1220422Sgabor/*	$NetBSD: file.c,v 1.5 2011/02/16 18:35:39 joerg Exp $	*/
2220422Sgabor/*	$FreeBSD$	*/
3210389Sgabor/*	$OpenBSD: file.c,v 1.11 2010/07/02 20:48:48 nicm Exp $	*/
4210389Sgabor
5210389Sgabor/*-
6211496Sdes * Copyright (c) 1999 James Howard and Dag-Erling Co��dan Sm��rgrav
7211463Sgabor * Copyright (C) 2008-2010 Gabor Kovesdan <gabor@FreeBSD.org>
8211463Sgabor * Copyright (C) 2010 Dimitry Andric <dimitry@andric.com>
9210389Sgabor * All rights reserved.
10210389Sgabor *
11210389Sgabor * Redistribution and use in source and binary forms, with or without
12210389Sgabor * modification, are permitted provided that the following conditions
13210389Sgabor * are met:
14210389Sgabor * 1. Redistributions of source code must retain the above copyright
15210389Sgabor *    notice, this list of conditions and the following disclaimer.
16210389Sgabor * 2. Redistributions in binary form must reproduce the above copyright
17210389Sgabor *    notice, this list of conditions and the following disclaimer in the
18210389Sgabor *    documentation and/or other materials provided with the distribution.
19210389Sgabor *
20210389Sgabor * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
21210389Sgabor * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22210389Sgabor * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23210389Sgabor * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
24210389Sgabor * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25210389Sgabor * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26210389Sgabor * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27210389Sgabor * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28210389Sgabor * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29210389Sgabor * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30210389Sgabor * SUCH DAMAGE.
31210389Sgabor */
32210389Sgabor
33210389Sgabor#include <sys/cdefs.h>
34210389Sgabor__FBSDID("$FreeBSD$");
35210389Sgabor
36210389Sgabor#include <sys/param.h>
37226261Sgabor#include <sys/mman.h>
38226261Sgabor#include <sys/stat.h>
39210389Sgabor#include <sys/types.h>
40210389Sgabor
41210389Sgabor#include <err.h>
42210389Sgabor#include <errno.h>
43211463Sgabor#include <fcntl.h>
44226261Sgabor#include <lzma.h>
45211463Sgabor#include <stddef.h>
46210389Sgabor#include <stdlib.h>
47210389Sgabor#include <string.h>
48210389Sgabor#include <unistd.h>
49210389Sgabor#include <wchar.h>
50210389Sgabor#include <wctype.h>
51210389Sgabor#include <zlib.h>
52210389Sgabor
53226573Sgabor#ifndef WITHOUT_BZIP2
54226573Sgabor#include <bzlib.h>
55226573Sgabor#endif
56226573Sgabor
57210389Sgabor#include "grep.h"
58210389Sgabor
59211463Sgabor#define	MAXBUFSIZ	(32 * 1024)
60211463Sgabor#define	LNBUFBUMP	80
61210389Sgabor
62211463Sgaborstatic gzFile gzbufdesc;
63226573Sgaborstatic lzma_stream lstrm = LZMA_STREAM_INIT;
64226573Sgabor#ifndef WITHOUT_BZIP2
65211463Sgaborstatic BZFILE* bzbufdesc;
66226573Sgabor#endif
67210389Sgabor
68226261Sgaborstatic unsigned char *buffer;
69211463Sgaborstatic unsigned char *bufpos;
70211463Sgaborstatic size_t bufrem;
71226261Sgaborstatic size_t fsiz;
72210389Sgabor
73211463Sgaborstatic unsigned char *lnbuf;
74211463Sgaborstatic size_t lnbuflen;
75210389Sgabor
76211364Sgaborstatic inline int
77211463Sgaborgrep_refill(struct file *f)
78210389Sgabor{
79211463Sgabor	ssize_t nr;
80210389Sgabor
81226261Sgabor	if (filebehave == FILE_MMAP)
82226261Sgabor		return (0);
83226261Sgabor
84211463Sgabor	bufpos = buffer;
85211463Sgabor	bufrem = 0;
86211463Sgabor
87226573Sgabor	if (filebehave == FILE_GZIP) {
88211463Sgabor		nr = gzread(gzbufdesc, buffer, MAXBUFSIZ);
89226573Sgabor#ifndef WITHOUT_BZIP2
90226573Sgabor	} else if (filebehave == FILE_BZIP && bzbufdesc != NULL) {
91226573Sgabor		int bzerr;
92226573Sgabor
93211463Sgabor		nr = BZ2_bzRead(&bzerr, bzbufdesc, buffer, MAXBUFSIZ);
94211463Sgabor		switch (bzerr) {
95211463Sgabor		case BZ_OK:
96211463Sgabor		case BZ_STREAM_END:
97211463Sgabor			/* No problem, nr will be okay */
98211463Sgabor			break;
99211463Sgabor		case BZ_DATA_ERROR_MAGIC:
100211463Sgabor			/*
101211463Sgabor			 * As opposed to gzread(), which simply returns the
102211463Sgabor			 * plain file data, if it is not in the correct
103211463Sgabor			 * compressed format, BZ2_bzRead() instead aborts.
