• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Create a squashfs filesystem.  This is a highly compressed read only
3  * filesystem.
4  *
5  * Copyright (c) 2014
6  * Phillip Lougher <phillip@squashfs.org.uk>
7  *
8  * This program is free software; you can redistribute it and/or
9  * modify it under the terms of the GNU General Public License
10  * as published by the Free Software Foundation; either version 2,
11  * or (at your option) any later version.
12  *
13  * This program is distributed in the hope that it will be useful,
14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
16  * GNU General Public License for more details.
17  *
18  * You should have received a copy of the GNU General Public License
19  * along with this program; if not, write to the Free Software
20  * Foundation, 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
21  *
22  * process_fragments.c
23  */
24 
25 #include <pthread.h>
26 #include <sys/ioctl.h>
27 #include <unistd.h>
28 #include <signal.h>
29 #include <sys/time.h>
30 #include <string.h>
31 #include <stdio.h>
32 #include <math.h>
33 #include <stdarg.h>
34 #include <errno.h>
35 #include <stdlib.h>
36 #include <dirent.h>
37 #include <sys/types.h>
38 #include <sys/stat.h>
39 #include <fcntl.h>
40 
41 #include "caches-queues-lists.h"
42 #include "squashfs_fs.h"
43 #include "mksquashfs.h"
44 #include "error.h"
45 #include "progressbar.h"
46 #include "info.h"
47 #include "compressor.h"
48 #include "process_fragments.h"
49 
50 #define FALSE 0
51 #define TRUE 1
52 
53 extern struct queue *to_process_frag;
54 extern struct seq_queue *to_main;
55 extern int sparse_files;
56 
57 /*
58  * Compute 16 bit BSD checksum over the data, and check for sparseness
59  */
checksum_sparse(struct file_buffer * file_buffer)60 static int checksum_sparse(struct file_buffer *file_buffer)
61 {
62 	unsigned char *b = (unsigned char *) file_buffer->data;
63 	unsigned short chksum = 0;
64 	int bytes = file_buffer->size, sparse = TRUE, value;
65 
66 	while(bytes --) {
67 		chksum = (chksum & 1) ? (chksum >> 1) | 0x8000 : chksum >> 1;
68 		value = *b++;
69 		if(value) {
70 			sparse = FALSE;
71 			chksum += value;
72 		}
73 	}
74 
75 	file_buffer->checksum = chksum;
76 	return sparse;
77 }
78 
79 
read_filesystem(int fd,long long byte,int bytes,void * buff)80 static int read_filesystem(int fd, long long byte, int bytes, void *buff)
81 {
82 	off_t off = byte;
83 
84 	TRACE("read_filesystem: reading from position 0x%llx, bytes %d\n",
85 		byte, bytes);
86 
87 	if(lseek(fd, off, SEEK_SET) == -1) {
88 		ERROR("read_filesystem: Lseek on destination failed because %s, "
89 			"offset=0x%llx\n", strerror(errno), off);
90 		return 0;
91 	} else if(read_bytes(fd, buff, bytes) < bytes) {
92 		ERROR("Read on destination failed\n");
93 		return 0;
94 	}
95 
96 	return 1;
97 }
98 
99 
get_fragment(struct fragment * fragment,char * data_buffer,int fd)100 static struct file_buffer *get_fragment(struct fragment *fragment,
101 	char *data_buffer, int fd)
102 {
103 	struct squashfs_fragment_entry *disk_fragment;
104 	struct file_buffer *buffer, *compressed_buffer;
105 	long long start_block;
106 	int res, size, index = fragment->index;
107 	char locked;
108 
109 	/*
110 	 * Lookup fragment block in cache.
111 	 * If the fragment block doesn't exist, then get the compressed version
112 	 * from the writer cache or off disk, and decompress it.
113 	 *
114 	 * This routine has two things which complicate the code:
115 	 *
116 	 *	1. Multiple threads can simultaneously lookup/create the
117 	 *	   same buffer.  This means a buffer needs to be "locked"
118 	 *	   when it is being filled in, to prevent other threads from
119 	 *	   using it when it is not ready.  This is because we now do
120 	 *	   fragment duplicate checking in parallel.
121 	 *	2. We have two caches which need to be checked for the
122 	 *	   presence of fragment blocks: the normal fragment cache
123 	 *	   and a "reserve" cache.  The reserve cache is used to
124 	 *	   prevent an unnecessary pipeline stall when the fragment cache
125 	 *	   is full of fragments waiting to be compressed.
