hw_breakpoint: Breakpoint modification fixes and new modify ioctl

PATCH In review Jiri Olsa 4 participants 24 messages Mar 12, 2018 – Mar 14, 2018 perf events

Share & tools

Jiri Olsa

author reviewer

2018-03-12 13:45 UTC #0 lore raw

v3 cover hw_breakpoint: Breakpoint modification fixes and new modify ioctl

hi,
Milind Chabbi introduced new ioctl interface to change live breakpoint [1]. It allows to change its bp_addr, bp_len and bp_type throught new ioctl for perf breakpoint event.

We already have a kernel interface for this via modify_user_hw_breakpoint function. This function however does not update the breakpoint slot counts.

So when the same functionality was exposed to user space (Milind's change), with simple test program I could put wrong slots count on arm server [2] and ended up with no breakpoints available on the system. Note it's not an issue on x86, because it shares slot single counter for both data and inst types (CONFIG_HAVE_MIXED_BREAKPOINTS_REGS).

This patchset contains my fixes for keeping breakpoint slots count updated. On top of it there's Milind's change for new _IOC_MODIFY_ATTRIBUTES ioctl interface to change a breakpoint.

As I mentioned above there're kernel users of modify_user_hw_breakpoint function, all ptrace related AFAICS, which could got broken.. so cc-ing Oleg ;-)

I ran gdb and strace tests suites and got same amount of skip/fail tests as when I run them on unpatched machine, so I assume nothing new got broken.

It's also available in here:
https://git.kernel.org/pub/scm/linux/kernel/git/jolsa/perf.git
perf/bp

v3 changes:
- rebased on Arnaldo's perf/core
- minor format changes in PERF_EVENT_IOC_* hunks

v2 changes:
- added check for the rest of the perf_event_attr fields to be the same as for kernel event

thanks,
jirka

[1] https://marc.info/?l=linux-kernel&m=151012255331565&w=2
[2] https://marc.info/?l=linux-man&m=151172469807302&w=2
---
Jiri Olsa (7):
hw_breakpoint: Pass bp_type directly as find_slot_idx argument
hw_breakpoint: Pass bp_type argument to __reserve_bp_slot|__release_bp_slot
hw_breakpoint: Add modify_bp_slot function
hw_breakpoint: Factor out __modify_user_hw_breakpoint function
hw_breakpoint: Add perf_event_attr fields check in __modify_user_hw_breakpoint perf/core: Move perf_event_attr::sample_max_stack into perf_copy_attr perf tests: Add breakpoint accounting/modify test

Milind Chabbi (1):
perf/core: fast breakpoint modification via _IOC_MODIFY_ATTRIBUTES.

 include/linux/hw_breakpoint.h         |   7 +++++
 include/uapi/linux/perf_event.h       |  23 ++++++++-------
 kernel/events/core.c                  |  54 ++++++++++++++++++++++++++++++++--
 kernel/events/hw_breakpoint.c         | 115 +++++++++++++++++++++++++++++++++++++++++++++++++++---------------------
 tools/include/uapi/linux/perf_event.h |  23 ++++++++-------
 tools/perf/tests/Build                |   1 +
 tools/perf/tests/bp_account.c         | 195 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 tools/perf/tests/builtin-test.c       |   4 +++
 tools/perf/tests/tests.h              |   1 +
 9 files changed, 365 insertions(+), 58 deletions(-)
 create mode 100644 tools/perf/tests/bp_account.c

Jiri Olsa

author reviewer

2018-03-12 13:45 UTC #1 lore raw

PATCH v3 1/8 hw_breakpoint: Pass bp_type directly as find_slot_idx argument

Passing bp_type directly as find_slot_idx argument, so we don't need to have whole event to get the breakpoint slot type. It will be used in following changes.

Signed-off-by Jiri Olsa <[email protected]>

---
 kernel/events/hw_breakpoint.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

kernel/events/hw_breakpoint.c +5 -5

index 3f8cb1e14588..395ca07965af 100644
@@ -85,9 +85,9 @@ __weak int hw_breakpoint_weight(struct perf_event *bp)
 	return 1;
 }
 
-static inline enum bp_type_idx find_slot_idx(struct perf_event *bp)
+static inline enum bp_type_idx find_slot_idx(u64 bp_type)
 {
-	if (bp->attr.bp_type & HW_BREAKPOINT_RW)
+	if (bp_type & HW_BREAKPOINT_RW)
 		return TYPE_DATA;
 
 	return TYPE_INST;

@@ -122,7 +122,7 @@ static int task_bp_pinned(int cpu, struct perf_event *bp, enum bp_type_idx type)
 
 	list_for_each_entry(iter, &bp_task_head, hw.bp_list) {
 		if (iter->hw.target == tsk &&
-		    find_slot_idx(iter) == type &&
+		    find_slot_idx(iter->attr.bp_type) == type &&
 		    (iter->cpu < 0 || cpu == iter->cpu))
 			count += hw_breakpoint_weight(iter);
 	}

@@ -292,7 +292,7 @@ static int __reserve_bp_slot(struct perf_event *bp)
 	    bp->attr.bp_type == HW_BREAKPOINT_INVALID)
 		return -EINVAL;
 
-	type = find_slot_idx(bp);
+	type = find_slot_idx(bp->attr.bp_type);
 	weight = hw_breakpoint_weight(bp);
 
 	fetch_bp_busy_slots(&slots, bp, type);

@@ -329,7 +329,7 @@ static void __release_bp_slot(struct perf_event *bp)
 	enum bp_type_idx type;
 	int weight;
 
-	type = find_slot_idx(bp);
+	type = find_slot_idx(bp->attr.bp_type);
 	weight = hw_breakpoint_weight(bp);
 	toggle_bp_slot(bp, false, type, weight);
 }

Signature · 1 line

2.13.6

tip-bot for Jiri Olsa

merge bot

2018-03-13 06:19 UTC #2 lore raw

[tip:perf/core] hw_breakpoint: Pass bp_type directly as find_slot_idx() argument

hw_breakpoint: Pass bp_type directly as find_slot_idx() argument

Commit-ID:  cbd9d9f114f9e5931ba199f4450667656b892f32
Gitweb:     https://git.kernel.org/tip/cbd9d9f114f9e5931ba199f4450667656b892f32
Author:     Jiri Olsa <[email protected]>
AuthorDate: Mon, 12 Mar 2018 14:45:41 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Tue, 13 Mar 2018 06:56:07 +0100

hw_breakpoint: Pass bp_type directly as find_slot_idx() argument

Pass bp_type directly as a find_slot_idx() argument, so we don't need to have whole event to get the breakpoint slot type. It will be used in following changes.

Signed-off-by Jiri Olsa <[email protected]>

Cc Alexander Shishkin <[email protected]>

Cc Arnaldo Carvalho de Melo <[email protected]>

Cc David Ahern <[email protected]>

Cc Frederic Weisbecker <[email protected]>

Cc Hari Bathini <[email protected]>

Cc Jin Yao <[email protected]>

Cc Jiri Olsa <[email protected]>

Cc Kan Liang <[email protected]>

Cc Linus Torvalds <[email protected]>

Cc Michael Ellerman <[email protected]>

Cc Milind Chabbi <[email protected]>

Cc Namhyung Kim <[email protected]>

Cc Oleg Nesterov <[email protected]>

Cc Peter Zijlstra <[email protected]>

Cc Sukadev Bhattiprolu <[email protected]>

Cc Thomas Gleixner <[email protected]>

Cc Will Deacon <[email protected]>

Link http://lkml.kernel.org/r/[email protected] open here

Signed-off-by Ingo Molnar <[email protected]>

---
 kernel/events/hw_breakpoint.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

kernel/events/hw_breakpoint.c +5 -5

index 3f8cb1e14588..395ca07965af 100644
@@ -85,9 +85,9 @@ __weak int hw_breakpoint_weight(struct perf_event *bp)
 	return 1;
 }
 
-static inline enum bp_type_idx find_slot_idx(struct perf_event *bp)
+static inline enum bp_type_idx find_slot_idx(u64 bp_type)
 {
-	if (bp->attr.bp_type & HW_BREAKPOINT_RW)
+	if (bp_type & HW_BREAKPOINT_RW)
 		return TYPE_DATA;
 
 	return TYPE_INST;

@@ -122,7 +122,7 @@ static int task_bp_pinned(int cpu, struct perf_event *bp, enum bp_type_idx type)
 
 	list_for_each_entry(iter, &bp_task_head, hw.bp_list) {
 		if (iter->hw.target == tsk &&
-		    find_slot_idx(iter) == type &&
+		    find_slot_idx(iter->attr.bp_type) == type &&
 		    (iter->cpu < 0 || cpu == iter->cpu))
 			count += hw_breakpoint_weight(iter);
 	}

@@ -292,7 +292,7 @@ static int __reserve_bp_slot(struct perf_event *bp)
 	    bp->attr.bp_type == HW_BREAKPOINT_INVALID)
 		return -EINVAL;
 
-	type = find_slot_idx(bp);
+	type = find_slot_idx(bp->attr.bp_type);
 	weight = hw_breakpoint_weight(bp);
 
 	fetch_bp_busy_slots(&slots, bp, type);

@@ -329,7 +329,7 @@ static void __release_bp_slot(struct perf_event *bp)
 	enum bp_type_idx type;
 	int weight;
 
-	type = find_slot_idx(bp);
+	type = find_slot_idx(bp->attr.bp_type);
 	weight = hw_breakpoint_weight(bp);
 	toggle_bp_slot(bp, false, type, weight);
 }

Jiri Olsa

author reviewer

2018-03-12 13:46 UTC #3 lore raw

PATCH v3 2/8 hw_breakpoint: Pass bp_type argument to __reserve_bp_slot|__release_bp_slot

In reply to Jiri Olsa

Passing bp_type argument to __reserve_bp_slot and __release_bp_slot functions, so we can pass another bp_type than the one defined in bp->attr.bp_type. This will be handy in following change that fixes breakpoint slot counts during its modification.

Signed-off-by Jiri Olsa <[email protected]>

---
 kernel/events/hw_breakpoint.c | 20 ++++++++++----------
 1 file changed, 10 insertions(+), 10 deletions(-)

kernel/events/hw_breakpoint.c +10 ~~-10~~

index 395ca07965af..9b31fbcc3305 100644
@@ -277,7 +277,7 @@ __weak void arch_unregister_hw_breakpoint(struct perf_event *bp)
  *       ((per_cpu(info->flexible, *) > 1) + max(per_cpu(info->cpu_pinned, *))
  *            + max(per_cpu(info->tsk_pinned, *))) < HBP_NUM
  */
-static int __reserve_bp_slot(struct perf_event *bp)
+static int __reserve_bp_slot(struct perf_event *bp, u64 bp_type)
 {
 	struct bp_busy_slots slots = {0};
 	enum bp_type_idx type;

@@ -288,11 +288,11 @@ static int __reserve_bp_slot(struct perf_event *bp)
 		return -ENOMEM;
 
 	/* Basic checks */
-	if (bp->attr.bp_type == HW_BREAKPOINT_EMPTY ||
-	    bp->attr.bp_type == HW_BREAKPOINT_INVALID)
+	if (bp_type == HW_BREAKPOINT_EMPTY ||
+	    bp_type == HW_BREAKPOINT_INVALID)
 		return -EINVAL;
 
-	type = find_slot_idx(bp->attr.bp_type);
+	type = find_slot_idx(bp_type);
 	weight = hw_breakpoint_weight(bp);
 
 	fetch_bp_busy_slots(&slots, bp, type);

@@ -317,19 +317,19 @@ int reserve_bp_slot(struct perf_event *bp)
 
 	mutex_lock(&nr_bp_mutex);
 
