2
0
mirror of https://github.com/edk2-porting/linux-next.git synced 2024-12-25 05:34:00 +08:00
linux-next/kernel/events/ring_buffer.c
Peter Zijlstra c7f2e3cd6c perf: Optimize ring-buffer write by depending on control dependencies
Remove a full barrier from the ring-buffer write path by relying on
a control dependency to order a LOAD -> STORE scenario.

Cc: "Paul E. McKenney" <paulmck@us.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/n/tip-8alv40z6ikk57jzbaobnxrjl@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
2013-12-11 15:53:22 +01:00

418 lines
9.0 KiB
C

/*
* Performance events ring-buffer code:
*
* Copyright (C) 2008 Thomas Gleixner <tglx@linutronix.de>
* Copyright (C) 2008-2011 Red Hat, Inc., Ingo Molnar
* Copyright (C) 2008-2011 Red Hat, Inc., Peter Zijlstra <pzijlstr@redhat.com>
* Copyright © 2009 Paul Mackerras, IBM Corp. <paulus@au1.ibm.com>
*
* For licensing details see kernel-base/COPYING
*/
#include <linux/perf_event.h>
#include <linux/vmalloc.h>
#include <linux/slab.h>
#include <linux/circ_buf.h>
#include "internal.h"
static void perf_output_wakeup(struct perf_output_handle *handle)
{
atomic_set(&handle->rb->poll, POLL_IN);
handle->event->pending_wakeup = 1;
irq_work_queue(&handle->event->pending);
}
/*
* We need to ensure a later event_id doesn't publish a head when a former
* event isn't done writing. However since we need to deal with NMIs we
* cannot fully serialize things.
*
* We only publish the head (and generate a wakeup) when the outer-most
* event completes.
*/
static void perf_output_get_handle(struct perf_output_handle *handle)
{
struct ring_buffer *rb = handle->rb;
preempt_disable();
local_inc(&rb->nest);
handle->wakeup = local_read(&rb->wakeup);
}
static void perf_output_put_handle(struct perf_output_handle *handle)
{
struct ring_buffer *rb = handle->rb;
unsigned long head;
again:
head = local_read(&rb->head);
/*
* IRQ/NMI can happen here, which means we can miss a head update.
*/
if (!local_dec_and_test(&rb->nest))
goto out;
/*
* Since the mmap() consumer (userspace) can run on a different CPU:
*
* kernel user
*
* if (LOAD ->data_tail) { LOAD ->data_head
* (A) smp_rmb() (C)
* STORE $data LOAD $data
* smp_wmb() (B) smp_mb() (D)
* STORE ->data_head STORE ->data_tail
* }
*
* Where A pairs with D, and B pairs with C.
*
* In our case (A) is a control dependency that separates the load of
* the ->data_tail and the stores of $data. In case ->data_tail
* indicates there is no room in the buffer to store $data we do not.
*
* D needs to be a full barrier since it separates the data READ
* from the tail WRITE.
*
* For B a WMB is sufficient since it separates two WRITEs, and for C
* an RMB is sufficient since it separates two READs.
*
* See perf_output_begin().
*/
smp_wmb(); /* B, matches C */
rb->user_page->data_head = head;
/*
* Now check if we missed an update -- rely on previous implied
* compiler barriers to force a re-read.
*/
if (unlikely(head != local_read(&rb->head))) {
local_inc(&rb->nest);
goto again;
}
if (handle->wakeup != local_read(&rb->wakeup))
perf_output_wakeup(handle);
out:
preempt_enable();
}
int perf_output_begin(struct perf_output_handle *handle,
struct perf_event *event, unsigned int size)
{
struct ring_buffer *rb;
unsigned long tail, offset, head;
int have_lost, page_shift;
struct {
struct perf_event_header header;
u64 id;
u64 lost;
} lost_event;
rcu_read_lock();
/*
* For inherited events we send all the output towards the parent.
*/
if (event->parent)
event = event->parent;
rb = rcu_dereference(event->rb);
if (unlikely(!rb))
goto out;
if (unlikely(!rb->nr_pages))
goto out;
handle->rb = rb;
handle->event = event;
have_lost = local_read(&rb->lost);
if (unlikely(have_lost)) {
size += sizeof(lost_event);
if (event->attr.sample_id_all)
size += event->id_header_size;
}
perf_output_get_handle(handle);
do {
tail = ACCESS_ONCE(rb->user_page->data_tail);
offset = head = local_read(&rb->head);
if (!rb->overwrite &&
unlikely(CIRC_SPACE(head, tail, perf_data_size(rb)) < size))
goto fail;
/*
* The above forms a control dependency barrier separating the
* @tail load above from the data stores below. Since the @tail
* load is required to compute the branch to fail below.
*
* A, matches D; the full memory barrier userspace SHOULD issue
* after reading the data and before storing the new tail
* position.
*
* See perf_output_put_handle().
*/
head += size;
} while (local_cmpxchg(&rb->head, offset, head) != offset);
/*
* We rely on the implied barrier() by local_cmpxchg() to ensure
* none of the data stores below can be lifted up by the compiler.
*/
if (unlikely(head - local_read(&rb->wakeup) > rb->watermark))
local_add(rb->watermark, &rb->wakeup);
page_shift = PAGE_SHIFT + page_order(rb);
handle->page = (offset >> page_shift) & (rb->nr_pages - 1);
offset &= (1UL << page_shift) - 1;
handle->addr = rb->data_pages[handle->page] + offset;
handle->size = (1UL << page_shift) - offset;
if (unlikely(have_lost)) {
struct perf_sample_data sample_data;
lost_event.header.size = sizeof(lost_event);
lost_event.header.type = PERF_RECORD_LOST;
lost_event.header.misc = 0;
lost_event.id = event->id;
lost_event.lost = local_xchg(&rb->lost, 0);
perf_event_header__init_id(&lost_event.header,
&sample_data, event);
perf_output_put(handle, lost_event);
perf_event__output_id_sample(event, handle, &sample_data);
}
return 0;
fail:
local_inc(&rb->lost);
perf_output_put_handle(handle);
out:
rcu_read_unlock();
return -ENOSPC;
}
unsigned int perf_output_copy(struct perf_output_handle *handle,
const void *buf, unsigned int len)
{
return __output_copy(handle, buf, len);
}
unsigned int perf_output_skip(struct perf_output_handle *handle,
unsigned int len)
{
return __output_skip(handle, NULL, len);
}
void perf_output_end(struct perf_output_handle *handle)
{
perf_output_put_handle(handle);
rcu_read_unlock();
}
static void
ring_buffer_init(struct ring_buffer *rb, long watermark, int flags)
{
long max_size = perf_data_size(rb);
if (watermark)
rb->watermark = min(max_size, watermark);
if (!rb->watermark)
rb->watermark = max_size / 2;
if (flags & RING_BUFFER_WRITABLE)
rb->overwrite = 0;
else
rb->overwrite = 1;
atomic_set(&rb->refcount, 1);
INIT_LIST_HEAD(&rb->event_list);
spin_lock_init(&rb->event_lock);
}
#ifndef CONFIG_PERF_USE_VMALLOC
/*
* Back perf_mmap() with regular GFP_KERNEL-0 pages.
*/
struct page *
perf_mmap_to_page(struct ring_buffer *rb, unsigned long pgoff)
{
if (pgoff > rb->nr_pages)
return NULL;
if (pgoff == 0)
return virt_to_page(rb->user_page);
return virt_to_page(rb->data_pages[pgoff - 1]);
}
static void *perf_mmap_alloc_page(int cpu)
{
struct page *page;
int node;
node = (cpu == -1) ? cpu : cpu_to_node(cpu);
page = alloc_pages_node(node, GFP_KERNEL | __GFP_ZERO, 0);
if (!page)
return NULL;
return page_address(page);
}
struct ring_buffer *rb_alloc(int nr_pages, long watermark, int cpu, int flags)
{
struct ring_buffer *rb;
unsigned long size;
int i;
size = sizeof(struct ring_buffer);
size += nr_pages * sizeof(void *);
rb = kzalloc(size, GFP_KERNEL);
if (!rb)
goto fail;
rb->user_page = perf_mmap_alloc_page(cpu);
if (!rb->user_page)
goto fail_user_page;
for (i = 0; i < nr_pages; i++) {
rb->data_pages[i] = perf_mmap_alloc_page(cpu);
if (!rb->data_pages[i])
goto fail_data_pages;
}
rb->nr_pages = nr_pages;
ring_buffer_init(rb, watermark, flags);
return rb;
fail_data_pages:
for (i--; i >= 0; i--)
free_page((unsigned long)rb->data_pages[i]);
free_page((unsigned long)rb->user_page);
fail_user_page:
kfree(rb);
fail:
return NULL;
}
static void perf_mmap_free_page(unsigned long addr)
{
struct page *page = virt_to_page((void *)addr);
page->mapping = NULL;
__free_page(page);
}
void rb_free(struct ring_buffer *rb)
{
int i;
perf_mmap_free_page((unsigned long)rb->user_page);
for (i = 0; i < rb->nr_pages; i++)
perf_mmap_free_page((unsigned long)rb->data_pages[i]);
kfree(rb);
}
#else
static int data_page_nr(struct ring_buffer *rb)
{
return rb->nr_pages << page_order(rb);
}
struct page *
perf_mmap_to_page(struct ring_buffer *rb, unsigned long pgoff)
{
/* The '>' counts in the user page. */
if (pgoff > data_page_nr(rb))
return NULL;
return vmalloc_to_page((void *)rb->user_page + pgoff * PAGE_SIZE);
}
static void perf_mmap_unmark_page(void *addr)
{
struct page *page = vmalloc_to_page(addr);
page->mapping = NULL;
}
static void rb_free_work(struct work_struct *work)
{
struct ring_buffer *rb;
void *base;
int i, nr;
rb = container_of(work, struct ring_buffer, work);
nr = data_page_nr(rb);
base = rb->user_page;
/* The '<=' counts in the user page. */
for (i = 0; i <= nr; i++)
perf_mmap_unmark_page(base + (i * PAGE_SIZE));
vfree(base);
kfree(rb);
}
void rb_free(struct ring_buffer *rb)
{
schedule_work(&rb->work);
}
struct ring_buffer *rb_alloc(int nr_pages, long watermark, int cpu, int flags)
{
struct ring_buffer *rb;
unsigned long size;
void *all_buf;
size = sizeof(struct ring_buffer);
size += sizeof(void *);
rb = kzalloc(size, GFP_KERNEL);
if (!rb)
goto fail;
INIT_WORK(&rb->work, rb_free_work);
all_buf = vmalloc_user((nr_pages + 1) * PAGE_SIZE);
if (!all_buf)
goto fail_all_buf;
rb->user_page = all_buf;
rb->data_pages[0] = all_buf + PAGE_SIZE;
rb->page_order = ilog2(nr_pages);
rb->nr_pages = !!nr_pages;
ring_buffer_init(rb, watermark, flags);
return rb;
fail_all_buf:
kfree(rb);
fail:
return NULL;
}
#endif