diff options
author | Mark Adler <madler@alumni.caltech.edu> | 2011-09-09 23:26:40 -0700 |
---|---|---|
committer | Mark Adler <madler@alumni.caltech.edu> | 2011-09-09 23:26:40 -0700 |
commit | f6194ef39af5864f792412460c354cc339dde7d1 (patch) | |
tree | 5ea1e6849128e9b2194c66ee3d82afa36b4ac07c /examples/gzlog.c | |
parent | 639be997883d9016baaf46017a2802b2ce1698bd (diff) | |
download | zlib-1.2.3.4.tar.gz zlib-1.2.3.4.tar.bz2 zlib-1.2.3.4.zip |
zlib 1.2.3.4v1.2.3.4
Diffstat (limited to 'examples/gzlog.c')
-rw-r--r-- | examples/gzlog.c | 1303 |
1 files changed, 974 insertions, 329 deletions
diff --git a/examples/gzlog.c b/examples/gzlog.c index b6acdef..4daf1c2 100644 --- a/examples/gzlog.c +++ b/examples/gzlog.c | |||
@@ -1,413 +1,1058 @@ | |||
1 | /* | 1 | /* |
2 | * gzlog.c | 2 | * gzlog.c |
3 | * Copyright (C) 2004 Mark Adler | 3 | * Copyright (C) 2004, 2008 Mark Adler, all rights reserved |
4 | * For conditions of distribution and use, see copyright notice in gzlog.h | 4 | * For conditions of distribution and use, see copyright notice in gzlog.h |
5 | * version 1.0, 26 Nov 2004 | 5 | * version 2.0, 25 Apr 2008 |
6 | * | ||
7 | */ | 6 | */ |
8 | 7 | ||
9 | #include <string.h> /* memcmp() */ | 8 | /* |
10 | #include <stdlib.h> /* malloc(), free(), NULL */ | 9 | gzlog provides a mechanism for frequently appending short strings to a gzip |
11 | #include <sys/types.h> /* size_t, off_t */ | 10 | file that is efficient both in execution time and compression ratio. The |
12 | #include <unistd.h> /* read(), close(), sleep(), ftruncate(), */ | 11 | strategy is to write the short strings in an uncompressed form to the end of |
13 | /* lseek() */ | 12 | the gzip file, only compressing when the amount of uncompressed data has |
14 | #include <fcntl.h> /* open() */ | 13 | reached a given threshold. |
15 | #include <sys/file.h> /* flock() */ | 14 | |
16 | #include "zlib.h" /* deflateInit2(), deflate(), deflateEnd() */ | 15 | gzlog also provides protection against interruptions in the process due to |
16 | system crashes. The status of the operation is recorded in an extra field | ||
17 | in the gzip file, and is only updated once the gzip file is brought to a | ||
18 | valid state. The last data to be appended or compressed is saved in an | ||
19 | auxiliary file, so that if the operation is interrupted, it can be completed | ||
20 | the next time an append operation is attempted. | ||
21 | |||
22 | gzlog maintains another auxiliary file with the last 32K of data from the | ||
23 | compressed portion, which is preloaded for the compression of the subsequent | ||
24 | data. This minimizes the impact to the compression ratio of appending. | ||
25 | */ | ||
26 | |||
27 | /* | ||
28 | Operations Concept: | ||
29 | |||
30 | Files (log name "foo"): | ||
31 | foo.gz -- gzip file with the complete log | ||
32 | foo.add -- last message to append or last data to compress | ||
33 | foo.dict -- dictionary of the last 32K of data for next compression | ||
34 | foo.temp -- temporary dictionary file for compression after this one | ||
35 | foo.lock -- lock file for reading and writing the other files | ||
36 | foo.repairs -- log file for log file recovery operations (not compressed) | ||
37 | |||
38 | gzip file structure: | ||
39 | - fixed-length (no file name) header with extra field (see below) | ||
40 | - compressed data ending initially with empty stored block | ||
41 | - uncompressed data filling out originally empty stored block and | ||
42 | subsequent stored blocks as needed (16K max each) | ||
43 | - gzip trailer | ||
44 | - no junk at end (no other gzip streams) | ||
45 | |||
46 | When appending data, the information in the first three items above plus the | ||
47 | foo.add file are sufficient to recover an interrupted append operation. The | ||
48 | extra field has the necessary information to restore the start of the last | ||
49 | stored block and determine where to append the data in the foo.add file, as | ||
50 | well as the crc and length of the gzip data before the append operation. | ||
51 | |||
52 | The foo.add file is created before the gzip file is marked for append, and | ||
53 | deleted after the gzip file is marked as complete. So if the append | ||
54 | operation is interrupted, the data to add will still be there. If due to | ||
55 | some external force, the foo.add file gets deleted between when the append | ||
56 | operation was interrupted and when recovery is attempted, the gzip file will | ||
57 | still be restored, but without the appended data. | ||
58 | |||
59 | When compressing data, the information in the first two items above plus the | ||
