pci root bus scanning
and start.

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
---
 drivers/iommu/dmar.c |  218 ++++++++++++++++++++++++++++++++++++++++++++++++++
 1 files changed, 218 insertions(+), 0 deletions(-)

diff --git a/drivers/iommu/dmar.c b/drivers/iommu/dmar.c
index f96b8ce..918a063 100644
--- a/drivers/iommu/dmar.c
+++ b/drivers/iommu/dmar.c
@@ -36,6 +36,7 @@
 #include <linux/tboot.h>
 #include <linux/dmi.h>
 #include <linux/slab.h>
+#include <acpi/acpi.h>
 #include <asm/irq_remapping.h>
 #include <asm/iommu_table.h>
 
@@ -1480,3 +1481,220 @@ int __init dmar_ir_support(void)
 	return dmar->flags & 0x1;
 }
 IOMMU_INIT_POST(detect_intel_iommu);
+
+#ifdef CONFIG_HOTPLUG
+static u8 dmar_uuid_str[] = "D8C1A3A6-BE9B-4C9B-91BF-C3CB81FC5DAF";
+
+static int acpi_dmar_dsm_support(acpi_handle handle)
+{
+	int support = 0;
+	struct acpi_dsm_context context = {
+		.uuid_str = dmar_uuid_str,
+		.rev = 1,
+		.func_idx = 0,
+	};
+
+	if (ACPI_SUCCESS(acpi_run_dsm(handle, &context))) {
+		u32 *ret = context.ret.pointer;
+		support = ret[0];
+		kfree(context.ret.pointer);
+	}
+
+	if (support & 1)
+		return support & 0x0e;
+
+	return 0;
+}
+
+static void *acpi_dmar_dsm_run(acpi_handle handle, int idx)
+{
+	struct acpi_dsm_context context = {
+		.uuid_str = dmar_uuid_str,
+		.rev = 1,
+		.func_idx = idx,
+	};
+
+	if (ACPI_SUCCESS(acpi_run_dsm(handle, &context)))
+		return context.ret.pointer;
+
+	return NULL;
+}
+
+static void handle_iommu_add(acpi_handle handle, void **d, void **a)
+{
+	int support;
+	struct acpi_dmar_header *header;
+	struct dmar_drhd_unit *dmaru = NULL;
+	struct dmar_atsr_unit *atsru = NULL;
+
+	*d = NULL;
+	*a = NULL;
+
+	support = acpi_dmar_dsm_support(handle);
+
+	if (!support)
+		return;
+
+	/* DRHD */
+	if (support & (1<<1)) {
+		header = acpi_dmar_dsm_run(handle, 1);
+		dmar_table_print_dmar_entry(header);
+		__dmar_parse_one_drhd(header, &dmaru);
+	}
+
+	if (!dmaru)
+		return;
+
+	/* ATSR */
+	if (support & (1<<2)) {
+		header = acpi_dmar_dsm_run(handle, 2);
+		dmar_table_print_dmar_entry(header);
+		__dmar_parse_one_atsr(header, &atsru);
+	}
+
+	/* RHSA */
+	if (support & (1<<3)) {
+		header = acpi_dmar_dsm_run(handle, 3);
+		dmar_table_print_dmar_entry(header);
+		dmar_parse_one_rhsa(header);
+		kfree(header);
+	}
+
+	/*
+	 *  only need to init intr_remap and dmar for hot-add ones
+	 *  after enable_IR() or pci_iommu_init
+	 */
+	/*
+	 * TODO: handle parsing failure for pre-installed hotplug one
+	 *	Could make every parse_one duplicate the entry table?
+	 */
+	if (irq_remapping_enabled == 1)
+		intel_enable_irq_remapping_one(dmaru);
+	if (irq_remapping_enabled == 1 || intel_iommu_enabled == 1)
+		dmar_parse_dev(dmaru);
+	if (intel_iommu_enabled == 1) {
+		init_dmar_one(dmaru);
