On Thu, Nov 06, 2014 at 09:58:06AM -0500, [email protected] wrote:
> From: Kan Liang <[email protected]>
>
SNIP
> + /* LBR call stack */
> + if (lbr) {
> + struct branch_stack *lbr_stack = sample->branch_stack;
> + int lbr_nr = lbr_stack->nr;
> + int mix_chain_nr;
>
> - if (callchain_param.order == ORDER_CALLEE)
> - j = i;
> - else
> - j = chain->nr - i - 1;
> + for (i = 0; i < chain_nr; i++) {
> + if (chain->ips[i] == PERF_CONTEXT_USER)
> + break;
> + }
>
> -#ifdef HAVE_SKIP_CALLCHAIN_IDX
> - if (j == skip_idx)
> - continue;
> -#endif
> - ip = chain->ips[j];
> + /* LBR only affects the user callchain */
> + if (i == chain_nr) {
> + lbr = 0;
> + goto again;
> + }
'goto again' sounds like u want to try something again,
but you prepare the condition already with lbr = 0,
could you please restruct the code in another way?
SNIP
> + goto exit;
> }
> + } else {
>
> - err = callchain_cursor_append(&callchain_cursor,
> - ip, al.map, al.sym);
> - if (err)
> - return err;
> - }
> + /*
> + * Based on DWARF debug information, some architectures skip
> + * a callchain entry saved by the kernel.
> + */
> + skip_idx = arch_skip_callchain_idx(thread, chain);
>
> - return 0;
> + for (i = 0; i < chain_nr; i++) {
> + struct addr_location al;
> +
> + if (callchain_param.order == ORDER_CALLEE)
> + j = i;
> + else
> + j = chain->nr - i - 1;
> +
> +#ifdef HAVE_SKIP_CALLCHAIN_IDX
> + if (j == skip_idx)
> + continue;
> +#endif
> + ip = chain->ips[j];
> + err = __thread__resolve_callchain_sample(thread,
Uou factored out the common functionality into
__thread__resolve_callchain_sample
function and added your own functionality.. could you please split this into 2
separate patches? (first the new function, then your change)
IMO It'll make the change simple and more obvious.
thanks,
jirka
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