60 | foo.add file are sufficient to recover an interrupted compress operation. | ||
61 | The extra field has the necessary information to find the end of the | ||
62 | compressed data, and contains both the crc and length of just the compressed | ||
63 | data and of the complete set of data including the contents of the foo.add | ||
64 | file. | ||
65 | |||
66 | Again, the foo.add file is maintained during the compress operation in case | ||
67 | of an interruption. If in the unlikely event the foo.add file with the data | ||
68 | to be compressed is missing due to some external force, a gzip file with | ||
69 | just the previous compressed data will be reconstructed. In this case, all | ||
70 | of the data that was to be compressed is lost (approximately one megabyte). | ||
71 | This will not occur if all that happened was an interruption of the compress | ||
72 | operation. | ||
73 | |||
74 | The third state that is marked is the replacement of the old dictionary with | ||
75 | the new dictionary after a compress operation. Once compression is | ||
76 | complete, the gzip file is marked as being in the replace state. This | ||
77 | completes the gzip file, so an interrupt after being so marked does not | ||
78 | result in recompression. Then the dictionary file is replaced, and the gzip | ||
79 | file is marked as completed. This state prevents the possibility of | ||
80 | restarting compression with the wrong dictionary file. | ||
81 | |||
82 | All three operations are wrapped by a lock/unlock procedure. In order to | ||
83 | gain exclusive access to the log files, first a foo.lock file must be | ||
84 | exclusively created. When all operations are complete, the lock is | ||
85 | released by deleting the foo.lock file. If when attempting to create the | ||
86 | lock file, it already exists and the modify time of the lock file is more | ||
87 | than five minutes old (set by the PATIENCE define below), then the old | ||
88 | lock file is considered stale and deleted, and the exclusive creation of | ||
89 | the lock file is retried. To assure that there are no false assessments | ||
90 | of the staleness of the lock file, the operations periodically touch the | ||
91 | lock file to update the modified date. | ||
92 | |||
93 | Following is the definition of the extra field with all of the information | ||
94 | required to enable the above append and compress operations and their | ||
95 | recovery if interrupted. Multi-byte values are stored little endian | ||
96 | (consistent with the gzip format). File pointers are eight bytes long. | ||
97 | The crc's and lengths for the gzip trailer are four bytes long. (Note that | ||
98 | the length at the end of a gzip file is used for error checking only, and | ||
99 | for large files is actually the length modulo 2^32.) The stored block | ||
100 | length is two bytes long. The gzip extra field two-byte identification is | ||
101 | "ap" for append. It is assumed that writing the extra field to the file is | ||
102 | an "atomic" operation. That is, either all of the extra field is written | ||
103 | to the file, or none of it is, if the operation is interrupted right at the | ||
104 | point of updating the extra field. This is a reasonable assumption, since | ||
105 | the extra field is within the first 52 bytes of the file, which is smaller | ||
106 | than any expected block size for a mass storage device (usually 512 bytes or | ||
107 | larger). | ||
108 | |||
109 | Extra field (35 bytes): | ||
110 | - Pointer to first stored block length -- this points to the two-byte length | ||
111 | of the first stored block, which is followed by the two-byte, one's | ||
112 | complement of that length. The stored block length is preceded by the | ||
113 | three-bit header of the stored block, which is the actual start of the | ||
114 | stored block in the deflate format. See the bit offset field below. | ||
115 | - Pointer to the last stored block length. This is the same as above, but | ||
116 | for the last stored block of the uncompressed data in the gzip file. | ||
117 | Initially this is the same as the first stored block length pointer. | ||
118 | When the stored block gets to 16K (see the MAX_STORE define), then a new | ||
119 | stored block as added, at which point the last stored block length pointer | ||
120 | is different from the first stored block length pointer. When they are | ||
121 | different, the first bit of the last stored block header is eight bits, or | ||
122 | one byte back from the block length. | ||
123 | - Compressed data crc and length. This is the crc and length of the data | ||
