kernel - Fix slice open mask tracking
[dragonfly.git] / sys / kern / subr_disk.c
CommitLineData
984263bc 1/*
149e86b9 2 * Copyright (c) 2003,2004,2009 The DragonFly Project. All rights reserved.
d7d5e114 3 *
8c10bfcf
MD
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@backplane.com>
81b331be 6 * and Alex Hornung <ahornung@gmail.com>
d7d5e114 7 *
8c10bfcf
MD
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
d7d5e114 11 *
8c10bfcf
MD
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in
16 * the documentation and/or other materials provided with the
17 * distribution.
18 * 3. Neither the name of The DragonFly Project nor the names of its
19 * contributors may be used to endorse or promote products derived
20 * from this software without specific, prior written permission.
d7d5e114 21 *
8c10bfcf
MD
22 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
24 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
25 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
26 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
27 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
28 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
29 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
30 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
31 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
32 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * SUCH DAMAGE.
d7d5e114 34 *
984263bc
MD
35 * ----------------------------------------------------------------------------
36 * "THE BEER-WARE LICENSE" (Revision 42):
37 * <phk@FreeBSD.ORG> wrote this file. As long as you retain this notice you
38 * can do whatever you want with this stuff. If we meet some day, and you think
39 * this stuff is worth it, you can buy me a beer in return. Poul-Henning Kamp
40 * ----------------------------------------------------------------------------
41 *
7a9e53ad
MD
42 * Copyright (c) 1982, 1986, 1988, 1993
43 * The Regents of the University of California. All rights reserved.
44 * (c) UNIX System Laboratories, Inc.
45 * All or some portions of this file are derived from material licensed
46 * to the University of California by American Telephone and Telegraph
47 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
48 * the permission of UNIX System Laboratories, Inc.
49 *
50 * Redistribution and use in source and binary forms, with or without
51 * modification, are permitted provided that the following conditions
52 * are met:
53 * 1. Redistributions of source code must retain the above copyright
54 * notice, this list of conditions and the following disclaimer.
55 * 2. Redistributions in binary form must reproduce the above copyright
56 * notice, this list of conditions and the following disclaimer in the
57 * documentation and/or other materials provided with the distribution.
58 * 3. All advertising materials mentioning features or use of this software
59 * must display the following acknowledgement:
60 * This product includes software developed by the University of
61 * California, Berkeley and its contributors.
62 * 4. Neither the name of the University nor the names of its contributors
63 * may be used to endorse or promote products derived from this software
64 * without specific prior written permission.
65 *
66 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
67 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
68 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
69 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
70 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
71 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
72 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
73 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
74 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
75 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
76 * SUCH DAMAGE.
984263bc 77 *
7a9e53ad
MD
78 * @(#)ufs_disksubr.c 8.5 (Berkeley) 1/21/94
79 * $FreeBSD: src/sys/kern/subr_disk.c,v 1.20.2.6 2001/10/05 07:14:57 peter Exp $
80 * $FreeBSD: src/sys/ufs/ufs/ufs_disksubr.c,v 1.44.2.3 2001/03/05 05:42:19 obrien Exp $
3641b7ca 81 * $DragonFly: src/sys/kern/subr_disk.c,v 1.40 2008/06/05 18:06:32 swildner Exp $
984263bc
MD
82 */
83
84#include <sys/param.h>
85#include <sys/systm.h>
86#include <sys/kernel.h>
7a9e53ad 87#include <sys/proc.h>
984263bc
MD
88#include <sys/sysctl.h>
89#include <sys/buf.h>
90#include <sys/conf.h>
cd29885a
MD
91#include <sys/disklabel.h>
92#include <sys/disklabel32.h>
93#include <sys/disklabel64.h>
7a9e53ad 94#include <sys/diskslice.h>
cd29885a 95#include <sys/diskmbr.h>
984263bc 96#include <sys/disk.h>
b24cd69c 97#include <sys/kerneldump.h>
984263bc
MD
98#include <sys/malloc.h>
99#include <sys/sysctl.h>
100#include <machine/md_var.h>
101#include <sys/ctype.h>
7a9e53ad
MD
102#include <sys/syslog.h>
103#include <sys/device.h>
335dda38 104#include <sys/msgport.h>
2c1e28dd 105#include <sys/devfs.h>
be755ff9 106#include <sys/thread.h>
8c72e3d5 107#include <sys/dsched.h>
cd29885a
MD
108#include <sys/queue.h>
109#include <sys/lock.h>
f5d8307c 110#include <sys/udev.h>
e33b65d9 111#include <sys/uuid.h>
984263bc 112
9f889dc4
MD
113#include <sys/buf2.h>
114#include <sys/mplock2.h>
115#include <sys/msgport2.h>
116#include <sys/thread2.h>
117
984263bc 118static MALLOC_DEFINE(M_DISK, "disk", "disk data");
8c05caab 119static int disk_debug_enable = 0;
984263bc 120
cd29885a
MD
121static void disk_msg_autofree_reply(lwkt_port_t, lwkt_msg_t);
122static void disk_msg_core(void *);
aec8eea4
MD
123static int disk_probe_slice(struct disk *dp, cdev_t dev, int slice, int reprobe);
124static void disk_probe(struct disk *dp, int reprobe);
125static void _setdiskinfo(struct disk *disk, struct disk_info *info);
30e5862e 126static void bioqwritereorder(struct bio_queue_head *bioq);
fbbbca99 127static void disk_cleanserial(char *serno);
c6b20585 128static int disk_debug(int, char *, ...) __printflike(2, 3);
4064300e
AH
129static cdev_t _disk_create_named(const char *name, int unit, struct disk *dp,
130 struct dev_ops *raw_ops, int clone);
cd29885a 131
984263bc 132static d_open_t diskopen;
d7d5e114 133static d_close_t diskclose;
984263bc 134static d_ioctl_t diskioctl;
fef8985e 135static d_strategy_t diskstrategy;
984263bc 136static d_psize_t diskpsize;
fef8985e 137static d_dump_t diskdump;
984263bc
MD
138
139static LIST_HEAD(, disk) disklist = LIST_HEAD_INITIALIZER(&disklist);
be755ff9 140static struct lwkt_token disklist_token;
984263bc 141
fef8985e 142static struct dev_ops disk_ops = {
8d51c2a2 143 { "disk", 0, D_DISK | D_MPSAFE | D_TRACKCLOSE },
fef8985e
MD
144 .d_open = diskopen,
145 .d_close = diskclose,
146 .d_read = physread,
147 .d_write = physwrite,
148 .d_ioctl = diskioctl,
149 .d_strategy = diskstrategy,
150 .d_dump = diskdump,
151 .d_psize = diskpsize,
fef8985e
MD
152};
153
cd29885a
MD
154static struct objcache *disk_msg_cache;
155
156struct objcache_malloc_args disk_msg_malloc_args = {
157 sizeof(struct disk_msg), M_DISK };
158
159static struct lwkt_port disk_dispose_port;
160static struct lwkt_port disk_msg_port;
161
8c05caab
AH
162static int
163disk_debug(int level, char *fmt, ...)
