xref: /freebsd/usr.bin/mkuzip/mkuzip.c (revision 7937bfbc0ca53fe7cdd0d54414f9296e273a518e)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2004-2016 Maxim Sobolev <sobomax@FreeBSD.org>
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  */
28 
29 #include <sys/types.h>
30 #include <sys/endian.h>
31 #include <sys/param.h>
32 #include <sys/sysctl.h>
33 #include <sys/stat.h>
34 #include <sys/uio.h>
35 #include <netinet/in.h>
36 #include <assert.h>
37 #include <ctype.h>
38 #include <err.h>
39 #include <fcntl.h>
40 #include <pthread.h>
41 #include <signal.h>
42 #include <stdint.h>
43 #include <stdio.h>
44 #include <stdlib.h>
45 #include <string.h>
46 #include <unistd.h>
47 
48 #include "mkuzip.h"
49 #include "mkuz_cloop.h"
50 #include "mkuz_blockcache.h"
51 #include "mkuz_lzma.h"
52 #include "mkuz_zlib.h"
53 #include "mkuz_zstd.h"
54 #include "mkuz_blk.h"
55 #include "mkuz_cfg.h"
56 #include "mkuz_conveyor.h"
57 #include "mkuz_format.h"
58 #include "mkuz_fqueue.h"
59 #include "mkuz_time.h"
60 #include "mkuz_insize.h"
61 
62 #define DEFAULT_CLSTSIZE	16384
63 
64 enum UZ_ALGORITHM {
65 	UZ_ZLIB = 0,
66 	UZ_LZMA,
67 	UZ_ZSTD,
68 	UZ_INVALID
69 };
70 
71 static const struct mkuz_format uzip_fmts[] = {
72 	[UZ_ZLIB] = {
73 		.option = "zlib",
74 		.magic = CLOOP_MAGIC_ZLIB,
75 		.default_sufx = DEFAULT_SUFX_ZLIB,
76 		.f_compress_bound = mkuz_zlib_cbound,
77 		.f_init = mkuz_zlib_init,
78 		.f_compress = mkuz_zlib_compress,
79 	},
80 	[UZ_LZMA] = {
81 		.option = "lzma",
82 		.magic = CLOOP_MAGIC_LZMA,
83 		.default_sufx = DEFAULT_SUFX_LZMA,
84 		.f_compress_bound = mkuz_lzma_cbound,
85 		.f_init = mkuz_lzma_init,
86 		.f_compress = mkuz_lzma_compress,
87 	},
88 	[UZ_ZSTD] = {
89 		.option = "zstd",
90 		.magic = CLOOP_MAGIC_ZSTD,
91 		.default_sufx = DEFAULT_SUFX_ZSTD,
92 		.f_compress_bound = mkuz_zstd_cbound,
93 		.f_init = mkuz_zstd_init,
94 		.f_compress = mkuz_zstd_compress,
95 	},
96 };
97 
98 static struct mkuz_blk *readblock(int, u_int32_t);
99 static void usage(void) __dead2;
100 static void cleanup(void);
101 
102 static char *cleanfile = NULL;
103 
104 static int
105 cmp_blkno(const struct mkuz_blk *bp, void *p)
106 {
107 	uint32_t *ap;
108 
109 	ap = (uint32_t *)p;
110 
111 	return (bp->info.blkno == *ap);
112 }
113 
114 int main(int argc, char **argv)
115 {
116 	struct mkuz_cfg cfs;
117 	char *oname;
118 	uint64_t *toc;
119 	int i, io, opt, tmp;
120 	struct {
121 		int en;
122 		FILE *f;
123 	} summary;
124 	struct iovec iov[2];
125 	uint64_t offset, last_offset;
126 	struct cloop_header hdr;
127 	struct mkuz_conveyor *cvp;
128 	struct mkuz_blk_info *chit;
129 	size_t ncpusz, ncpu, magiclen;
130 	double st, et;
131 	enum UZ_ALGORITHM comp_alg;
132 	int comp_level;
133 
134 	st = getdtime();
135 
136 	ncpusz = sizeof(size_t);
137 	if (sysctlbyname("hw.ncpu", &ncpu, &ncpusz, NULL, 0) < 0) {
138 		ncpu = 1;
139 	} else if (ncpu > MAX_WORKERS_AUTO) {
140 		ncpu = MAX_WORKERS_AUTO;
141 	}
142 
143 	memset(&hdr, 0, sizeof(hdr));
144 	cfs.blksz = DEFAULT_CLSTSIZE;
145 	oname = NULL;
146 	cfs.verbose = 0;
147 	cfs.no_zcomp = 0;
148 	cfs.en_dedup = 0;
149 	summary.en = 0;
150 	summary.f = stderr;
151 	comp_alg = UZ_ZLIB;
152 	comp_level = USE_DEFAULT_LEVEL;
153 	cfs.nworkers = ncpu;
154 	struct mkuz_blk *iblk, *oblk;
155 
156 	while((opt = getopt(argc, argv, "A:C:o:s:vZdLSj:")) != -1) {
157 		switch(opt) {
158 		case 'A':
159 			for (tmp = UZ_ZLIB; tmp < UZ_INVALID; tmp++) {
160 				if (strcmp(uzip_fmts[tmp].option, optarg) == 0)
