Messages in this thread | | | Date | Wed, 23 Dec 2009 13:36:58 -0500 | From | Neil Horman <> | Subject | Re: [PATCH] Fix tracing infrastructure to support multiple includes when defining CREATE_TRACE_POINTS |
| |
On Wed, Dec 23, 2009 at 08:58:33AM -0500, Steven Rostedt wrote: > On Wed, 2009-12-23 at 07:02 -0500, Neil Horman wrote: > > On Tue, Dec 22, 2009 at 01:03:59PM -0500, Steven Rostedt wrote: > > > On Wed, 2009-12-16 at 11:04 -0500, Neil Horman wrote: > > > > > > > > > > +#if !defined(_TRACE_NAPI_H_) || defined(TRACE_HEADER_MULTI_READ) > > > > #define _TRACE_NAPI_H_ > > > > > > > > > > Ug, I'm still getting panics because of this. I found out that it's > > > because napi->dev can be NULL. > > > > > > Here's what I did to solve it: > > Gahhh, you're right, We had to do the same thing in net-next's commit > > f2798eb4e01b095f273f4bf40f511c9d69c0e1da. If you queue to the backlog, the > > backlog napi instance has no net device and so its null. This is exactly the > > right fix. > > OK, I'll update the patch and apply it. > > But this doesn't need to go into 33 correct. We can let it brew for 34? > Yeah, the thing I was depending on it for got NAK'ed anyway, so its just a straight bugfix + a cleanup of the napi_poll tracepoint. It can definately wait until 34.
Thanks! Neil
> -- Steve > > >
| |