164{
165 __va_list ap;
166
167 __va_start(ap, fmt);
168 if (level <= disk_debug_enable)
169 kvprintf(fmt, ap);
170 __va_end(ap);
171
172 return 0;
173}
cd29885a
MD
174
175static int
aec8eea4 176disk_probe_slice(struct disk *dp, cdev_t dev, int slice, int reprobe)
cd29885a
MD
177{
178 struct disk_info *info = &dp->d_info;
179 struct diskslice *sp = &dp->d_slice->dss_slices[slice];
180 disklabel_ops_t ops;
181 struct partinfo part;
182 const char *msg;
e33b65d9 183 char uuid_buf[128];
cd29885a 184 cdev_t ndev;
440a040b 185 int sno;
539f339e 186 u_int i;
cd29885a 187
8c05caab
AH
188 disk_debug(2,
189 "disk_probe_slice (begin): %s (%s)\n",
190 dev->si_name, dp->d_cdev->si_name);
191
440a040b
MD
192 sno = slice ? slice - 1 : 0;
193
cd29885a
MD
194 ops = &disklabel32_ops;
195 msg = ops->op_readdisklabel(dev, sp, &sp->ds_label, info);
196 if (msg && !strcmp(msg, "no disk label")) {
cd29885a
MD
197 ops = &disklabel64_ops;
198 msg = ops->op_readdisklabel(dev, sp, &sp->ds_label, info);
199 }
cd29885a
MD
200 if (msg == NULL) {
201 if (slice != WHOLE_DISK_SLICE)
202 ops->op_adjust_label_reserved(dp->d_slice, slice, sp);
203 else
204 sp->ds_reserved = 0;
205
206 sp->ds_ops = ops;
cd29885a
MD
207 for (i = 0; i < ops->op_getnumparts(sp->ds_label); i++) {
208 ops->op_loadpartinfo(sp->ds_label, i, &part);
cd29885a 209 if (part.fstype) {
aec8eea4 210 if (reprobe &&
149e86b9 211 (ndev = devfs_find_device_by_name("%s%c",
440a040b 212 dev->si_name, 'a' + i))
149e86b9
MD
213 ) {
214 /*
215 * Device already exists and
216 * is still valid.
217 */
aec8eea4 218 ndev->si_flags |= SI_REPROBE_TEST;
e33b65d9
AH
219
220 /*
221 * Destroy old UUID alias
222 */
223 destroy_dev_alias(ndev, "part-by-uuid/*");
224
225 /* Create UUID alias */
226 if (!kuuid_is_nil(&part.storage_uuid)) {
227 snprintf_uuid(uuid_buf,
228 sizeof(uuid_buf),
229 &part.storage_uuid);
230 make_dev_alias(ndev,
231 "part-by-uuid/%s",
232 uuid_buf);
233 }
aec8eea4 234 } else {
8f960aa9 235 ndev = make_dev_covering(&disk_ops, dp->d_rawdev->si_ops,
149e86b9
MD
236 dkmakeminor(dkunit(dp->d_cdev),
237 slice, i),
aec8eea4 238 UID_ROOT, GID_OPERATOR, 0640,
440a040b 239 "%s%c", dev->si_name, 'a'+ i);
aec8eea4 240 ndev->si_disk = dp;
f5d8307c
AH
241 udev_dict_set_cstr(ndev, "subsystem", "disk");
242 /* Inherit parent's disk type */
243 if (dp->d_disktype) {
3af01d56 244 udev_dict_set_cstr(ndev, "disk-type",
f5d8307c
AH
245 __DECONST(char *, dp->d_disktype));
246 }
3af01d56
AH
247
248 /* Create serno alias */
55230951 249 if (dp->d_info.d_serialno) {
149e86b9
MD
250 make_dev_alias(ndev,
251 "serno/%s.s%d%c",
252 dp->d_info.d_serialno,
440a040b 253 sno, 'a' + i);
55230951 254 }
3af01d56 255
e33b65d9
AH
256 /* Create UUID alias */
257 if (!kuuid_is_nil(&part.storage_uuid)) {
258 snprintf_uuid(uuid_buf,
259 sizeof(uuid_buf),
260 &part.storage_uuid);
261 make_dev_alias(ndev,
262 "part-by-uuid/%s",
263 uuid_buf);
264 }
aec8eea4
MD
265 ndev->si_flags |= SI_REPROBE_TEST;
266 }
cd29885a
MD
267 }
268 }
269 } else if (info->d_dsflags & DSO_COMPATLABEL) {
270 msg = NULL;
271 if (sp->ds_size >= 0x100000000ULL)
272 ops = &disklabel64_ops;
273 else
274 ops = &disklabel32_ops;
275 sp->ds_label = ops->op_clone_label(info, sp);
276 } else {
dc6af901
TN
277 if (sp->ds_type == DOSPTYP_386BSD || /* XXX */
278 sp->ds_type == DOSPTYP_NETBSD ||
279 sp->ds_type == DOSPTYP_OPENBSD) {
cd29885a
MD
280 log(LOG_WARNING, "%s: cannot find label (%s)\n",
281 dev->si_name, msg);
149e86b9 282 }
cd29885a
MD
283 }
284
285 if (msg == NULL) {
286 sp->ds_wlabel = FALSE;
287 }
288
289 return (msg ? EINVAL : 0);
290}
291
666ec833
MD
292/*
293 * This routine is only called for newly minted drives or to reprobe
294 * a drive with no open slices. disk_probe_slice() is called directly
295 * when reprobing partition changes within slices.
296 */
cd29885a 297static void
aec8eea4 298disk_probe(struct disk *dp, int reprobe)
cd29885a
MD
299{
300 struct disk_info *info = &dp->d_info;
301 cdev_t dev = dp->d_cdev;
302 cdev_t ndev;
0831f2ab 303 int error, i, sno;
666ec833 304 struct diskslices *osp;
0831f2ab 305 struct diskslice *sp;
e33b65d9 306 char uuid_buf[128];
cd29885a 307
cd29885a 308 KKASSERT (info->d_media_blksize != 0);
cd29885a 309
666ec833 310 osp = dp->d_slice;
cd29885a 311 dp->d_slice = dsmakeslicestruct(BASE_SLICE, info);
666ec833 312 disk_debug(1, "disk_probe (begin): %s\n", dp->d_cdev->si_name);
cd29885a
MD
313
314 error = mbrinit(dev, info, &(dp->d_slice));
666ec833
MD
315 if (error) {
316 dsgone(&osp);
cd29885a 317 return;
666ec833 318 }
aec8eea4 319
0831f2ab
MD
320 for (i = 0; i < dp->d_slice->dss_nslices; i++) {
321 /*
322 * Ignore the whole-disk slice, it has already been created.
323 */
324 if (i == WHOLE_DISK_SLICE)
325 continue;
3af01d56 326
81b331be 327#if 1
3af01d56
AH
328 /*
329 * Ignore the compatibility slice s0 if it's a device mapper
330 * volume.
331 */
332 if ((i == COMPATIBILITY_SLICE) &&
333 (info->d_dsflags & DSO_DEVICEMAPPER))
334 continue;
335#endif
336
0831f2ab
MD
337 sp = &dp->d_slice->dss_slices[i];
338
339 /*
340 * Handle s0. s0 is a compatibility slice if there are no
341 * other slices and it has not otherwise been set up, else
342 * we ignore it.
343 */
344 if (i == COMPATIBILITY_SLICE) {
345 sno = 0;
346 if (sp->ds_type == 0 &&
347 dp->d_slice->dss_nslices == BASE_SLICE) {
348 sp->ds_size = info->d_media_blocks;
349 sp->ds_reserved = 0;
aec8eea4 350 }
0831f2ab
MD
351 } else {
352 sno = i - 1;
353 sp->ds_reserved = 0;
cd29885a 354 }
0831f2ab
MD
355
356 /*
357 * Ignore 0-length slices
358 */
359 if (sp->ds_size == 0)
360 continue;
361
362 if (reprobe &&
363 (ndev = devfs_find_device_by_name("%ss%d",
364 dev->si_name, sno))) {
365 /*
366 * Device already exists and is still valid
367 */
368 ndev->si_flags |= SI_REPROBE_TEST;
e33b65d9
AH
369
370 /*
371 * Destroy old UUID alias
372 */
373 destroy_dev_alias(ndev, "slice-by-uuid/*");
374
375 /* Create UUID alias */
376 if (!kuuid_is_nil(&sp->ds_stor_uuid)) {
377 snprintf_uuid(uuid_buf, sizeof(uuid_buf),
378 &sp->ds_stor_uuid);
379 make_dev_alias(ndev, "slice-by-uuid/%s",
380 uuid_buf);
381 }
0831f2ab
MD
382 } else {
383 /*
384 * Else create new device
385 */
8f960aa9 386 ndev = make_dev_covering(&disk_ops, dp->d_rawdev->si_ops,
aec8eea4
MD
387 dkmakewholeslice(dkunit(dev), i),
388 UID_ROOT, GID_OPERATOR, 0640,
3af01d56
AH
389 (info->d_dsflags & DSO_DEVICEMAPPER)?