-	ret = __reserve_bp_slot(bp);
+	ret = __reserve_bp_slot(bp, bp->attr.bp_type);
 
 	mutex_unlock(&nr_bp_mutex);
 
 	return ret;
 }
 
-static void __release_bp_slot(struct perf_event *bp)
+static void __release_bp_slot(struct perf_event *bp, u64 bp_type)
 {
 	enum bp_type_idx type;
 	int weight;
 
-	type = find_slot_idx(bp->attr.bp_type);
+	type = find_slot_idx(bp_type);
 	weight = hw_breakpoint_weight(bp);
 	toggle_bp_slot(bp, false, type, weight);
 }

@@ -339,7 +339,7 @@ void release_bp_slot(struct perf_event *bp)
 	mutex_lock(&nr_bp_mutex);
 
 	arch_unregister_hw_breakpoint(bp);
-	__release_bp_slot(bp);
+	__release_bp_slot(bp, bp->attr.bp_type);
 
 	mutex_unlock(&nr_bp_mutex);
 }

@@ -354,7 +354,7 @@ int dbg_reserve_bp_slot(struct perf_event *bp)
 	if (mutex_is_locked(&nr_bp_mutex))
 		return -1;
 
-	return __reserve_bp_slot(bp);
+	return __reserve_bp_slot(bp, bp->attr.bp_type);
 }
 
 int dbg_release_bp_slot(struct perf_event *bp)

@@ -362,7 +362,7 @@ int dbg_release_bp_slot(struct perf_event *bp)
 	if (mutex_is_locked(&nr_bp_mutex))
 		return -1;
 
-	__release_bp_slot(bp);
+	__release_bp_slot(bp, bp->attr.bp_type);
 
 	return 0;
 }

Signature · 1 line

2.13.6

tip-bot for Jiri Olsa

merge bot

2018-03-13 06:20 UTC #4 lore raw

[tip:perf/core] hw_breakpoint: Pass bp_type argument to __reserve_bp_slot|__release_bp_slot()

hw_breakpoint: Pass bp_type argument to __reserve_bp_slot|__release_bp_slot()

Commit-ID:  1ad9ff7dea4c42bee43f98b7d7ce8037ee22e133
Gitweb:     https://git.kernel.org/tip/1ad9ff7dea4c42bee43f98b7d7ce8037ee22e133
Author:     Jiri Olsa <[email protected]>
AuthorDate: Mon, 12 Mar 2018 14:45:42 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Tue, 13 Mar 2018 06:56:07 +0100

hw_breakpoint: Pass bp_type argument to __reserve_bp_slot|__release_bp_slot()

Passing bp_type argument to __reserve_bp_slot() and __release_bp_slot() functions, so we can pass another bp_type than the one defined in bp->attr.bp_type. This will be handy in following change that fixes breakpoint slot counts during its modification.

Signed-off-by Jiri Olsa <[email protected]>

Cc Alexander Shishkin <[email protected]>

Cc Arnaldo Carvalho de Melo <[email protected]>

Cc David Ahern <[email protected]>

Cc Frederic Weisbecker <[email protected]>

Cc Hari Bathini <[email protected]>

Cc Jin Yao <[email protected]>

Cc Jiri Olsa <[email protected]>

Cc Kan Liang <[email protected]>

Cc Linus Torvalds <[email protected]>

Cc Michael Ellerman <[email protected]>

Cc Milind Chabbi <[email protected]>

Cc Namhyung Kim <[email protected]>

Cc Oleg Nesterov <[email protected]>

Cc Peter Zijlstra <[email protected]>

Cc Sukadev Bhattiprolu <[email protected]>

Cc Thomas Gleixner <[email protected]>

Cc Will Deacon <[email protected]>

Link http://lkml.kernel.org/r/[email protected] open here

Signed-off-by Ingo Molnar <[email protected]>

---
 kernel/events/hw_breakpoint.c | 20 ++++++++++----------
 1 file changed, 10 insertions(+), 10 deletions(-)

kernel/events/hw_breakpoint.c +10 ~~-10~~

index 395ca07965af..9b31fbcc3305 100644
@@ -277,7 +277,7 @@ __weak void arch_unregister_hw_breakpoint(struct perf_event *bp)
  *       ((per_cpu(info->flexible, *) > 1) + max(per_cpu(info->cpu_pinned, *))
  *            + max(per_cpu(info->tsk_pinned, *))) < HBP_NUM
  */
-static int __reserve_bp_slot(struct perf_event *bp)
+static int __reserve_bp_slot(struct perf_event *bp, u64 bp_type)
 {
 	struct bp_busy_slots slots = {0};
 	enum bp_type_idx type;

@@ -288,11 +288,11 @@ static int __reserve_bp_slot(struct perf_event *bp)
 		return -ENOMEM;
 
 	/* Basic checks */
-	if (bp->attr.bp_type == HW_BREAKPOINT_EMPTY ||
-	    bp->attr.bp_type == HW_BREAKPOINT_INVALID)
+	if (bp_type == HW_BREAKPOINT_EMPTY ||
+	    bp_type == HW_BREAKPOINT_INVALID)
 		return -EINVAL;
 
-	type = find_slot_idx(bp->attr.bp_type);
+	type = find_slot_idx(bp_type);
 	weight = hw_breakpoint_weight(bp);
 
 	fetch_bp_busy_slots(&slots, bp, type);

@@ -317,19 +317,19 @@ int reserve_bp_slot(struct perf_event *bp)
 
 	mutex_lock(&nr_bp_mutex);
 
-	ret = __reserve_bp_slot(bp);
+	ret = __reserve_bp_slot(bp, bp->attr.bp_type);
 
 	mutex_unlock(&nr_bp_mutex);
 
 	return ret;
 }
 
-static void __release_bp_slot(struct perf_event *bp)
+static void __release_bp_slot(struct perf_event *bp, u64 bp_type)
 {
 	enum bp_type_idx type;
 	int weight;
 
-	type = find_slot_idx(bp->attr.bp_type);
+	type = find_slot_idx(bp_type);
 	weight = hw_breakpoint_weight(bp);
 	toggle_bp_slot(bp, false, type, weight);
 }

@@ -339,7 +339,7 @@ void release_bp_slot(struct perf_event *bp)
 	mutex_lock(&nr_bp_mutex);
 
 	arch_unregister_hw_breakpoint(bp);
-	__release_bp_slot(bp);
+	__release_bp_slot(bp, bp->attr.bp_type);
 
 	mutex_unlock(&nr_bp_mutex);
 }

@@ -354,7 +354,7 @@ int dbg_reserve_bp_slot(struct perf_event *bp)
 	if (mutex_is_locked(&nr_bp_mutex))
 		return -1;
 
-	return __reserve_bp_slot(bp);
+	return __reserve_bp_slot(bp, bp->attr.bp_type);
 }
 
 int dbg_release_bp_slot(struct perf_event *bp)

@@ -362,7 +362,7 @@ int dbg_release_bp_slot(struct perf_event *bp)
 	if (mutex_is_locked(&nr_bp_mutex))
 		return -1;
 
-	__release_bp_slot(bp);
+	__release_bp_slot(bp, bp->attr.bp_type);
 
 	return 0;
 }

Jiri Olsa

author reviewer

2018-03-12 13:46 UTC #5 lore raw

PATCH v3 3/8 hw_breakpoint: Add modify_bp_slot function

In reply to Jiri Olsa

Adding modify_bp_slot function to keep slot numbers correct when changing the breakpoint type.

Using existing __release_bp_slot/__reserve_bp_slot call sequence to update the slot counts.

Signed-off-by Jiri Olsa <[email protected]>

---
 kernel/events/hw_breakpoint.c | 46 ++++++++++++++++++++++++++++++++++++-------
 1 file changed, 39 insertions(+), 7 deletions(-)

kernel/events/hw_breakpoint.c +39 -7

index 9b31fbcc3305..776948beb4ac 100644
@@ -44,6 +44,7 @@
 #include <linux/list.h>
 #include <linux/cpu.h>
 #include <linux/smp.h>
+#include <linux/bug.h>
 
 #include <linux/hw_breakpoint.h>
 /*

@@ -344,6 +345,38 @@ void release_bp_slot(struct perf_event *bp)
 	mutex_unlock(&nr_bp_mutex);
 }
 
+static int __modify_bp_slot(struct perf_event *bp, u64 old_type)
+{
+	int err;
+
+	__release_bp_slot(bp, old_type);
+
+	err = __reserve_bp_slot(bp, bp->attr.bp_type);
+	if (err) {
+		/*
+		 * Reserve the old_type slot back in case
+		 * there's no space for the new type.
+		 *
+		 * This must succeed, because we just released
+		 * the old_type slot in the __release_bp_slot
+		 * call above. If not, something is broken.
+		 */
+		WARN_ON(__reserve_bp_slot(bp, old_type));
+	}
+
+	return err;
+}
+
+static int modify_bp_slot(struct perf_event *bp, u64 old_type)
+{
+	int ret;
+
+	mutex_lock(&nr_bp_mutex);
+	ret = __modify_bp_slot(bp, old_type);
+	mutex_unlock(&nr_bp_mutex);
+	return ret;
+}
+
 /*
  * Allow the kernel debugger to reserve breakpoint slots without
  * taking a lock using the dbg_* variant of for the reserve and

@@ -435,6 +468,7 @@ int modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *att
 	u64 old_addr = bp->attr.bp_addr;
 	u64 old_len = bp->attr.bp_len;
 	int old_type = bp->attr.bp_type;
+	bool modify = attr->bp_type != old_type;
 	int err = 0;
 
 	/*

@@ -452,12 +486,9 @@ int modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *att
 	bp->attr.bp_type = attr->bp_type;
 	bp->attr.bp_len = attr->bp_len;
 
-	if (attr->disabled)
-		goto end;
-
 	err = validate_hw_breakpoint(bp);
-	if (!err)
-		perf_event_enable(bp);
+	if (!err && modify)
+		err = modify_bp_slot(bp, old_type);
 
 	if (err) {
 		bp->attr.bp_addr = old_addr;

@@ -469,9 +500,10 @@ int modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *att
 		return err;
 	}
 
-end:
-	bp->attr.disabled = attr->disabled;
+	if (!attr->disabled)
+		perf_event_enable(bp);
 
+	bp->attr.disabled = attr->disabled;
 	return 0;
 }
 EXPORT_SYMBOL_GPL(modify_user_hw_breakpoint);

Signature · 1 line

2.13.6

tip-bot for Jiri Olsa

merge bot

2018-03-13 06:20 UTC #6 lore raw

[tip:perf/core] hw_breakpoint: Add modify_bp_slot() function

hw_breakpoint: Add modify_bp_slot() function

Commit-ID:  ea6a9d530c1779b9bd30944b803820c462f01eac
Gitweb:     https://git.kernel.org/tip/ea6a9d530c1779b9bd30944b803820c462f01eac
Author:     Jiri Olsa <[email protected]>
AuthorDate: Mon, 12 Mar 2018 14:45:43 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Tue, 13 Mar 2018 06:56:07 +0100

hw_breakpoint: Add modify_bp_slot() function

Add the modify_bp_slot() function to keep slot numbers correct when changing the breakpoint type.

Using existing __release_bp_slot()/__reserve_bp_slot() call sequence to update the slot counts.

