iommu: Add event tracing feature to iommu

iommu: Add event tracing feature to iommu

This patch set adds tracing feature to iommu driver to report various iommu events. Classes iommu_group, iommu_device, iommu_map_unmap, and iommu_amd_event are defined.

iommu_group class events can be enabled to trigger when devices get added to and removed from an iommu group. Trace information includes iommu group id and device name.

iommu:add_device_to_group
iommu:remove_device_from_group

iommu_device class events can be enabled to trigger when devices are attached to and detached from a domain. Trace information includes device name.

iommu:attach_device_to_domain
iommu:detach_device_from_domain

iommu_map_unmap class events can be enabled to trigger when iommu map and unmap iommu ops complete. Trace information includes iova, physical address (map event only), and size.

iommu:unmap
iommu:map

iommu_amd_event class can be enabled to trigger when AMD IOMMU driver finds events when it polls the IOMMU Event Log. Trace information includes the event string derived from the event type in human friendly form, event type, id of the domain the device is placed in, device (bus number, slot number, and function number), address associated with the event, and flags.

iommu:amd_event

In addition to defining the new traces, iommu and amd_iommu drivers are changed to call trace events.

Testing:
The following is trace is generated when intel-iommu driver adds devices to to iommu groups during boot-time during its initialization:

#           TASK-PID   CPU#  ||||    TIMESTAMP  FUNCTION
#              | |       |   ||||       |         |
       swapper/0-1     [003] ....     1.854793: add_device_to_group: IOMMU: groupID=0 device=0000:00:00.0
       swapper/0-1     [003] ....     1.854797: add_device_to_group: IOMMU: groupID=1 device=0000:00:02.0
       swapper/0-1     [003] ....     1.854802: add_device_to_group: IOMMU: groupID=2 device=0000:00:16.0
       swapper/0-1     [003] ....     1.854803: add_device_to_group: IOMMU: groupID=2 device=0000:00:16.3
       swapper/0-1     [003] ....     1.854805: add_device_to_group: IOMMU: groupID=3 device=0000:00:19.0
       swapper/0-1     [003] ....     1.854807: add_device_to_group: IOMMU: groupID=4 device=0000:00:1a.0
       swapper/0-1     [003] ....     1.854810: add_device_to_group: IOMMU: groupID=5 device=0000:00:1b.0
       swapper/0-1     [003] ....     1.854817: add_device_to_group: IOMMU: groupID=6 device=0000:00:1c.0
       swapper/0-1     [003] ....     1.854821: add_device_to_group: IOMMU: groupID=6 device=0000:00:1c.2
       swapper/0-1     [003] ....     1.854823: add_device_to_group: IOMMU: groupID=7 device=0000:00:1d.0
       swapper/0-1     [003] ....     1.854825: add_device_to_group: IOMMU: groupID=8 device=0000:00:1e.0
       swapper/0-1     [003] ....     1.854828: add_device_to_group: IOMMU: groupID=9 device=0000:00:1f.0
       swapper/0-1     [003] ....     1.854830: add_device_to_group: IOMMU: groupID=9 device=0000:00:1f.2
       swapper/0-1     [003] ....     1.854831: add_device_to_group: IOMMU: groupID=9 device=0000:00:1f.3

Added trace calls to iommu_prepare_identity_map() for testing some of the conditions that are hard to trigger. Here is the trace from the testing:

#           TASK-PID   CPU#  ||||    TIMESTAMP  FUNCTION
#              | |       |   ||||       |         |
       swapper/0-1     [003] ....     1.854099: amd_event: IOMMU:Event 0x01 00:00.0 d=0x0010 a=0x00000000cb800000 f=0x0002
       swapper/0-1     [003] ....     1.854101: remove_device_from_group: IOMMU: groupID=0 device=0000:00:02.0
       swapper/0-1     [003] ....     1.854102: attach_device_to_domain: IOMMU: device=0000:00:02.0
       swapper/0-1     [003] ....     1.854102: detach_device_from_domain: IOMMU: device=0000:00:02.0
       swapper/0-1     [003] ....     1.854102: map: IOMMU: iova=0x00000000cb800000 paddr=0x00000000cf9fffff size=0x400
       swapper/0-1     [003] ....     1.854102: unmap: IOMMU: iova=0x00000000cb800000 size=0x400

