g_uzip.c revision 266220 
1/*-
2 * Copyright (c) 2004 Max Khon
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 *
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24 * SUCH DAMAGE.
25 */
26
27#include <sys/cdefs.h>
28__FBSDID("$FreeBSD: stable/10/sys/geom/uzip/g_uzip.c 266220 2014-05-16 14:28:55Z loos $");
29
30#include <sys/param.h>
31#include <sys/bio.h>
32#include <sys/endian.h>
33#include <sys/errno.h>
34#include <sys/kernel.h>
35#include <sys/lock.h>
36#include <sys/mutex.h>
37#include <sys/malloc.h>
38#include <sys/sysctl.h>
39#include <sys/systm.h>
40
41#include <geom/geom.h>
42#include <net/zlib.h>
43
44FEATURE(geom_uzip, "GEOM uzip read-only compressed disks support");
45
46#undef GEOM_UZIP_DEBUG
47#ifdef GEOM_UZIP_DEBUG
48#define	DPRINTF(a)	printf a
49#else
50#define	DPRINTF(a)
51#endif
52
53static MALLOC_DEFINE(M_GEOM_UZIP, "geom_uzip", "GEOM UZIP data structures");
54
55#define	UZIP_CLASS_NAME	"UZIP"
56
57/*
58 * Maximum allowed valid block size (to prevent foot-shooting)
59 */
60#define	MAX_BLKSZ	(MAXPHYS - MAXPHYS / 1000 - 12)
61
62/*
63 * Integer values (block size, number of blocks, offsets)
64 * are stored in big-endian (network) order on disk and struct cloop_header
65 * and in native order in struct g_uzip_softc
66 */
67
68#define	CLOOP_MAGIC_LEN	128
69static char CLOOP_MAGIC_START[] = "#!/bin/sh\n";
70
71struct cloop_header {
72	char magic[CLOOP_MAGIC_LEN];	/* cloop magic */
73	uint32_t blksz;			/* block size */
74	uint32_t nblocks;		/* number of blocks */
75};
76
77struct g_uzip_softc {
78	uint32_t blksz;			/* block size */
79	uint32_t nblocks;		/* number of blocks */
80	uint64_t *offsets;
81
82	struct mtx last_mtx;
83	uint32_t last_blk;		/* last blk no */
84	char *last_buf;			/* last blk data */
85	int req_total;			/* total requests */
86	int req_cached;			/* cached requests */
87};
88
89static void
90g_uzip_softc_free(struct g_uzip_softc *sc, struct g_geom *gp)
91{
92
93	if (gp != NULL) {
94		printf("%s: %d requests, %d cached\n",
95		    gp->name, sc->req_total, sc->req_cached);
96	}
97	if (sc->offsets != NULL) {
98		free(sc->offsets, M_GEOM_UZIP);
99		sc->offsets = NULL;
100	}
101	mtx_destroy(&sc->last_mtx);
102	free(sc->last_buf, M_GEOM_UZIP);
103	free(sc, M_GEOM_UZIP);
104}
105
106static void *
107z_alloc(void *nil, u_int type, u_int size)
108{
109	void *ptr;
110
111	ptr = malloc(type * size, M_GEOM_UZIP, M_NOWAIT);
112
113	return (ptr);
114}
115
116static void
117z_free(void *nil, void *ptr)
118{
119
120	free(ptr, M_GEOM_UZIP);
121}
122
123static void
124g_uzip_done(struct bio *bp)
125{
126	int err;
127	struct bio *bp2;
128	z_stream zs;
129	struct g_provider *pp, *pp2;
130	struct g_consumer *cp;
131	struct g_geom *gp;
132	struct g_uzip_softc *sc;
133	off_t iolen, pos, upos;
134	uint32_t start_blk, i;
135	size_t bsize;
136
137	bp2 = bp->bio_parent;
138	pp = bp2->bio_to;
139	gp = pp->geom;
140	cp = LIST_FIRST(&gp->consumer);
141	pp2 = cp->provider;
142	sc = gp->softc;
143	DPRINTF(("%s: done\n", gp->name));
144
145	bp2->bio_error = bp->bio_error;
146	if (bp2->bio_error != 0)
147		goto done;
148
149	/*
150	 * Uncompress data.
