g_uzip.c revision 265193 
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: head/sys/geom/uzip/g_uzip.c 265193 2014-05-01 14:47:27Z 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/systm.h>
39
40#include <geom/geom.h>
41#include <net/zlib.h>
42
43FEATURE(geom_uzip, "GEOM uzip read-only compressed disks support");
44
45#undef GEOM_UZIP_DEBUG
46#ifdef GEOM_UZIP_DEBUG
47#define	DPRINTF(a)	printf a
48#else
49#define	DPRINTF(a)
50#endif
51
52static MALLOC_DEFINE(M_GEOM_UZIP, "geom_uzip", "GEOM UZIP data structures");
53
54#define	UZIP_CLASS_NAME	"UZIP"
55
56/*
57 * Maximum allowed valid block size (to prevent foot-shooting)
58 */
59#define	MAX_BLKSZ	(MAXPHYS - MAXPHYS / 1000 - 12)
60
61/*
62 * Integer values (block size, number of blocks, offsets)
63 * are stored in big-endian (network) order on disk and struct cloop_header
64 * and in native order in struct g_uzip_softc
65 */
66
67#define	CLOOP_MAGIC_LEN	128
68static char CLOOP_MAGIC_START[] = "#!/bin/sh\n";
69
70struct cloop_header {
71	char magic[CLOOP_MAGIC_LEN];	/* cloop magic */
72	uint32_t blksz;			/* block size */
73	uint32_t nblocks;		/* number of blocks */
74};
75
76struct g_uzip_softc {
77	uint32_t blksz;			/* block size */
78	uint32_t nblocks;		/* number of blocks */
79	uint64_t *offsets;
80
81	struct mtx last_mtx;
82	uint32_t last_blk;		/* last blk no */
83	char *last_buf;			/* last blk data */
84	int req_total;			/* total requests */
85	int req_cached;			/* cached requests */
86};
87
88static void
89g_uzip_softc_free(struct g_uzip_softc *sc, struct g_geom *gp)
90{
91
92	if (gp != NULL) {
93		printf("%s: %d requests, %d cached\n",
94		    gp->name, sc->req_total, sc->req_cached);
95	}
96	if (sc->offsets != NULL) {
97		free(sc->offsets, M_GEOM_UZIP);
98		sc->offsets = NULL;
99	}
100	mtx_destroy(&sc->last_mtx);
101	free(sc->last_buf, M_GEOM_UZIP);
102	free(sc, M_GEOM_UZIP);
103}
104
105static void *
106z_alloc(void *nil, u_int type, u_int size)
107{
108	void *ptr;
109
110	ptr = malloc(type * size, M_GEOM_UZIP, M_NOWAIT);
111
112	return (ptr);
113}
114
115static void
116z_free(void *nil, void *ptr)
117{
118
119	free(ptr, M_GEOM_UZIP);
120}
121
122static void
123g_uzip_done(struct bio *bp)
124{
125	int err;
126	struct bio *bp2;
127	z_stream zs;
128	struct g_provider *pp, *pp2;
129	struct g_consumer *cp;
130	struct g_geom *gp;
131	struct g_uzip_softc *sc;
132	off_t iolen, pos, upos;
133	uint32_t start_blk, i;
134	size_t bsize;
135
136	bp2 = bp->bio_parent;
137	pp = bp2->bio_to;
138	gp = pp->geom;
139	cp = LIST_FIRST(&gp->consumer);
140	pp2 = cp->provider;
141	sc = gp->softc;
142	DPRINTF(("%s: done\n", gp->name));
143
144	bp2->bio_error = bp->bio_error;
145	if (bp2->bio_error != 0)
146		goto done;
147
148	/*
149	 * Uncompress data.
