[PATCH 1/2] perf: Add persistent event facilities

From: Borislav Petkov
Date: Wed Mar 21 2012 - 10:35:15 EST


From: Borislav Petkov <borislav.petkov@xxxxxxx>

Add a barebones implementation for registering persistent events with
perf. For that, we don't destroy the buffers when they're unmapped;
also, we map them read-only so that multiple agents can access them.

Signed-off-by: Borislav Petkov <borislav.petkov@xxxxxxx>
---
include/linux/perf_event.h | 24 ++++++++++-
kernel/events/Makefile | 2 +-
kernel/events/core.c | 25 +++++++++--
kernel/events/internal.h | 2 +
kernel/events/persistent.c | 98 ++++++++++++++++++++++++++++++++++++++++++++
5 files changed, 145 insertions(+), 6 deletions(-)
create mode 100644 kernel/events/persistent.c

diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index abb2776be1ba..9743ba46e12d 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -223,8 +223,9 @@ struct perf_event_attr {

exclude_host : 1, /* don't count in host */
exclude_guest : 1, /* don't count in guest */
+ persistent : 1, /* always-on event */

- __reserved_1 : 43;
+ __reserved_1 : 42;

union {
__u32 wakeup_events; /* wakeup every n events */
@@ -944,6 +945,14 @@ struct perf_output_handle {
int page;
};

+/* Persistent events descriptor */
+struct pers_event_desc {
+ struct perf_event_attr *pattr;
+ struct perf_event *event;
+ const char *dfs_name;
+ struct dentry *dfs_entry;
+};
+
#ifdef CONFIG_PERF_EVENTS

extern int perf_pmu_register(struct pmu *pmu, char *name, int type);
@@ -1149,6 +1158,11 @@ extern void perf_swevent_put_recursion_context(int rctx);
extern void perf_event_enable(struct perf_event *event);
extern void perf_event_disable(struct perf_event *event);
extern void perf_event_task_tick(void);
+extern int perf_add_persistent_on_cpu(int cpu, struct pers_event_desc *desc,
+ struct dentry *dfs_dr, unsigned nr_pages);
+extern void perf_rm_persistent_on_cpu(int cpu, struct pers_event_desc *desc);
+extern int perf_persistent_open(struct inode *inode, struct file *file);
+extern const struct file_operations perf_pers_fops;
#else
static inline void
perf_event_task_sched_in(struct task_struct *prev,
@@ -1187,6 +1201,14 @@ static inline void perf_swevent_put_recursion_context(int rctx) { }
static inline void perf_event_enable(struct perf_event *event) { }
static inline void perf_event_disable(struct perf_event *event) { }
static inline void perf_event_task_tick(void) { }
+extern int perf_add_persistent_on_cpu(int cpu, struct pers_event_desc *desc,
+ struct perf_event_attr *pattr,
+ struct dentry *dfs_dir,
+ unsigned nr_pages); { return -EINVAL; }
+extern void perf_rm_persistent_on_cpu(int cpu,
+ struct pers_event_desc *desc) { }
+static inline int
+perf_persistent_open(struct inode *inode, struct file *file) { return -1; }
#endif

#define perf_output_put(handle, x) perf_output_copy((handle), &(x), sizeof(x))
diff --git a/kernel/events/Makefile b/kernel/events/Makefile
index 22d901f9caf4..0d2267fb4336 100644
--- a/kernel/events/Makefile
+++ b/kernel/events/Makefile
@@ -2,5 +2,5 @@ ifdef CONFIG_FUNCTION_TRACER
CFLAGS_REMOVE_core.o = -pg
endif

-obj-y := core.o ring_buffer.o callchain.o
+obj-y := core.o ring_buffer.o callchain.o persistent.o
obj-$(CONFIG_HAVE_HW_BREAKPOINT) += hw_breakpoint.o
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 1b5c081d8b9f..9a39dc6755f4 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -2770,8 +2770,6 @@ static void free_event_rcu(struct rcu_head *head)
kfree(event);
}

-static void ring_buffer_put(struct ring_buffer *rb);
-
static void free_event(struct perf_event *event)
{
irq_work_sync(&event->pending);
@@ -3423,7 +3421,7 @@ static struct ring_buffer *ring_buffer_get(struct perf_event *event)
return rb;
}

-static void ring_buffer_put(struct ring_buffer *rb)
+void ring_buffer_put(struct ring_buffer *rb)
{
struct perf_event *event, *n;
unsigned long flags;
@@ -3452,6 +3450,11 @@ static void perf_mmap_close(struct vm_area_struct *vma)
{
struct perf_event *event = vma->vm_file->private_data;

+ if (event->attr.persistent) {
+ atomic_dec(&event->mmap_count);
+ return;
+ }
+
if (atomic_dec_and_mutex_lock(&event->mmap_count, &event->mmap_mutex)) {
unsigned long size = perf_data_size(event->rb);
struct user_struct *user = event->mmap_user;
@@ -3495,7 +3498,7 @@ static int perf_mmap(struct file *file, struct vm_area_struct *vma)
if (event->cpu == -1 && event->attr.inherit)
return -EINVAL;