126 	 */
127 	pthread_cleanup_push((void *) pthread_mutex_unlock, &dup_mutex);
128 	pthread_mutex_lock(&dup_mutex);
129 
130 again:
131 	buffer = cache_lookup_nowait(fragment_buffer, index, &locked);
132 	if(buffer) {
133 		pthread_mutex_unlock(&dup_mutex);
134 		if(locked)
135 			/* got a buffer being filled in.  Wait for it */
136 			cache_wait_unlock(buffer);
137 		goto finished;
138 	}
139 
140 	/* not in fragment cache, is it in the reserve cache? */
141 	buffer = cache_lookup_nowait(reserve_cache, index, &locked);
142 	if(buffer) {
143 		pthread_mutex_unlock(&dup_mutex);
144 		if(locked)
145 			/* got a buffer being filled in.  Wait for it */
146 			cache_wait_unlock(buffer);
147 		goto finished;
148 	}
149 
150 	/* in neither cache, try to get it from the fragment cache */
151 	buffer = cache_get_nowait(fragment_buffer, index);
152 	if(!buffer) {
153 		/*
154 		 * no room, get it from the reserve cache, this is
155 		 * dimensioned so it will always have space (no more than
156 		 * processors + 1 can have an outstanding reserve buffer)
157 		 */
158 		buffer = cache_get_nowait(reserve_cache, index);
159 		if(!buffer) {
160 			/* failsafe */
161 			ERROR("no space in reserve cache\n");
162 			goto again;
163 		}
164 	}
165 
166 	pthread_mutex_unlock(&dup_mutex);
167 
168 	compressed_buffer = cache_lookup(fwriter_buffer, index);
169 
170 	pthread_cleanup_push((void *) pthread_mutex_unlock, &fragment_mutex);
171 	pthread_mutex_lock(&fragment_mutex);
172 	disk_fragment = &fragment_table[index];
173 	size = SQUASHFS_COMPRESSED_SIZE_BLOCK(disk_fragment->size);
174 	start_block = disk_fragment->start_block;
175 	pthread_cleanup_pop(1);
176 
177 	if(SQUASHFS_COMPRESSED_BLOCK(disk_fragment->size)) {
178 		int error;
179 		char *data;
180 
181 		if(compressed_buffer)
182 			data = compressed_buffer->data;
183 		else {
184 			res = read_filesystem(fd, start_block, size, data_buffer);
185 			if(res == 0) {
186 				ERROR("Failed to read fragment from output"
187 					" filesystem\n");
188 				BAD_ERROR("Output filesystem corrupted?\n");
189 			}
190 			data = data_buffer;
191 		}
192 
193 		res = compressor_uncompress(comp, buffer->data, data, size,
194 			block_size, &error);
195 		if(res == -1)
196 			BAD_ERROR("%s uncompress failed with error code %d\n",
197 				comp->name, error);
198 	} else if(compressed_buffer)
199 		memcpy(buffer->data, compressed_buffer->data, size);
200 	else {
201 		res = read_filesystem(fd, start_block, size, buffer->data);
202 		if(res == 0) {
203 			ERROR("Failed to read fragment from output "
204 				"filesystem\n");
205 			BAD_ERROR("Output filesystem corrupted?\n");
206 		}
207 	}
208 
209 	cache_unlock(buffer);
210 	cache_block_put(compressed_buffer);
211 
212 finished:
213 	pthread_cleanup_pop(0);
214 
215 	return buffer;
216 }
217 
218 
get_fragment_cksum(struct file_info * file,char * data_buffer,int fd,unsigned short * checksum)219 struct file_buffer *get_fragment_cksum(struct file_info *file,
220 	char *data_buffer, int fd, unsigned short *checksum)
221 {
222 	struct file_buffer *frag_buffer;
223 	struct append_file *append;
224 	int index = file->fragment->index;
225 
226 	frag_buffer = get_fragment(file->fragment, data_buffer, fd);
227 
228 	pthread_cleanup_push((void *) pthread_mutex_unlock, &dup_mutex);
229 
230 	for(append = file_mapping[index]; append; append = append->next) {
231 		int offset = append->file->fragment->offset;
232 		int size = append->file->fragment->size;
233 		char *data = frag_buffer->data + offset;
234 		unsigned short cksum = get_checksum_mem(data, size);
235 
236 		if(file == append->file)
237 			*checksum = cksum;
238 
239 		pthread_mutex_lock(&dup_mutex);
240 		append->file->fragment_checksum = cksum;
241 		append->file->have_frag_checksum = TRUE;
242 		pthread_mutex_unlock(&dup_mutex);
243 	}
244 