Shuah Khan (4):
iommu: Add event tracing feature to iommu - Add iommu trace events
iommu: Add event tracing feature to iommu - export trace events
iommu: Add event tracing feature to iommu - amd_iommu driver trace
events
iommu: Add event tracing feature to iommu - iommu driver trace events

 drivers/iommu/Makefile       |    1 +
 drivers/iommu/amd_iommu.c    |   19 +++++
 drivers/iommu/iommu-traces.c |   27 +++++++
 drivers/iommu/iommu.c        |   17 ++++-
 include/trace/events/iommu.h |  174 ++++++++++++++++++++++++++++++++++++++++++
 5 files changed, 237 insertions(+), 1 deletion(-)
 create mode 100644 drivers/iommu/iommu-traces.c
 create mode 100644 include/trace/events/iommu.h

1.7.10.4

Add tracing feature to iommu driver to report various iommu events. Classes iommu_group, iommu_device, iommu_map_unmap, and iommu_amd_event are defined.

iommu_group class events can be enabled to trigger when devices get added to and removed from an iommu group. Trace information includes iommu group id and device name.

iommu:add_device_to_group
iommu:remove_device_from_group

iommu_device class events can be enabled to trigger when devices are attached to and detached from a domain. Trace information includes device name.

iommu:attach_device_to_domain
iommu:detach_device_from_domain

iommu_map_unmap class events can be enabled to trigger when iommu map and unmap iommu ops. Trace information includes iova, physical address (map event only), and size.

iommu:unmap
iommu:map

iommu_amd_event class can be enabled to trigger when AMD IOMMU driver finds events when it polls the IOMMU Event Log. Trace information includes the event string derived from the event type in human friendly form, event type, id of the domain the device is placed in, device (bus number, slot number, and function number), address associated with the event, and flags.

iommu:amd_event

---
 include/trace/events/iommu.h |  174 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 174 insertions(+)
 create mode 100644 include/trace/events/iommu.h

