1 /* 2 * Copyright (c) 1997, 1998, 1999 Kenneth D. Merry. 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. The name of the author may not be used to endorse or promote products 14 * derived from this software without specific prior written permission. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 26 * SUCH DAMAGE. 27 * 28 * $FreeBSD$ 29 */ 30 31 #include <sys/param.h> 32 #include <sys/kernel.h> 33 #include <sys/systm.h> 34 #include <sys/time.h> 35 #include <sys/types.h> 36 #include <sys/buf.h> 37 #include <sys/sysctl.h> 38 39 #include <sys/devicestat.h> 40 41 static int devstat_num_devs; 42 static long devstat_generation; 43 static int devstat_version = DEVSTAT_VERSION; 44 static int devstat_current_devnumber; 45 46 static STAILQ_HEAD(devstatlist, devstat) device_statq; 47 48 /* 49 * Take a malloced and zeroed devstat structure given to us, fill it in 50 * and add it to the queue of devices. 51 */ 52 void 53 devstat_add_entry(struct devstat *ds, const char *dev_name, 54 int unit_number, u_int32_t block_size, 55 devstat_support_flags flags, 56 devstat_type_flags device_type, 57 devstat_priority priority) 58 { 59 int s; 60 struct devstatlist *devstat_head; 61 struct devstat *ds_tmp; 62 63 if (ds == NULL) 64 return; 65 66 if (devstat_num_devs == 0) 67 STAILQ_INIT(&device_statq); 68 69 devstat_generation++; 70 devstat_num_devs++; 71 72 devstat_head = &device_statq; 73 74 /* 75 * Priority sort. Each driver passes in its priority when it adds 76 * its devstat entry. Drivers are sorted first by priority, and 77 * then by probe order. 78 * 79 * For the first device, we just insert it, since the priority 80 * doesn't really matter yet. Subsequent devices are inserted into 81 * the list using the order outlined above. 82 */ 83 if (devstat_num_devs == 1) 84 STAILQ_INSERT_TAIL(devstat_head, ds, dev_links); 85 else { 86 for (ds_tmp = STAILQ_FIRST(devstat_head); ds_tmp != NULL; 87 ds_tmp = STAILQ_NEXT(ds_tmp, dev_links)) { 88 struct devstat *ds_next; 89 90 ds_next = STAILQ_NEXT(ds_tmp, dev_links); 91 92 /* 93 * If we find a break between higher and lower 94 * priority items, and if this item fits in the 95 * break, insert it. This also applies if the 96 * "lower priority item" is the end of the list. 97 */ 98 if ((priority <= ds_tmp->priority) 99 && ((ds_next == NULL) 100 || (priority > ds_next->priority))) { 101 STAILQ_INSERT_AFTER(devstat_head, ds_tmp, ds, 102 dev_links); 103 break; 104 } else if (priority > ds_tmp->priority) { 105 /* 106 * If this is the case, we should be able 107 * to insert ourselves at the head of the 108 * list. If we can't, something is wrong. 109 */ 110 if (ds_tmp == STAILQ_FIRST(devstat_head)) { 111 STAILQ_INSERT_HEAD(devstat_head, 112 ds, dev_links); 113 break; 114 } else { 115 STAILQ_INSERT_TAIL(devstat_head, 116 ds, dev_links); 117 printf("devstat_add_entry: HELP! " 118 "sorting problem detected " 119 "for %s%d\n", dev_name, 120 unit_number); 121 break; 122 } 123 } 124 } 125 } 126 127 ds->device_number = devstat_current_devnumber++; 128 ds->unit_number = unit_number; 129 strncpy(ds->device_name, dev_name, DEVSTAT_NAME_LEN); 130 ds->device_name[DEVSTAT_NAME_LEN - 1] = '\0'; 131 ds->block_size = block_size; 132 ds->flags = flags; 133 ds->device_type = device_type; 134 ds->priority = priority; 135 136 s = splclock(); 137 getmicrotime(&ds->dev_creation_time); 138 splx(s); 139 } 140 141 /* 142 * Remove a devstat structure from the list of devices. 143 */ 144 void 145 devstat_remove_entry(struct devstat *ds) 146 { 147 struct devstatlist *devstat_head; 148 149 if (ds == NULL) 150 return; 151 152 devstat_generation++; 153 devstat_num_devs--; 154 155 devstat_head = &device_statq; 156 157 /* Remove this entry from the devstat queue */ 158 STAILQ_REMOVE(devstat_head, ds, devstat, dev_links); 159 } 160 161 /* 162 * Record a transaction start. 163 */ 164 void 165 devstat_start_transaction(struct devstat *ds) 166 { 167 int s; 168 169 /* sanity check */ 170 if (ds == NULL) 171 return; 172 173 /* 174 * We only want to set the start time when we are going from idle 175 * to busy. The start time is really the start of the latest busy 176 * period. 177 */ 178 if (ds->busy_count == 0) { 179 s = splclock(); 180 getmicrouptime(&ds->start_time); 181 splx(s); 182 } 183 ds->busy_count++; 184 } 185 186 /* 187 * Record the ending of a transaction, and incrment the various counters. 