124 | that is in the compressed portion of the deflate stream. These are used | ||
125 | only in the event that the foo.add file containing the data to compress is | ||
126 | lost after a compress operation is interrupted. | ||
127 | - Total data crc and length. This is the crc and length of all of the data | ||
128 | stored in the gzip file, compressed and uncompressed. It is used to | ||
129 | reconstruct the gzip trailer when compressing, as well as when recovering | ||
130 | interrupted operations. | ||
131 | - Final stored block length. This is used to quickly find where to append, | ||
132 | and allows the restoration of the original final stored block state when | ||
133 | an append operation is interrupted. | ||
134 | - First stored block start as the number of bits back from the final stored | ||
135 | block first length byte. This value is in the range of 3..10, and is | ||
136 | stored as the low three bits of the final byte of the extra field after | ||
137 | subtracting three (0..7). This allows the last-block bit of the stored | ||
138 | block header to be updated when a new stored block is added, for the case | ||
139 | when the first stored block and the last stored block are the same. (When | ||
140 | they are different, the numbers of bits back is known to be eight.) This | ||
141 | also allows for new compressed data to be appended to the old compressed | ||
142 | data in the compress operation, overwriting the previous first stored | ||
143 | block, or for the compressed data to be terminated and a valid gzip file | ||
144 | reconstructed on the off chance that a compression operation was | ||
145 | interrupted and the data to compress in the foo.add file was deleted. | ||
146 | - The operation in process. This is the next two bits in the last byte (the | ||
147 | bits under the mask 0x18). The are interpreted as 0: nothing in process, | ||
148 | 1: append in process, 2: compress in process, 3: replace in process. | ||
149 | - The top three bits of the last byte in the extra field are reserved and | ||
150 | are currently set to zero. | ||
151 | |||
152 | Main procedure: | ||
153 | - Exclusively create the foo.lock file using the O_CREAT and O_EXCL modes of | ||
154 | the system open() call. If the modify time of an existing lock file is | ||
155 | more than PATIENCE seconds old, then the lock file is deleted and the | ||
156 | exclusive create is retried. | ||
157 | - Load the extra field from the foo.gz file, and see if an operation was in | ||
158 | progress but not completed. If so, apply the recovery procedure below. | ||
159 | - Perform the append procedure with the provided data. | ||
160 | - If the uncompressed data in the foo.gz file is 1MB or more, apply the | ||
161 | compress procedure. | ||
162 | - Delete the foo.lock file. | ||
163 | |||
164 | Append procedure: | ||
165 | - Put what to append in the foo.add file so that the operation can be | ||
166 | restarted if this procedure is interrupted. | ||
167 | - Mark the foo.gz extra field with the append operation in progress. | ||
168 | + Restore the original last-block bit and stored block length of the last | ||
169 | stored block from the information in the extra field, in case a previous | ||
170 | append operation was interrupted. | ||
171 | - Append the provided data to the last stored block, creating new stored | ||
172 | blocks as needed and updating the stored blocks last-block bits and | ||
173 | lengths. | ||
174 | - Update the crc and length with the new data, and write the gzip trailer. | ||
175 | - Write over the extra field (with a single write operation) with the new | ||
176 | pointers, lengths, and crc's, and mark the gzip file as not in process. | ||
177 | Though there is still a foo.add file, it will be ignored since nothing | ||
178 | is in process. If a foo.add file is leftover from a previously | ||
179 | completed operation, it is truncated when writing new data to it. | ||
180 | - Delete the foo.add file. | ||
181 | |||
182 | Compress and replace procedures: | ||
183 | - Read all of the uncompressed data in the stored blocks in foo.gz and write | ||
184 | it to foo.add. Also write foo.temp with the last 32K of that data to | ||
185 | provide a dictionary for the next invocation of this procedure. | ||
186 | - Rewrite the extra field marking foo.gz with a compression in process. | ||
187 | * If there is no data provided to compress (due to a missing foo.add file | ||
188 | when recovering), reconstruct and truncate the foo.gz file to contain | ||
189 | only the previous compressed data and proceed to the step after the next | ||
