Re: [RFC][PATCH] tracing/x86: Save CR2 before tracing irqsoff on error_entry

[Peter Zijlstra]

On Thu, Mar 21, 2019 at 09:33:17AM +0100, Peter Zijlstra wrote:

I'm thinking this problem wasn't new.

> idtentry page_fault             do_page_fault           has_error_code=1
>   call error_entry
>     TRACE_IRQS_OFF
>       call trace_hardirqs_off*
>         <tracer stuff>
> 	  <fault> # modifies CR2

      CALL_enter_from_user_mode
        __context_tracking_exit()
	  trace_user_exit(0)
	    #PF

>   call do_page_fault
>     address = read_cr2(); /* whoopsie */

And that also isn't fixed by your patch.

I'm trying to make idtentry put cr2 in rdx, such that do_page_fault()
takes address as a 3rd argument, but I'm still fighting that context
tracking nonsense.