ring-buffer: Init waitqueue for blocked readers
authorSteven Rostedt (Red Hat) <srostedt@redhat.com>
Mon, 4 Mar 2013 22:33:05 +0000 (17:33 -0500)
committerSteven Rostedt <rostedt@goodmis.org>
Fri, 15 Mar 2013 04:34:55 +0000 (00:34 -0400)
The move of blocked readers to the ring buffer left out the
init of the wait queue that is used. Tests missed this due to running
stress tests against the buffers, which didn't allow for any
readers to end up waiting. Running a simple read and wait triggered
a bug.

Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
kernel/trace/ring_buffer.c

index 56b6ea32d2e78624d45229dd43ea3c0a31b658ec..65fe2a4f98241bc4295276aee1fbec95463d7ce9 100644 (file)
@@ -1185,6 +1185,7 @@ rb_allocate_cpu_buffer(struct ring_buffer *buffer, int nr_pages, int cpu)
        INIT_WORK(&cpu_buffer->update_pages_work, update_pages_handler);
        init_completion(&cpu_buffer->update_done);
        init_irq_work(&cpu_buffer->irq_work.work, rb_wake_up_waiters);
+       init_waitqueue_head(&cpu_buffer->irq_work.waiters);
 
        bpage = kzalloc_node(ALIGN(sizeof(*bpage), cache_line_size()),
                            GFP_KERNEL, cpu_to_node(cpu));
@@ -1281,6 +1282,7 @@ struct ring_buffer *__ring_buffer_alloc(unsigned long size, unsigned flags,
        buffer->reader_lock_key = key;
 
        init_irq_work(&buffer->irq_work.work, rb_wake_up_waiters);
+       init_waitqueue_head(&buffer->irq_work.waiters);
 
        /* need at least two pages */
        if (nr_pages < 2)