190 | one. Otherwise ... | ||
191 | - Compress the data with the dictionary in foo.dict, and write to the | ||
192 | foo.gz file starting at the bit immediately following the last previously | ||
193 | compressed block. If there is no foo.dict, proceed anyway with the | ||
194 | compression at slightly reduced efficiency. (For the foo.dict file to be | ||
195 | missing requires some external failure beyond simply the interruption of | ||
196 | a compress operation.) During this process, the foo.lock file is | ||
197 | periodically touched to assure that that file is not considered stale by | ||
198 | another process before we're done. The deflation is terminated with a | ||
199 | non-last empty static block (10 bits long), that is then located and | ||
200 | written over by a last-bit-set empty stored block. | ||
201 | - Append the crc and length of the data in the gzip file (previously | ||
202 | calculated during the append operations). | ||
203 | - Write over the extra field with the updated stored block offsets, bits | ||
204 | back, crc's, and lengths, and mark foo.gz as in process for a replacement | ||
205 | of the dictionary. | ||
206 | @ Delete the foo.add file. | ||
207 | - Replace foo.dict with foo.temp. | ||
208 | - Write over the extra field, marking foo.gz as complete. | ||
209 | |||
210 | Recovery procedure: | ||
211 | - If not a replace recovery, read in the foo.add file, and provide that data | ||
212 | to the appropriate recovery below. If there is no foo.add file, provide | ||
213 | a zero data length to the recovery. In that case, the append recovery | ||
214 | restores the foo.gz to the previous compressed + uncompressed data state. | ||
215 | For the the compress recovery, a missing foo.add file results in foo.gz | ||
216 | being restored to the previous compressed-only data state. | ||
217 | - Append recovery: | ||
218 | - Pick up append at + step above | ||
219 | - Compress recovery: | ||
220 | - Pick up compress at * step above | ||
221 | - Replace recovery: | ||
222 | - Pick up compress at @ step above | ||
223 | - Log the repair with a date stamp in foo.repairs | ||
224 | */ | ||
225 | |||
226 | #include <sys/types.h> | ||
227 | #include <stdio.h> /* rename, fopen, fprintf, fclose */ | ||
228 | #include <stdlib.h> /* malloc, free */ | ||
229 | #include <string.h> /* strlen, strrchr, strcpy, strncpy, strcmp */ | ||
230 | #include <fcntl.h> /* open */ | ||
231 | #include <unistd.h> /* lseek, read, write, close, unlink, sleep, */ | ||
232 | /* ftruncate, fsync */ | ||
233 | #include <errno.h> /* errno */ | ||
234 | #include <time.h> /* time, ctime */ | ||
235 | #include <sys/stat.h> /* stat */ | ||
236 | #include <sys/time.h> /* utimes */ | ||
237 | #include "zlib.h" /* crc32 */ | ||
238 | |||
239 | #include "gzlog.h" /* header for external access */ | ||
17 | 240 | ||
18 | #include "gzlog.h" /* interface */ | ||
19 | #define local static | 241 | #define local static |
242 | typedef unsigned int uint; | ||
243 | typedef unsigned long ulong; | ||
244 | |||
245 | /* Macro for debugging to deterministically force recovery operations */ | ||
246 | #ifdef DEBUG | ||
247 | #include <setjmp.h> /* longjmp */ | ||
248 | jmp_buf gzlog_jump; /* where to go back to */ | ||
249 | int gzlog_bail = 0; /* which point to bail at (1..8) */ | ||
250 | int gzlog_count = -1; /* number of times through to wait */ | ||
251 | # define BAIL(n) do { if (n == gzlog_bail && gzlog_count-- == 0) \ | ||
252 | longjmp(gzlog_jump, gzlog_bail); } while (0) | ||
253 | #else | ||
254 | # define BAIL(n) | ||
255 | #endif | ||
256 | |||
257 | /* how old the lock file can be in seconds before considering it stale */ | ||
258 | #define PATIENCE 300 | ||
259 | |||
260 | /* maximum stored block size in Kbytes -- must be in 1..63 */ | ||
261 | #define MAX_STORE 16 | ||
20 | 262 | ||
21 | /* log object structure */ | 263 | /* number of stored Kbytes to trigger compression (must be >= 32 to allow |
22 | typedef struct { | 264 | dictionary construction, and <= 204 * MAX_STORE, in order for >> 10 to |
23 | int id; /* object identifier */ | 265 | discard the stored block headers contribution of five bytes each) */ |
24 | int fd; /* log file descriptor */ | 266 | #define TRIGGER 1024 |
25 | off_t extra; /* offset of extra "ap" subfield */ | 267 | |
26 | off_t mark_off; /* offset of marked data */ | 268 | /* size of a deflate dictionary (this cannot be changed) */ |
27 | off_t last_off; /* offset of last block */ | 269 | #define DICT 32768U |
28 | unsigned long crc; /* uncompressed crc */ | 270 | |
29 | unsigned long len; /* uncompressed length (modulo 2^32) */ | 271 | /* values for the operation (2 bits) */ |