104211463Sgabor			 *
105211463Sgabor			 * So, just restart at the beginning of the file again,
106211463Sgabor			 * and use plain reads from now on.
107211463Sgabor			 */
108211463Sgabor			BZ2_bzReadClose(&bzerr, bzbufdesc);
109211463Sgabor			bzbufdesc = NULL;
110211463Sgabor			if (lseek(f->fd, 0, SEEK_SET) == -1)
111211463Sgabor				return (-1);
112211463Sgabor			nr = read(f->fd, buffer, MAXBUFSIZ);
113211463Sgabor			break;
114211463Sgabor		default:
115211463Sgabor			/* Make sure we exit with an error */
116211463Sgabor			nr = -1;
117211463Sgabor		}
118226573Sgabor#endif
119226261Sgabor	} else if ((filebehave == FILE_XZ) || (filebehave == FILE_LZMA)) {
120226261Sgabor		lzma_action action = LZMA_RUN;
121226261Sgabor		uint8_t in_buf[MAXBUFSIZ];
122226261Sgabor		lzma_ret ret;
123226261Sgabor
124226261Sgabor		ret = (filebehave == FILE_XZ) ?
125226261Sgabor		    lzma_stream_decoder(&lstrm, UINT64_MAX,
126226261Sgabor		    LZMA_CONCATENATED) :
127226261Sgabor		    lzma_alone_decoder(&lstrm, UINT64_MAX);
128226261Sgabor
129226261Sgabor		if (ret != LZMA_OK)
130226261Sgabor			return (-1);
131226261Sgabor
132226261Sgabor		lstrm.next_out = buffer;
133226261Sgabor		lstrm.avail_out = MAXBUFSIZ;
134226261Sgabor		lstrm.next_in = in_buf;
135226261Sgabor		nr = read(f->fd, in_buf, MAXBUFSIZ);
136226261Sgabor
137226261Sgabor		if (nr < 0)
138226261Sgabor			return (-1);
139226261Sgabor		else if (nr == 0)
140226261Sgabor			action = LZMA_FINISH;
141226261Sgabor
142226261Sgabor		lstrm.avail_in = nr;
143226261Sgabor		ret = lzma_code(&lstrm, action);
144226261Sgabor
145226261Sgabor		if (ret != LZMA_OK && ret != LZMA_STREAM_END)
146226261Sgabor			return (-1);
147226261Sgabor		bufrem = MAXBUFSIZ - lstrm.avail_out;
148226261Sgabor		return (0);
149211463Sgabor	} else
150211463Sgabor		nr = read(f->fd, buffer, MAXBUFSIZ);
151211463Sgabor
152211463Sgabor	if (nr < 0)
153211463Sgabor		return (-1);
154211463Sgabor
155211463Sgabor	bufrem = nr;
156211463Sgabor	return (0);
157210389Sgabor}
158210389Sgabor
159211364Sgaborstatic inline int
160211463Sgaborgrep_lnbufgrow(size_t newlen)
161210389Sgabor{
162210389Sgabor
163211463Sgabor	if (lnbuflen < newlen) {
164211463Sgabor		lnbuf = grep_realloc(lnbuf, newlen);
165211463Sgabor		lnbuflen = newlen;
166210389Sgabor	}
167211463Sgabor
168211463Sgabor	return (0);
169210389Sgabor}
