Re: Potential Issue in Tracing Ring Buffer

[Steven Rostedt]

On Thu, 26 Nov 2020 06:52:45 +0100
Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:

> On Tue, Nov 24, 2020 at 10:39:17PM +0000, J. Avila wrote:
> > Hello,
> > 
> > In the ftrace logs we've collected internally, we have found that there are
> > situations where time seems to go backwards; this breaks userspace tools which
> > expect time to always go forward in these logs. For example, in this snippet
> > from a db845c running a 5.10-rc4 kernel[1] (thanks for getting us the trace,
> > John!), we see:  
> 
> Does the patch at:
> 	https://lore.kernel.org/r/20201125225654.1618966-1-minchan@xxxxxxxxxx
> 
> resolve this issue for you?
> 

I think I found the bug. Can you apply this patch and let me know if it
fixes the issue for you?

-- Steve

diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c
index dc83b3fa9fe7..bccaf88d3706 100644
--- a/kernel/trace/ring_buffer.c
+++ b/kernel/trace/ring_buffer.c
@@ -3291,7 +3291,7 @@ __rb_reserve_next(struct ring_buffer_per_cpu *cpu_buffer,
 			/* Nothing came after this event between C and E */
 			info->delta = ts - info->after;
 			(void)rb_time_cmpxchg(&cpu_buffer->write_stamp,
-					      info->after, info->ts);
+					      info->after, ts);
 			info->ts = ts;
 		} else {
 			/*