30 | unsigned stored; /* length of current stored block */ | 272 | #define NO_OP 0 |
31 | } gz_log; | 273 | #define APPEND_OP 1 |
32 | 274 | #define COMPRESS_OP 2 | |
33 | #define GZLOGID 19334 /* gz_log object identifier */ | 275 | #define REPLACE_OP 3 |
34 | 276 | ||
35 | #define LOCK_RETRY 1 /* retry lock once a second */ | 277 | /* macros to extract little-endian integers from an unsigned byte buffer */ |
36 | #define LOCK_PATIENCE 1200 /* try about twenty minutes before forcing */ | 278 | #define PULL2(p) ((p)[0]+((uint)((p)[1])<<8)) |
37 | 279 | #define PULL4(p) (PULL2(p)+((ulong)PULL2(p+2)<<16)) | |
38 | /* acquire a lock on a file */ | 280 | #define PULL8(p) (PULL4(p)+((off_t)PULL4(p+4)<<32)) |
39 | local int lock(int fd) | 281 | |
282 | /* macros to store integers into a byte buffer in little-endian order */ | ||
283 | #define PUT2(p,a) do {(p)[0]=a;(p)[1]=(a)>>8;} while(0) | ||
284 | #define PUT4(p,a) do {PUT2(p,a);PUT2(p+2,a>>16);} while(0) | ||
285 | #define PUT8(p,a) do {PUT4(p,a);PUT4(p+4,a>>32);} while(0) | ||
286 | |||
287 | /* internal structure for log information */ | ||
288 | #define LOGID "\106\035\172" /* should be three non-zero characters */ | ||
289 | struct log { | ||
290 | char id[4]; /* contains LOGID to detect inadvertent overwrites */ | ||
291 | int fd; /* file descriptor for .gz file, opened read/write */ | ||
292 | char *path; /* allocated path, e.g. "/var/log/foo" or "foo" */ | ||
293 | char *end; /* end of path, for appending suffices such as ".gz" */ | ||
294 | off_t first; /* offset of first stored block first length byte */ | ||
295 | int back; /* location of first block id in bits back from first */ | ||
296 | uint stored; /* bytes currently in last stored block */ | ||
297 | off_t last; /* offset of last stored block first length byte */ | ||
298 | ulong ccrc; /* crc of compressed data */ | ||
299 | ulong clen; /* length (modulo 2^32) of compressed data */ | ||
300 | ulong tcrc; /* crc of total data */ | ||
301 | ulong tlen; /* length (modulo 2^32) of total data */ | ||
302 | time_t lock; /* last modify time of our lock file */ | ||
303 | }; | ||
304 | |||
305 | /* gzip header for gzlog */ | ||
306 | local unsigned char log_gzhead[] = { | ||
307 | 0x1f, 0x8b, /* magic gzip id */ | ||
308 | 8, /* compression method is deflate */ | ||
309 | 4, /* there is an extra field (no file name) */ | ||
310 | 0, 0, 0, 0, /* no modification time provided */ | ||
311 | 0, 0xff, /* no extra flags, no OS specified */ | ||
312 | 39, 0, 'a', 'p', 35, 0 /* extra field with "ap" subfield */ | ||
313 | /* 35 is EXTRA, 39 is EXTRA + 4 */ | ||
314 | }; | ||
315 | |||
316 | #define HEAD sizeof(log_gzhead) /* should be 16 */ | ||
317 | |||
318 | /* initial gzip extra field content (52 == HEAD + EXTRA + 1) */ | ||
319 | local unsigned char log_gzext[] = { | ||
320 | 52, 0, 0, 0, 0, 0, 0, 0, /* offset of first stored block length */ | ||
321 | 52, 0, 0, 0, 0, 0, 0, 0, /* offset of last stored block length */ | ||
322 | 0, 0, 0, 0, 0, 0, 0, 0, /* compressed data crc and length */ | ||
323 | 0, 0, 0, 0, 0, 0, 0, 0, /* total data crc and length */ | ||
324 | 0, 0, /* final stored block data length */ | ||
325 | 5 /* op is NO_OP, last bit 8 bits back */ | ||
326 | }; | ||
327 | |||
328 | #define EXTRA sizeof(log_gzext) /* should be 35 */ | ||
329 | |||
330 | /* initial gzip data and trailer */ | ||
331 | local unsigned char log_gzbody[] = { | ||
332 | 1, 0, 0, 0xff, 0xff, /* empty stored block (last) */ | ||
333 | 0, 0, 0, 0, /* crc */ | ||
334 | 0, 0, 0, 0 /* uncompressed length */ | ||
335 | }; | ||
336 | |||
337 | #define BODY sizeof(log_gzbody) | ||
338 | |||
339 | /* Exclusively create foo.lock in order to negotiate exclusive access to the | ||
340 | foo.* files. If the modify time of an existing lock file is greater than | ||
341 | PATIENCE seconds in the past, then consider the lock file to have been | ||
342 | abandoned, delete it, and try the exclusive create again. Save the lock | ||
343 | file modify time for verification of ownership. Return 0 on success, or -1 | ||
344 | on failure, usually due to an access restriction or invalid path. Note that | ||
345 | if stat() or unlink() fails, it may be due to another process noticing the | ||
346 | abandoned lock file a smidge sooner and deleting it, so those are not | ||
347 | flagged as an error. */ | ||
348 | local int log_lock(struct log *log) | ||
40 | { | 349 | { |
41 | int patience; | 350 | int fd; |
351 | struct stat st; | ||
42 | 352 | ||
43 | /* try to lock every LOCK_RETRY seconds for LOCK_PATIENCE seconds */ | 353 | strcpy(log->end, ".lock"); |