161 					break;
162 			}
163 			if (tmp == UZ_INVALID)
164 				errx(1, "invalid algorithm specified: %s",
165 				    optarg);
166 				/* Not reached */
167 			comp_alg = tmp;
168 			break;
169 		case 'C':
170 			comp_level = atoi(optarg);
171 			break;
172 		case 'o':
173 			oname = optarg;
174 			break;
175 
176 		case 's':
177 			tmp = atoi(optarg);
178 			if (tmp <= 0) {
179 				errx(1, "invalid cluster size specified: %s",
180 				    optarg);
181 				/* Not reached */
182 			}
183 			cfs.blksz = tmp;
184 			break;
185 
186 		case 'v':
187 			cfs.verbose = 1;
188 			break;
189 
190 		case 'Z':
191 			cfs.no_zcomp = 1;
192 			break;
193 
194 		case 'd':
195 			cfs.en_dedup = 1;
196 			break;
197 
198 		case 'L':
199 			comp_alg = UZ_LZMA;
200 			break;
201 
202 		case 'S':
203 			summary.en = 1;
204 			summary.f = stdout;
205 			break;
206 
207 		case 'j':
208 			tmp = atoi(optarg);
209 			if (tmp <= 0) {
210 				errx(1, "invalid number of compression threads"
211                                     " specified: %s", optarg);
212 				/* Not reached */
213 			}
214 			cfs.nworkers = tmp;
215 			break;
216 
217 		default:
218 			usage();
219 			/* Not reached */
220 		}
221 	}
222 	argc -= optind;
223 	argv += optind;
224 
225 	if (argc != 1) {
226 		usage();
227 		/* Not reached */
228 	}
229 
230 	cfs.handler = &uzip_fmts[comp_alg];
231 
232 	magiclen = strlcpy(hdr.magic, cfs.handler->magic, sizeof(hdr.magic));
233 	assert(magiclen < sizeof(hdr.magic));
234 
235 	if (cfs.en_dedup != 0) {
236 		/*
237 		 * Dedupe requires a version 3 format.  Don't downgrade newer
238 		 * formats.
239 		 */
240 		if (hdr.magic[CLOOP_OFS_VERSN] == CLOOP_MAJVER_2)
241 			hdr.magic[CLOOP_OFS_VERSN] = CLOOP_MAJVER_3;
242 		hdr.magic[CLOOP_OFS_COMPR] =
243 		    tolower(hdr.magic[CLOOP_OFS_COMPR]);
244 	}
245 
246 	if (cfs.blksz % DEV_BSIZE != 0)
247 		errx(1, "cluster size should be multiple of %d", DEV_BSIZE);
248 
249 	cfs.cbound_blksz = cfs.handler->f_compress_bound(cfs.blksz);
250 	if (cfs.cbound_blksz > MAXPHYS)
251 		errx(1, "maximal compressed cluster size %zu greater than MAXPHYS %zu",
252 		    cfs.cbound_blksz, (size_t)MAXPHYS);
253 
254 	cfs.handler->f_init(&comp_level);
255 	cfs.comp_level = comp_level;
256 
257 	cfs.iname = argv[0];
258 	if (oname == NULL) {
259 		asprintf(&oname, "%s%s", cfs.iname, cfs.handler->default_sufx);
260 		if (oname == NULL) {
261 			err(1, "can't allocate memory");
262 			/* Not reached */
263 		}
264 	}
265 
266 	signal(SIGHUP, exit);
267 	signal(SIGINT, exit);
268 	signal(SIGTERM, exit);
269 	signal(SIGXCPU, exit);
270 	signal(SIGXFSZ, exit);
271 	atexit(cleanup);
272 
273 	cfs.fdr = open(cfs.iname, O_RDONLY);
274 	if (cfs.fdr < 0) {
275 		err(1, "open(%s)", cfs.iname);
276 		/* Not reached */
277 	}
278 	cfs.isize = mkuz_get_insize(&cfs);
279 	if (cfs.isize < 0) {
280 		errx(1, "can't determine input image size");
281 		/* Not reached */
282 	}
283 	hdr.nblocks = cfs.isize / cfs.blksz;
284 	if ((cfs.isize % cfs.blksz) != 0) {
285 		if (cfs.verbose != 0)
286 			fprintf(stderr, "file size is not multiple "
287 			"of %d, padding data\n", cfs.blksz);
288 		hdr.nblocks++;
289 	}
290 	toc = mkuz_safe_malloc((hdr.nblocks + 1) * sizeof(*toc));
291 
292 	/*
293 	 * Initialize last+1 entry with non-heap trash.  If final padding is
294 	 * added later, it may or may not be overwritten with an offset
295 	 * representing the length of the final compressed block.  If not,
296 	 * initialize to a defined value.
