ring-buffer: Page size per ring buffer

Currently the size of one sub buffer page is global for all buffers and
it is hard coded to one system page. In order to introduce configurable
ring buffer sub page size, the internal logic should be refactored to
work with sub page size per ring buffer.

Link: https://lore.kernel.org/linux-trace-devel/20211213094825.61876-3-tz.stoyanov@gmail.com
Link: https://lore.kernel.org/linux-trace-kernel/20231219185628.009147038@goodmis.org

Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Vincent Donnefort <vdonnefort@google.com>
Cc: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Tzvetomir Stoyanov (VMware) <tz.stoyanov@gmail.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>

1 files changed, 38 insertions, 30 deletions

diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c
index 2400c8e68fd3..d9f656502400 100644
--- a/kernel/trace/ring_buffer.c
+++ b/kernel/trace/ring_buffer.c
@@ -374,11 +374,6 @@ static inline bool test_time_stamp(u64 delta)
 	return !!(delta & TS_DELTA_TEST);
 }
 
-#define BUF_PAGE_SIZE (PAGE_SIZE - BUF_PAGE_HDR_SIZE)
-
-/* Max payload is BUF_PAGE_SIZE - header (8bytes) */
-#define BUF_MAX_DATA_SIZE (BUF_PAGE_SIZE - (sizeof(u32) * 2))
-
 struct rb_irq_work {
 	struct irq_work			work;
 	wait_queue_head_t		waiters;
@@ -510,6 +505,9 @@ struct trace_buffer {
 
 	struct rb_irq_work		irq_work;
 	bool				time_stamp_abs;
+
+	unsigned int			subbuf_size;
+	unsigned int			max_data_size;
 };
 
 struct ring_buffer_iter {
@@ -523,10 +521,11 @@ struct ring_buffer_iter {
 	u64				read_stamp;
 	u64				page_stamp;
 	struct ring_buffer_event	*event;
+	size_t				event_size;
 	int				missed_events;
 };
 
-int ring_buffer_print_page_header(struct trace_seq *s)
+int ring_buffer_print_page_header(struct trace_buffer *buffer, struct trace_seq *s)
 {
 	struct buffer_data_page field;
 
@@ -550,7 +549,7 @@ int ring_buffer_print_page_header(struct trace_seq *s)
 	trace_seq_printf(s, "\tfield: char data;\t"
 			 "offset:%u;\tsize:%u;\tsigned:%u;\n",
 			 (unsigned int)offsetof(typeof(field), data),
-			 (unsigned int)BUF_PAGE_SIZE,
+			 (unsigned int)buffer->subbuf_size,
 			 (unsigned int)is_signed_type(char));
 
 	return !trace_seq_has_overflowed(s);
@@ -1625,7 +1624,13 @@ struct trace_buffer *__ring_buffer_alloc(unsigned long size, unsigned flags,
 	if (!zalloc_cpumask_var(&buffer->cpumask, GFP_KERNEL))
 		goto fail_free_buffer;
 
-	nr_pages = DIV_ROUND_UP(size, BUF_PAGE_SIZE);
+	/* Default buffer page size - one system page */
+	buffer->subbuf_size = PAGE_SIZE - BUF_PAGE_HDR_SIZE;
+
+	/* Max payload is buffer page size - header (8bytes) */
+	buffer->max_data_size = buffer->subbuf_size - (sizeof(u32) * 2);
+
+	nr_pages = DIV_ROUND_UP(size, buffer->subbuf_size);
 	buffer->flags = flags;
 	buffer->clock = trace_clock_local;
 	buffer->reader_lock_key = key;
@@ -1944,7 +1949,7 @@ static void update_pages_handler(struct work_struct *work)
  * @size: the new size.
  * @cpu_id: the cpu buffer to resize
  *
- * Minimum size is 2 * BUF_PAGE_SIZE.
+ * Minimum size is 2 * buffer->subbuf_size.
  *
  * Returns 0 on success and < 0 on failure.
  */
@@ -1966,7 +1971,7 @@ int ring_buffer_resize(struct trace_buffer *buffer, unsigned long size,
 	    !cpumask_test_cpu(cpu_id, buffer->cpumask))
 		return 0;
 
-	nr_pages = DIV_ROUND_UP(size, BUF_PAGE_SIZE);
+	nr_pages = DIV_ROUND_UP(size, buffer->subbuf_size);
 