151	 */
152	zs.zalloc = z_alloc;
153	zs.zfree = z_free;
154	err = inflateInit(&zs);
155	if (err != Z_OK) {
156		bp2->bio_error = EIO;
157		goto done;
158	}
159	start_blk = bp2->bio_offset / sc->blksz;
160	bsize = pp2->sectorsize;
161	iolen = bp->bio_completed;
162	pos = sc->offsets[start_blk] % bsize;
163	upos = 0;
164	DPRINTF(("%s: done: start_blk %d, pos %jd, upos %jd, iolen %jd "
165	    "(%jd, %d, %zd)\n",
166	    gp->name, start_blk, (intmax_t)pos, (intmax_t)upos,
167	    (intmax_t)iolen, (intmax_t)bp2->bio_offset, sc->blksz, bsize));
168	for (i = start_blk; upos < bp2->bio_length; i++) {
169		off_t len, ulen, uoff;
170
171		uoff = i == start_blk ? bp2->bio_offset % sc->blksz : 0;
172		ulen = MIN(sc->blksz - uoff, bp2->bio_length - upos);
173		len = sc->offsets[i + 1] - sc->offsets[i];
174
175		if (len == 0) {
176			/* All zero block: no cache update */
177			bzero(bp2->bio_data + upos, ulen);
178			upos += ulen;
179			bp2->bio_completed += ulen;
180			continue;
181		}
182		if (len > iolen) {
183			DPRINTF(("%s: done: early termination: len (%jd) > "
184			    "iolen (%jd)\n",
185			    gp->name, (intmax_t)len, (intmax_t)iolen));
186			break;
187		}
188		zs.next_in = bp->bio_data + pos;
189		zs.avail_in = len;
190		zs.next_out = sc->last_buf;
191		zs.avail_out = sc->blksz;
192		mtx_lock(&sc->last_mtx);
193		err = inflate(&zs, Z_FINISH);
194		if (err != Z_STREAM_END) {
195			sc->last_blk = -1;
196			mtx_unlock(&sc->last_mtx);
197			DPRINTF(("%s: done: inflate failed (%jd + %jd -> %jd + %jd + %jd)\n",
198			    gp->name, (intmax_t)pos, (intmax_t)len,
199			    (intmax_t)uoff, (intmax_t)upos, (intmax_t)ulen));
200			inflateEnd(&zs);
201			bp2->bio_error = EIO;
202			goto done;
203		}
204		sc->last_blk = i;
205		DPRINTF(("%s: done: inflated %jd + %jd -> %jd + %jd + %jd\n",
206		    gp->name, (intmax_t)pos, (intmax_t)len, (intmax_t)uoff,
207		    (intmax_t)upos, (intmax_t)ulen));
208		memcpy(bp2->bio_data + upos, sc->last_buf + uoff, ulen);
209		mtx_unlock(&sc->last_mtx);
210
211		pos += len;
212		iolen -= len;
213		upos += ulen;
214		bp2->bio_completed += ulen;
215		err = inflateReset(&zs);
216		if (err != Z_OK) {
217			inflateEnd(&zs);
218			bp2->bio_error = EIO;
219			goto done;
220		}
221	}
222	err = inflateEnd(&zs);
223	if (err != Z_OK) {
224		bp2->bio_error = EIO;
225		goto done;
226	}
227
228done:
229	/*
230	 * Finish processing the request.