170210389Sgabor
171210389Sgaborchar *
172211463Sgaborgrep_fgetln(struct file *f, size_t *lenp)
173210389Sgabor{
174211463Sgabor	unsigned char *p;
175211463Sgabor	char *ret;
176211463Sgabor	size_t len;
177211463Sgabor	size_t off;
178211463Sgabor	ptrdiff_t diff;
179210389Sgabor
180211463Sgabor	/* Fill the buffer, if necessary */
181211463Sgabor	if (bufrem == 0 && grep_refill(f) != 0)
182211463Sgabor		goto error;
183210389Sgabor
184211463Sgabor	if (bufrem == 0) {
185211463Sgabor		/* Return zero length to indicate EOF */
186211463Sgabor		*lenp = 0;
187211463Sgabor		return (bufpos);
188211463Sgabor	}
189210389Sgabor
190211463Sgabor	/* Look for a newline in the remaining part of the buffer */
191211463Sgabor	if ((p = memchr(bufpos, '\n', bufrem)) != NULL) {
192211463Sgabor		++p; /* advance over newline */
193211463Sgabor		ret = bufpos;
194211463Sgabor		len = p - bufpos;
195211463Sgabor		bufrem -= len;
196211463Sgabor		bufpos = p;
197211463Sgabor		*lenp = len;
198211463Sgabor		return (ret);
199210389Sgabor	}
200210389Sgabor
201211463Sgabor	/* We have to copy the current buffered data to the line buffer */
202211463Sgabor	for (len = bufrem, off = 0; ; len += bufrem) {
203211463Sgabor		/* Make sure there is room for more data */
204211463Sgabor		if (grep_lnbufgrow(len + LNBUFBUMP))
205211463Sgabor			goto error;
206211463Sgabor		memcpy(lnbuf + off, bufpos, len - off);
207211463Sgabor		off = len;
208211463Sgabor		if (grep_refill(f) != 0)
209211463Sgabor			goto error;
210211463Sgabor		if (bufrem == 0)
211211463Sgabor			/* EOF: return partial line */
212210389Sgabor			break;
213211463Sgabor		if ((p = memchr(bufpos, '\n', bufrem)) == NULL)
214211463Sgabor			continue;
215211463Sgabor		/* got it: finish up the line (like code above) */
216211463Sgabor		++p;
217211463Sgabor		diff = p - bufpos;
218211463Sgabor		len += diff;
219211463Sgabor		if (grep_lnbufgrow(len))
220211463Sgabor		    goto error;
221211463Sgabor		memcpy(lnbuf + off, bufpos, diff);
222211463Sgabor		bufrem -= diff;
223211463Sgabor		bufpos = p;
224211463Sgabor		break;
225210389Sgabor	}
226211463Sgabor	*lenp = len;
227210389Sgabor	return (lnbuf);
228211463Sgabor
229211463Sgaborerror:
230211463Sgabor	*lenp = 0;
231211463Sgabor	return (NULL);
232210389Sgabor}
233210389Sgabor
234226261Sgabor/*
235226261Sgabor * Opens a file for processing.