Signed-off-by Jiri Olsa <[email protected]>

Cc Alexander Shishkin <[email protected]>

Cc Arnaldo Carvalho de Melo <[email protected]>

Cc David Ahern <[email protected]>

Cc Frederic Weisbecker <[email protected]>

Cc Hari Bathini <[email protected]>

Cc Jin Yao <[email protected]>

Cc Jiri Olsa <[email protected]>

Cc Kan Liang <[email protected]>

Cc Linus Torvalds <[email protected]>

Cc Michael Ellerman <[email protected]>

Cc Milind Chabbi <[email protected]>

Cc Namhyung Kim <[email protected]>

Cc Oleg Nesterov <[email protected]>

Cc Peter Zijlstra <[email protected]>

Cc Sukadev Bhattiprolu <[email protected]>

Cc Thomas Gleixner <[email protected]>

Cc Will Deacon <[email protected]>

Link http://lkml.kernel.org/r/[email protected] open here

Signed-off-by Ingo Molnar <[email protected]>

---
 kernel/events/hw_breakpoint.c | 46 ++++++++++++++++++++++++++++++++++++-------
 1 file changed, 39 insertions(+), 7 deletions(-)

kernel/events/hw_breakpoint.c +39 -7

index 9b31fbcc3305..776948beb4ac 100644
@@ -44,6 +44,7 @@
 #include <linux/list.h>
 #include <linux/cpu.h>
 #include <linux/smp.h>
+#include <linux/bug.h>
 
 #include <linux/hw_breakpoint.h>
 /*

@@ -344,6 +345,38 @@ void release_bp_slot(struct perf_event *bp)
 	mutex_unlock(&nr_bp_mutex);
 }
 
+static int __modify_bp_slot(struct perf_event *bp, u64 old_type)
+{
+	int err;
+
+	__release_bp_slot(bp, old_type);
+
+	err = __reserve_bp_slot(bp, bp->attr.bp_type);
+	if (err) {
+		/*
+		 * Reserve the old_type slot back in case
+		 * there's no space for the new type.
+		 *
+		 * This must succeed, because we just released
+		 * the old_type slot in the __release_bp_slot
+		 * call above. If not, something is broken.
+		 */
+		WARN_ON(__reserve_bp_slot(bp, old_type));
+	}
+
+	return err;
+}
+
+static int modify_bp_slot(struct perf_event *bp, u64 old_type)
+{
+	int ret;
+
+	mutex_lock(&nr_bp_mutex);
+	ret = __modify_bp_slot(bp, old_type);
+	mutex_unlock(&nr_bp_mutex);
+	return ret;
+}
+
 /*
  * Allow the kernel debugger to reserve breakpoint slots without
  * taking a lock using the dbg_* variant of for the reserve and

@@ -435,6 +468,7 @@ int modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *att
 	u64 old_addr = bp->attr.bp_addr;
 	u64 old_len = bp->attr.bp_len;
 	int old_type = bp->attr.bp_type;
+	bool modify = attr->bp_type != old_type;
 	int err = 0;
 
 	/*

@@ -452,12 +486,9 @@ int modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *att
 	bp->attr.bp_type = attr->bp_type;
 	bp->attr.bp_len = attr->bp_len;
 
-	if (attr->disabled)
-		goto end;
-
 	err = validate_hw_breakpoint(bp);
-	if (!err)
-		perf_event_enable(bp);
+	if (!err && modify)
+		err = modify_bp_slot(bp, old_type);
 
 	if (err) {
 		bp->attr.bp_addr = old_addr;

@@ -469,9 +500,10 @@ int modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *att
 		return err;
 	}
 
-end:
-	bp->attr.disabled = attr->disabled;
+	if (!attr->disabled)
+		perf_event_enable(bp);
 
+	bp->attr.disabled = attr->disabled;
 	return 0;
 }
 EXPORT_SYMBOL_GPL(modify_user_hw_breakpoint);

Jiri Olsa

author reviewer

2018-03-12 13:47 UTC #7 lore raw

PATCH v3 4/8 hw_breakpoint: Factor out __modify_user_hw_breakpoint function

In reply to Jiri Olsa

Moving out the all the functionality without the events disabling/enabling calls, because we want to call another disabling/enabling functions in following change.

Signed-off-by Jiri Olsa <[email protected]>

---
 kernel/events/hw_breakpoint.c | 46 +++++++++++++++++++++++++++----------------
 1 file changed, 29 insertions(+), 17 deletions(-)

kernel/events/hw_breakpoint.c +29 ~~-17~~

index 776948beb4ac..a556aba223da 100644
@@ -456,6 +456,33 @@ register_user_hw_breakpoint(struct perf_event_attr *attr,
 }
 EXPORT_SYMBOL_GPL(register_user_hw_breakpoint);
 
+static int __modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *attr)
+{
+	u64 old_addr = bp->attr.bp_addr;
+	u64 old_len  = bp->attr.bp_len;
+	int old_type = bp->attr.bp_type;
+	bool modify  = attr->bp_type != old_type;
+	int err = 0;
+
+	bp->attr.bp_addr = attr->bp_addr;
+	bp->attr.bp_type = attr->bp_type;
+	bp->attr.bp_len  = attr->bp_len;
+
+	err = validate_hw_breakpoint(bp);
+	if (!err && modify)
+		err = modify_bp_slot(bp, old_type);
+
+	if (err) {
+		bp->attr.bp_addr = old_addr;
+		bp->attr.bp_type = old_type;
+		bp->attr.bp_len  = old_len;
+		return err;
+	}
+
+	bp->attr.disabled = attr->disabled;
+	return 0;
+}
+
 /**
  * modify_user_hw_breakpoint - modify a user-space hardware breakpoint
  * @bp: the breakpoint structure to modify

@@ -465,11 +492,7 @@ EXPORT_SYMBOL_GPL(register_user_hw_breakpoint);
  */
 int modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *attr)
 {
-	u64 old_addr = bp->attr.bp_addr;
-	u64 old_len = bp->attr.bp_len;
-	int old_type = bp->attr.bp_type;
-	bool modify = attr->bp_type != old_type;
-	int err = 0;
+	int err;
 
 	/*
 	 * modify_user_hw_breakpoint can be invoked with IRQs disabled and hence it

@@ -482,18 +505,9 @@ int modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *att
 	else
 		perf_event_disable(bp);
 
-	bp->attr.bp_addr = attr->bp_addr;
-	bp->attr.bp_type = attr->bp_type;
-	bp->attr.bp_len = attr->bp_len;
-
-	err = validate_hw_breakpoint(bp);
-	if (!err && modify)
-		err = modify_bp_slot(bp, old_type);
+	err = __modify_user_hw_breakpoint(bp, attr);
 
 	if (err) {
-		bp->attr.bp_addr = old_addr;
-		bp->attr.bp_type = old_type;
-		bp->attr.bp_len = old_len;
 		if (!bp->attr.disabled)
 			perf_event_enable(bp);

@@ -502,8 +516,6 @@ int modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *att
 
 	if (!attr->disabled)
 		perf_event_enable(bp);
-
-	bp->attr.disabled = attr->disabled;
 	return 0;
 }
 EXPORT_SYMBOL_GPL(modify_user_hw_breakpoint);

Signature · 1 line

2.13.6

tip-bot for Jiri Olsa

merge bot

2018-03-13 06:21 UTC #8 lore raw

[tip:perf/core] hw_breakpoint: Factor out __modify_user_hw_breakpoint() function

hw_breakpoint: Factor out __modify_user_hw_breakpoint() function

Commit-ID:  18ff57b220610a699947f20b156a8245ca7eee98
Gitweb:     https://git.kernel.org/tip/18ff57b220610a699947f20b156a8245ca7eee98
Author:     Jiri Olsa <[email protected]>
AuthorDate: Mon, 12 Mar 2018 14:45:44 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Tue, 13 Mar 2018 06:56:08 +0100

hw_breakpoint: Factor out __modify_user_hw_breakpoint() function

Moving out all the functionality without the events disabling/enabling calls, because we want to call another disabling/enabling functions in following change.

Signed-off-by Jiri Olsa <[email protected]>

Cc Alexander Shishkin <[email protected]>

Cc Arnaldo Carvalho de Melo <[email protected]>

Cc David Ahern <[email protected]>

Cc Frederic Weisbecker <[email protected]>

Cc Hari Bathini <[email protected]>

Cc Jin Yao <[email protected]>

Cc Jiri Olsa <[email protected]>

Cc Kan Liang <[email protected]>

Cc Linus Torvalds <[email protected]>

Cc Michael Ellerman <[email protected]>

Cc Milind Chabbi <[email protected]>

Cc Namhyung Kim <[email protected]>

Cc Oleg Nesterov <[email protected]>

Cc Peter Zijlstra <[email protected]>

Cc Sukadev Bhattiprolu <[email protected]>

Cc Thomas Gleixner <[email protected]>

Cc Will Deacon <[email protected]>

Link http://lkml.kernel.org/r/[email protected] open here

Signed-off-by Ingo Molnar <[email protected]>

---
 kernel/events/hw_breakpoint.c | 46 +++++++++++++++++++++++++++----------------
 1 file changed, 29 insertions(+), 17 deletions(-)

kernel/events/hw_breakpoint.c +29 ~~-17~~

index 776948beb4ac..a556aba223da 100644
@@ -456,6 +456,33 @@ register_user_hw_breakpoint(struct perf_event_attr *attr,
 }
 EXPORT_SYMBOL_GPL(register_user_hw_breakpoint);
 
+static int __modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *attr)
+{
+	u64 old_addr = bp->attr.bp_addr;
+	u64 old_len  = bp->attr.bp_len;
+	int old_type = bp->attr.bp_type;
+	bool modify  = attr->bp_type != old_type;
+	int err = 0;
+
+	bp->attr.bp_addr = attr->bp_addr;
+	bp->attr.bp_type = attr->bp_type;
+	bp->attr.bp_len  = attr->bp_len;
+
+	err = validate_hw_breakpoint(bp);
+	if (!err && modify)
+		err = modify_bp_slot(bp, old_type);
+
+	if (err) {
+		bp->attr.bp_addr = old_addr;
+		bp->attr.bp_type = old_type;
+		bp->attr.bp_len  = old_len;
+		return err;
+	}
+
+	bp->attr.disabled = attr->disabled;
+	return 0;
+}
+
 /**
  * modify_user_hw_breakpoint - modify a user-space hardware breakpoint
  * @bp: the breakpoint structure to modify

@@ -465,11 +492,7 @@ EXPORT_SYMBOL_GPL(register_user_hw_breakpoint);
  */
 int modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *attr)
 {
-	u64 old_addr = bp->attr.bp_addr;
-	u64 old_len = bp->attr.bp_len;
-	int old_type = bp->attr.bp_type;
-	bool modify = attr->bp_type != old_type;
-	int err = 0;
+	int err;
 
 	/*
 	 * modify_user_hw_breakpoint can be invoked with IRQs disabled and hence it

@@ -482,18 +505,9 @@ int modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *att
 	else
 		perf_event_disable(bp);
 
-	bp->attr.bp_addr = attr->bp_addr;
-	bp->attr.bp_type = attr->bp_type;
-	bp->attr.bp_len = attr->bp_len;
-
-	err = validate_hw_breakpoint(bp);
-	if (!err && modify)
-		err = modify_bp_slot(bp, old_type);
+	err = __modify_user_hw_breakpoint(bp, attr);
 
 	if (err) {
-		bp->attr.bp_addr = old_addr;
-		bp->attr.bp_type = old_type;
-		bp->attr.bp_len = old_len;
 		if (!bp->attr.disabled)
 			perf_event_enable(bp);

@@ -502,8 +516,6 @@ int modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *att
 
 	if (!attr->disabled)
 		perf_event_enable(bp);
-
-	bp->attr.disabled = attr->disabled;
 	return 0;
 }
 EXPORT_SYMBOL_GPL(modify_user_hw_breakpoint);

Jiri Olsa

author reviewer

2018-03-12 13:46 UTC #9 lore raw

PATCH v3 5/8 hw_breakpoint: Add perf_event_attr fields check in __modify_user_hw_breakpoint

In reply to Jiri Olsa

And rename it to modify_user_hw_breakpoint_check.