231	 */
232	DPRINTF(("%s: done: (%d, %jd, %ld)\n",
233	    gp->name, bp2->bio_error, (intmax_t)bp2->bio_completed,
234	    bp2->bio_resid));
235	free(bp->bio_data, M_GEOM_UZIP);
236	g_destroy_bio(bp);
237	g_io_deliver(bp2, bp2->bio_error);
238}
239
240static void
241g_uzip_start(struct bio *bp)
242{
243	struct bio *bp2;
244	struct g_provider *pp, *pp2;
245	struct g_geom *gp;
246	struct g_consumer *cp;
247	struct g_uzip_softc *sc;
248	uint32_t start_blk, end_blk;
249	size_t bsize;
250
251	pp = bp->bio_to;
252	gp = pp->geom;
253	DPRINTF(("%s: start (%d)\n", gp->name, bp->bio_cmd));
254
255	if (bp->bio_cmd != BIO_READ) {
256		g_io_deliver(bp, EOPNOTSUPP);
257		return;
258	}
259
260	cp = LIST_FIRST(&gp->consumer);
261	pp2 = cp->provider;
262	sc = gp->softc;
263
264	start_blk = bp->bio_offset / sc->blksz;
265	end_blk = (bp->bio_offset + bp->bio_length + sc->blksz - 1) / sc->blksz;
266	KASSERT(start_blk < sc->nblocks, ("start_blk out of range"));
267	KASSERT(end_blk <= sc->nblocks, ("end_blk out of range"));
268
269	sc->req_total++;
270	if (start_blk + 1 == end_blk) {
271		mtx_lock(&sc->last_mtx);
272		if (start_blk == sc->last_blk) {
273			off_t uoff;
274
275			uoff = bp->bio_offset % sc->blksz;
276			KASSERT(bp->bio_length <= sc->blksz - uoff,
277			    ("cached data error"));
278			memcpy(bp->bio_data, sc->last_buf + uoff,
279			    bp->bio_length);
280			sc->req_cached++;
281			mtx_unlock(&sc->last_mtx);
282
283			DPRINTF(("%s: start: cached 0 + %jd, %jd + 0 + %jd\n",
284			    gp->name, (intmax_t)bp->bio_length, (intmax_t)uoff,
285			    (intmax_t)bp->bio_length));
286			bp->bio_completed = bp->bio_length;
287			g_io_deliver(bp, 0);
288			return;
289		}
290		mtx_unlock(&sc->last_mtx);
291	}
292
293	bp2 = g_clone_bio(bp);
294	if (bp2 == NULL) {
295		g_io_deliver(bp, ENOMEM);
296		return;
297	}
298	bp2->bio_done = g_uzip_done;
299	DPRINTF(("%s: start (%d..%d), %s: %d + %jd, %s: %d + %jd\n",
300	    gp->name, start_blk, end_blk,
301	    pp->name, pp->sectorsize, (intmax_t)pp->mediasize,
302	    pp2->name, pp2->sectorsize, (intmax_t)pp2->mediasize));
303	bsize = pp2->sectorsize;
304	bp2->bio_offset = sc->offsets[start_blk] - sc->offsets[start_blk] % bsize;
305	while (1) {
306		bp2->bio_length = sc->offsets[end_blk] - bp2->bio_offset;
307		bp2->bio_length = (bp2->bio_length + bsize - 1) / bsize * bsize;
308		if (bp2->bio_length < MAXPHYS)
309			break;
310
311		end_blk--;
312		DPRINTF(("%s: bio_length (%jd) > MAXPHYS: lowering end_blk "
313		    "to %u\n", gp->name, (intmax_t)bp2->bio_length, end_blk));
314	}
315	DPRINTF(("%s: start %jd + %jd -> %ju + %ju -> %jd + %jd\n",
316	    gp->name,
317	    (intmax_t)bp->bio_offset, (intmax_t)bp->bio_length,
318	    (uintmax_t)sc->offsets[start_blk],
319	    (uintmax_t)sc->offsets[end_blk] - sc->offsets[start_blk],
320	    (intmax_t)bp2->bio_offset, (intmax_t)bp2->bio_length));
