disk.c revision 344295
1/*-
2 * Copyright (c) 1998 Michael Smith <msmith@freebsd.org>
3 * Copyright (c) 2012 Andrey V. Elsukov <ae@FreeBSD.org>
4 * All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
8 * are met:
9 * 1. Redistributions of source code must retain the above copyright
10 *    notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 *    notice, this list of conditions and the following disclaimer in the
13 *    documentation and/or other materials provided with the distribution.
14 *
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
16 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
18 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
21 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
25 * SUCH DAMAGE.
26 */
27
28#include <sys/cdefs.h>
29__FBSDID("$FreeBSD: stable/11/stand/common/disk.c 344295 2019-02-19 19:16:28Z kevans $");
30
31#include <sys/disk.h>
32#include <sys/queue.h>
33#include <stand.h>
34#include <stdarg.h>
35#include <bootstrap.h>
36#include <part.h>
37
38#include "disk.h"
39
40#ifdef DISK_DEBUG
41# define DEBUG(fmt, args...)	printf("%s: " fmt "\n" , __func__ , ## args)
42#else
43# define DEBUG(fmt, args...)
44#endif
45
46struct open_disk {
47	struct ptable		*table;
48	uint64_t		mediasize;
49	uint64_t		entrysize;
50	u_int			sectorsize;
51};
52
53struct print_args {
54	struct disk_devdesc	*dev;
55	const char		*prefix;
56	int			verbose;
57};
58
59/* Convert size to a human-readable number. */
60static char *
61display_size(uint64_t size, u_int sectorsize)
62{
63	static char buf[80];
64	char unit;
65
66	size = size * sectorsize / 1024;
67	unit = 'K';
68	if (size >= 10485760000LL) {
69		size /= 1073741824;
70		unit = 'T';
71	} else if (size >= 10240000) {
72		size /= 1048576;
73		unit = 'G';
74	} else if (size >= 10000) {
75		size /= 1024;
76		unit = 'M';
77	}
78	sprintf(buf, "%ld%cB", (long)size, unit);
79	return (buf);
80}
81
82int
83ptblread(void *d, void *buf, size_t blocks, uint64_t offset)
84{
85	struct disk_devdesc *dev;
86	struct open_disk *od;
87
88	dev = (struct disk_devdesc *)d;
89	od = (struct open_disk *)dev->dd.d_opendata;
90
91	/*
92	 * The strategy function assumes the offset is in units of 512 byte
93	 * sectors. For larger sector sizes, we need to adjust the offset to
94	 * match the actual sector size.
95	 */
96	offset *= (od->sectorsize / 512);
97	/*
98	 * As the GPT backup partition is located at the end of the disk,
99	 * to avoid reading past disk end, flag bcache not to use RA.
100	 */
101	return (dev->dd.d_dev->dv_strategy(dev, F_READ | F_NORA, offset,
102	    blocks * od->sectorsize, (char *)buf, NULL));
103}
104
105#define	PWIDTH	35
106static int
107ptable_print(void *arg, const char *pname, const struct ptable_entry *part)
108{
109	struct disk_devdesc dev;
110	struct print_args *pa, bsd;
111	struct open_disk *od;
112	struct ptable *table;
113	char line[80];
114	int res;
115
116	pa = (struct print_args *)arg;
117	od = (struct open_disk *)pa->dev->dd.d_opendata;
118	sprintf(line, "  %s%s: %s", pa->prefix, pname,
119	    parttype2str(part->type));
120	if (pa->verbose)
121		sprintf(line, "%-*s%s", PWIDTH, line,
122		    display_size(part->end - part->start + 1,
123		    od->sectorsize));
124	strcat(line, "\n");
125	if (pager_output(line))
126		return 1;
127	res = 0;
128	if (part->type == PART_FREEBSD) {
129		/* Open slice with BSD label */
130		dev.dd.d_dev = pa->dev->dd.d_dev;
131		dev.dd.d_unit = pa->dev->dd.d_unit;
132		dev.d_slice = part->index;
133		dev.d_partition = -1;
134		if (disk_open(&dev, part->end - part->start + 1,
135		    od->sectorsize) == 0) {
136			table = ptable_open(&dev, part->end - part->start + 1,
137			    od->sectorsize, ptblread);
138			if (table != NULL) {
139				sprintf(line, "  %s%s", pa->prefix, pname);
140				bsd.dev = pa->dev;
141				bsd.prefix = line;
142				bsd.verbose = pa->verbose;
143				res = ptable_iterate(table, &bsd, ptable_print);
144				ptable_close(table);
145			}
146			disk_close(&dev);
147		}
148	}
149
150	return (res);
151}
152#undef PWIDTH
153
154int