We are about to use modify_user_hw_breakpoint_check for user space breakpoints modification, we must be very strict to check only the fields we can change have changed. As Peter explained:

  Suppose someone does:
        attr = malloc(sizeof(*attr)); // uninitialized memory
        attr->type = BP;
        attr->bp_addr = new_addr;
        attr->bp_type = bp_type;
        attr->bp_len = bp_len;
        ioctl(fd, PERF_IOC_MOD_ATTR, &attr);

And feeds absolute shite for the rest of the fields. Then we later want to extend IOC_MOD_ATTR to allow changing attr::sample_type but we can't, because that would break the above application.

I'm making this check optional because we already export modify_user_hw_breakpoint and with this check we could break existing users.

Signed-off-by Jiri Olsa <[email protected]>

---
 kernel/events/hw_breakpoint.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

kernel/events/hw_breakpoint.c +7 -2

index a556aba223da..0c82663395f7 100644
@@ -456,7 +456,9 @@ register_user_hw_breakpoint(struct perf_event_attr *attr,
 }
 EXPORT_SYMBOL_GPL(register_user_hw_breakpoint);
 
-static int __modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *attr)
+static int
+modify_user_hw_breakpoint_check(struct perf_event *bp, struct perf_event_attr *attr,
+			        bool check)
 {
 	u64 old_addr = bp->attr.bp_addr;
 	u64 old_len  = bp->attr.bp_len;

@@ -468,6 +470,9 @@ static int __modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_
 	bp->attr.bp_type = attr->bp_type;
 	bp->attr.bp_len  = attr->bp_len;
 
+	if (check && memcmp(&bp->attr, attr, sizeof(*attr)))
+		return -EINVAL;
+
 	err = validate_hw_breakpoint(bp);
 	if (!err && modify)
 		err = modify_bp_slot(bp, old_type);

@@ -505,7 +510,7 @@ int modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *att
 	else
 		perf_event_disable(bp);
 
-	err = __modify_user_hw_breakpoint(bp, attr);
+	err = modify_user_hw_breakpoint_check(bp, attr, false);
 
 	if (err) {
 		if (!bp->attr.disabled)

Signature · 1 line

2.13.6

tip-bot for Jiri Olsa

merge bot

2018-03-13 06:21 UTC #10 lore raw

[tip:perf/core] hw_breakpoint: Add perf_event_attr fields check in __modify_user_hw_breakpoint()

hw_breakpoint: Add perf_event_attr fields check in __modify_user_hw_breakpoint()

Commit-ID:  705feaf321c37e4dca3637fd5cb3b275f17a06c9
Gitweb:     https://git.kernel.org/tip/705feaf321c37e4dca3637fd5cb3b275f17a06c9
Author:     Jiri Olsa <[email protected]>
AuthorDate: Mon, 12 Mar 2018 14:45:45 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Tue, 13 Mar 2018 06:56:08 +0100

hw_breakpoint: Add perf_event_attr fields check in __modify_user_hw_breakpoint()

And rename it to modify_user_hw_breakpoint_check().

We are about to use modify_user_hw_breakpoint_check() for user space breakpoints modification, we must be very strict to check only the fields we can change have changed. As Peter explained:

"Suppose someone does:

        attr = malloc(sizeof(*attr)); // uninitialized memory
        attr->type = BP;
        attr->bp_addr = new_addr;
        attr->bp_type = bp_type;
        attr->bp_len = bp_len;
        ioctl(fd, PERF_IOC_MOD_ATTR, &attr);

And feeds absolute shite for the rest of the fields. Then we later want to extend IOC_MOD_ATTR to allow changing attr::sample_type but we can't, because that would break the above application."

I'm making this check optional because we already export modify_user_hw_breakpoint() and with this check we could break existing users.

Signed-off-by Jiri Olsa <[email protected]>

Cc Alexander Shishkin <[email protected]>

Cc Arnaldo Carvalho de Melo <[email protected]>

Cc David Ahern <[email protected]>

Cc Frederic Weisbecker <[email protected]>

Cc Hari Bathini <[email protected]>

Cc Jin Yao <[email protected]>

Cc Jiri Olsa <[email protected]>

Cc Kan Liang <[email protected]>

Cc Linus Torvalds <[email protected]>

Cc Michael Ellerman <[email protected]>

Cc Milind Chabbi <[email protected]>

Cc Namhyung Kim <[email protected]>

Cc Oleg Nesterov <[email protected]>

Cc Peter Zijlstra <[email protected]>

Cc Sukadev Bhattiprolu <[email protected]>

Cc Thomas Gleixner <[email protected]>

Cc Will Deacon <[email protected]>

Link http://lkml.kernel.org/r/[email protected] open here

Signed-off-by Ingo Molnar <[email protected]>

---
 kernel/events/hw_breakpoint.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

kernel/events/hw_breakpoint.c +7 -2

index a556aba223da..0c82663395f7 100644
@@ -456,7 +456,9 @@ register_user_hw_breakpoint(struct perf_event_attr *attr,
 }
 EXPORT_SYMBOL_GPL(register_user_hw_breakpoint);
 
-static int __modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *attr)
+static int
+modify_user_hw_breakpoint_check(struct perf_event *bp, struct perf_event_attr *attr,
+			        bool check)
 {
 	u64 old_addr = bp->attr.bp_addr;
 	u64 old_len  = bp->attr.bp_len;

@@ -468,6 +470,9 @@ static int __modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_
 	bp->attr.bp_type = attr->bp_type;
 	bp->attr.bp_len  = attr->bp_len;
 
+	if (check && memcmp(&bp->attr, attr, sizeof(*attr)))
+		return -EINVAL;
+
 	err = validate_hw_breakpoint(bp);
 	if (!err && modify)
 		err = modify_bp_slot(bp, old_type);

@@ -505,7 +510,7 @@ int modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *att
 	else
 		perf_event_disable(bp);
 
-	err = __modify_user_hw_breakpoint(bp, attr);
+	err = modify_user_hw_breakpoint_check(bp, attr, false);
 
 	if (err) {
 		if (!bp->attr.disabled)

Jiri Olsa

author reviewer

2018-03-12 13:47 UTC #11 lore raw

PATCH v3 6/8 perf/core: Move perf_event_attr::sample_max_stack into perf_copy_attr

In reply to Jiri Olsa

Moving sample_max_stack check and setup into perf_copy_attr, so we have all perf_event_attr initial setup in one place and can easily compare attrs in the new ioctl introduced in following change.

Signed-off-by Jiri Olsa <[email protected]>

---
 kernel/events/core.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

kernel/events/core.c +3 -3

index 57898102847f..bd61b4d68a25 100644
@@ -9834,6 +9834,9 @@ static int perf_copy_attr(struct perf_event_attr __user *uattr,
 			ret = -EINVAL;
 	}
 
+	if (!attr->sample_max_stack)
+		attr->sample_max_stack = sysctl_perf_event_max_stack;
+
 	if (attr->sample_type & PERF_SAMPLE_REGS_INTR)
 		ret = perf_reg_validate(attr->sample_regs_intr);
 out:

@@ -10047,9 +10050,6 @@ SYSCALL_DEFINE5(perf_event_open,
 	    perf_paranoid_kernel() && !capable(CAP_SYS_ADMIN))
 		return -EACCES;
 
-	if (!attr.sample_max_stack)
-		attr.sample_max_stack = sysctl_perf_event_max_stack;
-
 	/*
 	 * In cgroup mode, the pid argument is used to pass the fd
 	 * opened to the cgroup directory in cgroupfs. The cpu argument

Signature · 1 line

2.13.6

tip-bot for Jiri Olsa

merge bot

2018-03-13 06:21 UTC #12 lore raw

[tip:perf/core] perf/core: Move perf_event_attr::sample_max_stack into perf_copy_attr()

perf/core: Move perf_event_attr::sample_max_stack into perf_copy_attr()

Commit-ID:  5f970521d3279d99adcdebf329631e36cb9f0deb
Gitweb:     https://git.kernel.org/tip/5f970521d3279d99adcdebf329631e36cb9f0deb
Author:     Jiri Olsa <[email protected]>
AuthorDate: Mon, 12 Mar 2018 14:45:46 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Tue, 13 Mar 2018 06:56:08 +0100

perf/core: Move perf_event_attr::sample_max_stack into perf_copy_attr()

Move the sample_max_stack check and setup into perf_copy_attr(), so we have all perf_event_attr initial setup in one place and can easily compare attrs in the new ioctl introduced in following change.

Signed-off-by Jiri Olsa <[email protected]>

Cc Alexander Shishkin <[email protected]>

Cc Arnaldo Carvalho de Melo <[email protected]>

Cc David Ahern <[email protected]>

Cc Frederic Weisbecker <[email protected]>

Cc Hari Bathini <[email protected]>

Cc Jin Yao <[email protected]>

Cc Jiri Olsa <[email protected]>

Cc Kan Liang <[email protected]>

Cc Linus Torvalds <[email protected]>

Cc Michael Ellerman <[email protected]>

Cc Milind Chabbi <[email protected]>

Cc Namhyung Kim <[email protected]>

Cc Oleg Nesterov <[email protected]>

Cc Peter Zijlstra <[email protected]>

Cc Sukadev Bhattiprolu <[email protected]>

Cc Thomas Gleixner <[email protected]>

Cc Will Deacon <[email protected]>

Link http://lkml.kernel.org/r/[email protected] open here

Signed-off-by Ingo Molnar <[email protected]>

---
 kernel/events/core.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

kernel/events/core.c +3 -3

index 969f865f9f1c..ee145bdee6ed 100644
@@ -10125,6 +10125,9 @@ static int perf_copy_attr(struct perf_event_attr __user *uattr,
 			ret = -EINVAL;
 	}
 
+	if (!attr->sample_max_stack)
+		attr->sample_max_stack = sysctl_perf_event_max_stack;
+
 	if (attr->sample_type & PERF_SAMPLE_REGS_INTR)
 		ret = perf_reg_validate(attr->sample_regs_intr);
 out:

@@ -10338,9 +10341,6 @@ SYSCALL_DEFINE5(perf_event_open,
 	    perf_paranoid_kernel() && !capable(CAP_SYS_ADMIN))
 		return -EACCES;
 
-	if (!attr.sample_max_stack)
-		attr.sample_max_stack = sysctl_perf_event_max_stack;
-
 	/*
 	 * In cgroup mode, the pid argument is used to pass the fd
 	 * opened to the cgroup directory in cgroupfs. The cpu argument

Jiri Olsa

author reviewer

2018-03-12 13:46 UTC #13 lore raw

PATCH v3 7/8 perf/core: fast breakpoint modification via _IOC_MODIFY_ATTRIBUTES.

In reply to Jiri Olsa

From: Milind Chabbi <[email protected]>

Problem and motivation: Once a breakpoint perf event (PERF_TYPE_BREAKPOINT) is created, there is no flexibility to change the breakpoint type (bp_type), breakpoint address (bp_addr), or breakpoint length (bp_len). The only option is to close the perf event and configure a new breakpoint event. This inflexibility has a significant performance overhead. For example, sampling-based, lightweight performance profilers (and also concurrency bug detection tools), monitor different addresses for a short duration using PERF_TYPE_BREAKPOINT and change the address (bp_addr) to another address or change the kind of breakpoint (bp_type) from "write" to a "read" or vice-versa or change the length (bp_len) of the address being monitored. The cost of these modifications is prohibitive since it involves unmapping the circular buffer associated with the perf event, closing the perf event, opening another perf event and mmaping another circular buffer.

Solution: The new ioctl flag for perf events, PERF_EVENT_IOC_MODIFY_ATTRIBUTES, introduced in this patch takes a pointer to a struct perf_event_attr as an argument to update an old breakpoint event with new address, type, and size. This facility allows retaining a previous mmaped perf events ring buffer and avoids having to close and reopen another perf event.