150	 */
151	zs.zalloc = z_alloc;
152	zs.zfree = z_free;
153	err = inflateInit(&zs);
154	if (err != Z_OK) {
155		bp2->bio_error = EIO;
156		goto done;
157	}
158	start_blk = bp2->bio_offset / sc->blksz;
159	bsize = pp2->sectorsize;
160	iolen = bp->bio_completed;
161	pos = sc->offsets[start_blk] % bsize;
162	upos = 0;
163	DPRINTF(("%s: done: start_blk %d, pos %jd, upos %jd, iolen %jd "
164	    "(%jd, %d, %zd)\n",
165	    gp->name, start_blk, (intmax_t)pos, (intmax_t)upos,
166	    (intmax_t)iolen, (intmax_t)bp2->bio_offset, sc->blksz, bsize));
167	for (i = start_blk; upos < bp2->bio_length; i++) {
168		off_t len, ulen, uoff;
169
170		uoff = i == start_blk ? bp2->bio_offset % sc->blksz : 0;
171		ulen = MIN(sc->blksz - uoff, bp2->bio_length - upos);
172		len = sc->offsets[i + 1] - sc->offsets[i];
173
174		if (len == 0) {
175			/* All zero block: no cache update */
176			bzero(bp2->bio_data + upos, ulen);
177			upos += ulen;
178			bp2->bio_completed += ulen;
179			continue;
180		}
181		if (len > iolen) {
182			DPRINTF(("%s: done: early termination: len (%jd) > "
183			    "iolen (%jd)\n",
184			    gp->name, (intmax_t)len, (intmax_t)iolen));
185			break;
186		}
187		zs.next_in = bp->bio_data + pos;
188		zs.avail_in = len;
189		zs.next_out = sc->last_buf;
190		zs.avail_out = sc->blksz;
191		mtx_lock(&sc->last_mtx);
192		err = inflate(&zs, Z_FINISH);
193		if (err != Z_STREAM_END) {
194			sc->last_blk = -1;
195			mtx_unlock(&sc->last_mtx);
196			DPRINTF(("%s: done: inflate failed (%jd + %jd -> %jd + %jd + %jd)\n",
197			    gp->name, (intmax_t)pos, (intmax_t)len,
198			    (intmax_t)uoff, (intmax_t)upos, (intmax_t)ulen));
199			inflateEnd(&zs);
200			bp2->bio_error = EIO;
201			goto done;
202		}
203		sc->last_blk = i;
204		DPRINTF(("%s: done: inflated %jd + %jd -> %jd + %jd + %jd\n",
205		    gp->name, (intmax_t)pos, (intmax_t)len, (intmax_t)uoff,
206		    (intmax_t)upos, (intmax_t)ulen));
207		memcpy(bp2->bio_data + upos, sc->last_buf + uoff, ulen);
208		mtx_unlock(&sc->last_mtx);
209
210		pos += len;
211		iolen -= len;
212		upos += ulen;
213		bp2->bio_completed += ulen;
214		err = inflateReset(&zs);
215		if (err != Z_OK) {
216			inflateEnd(&zs);
217			bp2->bio_error = EIO;
218			goto done;
219		}
220	}
221	err = inflateEnd(&zs);
222	if (err != Z_OK) {
223		bp2->bio_error = EIO;
224		goto done;
225	}
226
227done:
228	/*
229	 * Finish processing the request.