44 | patience = LOCK_PATIENCE; | 354 | while ((fd = open(log->path, O_CREAT | O_EXCL, 0644)) < 0) { |
45 | do { | 355 | if (errno != EEXIST) |
46 | if (flock(fd, LOCK_EX + LOCK_NB) == 0) | 356 | return -1; |
47 | return 0; | 357 | if (stat(log->path, &st) == 0 && time(NULL) - st.st_mtime > PATIENCE) { |
48 | (void)sleep(LOCK_RETRY); | 358 | unlink(log->path); |
49 | patience -= LOCK_RETRY; | 359 | continue; |
50 | } while (patience > 0); | 360 | } |
361 | sleep(2); /* relinquish the CPU for two seconds while waiting */ | ||
362 | } | ||
363 | close(fd); | ||
364 | if (stat(log->path, &st) == 0) | ||
365 | log->lock = st.st_mtime; | ||
366 | return 0; | ||
367 | } | ||
51 | 368 | ||
52 | /* we've run out of patience -- give up */ | 369 | /* Update the modify time of the lock file to now, in order to prevent another |
53 | return -1; | 370 | task from thinking that the lock is stale. Save the lock file modify time |
371 | for verification of ownership. */ | ||
372 | local void log_touch(struct log *log) | ||
373 | { | ||
374 | struct stat st; | ||
375 | |||
376 | strcpy(log->end, ".lock"); | ||
377 | utimes(log->path, NULL); | ||
378 | if (stat(log->path, &st) == 0) | ||
379 | log->lock = st.st_mtime; | ||
54 | } | 380 | } |
55 | 381 | ||
56 | /* release lock */ | 382 | /* Check the log file modify time against what is expected. Return true if |
57 | local void unlock(int fd) | 383 | this is not our lock. If it is our lock, touch it to keep it. */ |
384 | local int log_check(struct log *log) | ||
58 | { | 385 | { |
59 | (void)flock(fd, LOCK_UN); | 386 | struct stat st; |
387 | |||
388 | strcpy(log->end, ".lock"); | ||
389 | if (stat(log->path, &st) || st.st_mtime != log->lock) | ||
390 | return 1; | ||
391 | log_touch(log); | ||
392 | return 0; | ||
60 | } | 393 | } |
61 | 394 | ||
62 | /* release a log object */ | 395 | /* Unlock a previously acquired lock, but only if it's ours. */ |
63 | local void log_clean(gz_log *log) | 396 | local void log_unlock(struct log *log) |
64 | { | 397 | { |
65 | unlock(log->fd); | 398 | if (log_check(log)) |
66 | (void)close(log->fd); | 399 | return; |
67 | free(log); | 400 | strcpy(log->end, ".lock"); |
401 | unlink(log->path); | ||
402 | log->lock = 0; | ||
68 | } | 403 | } |
69 | 404 | ||
70 | /* read an unsigned long from a byte buffer little-endian */ | 405 | /* Check the gzip header and read in the extra field, filling in the values in |
71 | local unsigned long make_ulg(unsigned char *buf) | 406 | the log structure. Return op on success or -1 if the gzip header was not as |
407 | expected. op is the current operation in progress last written to the extra | ||
408 | field. This assumes that the gzip file has already been opened, with the | ||
409 | file descriptor log->fd. */ | ||
410 | local int log_head(struct log *log) | ||
72 | { | 411 | { |
73 | int n; | 412 | int op; |
74 | unsigned long val; | 413 | unsigned char buf[HEAD + EXTRA]; |
75 | 414 | ||
76 | val = (unsigned long)(*buf++); | 415 | if (lseek(log->fd, 0, SEEK_SET) < 0 || |
77 | for (n = 8; n < 32; n += 8) | 416 | read(log->fd, buf, HEAD + EXTRA) != HEAD + EXTRA || |
78 | val += (unsigned long)(*buf++) << n; | 417 | memcmp(buf, log_gzhead, HEAD)) { |
79 | return val; | 418 | return -1; |
419 | } | ||
420 | log->first = PULL8(buf + HEAD); | ||
421 | log->last = PULL8(buf + HEAD + 8); | ||
422 | log->ccrc = PULL4(buf + HEAD + 16); | ||
423 | log->clen = PULL4(buf + HEAD + 20); | ||
424 | log->tcrc = PULL4(buf + HEAD + 24); | ||
425 | log->tlen = PULL4(buf + HEAD + 28); | ||
426 | log->stored = PULL2(buf + HEAD + 32); | ||
427 | log->back = 3 + (buf[HEAD + 34] & 7); | ||
428 | op = (buf[HEAD + 34] >> 3) & 3; | ||
429 | return op; | ||
80 | } | 430 | } |
81 | 431 | ||
82 | /* read an off_t from a byte buffer little-endian */ | 432 | /* Write over the extra field contents, marking the operation as op. Use fsync |
83 | local off_t make_off(unsigned char *buf) | 433 | to assure that the device is written to, and in the requested order. This |
434 | operation, and only this operation, is assumed to be atomic in order to | ||
435 | assure that the log is recoverable in the event of an interruption at any | ||
436 | point in the process. Return -1 if the write to foo.gz failed. */ | ||
437 | local int log_mark(struct log *log, int op) | ||
84 | { | 438 | { |
85 | int n; | 439 | int ret; |
86 | off_t val; | 440 | unsigned char ext[EXTRA]; |
87 | 441 | ||
88 | val = (off_t)(*buf++); | 442 | PUT8(ext, log->first); |
89 | for (n = 8; n < 64; n += 8) | 443 | PUT8(ext + 8, log->last); |
90 | val += (off_t)(*buf++) << n; | 444 | PUT4(ext + 16, log->ccrc); |