+		if (atsru) {
+			header = atsru->hdr;
+			if (atsr_parse_dev(atsru)) {
+				kfree(header);
+				atsru = NULL;
+			}
+			*a = atsru;
+		}
+	}
+	*d = dmaru;
+}
+
+static void handle_iommu_remove(void *drhd, void *atsr)
+{
+	struct dmar_drhd_unit *dmaru = drhd;
+	struct dmar_atsr_unit *atsru = atsr;
+
+	if (!dmaru)
+		return;
+
+	if (irq_remapping_enabled == 1)
+		disable_irq_remapping_one(dmaru);
+	if (intel_iommu_enabled == 1) {
+		free_dmar_iommu(dmaru->iommu);
+		if (atsru) {
+			kfree(atsru->devices);
+			remove_atsru_from_saved_dev_atsru_list(atsru);
+		}
+	}
+	if (irq_remapping_enabled == 1 || intel_iommu_enabled == 1) {
+		kfree(dmaru->devices);
+		remove_dmaru_from_saved_dev_drhd_list(dmaru);
+	}
+	if (atsru) {
+		kfree(atsru->hdr);
+		list_del(&atsru->list);
+		kfree(atsru);
+	}
+	free_iommu(dmaru->iommu);
+	kfree(dmaru->hdr);
+	list_del(&dmaru->list);
+	kfree(dmaru);
+}
+
+struct acpi_pci_iommu {
+	acpi_handle root_handle;
+	void *dmaru;
+	void *atsru;
+	struct list_head list;
+};
+
+static LIST_HEAD(iommu_list);
+static DEFINE_MUTEX(iommu_list_lock);
+
+static acpi_status register_iommu(acpi_handle handle, u32 lvl,
+				  void *context, void **rv)
+{
+	acpi_handle root_handle = context;
+	void *dmaru, *atsru;
+	struct acpi_pci_iommu *iommu;
+
+	handle_iommu_add(handle, &dmaru, &atsru);
+	if (!dmaru)
+		return AE_OK;
+
+	iommu = kzalloc(sizeof(*iommu), GFP_KERNEL);
+	if (!iommu) {
+		printk(KERN_ERR "%s: cannot allocate memory\n", __func__);
+		handle_iommu_remove(dmaru, atsru);
+		return AE_OK;
+	}
+	iommu->root_handle = root_handle;
+	iommu->dmaru = dmaru;
+	iommu->atsru = atsru;
+
+	mutex_lock(&iommu_list_lock);
+	list_add(&iommu->list, &iommu_list);
+	mutex_unlock(&iommu_list_lock);
+
+	return AE_OK;
+}
+
+static int acpi_pci_iommu_add(acpi_handle handle)
+{
+	acpi_status status;
+
+	status = acpi_walk_namespace(ACPI_TYPE_DEVICE, handle, (u32)1,
+				     register_iommu, NULL, handle, NULL);
+	if (ACPI_FAILURE(status))
+		printk(KERN_ERR "%s: register_iommu failure - %d", __func__,
+			status);
+
+	return status;
+}
+
+static void acpi_pci_iommu_remove(acpi_handle handle)
+{
+	struct acpi_pci_iommu *iommu, *tmp;
+
+	mutex_lock(&iommu_list_lock);
+	list_for_each_entry_safe(iommu, tmp, &iommu_list, list) {
+		if (handle != iommu->root_handle)
+			continue;
+		list_del(&iommu->list);
+		handle_iommu_remove(iommu->dmaru, iommu->atsru);
+		kfree(iommu);
+	}
+	mutex_unlock(&iommu_list_lock);
+}
+
+static struct acpi_pci_driver acpi_pci_iommu_driver = {
+	.add = acpi_pci_iommu_add,
+	.remove = acpi_pci_iommu_remove,
+};
+
+static int __init acpi_pci_iommu_init(void)
+{
+	acpi_pci_register_driver(&acpi_pci_iommu_driver);
+
+	return 0;
+}
+
+subsys_initcall(acpi_pci_iommu_init);
+#endif
-- 
1.7.7