This patch supports only changing PERF_TYPE_BREAKPOINT event type; future implementations can extend this feature. The patch replicates some of its functionality of modify_user_hw_breakpoint() in kernel/events/hw_breakpoint.c. modify_user_hw_breakpoint cannot be called directly since perf_event_ctx_lock() is already held in _perf_ioctl().

Evidence: Experiments show that the baseline (not able to modify an already created breakpoint) costs an order of magnitude (~10x) more than the suggested optimization (having the ability to dynamically modifying a configured breakpoint via ioctl). When the breakpoints typically do not trap, the speedup due to the suggested optimization is ~10x; even when the breakpoints always trap, the speedup is ~4x due to the suggested optimization.

Testing: tests posted at
https://github.com/linux-contrib/perf_event_modify_bp demonstrate the performance significance of this patch. Tests also check the functional correctness of the patch.

Signed-off-by Milind Chabbi <[email protected]>

[ Using modify_user_hw_breakpoint_check function. ]

[ Reformated PERF_EVENT_IOC_*, so the values are all in one column. ]

Signed-off-by Jiri Olsa <[email protected]>

---
 include/linux/hw_breakpoint.h         |  7 +++++
 include/uapi/linux/perf_event.h       | 23 +++++++++--------
 kernel/events/core.c                  | 48 +++++++++++++++++++++++++++++++++++
 kernel/events/hw_breakpoint.c         |  2 +-
 tools/include/uapi/linux/perf_event.h | 23 +++++++++--------
 5 files changed, 80 insertions(+), 23 deletions(-)

include/linux/hw_breakpoint.h +7 -0

index cf045885a499..abeba094f080 100644
@@ -53,6 +53,9 @@ register_user_hw_breakpoint(struct perf_event_attr *attr,
 /* FIXME: only change from the attr, and don't unregister */
 extern int
 modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *attr);
+extern int
+modify_user_hw_breakpoint_check(struct perf_event *bp, struct perf_event_attr *attr,
+				bool check);
 
 /*
  * Kernel breakpoints are not associated with any particular thread.

@@ -97,6 +100,10 @@ register_user_hw_breakpoint(struct perf_event_attr *attr,
 static inline int
 modify_user_hw_breakpoint(struct perf_event *bp,
 			  struct perf_event_attr *attr)	{ return -ENOSYS; }
+static inline int
+__modify_user_hw_breakpoint(struct perf_event *bp,
+			    struct perf_event_attr *attr) { return -ENOSYS; }
+
 static inline struct perf_event *
 register_wide_hw_breakpoint_cpu(struct perf_event_attr *attr,
 				perf_overflow_handler_t	 triggered,

include/uapi/linux/perf_event.h +12 ~~-11~~

index 6f873503552d..912b85b52344 100644
@@ -448,17 +448,18 @@ struct perf_event_query_bpf {
 /*
  * Ioctls that can be done on a perf event fd:
  */
-#define PERF_EVENT_IOC_ENABLE		_IO ('$', 0)
-#define PERF_EVENT_IOC_DISABLE		_IO ('$', 1)
-#define PERF_EVENT_IOC_REFRESH		_IO ('$', 2)
-#define PERF_EVENT_IOC_RESET		_IO ('$', 3)
-#define PERF_EVENT_IOC_PERIOD		_IOW('$', 4, __u64)
-#define PERF_EVENT_IOC_SET_OUTPUT	_IO ('$', 5)
-#define PERF_EVENT_IOC_SET_FILTER	_IOW('$', 6, char *)
-#define PERF_EVENT_IOC_ID		_IOR('$', 7, __u64 *)
-#define PERF_EVENT_IOC_SET_BPF		_IOW('$', 8, __u32)
-#define PERF_EVENT_IOC_PAUSE_OUTPUT	_IOW('$', 9, __u32)
-#define PERF_EVENT_IOC_QUERY_BPF	_IOWR('$', 10, struct perf_event_query_bpf *)
+#define PERF_EVENT_IOC_ENABLE			_IO ('$', 0)
+#define PERF_EVENT_IOC_DISABLE			_IO ('$', 1)
+#define PERF_EVENT_IOC_REFRESH			_IO ('$', 2)
+#define PERF_EVENT_IOC_RESET			_IO ('$', 3)
+#define PERF_EVENT_IOC_PERIOD			_IOW('$', 4, __u64)
+#define PERF_EVENT_IOC_SET_OUTPUT		_IO ('$', 5)
+#define PERF_EVENT_IOC_SET_FILTER		_IOW('$', 6, char *)
+#define PERF_EVENT_IOC_ID			_IOR('$', 7, __u64 *)
+#define PERF_EVENT_IOC_SET_BPF			_IOW('$', 8, __u32)
+#define PERF_EVENT_IOC_PAUSE_OUTPUT		_IOW('$', 9, __u32)
+#define PERF_EVENT_IOC_QUERY_BPF		_IOWR('$', 10, struct perf_event_query_bpf *)
+#define PERF_EVENT_IOC_MODIFY_ATTRIBUTES	_IOW('$', 11, struct perf_event_attr *)
 
 enum perf_event_ioc_flags {
 	PERF_IOC_FLAG_GROUP		= 1U << 0,

kernel/events/core.c +48 -0

66 lines not on this page · raw message

kernel/events/hw_breakpoint.c +1 -1

index 0c82663395f7..6253d5519cd8 100644
@@ -456,7 +456,7 @@ register_user_hw_breakpoint(struct perf_event_attr *attr,
 }
 EXPORT_SYMBOL_GPL(register_user_hw_breakpoint);
 
-static int
+int
 modify_user_hw_breakpoint_check(struct perf_event *bp, struct perf_event_attr *attr,
 			        bool check)
 {

tools/include/uapi/linux/perf_event.h +12 ~~-11~~

index 6f873503552d..912b85b52344 100644
@@ -448,17 +448,18 @@ struct perf_event_query_bpf {
 /*
  * Ioctls that can be done on a perf event fd:
  */
-#define PERF_EVENT_IOC_ENABLE		_IO ('$', 0)
-#define PERF_EVENT_IOC_DISABLE		_IO ('$', 1)
-#define PERF_EVENT_IOC_REFRESH		_IO ('$', 2)
-#define PERF_EVENT_IOC_RESET		_IO ('$', 3)
-#define PERF_EVENT_IOC_PERIOD		_IOW('$', 4, __u64)
-#define PERF_EVENT_IOC_SET_OUTPUT	_IO ('$', 5)
-#define PERF_EVENT_IOC_SET_FILTER	_IOW('$', 6, char *)
-#define PERF_EVENT_IOC_ID		_IOR('$', 7, __u64 *)
-#define PERF_EVENT_IOC_SET_BPF		_IOW('$', 8, __u32)
-#define PERF_EVENT_IOC_PAUSE_OUTPUT	_IOW('$', 9, __u32)
-#define PERF_EVENT_IOC_QUERY_BPF	_IOWR('$', 10, struct perf_event_query_bpf *)
+#define PERF_EVENT_IOC_ENABLE			_IO ('$', 0)
+#define PERF_EVENT_IOC_DISABLE			_IO ('$', 1)
+#define PERF_EVENT_IOC_REFRESH			_IO ('$', 2)
+#define PERF_EVENT_IOC_RESET			_IO ('$', 3)
+#define PERF_EVENT_IOC_PERIOD			_IOW('$', 4, __u64)
+#define PERF_EVENT_IOC_SET_OUTPUT		_IO ('$', 5)
+#define PERF_EVENT_IOC_SET_FILTER		_IOW('$', 6, char *)
+#define PERF_EVENT_IOC_ID			_IOR('$', 7, __u64 *)
+#define PERF_EVENT_IOC_SET_BPF			_IOW('$', 8, __u32)
+#define PERF_EVENT_IOC_PAUSE_OUTPUT		_IOW('$', 9, __u32)
+#define PERF_EVENT_IOC_QUERY_BPF		_IOWR('$', 10, struct perf_event_query_bpf *)
+#define PERF_EVENT_IOC_MODIFY_ATTRIBUTES	_IOW('$', 11, struct perf_event_attr *)
 
 enum perf_event_ioc_flags {
 	PERF_IOC_FLAG_GROUP		= 1U << 0,

Signature · 1 line

2.13.6

tip-bot for Milind Chabbi

merge bot

2018-03-13 06:22 UTC #14 lore raw

[tip:perf/core] perf/core: Implement fast breakpoint modification via _IOC_MODIFY_ATTRIBUTES

perf/core: Implement fast breakpoint modification via _IOC_MODIFY_ATTRIBUTES

Commit-ID:  f30b09b7f8aed3180d6e2f2984e32e91c7a7fcd1
Gitweb:     https://git.kernel.org/tip/f30b09b7f8aed3180d6e2f2984e32e91c7a7fcd1
Author:     Milind Chabbi <[email protected]>
AuthorDate: Mon, 12 Mar 2018 14:45:47 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Tue, 13 Mar 2018 06:56:08 +0100

perf/core: Implement fast breakpoint modification via _IOC_MODIFY_ATTRIBUTES

Testing: tests posted at
https://github.com/linux-contrib/perf_event_modify_bp demonstrate the performance significance of this patch. Tests also check the functional correctness of the patch.

Signed-off-by Milind Chabbi <[email protected]>

[ Using modify_user_hw_breakpoint_check function. ]

[ Reformated PERF_EVENT_IOC_*, so the values are all in one column. ]

Signed-off-by Jiri Olsa <[email protected]>

Cc Alexander Shishkin <[email protected]>

Cc Arnaldo Carvalho de Melo <[email protected]>

Cc David Ahern <[email protected]>

Cc Frederic Weisbecker <[email protected]>

Cc Hari Bathini <[email protected]>

Cc Jin Yao <[email protected]>

Cc Jiri Olsa <[email protected]>

Cc Kan Liang <[email protected]>

Cc Linus Torvalds <[email protected]>

Cc Michael Ellerman <[email protected]>

Cc Namhyung Kim <[email protected]>

Cc Oleg Nesterov <[email protected]>

Cc Peter Zijlstra <[email protected]>

Cc Sukadev Bhattiprolu <[email protected]>

Cc Thomas Gleixner <[email protected]>

Cc Will Deacon <[email protected]>

Link http://lkml.kernel.org/r/[email protected] open here

Signed-off-by Ingo Molnar <[email protected]>

---
 include/linux/hw_breakpoint.h         |  7 +++++
 include/uapi/linux/perf_event.h       | 23 +++++++++--------
 kernel/events/core.c                  | 48 +++++++++++++++++++++++++++++++++++
 kernel/events/hw_breakpoint.c         |  2 +-
 tools/include/uapi/linux/perf_event.h | 23 +++++++++--------
 5 files changed, 80 insertions(+), 23 deletions(-)

include/linux/hw_breakpoint.h +7 -0

index cf045885a499..abeba094f080 100644
@@ -53,6 +53,9 @@ register_user_hw_breakpoint(struct perf_event_attr *attr,
 /* FIXME: only change from the attr, and don't unregister */
 extern int
 modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *attr);
+extern int
+modify_user_hw_breakpoint_check(struct perf_event *bp, struct perf_event_attr *attr,
+				bool check);
 
 /*
  * Kernel breakpoints are not associated with any particular thread.