new file mode 100644
index 0000000..3d8d5e8
@@ -0,0 +1,174 @@
+/*
+ * iommu trace points
+ *
+ * Copyright (C) 2013 Shuah Khan <[email protected]>
+ *
+ */
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM iommu
+
+#if !defined(_TRACE_IOMMU_H) || defined(TRACE_HEADER_MULTI_READ)
+#define _TRACE_IOMMU_H
+
+#include <linux/tracepoint.h>
+#include <linux/pci.h>
+
+struct device;
+
+DECLARE_EVENT_CLASS(iommu_group_event,
+
+	TP_PROTO(int group_id, struct device *dev),
+
+	TP_ARGS(group_id, dev),
+
+	TP_STRUCT__entry(
+		__field(int, gid)
+		__string(device, dev_name(dev))
+	),
+
+	TP_fast_assign(
+		__entry->gid = group_id;
+		__assign_str(device, dev_name(dev));
+	),
+
+	TP_printk("IOMMU: groupID=%d device=%s",
+			__entry->gid, __get_str(device)
+	)
+);
+
+DEFINE_EVENT(iommu_group_event, add_device_to_group,
+
+	TP_PROTO(int group_id, struct device *dev),
+
+	TP_ARGS(group_id, dev)
+
+);
+
+DEFINE_EVENT(iommu_group_event, remove_device_from_group,
+
+	TP_PROTO(int group_id, struct device *dev),
+
+	TP_ARGS(group_id, dev)
+);
+
+DECLARE_EVENT_CLASS(iommu_device_event,
+
+	TP_PROTO(struct device *dev),
+
+	TP_ARGS(dev),
+
+	TP_STRUCT__entry(
+		__string(device, dev_name(dev))
+	),
+
+	TP_fast_assign(
+		__assign_str(device, dev_name(dev));
+	),
+
+	TP_printk("IOMMU: device=%s", __get_str(device)
+	)
+);
+
+DEFINE_EVENT(iommu_device_event, attach_device_to_domain,
+
+	TP_PROTO(struct device *dev),
+
+	TP_ARGS(dev)
+);
+
+DEFINE_EVENT(iommu_device_event, detach_device_from_domain,
+
+	TP_PROTO(struct device *dev),
+
+	TP_ARGS(dev)
+);
+
+DECLARE_EVENT_CLASS(iommu_map_unmap,
+
+	TP_PROTO(unsigned long iova, phys_addr_t paddr, size_t size),
+
+	TP_ARGS(iova, paddr, size),
+
+	TP_STRUCT__entry(
+		__field(u64, iova)
+		__field(u64, paddr)
+		__field(int, size)
+	),
+
+	TP_fast_assign(
+		__entry->iova = iova;
+		__entry->paddr = paddr;
+		__entry->size = size;
+	),
+
+	TP_printk("IOMMU: iova=0x%016llx paddr=0x%016llx size=0x%x",
+			__entry->iova, __entry->paddr, __entry->size
+	)
+);
+
+DEFINE_EVENT(iommu_map_unmap, map,
+
+	TP_PROTO(unsigned long iova, phys_addr_t paddr, size_t size),
+
+	TP_ARGS(iova, paddr, size)
+);
+
+DEFINE_EVENT_PRINT(iommu_map_unmap, unmap,
+
+	TP_PROTO(unsigned long iova, phys_addr_t paddr, size_t size),
+
+	TP_ARGS(iova, paddr, size),
+
+	TP_printk("IOMMU: iova=0x%016llx size=0x%x",
+			__entry->iova, __entry->size
+	)
+);
+
+DECLARE_EVENT_CLASS(iommu_amd_event,
+
+	TP_PROTO(char *event_str, int type, int domain_id, int device_id,
+		 u64 address, int flags),
+
+	TP_ARGS(event_str, type, domain_id, device_id, address, flags),
+
+	TP_STRUCT__entry(
+		__string(event_str, event_str)
+		__field(int, type)
+		__field(int, domain_id)
+		__field(int, device_id)
+		__field(u64, address)
+		__field(int, flags)
+	),
+
+	TP_fast_assign(
+		__assign_str(event_str, event_str);
+		__entry->type = type;
+		__entry->domain_id = domain_id;
+		__entry->device_id = device_id;
+		__entry->address = address;
+		__entry->flags = flags;
+	),
+
+	TP_printk("IOMMU:%s 0x%02x %02x:%02x.%x d=0x%04x a=0x%016llx f=0x%04x",
+			__get_str(event_str),
+			__entry->type,
+			PCI_BUS_NUM(__entry->device_id),
+			PCI_SLOT(__entry->device_id),
+			PCI_FUNC(__entry->device_id),
+			__entry->domain_id,
+			__entry->address, __entry->flags
+	)
+);
+
+DEFINE_EVENT(iommu_amd_event, amd_event,
+
+	TP_PROTO(char *event_str, int type, int domain_id, int device_id,
+		 u64 address, int flags),
+
+	TP_ARGS(event_str, type, domain_id, device_id, address, flags)
+);
+
+#endif /* _TRACE_IOMMU_H */
+
+/* This part must be outside protection */
+#include <trace/define_trace.h>

1.7.10.4

Add tracing feature to iommu driver to report various iommu events. Classes iommu_group, iommu_device, iommu_map_unmap, and iommu_amd_event are defined.

iommu_group class events can be enabled to trigger when devices get added to and removed from an iommu group. Trace information includes iommu group id and device name.

iommu:add_device_to_group
iommu:remove_device_from_group

iommu_device class events can be enabled to trigger when devices are attached to and detached from a domain. Trace information includes device name.

