#include <linux/module.h>
#include <linux/percpu.h>
#include <linux/mutex.h>
+#include <linux/slab.h>
#include <linux/init.h>
#include <linux/hash.h>
#include <linux/list.h>
#include <linux/cpu.h>
#include <linux/fs.h>
+#include <asm/local.h>
#include "trace.h"
/*
#define RB_MAX_SMALL_DATA (RB_ALIGNMENT * RINGBUF_TYPE_DATA_TYPE_LEN_MAX)
#define RB_EVNT_MIN_SIZE 8U /* two 32bit words */
+#if !defined(CONFIG_64BIT) || defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS)
+# define RB_FORCE_8BYTE_ALIGNMENT 0
+# define RB_ARCH_ALIGNMENT RB_ALIGNMENT
+#else
+# define RB_FORCE_8BYTE_ALIGNMENT 1
+# define RB_ARCH_ALIGNMENT 8U
+#endif
+
/* define RINGBUF_TYPE_DATA for 'case RINGBUF_TYPE_DATA:' */
#define RINGBUF_TYPE_DATA 0 ... RINGBUF_TYPE_DATA_TYPE_LEN_MAX
#define TS_MASK ((1ULL << TS_SHIFT) - 1)
#define TS_DELTA_TEST (~TS_MASK)
+/* Flag when events were overwritten */
+#define RB_MISSED_EVENTS (1 << 31)
+/* Missed count stored at end */
+#define RB_MISSED_STORED (1 << 30)
+
struct buffer_data_page {
u64 time_stamp; /* page time stamp */
local_t commit; /* write committed index */
local_t write; /* index for next write */
unsigned read; /* index for next read */
local_t entries; /* entries on this page */
+ unsigned long real_end; /* real end of data */
struct buffer_data_page *page; /* Actual data page */
};
(unsigned int)sizeof(field.commit),
(unsigned int)is_signed_type(long));
+ ret = trace_seq_printf(s, "\tfield: int overwrite;\t"
+ "offset:%u;\tsize:%u;\tsigned:%u;\n",
+ (unsigned int)offsetof(typeof(field), commit),
+ 1,
+ (unsigned int)is_signed_type(long));
+
ret = trace_seq_printf(s, "\tfield: char data;\t"
"offset:%u;\tsize:%u;\tsigned:%u;\n",
(unsigned int)offsetof(typeof(field), data),
struct buffer_page *tail_page; /* write to tail */
struct buffer_page *commit_page; /* committed pages */
struct buffer_page *reader_page;
+ unsigned long lost_events;
+ unsigned long last_overrun;
local_t commit_overrun;
local_t overrun;
local_t entries;
struct ring_buffer_per_cpu *cpu_buffer;
unsigned long head;
struct buffer_page *head_page;
+ struct buffer_page *cache_reader_page;
+ unsigned long cache_read;
u64 read_stamp;
};
for (i = 0; i < nr_pages; i++) {
if (RB_WARN_ON(cpu_buffer, list_empty(cpu_buffer->pages)))
- return;
+ goto out;
p = cpu_buffer->pages->next;
bpage = list_entry(p, struct buffer_page, list);
list_del_init(&bpage->list);
free_buffer_page(bpage);
}
if (RB_WARN_ON(cpu_buffer, list_empty(cpu_buffer->pages)))
- return;
+ goto out;
rb_reset_cpu(cpu_buffer);
rb_check_pages(cpu_buffer);
+out:
spin_unlock_irq(&cpu_buffer->reader_lock);
}
for (i = 0; i < nr_pages; i++) {
if (RB_WARN_ON(cpu_buffer, list_empty(pages)))
- return;
+ goto out;
p = pages->next;
bpage = list_entry(p, struct buffer_page, list);
list_del_init(&bpage->list);
rb_reset_cpu(cpu_buffer);
rb_check_pages(cpu_buffer);
+out:
spin_unlock_irq(&cpu_buffer->reader_lock);
}
case 0:
length -= RB_EVNT_HDR_SIZE;
- if (length > RB_MAX_SMALL_DATA)
+ if (length > RB_MAX_SMALL_DATA || RB_FORCE_8BYTE_ALIGNMENT)
event->array[0] = length;
else
event->type_len = DIV_ROUND_UP(length, RB_ALIGNMENT);
if (!length)
length = 1;
- if (length > RB_MAX_SMALL_DATA)
+ if (length > RB_MAX_SMALL_DATA || RB_FORCE_8BYTE_ALIGNMENT)
length += sizeof(event.array[0]);
length += RB_EVNT_HDR_SIZE;
- length = ALIGN(length, RB_ALIGNMENT);
+ length = ALIGN(length, RB_ARCH_ALIGNMENT);
return length;
}
event = __rb_page_index(tail_page, tail);
kmemcheck_annotate_bitfield(event, bitfield);
+ /*
+ * Save the original length to the meta data.