@@ -97,6 +100,10 @@ register_user_hw_breakpoint(struct perf_event_attr *attr,
 static inline int
 modify_user_hw_breakpoint(struct perf_event *bp,
 			  struct perf_event_attr *attr)	{ return -ENOSYS; }
+static inline int
+__modify_user_hw_breakpoint(struct perf_event *bp,
+			    struct perf_event_attr *attr) { return -ENOSYS; }
+
 static inline struct perf_event *
 register_wide_hw_breakpoint_cpu(struct perf_event_attr *attr,
 				perf_overflow_handler_t	 triggered,

include/uapi/linux/perf_event.h +12 ~~-11~~

index 6f873503552d..912b85b52344 100644
@@ -448,17 +448,18 @@ struct perf_event_query_bpf {
 /*
  * Ioctls that can be done on a perf event fd:
  */
-#define PERF_EVENT_IOC_ENABLE		_IO ('$', 0)
-#define PERF_EVENT_IOC_DISABLE		_IO ('$', 1)
-#define PERF_EVENT_IOC_REFRESH		_IO ('$', 2)
-#define PERF_EVENT_IOC_RESET		_IO ('$', 3)
-#define PERF_EVENT_IOC_PERIOD		_IOW('$', 4, __u64)
-#define PERF_EVENT_IOC_SET_OUTPUT	_IO ('$', 5)
-#define PERF_EVENT_IOC_SET_FILTER	_IOW('$', 6, char *)
-#define PERF_EVENT_IOC_ID		_IOR('$', 7, __u64 *)
-#define PERF_EVENT_IOC_SET_BPF		_IOW('$', 8, __u32)
-#define PERF_EVENT_IOC_PAUSE_OUTPUT	_IOW('$', 9, __u32)
-#define PERF_EVENT_IOC_QUERY_BPF	_IOWR('$', 10, struct perf_event_query_bpf *)
+#define PERF_EVENT_IOC_ENABLE			_IO ('$', 0)
+#define PERF_EVENT_IOC_DISABLE			_IO ('$', 1)
+#define PERF_EVENT_IOC_REFRESH			_IO ('$', 2)
+#define PERF_EVENT_IOC_RESET			_IO ('$', 3)
+#define PERF_EVENT_IOC_PERIOD			_IOW('$', 4, __u64)
+#define PERF_EVENT_IOC_SET_OUTPUT		_IO ('$', 5)
+#define PERF_EVENT_IOC_SET_FILTER		_IOW('$', 6, char *)
+#define PERF_EVENT_IOC_ID			_IOR('$', 7, __u64 *)
+#define PERF_EVENT_IOC_SET_BPF			_IOW('$', 8, __u32)
+#define PERF_EVENT_IOC_PAUSE_OUTPUT		_IOW('$', 9, __u32)
+#define PERF_EVENT_IOC_QUERY_BPF		_IOWR('$', 10, struct perf_event_query_bpf *)
+#define PERF_EVENT_IOC_MODIFY_ATTRIBUTES	_IOW('$', 11, struct perf_event_attr *)
 
 enum perf_event_ioc_flags {
 	PERF_IOC_FLAG_GROUP		= 1U << 0,

kernel/events/core.c +48 -0

66 lines not on this page · raw message

kernel/events/hw_breakpoint.c +1 -1

index 0c82663395f7..6253d5519cd8 100644
@@ -456,7 +456,7 @@ register_user_hw_breakpoint(struct perf_event_attr *attr,
 }
 EXPORT_SYMBOL_GPL(register_user_hw_breakpoint);
 
-static int
+int
 modify_user_hw_breakpoint_check(struct perf_event *bp, struct perf_event_attr *attr,
 			        bool check)
 {

tools/include/uapi/linux/perf_event.h +12 ~~-11~~

index 6f873503552d..912b85b52344 100644
@@ -448,17 +448,18 @@ struct perf_event_query_bpf {
 /*
  * Ioctls that can be done on a perf event fd:
  */
-#define PERF_EVENT_IOC_ENABLE		_IO ('$', 0)
-#define PERF_EVENT_IOC_DISABLE		_IO ('$', 1)
-#define PERF_EVENT_IOC_REFRESH		_IO ('$', 2)
-#define PERF_EVENT_IOC_RESET		_IO ('$', 3)
-#define PERF_EVENT_IOC_PERIOD		_IOW('$', 4, __u64)
-#define PERF_EVENT_IOC_SET_OUTPUT	_IO ('$', 5)
-#define PERF_EVENT_IOC_SET_FILTER	_IOW('$', 6, char *)
-#define PERF_EVENT_IOC_ID		_IOR('$', 7, __u64 *)
-#define PERF_EVENT_IOC_SET_BPF		_IOW('$', 8, __u32)
-#define PERF_EVENT_IOC_PAUSE_OUTPUT	_IOW('$', 9, __u32)
-#define PERF_EVENT_IOC_QUERY_BPF	_IOWR('$', 10, struct perf_event_query_bpf *)
+#define PERF_EVENT_IOC_ENABLE			_IO ('$', 0)
+#define PERF_EVENT_IOC_DISABLE			_IO ('$', 1)
+#define PERF_EVENT_IOC_REFRESH			_IO ('$', 2)
+#define PERF_EVENT_IOC_RESET			_IO ('$', 3)
+#define PERF_EVENT_IOC_PERIOD			_IOW('$', 4, __u64)
+#define PERF_EVENT_IOC_SET_OUTPUT		_IO ('$', 5)
+#define PERF_EVENT_IOC_SET_FILTER		_IOW('$', 6, char *)
+#define PERF_EVENT_IOC_ID			_IOR('$', 7, __u64 *)
+#define PERF_EVENT_IOC_SET_BPF			_IOW('$', 8, __u32)
+#define PERF_EVENT_IOC_PAUSE_OUTPUT		_IOW('$', 9, __u32)
+#define PERF_EVENT_IOC_QUERY_BPF		_IOWR('$', 10, struct perf_event_query_bpf *)
+#define PERF_EVENT_IOC_MODIFY_ATTRIBUTES	_IOW('$', 11, struct perf_event_attr *)
 
 enum perf_event_ioc_flags {
 	PERF_IOC_FLAG_GROUP		= 1U << 0,

tip-bot for Milind Chabbi

merge bot

2018-03-13 15:15 UTC #15 lore raw

[tip:perf/core] perf/core: Implement fast breakpoint modification via _IOC_MODIFY_ATTRIBUTES

In reply to Jiri Olsa

perf/core: Implement fast breakpoint modification via _IOC_MODIFY_ATTRIBUTES

Commit-ID:  32ff77e8cc9e66cc4fb38098f64fd54cc8f54573
Gitweb:     https://git.kernel.org/tip/32ff77e8cc9e66cc4fb38098f64fd54cc8f54573
Author:     Milind Chabbi <[email protected]>
AuthorDate: Mon, 12 Mar 2018 14:45:47 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Tue, 13 Mar 2018 15:24:02 +0100

perf/core: Implement fast breakpoint modification via _IOC_MODIFY_ATTRIBUTES

Testing: tests posted at
https://github.com/linux-contrib/perf_event_modify_bp demonstrate the performance significance of this patch. Tests also check the functional correctness of the patch.

Signed-off-by Milind Chabbi <[email protected]>

[ Using modify_user_hw_breakpoint_check function. ]

[ Reformated PERF_EVENT_IOC_*, so the values are all in one column. ]

Signed-off-by Jiri Olsa <[email protected]>

Cc Alexander Shishkin <[email protected]>

Cc Arnaldo Carvalho de Melo <[email protected]>

Cc David Ahern <[email protected]>

Cc Frederic Weisbecker <[email protected]>

Cc Hari Bathini <[email protected]>

Cc Jin Yao <[email protected]>

Cc Jiri Olsa <[email protected]>

Cc Kan Liang <[email protected]>

Cc Linus Torvalds <[email protected]>

Cc Michael Ellerman <[email protected]>

Cc Namhyung Kim <[email protected]>

Cc Oleg Nesterov <[email protected]>

Cc Peter Zijlstra <[email protected]>

Cc Sukadev Bhattiprolu <[email protected]>

Cc Thomas Gleixner <[email protected]>

Cc Will Deacon <[email protected]>

Link http://lkml.kernel.org/r/[email protected] open here

Signed-off-by Ingo Molnar <[email protected]>

---
 include/linux/hw_breakpoint.h         |  7 +++++
 include/uapi/linux/perf_event.h       | 23 +++++++++--------
 kernel/events/core.c                  | 48 +++++++++++++++++++++++++++++++++++
 kernel/events/hw_breakpoint.c         |  2 +-
 tools/include/uapi/linux/perf_event.h | 23 +++++++++--------
 5 files changed, 80 insertions(+), 23 deletions(-)

include/linux/hw_breakpoint.h +7 -0

index cf045885a499..6058c3844a76 100644
@@ -53,6 +53,9 @@ register_user_hw_breakpoint(struct perf_event_attr *attr,
 /* FIXME: only change from the attr, and don't unregister */
 extern int
 modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *attr);
+extern int
+modify_user_hw_breakpoint_check(struct perf_event *bp, struct perf_event_attr *attr,
+				bool check);
 
 /*
  * Kernel breakpoints are not associated with any particular thread.

@@ -97,6 +100,10 @@ register_user_hw_breakpoint(struct perf_event_attr *attr,
 static inline int
 modify_user_hw_breakpoint(struct perf_event *bp,
 			  struct perf_event_attr *attr)	{ return -ENOSYS; }
+static inline int
+modify_user_hw_breakpoint_check(struct perf_event *bp, struct perf_event_attr *attr,
+				bool check)	{ return -ENOSYS; }
+
 static inline struct perf_event *
 register_wide_hw_breakpoint_cpu(struct perf_event_attr *attr,
 				perf_overflow_handler_t	 triggered,

include/uapi/linux/perf_event.h +12 ~~-11~~

index 6f873503552d..912b85b52344 100644
@@ -448,17 +448,18 @@ struct perf_event_query_bpf {
 /*
  * Ioctls that can be done on a perf event fd:
  */
-#define PERF_EVENT_IOC_ENABLE		_IO ('$', 0)
-#define PERF_EVENT_IOC_DISABLE		_IO ('$', 1)
-#define PERF_EVENT_IOC_REFRESH		_IO ('$', 2)
-#define PERF_EVENT_IOC_RESET		_IO ('$', 3)
-#define PERF_EVENT_IOC_PERIOD		_IOW('$', 4, __u64)
-#define PERF_EVENT_IOC_SET_OUTPUT	_IO ('$', 5)
-#define PERF_EVENT_IOC_SET_FILTER	_IOW('$', 6, char *)
-#define PERF_EVENT_IOC_ID		_IOR('$', 7, __u64 *)
-#define PERF_EVENT_IOC_SET_BPF		_IOW('$', 8, __u32)
-#define PERF_EVENT_IOC_PAUSE_OUTPUT	_IOW('$', 9, __u32)
-#define PERF_EVENT_IOC_QUERY_BPF	_IOWR('$', 10, struct perf_event_query_bpf *)
+#define PERF_EVENT_IOC_ENABLE			_IO ('$', 0)
+#define PERF_EVENT_IOC_DISABLE			_IO ('$', 1)
+#define PERF_EVENT_IOC_REFRESH			_IO ('$', 2)
+#define PERF_EVENT_IOC_RESET			_IO ('$', 3)
+#define PERF_EVENT_IOC_PERIOD			_IOW('$', 4, __u64)
+#define PERF_EVENT_IOC_SET_OUTPUT		_IO ('$', 5)
+#define PERF_EVENT_IOC_SET_FILTER		_IOW('$', 6, char *)
+#define PERF_EVENT_IOC_ID			_IOR('$', 7, __u64 *)
+#define PERF_EVENT_IOC_SET_BPF			_IOW('$', 8, __u32)
+#define PERF_EVENT_IOC_PAUSE_OUTPUT		_IOW('$', 9, __u32)
+#define PERF_EVENT_IOC_QUERY_BPF		_IOWR('$', 10, struct perf_event_query_bpf *)
+#define PERF_EVENT_IOC_MODIFY_ATTRIBUTES	_IOW('$', 11, struct perf_event_attr *)
 
 enum perf_event_ioc_flags {
 	PERF_IOC_FLAG_GROUP		= 1U << 0,

kernel/events/core.c +48 -0

66 lines not on this page · raw message

kernel/events/hw_breakpoint.c +1 -1

index 0c82663395f7..6253d5519cd8 100644
@@ -456,7 +456,7 @@ register_user_hw_breakpoint(struct perf_event_attr *attr,
 }
 EXPORT_SYMBOL_GPL(register_user_hw_breakpoint);
 