390 "%s.s%d" : "%ss%d", dev->si_name, sno);
f5d8307c
AH
391 udev_dict_set_cstr(ndev, "subsystem", "disk");
392 /* Inherit parent's disk type */
393 if (dp->d_disktype) {
3af01d56 394 udev_dict_set_cstr(ndev, "disk-type",
f5d8307c
AH
395 __DECONST(char *, dp->d_disktype));
396 }
3af01d56
AH
397
398 /* Create serno alias */
0831f2ab
MD
399 if (dp->d_info.d_serialno) {
400 make_dev_alias(ndev, "serno/%s.s%d",
401 dp->d_info.d_serialno, sno);
cd29885a 402 }
3af01d56 403
e33b65d9
AH
404 /* Create UUID alias */
405 if (!kuuid_is_nil(&sp->ds_stor_uuid)) {
406 snprintf_uuid(uuid_buf, sizeof(uuid_buf),
407 &sp->ds_stor_uuid);
408 make_dev_alias(ndev, "slice-by-uuid/%s",
409 uuid_buf);
410 }
411
0831f2ab
MD
412 ndev->si_disk = dp;
413 ndev->si_flags |= SI_REPROBE_TEST;
414 }
415 sp->ds_dev = ndev;
149e86b9 416
1cb0bdb6
MD
417 /*
418 * Probe appropriate slices for a disklabel
419 *
420 * XXX slice type 1 used by our gpt probe code.
d7d5e114 421 * XXX slice type 0 used by mbr compat slice.
1cb0bdb6 422 */
dc6af901
TN
423 if (sp->ds_type == DOSPTYP_386BSD ||
424 sp->ds_type == DOSPTYP_NETBSD ||
425 sp->ds_type == DOSPTYP_OPENBSD ||
426 sp->ds_type == 0 ||
427 sp->ds_type == 1) {
0831f2ab
MD
428 if (dp->d_slice->dss_first_bsd_slice == 0)
429 dp->d_slice->dss_first_bsd_slice = i;
430 disk_probe_slice(dp, ndev, i, reprobe);
cd29885a
MD
431 }
432 }
666ec833
MD
433 dsgone(&osp);
434 disk_debug(1, "disk_probe (end): %s\n", dp->d_cdev->si_name);
cd29885a
MD
435}
436
437
438static void
439disk_msg_core(void *arg)
440{
cd29885a
MD
441 struct disk *dp;
442 struct diskslice *sp;
149e86b9
MD
443 disk_msg_t msg;
444 int run;
cd29885a 445
c9e9fb21 446 lwkt_gettoken(&disklist_token);
cd29885a 447 lwkt_initport_thread(&disk_msg_port, curthread);
c9e9fb21
MD
448 wakeup(curthread); /* synchronous startup */
449 lwkt_reltoken(&disklist_token);
450
451 get_mplock(); /* not mpsafe yet? */
149e86b9 452 run = 1;
cd29885a 453
149e86b9
MD
454 while (run) {
455 msg = (disk_msg_t)lwkt_waitport(&disk_msg_port, 0);
cd29885a 456
149e86b9
MD
457 switch (msg->hdr.u.ms_result) {
458 case DISK_DISK_PROBE:
cd29885a 459 dp = (struct disk *)msg->load;
8c05caab
AH
460 disk_debug(1,
461 "DISK_DISK_PROBE: %s\n",
462 dp->d_cdev->si_name);
aec8eea4 463 disk_probe(dp, 0);
cd29885a 464 break;
cd29885a
MD
465 case DISK_DISK_DESTROY:
466 dp = (struct disk *)msg->load;
8c05caab
AH
467 disk_debug(1,
468 "DISK_DISK_DESTROY: %s\n",
469 dp->d_cdev->si_name);
cd29885a 470 devfs_destroy_subnames(dp->d_cdev->si_name);
4064300e
AH
471 destroy_dev(dp->d_cdev);
472 destroy_only_dev(dp->d_rawdev);
3b998fa9 473 lwkt_gettoken(&disklist_token);
aec8eea4 474 LIST_REMOVE(dp, d_list);
3b998fa9 475 lwkt_reltoken(&disklist_token);
55230951
MD
476 if (dp->d_info.d_serialno) {
477 kfree(dp->d_info.d_serialno, M_TEMP);
478 dp->d_info.d_serialno = NULL;
479 }
cd29885a 480 break;
aec8eea4
MD
481 case DISK_UNPROBE:
482 dp = (struct disk *)msg->load;
8c05caab
AH
483 disk_debug(1,
484 "DISK_DISK_UNPROBE: %s\n",
485 dp->d_cdev->si_name);
aec8eea4
MD
486 devfs_destroy_subnames(dp->d_cdev->si_name);
487 break;
cd29885a
MD
488 case DISK_SLICE_REPROBE:
489 dp = (struct disk *)msg->load;
490 sp = (struct diskslice *)msg->load2;
149e86b9
MD
491 devfs_clr_subnames_flag(sp->ds_dev->si_name,
492 SI_REPROBE_TEST);
8c05caab 493 disk_debug(1,
aec8eea4
MD
494 "DISK_SLICE_REPROBE: %s\n",
495 sp->ds_dev->si_name);
149e86b9
MD
496 disk_probe_slice(dp, sp->ds_dev,
497 dkslice(sp->ds_dev), 1);
498 devfs_destroy_subnames_without_flag(
499 sp->ds_dev->si_name, SI_REPROBE_TEST);
cd29885a 500 break;
cd29885a
MD
501 case DISK_DISK_REPROBE:
502 dp = (struct disk *)msg->load;
aec8eea4 503 devfs_clr_subnames_flag(dp->d_cdev->si_name, SI_REPROBE_TEST);
8c05caab 504 disk_debug(1,
aec8eea4
MD
505 "DISK_DISK_REPROBE: %s\n",
506 dp->d_cdev->si_name);
507 disk_probe(dp, 1);
149e86b9
MD
508 devfs_destroy_subnames_without_flag(
509 dp->d_cdev->si_name, SI_REPROBE_TEST);
cd29885a 510 break;
cd29885a 511 case DISK_SYNC:
8c05caab 512 disk_debug(1, "DISK_SYNC\n");
cd29885a 513 break;
149e86b9
MD
514 default:
515 devfs_debug(DEVFS_DEBUG_WARNING,
516 "disk_msg_core: unknown message "
517 "received at core\n");
518 break;
519 }
a9177e09 520 lwkt_replymsg(&msg->hdr, 0);
149e86b9 521 }
cd29885a
MD
522 lwkt_exit();
523}
524
525
149e86b9
MD
526/*
527 * Acts as a message drain. Any message that is replied to here gets
528 * destroyed and the memory freed.