iommu:attach_device_to_domain
iommu:detach_device_from_domain

iommu_map_unmap class events can be enabled to trigger when iommu map and unmap iommu ops complete. Trace information includes iova, physical address (map event only), and size.

iommu:unmap
iommu:map

iommu_amd_event class can be enabled to trigger when AMD IOMMU driver finds events when it polls the IOMMU Event Log. Trace information includes the event string derived from the event type in human friendly form, event type, id of the domain the device is placed in, device (bus number, slot number, and function number), address associated with the event, and flags.

iommu:amd_event

---
 drivers/iommu/Makefile       |    1 +
 drivers/iommu/iommu-traces.c |   27 +++++++++++++++++++++++++++
 2 files changed, 28 insertions(+)
 create mode 100644 drivers/iommu/iommu-traces.c

index bbe7041..9015c45 100644
@@ -1,4 +1,5 @@
 obj-$(CONFIG_IOMMU_API) += iommu.o
+obj-$(CONFIG_IOMMU_API) += iommu-traces.o
 obj-$(CONFIG_OF_IOMMU)	+= of_iommu.o
 obj-$(CONFIG_MSM_IOMMU) += msm_iommu.o msm_iommu_dev.o
 obj-$(CONFIG_AMD_IOMMU) += amd_iommu.o amd_iommu_init.o

new file mode 100644
index 0000000..5b92e4c
@@ -0,0 +1,27 @@
+/*
+ * iommu trace points
+ *
+ * Copyright (C) 2013 Shuah Khan <[email protected]>
+ *
+ */
+
+#include <linux/string.h>
+#include <linux/types.h>
+
+#define CREATE_TRACE_POINTS
+#include <trace/events/iommu.h>
+
+/* iommu_group_event */
+EXPORT_TRACEPOINT_SYMBOL_GPL(add_device_to_group);
+EXPORT_TRACEPOINT_SYMBOL_GPL(remove_device_from_group);
+
+/* iommu_device_event */
+EXPORT_TRACEPOINT_SYMBOL_GPL(attach_device_to_domain);
+EXPORT_TRACEPOINT_SYMBOL_GPL(detach_device_from_domain);
+
+/* iommu_map_unmap */
+EXPORT_TRACEPOINT_SYMBOL_GPL(map);
+EXPORT_TRACEPOINT_SYMBOL_GPL(unmap);
+
+/* iommu_amd_event */
+EXPORT_TRACEPOINT_SYMBOL_GPL(amd_event);

1.7.10.4

Add tracing feature to iommu driver to report various iommu events. Classes iommu_group, iommu_device, iommu_map_unmap, and iommu_amd_event are defined.

iommu_group class events can be enabled to trigger when devices get added to and removed from an iommu group. Trace information includes iommu group id and device name.

iommu:add_device_to_group
iommu:remove_device_from_group

iommu_device class events can be enabled to trigger when devices are attached to and detached from a domain. Trace information includes device name.

iommu:attach_device_to_domain
iommu:detach_device_from_domain

iommu_map_unmap class events can be enabled to trigger when iommu map and unmap iommu ops complete. Trace information includes iova, physical address (map event only), and size.

iommu:unmap
iommu:map

iommu_amd_event class can be enabled to trigger when AMD IOMMU driver finds events when it polls the IOMMU Event Log. Trace information includes the event string derived from the event type in human friendly form, event type, id of the domain the device is placed in, device (bus number, slot number, and function number), address associated with the event, and flags.

iommu:amd_event

---
 drivers/iommu/amd_iommu.c |   19 +++++++++++++++++++
 1 file changed, 19 insertions(+)

index 6dc6594..10ac042 100644
@@ -42,6 +42,7 @@
 #include <asm/iommu.h>
 #include <asm/gart.h>
 #include <asm/dma.h>
+#include <trace/events/iommu.h>
 
 #include "amd_iommu_proto.h"
 #include "amd_iommu_types.h"