245 	pthread_cleanup_pop(0);
246 
247 	return frag_buffer;
248 }
249 
250 
frag_thrd(void * destination_file)251 void *frag_thrd(void *destination_file)
252 {
253 	sigset_t sigmask, old_mask;
254 	char *data_buffer;
255 	int fd;
256 
257 	sigemptyset(&sigmask);
258 	sigaddset(&sigmask, SIGINT);
259 	sigaddset(&sigmask, SIGTERM);
260 	sigaddset(&sigmask, SIGUSR1);
261 	pthread_sigmask(SIG_BLOCK, &sigmask, &old_mask);
262 
263 	fd = open(destination_file, O_RDONLY);
264 	if(fd == -1)
265 		BAD_ERROR("frag_thrd: can't open destination for reading\n");
266 
267 	data_buffer = malloc(SQUASHFS_FILE_MAX_SIZE);
268 	if(data_buffer == NULL)
269 		MEM_ERROR();
270 
271 	pthread_cleanup_push((void *) pthread_mutex_unlock, &dup_mutex);
272 
273 	while(1) {
274 		struct file_buffer *file_buffer = queue_get(to_process_frag);
275 		struct file_buffer *buffer;
276 		int sparse = checksum_sparse(file_buffer);
277 		struct file_info *dupl_ptr;
278 		long long file_size;
279 		unsigned short checksum;
280 		char flag;
281 		int res;
282 
283 		if(sparse_files && sparse) {
284 			file_buffer->c_byte = 0;
285 			file_buffer->fragment = FALSE;
286 		} else
287 			file_buffer->c_byte = file_buffer->size;
288 
289 		/*
290 		 * Specutively pull into the fragment cache any fragment blocks
291 		 * which contain fragments which *this* fragment may be
292 		 * be a duplicate.
293 		 *
294 		 * By ensuring the fragment block is in cache ahead of time
295 		 * should eliminate the parallelisation stall when the
296 		 * main thread needs to read the fragment block to do a
297 		 * duplicate check on it.
298 		 *
299 		 * If this is a fragment belonging to a larger file
300 		 * (with additional blocks) then ignore it.  Here we're
301 		 * interested in the "low hanging fruit" of files which
302 		 * consist of only a fragment
303 		 */
304 		if(file_buffer->file_size != file_buffer->size) {
305 			seq_queue_put(to_main, file_buffer);
306 			continue;
307 		}
308 
309 		file_size = file_buffer->file_size;
310 
311 		pthread_mutex_lock(&dup_mutex);
312 		dupl_ptr = dupl[DUP_HASH(file_size)];
313 		pthread_mutex_unlock(&dup_mutex);
314 
315 		file_buffer->dupl_start = dupl_ptr;
316 		file_buffer->duplicate = FALSE;
317 
318 		for(; dupl_ptr; dupl_ptr = dupl_ptr->next) {
319 			if(file_size != dupl_ptr->file_size ||
320 					file_size != dupl_ptr->fragment->size)
321 				continue;
322 
323 			pthread_mutex_lock(&dup_mutex);
324 			flag = dupl_ptr->have_frag_checksum;
325 			checksum = dupl_ptr->fragment_checksum;
326 			pthread_mutex_unlock(&dup_mutex);
327 
328 			/*
329 			 * If we have the checksum and it matches then
330 			 * read in the fragment block.
331 			 *
332 			 * If we *don't* have the checksum, then we are
333 			 * appending, and the fragment block is on the
334 			 * "old" filesystem.  Read it in and checksum
335 			 * the entire fragment buffer
336 			 */
337 			if(!flag) {
338 				buffer = get_fragment_cksum(dupl_ptr,
339 					data_buffer, fd, &checksum);
340 				if(checksum != file_buffer->checksum) {
341 					cache_block_put(buffer);
342 					continue;
343 				}
344 			} else if(checksum == file_buffer->checksum)
345 				buffer = get_fragment(dupl_ptr->fragment,
346 					data_buffer, fd);
347 			else
348 				continue;
349 
350 			res = memcmp(file_buffer->data, buffer->data +
351 				dupl_ptr->fragment->offset, file_size);
352 			cache_block_put(buffer);
353 			if(res == 0) {
354 				struct file_buffer *dup = malloc(sizeof(*dup));
355 				if(dup == NULL)
356 					MEM_ERROR();
357 				memcpy(dup, file_buffer, sizeof(*dup));
358 				cache_block_put(file_buffer);
359 				dup->dupl_start = dupl_ptr;
360 				dup->duplicate = TRUE;
361 				file_buffer = dup;
362 				break;
363 			}
364 		}
365 
366 		seq_queue_put(to_main, file_buffer);
367 	}
368 
369 	pthread_cleanup_pop(0);
370 }
371