188 */ 189 void 190 devstat_end_transaction(struct devstat *ds, u_int32_t bytes, 191 devstat_tag_type tag_type, devstat_trans_flags flags) 192 { 193 int s; 194 struct timeval busy_time; 195 196 /* sanity check */ 197 if (ds == NULL) 198 return; 199 200 s = splclock(); 201 getmicrouptime(&ds->last_comp_time); 202 splx(s); 203 204 ds->busy_count--; 205 206 /* 207 * There might be some transactions (DEVSTAT_NO_DATA) that don't 208 * transfer any data. 209 */ 210 if (flags == DEVSTAT_READ) { 211 ds->bytes_read += bytes; 212 ds->num_reads++; 213 } else if (flags == DEVSTAT_WRITE) { 214 ds->bytes_written += bytes; 215 ds->num_writes++; 216 } else if (flags == DEVSTAT_FREE) { 217 ds->bytes_freed += bytes; 218 ds->num_frees++; 219 } else 220 ds->num_other++; 221 222 /* 223 * Keep a count of the various tag types sent. 224 */ 225 if ((ds->flags & DEVSTAT_NO_ORDERED_TAGS) == 0 && 226 tag_type != DEVSTAT_TAG_NONE) 227 ds->tag_types[tag_type]++; 228 229 /* 230 * We only update the busy time when we go idle. Otherwise, this 231 * calculation would require many more clock cycles. 232 */ 233 if (ds->busy_count == 0) { 234 /* Calculate how long we were busy */ 235 busy_time = ds->last_comp_time; 236 timevalsub(&busy_time, &ds->start_time); 237 238 /* Add our busy time to the total busy time. */ 239 timevaladd(&ds->busy_time, &busy_time); 240 } else if (ds->busy_count < 0) 241 printf("devstat_end_transaction: HELP!! busy_count " 242 "for %s%d is < 0 (%d)!\n", ds->device_name, 243 ds->unit_number, ds->busy_count); 244 } 245 246 void 247 devstat_end_transaction_buf(struct devstat *ds, struct buf *bp) 248 { 249 devstat_trans_flags flg; 250 251 if (bp->b_flags & B_FREEBUF) 252 flg = DEVSTAT_FREE; 253 else if (bp->b_flags & B_READ) 254 flg = DEVSTAT_READ; 255 else 256 flg = DEVSTAT_WRITE; 257 258 devstat_end_transaction(ds, bp->b_bcount - bp->b_resid, 259 (bp->b_flags & B_ORDERED) ? 260 DEVSTAT_TAG_ORDERED : DEVSTAT_TAG_SIMPLE, flg); 261 } 262 263 /* 264 * This is the sysctl handler for the devstat package. The data pushed out 265 * on the kern.devstat.all sysctl variable consists of the current devstat 266 * generation number, and then an array of devstat structures, one for each 267 * device in the system. 268 * 269 * I'm really not too fond of this method of doing things, but there really 270 * aren't that many alternatives. We must have some method of making sure 271 * that the generation number the user gets corresponds with the data the 272 * user gets. If the user makes a separate sysctl call to get the 273 * generation, and then a sysctl call to get the device statistics, the 274 * device list could have changed in that brief period of time. By 275 * supplying the generation number along with the statistics output, we can 276 * guarantee that the generation number and the statistics match up. 277 */ 278 static int 279 sysctl_devstat SYSCTL_HANDLER_ARGS 280 { 281 int error, i; 282 struct devstat *nds; 283 struct devstatlist *devstat_head; 284 285 if (devstat_num_devs == 0) 286 return(EINVAL); 287 288 error = 0; 289 devstat_head = &device_statq; 290 291 /* 292 * First push out the generation number. 293 */ 294 error = SYSCTL_OUT(req, &devstat_generation, sizeof(long)); 295 296 /* 297 * Now push out all the devices. 298 */ 299 for (i = 0, nds = devstat_head->stqh_first; 300 (nds != NULL) && (i < devstat_num_devs) && (error == 0); 301 nds = nds->dev_links.stqe_next, i++) 302 error = SYSCTL_OUT(req, nds, sizeof(struct devstat)); 303 304 return(error); 305 } 306 307 /* 308 * Sysctl entries for devstat. The first one is a node that all the rest 309 * hang off of. 310 */ 311 SYSCTL_NODE(_kern, OID_AUTO, devstat, CTLFLAG_RD, 0, "Device Statistics"); 312 313 SYSCTL_PROC(_kern_devstat, OID_AUTO, all, CTLFLAG_RD|CTLTYPE_OPAQUE, 314 0, 0, sysctl_devstat, "S,devstat", "All devices in the devstat list"); 315 /* 316 * Export the number of devices in the system so that userland utilities 317 * can determine how much memory to allocate to hold all the devices. 318 */ 319 SYSCTL_INT(_kern_devstat, OID_AUTO, numdevs, CTLFLAG_RD, 320 &devstat_num_devs, 0, "Number of devices in the devstat list"); 321 SYSCTL_LONG(_kern_devstat, OID_AUTO, generation, CTLFLAG_RD, 322 &devstat_generation, "Devstat list generation"); 323 SYSCTL_INT(_kern_devstat, OID_AUTO, version, CTLFLAG_RD, 324 &devstat_version, 0, "Devstat list version number"); 325