529 */
cd29885a
MD
530static void
531disk_msg_autofree_reply(lwkt_port_t port, lwkt_msg_t msg)
532{
149e86b9 533 objcache_put(disk_msg_cache, msg);
cd29885a
MD
534}
535
536
537void
538disk_msg_send(uint32_t cmd, void *load, void *load2)
539{
149e86b9 540 disk_msg_t disk_msg;
cd29885a
MD
541 lwkt_port_t port = &disk_msg_port;
542
149e86b9 543 disk_msg = objcache_get(disk_msg_cache, M_WAITOK);
cd29885a 544
149e86b9 545 lwkt_initmsg(&disk_msg->hdr, &disk_dispose_port, 0);
cd29885a
MD
546
547 disk_msg->hdr.u.ms_result = cmd;
548 disk_msg->load = load;
549 disk_msg->load2 = load2;
550 KKASSERT(port);
a9177e09 551 lwkt_sendmsg(port, &disk_msg->hdr);
cd29885a
MD
552}
553
aec8eea4
MD
554void
555disk_msg_send_sync(uint32_t cmd, void *load, void *load2)
556{
557 struct lwkt_port rep_port;
a9177e09
MD
558 disk_msg_t disk_msg;
559 lwkt_port_t port;
aec8eea4 560
a9177e09
MD
561 disk_msg = objcache_get(disk_msg_cache, M_WAITOK);
562 port = &disk_msg_port;
563
564 /* XXX could probably use curthread's built-in msgport */
aec8eea4
MD
565 lwkt_initport_thread(&rep_port, curthread);
566 lwkt_initmsg(&disk_msg->hdr, &rep_port, 0);
567
568 disk_msg->hdr.u.ms_result = cmd;
569 disk_msg->load = load;
570 disk_msg->load2 = load2;
571
a9177e09
MD
572 lwkt_sendmsg(port, &disk_msg->hdr);
573 lwkt_waitmsg(&disk_msg->hdr, 0);
574 objcache_put(disk_msg_cache, disk_msg);
aec8eea4
MD
575}
576
335dda38 577/*
fef8985e
MD
578 * Create a raw device for the dev_ops template (which is returned). Also
579 * create a slice and unit managed disk and overload the user visible
580 * device space with it.
e4c9c0c8 581 *
fef8985e
MD
582 * NOTE: The returned raw device is NOT a slice and unit managed device.
583 * It is an actual raw device representing the raw disk as specified by
584 * the passed dev_ops. The disk layer not only returns such a raw device,
585 * it also uses it internally when passing (modified) commands through.
335dda38 586 */
b13267a5 587cdev_t
a688b15c 588disk_create(int unit, struct disk *dp, struct dev_ops *raw_ops)
984263bc 589{
4064300e
AH
590 return _disk_create_named(NULL, unit, dp, raw_ops, 0);
591}
592
593cdev_t
594disk_create_clone(int unit, struct disk *dp, struct dev_ops *raw_ops)
595{
596 return _disk_create_named(NULL, unit, dp, raw_ops, 1);
c6ef65ea
AH
597}
598
599cdev_t
600disk_create_named(const char *name, int unit, struct disk *dp, struct dev_ops *raw_ops)
601{
4064300e
AH
602 return _disk_create_named(name, unit, dp, raw_ops, 0);
603}
604
605cdev_t
606disk_create_named_clone(const char *name, int unit, struct disk *dp, struct dev_ops *raw_ops)
607{
608 return _disk_create_named(name, unit, dp, raw_ops, 1);
609}
610
611static cdev_t
612_disk_create_named(const char *name, int unit, struct disk *dp, struct dev_ops *raw_ops, int clone)
613{
b13267a5 614 cdev_t rawdev;
e4c9c0c8 615
c6ef65ea 616 disk_debug(1, "disk_create (begin): %s%d\n", name, unit);
8c05caab 617
a812bf57
AH
618 if (name) {
619 rawdev = make_only_dev(raw_ops, dkmakewholedisk(unit),
620 UID_ROOT, GID_OPERATOR, 0640, "%s", name);
621 } else {
622 rawdev = make_only_dev(raw_ops, dkmakewholedisk(unit),
623 UID_ROOT, GID_OPERATOR, 0640,
624 "%s%d", raw_ops->head.name, unit);
625 }
984263bc 626
cd29885a 627 bzero(dp, sizeof(*dp));
fef8985e 628
e4c9c0c8 629 dp->d_rawdev = rawdev;
fef8985e 630 dp->d_raw_ops = raw_ops;
cd29885a 631 dp->d_dev_ops = &disk_ops;
a812bf57
AH
632
633 if (name) {
4064300e
AH
634 if (clone) {
635 dp->d_cdev = make_only_dev_covering(&disk_ops, dp->d_rawdev->si_ops,
636 dkmakewholedisk(unit), UID_ROOT, GID_OPERATOR, 0640,
637 "%s", name);
638 } else {
639 dp->d_cdev = make_dev_covering(&disk_ops, dp->d_rawdev->si_ops,
640 dkmakewholedisk(unit), UID_ROOT, GID_OPERATOR, 0640,
641 "%s", name);
642 }
a812bf57 643 } else {
4064300e
AH
644 if (clone) {
645 dp->d_cdev = make_only_dev_covering(&disk_ops, dp->d_rawdev->si_ops,
646 dkmakewholedisk(unit),
647 UID_ROOT, GID_OPERATOR, 0640,
648 "%s%d", raw_ops->head.name, unit);
649 } else {
650 dp->d_cdev = make_dev_covering(&disk_ops, dp->d_rawdev->si_ops,
651 dkmakewholedisk(unit),
652 UID_ROOT, GID_OPERATOR, 0640,
653 "%s%d", raw_ops->head.name, unit);
654 }
a812bf57
AH
655 }
656
f5d8307c 657 udev_dict_set_cstr(dp->d_cdev, "subsystem", "disk");
cd29885a
MD
658 dp->d_cdev->si_disk = dp;
659
a812bf57
AH
660 if (name)
661 dsched_disk_create_callback(dp, name, unit);
662 else
663 dsched_disk_create_callback(dp, raw_ops->head.name, unit);
e02e815e 664
3b998fa9 665 lwkt_gettoken(&disklist_token);
984263bc 666 LIST_INSERT_HEAD(&disklist, dp, d_list);
3b998fa9 667 lwkt_reltoken(&disklist_token);
8c05caab 668
a812bf57
AH
669 disk_debug(1, "disk_create (end): %s%d\n",
670 (name != NULL)?(name):(raw_ops->head.name), unit);
8c05caab 671
e4c9c0c8 672 return (dp->d_rawdev);
984263bc
MD
673}
674
f5d8307c
AH
675int
676disk_setdisktype(struct disk *disk, const char *type)
677{
678 KKASSERT(disk != NULL);
679
680 disk->d_disktype = type;
681 return udev_dict_set_cstr(disk->d_cdev, "disk-type", __DECONST(char *, type));
682}
683
8d51c2a2
AH
684int
685disk_getopencount(struct disk *disk)
686{
687 return disk->d_opencount;
688}
689
aec8eea4
MD
690static void
691_setdiskinfo(struct disk *disk, struct disk_info *info)
a688b15c 692{
55230951
MD
693 char *oldserialno;
694
55230951 695 oldserialno = disk->d_info.d_serialno;
a688b15c
MD
696 bcopy(info, &disk->d_info, sizeof(disk->d_info));
697 info = &disk->d_info;
698
8c05caab
AH
699 disk_debug(1,
700 "_setdiskinfo: %s\n",
701 disk->d_cdev->si_name);
702
55230951
MD
703 /*
704 * The serial number is duplicated so the caller can throw
705 * their copy away.