297 	 */
298 	toc[hdr.nblocks] = 0;
299 
300 	cfs.fdw = open(oname, (cfs.en_dedup ? O_RDWR : O_WRONLY) | O_TRUNC | O_CREAT,
301 		   S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
302 	if (cfs.fdw < 0) {
303 		err(1, "open(%s)", oname);
304 		/* Not reached */
305 	}
306 	cleanfile = oname;
307 
308 	/* Prepare header that we will write later when we have index ready. */
309 	iov[0].iov_base = (char *)&hdr;
310 	iov[0].iov_len = sizeof(hdr);
311 	iov[1].iov_base = (char *)toc;
312 	iov[1].iov_len = (hdr.nblocks + 1) * sizeof(*toc);
313 	offset = iov[0].iov_len + iov[1].iov_len;
314 
315 	/* Reserve space for header */
316 	lseek(cfs.fdw, offset, SEEK_SET);
317 
318 	if (cfs.verbose != 0) {
319 		fprintf(stderr, "data size %ju bytes, number of clusters "
320 		    "%u, index length %zu bytes\n", cfs.isize,
321 		    hdr.nblocks, iov[1].iov_len);
322 	}
323 
324 	cvp = mkuz_conveyor_ctor(&cfs);
325 
326 	last_offset = 0;
327         iblk = oblk = NULL;
328 	for(i = io = 0; iblk != MKUZ_BLK_EOF; i++) {
329 		iblk = readblock(cfs.fdr, cfs.blksz);
330 		mkuz_fqueue_enq(cvp->wrk_queue, iblk);
331 		if (iblk != MKUZ_BLK_EOF &&
332 		    (i < (cfs.nworkers * ITEMS_PER_WORKER))) {
333 			continue;
334 		}
335 drain:
336 		oblk = mkuz_fqueue_deq_when(cvp->results, cmp_blkno, &io);
337 		assert(oblk->info.blkno == (unsigned)io);
338 		oblk->info.offset = offset;
339 		chit = NULL;
340 		if (cfs.en_dedup != 0 && oblk->info.len > 0) {
341 			chit = mkuz_blkcache_regblock(cfs.fdw, oblk);
342 			/*
343 			 * There should be at least one non-empty block
344 			 * between us and the backref'ed offset, otherwise
345 			 * we won't be able to parse that sequence correctly
346 			 * as it would be indistinguishible from another
347 			 * empty block.