155disk_print(struct disk_devdesc *dev, char *prefix, int verbose)
156{
157	struct open_disk *od;
158	struct print_args pa;
159
160	/* Disk should be opened */
161	od = (struct open_disk *)dev->dd.d_opendata;
162	pa.dev = dev;
163	pa.prefix = prefix;
164	pa.verbose = verbose;
165	return (ptable_iterate(od->table, &pa, ptable_print));
166}
167
168int
169disk_read(struct disk_devdesc *dev, void *buf, uint64_t offset, u_int blocks)
170{
171	struct open_disk *od;
172	int ret;
173
174	od = (struct open_disk *)dev->dd.d_opendata;
175	ret = dev->dd.d_dev->dv_strategy(dev, F_READ, dev->d_offset + offset,
176	    blocks * od->sectorsize, buf, NULL);
177
178	return (ret);
179}
180
181int
182disk_write(struct disk_devdesc *dev, void *buf, uint64_t offset, u_int blocks)
183{
184	struct open_disk *od;
185	int ret;
186
187	od = (struct open_disk *)dev->dd.d_opendata;
188	ret = dev->dd.d_dev->dv_strategy(dev, F_WRITE, dev->d_offset + offset,
189	    blocks * od->sectorsize, buf, NULL);
190
191	return (ret);
192}
193
194int
195disk_ioctl(struct disk_devdesc *dev, u_long cmd, void *data)
196{
197	struct open_disk *od = dev->dd.d_opendata;
198
199	if (od == NULL)
200		return (ENOTTY);
201
202	switch (cmd) {
203	case DIOCGSECTORSIZE:
204		*(u_int *)data = od->sectorsize;
205		break;
206	case DIOCGMEDIASIZE:
207		if (dev->d_offset == 0)
208			*(uint64_t *)data = od->mediasize;
209		else
210			*(uint64_t *)data = od->entrysize * od->sectorsize;
211		break;
212	default:
213		return (ENOTTY);
214	}
215
216	return (0);
217}
218
219int
220disk_open(struct disk_devdesc *dev, uint64_t mediasize, u_int sectorsize)
221{
222	struct disk_devdesc partdev;
223	struct open_disk *od;
224	struct ptable *table;
225	struct ptable_entry part;
226	int rc, slice, partition;
227
228	rc = 0;
229	od = (struct open_disk *)malloc(sizeof(struct open_disk));
230	if (od == NULL) {
231		DEBUG("no memory");
232		return (ENOMEM);
233	}
234	dev->dd.d_opendata = od;
235	od->entrysize = 0;
236	od->mediasize = mediasize;
237	od->sectorsize = sectorsize;
238	/*
239	 * While we are reading disk metadata, make sure we do it relative
240	 * to the start of the disk
241	 */
242	memcpy(&partdev, dev, sizeof(partdev));
243	partdev.d_offset = 0;
244	partdev.d_slice = -1;
245	partdev.d_partition = -1;
246
247	dev->d_offset = 0;
248	table = NULL;
249	slice = dev->d_slice;
250	partition = dev->d_partition;
251
252	DEBUG("%s unit %d, slice %d, partition %d => %p",
253	    disk_fmtdev(dev), dev->dd.d_unit, dev->d_slice, dev->d_partition, od);
254
255	/* Determine disk layout. */
256	od->table = ptable_open(&partdev, mediasize / sectorsize, sectorsize,
257	    ptblread);
258	if (od->table == NULL) {
259		DEBUG("Can't read partition table");
260		rc = ENXIO;
261		goto out;
262	}
263
264	if (ptable_getsize(od->table, &mediasize) != 0) {
265		rc = ENXIO;
266		goto out;
267	}
268	if (mediasize > od->mediasize) {
269		od->mediasize = mediasize;
270	}
271
272	if (ptable_gettype(od->table) == PTABLE_BSD &&
273	    partition >= 0) {
274		/* It doesn't matter what value has d_slice */
275		rc = ptable_getpart(od->table, &part, partition);
276		if (rc == 0) {
277			dev->d_offset = part.start;
278			od->entrysize = part.end - part.start + 1;
279		}
280	} else if (ptable_gettype(od->table) == PTABLE_ISO9660) {
281		dev->d_offset = 0;
282		od->entrysize = mediasize;
283	} else if (slice >= 0) {
284		/* Try to get information about partition */
285		if (slice == 0)
286			rc = ptable_getbestpart(od->table, &part);
287		else
288			rc = ptable_getpart(od->table, &part, slice);
289		if (rc != 0) /* Partition doesn't exist */
290			goto out;
291		dev->d_offset = part.start;
292		od->entrysize = part.end - part.start + 1;
293		slice = part.index;
294		if (ptable_gettype(od->table) == PTABLE_GPT) {
295			partition = 255;
296			goto out; /* Nothing more to do */
297		} else if (partition == 255) {
298			/*
299			 * When we try to open GPT partition, but partition
300			 * table isn't GPT, reset d_partition value to -1
301			 * and try to autodetect appropriate value.
