Commit-ID: a4be7c2778d1fd8e0a8a5607bec91fa221ab2c91
Gitweb: http://git.kernel.org/tip/a4be7c2778d1fd8e0a8a5607bec91fa221ab2c91
Author: Peter Zijlstra <a.p.zijlstra@xxxxxxxxx>
AuthorDate: Wed, 19 Aug 2009 11:18:27 +0200
Committer: Ingo Molnar <mingo@xxxxxxx>
CommitDate: Tue, 25 Aug 2009 09:36:13 +0200
perf_counter: Allow sharing of output channels
Provide the ability to configure a counter to send its output
to another (already existing) counter's output stream.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@xxxxxxxxx>
Cc: Arnaldo Carvalho de Melo <acme@xxxxxxxxxx>
Cc: Frederic Weisbecker <fweisbec@xxxxxxxxx>
Cc: Mike Galbraith <efault@xxxxxx>
Cc: stephane eranian <eranian@xxxxxxxxxxxxxx>
Cc: Paul Mackerras <paulus@xxxxxxxxx>
LKML-Reference: <20090819092023.980284148@xxxxxxxxx>
Signed-off-by: Ingo Molnar <mingo@xxxxxxx>
---
include/linux/perf_counter.h | 5 +++
kernel/perf_counter.c | 83 ++++++++++++++++++++++++++++++++++++++++--
2 files changed, 85 insertions(+), 3 deletions(-)
diff --git a/include/linux/perf_counter.h b/include/linux/perf_counter.h
index b53f700..e022b84 100644
--- a/include/linux/perf_counter.h
+++ b/include/linux/perf_counter.h
@@ -216,6 +216,7 @@ struct perf_counter_attr {
#define PERF_COUNTER_IOC_REFRESH _IO ('$', 2)
#define PERF_COUNTER_IOC_RESET _IO ('$', 3)
#define PERF_COUNTER_IOC_PERIOD _IOW('$', 4, u64)
+#define PERF_COUNTER_IOC_SET_OUTPUT _IO ('$', 5)
enum perf_counter_ioc_flags {
PERF_IOC_FLAG_GROUP = 1U << 0,
@@ -415,6 +416,9 @@ enum perf_callchain_context {
PERF_CONTEXT_MAX = (__u64)-4095,
};
+#define PERF_FLAG_FD_NO_GROUP (1U << 0)
+#define PERF_FLAG_FD_OUTPUT (1U << 1)
+
#ifdef __KERNEL__
/*
* Kernel-internal data types and definitions:
@@ -536,6 +540,7 @@ struct perf_counter {
struct list_head sibling_list;
int nr_siblings;
struct perf_counter *group_leader;
+ struct perf_counter *output;
const struct pmu *pmu;
enum perf_counter_active_state state;
diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c
index 06bf6a4..53abcbe 100644
--- a/kernel/perf_counter.c
+++ b/kernel/perf_counter.c
@@ -1692,6 +1692,11 @@ static void free_counter(struct perf_counter *counter)
atomic_dec(&nr_task_counters);
}
+ if (counter->output) {
+ fput(counter->output->filp);
+ counter->output = NULL;
+ }
+
if (counter->destroy)
counter->destroy(counter);
@@ -1977,6 +1982,8 @@ unlock:
return ret;
}
+int perf_counter_set_output(struct perf_counter *counter, int output_fd);
+
static long perf_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
{
struct perf_counter *counter = file->private_data;
@@ -2000,6 +2007,9 @@ static long perf_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
case PERF_COUNTER_IOC_PERIOD:
return perf_counter_period(counter, (u64 __user *)arg);
+ case PERF_COUNTER_IOC_SET_OUTPUT:
+ return perf_counter_set_output(counter, arg);
+
default:
return -ENOTTY;
}
@@ -2270,6 +2280,11 @@ static int perf_mmap(struct file *file, struct vm_area_struct *vma)
WARN_ON_ONCE(counter->ctx->parent_ctx);
mutex_lock(&counter->mmap_mutex);
+ if (counter->output) {
+ ret = -EINVAL;
+ goto unlock;
+ }
+
if (atomic_inc_not_zero(&counter->mmap_count)) {
if (nr_pages != counter->data->nr_pages)
ret = -EINVAL;
@@ -2655,6 +2670,7 @@ static int perf_output_begin(struct perf_output_handle *handle,
struct perf_counter *counter, unsigned int size,
int nmi, int sample)
{
+ struct perf_counter *output_counter;
struct perf_mmap_data *data;
unsigned int offset, head;
int have_lost;
@@ -2664,13 +2680,17 @@ static int perf_output_begin(struct perf_output_handle *handle,
u64 lost;
} lost_event;
+ rcu_read_lock();
/*
* For inherited counters we send all the output towards the parent.