91 | return val; | 445 | PUT4(ext + 20, log->clen); |
446 | PUT4(ext + 24, log->tcrc); | ||
447 | PUT4(ext + 28, log->tlen); | ||
448 | PUT2(ext + 32, log->stored); | ||
449 | ext[34] = log->back - 3 + (op << 3); | ||
450 | fsync(log->fd); | ||
451 | ret = lseek(log->fd, HEAD, SEEK_SET) < 0 || | ||
452 | write(log->fd, ext, EXTRA) != EXTRA ? -1 : 0; | ||
453 | fsync(log->fd); | ||
454 | return ret; | ||
92 | } | 455 | } |
93 | 456 | ||
94 | /* write an unsigned long little-endian to byte buffer */ | 457 | /* Rewrite the last block header bits and subsequent zero bits to get to a byte |
95 | local void dice_ulg(unsigned long val, unsigned char *buf) | 458 | boundary, setting the last block bit if last is true, and then write the |
459 | remainder of the stored block header (length and one's complement). Leave | ||
460 | the file pointer after the end of the last stored block data. Return -1 if | ||
461 | there is a read or write failure on the foo.gz file */ | ||
462 | local int log_last(struct log *log, int last) | ||
96 | { | 463 | { |
97 | int n; | 464 | int back, len, mask; |
465 | unsigned char buf[6]; | ||
466 | |||
467 | /* determine the locations of the bytes and bits to modify */ | ||
468 | back = log->last == log->first ? log->back : 8; | ||
469 | len = back > 8 ? 2 : 1; /* bytes back from log->last */ | ||
470 | mask = 0x80 >> ((back - 1) & 7); /* mask for block last-bit */ | ||
471 | |||
472 | /* get the byte to modify (one or two back) into buf[0] -- don't need to | ||
473 | read the byte if the last-bit is eight bits back, since in that case | ||
474 | the entire byte will be modified */ | ||
475 | buf[0] = 0; | ||
476 | if (back != 8 && (lseek(log->fd, log->last - len, SEEK_SET) < 0 || | ||
477 | read(log->fd, buf, 1) != 1)) | ||
478 | return -1; | ||
479 | |||
480 | /* change the last-bit of the last stored block as requested -- note | ||
481 | that all bits above the last-bit are set to zero, per the type bits | ||
482 | of a stored block being 00 and per the convention that the bits to | ||
483 | bring the stream to a byte boundary are also zeros */ | ||
484 | buf[1] = 0; | ||
485 | buf[2 - len] = (*buf & (mask - 1)) + (last ? mask : 0); | ||
98 | 486 | ||
99 | for (n = 0; n < 4; n++) { | 487 | /* write the modified stored block header and lengths, move the file |
100 | *buf++ = val & 0xff; | 488 | pointer to after the last stored block data */ |
101 | val >>= 8; | 489 | PUT2(buf + 2, log->stored); |
490 | PUT2(buf + 4, log->stored ^ 0xffff); | ||
491 | return lseek(log->fd, log->last - len, SEEK_SET) < 0 || | ||
492 | write(log->fd, buf + 2 - len, len + 4) != len + 4 || | ||
493 | lseek(log->fd, log->stored, SEEK_CUR) < 0 ? -1 : 0; | ||
494 | } | ||
495 | |||
496 | /* Append len bytes from data to the locked and open log file. len may be zero | ||
497 | if recovering and no .add file was found. In that case, the previous state | ||
498 | of the foo.gz file is restored. The data is appended uncompressed in | ||
499 | deflate stored blocks. Return -1 if there was an error reading or writing | ||
500 | the foo.gz file. */ | ||
501 | local int log_append(struct log *log, unsigned char *data, size_t len) | ||
502 | { | ||
503 | uint put; | ||
504 | off_t end; | ||
505 | unsigned char buf[8]; | ||
506 | |||
507 | /* set the last block last-bit and length, in case recovering an | ||
508 | interrupted append, then position the file pointer to append to the | ||
509 | block */ | ||
510 | if (log_last(log, 1)) | ||
511 | return -1; | ||
512 | |||
513 | /* append, adding stored blocks and updating the offset of the last stored | ||
514 | block as needed, and update the total crc and length */ | ||
515 | while (len) { | ||
516 | /* append as much as we can to the last block */ | ||
517 | put = (MAX_STORE << 10) - log->stored; | ||
518 | if (put > len) | ||
519 | put = (uint)len; | ||
520 | if (put) { | ||
521 | if (write(log->fd, data, put) != put) | ||
522 | return -1; | ||
523 | BAIL(1); | ||
524 | log->tcrc = crc32(log->tcrc, data, put); | ||
525 | log->tlen += put; | ||
526 | log->stored += put; | ||
527 | data += put; | ||
528 | len -= put; | ||
529 | } | ||
530 | |||
531 | /* if we need to, add a new empty stored block */ | ||
532 | if (len) { | ||
533 | /* mark current block as not last */ | ||
534 | if (log_last(log, 0)) | ||
535 | return -1; | ||
536 | |||
537 | /* point to new, empty stored block */ | ||
538 | log->last += 4 + log->stored + 1; | ||
539 | log->stored = 0; | ||
540 | } | ||
541 | |||
542 | /* mark last block as last, update its length */ | ||
543 | if (log_last(log, 1)) | ||
544 | return -1; | ||
