[PATCH 4/4] perf/ring-buffer: Use regular variables for nesting

From: Peter Zijlstra
Date: Fri May 17 2019 - 07:57:11 EST


While the IRQ/NMI will nest, the nest-count will be invariant over the
actual exception, since it will decrement equal to increment.

This means we can -- carefully -- use a regular variable since the
typical LOAD-STORE race doesn't exist (similar to preempt_count).

This optimizes the ring-buffer for all LOAD-STORE architectures, since
they need to use atomic ops to implement local_t.

Suggested-by: Alexander Shishkin <alexander.shishkin@xxxxxxxxxxxxxxx>
Signed-off-by: Peter Zijlstra (Intel) <peterz@xxxxxxxxxxxxx>
---
kernel/events/internal.h | 4 ++--
kernel/events/ring_buffer.c | 41 ++++++++++++++++++++++++++---------------
2 files changed, 28 insertions(+), 17 deletions(-)

--- a/kernel/events/internal.h
+++ b/kernel/events/internal.h
@@ -24,7 +24,7 @@ struct ring_buffer {
atomic_t poll; /* POLL_ for wakeups */

local_t head; /* write position */
- local_t nest; /* nested writers */
+ unsigned int nest; /* nested writers */
local_t events; /* event limit */
local_t wakeup; /* wakeup stamp */
local_t lost; /* nr records lost */
@@ -41,7 +41,7 @@ struct ring_buffer {

/* AUX area */
long aux_head;
- local_t aux_nest;
+ unsigned int aux_nest;
long aux_wakeup; /* last aux_watermark boundary crossed by aux_head */
unsigned long aux_pgoff;
int aux_nr_pages;
--- a/kernel/events/ring_buffer.c
+++ b/kernel/events/ring_buffer.c
@@ -38,7 +38,12 @@ static void perf_output_get_handle(struc
struct ring_buffer *rb = handle->rb;

preempt_disable();
- local_inc(&rb->nest);
+
+ /*
+ * Avoid an explicit LOAD/STORE such that architectures with memops
+ * can use them.
+ */
+ (*(volatile unsigned int *)&rb->nest)++;
handle->wakeup = local_read(&rb->wakeup);
}

@@ -46,6 +51,17 @@ static void perf_output_put_handle(struc
{
struct ring_buffer *rb = handle->rb;
unsigned long head;
+ unsigned int nest;
+
+ /*
+ * If this isn't the outermost nesting, we don't have to update
+ * @rb->user_page->data_head.
+ */
+ nest = READ_ONCE(rb->nest);
+ if (nest > 1) {
+ WRITE_ONCE(rb->nest, nest - 1);
+ goto out;
+ }

again:
/*
@@ -65,15 +81,6 @@ static void perf_output_put_handle(struc
*/

/*
- * If this isn't the outermost nesting, we don't have to update
- * @rb->user_page->data_head.
- */
- if (local_read(&rb->nest) > 1) {
- local_dec(&rb->nest);
- goto out;
- }
-
- /*
* Since the mmap() consumer (userspace) can run on a different CPU:
*
* kernel user
@@ -108,7 +115,7 @@ static void perf_output_put_handle(struc
* write will (temporarily) publish a stale value.
*/
barrier();
- local_set(&rb->nest, 0);
+ WRITE_ONCE(rb->nest, 0);

/*
* Ensure we decrement @rb->nest before we validate the @rb->head.
@@ -116,7 +123,7 @@ static void perf_output_put_handle(struc
*/
barrier();
if (unlikely(head != local_read(&rb->head))) {
- local_inc(&rb->nest);
+ WRITE_ONCE(rb->nest, 1);
goto again;
}

@@ -355,6 +362,7 @@ void *perf_aux_output_begin(struct perf_
struct perf_event *output_event = event;
unsigned long aux_head, aux_tail;
struct ring_buffer *rb;
+ unsigned int nest;

if (output_event->parent)
output_event = output_event->parent;
@@ -385,13 +393,16 @@ void *perf_aux_output_begin(struct perf_
if (!refcount_inc_not_zero(&rb->aux_refcount))
goto err;

+ nest = READ_ONCE(rb->aux_nest);
/*
* Nesting is not supported for AUX area, make sure nested
* writers are caught early
*/
- if (WARN_ON_ONCE(local_xchg(&rb->aux_nest, 1)))
+ if (WARN_ON_ONCE(nest))
goto err_put;

+ WRITE_ONCE(rb->aux_nest, nest + 1);
+
aux_head = rb->aux_head;

handle->rb = rb;
@@ -419,7 +430,7 @@ void *perf_aux_output_begin(struct perf_
if (!handle->size) { /* A, matches D */
event->pending_disable = smp_processor_id();
perf_output_wakeup(handle);
- local_set(&rb->aux_nest, 0);
+ WRITE_ONCE(rb->aux_nest, 0);
goto err_put;
}
}
@@ -508,7 +519,7 @@ void perf_aux_output_end(struct perf_out

handle->event = NULL;

- local_set(&rb->aux_nest, 0);
+ WRITE_ONCE(rb->aux_nest, 0);
/* can't be last */
rb_free_aux(rb);
ring_buffer_put(rb);