706 */
5300ce8f
SW
707 if (info->d_serialno && info->d_serialno[0] &&
708 (info->d_serialno[0] != ' ' || strlen(info->d_serialno) > 1)) {
55230951 709 info->d_serialno = kstrdup(info->d_serialno, M_TEMP);
fbbbca99 710 disk_cleanserial(info->d_serialno);
55230951
MD
711 if (disk->d_cdev) {
712 make_dev_alias(disk->d_cdev, "serno/%s",
713 info->d_serialno);
714 }
715 } else {
716 info->d_serialno = NULL;
717 }
718 if (oldserialno)
719 kfree(oldserialno, M_TEMP);
720
279e9fd5
AH
721 dsched_disk_update_callback(disk, info);
722
55230951
MD
723 /*
724 * The caller may set d_media_size or d_media_blocks and we
725 * calculate the other.
726 */
b9f2d452 727 KKASSERT(info->d_media_size == 0 || info->d_media_blocks == 0);
a688b15c 728 if (info->d_media_size == 0 && info->d_media_blocks) {
d7d5e114 729 info->d_media_size = (u_int64_t)info->d_media_blocks *
a688b15c 730 info->d_media_blksize;
d7d5e114 731 } else if (info->d_media_size && info->d_media_blocks == 0 &&
a688b15c 732 info->d_media_blksize) {
d7d5e114 733 info->d_media_blocks = info->d_media_size /
a688b15c
MD
734 info->d_media_blksize;
735 }
5d6c6885
MD
736
737 /*
738 * The si_* fields for rawdev are not set until after the
739 * disk_create() call, so someone using the cooked version
740 * of the raw device (i.e. da0s0) will not get the right
741 * si_iosize_max unless we fix it up here.
742 */
743 if (disk->d_cdev && disk->d_rawdev &&
744 disk->d_cdev->si_iosize_max == 0) {
745 disk->d_cdev->si_iosize_max = disk->d_rawdev->si_iosize_max;
746 disk->d_cdev->si_bsize_phys = disk->d_rawdev->si_bsize_phys;
747 disk->d_cdev->si_bsize_best = disk->d_rawdev->si_bsize_best;
748 }
d5cc18b0
AHJ
749
750 /* Add the serial number to the udev_dictionary */
751 if (info->d_serialno)
752 udev_dict_set_cstr(disk->d_cdev, "serno", info->d_serialno);
aec8eea4 753}
cd29885a 754
aec8eea4
MD
755/*
756 * Disk drivers must call this routine when media parameters are available
757 * or have changed.
758 */
759void
760disk_setdiskinfo(struct disk *disk, struct disk_info *info)
761{
762 _setdiskinfo(disk, info);
cd29885a 763 disk_msg_send(DISK_DISK_PROBE, disk, NULL);
8c05caab
AH
764 disk_debug(1,
765 "disk_setdiskinfo: sent probe for %s\n",
766 disk->d_cdev->si_name);
a688b15c
MD
767}
768
aec8eea4
MD
769void
770disk_setdiskinfo_sync(struct disk *disk, struct disk_info *info)
771{
772 _setdiskinfo(disk, info);
aec8eea4 773 disk_msg_send_sync(DISK_DISK_PROBE, disk, NULL);
8c05caab
AH
774 disk_debug(1,
775 "disk_setdiskinfo_sync: sent probe for %s\n",
776 disk->d_cdev->si_name);
aec8eea4
MD
777}
778
a688b15c 779/*
e4c9c0c8
MD
780 * This routine is called when an adapter detaches. The higher level
781 * managed disk device is destroyed while the lower level raw device is
782 * released.
783 */
335dda38
MD
784void
785disk_destroy(struct disk *disk)
786{
0160356d 787 dsched_disk_destroy_callback(disk);
aec8eea4 788 disk_msg_send_sync(DISK_DISK_DESTROY, disk, NULL);
cd29885a 789 return;
335dda38
MD
790}
791
984263bc 792int
b24cd69c 793disk_dumpcheck(cdev_t dev, u_int64_t *size, u_int64_t *blkno, u_int32_t *secsize)
984263bc 794{
a6c0f342
MD
795 struct partinfo pinfo;
796 int error;
984263bc 797
a6c0f342 798 bzero(&pinfo, sizeof(pinfo));
87baaf0c
MD
799 error = dev_dioctl(dev, DIOCGPART, (void *)&pinfo, 0,
800 proc0.p_ucred, NULL);
a6c0f342
MD
801 if (error)
802 return (error);
b24cd69c 803
a6c0f342 804 if (pinfo.media_blksize == 0)
984263bc 805 return (ENXIO);
b24cd69c
AH
806
807 if (blkno) /* XXX: make sure this reserved stuff is right */
808 *blkno = pinfo.reserved_blocks +
809 pinfo.media_offset / pinfo.media_blksize;
810 if (secsize)
811 *secsize = pinfo.media_blksize;
812 if (size)
813 *size = (pinfo.media_blocks - pinfo.reserved_blocks);
814
984263bc 815 return (0);
984263bc
MD
816}
817
b24cd69c
AH
818int
819disk_dumpconf(cdev_t dev, u_int onoff)
820{
821 struct dumperinfo di;
822 u_int64_t size, blkno;
823 u_int32_t secsize;
824 int error;
825
826 if (!onoff)
827 return set_dumper(NULL);
828
829 error = disk_dumpcheck(dev, &size, &blkno, &secsize);
830
831 if (error)
832 return ENXIO;
833
834 bzero(&di, sizeof(struct dumperinfo));
835 di.dumper = diskdump;
836 di.priv = dev;
837 di.blocksize = secsize;
838 di.mediaoffset = blkno * DEV_BSIZE;
839 di.mediasize = size * DEV_BSIZE;
840
841 return set_dumper(&di);
842}
843
aec8eea4
MD
844void
845disk_unprobe(struct disk *disk)
846{
847 if (disk == NULL)
848 return;
849
850 disk_msg_send_sync(DISK_UNPROBE, disk, NULL);
851}
852
d7d5e114 853void
984263bc
MD
854disk_invalidate (struct disk *disk)
855{
666ec833 856 dsgone(&disk->d_slice);
984263bc
MD
857}
858
984263bc
MD
859struct disk *
860disk_enumerate(struct disk *disk)
861{
be755ff9 862 struct disk *dp;
be755ff9 863
3b998fa9 864 lwkt_gettoken(&disklist_token);
984263bc 865 if (!disk)
be755ff9 866 dp = (LIST_FIRST(&disklist));
984263bc 867 else
be755ff9 868 dp = (LIST_NEXT(disk, d_list));
3b998fa9 869 lwkt_reltoken(&disklist_token);
be755ff9
AH
870
871 return dp;
984263bc
MD
872}
873
d7d5e114 874static
fbda7fa6 875int
984263bc
MD
876sysctl_disks(SYSCTL_HANDLER_ARGS)
877{
878 struct disk *disk;
879 int error, first;
880
881 disk = NULL;
882 first = 1;
883
884 while ((disk = disk_enumerate(disk))) {
885 if (!first) {
886 error = SYSCTL_OUT(req, " ", 1);
887 if (error)
888 return error;
889 } else {
890 first = 0;
891 }
95ce4036
HP
892 error = SYSCTL_OUT(req, disk->d_rawdev->si_name,
893 strlen(disk->d_rawdev->si_name));
984263bc
MD
894 if (error)
895 return error;
896 }
897 error = SYSCTL_OUT(req, "", 1);
898 return error;
899}
d7d5e114 900
3641b7ca 901SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD, NULL, 0,
984263bc
MD
902 sysctl_disks, "A", "names of available disks");
903
904/*
e4c9c0c8
MD
905 * Open a disk device or partition.
