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 *
| 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: head/sys/kern/subr_devstat.c 62454 2000-07-03 09:35:31Z phk $
| 28 * $FreeBSD: head/sys/kern/subr_devstat.c 62573 2000-07-04 11:25:35Z phk $
|
29 */
30
31#include <sys/param.h>
32#include <sys/kernel.h>
33#include <sys/systm.h>
34#include <sys/bio.h>
35#include <sys/sysctl.h>
36
37#include <sys/devicestat.h>
38
39static int devstat_num_devs;
40static long devstat_generation;
41static int devstat_version = DEVSTAT_VERSION;
42static int devstat_current_devnumber;
43
44static STAILQ_HEAD(devstatlist, devstat) device_statq;
45
46/*
47 * Take a malloced and zeroed devstat structure given to us, fill it in
48 * and add it to the queue of devices.
49 */
50void
51devstat_add_entry(struct devstat *ds, const char *dev_name,
52 int unit_number, u_int32_t block_size,
53 devstat_support_flags flags,
54 devstat_type_flags device_type,
55 devstat_priority priority)
56{
57 struct devstatlist *devstat_head;
58 struct devstat *ds_tmp;
59
60 if (ds == NULL)
61 return;
62
63 if (devstat_num_devs == 0)
64 STAILQ_INIT(&device_statq);
65
66 devstat_generation++;
67 devstat_num_devs++;
68
69 devstat_head = &device_statq;
70
71 /*
72 * Priority sort. Each driver passes in its priority when it adds
73 * its devstat entry. Drivers are sorted first by priority, and
74 * then by probe order.
75 *
76 * For the first device, we just insert it, since the priority
77 * doesn't really matter yet. Subsequent devices are inserted into
78 * the list using the order outlined above.
79 */
80 if (devstat_num_devs == 1)
81 STAILQ_INSERT_TAIL(devstat_head, ds, dev_links);
82 else {
83 for (ds_tmp = STAILQ_FIRST(devstat_head); ds_tmp != NULL;
84 ds_tmp = STAILQ_NEXT(ds_tmp, dev_links)) {
85 struct devstat *ds_next;
86
87 ds_next = STAILQ_NEXT(ds_tmp, dev_links);
88
89 /*
90 * If we find a break between higher and lower
91 * priority items, and if this item fits in the
92 * break, insert it. This also applies if the
93 * "lower priority item" is the end of the list.
94 */
95 if ((priority <= ds_tmp->priority)
96 && ((ds_next == NULL)
97 || (priority > ds_next->priority))) {
98 STAILQ_INSERT_AFTER(devstat_head, ds_tmp, ds,
99 dev_links);
100 break;
101 } else if (priority > ds_tmp->priority) {
102 /*
103 * If this is the case, we should be able
104 * to insert ourselves at the head of the
105 * list. If we can't, something is wrong.
106 */
107 if (ds_tmp == STAILQ_FIRST(devstat_head)) {
108 STAILQ_INSERT_HEAD(devstat_head,
109 ds, dev_links);
110 break;
111 } else {
112 STAILQ_INSERT_TAIL(devstat_head,
113 ds, dev_links);
114 printf("devstat_add_entry: HELP! "
115 "sorting problem detected "
116 "for %s%d\n", dev_name,
117 unit_number);
118 break;
119 }
120 }
121 }
122 }
123
124 ds->device_number = devstat_current_devnumber++;
125 ds->unit_number = unit_number;
126 strncpy(ds->device_name, dev_name, DEVSTAT_NAME_LEN);
127 ds->device_name[DEVSTAT_NAME_LEN - 1] = '\0';
128 ds->block_size = block_size;
129 ds->flags = flags;
130 ds->device_type = device_type;
131 ds->priority = priority;
132 getmicrotime(&ds->dev_creation_time);
133}
134
135/*
136 * Remove a devstat structure from the list of devices.
