x86, trace: Further robustify CR2 handling vs tracing
authorPeter Zijlstra <peterz@infradead.org>
Wed, 5 Mar 2014 13:07:49 +0000 (14:07 +0100)
committerH. Peter Anvin <hpa@linux.intel.com>
Thu, 6 Mar 2014 18:58:18 +0000 (10:58 -0800)
commitd4078e232267ff53f3b030b9698a3c001db4dbec
treebb8658ced4e6610b36fbf3de05ae85c3585df850
parent0ac09f9f8cd1fb028a48330edba6023d347d3cea
x86, trace: Further robustify CR2 handling vs tracing

Building on commit 0ac09f9f8cd1 ("x86, trace: Fix CR2 corruption when
tracing page faults") this patch addresses another few issues:

 - Now that read_cr2() is lifted into trace_do_page_fault(), we should
   pass the address to trace_page_fault_entries() to avoid it
   re-reading a potentially changed cr2.

 - Put both trace_do_page_fault() and trace_page_fault_entries() under
   CONFIG_TRACING.

 - Mark both fault entry functions {,trace_}do_page_fault() as notrace
   to avoid getting __mcount or other function entry trace callbacks
   before we've observed CR2.

 - Mark __do_page_fault() as noinline to guarantee the function tracer
   does get to see the fault.

Cc: <jolsa@redhat.com>
Cc: <vincent.weaver@maine.edu>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20140306145300.GO9987@twins.programming.kicks-ass.net
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
arch/x86/mm/fault.c