230	 */
231	DPRINTF(("%s: done: (%d, %jd, %ld)\n",
232	    gp->name, bp2->bio_error, (intmax_t)bp2->bio_completed,
233	    bp2->bio_resid));
234	free(bp->bio_data, M_GEOM_UZIP);
235	g_destroy_bio(bp);
236	g_io_deliver(bp2, bp2->bio_error);
237}
238
239static void
240g_uzip_start(struct bio *bp)
241{
242	struct bio *bp2;
243	struct g_provider *pp, *pp2;
244	struct g_geom *gp;
245	struct g_consumer *cp;
246	struct g_uzip_softc *sc;
247	uint32_t start_blk, end_blk;
248	size_t bsize;
249
250	pp = bp->bio_to;
251	gp = pp->geom;
252	DPRINTF(("%s: start (%d)\n", gp->name, bp->bio_cmd));
253
254	if (bp->bio_cmd != BIO_READ) {
255		g_io_deliver(bp, EOPNOTSUPP);
256		return;
257	}
258
259	cp = LIST_FIRST(&gp->consumer);
260	pp2 = cp->provider;
261	sc = gp->softc;
262
263	start_blk = bp->bio_offset / sc->blksz;
264	end_blk = (bp->bio_offset + bp->bio_length + sc->blksz - 1) / sc->blksz;
265	KASSERT(start_blk < sc->nblocks, ("start_blk out of range"));
266	KASSERT(end_blk <= sc->nblocks, ("end_blk out of range"));
267
268	sc->req_total++;
269	if (start_blk + 1 == end_blk) {
270		mtx_lock(&sc->last_mtx);
271		if (start_blk == sc->last_blk) {
272			off_t uoff;
273
274			uoff = bp->bio_offset % sc->blksz;
275			KASSERT(bp->bio_length <= sc->blksz - uoff,
276			    ("cached data error"));
277			memcpy(bp->bio_data, sc->last_buf + uoff,
278			    bp->bio_length);
279			sc->req_cached++;
280			mtx_unlock(&sc->last_mtx);
281
282			DPRINTF(("%s: start: cached 0 + %jd, %jd + 0 + %jd\n",
283			    gp->name, (intmax_t)bp->bio_length, (intmax_t)uoff,
284			    (intmax_t)bp->bio_length));
285			bp->bio_completed = bp->bio_length;
286			g_io_deliver(bp, 0);
287			return;
288		}
289		mtx_unlock(&sc->last_mtx);
290	}
291
292	bp2 = g_clone_bio(bp);
293	if (bp2 == NULL) {
294		g_io_deliver(bp, ENOMEM);
295		return;
296	}
297	bp2->bio_done = g_uzip_done;
298	DPRINTF(("%s: start (%d..%d), %s: %d + %jd, %s: %d + %jd\n",
299	    gp->name, start_blk, end_blk,
300	    pp->name, pp->sectorsize, (intmax_t)pp->mediasize,
301	    pp2->name, pp2->sectorsize, (intmax_t)pp2->mediasize));
302	bsize = pp2->sectorsize;
303	bp2->bio_offset = sc->offsets[start_blk] - sc->offsets[start_blk] % bsize;
304	while (1) {
305		bp2->bio_length = sc->offsets[end_blk] - bp2->bio_offset;
306		bp2->bio_length = (bp2->bio_length + bsize - 1) / bsize * bsize;
307		if (bp2->bio_length < MAXPHYS)
308			break;
309
310		end_blk--;
311		DPRINTF(("%s: bio_length (%jd) > MAXPHYS: lowering end_blk "
312		    "to %u\n", gp->name, (intmax_t)bp2->bio_length, end_blk));
313	}
314	DPRINTF(("%s: start %jd + %jd -> %ju + %ju -> %jd + %jd\n",
315	    gp->name,
316	    (intmax_t)bp->bio_offset, (intmax_t)bp->bio_length,
317	    (uintmax_t)sc->offsets[start_blk],
318	    (uintmax_t)sc->offsets[end_blk] - sc->offsets[start_blk],
319	    (intmax_t)bp2->bio_offset, (intmax_t)bp2->bio_length));
320	bp2->bio_data = malloc(bp2->bio_length, M_GEOM_UZIP, M_NOWAIT);
321	if (bp2->bio_data == NULL) {
322		g_destroy_bio(bp2);
323		g_io_deliver(bp, ENOMEM);
324		return;
325	}
326
327	g_io_request(bp2, cp);
328	DPRINTF(("%s: start ok\n", gp->name));
329}
330
331static void
332g_uzip_orphan(struct g_consumer *cp)