321	bp2->bio_data = malloc(bp2->bio_length, M_GEOM_UZIP, M_NOWAIT);
322	if (bp2->bio_data == NULL) {
323		g_destroy_bio(bp2);
324		g_io_deliver(bp, ENOMEM);
325		return;
326	}
327
328	g_io_request(bp2, cp);
329	DPRINTF(("%s: start ok\n", gp->name));
330}
331
332static void
333g_uzip_orphan(struct g_consumer *cp)
334{
335	struct g_geom *gp;
336
337	g_trace(G_T_TOPOLOGY, "%s(%p/%s)", __func__, cp, cp->provider->name);
338	g_topology_assert();
339
340	gp = cp->geom;
341	g_uzip_softc_free(gp->softc, gp);
342	gp->softc = NULL;
343	g_wither_geom(gp, ENXIO);
344}
345
346static int
347g_uzip_access(struct g_provider *pp, int dr, int dw, int de)
348{
349	struct g_geom *gp;
350	struct g_consumer *cp;
351
352	gp = pp->geom;
353	cp = LIST_FIRST(&gp->consumer);
354	KASSERT (cp != NULL, ("g_uzip_access but no consumer"));
355
356	if (cp->acw + dw > 0)
357		return (EROFS);
358
359	return (g_access(cp, dr, dw, de));
360}
361
362static void
363g_uzip_spoiled(struct g_consumer *cp)
364{
365	struct g_geom *gp;
366
367	gp = cp->geom;
368	g_trace(G_T_TOPOLOGY, "%s(%p/%s)", __func__, cp, gp->name);
369	g_topology_assert();
370
371	g_uzip_softc_free(gp->softc, gp);
372	gp->softc = NULL;
373	g_wither_geom(gp, ENXIO);
374}
375
376static struct g_geom *
377g_uzip_taste(struct g_class *mp, struct g_provider *pp, int flags)
378{
379	int error;
380	uint32_t i, total_offsets, offsets_read, blk;
381	void *buf;
382	struct cloop_header *header;
383	struct g_consumer *cp;
384	struct g_geom *gp;
385	struct g_provider *pp2;
386	struct g_uzip_softc *sc;
387
388	g_trace(G_T_TOPOLOGY, "%s(%s,%s)", __func__, mp->name, pp->name);
389	g_topology_assert();
390
391	/* Skip providers that are already open for writing. */
392	if (pp->acw > 0)
393		return (NULL);
394
395	buf = NULL;
396
397	/*
398	 * Create geom instance.
399	 */
400	gp = g_new_geomf(mp, "%s.uzip", pp->name);
401	cp = g_new_consumer(gp);
402	error = g_attach(cp, pp);
403	if (error == 0)
404		error = g_access(cp, 1, 0, 0);
405	if (error) {
406		g_detach(cp);
407		g_destroy_consumer(cp);
408		g_destroy_geom(gp);
409		return (NULL);
410	}
411	g_topology_unlock();
412
413	/*
414	 * Read cloop header, look for CLOOP magic, perform
415	 * other validity checks.
416	 */
417	DPRINTF(("%s: media sectorsize %u, mediasize %jd\n",
418	    gp->name, pp->sectorsize, (intmax_t)pp->mediasize));
419	buf = g_read_data(cp, 0, pp->sectorsize, NULL);
420	if (buf == NULL)
421		goto err;
422	header = (struct cloop_header *) buf;
423	if (strncmp(header->magic, CLOOP_MAGIC_START,
424	    sizeof(CLOOP_MAGIC_START) - 1) != 0) {
425		DPRINTF(("%s: no CLOOP magic\n", gp->name));
426		goto err;
427	}
428	if (header->magic[0x0b] != 'V' || header->magic[0x0c] < '2') {
429		DPRINTF(("%s: image version too old\n", gp->name));
430		goto err;
431	}
432
433	/*
434	 * Initialize softc and read offsets.