906 */
fbda7fa6
MD
907static
908int
fef8985e 909diskopen(struct dev_open_args *ap)
984263bc 910{
b13267a5 911 cdev_t dev = ap->a_head.a_dev;
984263bc
MD
912 struct disk *dp;
913 int error;
914
e4c9c0c8
MD
915 /*
916 * dp can't be NULL here XXX.
7ba1363d
MD
917 *
918 * d_slice will be NULL if setdiskinfo() has not been called yet.
919 * setdiskinfo() is typically called whether the disk is present
920 * or not (e.g. CD), but the base disk device is created first
921 * and there may be a race.
e4c9c0c8 922 */
e4c9c0c8 923 dp = dev->si_disk;
7ba1363d 924 if (dp == NULL || dp->d_slice == NULL)
984263bc 925 return (ENXIO);
fef8985e 926 error = 0;
984263bc 927
e4c9c0c8
MD
928 /*
929 * Deal with open races
930 */
9f889dc4 931 get_mplock();
984263bc
MD
932 while (dp->d_flags & DISKFLAG_LOCK) {
933 dp->d_flags |= DISKFLAG_WANTED;
377d4740 934 error = tsleep(dp, PCATCH, "diskopen", hz);
9f889dc4
MD
935 if (error) {
936 rel_mplock();
984263bc 937 return (error);
9f889dc4 938 }
984263bc
MD
939 }
940 dp->d_flags |= DISKFLAG_LOCK;
941
e4c9c0c8
MD
942 /*
943 * Open the underlying raw device.
944 */
984263bc 945 if (!dsisopen(dp->d_slice)) {
e4c9c0c8 946#if 0
984263bc
MD
947 if (!pdev->si_iosize_max)
948 pdev->si_iosize_max = dev->si_iosize_max;
e4c9c0c8 949#endif
fef8985e
MD
950 error = dev_dopen(dp->d_rawdev, ap->a_oflags,
951 ap->a_devtype, ap->a_cred);
984263bc 952 }
984263bc
MD
953
954 if (error)
955 goto out;
a688b15c 956 error = dsopen(dev, ap->a_devtype, dp->d_info.d_dsflags,
84f8b009 957 &dp->d_slice, &dp->d_info);
cd29885a 958 if (!dsisopen(dp->d_slice)) {
fef8985e 959 dev_dclose(dp->d_rawdev, ap->a_oflags, ap->a_devtype);
cd29885a 960 }
d7d5e114 961out:
984263bc
MD
962 dp->d_flags &= ~DISKFLAG_LOCK;
963 if (dp->d_flags & DISKFLAG_WANTED) {
964 dp->d_flags &= ~DISKFLAG_WANTED;
965 wakeup(dp);
966 }
9f889dc4 967 rel_mplock();
d7d5e114 968
8d51c2a2
AH
969 KKASSERT(dp->d_opencount >= 0);
970 /* If the open was successful, bump open count */
971 if (error == 0)
972 atomic_add_int(&dp->d_opencount, 1);
973
984263bc
MD
974 return(error);
975}
976
e4c9c0c8
MD
977/*
978 * Close a disk device or partition
979 */
fbda7fa6
MD
980static
981int
fef8985e 982diskclose(struct dev_close_args *ap)
984263bc 983{
b13267a5 984 cdev_t dev = ap->a_head.a_dev;
984263bc
MD
985 struct disk *dp;
986 int error;
2d5beb81 987 int lcount;
984263bc
MD
988
989 error = 0;
e4c9c0c8
MD
990 dp = dev->si_disk;
991
2d5beb81
MD
992 /*
993 * The cdev_t represents the disk/slice/part. The shared
994 * dp structure governs all cdevs associated with the disk.
995 *
996 * As a safety only close the underlying raw device on the last
997 * close the disk device if our tracking of the slices/partitions
998 * also indicates nothing is open.
999 */
8d51c2a2 1000 KKASSERT(dp->d_opencount >= 1);
2d5beb81 1001 lcount = atomic_fetchadd_int(&dp->d_opencount, -1);
8d51c2a2 1002
9f889dc4 1003 get_mplock();
fef8985e 1004 dsclose(dev, ap->a_devtype, dp->d_slice);
2d5beb81 1005 if (lcount <= 1 && !dsisopen(dp->d_slice)) {
fef8985e 1006 error = dev_dclose(dp->d_rawdev, ap->a_fflag, ap->a_devtype);
cd29885a 1007 }
9f889dc4 1008 rel_mplock();
fef8985e
MD
1009 return (error);
1010}
1011
1012/*
d7d5e114 1013 * First execute the ioctl on the disk device, and if it isn't supported
fef8985e
MD
1014 * try running it on the backing device.
1015 */
1016static
1017int
1018diskioctl(struct dev_ioctl_args *ap)
1019{
b13267a5 1020 cdev_t dev = ap->a_head.a_dev;
fef8985e
MD
1021 struct disk *dp;
1022 int error;
b24cd69c 1023 u_int u;
fef8985e
MD
1024
1025 dp = dev->si_disk;
1026 if (dp == NULL)
1027 return (ENXIO);
cd29885a 1028
149e86b9 1029 devfs_debug(DEVFS_DEBUG_DEBUG,
fcefa6f2 1030 "diskioctl: cmd is: %lx (name: %s)\n",
149e86b9
MD
1031 ap->a_cmd, dev->si_name);
1032 devfs_debug(DEVFS_DEBUG_DEBUG,
fcefa6f2 1033 "diskioctl: &dp->d_slice is: %p, %p\n",
149e86b9 1034 &dp->d_slice, dp->d_slice);
cd29885a 1035
b24cd69c
AH
1036 if (ap->a_cmd == DIOCGKERNELDUMP) {
1037 u = *(u_int *)ap->a_data;
1038 return disk_dumpconf(dev, u);
1039 }
1040
81b331be
AH
1041 if (&dp->d_slice == NULL || dp->d_slice == NULL ||
1042 ((dp->d_info.d_dsflags & DSO_DEVICEMAPPER) &&
1043 dkslice(dev) == WHOLE_DISK_SLICE)) {
c6ef65ea
AH
1044 error = ENOIOCTL;
1045 } else {
9f889dc4 1046 get_mplock();
c6ef65ea
AH
1047 error = dsioctl(dev, ap->a_cmd, ap->a_data, ap->a_fflag,
1048 &dp->d_slice, &dp->d_info);
9f889dc4 1049 rel_mplock();
c6ef65ea 1050 }
cd29885a 1051
fef8985e
MD
1052 if (error == ENOIOCTL) {
1053 error = dev_dioctl(dp->d_rawdev, ap->a_cmd, ap->a_data,
87baaf0c 1054 ap->a_fflag, ap->a_cred, NULL);
fef8985e 1055 }
984263bc
MD
1056 return (error);
1057}
1058
e4c9c0c8
MD
1059/*
1060 * Execute strategy routine
1061 */
fbda7fa6 1062static
fef8985e
MD
1063int
1064diskstrategy(struct dev_strategy_args *ap)
984263bc 1065{
b13267a5 1066 cdev_t dev = ap->a_head.a_dev;
fef8985e 1067 struct bio *bio = ap->a_bio;
81b5c339 1068 struct bio *nbio;
984263bc
MD
1069 struct disk *dp;
1070
81b5c339 1071 dp = dev->si_disk;
984263bc 1072
e4c9c0c8 1073 if (dp == NULL) {
81b5c339
MD
1074 bio->bio_buf->b_error = ENXIO;
1075 bio->bio_buf->b_flags |= B_ERROR;
1076 biodone(bio);
fef8985e 1077 return(0);
984263bc 1078 }
81b5c339 1079 KKASSERT(dev->si_disk == dp);
984263bc 1080
6f76c57e
HP
1081 /*
1082 * The dscheck() function will also transform the slice relative
54078292 1083 * block number i.e. bio->bio_offset into a block number that can be
9a71d53f
MD
1084 * passed directly to the underlying raw device. If dscheck()
1085 * returns NULL it will have handled the bio for us (e.g. EOF
1086 * or error due to being beyond the device size).