*/
if (counter->parent)
counter = counter->parent;
- rcu_read_lock();
+ output_counter = rcu_dereference(counter->output);
+ if (output_counter)
+ counter = output_counter;
+
data = rcu_dereference(counter->data);
if (!data)
goto out;
@@ -4218,6 +4238,57 @@ err_size:
goto out;
}
+int perf_counter_set_output(struct perf_counter *counter, int output_fd)
+{
+ struct perf_counter *output_counter = NULL;
+ struct file *output_file = NULL;
+ struct perf_counter *old_output;
+ int fput_needed = 0;
+ int ret = -EINVAL;
+
+ if (!output_fd)
+ goto set;
+
+ output_file = fget_light(output_fd, &fput_needed);
+ if (!output_file)
+ return -EBADF;
+
+ if (output_file->f_op != &perf_fops)
+ goto out;
+
+ output_counter = output_file->private_data;
+
+ /* Don't chain output fds */
+ if (output_counter->output)
+ goto out;
+
+ /* Don't set an output fd when we already have an output channel */
+ if (counter->data)
+ goto out;
+
+ atomic_long_inc(&output_file->f_count);
+
+set:
+ mutex_lock(&counter->mmap_mutex);
+ old_output = counter->output;
+ rcu_assign_pointer(counter->output, output_counter);
+ mutex_unlock(&counter->mmap_mutex);
+
+ if (old_output) {
+ /*
+ * we need to make sure no existing perf_output_*()
+ * is still referencing this counter.
+ */
+ synchronize_rcu();
+ fput(old_output->filp);
+ }
+
+ ret = 0;
+out:
+ fput_light(output_file, fput_needed);
+ return ret;
+}
+
/**
* sys_perf_counter_open - open a performance counter, associate it to a task/cpu
*
@@ -4240,7 +4311,7 @@ SYSCALL_DEFINE5(perf_counter_open,
int ret;
/* for future expandability... */
- if (flags)
+ if (flags & ~(PERF_FLAG_FD_NO_GROUP | PERF_FLAG_FD_OUTPUT))
return -EINVAL;
ret = perf_copy_attr(attr_uptr, &attr);
@@ -4268,7 +4339,7 @@ SYSCALL_DEFINE5(perf_counter_open,
* Look up the group leader (we will attach this counter to it):
*/
group_leader = NULL;
- if (group_fd != -1) {
+ if (group_fd != -1 && !(flags & PERF_FLAG_FD_NO_GROUP)) {
ret = -EINVAL;
group_file = fget_light(group_fd, &fput_needed);
if (!group_file)
@@ -4310,6 +4381,12 @@ SYSCALL_DEFINE5(perf_counter_open,
if (!counter_file)
goto err_free_put_context;
+ if (flags & PERF_FLAG_FD_OUTPUT) {
+ ret = perf_counter_set_output(counter, group_fd);
+ if (ret)
+ goto err_free_put_context;
+ }
+
counter->filp = counter_file;
WARN_ON_ONCE(ctx->parent_ctx);
mutex_lock(&ctx->mutex);
--
To unsubscribe from this list: send the line "unsubscribe linux-tip-commits" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
|