@@ -659,6 +660,8 @@ retry:
 		       "address=0x%016llx flags=0x%04x]\n",
 		       PCI_BUS_NUM(devid), PCI_SLOT(devid), PCI_FUNC(devid),
 		       address, flags);
+		trace_amd_event("ILLEGAL_DEV_TABLE_ENTRY", type, domid,
+				      devid, address, flags);
 		dump_dte_entry(devid);
 		break;
 	case EVENT_TYPE_IO_FAULT:

@@ -666,41 +669,57 @@ retry:
 		       "domain=0x%04x address=0x%016llx flags=0x%04x]\n",
 		       PCI_BUS_NUM(devid), PCI_SLOT(devid), PCI_FUNC(devid),
 		       domid, address, flags);
+		trace_amd_event("IO_PAGE_FAULT", type, domid,
+				      devid, address, flags);
 		break;
 	case EVENT_TYPE_DEV_TAB_ERR:
 		printk("DEV_TAB_HARDWARE_ERROR device=%02x:%02x.%x "
 		       "address=0x%016llx flags=0x%04x]\n",
 		       PCI_BUS_NUM(devid), PCI_SLOT(devid), PCI_FUNC(devid),
 		       address, flags);
+		trace_amd_event("DEV_TAB_HARDWARE_ERROR", type, domid,
+				      devid, address, flags);
 		break;
 	case EVENT_TYPE_PAGE_TAB_ERR:
 		printk("PAGE_TAB_HARDWARE_ERROR device=%02x:%02x.%x "
 		       "domain=0x%04x address=0x%016llx flags=0x%04x]\n",
 		       PCI_BUS_NUM(devid), PCI_SLOT(devid), PCI_FUNC(devid),
 		       domid, address, flags);
+		trace_amd_event("PAGE_TAB_HARDWARE_ERROR", type, domid,
+				      devid, address, flags);
 		break;
 	case EVENT_TYPE_ILL_CMD:
 		printk("ILLEGAL_COMMAND_ERROR address=0x%016llx]\n", address);
+		trace_amd_event("ILLEGAL_COMMAND_ERROR", type, domid,
+				      devid, address, flags);
 		dump_command(address);
 		break;
 	case EVENT_TYPE_CMD_HARD_ERR:
 		printk("COMMAND_HARDWARE_ERROR address=0x%016llx "
 		       "flags=0x%04x]\n", address, flags);
+		trace_amd_event("COMMAND_HARDWARE_ERROR", type, domid,
+				      devid, address, flags);
 		break;
 	case EVENT_TYPE_IOTLB_INV_TO:
 		printk("IOTLB_INV_TIMEOUT device=%02x:%02x.%x "
 		       "address=0x%016llx]\n",
 		       PCI_BUS_NUM(devid), PCI_SLOT(devid), PCI_FUNC(devid),
 		       address);
+		trace_amd_event("IOTLB_INV_TIMEOUT", type, domid,
+				      devid, address, flags);
 		break;
 	case EVENT_TYPE_INV_DEV_REQ:
 		printk("INVALID_DEVICE_REQUEST device=%02x:%02x.%x "
 		       "address=0x%016llx flags=0x%04x]\n",
 		       PCI_BUS_NUM(devid), PCI_SLOT(devid), PCI_FUNC(devid),
 		       address, flags);
+		trace_amd_event("INVALID_DEVICE_REQUEST", type, domid,
+				      devid, address, flags);
 		break;
 	default:
 		printk(KERN_ERR "UNKNOWN type=0x%02x]\n", type);
+		trace_amd_event("UNKNOWN", type, domid,
+				      devid, address, flags);
 	}
 
 	memset(__evt, 0, 4 * sizeof(u32));

1.7.10.4

Add tracing feature to iommu driver to report various iommu events. Classes iommu_group, iommu_device, iommu_map_unmap, and iommu_amd_event are defined.

iommu_group class events can be enabled to trigger when devices get added to and removed from an iommu group. Trace information includes iommu group id and device name.