137 */
138void
139devstat_remove_entry(struct devstat *ds)
140{
141 struct devstatlist *devstat_head;
142
143 if (ds == NULL)
144 return;
145
146 devstat_generation++;
147 devstat_num_devs--;
148
149 devstat_head = &device_statq;
150
151 /* Remove this entry from the devstat queue */
152 STAILQ_REMOVE(devstat_head, ds, devstat, dev_links);
153}
154
155/*
156 * Record a transaction start.
157 */
158void
159devstat_start_transaction(struct devstat *ds)
160{
161 /* sanity check */
162 if (ds == NULL)
163 return;
164
165 /*
166 * We only want to set the start time when we are going from idle
167 * to busy. The start time is really the start of the latest busy
168 * period.
169 */
170 if (ds->busy_count == 0)
171 getmicrouptime(&ds->start_time);
172 ds->busy_count++;
173}
174
175/*
176 * Record the ending of a transaction, and incrment the various counters.
177 */
178void
179devstat_end_transaction(struct devstat *ds, u_int32_t bytes,
180 devstat_tag_type tag_type, devstat_trans_flags flags)
181{
182 struct timeval busy_time;
183
184 /* sanity check */
185 if (ds == NULL)
186 return;
187
188 getmicrouptime(&ds->last_comp_time);
189 ds->busy_count--;
190
191 /*
192 * There might be some transactions (DEVSTAT_NO_DATA) that don't
193 * transfer any data.
194 */
195 if (flags == DEVSTAT_READ) {
196 ds->bytes_read += bytes;
197 ds->num_reads++;
198 } else if (flags == DEVSTAT_WRITE) {
199 ds->bytes_written += bytes;
200 ds->num_writes++;
201 } else if (flags == DEVSTAT_FREE) {
202 ds->bytes_freed += bytes;
203 ds->num_frees++;
204 } else
205 ds->num_other++;
206
207 /*
208 * Keep a count of the various tag types sent.
209 */
210 if ((ds->flags & DEVSTAT_NO_ORDERED_TAGS) == 0 &&
211 tag_type != DEVSTAT_TAG_NONE)
212 ds->tag_types[tag_type]++;
213
214 /*
215 * We only update the busy time when we go idle. Otherwise, this
216 * calculation would require many more clock cycles.
217 */
218 if (ds->busy_count == 0) {
219 /* Calculate how long we were busy */
220 busy_time = ds->last_comp_time;
221 timevalsub(&busy_time, &ds->start_time);
222
223 /* Add our busy time to the total busy time. */
224 timevaladd(&ds->busy_time, &busy_time);
225 } else if (ds->busy_count < 0)
226 printf("devstat_end_transaction: HELP!! busy_count "
227 "for %s%d is < 0 (%d)!\n", ds->device_name,
228 ds->unit_number, ds->busy_count);
229}
230
231void
232devstat_end_transaction_bio(struct devstat *ds, struct bio *bp)
233{
234 devstat_trans_flags flg;
235
236 if (bp->bio_cmd == BIO_DELETE)
237 flg = DEVSTAT_FREE;
238 else if (bp->bio_cmd == BIO_READ)
239 flg = DEVSTAT_READ;
240 else
241 flg = DEVSTAT_WRITE;
242
243 devstat_end_transaction(ds, bp->bio_bcount - bp->bio_resid,
244 (bp->bio_flags & BIO_ORDERED) ?
245 DEVSTAT_TAG_ORDERED : DEVSTAT_TAG_SIMPLE, flg);
246}
247
248/*
249 * This is the sysctl handler for the devstat package. The data pushed out
250 * on the kern.devstat.all sysctl variable consists of the current devstat
251 * generation number, and then an array of devstat structures, one for each
252 * device in the system.
253 *
254 * I'm really not too fond of this method of doing things, but there really
255 * aren't that many alternatives. We must have some method of making sure
256 * that the generation number the user gets corresponds with the data the
257 * user gets. If the user makes a separate sysctl call to get the
258 * generation, and then a sysctl call to get the device statistics, the
259 * device list could have changed in that brief period of time. By
260 * supplying the generation number along with the statistics output, we can
261 * guarantee that the generation number and the statistics match up.