-static int
+int
 modify_user_hw_breakpoint_check(struct perf_event *bp, struct perf_event_attr *attr,
 			        bool check)
 {

tools/include/uapi/linux/perf_event.h +12 ~~-11~~

index 6f873503552d..912b85b52344 100644
@@ -448,17 +448,18 @@ struct perf_event_query_bpf {
 /*
  * Ioctls that can be done on a perf event fd:
  */
-#define PERF_EVENT_IOC_ENABLE		_IO ('$', 0)
-#define PERF_EVENT_IOC_DISABLE		_IO ('$', 1)
-#define PERF_EVENT_IOC_REFRESH		_IO ('$', 2)
-#define PERF_EVENT_IOC_RESET		_IO ('$', 3)
-#define PERF_EVENT_IOC_PERIOD		_IOW('$', 4, __u64)
-#define PERF_EVENT_IOC_SET_OUTPUT	_IO ('$', 5)
-#define PERF_EVENT_IOC_SET_FILTER	_IOW('$', 6, char *)
-#define PERF_EVENT_IOC_ID		_IOR('$', 7, __u64 *)
-#define PERF_EVENT_IOC_SET_BPF		_IOW('$', 8, __u32)
-#define PERF_EVENT_IOC_PAUSE_OUTPUT	_IOW('$', 9, __u32)
-#define PERF_EVENT_IOC_QUERY_BPF	_IOWR('$', 10, struct perf_event_query_bpf *)
+#define PERF_EVENT_IOC_ENABLE			_IO ('$', 0)
+#define PERF_EVENT_IOC_DISABLE			_IO ('$', 1)
+#define PERF_EVENT_IOC_REFRESH			_IO ('$', 2)
+#define PERF_EVENT_IOC_RESET			_IO ('$', 3)
+#define PERF_EVENT_IOC_PERIOD			_IOW('$', 4, __u64)
+#define PERF_EVENT_IOC_SET_OUTPUT		_IO ('$', 5)
+#define PERF_EVENT_IOC_SET_FILTER		_IOW('$', 6, char *)
+#define PERF_EVENT_IOC_ID			_IOR('$', 7, __u64 *)
+#define PERF_EVENT_IOC_SET_BPF			_IOW('$', 8, __u32)
+#define PERF_EVENT_IOC_PAUSE_OUTPUT		_IOW('$', 9, __u32)
+#define PERF_EVENT_IOC_QUERY_BPF		_IOWR('$', 10, struct perf_event_query_bpf *)
+#define PERF_EVENT_IOC_MODIFY_ATTRIBUTES	_IOW('$', 11, struct perf_event_attr *)
 
 enum perf_event_ioc_flags {
 	PERF_IOC_FLAG_GROUP		= 1U << 0,

Jiri Olsa

author reviewer

2018-03-12 13:46 UTC #16 lore raw

PATCH v3 8/8 perf tests: Add breakpoint accounting/modify test

In reply to Jiri Olsa

Adding test that:
- detects the number of watch/break-points, skip test if any is missing
- detects PERF_EVENT_IOC_MODIFY_ATTRIBUTES ioctl, skip test if it's missing
- detects if watchpoints and breakpoints share same slots
- create all possible watchpoints on cpu 0
- change one of it to breakpoint
- in case wp and bp do not share slots, we create another watchpoint to ensure the slot accounting is correct

Signed-off-by Jiri Olsa <[email protected]>

---
 tools/perf/tests/Build          |   1 +
 tools/perf/tests/bp_account.c   | 195 ++++++++++++++++++++++++++++++++++++++++
 tools/perf/tests/builtin-test.c |   4 +
 tools/perf/tests/tests.h        |   1 +
 4 files changed, 201 insertions(+)
 create mode 100644 tools/perf/tests/bp_account.c

tools/perf/tests/Build +1 -0

index 45782220ac23..6c108fa79ae3 100644
@@ -20,6 +20,7 @@ perf-y += hists_cumulate.o
 perf-y += python-use.o
 perf-y += bp_signal.o
 perf-y += bp_signal_overflow.o
+perf-y += bp_account.o
 perf-y += task-exit.o
 perf-y += sw-clock.o
 perf-y += mmap-thread-lookup.o

tools/perf/tests/bp_account.c new file +195 -0

195 lines not on this page · raw message

tools/perf/tests/builtin-test.c +4 -0

index 09071ef4434f..625f5a6772af 100644
@@ -116,6 +116,10 @@ static struct test generic_tests[] = {
 		.is_supported = test__bp_signal_is_supported,
 	},
 	{
+		.desc = "Breakpoint accounting",
+		.func = test__bp_accounting,
+	},
+	{
 		.desc = "Number of exit events of a simple workload",
 		.func = test__task_exit,
 	},

tools/perf/tests/tests.h +1 -0

index 2e169819e647..a9760e790563 100644
@@ -58,6 +58,7 @@ int test__hists_link(struct test *test, int subtest);
 int test__python_use(struct test *test, int subtest);
 int test__bp_signal(struct test *test, int subtest);
 int test__bp_signal_overflow(struct test *test, int subtest);
+int test__bp_accounting(struct test *test, int subtest);
 int test__task_exit(struct test *test, int subtest);
 int test__mem(struct test *test, int subtest);
 int test__sw_clock_freq(struct test *test, int subtest);

Signature · 1 line

2.13.6

tip-bot for Jiri Olsa

merge bot

2018-03-13 06:23 UTC #17 lore raw

[tip:perf/core] perf tests: Add breakpoint accounting/modify test

perf tests: Add breakpoint accounting/modify test

Commit-ID:  5acc575fec9f424f0b557844dbda8aed57baae1e
Gitweb:     https://git.kernel.org/tip/5acc575fec9f424f0b557844dbda8aed57baae1e
Author:     Jiri Olsa <[email protected]>
AuthorDate: Mon, 12 Mar 2018 14:45:48 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Tue, 13 Mar 2018 06:56:09 +0100

perf tests: Add breakpoint accounting/modify test

Adding test that:

- detects the number of watch/break-points, skip test if any is missing
- detects PERF_EVENT_IOC_MODIFY_ATTRIBUTES ioctl, skip test if it's missing
- detects if watchpoints and breakpoints share same slots
- create all possible watchpoints on cpu 0
- change one of it to breakpoint
- in case wp and bp do not share slots, we create another watchpoint to ensure the slot accounting is correct

Signed-off-by Jiri Olsa <[email protected]>

Cc Alexander Shishkin <[email protected]>

Cc Arnaldo Carvalho de Melo <[email protected]>

Cc David Ahern <[email protected]>

Cc Frederic Weisbecker <[email protected]>

Cc Hari Bathini <[email protected]>

Cc Jin Yao <[email protected]>

Cc Jiri Olsa <[email protected]>

Cc Kan Liang <[email protected]>

Cc Linus Torvalds <[email protected]>

Cc Michael Ellerman <[email protected]>

Cc Milind Chabbi <[email protected]>

Cc Namhyung Kim <[email protected]>

Cc Oleg Nesterov <[email protected]>

Cc Peter Zijlstra <[email protected]>

Cc Sukadev Bhattiprolu <[email protected]>

Cc Thomas Gleixner <[email protected]>

Cc Will Deacon <[email protected]>

Link http://lkml.kernel.org/r/[email protected] open here

Signed-off-by Ingo Molnar <[email protected]>

---
 tools/perf/tests/Build          |   1 +
 tools/perf/tests/bp_account.c   | 195 ++++++++++++++++++++++++++++++++++++++++
 tools/perf/tests/builtin-test.c |   4 +
 tools/perf/tests/tests.h        |   1 +
 4 files changed, 201 insertions(+)

tools/perf/tests/Build +1 -0

index 87bf3edb037c..62ca0174d5e1 100644
@@ -20,6 +20,7 @@ perf-y += hists_cumulate.o
 perf-y += python-use.o
 perf-y += bp_signal.o
 perf-y += bp_signal_overflow.o
+perf-y += bp_account.o
 perf-y += task-exit.o
 perf-y += sw-clock.o
 perf-y += mmap-thread-lookup.o

tools/perf/tests/bp_account.c new file +195 -0

195 lines not on this page · raw message

tools/perf/tests/builtin-test.c +4 -0

index fafa014240cd..38bf109ce106 100644
@@ -115,6 +115,10 @@ static struct test generic_tests[] = {
 		.func = test__bp_signal_overflow,
 		.is_supported = test__bp_signal_is_supported,
 	},
+	{
+		.desc = "Breakpoint accounting",
+		.func = test__bp_accounting,
+	},
 	{
 		.desc = "Number of exit events of a simple workload",
 		.func = test__task_exit,

tools/perf/tests/tests.h +1 -0

index 2862b80bc288..9f51edac44ae 100644
@@ -58,6 +58,7 @@ int test__hists_link(struct test *test, int subtest);
 int test__python_use(struct test *test, int subtest);
 int test__bp_signal(struct test *test, int subtest);
 int test__bp_signal_overflow(struct test *test, int subtest);
+int test__bp_accounting(struct test *test, int subtest);
 int test__task_exit(struct test *test, int subtest);
 int test__mem(struct test *test, int subtest);
 int test__sw_clock_freq(struct test *test, int subtest);

tip-bot for Jiri Olsa

merge bot

2018-03-13 15:14 UTC #18 lore raw

[tip:perf/core] perf tests: Add breakpoint accounting/modify test

In reply to Jiri Olsa

perf tests: Add breakpoint accounting/modify test

Commit-ID:  032db28e5fa3594dfa3df585c54d8b612657f537
Gitweb:     https://git.kernel.org/tip/032db28e5fa3594dfa3df585c54d8b612657f537
Author:     Jiri Olsa <[email protected]>
AuthorDate: Mon, 12 Mar 2018 14:45:48 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Tue, 13 Mar 2018 15:23:37 +0100

perf tests: Add breakpoint accounting/modify test

Adding test that:

Signed-off-by Jiri Olsa <[email protected]>

Cc Alexander Shishkin <[email protected]>

Cc Arnaldo Carvalho de Melo <[email protected]>

Cc David Ahern <[email protected]>

Cc Frederic Weisbecker <[email protected]>

Cc Hari Bathini <[email protected]>

Cc Jin Yao <[email protected]>

Cc Jiri Olsa <[email protected]>

Cc Kan Liang <[email protected]>

Cc Linus Torvalds <[email protected]>

Cc Michael Ellerman <[email protected]>

Cc Milind Chabbi <[email protected]>

Cc Namhyung Kim <[email protected]>

Cc Oleg Nesterov <[email protected]>

Cc Peter Zijlstra <[email protected]>

Cc Sukadev Bhattiprolu <[email protected]>

Cc Thomas Gleixner <[email protected]>

Cc Will Deacon <[email protected]>

Link http://lkml.kernel.org/r/[email protected] open here

Signed-off-by Ingo Molnar <[email protected]>

---
 tools/perf/tests/Build          |   1 +
 tools/perf/tests/bp_account.c   | 195 ++++++++++++++++++++++++++++++++++++++++
 tools/perf/tests/builtin-test.c |   4 +
 tools/perf/tests/tests.h        |   1 +
 4 files changed, 201 insertions(+)

tools/perf/tests/Build +1 -0

index 87bf3edb037c..62ca0174d5e1 100644
@@ -20,6 +20,7 @@ perf-y += hists_cumulate.o
 perf-y += python-use.o
 perf-y += bp_signal.o
 perf-y += bp_signal_overflow.o
+perf-y += bp_account.o
 perf-y += task-exit.o
 perf-y += sw-clock.o
 perf-y += mmap-thread-lookup.o

tools/perf/tests/bp_account.c new file +195 -0

195 lines not on this page · raw message

tools/perf/tests/builtin-test.c +4 -0

index fafa014240cd..38bf109ce106 100644
@@ -115,6 +115,10 @@ static struct test generic_tests[] = {
 		.func = test__bp_signal_overflow,
 		.is_supported = test__bp_signal_is_supported,
 	},
+	{
+		.desc = "Breakpoint accounting",
+		.func = test__bp_accounting,
+	},
 	{
 		.desc = "Number of exit events of a simple workload",
 		.func = test__task_exit,

tools/perf/tests/tests.h +1 -0

index 2862b80bc288..9f51edac44ae 100644
@@ -58,6 +58,7 @@ int test__hists_link(struct test *test, int subtest);
 int test__python_use(struct test *test, int subtest);
 int test__bp_signal(struct test *test, int subtest);
 int test__bp_signal_overflow(struct test *test, int subtest);
+int test__bp_accounting(struct test *test, int subtest);
 int test__task_exit(struct test *test, int subtest);
 int test__mem(struct test *test, int subtest);
 int test__sw_clock_freq(struct test *test, int subtest);