302			 */
303			partition = -1;
304		}
305		/*
306		 * If d_partition < 0 and we are looking at a BSD slice,
307		 * then try to read BSD label, otherwise return the
308		 * whole MBR slice.
309		 */
310		if (partition == -1 &&
311		    part.type != PART_FREEBSD)
312			goto out;
313		/* Try to read BSD label */
314		table = ptable_open(dev, part.end - part.start + 1,
315		    od->sectorsize, ptblread);
316		if (table == NULL) {
317			DEBUG("Can't read BSD label");
318			rc = ENXIO;
319			goto out;
320		}
321		/*
322		 * If slice contains BSD label and d_partition < 0, then
323		 * assume the 'a' partition. Otherwise just return the
324		 * whole MBR slice, because it can contain ZFS.
325		 */
326		if (partition < 0) {
327			if (ptable_gettype(table) != PTABLE_BSD)
328				goto out;
329			partition = 0;
330		}
331		rc = ptable_getpart(table, &part, partition);
332		if (rc != 0)
333			goto out;
334		dev->d_offset += part.start;
335		od->entrysize = part.end - part.start + 1;
336	}
337out:
338	if (table != NULL)
339		ptable_close(table);
340
341	if (rc != 0) {
342		if (od->table != NULL)
343			ptable_close(od->table);
344		free(od);
345		DEBUG("%s could not open", disk_fmtdev(dev));
346	} else {
347		/* Save the slice and partition number to the dev */
348		dev->d_slice = slice;
349		dev->d_partition = partition;
350		DEBUG("%s offset %lld => %p", disk_fmtdev(dev),
351		    (long long)dev->d_offset, od);
352	}
353	return (rc);
354}
355
356int
357disk_close(struct disk_devdesc *dev)
358{
359	struct open_disk *od;
360
361	od = (struct open_disk *)dev->dd.d_opendata;
362	DEBUG("%s closed => %p", disk_fmtdev(dev), od);
363	ptable_close(od->table);
364	free(od);
365	return (0);
366}
367
368char*
369disk_fmtdev(struct disk_devdesc *dev)
370{
371	static char buf[128];
372	char *cp;
373
374	cp = buf + sprintf(buf, "%s%d", dev->dd.d_dev->dv_name, dev->dd.d_unit);
375	if (dev->d_slice >= 0) {
376#ifdef LOADER_GPT_SUPPORT
377		if (dev->d_partition == 255) {
378			sprintf(cp, "p%d:", dev->d_slice);
379			return (buf);
380		} else
381#endif
382#ifdef LOADER_MBR_SUPPORT
383			cp += sprintf(cp, "s%d", dev->d_slice);
384#endif
385	}
386	if (dev->d_partition >= 0)
387		cp += sprintf(cp, "%c", dev->d_partition + 'a');
388	strcat(cp, ":");
389	return (buf);
390}
391
392int
393disk_parsedev(struct disk_devdesc *dev, const char *devspec, const char **path)
394{
395	int unit, slice, partition;
396	const char *np;
397	char *cp;
398
399	np = devspec;
400	unit = slice = partition = -1;
401	if (*np != '\0' && *np != ':') {
402		unit = strtol(np, &cp, 10);
403		if (cp == np)
404			return (EUNIT);
405#ifdef LOADER_GPT_SUPPORT
406		if (*cp == 'p') {
407			np = cp + 1;
408			slice = strtol(np, &cp, 10);
409			if (np == cp)
410				return (ESLICE);
411			/* we don't support nested partitions on GPT */
412			if (*cp != '\0' && *cp != ':')
413				return (EINVAL);
414			partition = 255;
415		} else
416#endif
417#ifdef LOADER_MBR_SUPPORT
418		if (*cp == 's') {
419			np = cp + 1;
420			slice = strtol(np, &cp, 10);
421			if (np == cp)
422				return (ESLICE);
423		}
424#endif
425		if (*cp != '\0' && *cp != ':') {
426			partition = *cp - 'a';
427			if (partition < 0)
428				return (EPART);
429			cp++;
430		}
431	} else
432		return (EINVAL);
433
434	if (*cp != '\0' && *cp != ':')
435		return (EINVAL);
436	dev->dd.d_unit = unit;
437	dev->d_slice = slice;
438	dev->d_partition = partition;
439	if (path != NULL)
440		*path = (*cp == '\0') ? cp: cp + 1;
441	return (0);
442}
443