+ * This will be used by the reader to add lost event
+ * counter.
+ */
+ tail_page->real_end = tail;
+
/*
* If this event is bigger than the minimum size, then
* we need to be careful that we don't subtract the
if (ring_buffer_flags != RB_BUFFERS_ON)
return NULL;
- if (atomic_read(&buffer->record_disabled))
- return NULL;
-
/* If we are tracing schedule, we don't want to recurse */
resched = ftrace_preempt_disable();
+ if (atomic_read(&buffer->record_disabled))
+ goto out_nocheck;
+
if (trace_recursive_lock())
goto out_nocheck;
if (ring_buffer_flags != RB_BUFFERS_ON)
return -EBUSY;
- if (atomic_read(&buffer->record_disabled))
- return -EBUSY;
-
resched = ftrace_preempt_disable();
+ if (atomic_read(&buffer->record_disabled))
+ goto out;
+
cpu = raw_smp_processor_id();
if (!cpumask_test_cpu(cpu, buffer->cpumask))
* @buffer: The ring buffer to enable writes
*
* Note, multiple disables will need the same number of enables
- * to truely enable the writing (much like preempt_disable).
+ * to truly enable the writing (much like preempt_disable).
*/
void ring_buffer_record_enable(struct ring_buffer *buffer)
{
* @cpu: The CPU to enable.
*
* Note, multiple disables will need the same number of enables
- * to truely enable the writing (much like preempt_disable).
+ * to truly enable the writing (much like preempt_disable).
*/
void ring_buffer_record_enable_cpu(struct ring_buffer *buffer, int cpu)
{
iter->read_stamp = cpu_buffer->read_stamp;
else
iter->read_stamp = iter->head_page->page->time_stamp;
+ iter->cache_reader_page = cpu_buffer->reader_page;
+ iter->cache_read = cpu_buffer->read;
}
/**
rb_get_reader_page(struct ring_buffer_per_cpu *cpu_buffer)
{
struct buffer_page *reader = NULL;
+ unsigned long overwrite;
unsigned long flags;
int nr_loops = 0;
int ret;
local_set(&cpu_buffer->reader_page->write, 0);
local_set(&cpu_buffer->reader_page->entries, 0);
local_set(&cpu_buffer->reader_page->page->commit, 0);
+ cpu_buffer->reader_page->real_end = 0;
spin:
/*
* Splice the empty reader page into the list around the head.
*/
reader = rb_set_head_page(cpu_buffer);
- cpu_buffer->reader_page->list.next = reader->list.next;
+ cpu_buffer->reader_page->list.next = rb_list_head(reader->list.next);
cpu_buffer->reader_page->list.prev = reader->list.prev;
/*
/* The reader page will be pointing to the new head */
rb_set_list_to_head(cpu_buffer, &cpu_buffer->reader_page->list);
+ /*
+ * We want to make sure we read the overruns after we set up our
+ * pointers to the next object. The writer side does a
+ * cmpxchg to cross pages which acts as the mb on the writer
+ * side. Note, the reader will constantly fail the swap
+ * while the writer is updating the pointers, so this
+ * guarantees that the overwrite recorded here is the one we
+ * want to compare with the last_overrun.
+ */
+ smp_mb();
+ overwrite = local_read(&(cpu_buffer->overrun));
+
/*
* Here's the tricky part.
*
*
* Now make the new head point back to the reader page.