- if (!(vma->vm_flags & VM_SHARED))
+ if (!(vma->vm_flags & VM_SHARED) && !event->attr.persistent)
return -EINVAL;

vma_size = vma->vm_end - vma->vm_start;
@@ -3606,6 +3609,16 @@ static const struct file_operations perf_fops = {
.fasync = perf_fasync,
};

+const struct file_operations perf_pers_fops = {
+ .llseek = no_llseek,
+ .open = perf_persistent_open,
+ .poll = perf_poll,
+ .unlocked_ioctl = perf_ioctl,
+ .compat_ioctl = perf_ioctl,
+ .mmap = perf_mmap,
+ .fasync = perf_fasync,
+};
+
/*
* Perf event wakeup
*
@@ -6319,6 +6332,10 @@ __perf_event_exit_task(struct perf_event *child_event,
raw_spin_unlock_irq(&child_ctx->lock);
}

+ /* do not remove persistent events on task exit */
+ if (child_event->attr.persistent)
+ return;
+
perf_remove_from_context(child_event);

/*
diff --git a/kernel/events/internal.h b/kernel/events/internal.h
index b0b107f90afc..d4e6181d5faf 100644
--- a/kernel/events/internal.h
+++ b/kernel/events/internal.h
@@ -37,6 +37,8 @@ struct ring_buffer {
extern void rb_free(struct ring_buffer *rb);
extern struct ring_buffer *
rb_alloc(int nr_pages, long watermark, int cpu, int flags);
+extern void ring_buffer_put(struct ring_buffer *rb);
+
extern void perf_event_wakeup(struct perf_event *event);

extern void
diff --git a/kernel/events/persistent.c b/kernel/events/persistent.c
new file mode 100644
index 000000000000..e6f146532e34
--- /dev/null
+++ b/kernel/events/persistent.c
@@ -0,0 +1,98 @@
+#include <linux/debugfs.h>
+#include <linux/perf_event.h>
+
+#include "internal.h"
+
+/*
+ * Create and enable the perf event described by @attr.
+ *
+ * Returns the @event pointer which receives the allocated event from
+ * perf on success. Make sure to check return code before touching @event
+ * any further.
+ *
+ * @attr: perf attr template
+ * @cpu: on which cpu
+ * @nr_pages: perf buffer size in pages
+ *
+ */
+static struct perf_event *perf_add_persistent(struct perf_event_attr *attr,
+ int cpu, unsigned nr_pages)
+{
+ struct ring_buffer *buffer;
+ struct perf_event *event;
+
+ event = perf_event_create_kernel_counter(attr, cpu, NULL, NULL, NULL);
+ if (IS_ERR(event))
+ return event;
+
+ buffer = rb_alloc(nr_pages, 0, cpu, RING_BUFFER_WRITABLE);
+ if (!buffer)
+ goto err;
+
+ rcu_assign_pointer(event->rb, buffer);
+ perf_event_enable(event);
+
+ return event;
+
+err:
+ perf_event_release_kernel(event);
+ return ERR_PTR(-EINVAL);
+}
+
+static void perf_rm_persistent(struct perf_event *event, int cpu)
+{
+ if (!event)
+ return;
+
+ perf_event_disable(event);
+
+ if (event->rb) {
+ ring_buffer_put(event->rb);
+ rcu_assign_pointer(event->rb, NULL);
+ }
+
+ perf_event_release_kernel(event);
+}
+
+static struct dentry *perf_add_debugfs_entry(struct pers_event_desc *d,
+ struct dentry *dir)
+{
+ return debugfs_create_file(d->dfs_name, S_IRUGO | S_IWUSR,
+ dir, d->event, &perf_pers_fops);
+}
+
+int perf_add_persistent_on_cpu(int cpu, struct pers_event_desc *desc,
+ struct dentry *dir, unsigned nr_pages)
+{
+ int err = -EINVAL;
+
+ desc->event = perf_add_persistent(desc->pattr, cpu, nr_pages);
+ if (IS_ERR(desc->event)) {
+ printk(KERN_ERR "Error enabling persistent event on cpu %d\n", cpu);
+ return err;
+ }
+
+ desc->dfs_entry = perf_add_debugfs_entry(desc, dir);
+ if (!desc->dfs_entry) {
+ printk(KERN_ERR "Error adding debugfs entry on cpu %d\n", cpu);
+ goto disable;
+ }
+ return 0;
+
+disable:
+ perf_rm_persistent(desc->event, cpu);
+ return err;
+}
+
+void perf_rm_persistent_on_cpu(int cpu, struct pers_event_desc *desc)
+{
+ debugfs_remove(desc->dfs_entry);
+ perf_rm_persistent(desc->event, cpu);
+}
+
+int perf_persistent_open(struct inode *inode, struct file *file)
+{
+ file->private_data = inode->i_private;
+
+ return 0;
+}
--
1.7.9.3.362.g71319

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/