#!/usr/bin/stap
/*
 * biolatency-nd.stp	Measure storage (bio) latency distribution.
 *			For Linux, uses SystemTap (non-debuginfo).
 *
 * USAGE: ./biolatency-nd.stp [interval [count]]
 *
 * This script uses the kernel tracepoints block_rq_issue and block_rq_complete,
 * to trace the time from storage request to completion. To include extra time
 * spent queued in the kernel, change block_rq_issue to block_rq_insert.
 *
 * From systemtap-lwtools: https://github.com/brendangregg/systemtap-lwtools
 *
 * See the corresponding man page (in systemtap-lwtools) for more info.
 *
 * Copyright (C) 2015 Brendan Gregg.
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * 01-Feb-2015	Brendan Gregg	Created this.
 */

global ts[65536];	# 65536 is max concurrent I/O
global lat, interval = 0, secs = 0, count = 0, output = 0;

probe begin
{
	printf("Tracing block I/O... ");
	if (argv_1 != "") {
		interval = strtol(argv_1, 10);
		if (argv_2 != "") {
			count = strtol(argv_2, 10);
		}
		printf("Output every %d secs.\n", interval);
	} else {
		printf("Hit Ctrl-C to end.\n");
	}
}

probe kernel.trace("block_rq_issue") {
	ts[$rq] = gettimeofday_ns();
}

probe kernel.trace("block_rq_complete") {
	if (ts[$rq]) {
		lat <<< gettimeofday_ns() - ts[$rq];
		delete ts[$rq];
	}
}

function print_report()
{
	printf("\nbio latency (ns):\n");
	if (@count(lat)) { print(@hist_log(lat)); }
	delete lat;
}

probe timer.s(1)
{
	if (interval) {
		if (++secs == interval) {
			print_report();
			secs = 0;
			count && (++output == count) && exit();
		}
	}
}

probe end
{
	!interval && print_report();
	delete secs;
	delete output;
}