435	 */
436	sc = malloc(sizeof(*sc), M_GEOM_UZIP, M_WAITOK | M_ZERO);
437	gp->softc = sc;
438	sc->blksz = ntohl(header->blksz);
439	sc->nblocks = ntohl(header->nblocks);
440	if (sc->blksz % 512 != 0) {
441		printf("%s: block size (%u) should be multiple of 512.\n",
442		    gp->name, sc->blksz);
443		goto err;
444	}
445	if (sc->blksz > MAX_BLKSZ) {
446		printf("%s: block size (%u) should not be larger than %d.\n",
447		    gp->name, sc->blksz, MAX_BLKSZ);
448	}
449	total_offsets = sc->nblocks + 1;
450	if (sizeof(struct cloop_header) +
451	    total_offsets * sizeof(uint64_t) > pp->mediasize) {
452		printf("%s: media too small for %u blocks\n",
453		    gp->name, sc->nblocks);
454		goto err;
455	}
456	sc->offsets = malloc(
457	    total_offsets * sizeof(uint64_t), M_GEOM_UZIP, M_WAITOK);
458	offsets_read = MIN(total_offsets,
459	    (pp->sectorsize - sizeof(*header)) / sizeof(uint64_t));
460	for (i = 0; i < offsets_read; i++)
461		sc->offsets[i] = be64toh(((uint64_t *) (header + 1))[i]);
462	DPRINTF(("%s: %u offsets in the first sector\n",
463	       gp->name, offsets_read));
464	for (blk = 1; offsets_read < total_offsets; blk++) {
465		uint32_t nread;
466
467		free(buf, M_GEOM);
468		buf = g_read_data(
469		    cp, blk * pp->sectorsize, pp->sectorsize, NULL);
470		if (buf == NULL)
471			goto err;
472		nread = MIN(total_offsets - offsets_read,
473		     pp->sectorsize / sizeof(uint64_t));
474		DPRINTF(("%s: %u offsets read from sector %d\n",
475		    gp->name, nread, blk));
476		for (i = 0; i < nread; i++) {
477			sc->offsets[offsets_read + i] =
478			    be64toh(((uint64_t *) buf)[i]);
479		}
480		offsets_read += nread;
481	}
482	free(buf, M_GEOM);
483	DPRINTF(("%s: done reading offsets\n", gp->name));
484	mtx_init(&sc->last_mtx, "geom_uzip cache", NULL, MTX_DEF);
485	sc->last_blk = -1;
486	sc->last_buf = malloc(sc->blksz, M_GEOM_UZIP, M_WAITOK);
487	sc->req_total = 0;
488	sc->req_cached = 0;
489
490	g_topology_lock();
491	pp2 = g_new_providerf(gp, "%s", gp->name);
492	pp2->sectorsize = 512;
493	pp2->mediasize = (off_t)sc->nblocks * sc->blksz;
494	pp2->stripesize = pp->stripesize;
495	pp2->stripeoffset = pp->stripeoffset;
496	g_error_provider(pp2, 0);
497	g_access(cp, -1, 0, 0);
498
499	DPRINTF(("%s: taste ok (%d, %jd), (%d, %d), %x\n",
500	    gp->name,
501	    pp2->sectorsize, (intmax_t)pp2->mediasize,
502	    pp2->stripeoffset, pp2->stripesize, pp2->flags));
503	printf("%s: %u x %u blocks\n", gp->name, sc->nblocks, sc->blksz);
504	return (gp);
505
506err:
507	g_topology_lock();
508	g_access(cp, -1, 0, 0);
509	if (buf != NULL)
510		free(buf, M_GEOM);
511	if (gp->softc != NULL) {
512		g_uzip_softc_free(gp->softc, NULL);
513		gp->softc = NULL;
514	}
515	g_detach(cp);
516	g_destroy_consumer(cp);
517	g_destroy_geom(gp);
518
519	return (NULL);
520}
521
522static int
523g_uzip_destroy_geom(struct gctl_req *req, struct g_class *mp, struct g_geom *gp)
524{
525	struct g_provider *pp;
526
527	g_trace(G_T_TOPOLOGY, "%s(%s, %s)", __func__, mp->name, gp->name);
528	g_topology_assert();
529
530	if (gp->softc == NULL) {
531		printf("%s(%s): gp->softc == NULL\n", __func__, gp->name);
532		return (ENXIO);
533	}
534
535	KASSERT(gp != NULL, ("NULL geom"));
536	pp = LIST_FIRST(&gp->provider);
537	KASSERT(pp != NULL, ("NULL provider"));
538	if (pp->acr > 0 || pp->acw > 0 || pp->ace > 0)
539		return (EBUSY);
540
541	g_uzip_softc_free(gp->softc, gp);
542	gp->softc = NULL;
543	g_wither_geom(gp, ENXIO);
544
545	return (0);
546}
547
548static struct g_class g_uzip_class = {
549	.name = UZIP_CLASS_NAME,
550	.version = G_VERSION,
551	.taste = g_uzip_taste,
552	.destroy_geom = g_uzip_destroy_geom,
553
554	.start = g_uzip_start,
555	.orphan = g_uzip_orphan,
556	.access = g_uzip_access,
557	.spoiled = g_uzip_spoiled,
558};
559
560DECLARE_GEOM_CLASS(g_uzip_class, g_uzip);
561MODULE_DEPEND(g_uzip, zlib, 1, 1, 1);
562