262 */
263static int
| 29 */
30
31#include <sys/param.h>
32#include <sys/kernel.h>
33#include <sys/systm.h>
34#include <sys/bio.h>
35#include <sys/sysctl.h>
36
37#include <sys/devicestat.h>
38
39static int devstat_num_devs;
40static long devstat_generation;
41static int devstat_version = DEVSTAT_VERSION;
42static int devstat_current_devnumber;
43
44static STAILQ_HEAD(devstatlist, devstat) device_statq;
45
46/*
47 * Take a malloced and zeroed devstat structure given to us, fill it in
48 * and add it to the queue of devices.
49 */
50void
51devstat_add_entry(struct devstat *ds, const char *dev_name,
52 int unit_number, u_int32_t block_size,
53 devstat_support_flags flags,
54 devstat_type_flags device_type,
55 devstat_priority priority)
56{
57 struct devstatlist *devstat_head;
58 struct devstat *ds_tmp;
59
60 if (ds == NULL)
61 return;
62
63 if (devstat_num_devs == 0)
64 STAILQ_INIT(&device_statq);
65
66 devstat_generation++;
67 devstat_num_devs++;
68
69 devstat_head = &device_statq;
70
71 /*
72 * Priority sort. Each driver passes in its priority when it adds
73 * its devstat entry. Drivers are sorted first by priority, and
74 * then by probe order.
75 *
76 * For the first device, we just insert it, since the priority
77 * doesn't really matter yet. Subsequent devices are inserted into
78 * the list using the order outlined above.
79 */
80 if (devstat_num_devs == 1)
81 STAILQ_INSERT_TAIL(devstat_head, ds, dev_links);
82 else {
83 for (ds_tmp = STAILQ_FIRST(devstat_head); ds_tmp != NULL;
84 ds_tmp = STAILQ_NEXT(ds_tmp, dev_links)) {
85 struct devstat *ds_next;
86
87 ds_next = STAILQ_NEXT(ds_tmp, dev_links);
88
89 /*
90 * If we find a break between higher and lower
91 * priority items, and if this item fits in the
92 * break, insert it. This also applies if the
93 * "lower priority item" is the end of the list.
94 */
95 if ((priority <= ds_tmp->priority)
96 && ((ds_next == NULL)
97 || (priority > ds_next->priority))) {
98 STAILQ_INSERT_AFTER(devstat_head, ds_tmp, ds,
99 dev_links);
100 break;
101 } else if (priority > ds_tmp->priority) {
102 /*
103 * If this is the case, we should be able
104 * to insert ourselves at the head of the
105 * list. If we can't, something is wrong.
106 */
107 if (ds_tmp == STAILQ_FIRST(devstat_head)) {
108 STAILQ_INSERT_HEAD(devstat_head,
109 ds, dev_links);
110 break;
111 } else {
112 STAILQ_INSERT_TAIL(devstat_head,
113 ds, dev_links);
114 printf("devstat_add_entry: HELP! "
115 "sorting problem detected "
116 "for %s%d\n", dev_name,
117 unit_number);
118 break;
119 }
120 }
121 }
122 }
123
124 ds->device_number = devstat_current_devnumber++;
125 ds->unit_number = unit_number;
126 strncpy(ds->device_name, dev_name, DEVSTAT_NAME_LEN);
127 ds->device_name[DEVSTAT_NAME_LEN - 1] = '\0';
128 ds->block_size = block_size;
129 ds->flags = flags;
130 ds->device_type = device_type;
131 ds->priority = priority;
132 getmicrotime(&ds->dev_creation_time);
133}
134
135/*
136 * Remove a devstat structure from the list of devices.