236226261Sgabor */
237226261Sgaborstruct file *
238226261Sgaborgrep_open(const char *path)
239210389Sgabor{
240226261Sgabor	struct file *f;
241210389Sgabor
242226261Sgabor	f = grep_malloc(sizeof *f);
243226261Sgabor	memset(f, 0, sizeof *f);
244226261Sgabor	if (path == NULL) {
245226261Sgabor		/* Processing stdin implies --line-buffered. */
246226261Sgabor		lbflag = true;
247226261Sgabor		f->fd = STDIN_FILENO;
248226261Sgabor	} else if ((f->fd = open(path, O_RDONLY)) == -1)
249226261Sgabor		goto error1;
250226261Sgabor
251226261Sgabor	if (filebehave == FILE_MMAP) {
252226261Sgabor		struct stat st;
253226261Sgabor
254226261Sgabor		if ((fstat(f->fd, &st) == -1) || (st.st_size > OFF_MAX) ||
255226261Sgabor		    (!S_ISREG(st.st_mode)))
256226261Sgabor			filebehave = FILE_STDIO;
257226261Sgabor		else {
258226261Sgabor			int flags = MAP_PRIVATE | MAP_NOCORE | MAP_NOSYNC;
259226261Sgabor#ifdef MAP_PREFAULT_READ
260226261Sgabor			flags |= MAP_PREFAULT_READ;
261226261Sgabor#endif
262226261Sgabor			fsiz = st.st_size;
263226261Sgabor			buffer = mmap(NULL, fsiz, PROT_READ, flags,
264226261Sgabor			     f->fd, (off_t)0);
265226261Sgabor			if (buffer == MAP_FAILED)
266226261Sgabor				filebehave = FILE_STDIO;
267226261Sgabor			else {
268226261Sgabor				bufrem = st.st_size;
269226261Sgabor				bufpos = buffer;
270226261Sgabor				madvise(buffer, st.st_size, MADV_SEQUENTIAL);
271226261Sgabor			}
272226261Sgabor		}
273226261Sgabor	}
274226261Sgabor
275226261Sgabor	if ((buffer == NULL) || (buffer == MAP_FAILED))
276226261Sgabor		buffer = grep_malloc(MAXBUFSIZ);
277226261Sgabor
278211463Sgabor	if (filebehave == FILE_GZIP &&
279211463Sgabor	    (gzbufdesc = gzdopen(f->fd, "r")) == NULL)
280226261Sgabor		goto error2;
281211364Sgabor
282226573Sgabor#ifndef WITHOUT_BZIP2
283211463Sgabor	if (filebehave == FILE_BZIP &&
284211463Sgabor	    (bzbufdesc = BZ2_bzdopen(f->fd, "r")) == NULL)
285226261Sgabor		goto error2;
286226573Sgabor#endif
287210389Sgabor
288211463Sgabor	/* Fill read buffer, also catches errors early */
289226261Sgabor	if (bufrem == 0 && grep_refill(f) != 0)
290226261Sgabor		goto error2;
291210389Sgabor
292211463Sgabor	/* Check for binary stuff, if necessary */
293211463Sgabor	if (binbehave != BINFILE_TEXT && memchr(bufpos, '\0', bufrem) != NULL)
294226261Sgabor	f->binary = true;
295210389Sgabor
296211463Sgabor	return (f);
297226261Sgabor
298226261Sgaborerror2:
299211463Sgabor	close(f->fd);
300226261Sgaborerror1:
301210389Sgabor	free(f);
302210389Sgabor	return (NULL);
303210389Sgabor}
304210389Sgabor
305210389Sgabor/*
306211463Sgabor * Closes a file.
307210389Sgabor */
308210389Sgaborvoid
309210389Sgaborgrep_close(struct file *f)
310210389Sgabor{
311210389Sgabor
312211463Sgabor	close(f->fd);
313210389Sgabor
314211463Sgabor	/* Reset read buffer and line buffer */
315226261Sgabor	if (filebehave == FILE_MMAP) {
316226261Sgabor		munmap(buffer, fsiz);
317226261Sgabor		buffer = NULL;
318226261Sgabor	}
319211463Sgabor	bufpos = buffer;
320211463Sgabor	bufrem = 0;
321211463Sgabor
322211463Sgabor	free(lnbuf);
323211463Sgabor	lnbuf = NULL;
324211463Sgabor	lnbuflen = 0;
325210389Sgabor}
326