[PATCH] ring_buffer: use list_replace() in rb_get_reader_page()

[PATCH] ring_buffer: use list_replace() in rb_get_reader_page()

[Xiao Guangrong]

Use list_replace() to splice list in rb_get_reader_page()

[ Impact: cleanup ]

Signed-off-by: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com>
---
 kernel/trace/ring_buffer.c |   12 ++----------
 1 files changed, 2 insertions(+), 10 deletions(-)

diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c
index f558b37..f88bfcc 100644
--- a/kernel/trace/ring_buffer.c
+++ b/kernel/trace/ring_buffer.c
@@ -2179,22 +2179,14 @@ rb_get_reader_page(struct ring_buffer_per_cpu *cpu_buffer)
 	if (cpu_buffer->commit_page == cpu_buffer->reader_page)
 		goto out;
 
-	/*
-	 * Splice the empty reader page into the list around the head.
-	 * Reset the reader page to size zero.
-	 */
-
+	/* Reset the reader page to size zero */
 	reader = cpu_buffer->head_page;
-	cpu_buffer->reader_page->list.next = reader->list.next;
-	cpu_buffer->reader_page->list.prev = reader->list.prev;
-
 	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);
 
 	/* Make the reader page now replace the head */
-	reader->list.prev->next = &cpu_buffer->reader_page->list;
-	reader->list.next->prev = &cpu_buffer->reader_page->list;
+	list_replace(&reader->list, &cpu_buffer->reader_page->list);
 
 	/*
 	 * If the tail is on the reader, then we must set the head
-- 
1.6.1.2

Re: [PATCH] ring_buffer: use list_replace() in rb_get_reader_page()

[Steven Rostedt]

On Mon, 25 May 2009, Xiao Guangrong wrote:

> Use list_replace() to splice list in rb_get_reader_page()

This code is very sensitive to change. But I think you actually uncovered 
a potential bug. There was a reason that I did not use list_replace, and 
it is because we need memory barriers here. And that is the bug you 
uncovered. We don't have memory barriers.

Ah, it's not a bug (yet). When we make this lockless we will need to add 
memory barriers here. Once reader->list.prev->next points to 
cpu_buffer->reader_page->list, the cpu_buffer->reader_page->list had 
better have both next and prev visible.

If we add this code now, we will just have to remove it when it goes 
lockless.

-- Steve



> 
> [ Impact: cleanup ]
> 
> Signed-off-by: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com>
> ---
>  kernel/trace/ring_buffer.c |   12 ++----------
>  1 files changed, 2 insertions(+), 10 deletions(-)
> 
> diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c
> index f558b37..f88bfcc 100644
> --- a/kernel/trace/ring_buffer.c
> +++ b/kernel/trace/ring_buffer.c
> @@ -2179,22 +2179,14 @@ rb_get_reader_page(struct ring_buffer_per_cpu *cpu_buffer)
>  	if (cpu_buffer->commit_page == cpu_buffer->reader_page)
>  		goto out;
>  
> -	/*
> -	 * Splice the empty reader page into the list around the head.
> -	 * Reset the reader page to size zero.
> -	 */
> -
> +	/* Reset the reader page to size zero */
>  	reader = cpu_buffer->head_page;
> -	cpu_buffer->reader_page->list.next = reader->list.next;
> -	cpu_buffer->reader_page->list.prev = reader->list.prev;
> -
>  	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);
>  
>  	/* Make the reader page now replace the head */
> -	reader->list.prev->next = &cpu_buffer->reader_page->list;
> -	reader->list.next->prev = &cpu_buffer->reader_page->list;
> +	list_replace(&reader->list, &cpu_buffer->reader_page->list);
>  
>  	/*
>  	 * If the tail is on the reader, then we must set the head
> -- 
> 1.6.1.2
> 
> 
>