6f76c57e 1087 */
cd29885a 1088 if ((nbio = dscheck(dev, bio, dp->d_slice)) != NULL) {
8c72e3d5 1089 dsched_queue(dp, nbio);
cd29885a 1090 } else {
81b5c339 1091 biodone(bio);
cd29885a 1092 }
fef8985e 1093 return(0);
984263bc
MD
1094}
1095
335dda38 1096/*
fef8985e 1097 * Return the partition size in ?blocks?
335dda38 1098 */
fbda7fa6
MD
1099static
1100int
fef8985e 1101diskpsize(struct dev_psize_args *ap)
984263bc 1102{
b13267a5 1103 cdev_t dev = ap->a_head.a_dev;
984263bc 1104 struct disk *dp;
984263bc 1105
e4c9c0c8
MD
1106 dp = dev->si_disk;
1107 if (dp == NULL)
fef8985e 1108 return(ENODEV);
81b331be 1109
fef8985e 1110 ap->a_result = dssize(dev, &dp->d_slice);
81b331be
AH
1111
1112 if ((ap->a_result == -1) &&
1113 (dp->d_info.d_dsflags & DSO_DEVICEMAPPER)) {
1114 ap->a_head.a_dev = dp->d_rawdev;
1115 return dev_doperate(&ap->a_head);
1116 }
fef8985e 1117 return(0);
984263bc
MD
1118}
1119
fef8985e
MD
1120int
1121diskdump(struct dev_dump_args *ap)
1122{
b13267a5 1123 cdev_t dev = ap->a_head.a_dev;
aec8eea4 1124 struct disk *dp = dev->si_disk;
b24cd69c 1125 u_int64_t size, offset;
fef8985e
MD
1126 int error;
1127
b24cd69c
AH
1128 error = disk_dumpcheck(dev, &size, &ap->a_blkno, &ap->a_secsize);
1129 /* XXX: this should probably go in disk_dumpcheck somehow */
1130 if (ap->a_length != 0) {
1131 size *= DEV_BSIZE;
1132 offset = ap->a_blkno * DEV_BSIZE;
1133 if ((ap->a_offset < offset) ||
1134 (ap->a_offset + ap->a_length - offset > size)) {
1135 kprintf("Attempt to write outside dump device boundaries.\n");
1136 error = ENOSPC;
1137 }
1138 }
1139
fef8985e
MD
1140 if (error == 0) {
1141 ap->a_head.a_dev = dp->d_rawdev;
1142 error = dev_doperate(&ap->a_head);
984263bc 1143 }
fef8985e
MD
1144
1145 return(error);
984263bc
MD
1146}
1147
fef8985e 1148
d7d5e114 1149SYSCTL_INT(_debug_sizeof, OID_AUTO, diskslices, CTLFLAG_RD,
984263bc
MD
1150 0, sizeof(struct diskslices), "sizeof(struct diskslices)");
1151
d7d5e114 1152SYSCTL_INT(_debug_sizeof, OID_AUTO, disk, CTLFLAG_RD,
984263bc 1153 0, sizeof(struct disk), "sizeof(struct disk)");
7a9e53ad 1154
ef548879 1155/*
4afeea0d
MD
1156 * Reorder interval for burst write allowance and minor write
1157 * allowance.
1158 *
1159 * We always want to trickle some writes in to make use of the
1160 * disk's zone cache. Bursting occurs on a longer interval and only
1161 * runningbufspace is well over the hirunningspace limit.
ef548879 1162 */
4afeea0d
MD
1163int bioq_reorder_burst_interval = 60; /* should be multiple of minor */
1164SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_burst_interval,
1165 CTLFLAG_RW, &bioq_reorder_burst_interval, 0, "");
1166int bioq_reorder_minor_interval = 5;
1167SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_minor_interval,
1168 CTLFLAG_RW, &bioq_reorder_minor_interval, 0, "");
1169
1170int bioq_reorder_burst_bytes = 3000000;
1171SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_burst_bytes,
1172 CTLFLAG_RW, &bioq_reorder_burst_bytes, 0, "");
1173int bioq_reorder_minor_bytes = 262144;
1174SYSCTL_INT(_kern, OID_AUTO, bioq_reorder_minor_bytes,
1175 CTLFLAG_RW, &bioq_reorder_minor_bytes, 0, "");
ef548879 1176
7a9e53ad
MD
1177
1178/*
30e5862e
MD
1179 * Order I/Os. Generally speaking this code is designed to make better
1180 * use of drive zone caches. A drive zone cache can typically track linear
1181 * reads or writes for around 16 zones simultaniously.
7a9e53ad 1182 *
30e5862e
MD
1183 * Read prioritization issues: It is possible for hundreds of megabytes worth
1184 * of writes to be queued asynchronously. This creates a huge bottleneck
1185 * for reads which reduce read bandwidth to a trickle.
7a9e53ad 1186 *
4afeea0d
MD
1187 * To solve this problem we generally reorder reads before writes.
1188 *
1189 * However, a large number of random reads can also starve writes and
1190 * make poor use of the drive zone cache so we allow writes to trickle
1191 * in every N reads.
7a9e53ad
MD
1192 */
1193void
81b5c339 1194bioqdisksort(struct bio_queue_head *bioq, struct bio *bio)
7a9e53ad 1195{
4afeea0d
MD
1196 /*
1197 * The BIO wants to be ordered. Adding to the tail also
1198 * causes transition to be set to NULL, forcing the ordering
1199 * of all prior I/O's.
1200 */
1201 if (bio->bio_buf->b_flags & B_ORDERED) {
1202 bioq_insert_tail(bioq, bio);
1203 return;
1204 }
1205
30e5862e
MD
1206 switch(bio->bio_buf->b_cmd) {
1207 case BUF_CMD_READ:
1208 if (bioq->transition) {
7a9e53ad 1209 /*
4afeea0d
MD
1210 * Insert before the first write. Bleedover writes
1211 * based on reorder intervals to prevent starvation.
7a9e53ad 1212 */
30e5862e 1213 TAILQ_INSERT_BEFORE(bioq->transition, bio, bio_act);
4afeea0d
MD
1214 ++bioq->reorder;
1215 if (bioq->reorder % bioq_reorder_minor_interval == 0) {
30e5862e 1216 bioqwritereorder(bioq);
4afeea0d
MD
1217 if (bioq->reorder >=
1218 bioq_reorder_burst_interval) {
1219 bioq->reorder = 0;
1220 }
7a9e53ad
MD
1221 }
1222 } else {
7a9e53ad 1223 /*
30e5862e
MD
1224 * No writes queued (or ordering was forced),
1225 * insert at tail.
7a9e53ad 1226 */
30e5862e 1227 TAILQ_INSERT_TAIL(&bioq->queue, bio, bio_act);
7a9e53ad 1228 }
30e5862e
MD
1229 break;
1230 case BUF_CMD_WRITE:
1231 /*
1232 * Writes are always appended. If no writes were previously
1233 * queued or an ordered tail insertion occured the transition
1234 * field will be NULL.
1235 */
1236 TAILQ_INSERT_TAIL(&bioq->queue, bio, bio_act);
1237 if (bioq->transition == NULL)
1238 bioq->transition = bio;
1239 break;
1240 default:
1241 /*
1242 * All other request types are forced to be ordered.