333{
334	struct g_geom *gp;
335
336	g_trace(G_T_TOPOLOGY, "%s(%p/%s)", __func__, cp, cp->provider->name);
337	g_topology_assert();
338
339	gp = cp->geom;
340	g_uzip_softc_free(gp->softc, gp);
341	gp->softc = NULL;
342	g_wither_geom(gp, ENXIO);
343}
344
345static int
346g_uzip_access(struct g_provider *pp, int dr, int dw, int de)
347{
348	struct g_geom *gp;
349	struct g_consumer *cp;
350
351	gp = pp->geom;
352	cp = LIST_FIRST(&gp->consumer);
353	KASSERT (cp != NULL, ("g_uzip_access but no consumer"));
354
355	if (cp->acw + dw > 0)
356		return (EROFS);
357
358	return (g_access(cp, dr, dw, de));
359}
360
361static void
362g_uzip_spoiled(struct g_consumer *cp)
363{
364	struct g_geom *gp;
365
366	gp = cp->geom;
367	g_trace(G_T_TOPOLOGY, "%s(%p/%s)", __func__, cp, gp->name);
368	g_topology_assert();
369
370	g_uzip_softc_free(gp->softc, gp);
371	gp->softc = NULL;
372	g_wither_geom(gp, ENXIO);
373}
374
375static struct g_geom *
376g_uzip_taste(struct g_class *mp, struct g_provider *pp, int flags)
377{
378	int error;
379	uint32_t i, total_offsets, offsets_read, blk;
380	void *buf;
381	struct cloop_header *header;
382	struct g_consumer *cp;
383	struct g_geom *gp;
384	struct g_provider *pp2;
385	struct g_uzip_softc *sc;
386
387	g_trace(G_T_TOPOLOGY, "%s(%s,%s)", __func__, mp->name, pp->name);
388	g_topology_assert();
389
390	/* Skip providers that are already open for writing. */
391	if (pp->acw > 0)
392		return (NULL);
393
394	buf = NULL;
395
396	/*
397	 * Create geom instance.
398	 */
399	gp = g_new_geomf(mp, "%s.uzip", pp->name);
400	cp = g_new_consumer(gp);
401	error = g_attach(cp, pp);
402	if (error == 0)
403		error = g_access(cp, 1, 0, 0);
404	if (error) {
405		g_detach(cp);
406		g_destroy_consumer(cp);
407		g_destroy_geom(gp);
408		return (NULL);
409	}
410	g_topology_unlock();
411
412	/*
413	 * Read cloop header, look for CLOOP magic, perform
414	 * other validity checks.
415	 */
416	DPRINTF(("%s: media sectorsize %u, mediasize %jd\n",
417	    gp->name, pp->sectorsize, (intmax_t)pp->mediasize));
418	buf = g_read_data(cp, 0, pp->sectorsize, NULL);
419	if (buf == NULL)
420		goto err;
421	header = (struct cloop_header *) buf;
422	if (strncmp(header->magic, CLOOP_MAGIC_START,
423	    sizeof(CLOOP_MAGIC_START) - 1) != 0) {
424		DPRINTF(("%s: no CLOOP magic\n", gp->name));
425		goto err;
426	}
427	if (header->magic[0x0b] != 'V' || header->magic[0x0c] < '2') {
428		DPRINTF(("%s: image version too old\n", gp->name));
429		goto err;
430	}
431
432	/*
433	 * Initialize softc and read offsets.
434	 */
435	sc = malloc(sizeof(*sc), M_GEOM_UZIP, M_WAITOK | M_ZERO);
436	gp->softc = sc;
437	sc->blksz = ntohl(header->blksz);
438	sc->nblocks = ntohl(header->nblocks);
439	if (sc->blksz % 512 != 0) {
440		printf("%s: block size (%u) should be multiple of 512.\n",
441		    gp->name, sc->blksz);
442		goto err;
443	}
444	if (sc->blksz > MAX_BLKSZ) {
445		printf("%s: block size (%u) should not be larger than %d.\n",
446		    gp->name, sc->blksz, MAX_BLKSZ);
447	}
448	total_offsets = sc->nblocks + 1;
449	if (sizeof(struct cloop_header) +
450	    total_offsets * sizeof(uint64_t) > pp->mediasize) {
451		printf("%s: media too small for %u blocks\n",