137 */
138void
139devstat_remove_entry(struct devstat *ds)
140{
141 struct devstatlist *devstat_head;
142
143 if (ds == NULL)
144 return;
145
146 devstat_generation++;
147 devstat_num_devs--;
148
149 devstat_head = &device_statq;
150
151 /* Remove this entry from the devstat queue */
152 STAILQ_REMOVE(devstat_head, ds, devstat, dev_links);
153}
154
155/*
156 * Record a transaction start.
157 */
158void
159devstat_start_transaction(struct devstat *ds)
160{
161 /* sanity check */
162 if (ds == NULL)
163 return;
164
165 /*
166 * We only want to set the start time when we are going from idle
167 * to busy. The start time is really the start of the latest busy
168 * period.
169 */
170 if (ds->busy_count == 0)
171 getmicrouptime(&ds->start_time);
172 ds->busy_count++;
173}
174
175/*
176 * Record the ending of a transaction, and incrment the various counters.
177 */
178void
179devstat_end_transaction(struct devstat *ds, u_int32_t bytes,
180 devstat_tag_type tag_type, devstat_trans_flags flags)
181{
182 struct timeval busy_time;
183
184 /* sanity check */
185 if (ds == NULL)
186 return;
187
188 getmicrouptime(&ds->last_comp_time);
189 ds->busy_count--;
190
191 /*
192 * There might be some transactions (DEVSTAT_NO_DATA) that don't
193 * transfer any data.
194 */
195 if (flags == DEVSTAT_READ) {
196 ds->bytes_read += bytes;
197 ds->num_reads++;
198 } else if (flags == DEVSTAT_WRITE) {
199 ds->bytes_written += bytes;
200 ds->num_writes++;
201 } else if (flags == DEVSTAT_FREE) {
202 ds->bytes_freed += bytes;
203 ds->num_frees++;
204 } else
205 ds->num_other++;
206
207 /*
208 * Keep a count of the various tag types sent.
209 */
210 if ((ds->flags & DEVSTAT_NO_ORDERED_TAGS) == 0 &&
211 tag_type != DEVSTAT_TAG_NONE)
212 ds->tag_types[tag_type]++;
213
214 /*
215 * We only update the busy time when we go idle. Otherwise, this
216 * calculation would require many more clock cycles.
217 */
218 if (ds->busy_count == 0) {
219 /* Calculate how long we were busy */
220 busy_time = ds->last_comp_time;
221 timevalsub(&busy_time, &ds->start_time);
222
223 /* Add our busy time to the total busy time. */
224 timevaladd(&ds->busy_time, &busy_time);
225 } else if (ds->busy_count < 0)
226 printf("devstat_end_transaction: HELP!! busy_count "
227 "for %s%d is < 0 (%d)!\n", ds->device_name,
228 ds->unit_number, ds->busy_count);
229}
230
231void
232devstat_end_transaction_bio(struct devstat *ds, struct bio *bp)
233{
234 devstat_trans_flags flg;
235
236 if (bp->bio_cmd == BIO_DELETE)
237 flg = DEVSTAT_FREE;
238 else if (bp->bio_cmd == BIO_READ)
239 flg = DEVSTAT_READ;
240 else
241 flg = DEVSTAT_WRITE;
242
243 devstat_end_transaction(ds, bp->bio_bcount - bp->bio_resid,
244 (bp->bio_flags & BIO_ORDERED) ?
245 DEVSTAT_TAG_ORDERED : DEVSTAT_TAG_SIMPLE, flg);
246}
247
248/*
249 * This is the sysctl handler for the devstat package. The data pushed out
250 * on the kern.devstat.all sysctl variable consists of the current devstat
251 * generation number, and then an array of devstat structures, one for each
252 * device in the system.
253 *
254 * I'm really not too fond of this method of doing things, but there really
255 * aren't that many alternatives. We must have some method of making sure
256 * that the generation number the user gets corresponds with the data the
257 * user gets. If the user makes a separate sysctl call to get the
258 * generation, and then a sysctl call to get the device statistics, the
259 * device list could have changed in that brief period of time. By
260 * supplying the generation number along with the statistics output, we can
261 * guarantee that the generation number and the statistics match up.