Ingo Molnar

2018-03-13 06:37 UTC #19 lore raw

In reply to Jiri Olsa

Quoted 12 lines · Jiri Olsa

Jiri Olsa (7):
hw_breakpoint: Pass bp_type directly as find_slot_idx argument
hw_breakpoint: Pass bp_type argument to __reserve_bp_slot|__release_bp_slot
hw_breakpoint: Add modify_bp_slot function
hw_breakpoint: Factor out __modify_user_hw_breakpoint function
hw_breakpoint: Add perf_event_attr fields check in __modify_user_hw_breakpoint perf/core: Move perf_event_attr::sample_max_stack into perf_copy_attr perf tests: Add breakpoint accounting/modify test
Milind Chabbi (1):
perf/core: fast breakpoint modification via _IOC_MODIFY_ATTRIBUTES.

Quoted 11 lines

 include/linux/hw_breakpoint.h         |   7 +++++
 include/uapi/linux/perf_event.h       |  23 ++++++++-------
 kernel/events/core.c                  |  54 ++++++++++++++++++++++++++++++++--
 kernel/events/hw_breakpoint.c         | 115 +++++++++++++++++++++++++++++++++++++++++++++++++++---------------------
 tools/include/uapi/linux/perf_event.h |  23 ++++++++-------
 tools/perf/tests/Build                |   1 +
 tools/perf/tests/bp_account.c         | 195 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 tools/perf/tests/builtin-test.c       |   4 +++
 tools/perf/tests/tests.h              |   1 +
 9 files changed, 365 insertions(+), 58 deletions(-)
 create mode 100644 tools/perf/tests/bp_account.c

Note, there's a new PARISC build failure:

home/mingo/tip/kernel/events/core.c:2858:2: error: implicit declaration of function 'modify_user_hw_breakpoint_check' [-Werror=implicit-function-declaration]
err = modify_user_hw_breakpoint_check(bp, attr, true);

will only be able to push it out to -next once this is fixed.

Thanks,

Ingo

Jiri Olsa

2018-03-13 09:28 UTC #20 lore raw

Quoted 27 lines · Ingo Molnar

Quoted 12 lines · Jiri Olsa
Jiri Olsa (7):
hw_breakpoint: Pass bp_type directly as find_slot_idx argument
hw_breakpoint: Pass bp_type argument to __reserve_bp_slot|__release_bp_slot
hw_breakpoint: Add modify_bp_slot function
hw_breakpoint: Factor out __modify_user_hw_breakpoint function
hw_breakpoint: Add perf_event_attr fields check in __modify_user_hw_breakpoint perf/core: Move perf_event_attr::sample_max_stack into perf_copy_attr perf tests: Add breakpoint accounting/modify test
Milind Chabbi (1):
perf/core: fast breakpoint modification via _IOC_MODIFY_ATTRIBUTES.
Quoted 11 lines
 include/linux/hw_breakpoint.h         |   7 +++++
 include/uapi/linux/perf_event.h       |  23 ++++++++-------
 kernel/events/core.c                  |  54 ++++++++++++++++++++++++++++++++--
 kernel/events/hw_breakpoint.c         | 115 +++++++++++++++++++++++++++++++++++++++++++++++++++---------------------
 tools/include/uapi/linux/perf_event.h |  23 ++++++++-------
 tools/perf/tests/Build                |   1 +
 tools/perf/tests/bp_account.c         | 195 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 tools/perf/tests/builtin-test.c       |   4 +++
 tools/perf/tests/tests.h              |   1 +
 9 files changed, 365 insertions(+), 58 deletions(-)
 create mode 100644 tools/perf/tests/bp_account.c

Quoted 6 lines

Note, there's a new PARISC build failure:
home/mingo/tip/kernel/events/core.c:2858:2: error: implicit declaration of function 'modify_user_hw_breakpoint_check' [-Werror=implicit-function-declaration]
err = modify_user_hw_breakpoint_check(bp, attr, true);
will only be able to push it out to -next once this is fixed.

ok, checking on that

jirka

Jiri Olsa

2018-03-13 09:50 UTC #21 lore raw

Quoted 28 lines · Jiri Olsa

Quoted 27 lines · Ingo Molnar
Quoted 23 lines · Jiri Olsa
Jiri Olsa (7):
hw_breakpoint: Pass bp_type directly as find_slot_idx argument
hw_breakpoint: Pass bp_type argument to __reserve_bp_slot|__release_bp_slot
hw_breakpoint: Add modify_bp_slot function
hw_breakpoint: Factor out __modify_user_hw_breakpoint function
hw_breakpoint: Add perf_event_attr fields check in __modify_user_hw_breakpoint perf/core: Move perf_event_attr::sample_max_stack into perf_copy_attr perf tests: Add breakpoint accounting/modify test
Milind Chabbi (1):
perf/core: fast breakpoint modification via _IOC_MODIFY_ATTRIBUTES.
 include/linux/hw_breakpoint.h         |   7 +++++
 include/uapi/linux/perf_event.h       |  23 ++++++++-------
 kernel/events/core.c                  |  54 ++++++++++++++++++++++++++++++++--
 kernel/events/hw_breakpoint.c         | 115 +++++++++++++++++++++++++++++++++++++++++++++++++++---------------------
 tools/include/uapi/linux/perf_event.h |  23 ++++++++-------
 tools/perf/tests/Build                |   1 +
 tools/perf/tests/bp_account.c         | 195 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 tools/perf/tests/builtin-test.c       |   4 +++
 tools/perf/tests/tests.h              |   1 +
 9 files changed, 365 insertions(+), 58 deletions(-)
 create mode 100644 tools/perf/tests/bp_account.c

Quoted 8 lines

Quoted 6 lines
Note, there's a new PARISC build failure:
home/mingo/tip/kernel/events/core.c:2858:2: error: implicit declaration of function 'modify_user_hw_breakpoint_check' [-Werror=implicit-function-declaration]
err = modify_user_hw_breakpoint_check(bp, attr, true);
will only be able to push it out to -next once this is fixed.
ok, checking on that

should be this one.. I'm checking on s390 which is also without breakpoint support, I'll send full patch after

jirka

---

include/linux/hw_breakpoint.h +2 -2

index abeba094f080..6058c3844a76 100644
@@ -101,8 +101,8 @@ static inline int
 modify_user_hw_breakpoint(struct perf_event *bp,
 			  struct perf_event_attr *attr)	{ return -ENOSYS; }
 static inline int
-__modify_user_hw_breakpoint(struct perf_event *bp,
-			    struct perf_event_attr *attr) { return -ENOSYS; }
+modify_user_hw_breakpoint_check(struct perf_event *bp, struct perf_event_attr *attr,
+				bool check)	{ return -ENOSYS; }
 
 static inline struct perf_event *
 register_wide_hw_breakpoint_cpu(struct perf_event_attr *attr,

Jiri Olsa

2018-03-14 08:45 UTC #22 lore raw

PATCH 1/1 hw_breakpoint: Fix build for disabled CONFIG_HAVE_HW_BREAKPOINT

Quoted 34 lines · Jiri Olsa

Quoted 33 lines · Jiri Olsa
Quoted 32 lines · Ingo Molnar
Quoted 12 lines · Jiri Olsa
Jiri Olsa (7):
hw_breakpoint: Pass bp_type directly as find_slot_idx argument
hw_breakpoint: Pass bp_type argument to __reserve_bp_slot|__release_bp_slot
hw_breakpoint: Add modify_bp_slot function
hw_breakpoint: Factor out __modify_user_hw_breakpoint function
hw_breakpoint: Add perf_event_attr fields check in __modify_user_hw_breakpoint perf/core: Move perf_event_attr::sample_max_stack into perf_copy_attr perf tests: Add breakpoint accounting/modify test
Milind Chabbi (1):
perf/core: fast breakpoint modification via _IOC_MODIFY_ATTRIBUTES.
Quoted 11 lines
 include/linux/hw_breakpoint.h         |   7 +++++
 include/uapi/linux/perf_event.h       |  23 ++++++++-------
 kernel/events/core.c                  |  54 ++++++++++++++++++++++++++++++++--
 kernel/events/hw_breakpoint.c         | 115 +++++++++++++++++++++++++++++++++++++++++++++++++++---------------------
 tools/include/uapi/linux/perf_event.h |  23 ++++++++-------
 tools/perf/tests/Build                |   1 +
 tools/perf/tests/bp_account.c         | 195 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 tools/perf/tests/builtin-test.c       |   4 +++
 tools/perf/tests/tests.h              |   1 +
 9 files changed, 365 insertions(+), 58 deletions(-)
 create mode 100644 tools/perf/tests/bp_account.c
Note, there's a new PARISC build failure:
home/mingo/tip/kernel/events/core.c:2858:2: error: implicit declaration of function 'modify_user_hw_breakpoint_check' [-Werror=implicit-function-declaration]
err = modify_user_hw_breakpoint_check(bp, attr, true);

Quoted 6 lines

Quoted 3 lines
Quoted 1 line
will only be able to push it out to -next once this is fixed.
ok, checking on that
should be this one.. I'm checking on s390 which is also without breakpoint support, I'll send full patch after

and here it is

thanks,
jirka

---
We're missing stub for modify_user_hw_breakpoint_check function when option CONFIG_HAVE_HW_BREAKPOINT is disabled. It was mixed up in recent changes.

Fixes f30b09b7f8ae ("perf/core: Implement fast breakpoint modification via _IOC_MODIFY_ATTRIBUTES")

Signed-off-by Jiri Olsa <[email protected]>

---
 include/linux/hw_breakpoint.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

include/linux/hw_breakpoint.h +2 -2

index abeba094f080..6058c3844a76 100644
@@ -101,8 +101,8 @@ static inline int
 modify_user_hw_breakpoint(struct perf_event *bp,
 			  struct perf_event_attr *attr)	{ return -ENOSYS; }
 static inline int
-__modify_user_hw_breakpoint(struct perf_event *bp,
-			    struct perf_event_attr *attr) { return -ENOSYS; }
+modify_user_hw_breakpoint_check(struct perf_event *bp, struct perf_event_attr *attr,
+				bool check)	{ return -ENOSYS; }
 
 static inline struct perf_event *
 register_wide_hw_breakpoint_cpu(struct perf_event_attr *attr,

Signature · 1 line

2.13.6

Ingo Molnar

2018-03-14 12:28 UTC #23 lore raw

Re: [PATCH] hw_breakpoint: Fix build for disabled CONFIG_HAVE_HW_BREAKPOINT

Quoted 4 lines · Jiri Olsa

Quoted 2 lines
should be this one.. I'm checking on s390 which is also without breakpoint support, I'll send full patch after
and here it is

Yeah, note that I back-merged this into the series so there's no separate commit for the fix.

Thanks,

Ingo