On 12/2/2015 4:11 PM, Shi, Yang wrote:
On 12/2/2015 3:36 PM, Dave Hansen wrote:
On 12/02/2015 02:53 PM, Yang Shi wrote:
diff --git a/mm/gup.c b/mm/gup.c index deafa2c..10245a4 100644 --- a/mm/gup.c +++ b/mm/gup.c @@ -13,6 +13,9 @@ #include <linux/rwsem.h> #include <linux/hugetlb.h>
+#define CREATE_TRACE_POINTS +#include <trace/events/gup.h>
- #include <asm/pgtable.h> #include <asm/tlbflush.h>
This needs to be _the_ last thing that gets #included. Otherwise, you risk colliding with any other trace header that gets implicitly included below.
Thanks for the suggestion, will move it to the last.
@@ -1340,6 +1346,8 @@ int __get_user_pages_fast(unsigned long start, int nr_pages, int write, start, len))) return 0;
- trace_gup_get_user_pages_fast(start, nr_pages, write, pages);
/* * Disable interrupts. We use the nested form as we canalready have * interrupts disabled by get_futex_key.
It would be _really_ nice to be able to see return values from the various gup calls as well. Is that feasible?
I think it should be feasible. kmem_cache_alloc trace event could show return value. I'm supposed gup trace events should be able to do the same thing.
Just did a quick test, it is definitely feasible. Please check the below test log:
trace-cmd-200 [000] 99.221486: gup_get_user_pages: start=8000000ff0 nr_pages=1 ret=1 trace-cmd-200 [000] 99.223215: gup_get_user_pages: start=8000000fdb nr_pages=1 ret=1 trace-cmd-200 [000] 99.223298: gup_get_user_pages: start=8000000ed0 nr_pages=1 ret=1
nr_pages is the number of pages requested by the gup, ret is the return value.
If nobody has objection, I will add it into V3.
Regards, Yang
Regards, Yang