Re: [PATCH] tracing: add trace event for memory-failure

[Naoya Horiguchi]

On Fri, Mar 13, 2015 at 06:10:51PM +0800, Xie XiuQi wrote:
> Memory-failure as the high level machine check handler, it's necessary
> to report memory page recovery action result to user space by ftrace.
> 
> This patch add a event at ras group for memory-failure.
> 
> The output like below:
> # tracer: nop
> #
> # entries-in-buffer/entries-written: 2/2   #P:24
> #
> #                              _-----=> irqs-off
> #                             / _----=> need-resched
> #                            | / _---=> hardirq/softirq
> #                            || / _--=> preempt-depth
> #                            ||| /     delay
> #           TASK-PID   CPU#  ||||    TIMESTAMP  FUNCTION
> #              | |       |   ||||       |         |
>       mce-inject-13150 [001] ....   277.019359: memory_failure_event: pfn 0x19869: free buddy page recovery: Delayed
> 
> Signed-off-by: Xie XiuQi <xiexiuqi@xxxxxxxxxx>
> ---
>  include/ras/ras_event.h |   36 ++++++++++++++++++++++++++++++++++++
>  mm/memory-failure.c     |    3 +++
>  2 files changed, 39 insertions(+), 0 deletions(-)
> 
> diff --git a/include/ras/ras_event.h b/include/ras/ras_event.h
> index 79abb9c..0a6c8f3 100644
> --- a/include/ras/ras_event.h
> +++ b/include/ras/ras_event.h
> @@ -232,6 +232,42 @@ TRACE_EVENT(aer_event,
>  		__print_flags(__entry->status, "|", aer_uncorrectable_errors))
>  );
>  
> +/*
> + * memory-failure recovery action result event
> + *
> + * unsigned long pfn -	Page Number of the corrupted page
> + * char * action -	Recovery action: "free buddy", "free huge", "high
> + *			order kernel", "free buddy, 2nd try", "different
> + *			compound page after locking", "hugepage already
> + *			hardware poisoned", "unmapping failed", "already
> + *			truncated LRU", etc.

Listing all possible values here might be prone to become out of date when
someone try to add/remove/change action_result() call sites.
So I'd like to have some enumerator to bundle these strings in one place
in mm/memory-failure.c.
# I feel like doing this later.

> + * char * result -	Action result: Ignored, Failed, Delayed, Recovered.

mm/memory-failure.c has good explanation:

  /*
   * Error handlers for various types of pages.
   */
  enum outcome {
          IGNORED,        /* Error: cannot be handled */
          FAILED,         /* Error: handling failed */
          DELAYED,        /* Will be handled later */
          RECOVERED,      /* Successfully recovered */
  };

So adding a reference to here looks better to me.

Thanks,
Naoya Horiguchi

> + */
> +TRACE_EVENT(memory_failure_event,
> +	TP_PROTO(const unsigned long pfn,
> +		 const char *action,
> +		 const char *result),
> +
> +	TP_ARGS(pfn, action, result),
> +
> +	TP_STRUCT__entry(
> +		__field(unsigned long, pfn)
> +		__string(action, action)
> +		__string(result, result)
> +	),
> +
> +	TP_fast_assign(
> +		__entry->pfn = pfn;
> +		__assign_str(action, action);
> +		__assign_str(result, result);
> +	),
> +
> +	TP_printk("pfn %#lx: %s page recovery: %s",
> +		__entry->pfn,
> +		__get_str(action),
> +		__get_str(result)
> +	)
> +);
>  #endif /* _TRACE_HW_EVENT_MC_H */
>  
>  /* This part must be outside protection */
> diff --git a/mm/memory-failure.c b/mm/memory-failure.c
> index d487f8d..86a9cce 100644
> --- a/mm/memory-failure.c
> +++ b/mm/memory-failure.c
> @@ -56,6 +56,7 @@
>  #include <linux/mm_inline.h>
>  #include <linux/kfifo.h>
>  #include "internal.h"
> +#include <ras/ras_event.h>
>  
>  int sysctl_memory_failure_early_kill __read_mostly = 0;
>  
> @@ -837,6 +838,8 @@ static struct page_state {
>   */
>  static void action_result(unsigned long pfn, char *msg, int result)
>  {
> +	trace_memory_failure_event(pfn, msg, action_name[result]);
> +
>  	pr_err("MCE %#lx: %s page recovery: %s\n",
>  		pfn, msg, action_name[result]);
>  }
> -- 
> 1.7.1
> --
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/