Home | History | Annotate | Line # | Download | only in kern
subr_iostat.c revision 1.17
      1 /*	$NetBSD: subr_iostat.c,v 1.17 2009/04/04 07:30:10 ad Exp $	*/
      2 /*	NetBSD: subr_disk.c,v 1.69 2005/05/29 22:24:15 christos Exp	*/
      3 
      4 /*-
      5  * Copyright (c) 1996, 1997, 1999, 2000, 2009 The NetBSD Foundation, Inc.
      6  * All rights reserved.
      7  *
      8  * This code is derived from software contributed to The NetBSD Foundation
      9  * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
     10  * NASA Ames Research Center.
     11  *
     12  * Redistribution and use in source and binary forms, with or without
     13  * modification, are permitted provided that the following conditions
     14  * are met:
     15  * 1. Redistributions of source code must retain the above copyright
     16  *    notice, this list of conditions and the following disclaimer.
     17  * 2. Redistributions in binary form must reproduce the above copyright
     18  *    notice, this list of conditions and the following disclaimer in the
     19  *    documentation and/or other materials provided with the distribution.
     20  *
     21  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     22  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     23  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     24  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     25  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     26  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     27  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     28  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     29  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     30  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     31  * POSSIBILITY OF SUCH DAMAGE.
     32  */
     33 
     34 /*
     35  * Copyright (c) 1982, 1986, 1988, 1993
     36  *	The Regents of the University of California.  All rights reserved.
     37  * (c) UNIX System Laboratories, Inc.
     38  * All or some portions of this file are derived from material licensed
     39  * to the University of California by American Telephone and Telegraph
     40  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
     41  * the permission of UNIX System Laboratories, Inc.
     42  *
     43  * Redistribution and use in source and binary forms, with or without
     44  * modification, are permitted provided that the following conditions
     45  * are met:
     46  * 1. Redistributions of source code must retain the above copyright
     47  *    notice, this list of conditions and the following disclaimer.
     48  * 2. Redistributions in binary form must reproduce the above copyright
     49  *    notice, this list of conditions and the following disclaimer in the
     50  *    documentation and/or other materials provided with the distribution.
     51  * 3. Neither the name of the University nor the names of its contributors
     52  *    may be used to endorse or promote products derived from this software
     53  *    without specific prior written permission.
     54  *
     55  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     56  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     57  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     58  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     59  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     60  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     61  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     62  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     63  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     64  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     65  * SUCH DAMAGE.
     66  *
     67  *	@(#)ufs_disksubr.c	8.5 (Berkeley) 1/21/94
     68  */
     69 
     70 #include <sys/cdefs.h>
     71 __KERNEL_RCSID(0, "$NetBSD: subr_iostat.c,v 1.17 2009/04/04 07:30:10 ad Exp $");
     72 
     73 #include <sys/param.h>
     74 #include <sys/kernel.h>
     75 #include <sys/kmem.h>
     76 #include <sys/iostat.h>
     77 #include <sys/sysctl.h>
     78 #include <sys/rwlock.h>
     79 
     80 /*
     81  * Function prototypes for sysctl nodes
     82  */
     83 static int	sysctl_hw_disknames(SYSCTLFN_PROTO);
     84 static int	sysctl_hw_iostatnames(SYSCTLFN_PROTO);
     85 static int	sysctl_hw_iostats(SYSCTLFN_PROTO);
     86 
     87 static int
     88 iostati_getnames(int disk_only, char *oldp, size_t *oldlenp, const void *newp,
     89 		u_int namelen);
     90 
     91 /*
     92  * A global list of all drives attached to the system.  May grow or
     93  * shrink over time.
     94  */
     95 struct iostatlist_head iostatlist = TAILQ_HEAD_INITIALIZER(iostatlist);
     96 int iostat_count;		/* number of drives in global drivelist */
     97 krwlock_t iostatlist_lock;
     98 
     99 /*
    100  * Initialise the iostat subsystem.
    101  */
    102 void
    103 iostat_init(void)
    104 {
    105 
    106 	rw_init(&iostatlist_lock);
    107 }
    108 
    109 /*
    110  * Searches the iostatlist for the iostat corresponding to the
    111  * name provided.
    112  */
    113 struct io_stats *
    114 iostat_find(const char *name)
    115 {
    116 	struct io_stats *iostatp;
    117 
    118 	KASSERT(name != NULL);
    119 
    120 	rw_enter(&iostatlist_lock, RW_READER);
    121 	TAILQ_FOREACH(iostatp, &iostatlist, io_link) {
    122 		if (strcmp(iostatp->io_name, name) == 0) {
    123 			break;
    124 		}
    125 	}
    126 	rw_exit(&iostatlist_lock);
    127 
    128 	return iostatp;
    129 }
    130 
    131 /*
    132  * Allocate and initialise memory for the i/o statistics.