iommu:add_device_to_group
iommu:remove_device_from_group

iommu_device class events can be enabled to trigger when devices are attached to and detached from a domain. Trace information includes device name.

iommu:attach_device_to_domain
iommu:detach_device_from_domain

iommu_map_unmap class events can be enabled to trigger when iommu map and unmap iommu ops complete. Trace information includes iova, physical address (map event only), and size.

iommu:unmap
iommu:map

iommu_amd_event class can be enabled to trigger when AMD IOMMU driver finds events when it polls the IOMMU Event Log. Trace information includes the event string derived from the event type in human friendly form, event type, id of the domain the device is placed in, device (bus number, slot number, and function number), address associated with the event, and flags.

iommu:amd_event

---
 drivers/iommu/iommu.c |   17 ++++++++++++++++-
 1 file changed, 16 insertions(+), 1 deletion(-)

index fbe9ca7..8e77110 100644
@@ -29,6 +29,7 @@
 #include <linux/idr.h>
 #include <linux/notifier.h>
 #include <linux/err.h>
+#include <trace/events/iommu.h>
 
 static struct kset *iommu_group_kset;
 static struct ida iommu_group_ida;

@@ -363,6 +364,8 @@ rename:
 	/* Notify any listeners about change to group. */
 	blocking_notifier_call_chain(&group->notifier,
 				     IOMMU_GROUP_NOTIFY_ADD_DEVICE, dev);
+
+	trace_add_device_to_group(group->id, dev);
 	return 0;
 }
 EXPORT_SYMBOL_GPL(iommu_group_add_device);

@@ -399,6 +402,8 @@ void iommu_group_remove_device(struct device *dev)
 	sysfs_remove_link(group->devices_kobj, device->name);
 	sysfs_remove_link(&dev->kobj, "iommu_group");
 
+	trace_remove_device_from_group(group->id, dev);
+
 	kfree(device->name);
 	kfree(device);
 	dev->iommu_group = NULL;

@@ -680,10 +685,16 @@ EXPORT_SYMBOL_GPL(iommu_domain_free);
 
 int iommu_attach_device(struct iommu_domain *domain, struct device *dev)
 {
+	int ret;
 	if (unlikely(domain->ops->attach_dev == NULL))
 		return -ENODEV;
 
-	return domain->ops->attach_dev(domain, dev);
+	ret = domain->ops->attach_dev(domain, dev);
+
+	if (!ret)
+		trace_attach_device_to_domain(dev);
+
+	return ret;
 }
 EXPORT_SYMBOL_GPL(iommu_attach_device);
 

@@ -693,6 +704,7 @@ void iommu_detach_device(struct iommu_domain *domain, struct device *dev)
 		return;
 
 	domain->ops->detach_dev(domain, dev);
+	trace_detach_device_from_domain(dev);
 }
 EXPORT_SYMBOL_GPL(iommu_detach_device);
 

@@ -832,6 +844,8 @@ int iommu_map(struct iommu_domain *domain, unsigned long iova,
 	/* unroll mapping in case something went wrong */
 	if (ret)
 		iommu_unmap(domain, orig_iova, orig_size - size);
+	else
+		trace_map(iova, paddr, size);
 
 	return ret;
 }

@@ -880,6 +894,7 @@ size_t iommu_unmap(struct iommu_domain *domain, unsigned long iova, size_t size)
 		unmapped += unmapped_page;
 	}
 
+	trace_unmap(iova, 0, size);
 	return unmapped;
 }
 EXPORT_SYMBOL_GPL(iommu_unmap);

1.7.10.4

I don't know what the maintainer thinks, but usually new tracepoints like this are done as one patch. Not four.

-- Steve

If one single patch is preferred, I can easily generate one. Does the code look good otherwise?

-- Shuah

Shuah Khan, Linux Kernel Developer - Open Source Group Samsung Research America (Silicon Valley) [email protected] | (970) 672-0658