262 */
263static int
|
264sysctl_devstat (SYSCTL_HANDLER_ARGS)
| 264sysctl_devstat(SYSCTL_HANDLER_ARGS)
|
265{
266 int error, i;
267 struct devstat *nds;
268 struct devstatlist *devstat_head;
269
270 if (devstat_num_devs == 0)
271 return(EINVAL);
272
273 error = 0;
274 devstat_head = &device_statq;
275
276 /*
277 * First push out the generation number.
278 */
279 error = SYSCTL_OUT(req, &devstat_generation, sizeof(long));
280
281 /*
282 * Now push out all the devices.
283 */
284 for (i = 0, nds = STAILQ_FIRST(devstat_head);
285 (nds != NULL) && (i < devstat_num_devs) && (error == 0);
286 nds = STAILQ_NEXT(nds, dev_links), i++)
287 error = SYSCTL_OUT(req, nds, sizeof(struct devstat));
288
289 return(error);
290}
291
292/*
293 * Sysctl entries for devstat. The first one is a node that all the rest
294 * hang off of.
295 */
296SYSCTL_NODE(_kern, OID_AUTO, devstat, CTLFLAG_RD, 0, "Device Statistics");
297
298SYSCTL_PROC(_kern_devstat, OID_AUTO, all, CTLFLAG_RD|CTLTYPE_OPAQUE,
299 0, 0, sysctl_devstat, "S,devstat", "All devices in the devstat list");
300/*
301 * Export the number of devices in the system so that userland utilities
302 * can determine how much memory to allocate to hold all the devices.
303 */
304SYSCTL_INT(_kern_devstat, OID_AUTO, numdevs, CTLFLAG_RD,
305 &devstat_num_devs, 0, "Number of devices in the devstat list");
306SYSCTL_LONG(_kern_devstat, OID_AUTO, generation, CTLFLAG_RD,
307 &devstat_generation, "Devstat list generation");
308SYSCTL_INT(_kern_devstat, OID_AUTO, version, CTLFLAG_RD,
309 &devstat_version, 0, "Devstat list version number");
| 265{
266 int error, i;
267 struct devstat *nds;
268 struct devstatlist *devstat_head;
269
270 if (devstat_num_devs == 0)
271 return(EINVAL);
272
273 error = 0;
274 devstat_head = &device_statq;
275
276 /*
277 * First push out the generation number.
278 */
279 error = SYSCTL_OUT(req, &devstat_generation, sizeof(long));
280
281 /*
282 * Now push out all the devices.
283 */
284 for (i = 0, nds = STAILQ_FIRST(devstat_head);
285 (nds != NULL) && (i < devstat_num_devs) && (error == 0);
286 nds = STAILQ_NEXT(nds, dev_links), i++)
287 error = SYSCTL_OUT(req, nds, sizeof(struct devstat));
288
289 return(error);
290}
291
292/*
293 * Sysctl entries for devstat. The first one is a node that all the rest
294 * hang off of.
295 */
296SYSCTL_NODE(_kern, OID_AUTO, devstat, CTLFLAG_RD, 0, "Device Statistics");
297
298SYSCTL_PROC(_kern_devstat, OID_AUTO, all, CTLFLAG_RD|CTLTYPE_OPAQUE,
299 0, 0, sysctl_devstat, "S,devstat", "All devices in the devstat list");
300/*
301 * Export the number of devices in the system so that userland utilities
302 * can determine how much memory to allocate to hold all the devices.
303 */
304SYSCTL_INT(_kern_devstat, OID_AUTO, numdevs, CTLFLAG_RD,
305 &devstat_num_devs, 0, "Number of devices in the devstat list");
306SYSCTL_LONG(_kern_devstat, OID_AUTO, generation, CTLFLAG_RD,
307 &devstat_generation, "Devstat list generation");
308SYSCTL_INT(_kern_devstat, OID_AUTO, version, CTLFLAG_RD,
309 &devstat_version, 0, "Devstat list version number");
|