*/
- reader->list.next->prev = &cpu_buffer->reader_page->list;
+ rb_list_head(reader->list.next)->prev = &cpu_buffer->reader_page->list;
rb_inc_page(cpu_buffer, &cpu_buffer->head_page);
/* Finally update the reader page to the new head */
cpu_buffer->reader_page = reader;
rb_reset_reader_page(cpu_buffer);
+ if (overwrite != cpu_buffer->last_overrun) {
+ cpu_buffer->lost_events = overwrite - cpu_buffer->last_overrun;
+ cpu_buffer->last_overrun = overwrite;
+ }
+
goto again;
out:
rb_advance_iter(iter);
}
+static int rb_lost_events(struct ring_buffer_per_cpu *cpu_buffer)
+{
+ return cpu_buffer->lost_events;
+}
+
static struct ring_buffer_event *
-rb_buffer_peek(struct ring_buffer_per_cpu *cpu_buffer, u64 *ts)
+rb_buffer_peek(struct ring_buffer_per_cpu *cpu_buffer, u64 *ts,
+ unsigned long *lost_events)
{
struct ring_buffer_event *event;
struct buffer_page *reader;
ring_buffer_normalize_time_stamp(cpu_buffer->buffer,
cpu_buffer->cpu, ts);
}
+ if (lost_events)
+ *lost_events = rb_lost_events(cpu_buffer);
return event;
default:
struct ring_buffer_event *event;
int nr_loops = 0;
- if (ring_buffer_iter_empty(iter))
- return NULL;
-
cpu_buffer = iter->cpu_buffer;
buffer = cpu_buffer->buffer;
+ /*
+ * Check if someone performed a consuming read to
+ * the buffer. A consuming read invalidates the iterator
+ * and we need to reset the iterator in this case.
+ */
+ if (unlikely(iter->cache_read != cpu_buffer->read ||
+ iter->cache_reader_page != cpu_buffer->reader_page))
+ rb_iter_reset(iter);
+
again:
+ if (ring_buffer_iter_empty(iter))
+ return NULL;
+
/*
* We repeat when a timestamp is encountered.
* We can get multiple timestamps by nested interrupts or also
if (rb_per_cpu_empty(cpu_buffer))
return NULL;
+ if (iter->head >= local_read(&iter->head_page->page->commit)) {
+ rb_inc_iter(iter);
+ goto again;
+ }
+
event = rb_iter_head_event(iter);
switch (event->type_len) {
* @buffer: The ring buffer to read
* @cpu: The cpu to peak at
* @ts: The timestamp counter of this event.
+ * @lost_events: a variable to store if events were lost (may be NULL)
*
* This will return the event that will be read next, but does
* not consume the data.
*/
struct ring_buffer_event *
-ring_buffer_peek(struct ring_buffer *buffer, int cpu, u64 *ts)
+ring_buffer_peek(struct ring_buffer *buffer, int cpu, u64 *ts,
+ unsigned long *lost_events)
{
struct ring_buffer_per_cpu *cpu_buffer = buffer->buffers[cpu];
struct ring_buffer_event *event;
local_irq_save(flags);
if (dolock)
spin_lock(&cpu_buffer->reader_lock);
- event = rb_buffer_peek(cpu_buffer, ts);
+ event = rb_buffer_peek(cpu_buffer, ts, lost_events);
if (event && event->type_len == RINGBUF_TYPE_PADDING)
rb_advance_reader(cpu_buffer);
if (dolock)
/**
* ring_buffer_consume - return an event and consume it
* @buffer: The ring buffer to get the next event from
+ * @cpu: the cpu to read the buffer from
+ * @ts: a variable to store the timestamp (may be NULL)
+ * @lost_events: a variable to store if events were lost (may be NULL)
*
* Returns the next event in the ring buffer, and that event is consumed.
* Meaning, that sequential reads will keep returning a different event,
* and eventually empty the ring buffer if the producer is slower.
*/
struct ring_buffer_event *
-ring_buffer_consume(struct ring_buffer *buffer, int cpu, u64 *ts)
+ring_buffer_consume(struct ring_buffer *buffer, int cpu, u64 *ts,
+ unsigned long *lost_events)
{
struct ring_buffer_per_cpu *cpu_buffer;
struct ring_buffer_event *event = NULL;
if (dolock)
spin_lock(&cpu_buffer->reader_lock);
- event = rb_buffer_peek(cpu_buffer, ts);
- if (event)
+ event = rb_buffer_peek(cpu_buffer, ts, lost_events);
+ if (event) {
+ cpu_buffer->lost_events = 0;
rb_advance_reader(cpu_buffer);
+ }
if (dolock)
spin_unlock(&cpu_buffer->reader_lock);
EXPORT_SYMBOL_GPL(ring_buffer_consume);
/**
- * ring_buffer_read_start - start a non consuming read of the buffer
+ * ring_buffer_read_prepare - Prepare for a non consuming read of the buffer
* @buffer: The ring buffer to read from
* @cpu: The cpu buffer to iterate over
*
- * This starts up an iteration through the buffer. It also disables
- * the recording to the buffer until the reading is finished.
- * This prevents the reading from being corrupted. This is not
- * a consuming read, so a producer is not expected.