 	/* we need a minimum of two pages */
 	if (nr_pages < 2)
@@ -2213,7 +2218,7 @@ rb_iter_head_event(struct ring_buffer_iter *iter)
 	 */
 	barrier();
 
-	if ((iter->head + length) > commit || length > BUF_PAGE_SIZE)
+	if ((iter->head + length) > commit || length > iter->event_size)
 		/* Writer corrupted the read? */
 		goto reset;
 
@@ -2446,6 +2451,7 @@ static inline void
 rb_reset_tail(struct ring_buffer_per_cpu *cpu_buffer,
 	      unsigned long tail, struct rb_event_info *info)
 {
+	unsigned long bsize = READ_ONCE(cpu_buffer->buffer->subbuf_size);
 	struct buffer_page *tail_page = info->tail_page;
 	struct ring_buffer_event *event;
 	unsigned long length = info->length;
@@ -2454,13 +2460,13 @@ rb_reset_tail(struct ring_buffer_per_cpu *cpu_buffer,
 	 * Only the event that crossed the page boundary
 	 * must fill the old tail_page with padding.
 	 */
-	if (tail >= BUF_PAGE_SIZE) {
+	if (tail >= bsize) {
 		/*
 		 * If the page was filled, then we still need
 		 * to update the real_end. Reset it to zero
 		 * and the reader will ignore it.
 		 */
-		if (tail == BUF_PAGE_SIZE)
+		if (tail == bsize)
 			tail_page->real_end = 0;
 
 		local_sub(length, &tail_page->write);
@@ -2488,7 +2494,7 @@ rb_reset_tail(struct ring_buffer_per_cpu *cpu_buffer,
 	 * If we are less than the minimum size, we don't need to
 	 * worry about it.
 	 */
-	if (tail > (BUF_PAGE_SIZE - RB_EVNT_MIN_SIZE)) {
+	if (tail > (bsize - RB_EVNT_MIN_SIZE)) {
 		/* No room for any events */
 
 		/* Mark the rest of the page with padding */
@@ -2503,19 +2509,19 @@ rb_reset_tail(struct ring_buffer_per_cpu *cpu_buffer,
 	}
 
 	/* Put in a discarded event */
-	event->array[0] = (BUF_PAGE_SIZE - tail) - RB_EVNT_HDR_SIZE;
+	event->array[0] = (bsize - tail) - RB_EVNT_HDR_SIZE;
 	event->type_len = RINGBUF_TYPE_PADDING;
 	/* time delta must be non zero */
 	event->time_delta = 1;
 
 	/* account for padding bytes */
-	local_add(BUF_PAGE_SIZE - tail, &cpu_buffer->entries_bytes);
+	local_add(bsize - tail, &cpu_buffer->entries_bytes);
 
 	/* Make sure the padding is visible before the tail_page->write update */
 	smp_wmb();
 
 	/* Set write to end of buffer */
-	length = (tail + length) - BUF_PAGE_SIZE;
+	length = (tail + length) - bsize;
 	local_sub(length, &tail_page->write);
 }
 
@@ -3469,7 +3475,7 @@ __rb_reserve_next(struct ring_buffer_per_cpu *cpu_buffer,
 	tail = write - info->length;
 
 	/* See if we shot pass the end of this buffer page */
-	if (unlikely(write > BUF_PAGE_SIZE)) {
+	if (unlikely(write > cpu_buffer->buffer->subbuf_size)) {
 		check_buffer(cpu_buffer, info, CHECK_FULL_PAGE);
 		return rb_move_tail(cpu_buffer, tail, info);
 	}
@@ -3600,7 +3606,7 @@ rb_reserve_next_event(struct trace_buffer *buffer,
 	if (ring_buffer_time_stamp_abs(cpu_buffer->buffer)) {
 		add_ts_default = RB_ADD_STAMP_ABSOLUTE;
 		info.length += RB_LEN_TIME_EXTEND;
-		if (info.length > BUF_MAX_DATA_SIZE)
+		if (info.length > cpu_buffer->buffer->max_data_size)
 			goto out_fail;
 	} else {
 		add_ts_default = RB_ADD_STAMP_NONE;
@@ -3675,7 +3681,7 @@ ring_buffer_lock_reserve(struct trace_buffer *buffer, unsigned long length)
 	if (unlikely(atomic_read(&cpu_buffer->record_disabled)))
 		goto out;
 