    133  */
    134 struct io_stats *
    135 iostat_alloc(int32_t type, void *parent, const char *name)
    136 {
    137 	struct io_stats *stats;
    138 
    139 	stats = kmem_zalloc(sizeof(*stats), KM_SLEEP);
    140 	if (stats == NULL)
    141 		panic("iostat_alloc: cannot allocate memory for stats buffer");
    142 
    143 	stats->io_type = type;
    144 	stats->io_parent = parent;
    145 	(void)strlcpy(stats->io_name, name, sizeof(stats->io_name));
    146 
    147 	/*
    148 	 * Set the attached timestamp.
    149 	 */
    150 	getmicrouptime(&stats->io_attachtime);
    151 
    152 	/*
    153 	 * Link into the drivelist.
    154 	 */
    155 	rw_enter(&iostatlist_lock, RW_WRITER);
    156 	TAILQ_INSERT_TAIL(&iostatlist, stats, io_link);
    157 	iostat_count++;
    158 	rw_exit(&iostatlist_lock);
    159 
    160 	return stats;
    161 }
    162 
    163 /*
    164  * Remove i/o from stats collection.
    165  */
    166 void
    167 iostat_free(struct io_stats *stats)
    168 {
    169 
    170 	/*
    171 	 * Remove from the iostat list.
    172 	 */
    173 	if (iostat_count == 0)
    174 		panic("iostat_free: iostat_count == 0");
    175 	rw_enter(&iostatlist_lock, RW_WRITER);
    176 	TAILQ_REMOVE(&iostatlist, stats, io_link);
    177 	iostat_count--;
    178 	rw_exit(&iostatlist_lock);
    179 	kmem_free(stats, sizeof(*stats));
    180 }
    181 
    182 /*
    183  * Increment a iostat busy counter.  If the counter is going from
    184  * 0 to 1, set the timestamp.
    185  */
    186 void
    187 iostat_busy(struct io_stats *stats)
    188 {
    189 
    190 	if (stats->io_busy++ == 0)
    191 		getmicrouptime(&stats->io_timestamp);
    192 }
    193 
    194 /*
    195  * Decrement a iostat busy counter, increment the byte count, total busy
    196  * time, and reset the timestamp.
    197  */
    198 void
    199 iostat_unbusy(struct io_stats *stats, long bcount, int read)
    200 {
    201 	struct timeval dv_time, diff_time;
    202 
    203 	if (stats->io_busy-- == 0) {
    204 		printf("%s: busy < 0\n", stats->io_name);
    205 		panic("iostat_unbusy");
    206 	}
    207 
    208 	getmicrouptime(&dv_time);
    209 
    210 	timersub(&dv_time, &stats->io_timestamp, &diff_time);
    211 	timeradd(&stats->io_time, &diff_time, &stats->io_time);
    212 
    213 	stats->io_timestamp = dv_time;
    214 	if (bcount > 0) {
    215 		if (read) {
    216 			stats->io_rbytes += bcount;
    217 			stats->io_rxfer++;
    218 		} else {
    219 			stats->io_wbytes += bcount;
    220 			stats->io_wxfer++;
    221 		}
    222 	}
    223 }
    224 
    225 /*
    226  * Return non-zero if a device has an I/O request in flight.
    227  */
    228 bool
    229 iostat_isbusy(struct io_stats *stats)
    230 {
    231 
    232 	return stats->io_busy != 0;
    233 }
    234 
    235 /*
    236  * Increment the seek counter.  This does look almost redundant but it
    237  * abstracts the stats gathering.
    238  */
    239 void
    240 iostat_seek(struct io_stats *stats)
    241 {
    242 
    243 	stats->io_seek++;
    244 }
    245 
    246 static int
    247 sysctl_hw_disknames(SYSCTLFN_ARGS)
    248 {
    249 
    250 	return iostati_getnames(1, oldp, oldlenp, newp, namelen);
    251 }
    252 
    253 static int
    254 sysctl_hw_iostatnames(SYSCTLFN_ARGS)
    255 {
    256 
    257 	return iostati_getnames(0, oldp, oldlenp, newp, namelen);
    258 }
    259 
    260 static int
    261 iostati_getnames(int disk_only, char *oldp, size_t *oldlenp, const void *newp,
    262 		 u_int namelen)
    263 {
    264 	char bf[IOSTATNAMELEN + 1];
    265 	char *where = oldp;
    266 	struct io_stats *stats;
    267 	size_t needed, left, slen;
    268 	int error, first;
    269 
    270 	if (newp != NULL)
    271 		return (EPERM);
    272 	if (namelen != 0)
    273 		return (EINVAL);
    274 
    275 	first = 1;
    276 	error = 0;
    277 	needed = 0;
    278 	left = *oldlenp;
    279 
    280 	rw_enter(&iostatlist_lock, RW_READER);
    281 	for (stats = TAILQ_FIRST(&iostatlist); stats != NULL;
    282 	    stats = TAILQ_NEXT(stats, io_link)) {
    283 		if ((disk_only == 1) && (stats->io_type != IOSTAT_DISK))
    284 			continue;
    285 
    286 		if (where == NULL)
    287 			needed += strlen(stats->io_name) + 1;
    288 		else {
    289 			memset(bf, 0, sizeof(bf));
    290 			if (first) {
    291 				strncpy(bf, stats->io_name, sizeof(bf));
    292 				first = 0;
    293 			} else {
    294 				bf[0] = ' ';
    295 				strncpy(bf + 1, stats->io_name,
    296 				    sizeof(bf) - 1);
    297 			}
    298 			bf[IOSTATNAMELEN] = '\0';
    299 			slen = strlen(bf);
    300 			if (left < slen + 1)
    301 				break;
    302 			/* +1 to copy out the trailing NUL byte */
    303 			error = copyout(bf, where, slen + 1);
    304 			if (error)
    305 				break;
    306 			where += slen;
    307 			needed += slen;
    308 			left -= slen;
    309 		}
    310 	}
    311 	rw_exit(&iostatlist_lock);
    312 	*oldlenp = needed;
    313 	return (error);
    314 }
    315 
    316 static int
    317 sysctl_hw_iostats(SYSCTLFN_ARGS)
    318 {
    319 	struct io_sysctl sdrive;
    320 	struct io_stats *stats;
    321 	char *where = oldp;
    322 	size_t tocopy, left;
    323 	int error;
    324 
    325 	if (newp != NULL)
    326 		return (EPERM);
    327 
    328 	/*
    329 	 * The original hw.diskstats call was broken and did not require
    330 	 * the userland to pass in it's size of struct disk_sysctl.  This
    331 	 * was fixed after NetBSD 1.6 was released.