+ * This performs the initial preparations necessary to iterate
+ * through the buffer. Memory is allocated, buffer recording
+ * is disabled, and the iterator pointer is returned to the caller.
*
- * Must be paired with ring_buffer_finish.
+ * Disabling buffer recordng prevents the reading from being
+ * corrupted. This is not a consuming read, so a producer is not
+ * expected.
+ *
+ * After a sequence of ring_buffer_read_prepare calls, the user is
+ * expected to make at least one call to ring_buffer_prepare_sync.
+ * Afterwards, ring_buffer_read_start is invoked to get things going
+ * for real.
+ *
+ * This overall must be paired with ring_buffer_finish.
*/
struct ring_buffer_iter *
-ring_buffer_read_start(struct ring_buffer *buffer, int cpu)
+ring_buffer_read_prepare(struct ring_buffer *buffer, int cpu)
{
struct ring_buffer_per_cpu *cpu_buffer;
struct ring_buffer_iter *iter;
- unsigned long flags;
if (!cpumask_test_cpu(cpu, buffer->cpumask))
return NULL;
iter->cpu_buffer = cpu_buffer;
atomic_inc(&cpu_buffer->record_disabled);
+
+ return iter;
+}
+EXPORT_SYMBOL_GPL(ring_buffer_read_prepare);
+
+/**
+ * ring_buffer_read_prepare_sync - Synchronize a set of prepare calls
+ *
+ * All previously invoked ring_buffer_read_prepare calls to prepare
+ * iterators will be synchronized. Afterwards, read_buffer_read_start
+ * calls on those iterators are allowed.
+ */
+void
+ring_buffer_read_prepare_sync(void)
+{
synchronize_sched();
+}
+EXPORT_SYMBOL_GPL(ring_buffer_read_prepare_sync);
+
+/**
+ * ring_buffer_read_start - start a non consuming read of the buffer
+ * @iter: The iterator returned by ring_buffer_read_prepare
+ *
+ * This finalizes the startup of an iteration through the buffer.
+ * The iterator comes from a call to ring_buffer_read_prepare and
+ * an intervening ring_buffer_read_prepare_sync must have been
+ * performed.
+ *
+ * Must be paired with ring_buffer_finish.
+ */
+void
+ring_buffer_read_start(struct ring_buffer_iter *iter)
+{
+ struct ring_buffer_per_cpu *cpu_buffer;
+ unsigned long flags;
+
+ if (!iter)
+ return;
+
+ cpu_buffer = iter->cpu_buffer;
spin_lock_irqsave(&cpu_buffer->reader_lock, flags);
arch_spin_lock(&cpu_buffer->lock);
rb_iter_reset(iter);
arch_spin_unlock(&cpu_buffer->lock);
spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags);
-
- return iter;
}
EXPORT_SYMBOL_GPL(ring_buffer_read_start);
cpu_buffer->write_stamp = 0;
cpu_buffer->read_stamp = 0;
+ cpu_buffer->lost_events = 0;
+ cpu_buffer->last_overrun = 0;
+
rb_head_page_activate(cpu_buffer);
}
struct ring_buffer_event *event;
struct buffer_data_page *bpage;
struct buffer_page *reader;
+ unsigned long missed_events;
unsigned long flags;
unsigned int commit;
unsigned int read;
read = reader->read;
commit = rb_page_commit(reader);
+ /* Check if any events were dropped */
+ missed_events = cpu_buffer->lost_events;
+
/*
* If this page has been partially read or
* if len is not big enough to read the rest of the page or
local_set(&reader->entries, 0);
reader->read = 0;
*data_page = bpage;
+
+ /*
+ * Use the real_end for the data size,
+ * This gives us a chance to store the lost events
+ * on the page.
+ */
+ if (reader->real_end)
+ local_set(&bpage->commit, reader->real_end);
}
ret = read;
+ cpu_buffer->lost_events = 0;
+ /*
+ * Set a flag in the commit field if we lost events
+ */
+ if (missed_events) {
+ commit = local_read(&bpage->commit);
+
+ /* If there is room at the end of the page to save the
+ * missed events, then record it there.
+ */
+ if (BUF_PAGE_SIZE - commit >= sizeof(missed_events)) {
+ memcpy(&bpage->data[commit], &missed_events,
+ sizeof(missed_events));
+ local_add(RB_MISSED_STORED, &bpage->commit);
+ }
+ local_add(RB_MISSED_EVENTS, &bpage->commit);
+ }
+
out_unlock:
spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags);
Code Review / linux-2.6.git / blobdiff