-	if (unlikely(length > BUF_MAX_DATA_SIZE))
+	if (unlikely(length > buffer->max_data_size))
 		goto out;
 
 	if (unlikely(trace_recursive_lock(cpu_buffer)))
@@ -3825,7 +3831,7 @@ int ring_buffer_write(struct trace_buffer *buffer,
 	if (atomic_read(&cpu_buffer->record_disabled))
 		goto out;
 
-	if (length > BUF_MAX_DATA_SIZE)
+	if (length > buffer->max_data_size)
 		goto out;
 
 	if (unlikely(trace_recursive_lock(cpu_buffer)))
@@ -4405,6 +4411,7 @@ static struct buffer_page *
 rb_get_reader_page(struct ring_buffer_per_cpu *cpu_buffer)
 {
 	struct buffer_page *reader = NULL;
+	unsigned long bsize = READ_ONCE(cpu_buffer->buffer->subbuf_size);
 	unsigned long overwrite;
 	unsigned long flags;
 	int nr_loops = 0;
@@ -4540,7 +4547,7 @@ rb_get_reader_page(struct ring_buffer_per_cpu *cpu_buffer)
 #define USECS_WAIT	1000000
         for (nr_loops = 0; nr_loops < USECS_WAIT; nr_loops++) {
 		/* If the write is past the end of page, a writer is still updating it */
-		if (likely(!reader || rb_page_write(reader) <= BUF_PAGE_SIZE))
+		if (likely(!reader || rb_page_write(reader) <= bsize))
 			break;
 
 		udelay(1);
@@ -4984,7 +4991,8 @@ ring_buffer_read_prepare(struct trace_buffer *buffer, int cpu, gfp_t flags)
 		return NULL;
 
 	/* Holds the entire event: data and meta data */
-	iter->event = kmalloc(BUF_PAGE_SIZE, flags);
+	iter->event_size = buffer->subbuf_size;
+	iter->event = kmalloc(iter->event_size, flags);
 	if (!iter->event) {
 		kfree(iter);
 		return NULL;
@@ -5102,14 +5110,14 @@ unsigned long ring_buffer_size(struct trace_buffer *buffer, int cpu)
 {
 	/*
 	 * Earlier, this method returned
-	 *	BUF_PAGE_SIZE * buffer->nr_pages
+	 *	buffer->subbuf_size * buffer->nr_pages
 	 * Since the nr_pages field is now removed, we have converted this to
 	 * return the per cpu buffer value.
 	 */
 	if (!cpumask_test_cpu(cpu, buffer->cpumask))
 		return 0;
 
-	return BUF_PAGE_SIZE * buffer->buffers[cpu]->nr_pages;
+	return buffer->subbuf_size * buffer->buffers[cpu]->nr_pages;
 }
 EXPORT_SYMBOL_GPL(ring_buffer_size);
 
@@ -5123,8 +5131,8 @@ unsigned long ring_buffer_max_event_size(struct trace_buffer *buffer)
 {
 	/* If abs timestamp is requested, events have a timestamp too */
 	if (ring_buffer_time_stamp_abs(buffer))
-		return BUF_MAX_DATA_SIZE - RB_LEN_TIME_EXTEND;
-	return BUF_MAX_DATA_SIZE;
+		return buffer->max_data_size - RB_LEN_TIME_EXTEND;
+	return buffer->max_data_size;
 }
 EXPORT_SYMBOL_GPL(ring_buffer_max_event_size);
 
@@ -5730,7 +5738,7 @@ int ring_buffer_read_page(struct trace_buffer *buffer,
 		/* 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)) {
+		if (buffer->subbuf_size - commit >= sizeof(missed_events)) {
 			memcpy(&bpage->data[commit], &missed_events,
 			       sizeof(missed_events));
 			local_add(RB_MISSED_STORED, &bpage->commit);
@@ -5742,8 +5750,8 @@ int ring_buffer_read_page(struct trace_buffer *buffer,
 	/*
 	 * This page may be off to user land. Zero it out here.
 	 */
-	if (commit < BUF_PAGE_SIZE)
-		memset(&bpage->data[commit], 0, BUF_PAGE_SIZE - commit);
+	if (commit < buffer->subbuf_size)
+		memset(&bpage->data[commit], 0, buffer->subbuf_size - commit);
 
  out_unlock:
 	raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags);