xref: /netbsd-src/sys/dev/dksubr.c (revision a6f3f22f245acb8ee3bbf6871d7dce989204fa97)
1 /* $NetBSD: dksubr.c,v 1.81 2015/10/23 01:34:22 christos Exp $ */
2 
3 /*-
4  * Copyright (c) 1996, 1997, 1998, 1999, 2002, 2008 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Jason R. Thorpe and Roland C. Dowdeswell.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  */
31 
32 #include <sys/cdefs.h>
33 __KERNEL_RCSID(0, "$NetBSD: dksubr.c,v 1.81 2015/10/23 01:34:22 christos Exp $");
34 
35 #include <sys/param.h>
36 #include <sys/systm.h>
37 #include <sys/stat.h>
38 #include <sys/proc.h>
39 #include <sys/ioctl.h>
40 #include <sys/device.h>
41 #include <sys/disk.h>
42 #include <sys/disklabel.h>
43 #include <sys/buf.h>
44 #include <sys/bufq.h>
45 #include <sys/vnode.h>
46 #include <sys/fcntl.h>
47 #include <sys/namei.h>
48 #include <sys/module.h>
49 #include <sys/syslog.h>
50 
51 #include <dev/dkvar.h>
52 #include <miscfs/specfs/specdev.h> /* for v_rdev */
53 
54 int	dkdebug = 0;
55 
56 #ifdef DEBUG
57 #define DKDB_FOLLOW	0x1
58 #define DKDB_INIT	0x2
59 #define DKDB_VNODE	0x4
60 #define DKDB_DUMP	0x8
61 
62 #define IFDEBUG(x,y)		if (dkdebug & (x)) y
63 #define DPRINTF(x,y)		IFDEBUG(x, printf y)
64 #define DPRINTF_FOLLOW(y)	DPRINTF(DKDB_FOLLOW, y)
65 #else
66 #define IFDEBUG(x,y)
67 #define DPRINTF(x,y)
68 #define DPRINTF_FOLLOW(y)
69 #endif
70 
71 #define DKF_READYFORDUMP	(DKF_INITED|DKF_TAKEDUMP)
72 
73 static int dk_subr_modcmd(modcmd_t, void *);
74 
75 #define DKLABELDEV(dev)	\
76 	(MAKEDISKDEV(major((dev)), DISKUNIT((dev)), RAW_PART))
77 
78 static void	dk_makedisklabel(struct dk_softc *);
79 static int	dk_translate(struct dk_softc *, struct buf *);
80 static void	dk_done1(struct dk_softc *, struct buf *, bool);
81 
82 void
83 dk_init(struct dk_softc *dksc, device_t dev, int dtype)
84 {
85 
86 	memset(dksc, 0x0, sizeof(*dksc));
87 	dksc->sc_dtype = dtype;
88 	dksc->sc_dev = dev;
89 
90 	strlcpy(dksc->sc_xname, device_xname(dev), DK_XNAME_SIZE);
91 	dksc->sc_dkdev.dk_name = dksc->sc_xname;
92 }
93 
94 void
95 dk_attach(struct dk_softc *dksc)
96 {
97 	mutex_init(&dksc->sc_iolock, MUTEX_DEFAULT, IPL_VM);
98 	dksc->sc_flags |= DKF_READYFORDUMP;
99 #ifdef DIAGNOSTIC
100 	dksc->sc_flags |= DKF_WARNLABEL | DKF_LABELSANITY;
101 #endif
102 
103 	/* Attach the device into the rnd source list. */
104 	rnd_attach_source(&dksc->sc_rnd_source, dksc->sc_xname,
105 	    RND_TYPE_DISK, RND_FLAG_DEFAULT);
106 }
107 
108 void
109 dk_detach(struct dk_softc *dksc)
110 {
111 	/* Unhook the entropy source. */
112 	rnd_detach_source(&dksc->sc_rnd_source);
113 
114 	dksc->sc_flags &= ~DKF_READYFORDUMP;
115 	mutex_destroy(&dksc->sc_iolock);
116 }
117 
118 /* ARGSUSED */
119 int
120 dk_open(struct dk_softc *dksc, dev_t dev,
121     int flags, int fmt, struct lwp *l)
122 {
123 	struct	disklabel *lp = dksc->sc_dkdev.dk_label;
124 	int	part = DISKPART(dev);
125 	int	pmask = 1 << part;
126 	int	ret = 0;
127 	struct disk *dk = &dksc->sc_dkdev;
128 
129 	DPRINTF_FOLLOW(("%s(%s, %p, 0x%"PRIx64", 0x%x)\n", __func__,
130 	    dksc->sc_xname, dksc, dev, flags));
131 
132 	mutex_enter(&dk->dk_openlock);
133 
134 	/*
135 	 * If there are wedges, and this is not RAW_PART, then we
136 	 * need to fail.
137 	 */
138 	if (dk->dk_nwedges != 0 && part != RAW_PART) {
139 		ret = EBUSY;
140 		goto done;
141 	}
142 
143 	/*
144 	 * If we're init'ed and there are no other open partitions then
145 	 * update the in-core disklabel.
146 	 */
147 	if ((dksc->sc_flags & DKF_INITED)) {
148 		if ((dksc->sc_flags & DKF_VLABEL) == 0) {
149 			dksc->sc_flags |= DKF_VLABEL;
150 			dk_getdisklabel(dksc, dev);
151 		}
152 	}
153 
154 	/* Fail if we can't find the partition. */
155 	if (part != RAW_PART &&
156 	    ((dksc->sc_flags & DKF_VLABEL) == 0 ||
157 	     part >= lp->d_npartitions ||
158 	     lp->d_partitions[part].p_fstype == FS_UNUSED)) {
159 		ret = ENXIO;
160 		goto done;
161 	}
162 
163 	/* Mark our unit as open. */
164 	switch (fmt) {
165 	case S_IFCHR:
166 		dk->dk_copenmask |= pmask;
167 		break;
168 	case S_IFBLK:
169 		dk->dk_bopenmask |= pmask;
170 		break;
171 	}
172 
173 	dk->dk_openmask = dk->dk_copenmask | dk->dk_bopenmask;
174 
175 done:
176 	mutex_exit(&dk->dk_openlock);
177 	return ret;
178 }
179 
180 /* ARGSUSED */
181 int
182 dk_close(struct dk_softc *dksc, dev_t dev,
183     int flags, int fmt, struct lwp *l)
184 {
185 	const struct dkdriver *dkd = dksc->sc_dkdev.dk_driver;
186 	int	part = DISKPART(dev);
187 	int	pmask = 1 << part;
188 	struct disk *dk = &dksc->sc_dkdev;
189 
190 	DPRINTF_FOLLOW(("%s(%s, %p, 0x%"PRIx64", 0x%x)\n", __func__,
191 	    dksc->sc_xname, dksc, dev, flags));
192 
193 	mutex_enter(&dk->dk_openlock);
194 
195 	switch (fmt) {
196 	case S_IFCHR:
197 		dk->dk_copenmask &= ~pmask;
198 		break;
199 	case S_IFBLK:
200 		dk->dk_bopenmask &= ~pmask;
201 		break;
202 	}
203 	dk->dk_openmask = dk->dk_copenmask | dk->dk_bopenmask;
204 
205 	if (dk->dk_openmask == 0) {
206 		if (dkd->d_lastclose != NULL)
207 			(*dkd->d_lastclose)(dksc->sc_dev);
208 		if ((dksc->sc_flags & DKF_KLABEL) == 0)
209 			dksc->sc_flags &= ~DKF_VLABEL;
210 	}
211 
212 	mutex_exit(&dk->dk_openlock);
213 	return 0;
214 }
215 
216 static int
217 dk_translate(struct dk_softc *dksc, struct buf *bp)
218 {
219 	int	part;
220 	int	wlabel;
221 	daddr_t	blkno;
222 	struct disklabel *lp;
223 	struct disk *dk;
224 	uint64_t numsecs;
225 	unsigned secsize;
226 
227 	lp = dksc->sc_dkdev.dk_label;
228 	dk = &dksc->sc_dkdev;
229 
230 	part = DISKPART(bp->b_dev);
231 	numsecs = dk->dk_geom.dg_secperunit;
232 	secsize = dk->dk_geom.dg_secsize;
233 
234 	/*
235 	 * The transfer must be a whole number of blocks and the offset must
236 	 * not be negative.
237 	 */
238 	if ((bp->b_bcount % secsize) != 0 || bp->b_blkno < 0) {
239 		bp->b_error = EINVAL;
240 		goto done;
241 	}
242 
243 	/* If there is nothing to do, then we are done */
244 	if (bp->b_bcount == 0)
245 		goto done;
246 
247 	wlabel = dksc->sc_flags & (DKF_WLABEL|DKF_LABELLING);
248 	if (part == RAW_PART) {
249 		if (bounds_check_with_mediasize(bp, DEV_BSIZE, numsecs) <= 0)
250 			goto done;
251 	} else {
252 		if (bounds_check_with_label(&dksc->sc_dkdev, bp, wlabel) <= 0)
253 			goto done;
254 	}
255 
256 	/*
257 	 * Convert the block number to absolute and put it in terms
258 	 * of the device's logical block size.
259 	 */
260 	if (secsize >= DEV_BSIZE)
261 		blkno = bp->b_blkno / (secsize / DEV_BSIZE);
262 	else
263 		blkno = bp->b_blkno * (DEV_BSIZE / secsize);
264 
265 	if (part != RAW_PART)
266 		blkno += lp->d_partitions[DISKPART(bp->b_dev)].p_offset;
267 	bp->b_rawblkno = blkno;
268 
269 	return -1;
270 
271 done:
272 	bp->b_resid = bp->b_bcount;
273 	return bp->b_error;
274 }
275 
276 void
277 dk_strategy(struct dk_softc *dksc, struct buf *bp)
278 {
279 	int error;
280 
281 	DPRINTF_FOLLOW(("%s(%s, %p, %p)\n", __func__,
282 	    dksc->sc_xname, dksc, bp));
283 
284 	if (!(dksc->sc_flags & DKF_INITED)) {
285 		DPRINTF_FOLLOW(("%s: not inited\n", __func__));
286 		bp->b_error  = ENXIO;
287 		biodone(bp);
288 		return;
289 	}
290 
291 	error = dk_translate(dksc, bp);
292 	if (error >= 0) {
293 		biodone(bp);
294 		return;
295 	}
296 
297 	/*
298 	 * Queue buffer and start unit
299 	 */
300 	dk_start(dksc, bp);
301 }
302 
303 void
304 dk_start(struct dk_softc *dksc, struct buf *bp)
305 {
306 	const struct dkdriver *dkd = dksc->sc_dkdev.dk_driver;
307 	int error;
308 
309 	mutex_enter(&dksc->sc_iolock);
310 
311 	if (bp != NULL)
312 		bufq_put(dksc->sc_bufq, bp);
313 
314 	if (dksc->sc_busy)
315 		goto done;
316 	dksc->sc_busy = true;
317 
318 	/*
319 	 * Peeking at the buffer queue and committing the operation
320 	 * only after success isn't atomic.
321 	 *
322 	 * So when a diskstart fails, the buffer is saved
323 	 * and tried again before the next buffer is fetched.
324 	 * dk_drain() handles flushing of a saved buffer.
325 	 *
326 	 * This keeps order of I/O operations, unlike bufq_put.
327 	 */
328 
329 	bp = dksc->sc_deferred;
330 	dksc->sc_deferred = NULL;
331 
332 	if (bp == NULL)
333 		bp = bufq_get(dksc->sc_bufq);
334 
335 	while (bp != NULL) {
336 
337 		disk_busy(&dksc->sc_dkdev);
338 		mutex_exit(&dksc->sc_iolock);
339 		error = dkd->d_diskstart(dksc->sc_dev, bp);
340 		mutex_enter(&dksc->sc_iolock);
341 		if (error == EAGAIN) {
342 			dksc->sc_deferred = bp;
343 			disk_unbusy(&dksc->sc_dkdev, 0, (bp->b_flags & B_READ));
344 			break;
345 		}
346 
347 		if (error != 0) {
348 			bp->b_error = error;
349 			bp->b_resid = bp->b_bcount;
350 			dk_done1(dksc, bp, false);
351 		}
352 
353 		bp = bufq_get(dksc->sc_bufq);
354 	}
355 
356 	dksc->sc_busy = false;
357 done:
358 	mutex_exit(&dksc->sc_iolock);
359 }
360 
361 static void
362 dk_done1(struct dk_softc *dksc, struct buf *bp, bool lock)
363 {
364 	struct disk *dk = &dksc->sc_dkdev;
365 
366 	if (bp->b_error != 0) {
367 		struct cfdriver *cd = device_cfdriver(dksc->sc_dev);
368 
369 		diskerr(bp, cd->cd_name, "error", LOG_PRINTF, 0,
370 			dk->dk_label);
371 		printf("\n");
372 	}
373 
374 	if (lock)
375 		mutex_enter(&dksc->sc_iolock);
376 	disk_unbusy(dk, bp->b_bcount - bp->b_resid, (bp->b_flags & B_READ));
377 	if (lock)
378 		mutex_exit(&dksc->sc_iolock);
379 
380 	rnd_add_uint32(&dksc->sc_rnd_source, bp->b_rawblkno);
381 
382 	biodone(bp);
383 }
384 
385 void
386 dk_done(struct dk_softc *dksc, struct buf *bp)
387 {
388 	dk_done1(dksc, bp, true);
389 }
390 
391 void
392 dk_drain(struct dk_softc *dksc)
393 {
394 	struct buf *bp;
395 
396 	mutex_enter(&dksc->sc_iolock);
397 	bp = dksc->sc_deferred;
398 	if (bp != NULL) {
399 		bp->b_error = EIO;
400 		bp->b_resid = bp->b_bcount;
401 		biodone(bp);
402 	}
403 	bufq_drain(dksc->sc_bufq);
404 	mutex_exit(&dksc->sc_iolock);
405 }
406 
407 int
408 dk_discard(struct dk_softc *dksc, dev_t dev, off_t pos, off_t len)
409 {
410 	const struct dkdriver *dkd = dksc->sc_dkdev.dk_driver;
411 	unsigned secsize = dksc->sc_dkdev.dk_geom.dg_secsize;
412 	struct buf tmp, *bp = &tmp;
413 	int error;
414 
415 	DPRINTF_FOLLOW(("%s(%s, %p, 0x"PRIx64", %jd, %jd)\n", __func__,
416 	    dksc->sc_xname, dksc, (intmax_t)pos, (intmax_t)len));
417 
418 	if (!(dksc->sc_flags & DKF_INITED)) {
419 		DPRINTF_FOLLOW(("%s: not inited\n", __func__));
420 		return ENXIO;
421 	}
422 
423 	if (secsize == 0 || (pos % secsize) != 0)
424 		return EINVAL;
425 
426 	/* enough data to please the bounds checking code */
427 	bp->b_dev = dev;
428 	bp->b_blkno = (daddr_t)(pos / secsize);
429 	bp->b_bcount = len;
430 	bp->b_flags = B_WRITE;
431 
432 	error = dk_translate(dksc, bp);
433 	if (error >= 0)
434 		return error;
435 
436 	error = dkd->d_discard(dksc->sc_dev,
437 		(off_t)bp->b_rawblkno * secsize,
438 		(off_t)bp->b_bcount);
439 
440 	return error;
441 }
442 
443 int
444 dk_size(struct dk_softc *dksc, dev_t dev)
445 {
446 	const struct dkdriver *dkd = dksc->sc_dkdev.dk_driver;
447 	struct	disklabel *lp;
448 	int	is_open;
449 	int	part;
450 	int	size;
451 
452 	if ((dksc->sc_flags & DKF_INITED) == 0)
453 		return -1;
454 
455 	part = DISKPART(dev);
456 	is_open = dksc->sc_dkdev.dk_openmask & (1 << part);
457 
458 	if (!is_open && dkd->d_open(dev, 0, S_IFBLK, curlwp))
459 		return -1;
460 
461 	lp = dksc->sc_dkdev.dk_label;
462 	if (lp->d_partitions[part].p_fstype != FS_SWAP)
463 		size = -1;
464 	else
465 		size = lp->d_partitions[part].p_size *
466 		    (lp->d_secsize / DEV_BSIZE);
467 
468 	if (!is_open && dkd->d_close(dev, 0, S_IFBLK, curlwp))
469 		return -1;
470 
471 	return size;
472 }
473 
474 int
475 dk_ioctl(struct dk_softc *dksc, dev_t dev,
476 	    u_long cmd, void *data, int flag, struct lwp *l)
477 {
478 	const struct dkdriver *dkd = dksc->sc_dkdev.dk_driver;
479 	struct	disklabel *lp;
480 	struct	disk *dk = &dksc->sc_dkdev;
481 #ifdef __HAVE_OLD_DISKLABEL
482 	struct	disklabel newlabel;
483 #endif
484 	int	error;
485 
486 	DPRINTF_FOLLOW(("%s(%s, %p, 0x%"PRIx64", 0x%lx)\n", __func__,
487 	    dksc->sc_xname, dksc, dev, cmd));
488 
489 	/* ensure that the pseudo disk is open for writes for these commands */
490 	switch (cmd) {
491 	case DIOCSDINFO:
492 	case DIOCWDINFO:
493 #ifdef __HAVE_OLD_DISKLABEL
494 	case ODIOCSDINFO:
495 	case ODIOCWDINFO:
496 #endif
497 	case DIOCKLABEL:
498 	case DIOCWLABEL:
499 	case DIOCAWEDGE:
500 	case DIOCDWEDGE:
501 	case DIOCSSTRATEGY:
502 		if ((flag & FWRITE) == 0)
503 			return EBADF;
504 	}
505 
506 	/* ensure that the pseudo-disk is initialized for these */
507 	switch (cmd) {
508 	case DIOCGDINFO:
509 	case DIOCSDINFO:
510 	case DIOCWDINFO:
511 	case DIOCGPART:
512 	case DIOCKLABEL:
513 	case DIOCWLABEL:
514 	case DIOCGDEFLABEL:
515 	case DIOCAWEDGE:
516 	case DIOCDWEDGE:
517 	case DIOCLWEDGES:
518 	case DIOCMWEDGES:
519 	case DIOCCACHESYNC:
520 #ifdef __HAVE_OLD_DISKLABEL
521 	case ODIOCGDINFO:
522 	case ODIOCSDINFO:
523 	case ODIOCWDINFO:
524 	case ODIOCGDEFLABEL:
525 #endif
526 		if ((dksc->sc_flags & DKF_INITED) == 0)
527 			return ENXIO;
528 	}
529 
530 	error = disk_ioctl(dk, dev, cmd, data, flag, l);
531 	if (error != EPASSTHROUGH)
532 		return error;
533 	else
534 		error = 0;
535 
536 	switch (cmd) {
537 	case DIOCWDINFO:
538 	case DIOCSDINFO:
539 #ifdef __HAVE_OLD_DISKLABEL
540 	case ODIOCWDINFO:
541 	case ODIOCSDINFO:
542 #endif
543 #ifdef __HAVE_OLD_DISKLABEL
544 		if (cmd == ODIOCSDINFO || cmd == ODIOCWDINFO) {
545 			memset(&newlabel, 0, sizeof newlabel);
546 			memcpy(&newlabel, data, sizeof (struct olddisklabel));
547 			lp = &newlabel;
548 		} else
549 #endif
550 		lp = (struct disklabel *)data;
551 
552 		mutex_enter(&dk->dk_openlock);
553 		dksc->sc_flags |= DKF_LABELLING;
554 
555 		error = setdisklabel(dksc->sc_dkdev.dk_label,
556 		    lp, 0, dksc->sc_dkdev.dk_cpulabel);
557 		if (error == 0) {
558 			if (cmd == DIOCWDINFO
559 #ifdef __HAVE_OLD_DISKLABEL
560 			    || cmd == ODIOCWDINFO
561 #endif
562 			   )
563 				error = writedisklabel(DKLABELDEV(dev),
564 				    dkd->d_strategy, dksc->sc_dkdev.dk_label,
565 				    dksc->sc_dkdev.dk_cpulabel);
566 		}
567 
568 		dksc->sc_flags &= ~DKF_LABELLING;
569 		mutex_exit(&dk->dk_openlock);
570 		break;
571 
572 	case DIOCKLABEL:
573 		if (*(int *)data != 0)
574 			dksc->sc_flags |= DKF_KLABEL;
575 		else
576 			dksc->sc_flags &= ~DKF_KLABEL;
577 		break;
578 
579 	case DIOCWLABEL:
580 		if (*(int *)data != 0)
581 			dksc->sc_flags |= DKF_WLABEL;
582 		else
583 			dksc->sc_flags &= ~DKF_WLABEL;
584 		break;
585 
586 	case DIOCGDEFLABEL:
587 		dk_getdefaultlabel(dksc, (struct disklabel *)data);
588 		break;
589 
590 #ifdef __HAVE_OLD_DISKLABEL
591 	case ODIOCGDEFLABEL:
592 		dk_getdefaultlabel(dksc, &newlabel);
593 		if (newlabel.d_npartitions > OLDMAXPARTITIONS)
594 			return ENOTTY;
595 		memcpy(data, &newlabel, sizeof (struct olddisklabel));
596 		break;
597 #endif
598 
599 	case DIOCGSTRATEGY:
600 	    {
601 		struct disk_strategy *dks = (void *)data;
602 
603 		mutex_enter(&dksc->sc_iolock);
604 		strlcpy(dks->dks_name, bufq_getstrategyname(dksc->sc_bufq),
605 		    sizeof(dks->dks_name));
606 		mutex_exit(&dksc->sc_iolock);
607 		dks->dks_paramlen = 0;
608 
609 		return 0;
610 	    }
611 
612 	case DIOCSSTRATEGY:
613 	    {
614 		struct disk_strategy *dks = (void *)data;
615 		struct bufq_state *new;
616 		struct bufq_state *old;
617 
618 		if (dks->dks_param != NULL) {
619 			return EINVAL;
620 		}
621 		dks->dks_name[sizeof(dks->dks_name) - 1] = 0; /* ensure term */
622 		error = bufq_alloc(&new, dks->dks_name,
623 		    BUFQ_EXACT|BUFQ_SORT_RAWBLOCK);
624 		if (error) {
625 			return error;
626 		}
627 		mutex_enter(&dksc->sc_iolock);
628 		old = dksc->sc_bufq;
629 		bufq_move(new, old);
630 		dksc->sc_bufq = new;
631 		mutex_exit(&dksc->sc_iolock);
632 		bufq_free(old);
633 
634 		return 0;
635 	    }
636 
637 	default:
638 		error = ENOTTY;
639 	}
640 
641 	return error;
642 }
643 
644 /*
645  * dk_dump dumps all of physical memory into the partition specified.
646  * This requires substantially more framework than {s,w}ddump, and hence
647  * is probably much more fragile.
648  *
649  */
650 
651 #define DKFF_READYFORDUMP(x)	(((x) & DKF_READYFORDUMP) == DKF_READYFORDUMP)
652 static volatile int	dk_dumping = 0;
653 
654 /* ARGSUSED */
655 int
656 dk_dump(struct dk_softc *dksc, dev_t dev,
657     daddr_t blkno, void *vav, size_t size)
658 {
659 	const struct dkdriver *dkd = dksc->sc_dkdev.dk_driver;
660 	char *va = vav;
661 	struct disklabel *lp;
662 	struct partition *p;
663 	int part, towrt, nsects, sectoff, maxblkcnt, nblk;
664 	int maxxfer, rv = 0;
665 
666 	/*
667 	 * ensure that we consider this device to be safe for dumping,
668 	 * and that the device is configured.
669 	 */
670 	if (!DKFF_READYFORDUMP(dksc->sc_flags)) {
671 		DPRINTF(DKDB_DUMP, ("%s: bad dump flags 0x%x\n", __func__,
672 		    dksc->sc_flags));
673 		return ENXIO;
674 	}
675 
676 	/* ensure that we are not already dumping */
677 	if (dk_dumping)
678 		return EFAULT;
679 	dk_dumping = 1;
680 
681 	if (dkd->d_dumpblocks == NULL) {
682 		DPRINTF(DKDB_DUMP, ("%s: no dumpblocks\n", __func__));
683 		return ENXIO;
684 	}
685 
686 	/* device specific max transfer size */
687 	maxxfer = MAXPHYS;
688 	if (dkd->d_iosize != NULL)
689 		(*dkd->d_iosize)(dksc->sc_dev, &maxxfer);
690 
691 	/* Convert to disk sectors.  Request must be a multiple of size. */
692 	part = DISKPART(dev);
693 	lp = dksc->sc_dkdev.dk_label;
694 	if ((size % lp->d_secsize) != 0) {
695 		DPRINTF(DKDB_DUMP, ("%s: odd size %zu\n", __func__, size));
696 		return EFAULT;
697 	}
698 	towrt = size / lp->d_secsize;
699 	blkno = dbtob(blkno) / lp->d_secsize;   /* blkno in secsize units */
700 
701 	p = &lp->d_partitions[part];
702 	if (p->p_fstype != FS_SWAP) {
703 		DPRINTF(DKDB_DUMP, ("%s: bad fstype %d\n", __func__,
704 		    p->p_fstype));
705 		return ENXIO;
706 	}
707 	nsects = p->p_size;
708 	sectoff = p->p_offset;
709 
710 	/* Check transfer bounds against partition size. */
711 	if ((blkno < 0) || ((blkno + towrt) > nsects)) {
712 		DPRINTF(DKDB_DUMP, ("%s: out of bounds blkno=%jd, towrt=%d, "
713 		    "nsects=%d\n", __func__, (intmax_t)blkno, towrt, nsects));
714 		return EINVAL;
715 	}
716 
717 	/* Offset block number to start of partition. */
718 	blkno += sectoff;
719 
720 	/* Start dumping and return when done. */
721 	maxblkcnt = howmany(maxxfer, lp->d_secsize);
722 	while (towrt > 0) {
723 		nblk = min(maxblkcnt, towrt);
724 
725 		if ((rv = (*dkd->d_dumpblocks)(dksc->sc_dev, va, blkno, nblk))
726 		    != 0) {
727 			DPRINTF(DKDB_DUMP, ("%s: dumpblocks %d\n", __func__,
728 			    rv));
729 			return rv;
730 		}
731 
732 		towrt -= nblk;
733 		blkno += nblk;
734 		va += nblk * lp->d_secsize;
735 	}
736 
737 	dk_dumping = 0;
738 
739 	return 0;
740 }
741 
742 /* ARGSUSED */
743 void
744 dk_getdefaultlabel(struct dk_softc *dksc, struct disklabel *lp)
745 {
746 	struct disk_geom *dg = &dksc->sc_dkdev.dk_geom;
747 
748 	memset(lp, 0, sizeof(*lp));
749 
750 	if (dg->dg_secperunit > UINT32_MAX)
751 		lp->d_secperunit = UINT32_MAX;
752 	else
753 		lp->d_secperunit = dg->dg_secperunit;
754 	lp->d_secsize = dg->dg_secsize;
755 	lp->d_nsectors = dg->dg_nsectors;
756 	lp->d_ntracks = dg->dg_ntracks;
757 	lp->d_ncylinders = dg->dg_ncylinders;
758 	lp->d_secpercyl = lp->d_ntracks * lp->d_nsectors;
759 
760 	strlcpy(lp->d_typename, dksc->sc_xname, sizeof(lp->d_typename));
761 	lp->d_type = dksc->sc_dtype;
762 	strlcpy(lp->d_packname, "fictitious", sizeof(lp->d_packname));
763 	lp->d_rpm = 3600;
764 	lp->d_interleave = 1;
765 	lp->d_flags = 0;
766 
767 	lp->d_partitions[RAW_PART].p_offset = 0;
768 	lp->d_partitions[RAW_PART].p_size = lp->d_secperunit;
769 	lp->d_partitions[RAW_PART].p_fstype = FS_UNUSED;
770 	lp->d_npartitions = RAW_PART + 1;
771 
772 	lp->d_magic = DISKMAGIC;
773 	lp->d_magic2 = DISKMAGIC;
774 	lp->d_checksum = dkcksum(dksc->sc_dkdev.dk_label);
775 }
776 
777 /* ARGSUSED */
778 void
779 dk_getdisklabel(struct dk_softc *dksc, dev_t dev)
780 {
781 	const struct dkdriver *dkd = dksc->sc_dkdev.dk_driver;
782 	struct	 disklabel *lp = dksc->sc_dkdev.dk_label;
783 	struct	 cpu_disklabel *clp = dksc->sc_dkdev.dk_cpulabel;
784 	struct   disk_geom *dg = &dksc->sc_dkdev.dk_geom;
785 	struct	 partition *pp;
786 	int	 i;
787 	const char	*errstring;
788 
789 	memset(clp, 0x0, sizeof(*clp));
790 	dk_getdefaultlabel(dksc, lp);
791 	errstring = readdisklabel(DKLABELDEV(dev), dkd->d_strategy,
792 	    dksc->sc_dkdev.dk_label, dksc->sc_dkdev.dk_cpulabel);
793 	if (errstring) {
794 		dk_makedisklabel(dksc);
795 		if (dksc->sc_flags & DKF_WARNLABEL)
796 			printf("%s: %s\n", dksc->sc_xname, errstring);
797 		return;
798 	}
799 
800 	if ((dksc->sc_flags & DKF_LABELSANITY) == 0)
801 		return;
802 
803 	/* Sanity check */
804 	if (lp->d_secperunit < UINT32_MAX ?
805 		lp->d_secperunit != dg->dg_secperunit :
806 		lp->d_secperunit > dg->dg_secperunit)
807 		printf("WARNING: %s: total sector size in disklabel (%ju) "
808 		    "!= the size of %s (%ju)\n", dksc->sc_xname,
809 		    (uintmax_t)lp->d_secperunit, dksc->sc_xname,
810 		    (uintmax_t)dg->dg_secperunit);
811 
812 	for (i=0; i < lp->d_npartitions; i++) {
813 		pp = &lp->d_partitions[i];
814 		if (pp->p_offset + pp->p_size > dg->dg_secperunit)
815 			printf("WARNING: %s: end of partition `%c' exceeds "
816 			    "the size of %s (%ju)\n", dksc->sc_xname,
817 			    'a' + i, dksc->sc_xname,
818 			    (uintmax_t)dg->dg_secperunit);
819 	}
820 }
821 
822 /* ARGSUSED */
823 static void
824 dk_makedisklabel(struct dk_softc *dksc)
825 {
826 	struct	disklabel *lp = dksc->sc_dkdev.dk_label;
827 
828 	lp->d_partitions[RAW_PART].p_fstype = FS_BSDFFS;
829 	strlcpy(lp->d_packname, "default label", sizeof(lp->d_packname));
830 	lp->d_checksum = dkcksum(lp);
831 }
832 
833 /* This function is taken from ccd.c:1.76  --rcd */
834 
835 /*
836  * XXX this function looks too generic for dksubr.c, shouldn't we
837  *     put it somewhere better?
838  */
839 
840 /*
841  * Lookup the provided name in the filesystem.  If the file exists,
842  * is a valid block device, and isn't being used by anyone else,
843  * set *vpp to the file's vnode.
844  */
845 int
846 dk_lookup(struct pathbuf *pb, struct lwp *l, struct vnode **vpp)
847 {
848 	struct nameidata nd;
849 	struct vnode *vp;
850 	int     error;
851 
852 	if (l == NULL)
853 		return ESRCH;	/* Is ESRCH the best choice? */
854 
855 	NDINIT(&nd, LOOKUP, FOLLOW, pb);
856 	if ((error = vn_open(&nd, FREAD | FWRITE, 0)) != 0) {
857 		DPRINTF((DKDB_FOLLOW|DKDB_INIT),
858 		    ("%s: vn_open error = %d\n", __func__, error));
859 		return error;
860 	}
861 
862 	vp = nd.ni_vp;
863 	if (vp->v_type != VBLK) {
864 		error = ENOTBLK;
865 		goto out;
866 	}
867 
868 	/* Reopen as anonymous vnode to protect against forced unmount. */
869 	if ((error = bdevvp(vp->v_rdev, vpp)) != 0)
870 		goto out;
871 	VOP_UNLOCK(vp);
872 	if ((error = vn_close(vp, FREAD | FWRITE, l->l_cred)) != 0) {
873 		vrele(*vpp);
874 		return error;
875 	}
876 	if ((error = VOP_OPEN(*vpp, FREAD | FWRITE, l->l_cred)) != 0) {
877 		vrele(*vpp);
878 		return error;
879 	}
880 	mutex_enter((*vpp)->v_interlock);
881 	(*vpp)->v_writecount++;
882 	mutex_exit((*vpp)->v_interlock);
883 
884 	IFDEBUG(DKDB_VNODE, vprint("dk_lookup: vnode info", *vpp));
885 
886 	return 0;
887 out:
888 	VOP_UNLOCK(vp);
889 	(void) vn_close(vp, FREAD | FWRITE, l->l_cred);
890 	return error;
891 }
892 
893 MODULE(MODULE_CLASS_MISC, dk_subr, NULL);
894 
895 static int
896 dk_subr_modcmd(modcmd_t cmd, void *arg)
897 {
898 	switch (cmd) {
899 	case MODULE_CMD_INIT:
900 	case MODULE_CMD_FINI:
901 		return 0;
902 	case MODULE_CMD_STAT:
903 	case MODULE_CMD_AUTOUNLOAD:
904 	default:
905 		return ENOTTY;
906 	}
907 }
908