452		    gp->name, sc->nblocks);
453		goto err;
454	}
455	sc->offsets = malloc(
456	    total_offsets * sizeof(uint64_t), M_GEOM_UZIP, M_WAITOK);
457	offsets_read = MIN(total_offsets,
458	    (pp->sectorsize - sizeof(*header)) / sizeof(uint64_t));
459	for (i = 0; i < offsets_read; i++)
460		sc->offsets[i] = be64toh(((uint64_t *) (header + 1))[i]);
461	DPRINTF(("%s: %u offsets in the first sector\n",
462	       gp->name, offsets_read));
463	for (blk = 1; offsets_read < total_offsets; blk++) {
464		uint32_t nread;
465
466		free(buf, M_GEOM);
467		buf = g_read_data(
468		    cp, blk * pp->sectorsize, pp->sectorsize, NULL);
469		if (buf == NULL)
470			goto err;
471		nread = MIN(total_offsets - offsets_read,
472		     pp->sectorsize / sizeof(uint64_t));
473		DPRINTF(("%s: %u offsets read from sector %d\n",
474		    gp->name, nread, blk));
475		for (i = 0; i < nread; i++) {
476			sc->offsets[offsets_read + i] =
477			    be64toh(((uint64_t *) buf)[i]);
478		}
479		offsets_read += nread;
480	}
481	DPRINTF(("%s: done reading offsets\n", gp->name));
482	mtx_init(&sc->last_mtx, "geom_uzip cache", NULL, MTX_DEF);
483	sc->last_blk = -1;
484	sc->last_buf = malloc(sc->blksz, M_GEOM_UZIP, M_WAITOK);
485	sc->req_total = 0;
486	sc->req_cached = 0;
487
488	g_topology_lock();
489	pp2 = g_new_providerf(gp, "%s", gp->name);
490	pp2->sectorsize = 512;
491	pp2->mediasize = (off_t)sc->nblocks * sc->blksz;
492	pp2->stripesize = pp->stripesize;
493	pp2->stripeoffset = pp->stripeoffset;
494	g_error_provider(pp2, 0);
495	g_access(cp, -1, 0, 0);
496
497	DPRINTF(("%s: taste ok (%d, %jd), (%d, %d), %x\n",
498	    gp->name,
499	    pp2->sectorsize, (intmax_t)pp2->mediasize,
500	    pp2->stripeoffset, pp2->stripesize, pp2->flags));
501	printf("%s: %u x %u blocks\n", gp->name, sc->nblocks, sc->blksz);
502	return (gp);
503
504err:
505	g_topology_lock();
506	g_access(cp, -1, 0, 0);
507	if (buf != NULL)
508		free(buf, M_GEOM);
509	if (gp->softc != NULL) {
510		g_uzip_softc_free(gp->softc, NULL);
511		gp->softc = NULL;
512	}
513	g_detach(cp);
514	g_destroy_consumer(cp);
515	g_destroy_geom(gp);
516
517	return (NULL);
518}
519
520static int
521g_uzip_destroy_geom(struct gctl_req *req, struct g_class *mp, struct g_geom *gp)
522{
523	struct g_provider *pp;
524
525	g_trace(G_T_TOPOLOGY, "%s(%s, %s)", __func__, mp->name, gp->name);
526	g_topology_assert();
527
528	if (gp->softc == NULL) {
529		printf("%s(%s): gp->softc == NULL\n", __func__, gp->name);
530		return (ENXIO);
531	}
532
533	KASSERT(gp != NULL, ("NULL geom"));
534	pp = LIST_FIRST(&gp->provider);
535	KASSERT(pp != NULL, ("NULL provider"));
536	if (pp->acr > 0 || pp->acw > 0 || pp->ace > 0)
537		return (EBUSY);
538
539	g_uzip_softc_free(gp->softc, gp);
540	gp->softc = NULL;
541	g_wither_geom(gp, ENXIO);
542
543	return (0);
544}
545
546static struct g_class g_uzip_class = {
547	.name = UZIP_CLASS_NAME,
548	.version = G_VERSION,
549	.taste = g_uzip_taste,
550	.destroy_geom = g_uzip_destroy_geom,
551
552	.start = g_uzip_start,
553	.orphan = g_uzip_orphan,
554	.access = g_uzip_access,
555	.spoiled = g_uzip_spoiled,
556};
557
558DECLARE_GEOM_CLASS(g_uzip_class, g_uzip);
559MODULE_DEPEND(g_uzip, zlib, 1, 1, 1);
560