348 			 */
349 			if (chit != NULL && chit->offset == last_offset) {
350 				chit = NULL;
351 			}
352 		}
353 		if (chit != NULL) {
354 			toc[io] = htobe64(chit->offset);
355 			oblk->info.len = 0;
356 		} else {
357 			if (oblk->info.len > 0 && write(cfs.fdw, oblk->data,
358 			    oblk->info.len) < 0) {
359 				err(1, "write(%s)", oname);
360 				/* Not reached */
361 			}
362 			toc[io] = htobe64(offset);
363 			last_offset = offset;
364 			offset += oblk->info.len;
365 		}
366 		if (cfs.verbose != 0) {
367 			fprintf(stderr, "cluster #%d, in %u bytes, "
368 			    "out len=%lu offset=%lu", io, cfs.blksz,
369 			    (u_long)oblk->info.len, (u_long)be64toh(toc[io]));
370 			if (chit != NULL) {
371 				fprintf(stderr, " (backref'ed to #%d)",
372 				    chit->blkno);
373 			}
374 			fprintf(stderr, "\n");
375 		}
376 		free(oblk);
377 		io += 1;
378 		if (iblk == MKUZ_BLK_EOF) {
379 			if (io < i)
380 				goto drain;
381 			/* Last block, see if we need to add some padding */
382 			if ((offset % DEV_BSIZE) == 0)
383 				continue;
384 			oblk = mkuz_blk_ctor(DEV_BSIZE - (offset % DEV_BSIZE));
385 			oblk->info.blkno = io;
386 			oblk->info.len = oblk->alen;
387 			if (cfs.verbose != 0) {
388 				fprintf(stderr, "padding data with %lu bytes "
389 				    "so that file size is multiple of %d\n",
390 				    (u_long)oblk->alen, DEV_BSIZE);
391 			}
392 			mkuz_fqueue_enq(cvp->results, oblk);
393 			goto drain;
394 		}
395 	}
396 
397 	close(cfs.fdr);
398 
399 	if (cfs.verbose != 0 || summary.en != 0) {
400 		et = getdtime();
401 		fprintf(summary.f, "compressed data to %ju bytes, saved %lld "
402 		    "bytes, %.2f%% decrease, %.2f bytes/sec.\n", offset,
403 		    (long long)(cfs.isize - offset),
404 		    100.0 * (long long)(cfs.isize - offset) /
405 		    (float)cfs.isize, (float)cfs.isize / (et - st));
406 	}
407 
408 	/* Convert to big endian */
409 	hdr.blksz = htonl(cfs.blksz);
410 	hdr.nblocks = htonl(hdr.nblocks);
411 	/* Write headers into pre-allocated space */
412 	lseek(cfs.fdw, 0, SEEK_SET);
413 	if (writev(cfs.fdw, iov, 2) < 0) {
414 		err(1, "writev(%s)", oname);
415 		/* Not reached */
416 	}
417 	cleanfile = NULL;
418 	close(cfs.fdw);
419 
420 	exit(0);
421 }
422 
423 static struct mkuz_blk *
424 readblock(int fd, u_int32_t clstsize)
425 {
426 	int numread;
427 	struct mkuz_blk *rval;
428 	static int blockcnt;
429 	off_t cpos;
430 
431 	rval = mkuz_blk_ctor(clstsize);
432 
433 	rval->info.blkno = blockcnt;
434 	blockcnt += 1;
435 	cpos = lseek(fd, 0, SEEK_CUR);
436 	if (cpos < 0) {
437 		err(1, "readblock: lseek() failed");
438 		/* Not reached */
439 	}
440 	rval->info.offset = cpos;
441 
442 	numread = read(fd, rval->data, clstsize);
443 	if (numread < 0) {
444 		err(1, "readblock: read() failed");
445 		/* Not reached */
446 	}
447 	if (numread == 0) {
448 		free(rval);
449 		return MKUZ_BLK_EOF;
450 	}
451 	rval->info.len = numread;
452 	return rval;
453 }
454 
455 static void
456 usage(void)
457 {
458 
459 	fprintf(stderr, "usage: mkuzip [-vZdLS] [-o outfile] [-s cluster_size] "
460 	    "[-j ncompr] infile\n");
461 	exit(1);
462 }
463 
464 void *
465 mkuz_safe_malloc(size_t size)
466 {
467 	void *retval;
468 
469 	retval = malloc(size);
470 	if (retval == NULL) {
471 		err(1, "can't allocate memory");
472 		/* Not reached */
473 	}
474 	return retval;
475 }
476 
477 void *
478 mkuz_safe_zmalloc(size_t size)
479 {
480 	void *retval;
481 
482 	retval = mkuz_safe_malloc(size);
483 	bzero(retval, size);
484 	return retval;
485 }
486 
487 static void
488 cleanup(void)
489 {
490 
491 	if (cleanfile != NULL)
492 		unlink(cleanfile);
493 }
494 
495 int
496 mkuz_memvcmp(const void *memory, unsigned char val, size_t size)
497 {
498     const u_char *mm;
499 
500     mm = (const u_char *)memory;
501     return (*mm == val) && memcmp(mm, mm + 1, size - 1) == 0;
502 }
503