1243 */
1244 bioq_insert_tail(bioq, bio);
4afeea0d 1245 break;
7a9e53ad 1246 }
30e5862e 1247}
7a9e53ad 1248
30e5862e 1249/*
4afeea0d
MD
1250 * Move the read-write transition point to prevent reads from
1251 * completely starving our writes. This brings a number of writes into
30e5862e 1252 * the fold every N reads.
4afeea0d
MD
1253 *
1254 * We bring a few linear writes into the fold on a minor interval
1255 * and we bring a non-linear burst of writes into the fold on a major
1256 * interval. Bursting only occurs if runningbufspace is really high
1257 * (typically from syncs, fsyncs, or HAMMER flushes).
30e5862e
MD
1258 */
1259static
1260void
1261bioqwritereorder(struct bio_queue_head *bioq)
1262{
1263 struct bio *bio;
1264 off_t next_offset;
4afeea0d 1265 size_t left;
30e5862e 1266 size_t n;
4afeea0d
MD
1267 int check_off;
1268
1269 if (bioq->reorder < bioq_reorder_burst_interval ||
1270 !buf_runningbufspace_severe()) {
1271 left = (size_t)bioq_reorder_minor_bytes;
1272 check_off = 1;
1273 } else {
1274 left = (size_t)bioq_reorder_burst_bytes;
1275 check_off = 0;
1276 }
30e5862e
MD
1277
1278 next_offset = bioq->transition->bio_offset;
1279 while ((bio = bioq->transition) != NULL &&
4afeea0d
MD
1280 (check_off == 0 || next_offset == bio->bio_offset)
1281 ) {
30e5862e
MD
1282 n = bio->bio_buf->b_bcount;
1283 next_offset = bio->bio_offset + n;
1284 bioq->transition = TAILQ_NEXT(bio, bio_act);
1285 if (left < n)
7a9e53ad 1286 break;
30e5862e 1287 left -= n;
7a9e53ad 1288 }
7a9e53ad
MD
1289}
1290
7a9e53ad 1291/*
c6ef65ea
AH
1292 * Bounds checking against the media size, used for the raw partition.
1293 * secsize, mediasize and b_blkno must all be the same units.
1294 * Possibly this has to be DEV_BSIZE (512).
1295 */
1296int
1297bounds_check_with_mediasize(struct bio *bio, int secsize, uint64_t mediasize)
1298{
1299 struct buf *bp = bio->bio_buf;
1300 int64_t sz;
1301
1302 sz = howmany(bp->b_bcount, secsize);
1303
1304 if (bio->bio_offset/DEV_BSIZE + sz > mediasize) {
1305 sz = mediasize - bio->bio_offset/DEV_BSIZE;
1306 if (sz == 0) {
1307 /* If exactly at end of disk, return EOF. */
1308 bp->b_resid = bp->b_bcount;
1309 return 0;
1310 }
1311 if (sz < 0) {
1312 /* If past end of disk, return EINVAL. */
1313 bp->b_error = EINVAL;
1314 return 0;
1315 }
1316 /* Otherwise, truncate request. */
1317 bp->b_bcount = sz * secsize;
1318 }
1319
1320 return 1;
1321}
1322
1323/*
7a9e53ad
MD
1324 * Disk error is the preface to plaintive error messages
1325 * about failing disk transfers. It prints messages of the form
1326
1327hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
1328
1329 * if the offset of the error in the transfer and a disk label
1330 * are both available. blkdone should be -1 if the position of the error
1331 * is unknown; the disklabel pointer may be null from drivers that have not
6ea70f76 1332 * been converted to use them. The message is printed with kprintf
7a9e53ad 1333 * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
6ea70f76 1334 * The message should be completed (with at least a newline) with kprintf
a0a36cfd 1335 * or log(-1, ...), respectively. There is no trailing space.
7a9e53ad
MD
1336 */
1337void
a688b15c 1338diskerr(struct bio *bio, cdev_t dev, const char *what, int pri, int donecnt)
7a9e53ad 1339{
81b5c339 1340 struct buf *bp = bio->bio_buf;
c6f49b01
MD
1341 const char *term;
1342
1343 switch(bp->b_cmd) {
1344 case BUF_CMD_READ:
1345 term = "read";
1346 break;
1347 case BUF_CMD_WRITE:
1348 term = "write";
1349 break;
1350 default:
1351 term = "access";
1352 break;
1353 }
cd29885a 1354 kprintf("%s: %s %sing ", dev->si_name, what, term);
973c11b9
MD
1355 kprintf("offset %012llx for %d",
1356 (long long)bio->bio_offset,
1357 bp->b_bcount);
cd29885a 1358
54078292 1359 if (donecnt)
6ea70f76 1360 kprintf(" (%d bytes completed)", donecnt);
7a9e53ad 1361}
81b5c339 1362
a8873631
MD
1363/*
1364 * Locate a disk device
1365 */
1366cdev_t
1367disk_locate(const char *devname)
1368{
cd29885a
MD
1369 return devfs_find_device_by_name(devname);
1370}
a8873631 1371
cd29885a
MD
1372void
1373disk_config(void *arg)
1374{
aec8eea4 1375 disk_msg_send_sync(DISK_SYNC, NULL, NULL);
cd29885a
MD
1376}
1377
cd29885a
MD
1378static void
1379disk_init(void)
1380{
1381 struct thread* td_core;
cd29885a 1382
149e86b9
MD
1383 disk_msg_cache = objcache_create("disk-msg-cache", 0, 0,
1384 NULL, NULL, NULL,
1385 objcache_malloc_alloc,
1386 objcache_malloc_free,
1387 &disk_msg_malloc_args);
cd29885a 1388
a3c18566 1389 lwkt_token_init(&disklist_token, "disks");
be755ff9 1390
149e86b9
MD
1391 /*
1392 * Initialize the reply-only port which acts as a message drain
1393 */
cd29885a
MD
1394 lwkt_initport_replyonly(&disk_dispose_port, disk_msg_autofree_reply);
1395
c9e9fb21 1396 lwkt_gettoken(&disklist_token);
cd29885a 1397 lwkt_create(disk_msg_core, /*args*/NULL, &td_core, NULL,
fdce8919 1398 0, 0, "disk_msg_core");
cd29885a 1399 tsleep(td_core, 0, "diskcore", 0);
c9e9fb21 1400 lwkt_reltoken(&disklist_token);
cd29885a
MD
1401}
1402
cd29885a
MD
1403static void
1404disk_uninit(void)
1405{
cd29885a 1406 objcache_destroy(disk_msg_cache);
a8873631
MD
1407}
1408
fbbbca99
MD
1409/*
1410 * Clean out illegal characters in serial numbers.
1411 */
1412static void
1413disk_cleanserial(char *serno)
1414{
1415 char c;
1416
1417 while ((c = *serno) != 0) {
1418 if (c >= 'a' && c <= 'z')
1419 ;
1420 else if (c >= 'A' && c <= 'Z')
1421 ;
1422 else if (c >= '0' && c <= '9')
1423 ;
1424 else if (c == '-' || c == '@' || c == '+' || c == '.')
1425 ;
1426 else
1427 c = '_';
1428 *serno++= c;
1429 }
1430}
1431
8c05caab
AH
1432TUNABLE_INT("kern.disk_debug", &disk_debug_enable);
1433SYSCTL_INT(_kern, OID_AUTO, disk_debug, CTLFLAG_RW, &disk_debug_enable,
1434 0, "Enable subr_disk debugging");
1435
cd29885a
MD
1436SYSINIT(disk_register, SI_SUB_PRE_DRIVERS, SI_ORDER_FIRST, disk_init, NULL);
1437SYSUNINIT(disk_register, SI_SUB_PRE_DRIVERS, SI_ORDER_ANY, disk_uninit, NULL);