    332 	 */
    333 	if (namelen == 0)
    334 		tocopy = offsetof(struct io_sysctl, busy);
    335 	else
    336 		tocopy = name[0];
    337 
    338 	if (where == NULL) {
    339 		*oldlenp = iostat_count * tocopy;
    340 		return (0);
    341 	}
    342 
    343 	error = 0;
    344 	left = *oldlenp;
    345 	memset(&sdrive, 0, sizeof(sdrive));
    346 	*oldlenp = 0;
    347 
    348 	rw_enter(&iostatlist_lock, RW_READER);
    349 	TAILQ_FOREACH(stats, &iostatlist, io_link) {
    350 		if (left < tocopy)
    351 			break;
    352 		strncpy(sdrive.name, stats->io_name, sizeof(sdrive.name));
    353 		sdrive.xfer = stats->io_rxfer + stats->io_wxfer;
    354 		sdrive.rxfer = stats->io_rxfer;
    355 		sdrive.wxfer = stats->io_wxfer;
    356 		sdrive.seek = stats->io_seek;
    357 		sdrive.bytes = stats->io_rbytes + stats->io_wbytes;
    358 		sdrive.rbytes = stats->io_rbytes;
    359 		sdrive.wbytes = stats->io_wbytes;
    360 		sdrive.attachtime_sec = stats->io_attachtime.tv_sec;
    361 		sdrive.attachtime_usec = stats->io_attachtime.tv_usec;
    362 		sdrive.timestamp_sec = stats->io_timestamp.tv_sec;
    363 		sdrive.timestamp_usec = stats->io_timestamp.tv_usec;
    364 		sdrive.time_sec = stats->io_time.tv_sec;
    365 		sdrive.time_usec = stats->io_time.tv_usec;
    366 		sdrive.busy = stats->io_busy;
    367 
    368 		error = copyout(&sdrive, where, min(tocopy, sizeof(sdrive)));
    369 		if (error)
    370 			break;
    371 		where += tocopy;
    372 		*oldlenp += tocopy;
    373 		left -= tocopy;
    374 	}
    375 	rw_exit(&iostatlist_lock);
    376 	return (error);
    377 }
    378 
    379 SYSCTL_SETUP(sysctl_io_stats_setup, "sysctl i/o stats setup")
    380 {
    381 
    382 	sysctl_createv(clog, 0, NULL, NULL,
    383 		       CTLFLAG_PERMANENT,
    384 		       CTLTYPE_STRING, "disknames",
    385 		       SYSCTL_DESCR("List of disk drives present"),
    386 		       sysctl_hw_disknames, 0, NULL, 0,
    387 		       CTL_HW, HW_DISKNAMES, CTL_EOL);
    388 	sysctl_createv(clog, 0, NULL, NULL,
    389 		       CTLFLAG_PERMANENT,
    390 		       CTLTYPE_STRING, "iostatnames",
    391 		       SYSCTL_DESCR("I/O stats are being collected for these"
    392 				    " devices"),
    393 		       sysctl_hw_iostatnames, 0, NULL, 0,
    394 		       CTL_HW, HW_IOSTATNAMES, CTL_EOL);
    395 	sysctl_createv(clog, 0, NULL, NULL,
    396 		       CTLFLAG_PERMANENT,
    397 		       CTLTYPE_STRUCT, "iostats",
    398 		       SYSCTL_DESCR("Statistics on device I/O operations"),
    399 		       sysctl_hw_iostats, 0, NULL, 0,
    400 		       CTL_HW, HW_IOSTATS, CTL_EOL);
    401 }
    402