545 | BAIL(2); | ||
102 | } | 546 | } |
547 | |||
548 | /* write the new crc and length trailer, and truncate just in case (could | ||
549 | be recovering from partial append with a missing foo.add file) */ | ||
550 | PUT4(buf, log->tcrc); | ||
551 | PUT4(buf + 4, log->tlen); | ||
552 | if (write(log->fd, buf, 8) != 8 || | ||
553 | (end = lseek(log->fd, 0, SEEK_CUR)) < 0 || ftruncate(log->fd, end)) | ||
554 | return -1; | ||
555 | |||
556 | /* write the extra field, marking the log file as done, delete .add file */ | ||
557 | if (log_mark(log, NO_OP)) | ||
558 | return -1; | ||
559 | strcpy(log->end, ".add"); | ||
560 | unlink(log->path); /* ignore error, since may not exist */ | ||
561 | return 0; | ||
103 | } | 562 | } |
104 | 563 | ||
105 | /* write an off_t little-endian to byte buffer */ | 564 | /* Replace the foo.dict file with the foo.temp file. Also delete the foo.add |
106 | local void dice_off(off_t val, unsigned char *buf) | 565 | file, since the compress operation may have been interrupted before that was |
566 | done. Returns 1 if memory could not be allocated, or -1 if reading or | ||
567 | writing foo.gz fails, or if the rename fails for some reason other than | ||
568 | foo.temp not existing. foo.temp not existing is a permitted error, since | ||
569 | the replace operation may have been interrupted after the rename is done, | ||
570 | but before foo.gz is marked as complete. */ | ||
571 | local int log_replace(struct log *log) | ||
107 | { | 572 | { |
108 | int n; | 573 | int ret; |
574 | char *dest; | ||
575 | |||
576 | /* delete foo.add file */ | ||
577 | strcpy(log->end, ".add"); | ||
578 | unlink(log->path); /* ignore error, since may not exist */ | ||
579 | BAIL(3); | ||
580 | |||
581 | /* rename foo.name to foo.dict, replacing foo.dict if it exists */ | ||
582 | strcpy(log->end, ".dict"); | ||
583 | dest = malloc(strlen(log->path) + 1); | ||
584 | if (dest == NULL) | ||
585 | return -2; | ||
586 | strcpy(dest, log->path); | ||
587 | strcpy(log->end, ".temp"); | ||
588 | ret = rename(log->path, dest); | ||
589 | free(dest); | ||
590 | if (ret && errno != ENOENT) | ||
591 | return -1; | ||
592 | BAIL(4); | ||
109 | 593 | ||
110 | for (n = 0; n < 8; n++) { | 594 | /* mark the foo.gz file as done */ |
111 | *buf++ = val & 0xff; | 595 | return log_mark(log, NO_OP); |
112 | val >>= 8; | 596 | } |
597 | |||
598 | /* Compress the len bytes at data and append the compressed data to the | ||
599 | foo.gz deflate data immediately after the previous compressed data. This | ||
600 | overwrites the previous uncompressed data, which was stored in foo.add | ||
601 | and is the data provided in data[0..len-1]. If this operation is | ||
602 | interrupted, it picks up at the start of this routine, with the foo.add | ||
603 | file read in again. If there is no data to compress (len == 0), then we | ||
604 | simply terminate the foo.gz file after the previously compressed data, | ||
605 | appending a final empty stored block and the gzip trailer. Return -1 if | ||
606 | reading or writing the log.gz file failed, or -2 if there was a memory | ||
607 | allocation failure. */ | ||
608 | local int log_compress(struct log *log, unsigned char *data, size_t len) | ||
609 | { | ||
610 | int fd; | ||
611 | uint got, max; | ||
612 | ssize_t dict; | ||
613 | off_t end; | ||
614 | z_stream strm; | ||
615 | unsigned char buf[DICT]; | ||
616 | |||
617 | /* compress and append compressed data */ | ||
618 | if (len) { | ||
619 | /* set up for deflate, allocating memory */ | ||
620 | strm.zalloc = Z_NULL; | ||
621 | strm.zfree = Z_NULL; | ||
622 | strm.opaque = Z_NULL; | ||
623 | if (deflateInit2(&strm, Z_DEFAULT_COMPRESSION, Z_DEFLATED, -15, 8, | ||
624 | Z_DEFAULT_STRATEGY) != Z_OK) | ||
625 | return -2; | ||
626 | |||
627 | /* read in dictionary (last 32K of data that was compressed) */ | ||
628 | strcpy(log->end, ".dict"); | ||
629 | fd = open(log->path, O_RDONLY, 0); | ||
630 | if (fd >= 0) { | ||
631 | dict = read(fd, buf, DICT); | ||
632 | close(fd); | ||
633 | if (dict < 0) { | ||
634 | deflateEnd(&strm); | ||
635 | return -1; | ||
636 | } | ||
637 | if (dict) | ||
638 | deflateSetDictionary(&strm, buf, (uint)dict); | ||
639 | } | ||
640 | log_touch(log); | ||
641 | |||
642 | /* prime deflate with last bits of previous block, position write | ||
643 | pointer to write those bits and overwrite what follows */ | ||
644 | if (lseek(log->fd, log->first - (log->back > 8 ? 2 : 1), | ||
645 | SEEK_SET) < 0 || | ||
646 | read(log->fd, buf, 1) != 1 || lseek(log->fd, -1, SEEK_CUR) < 0) { | ||
647 | deflateEnd(&strm); | ||
648 | return -1; | ||
649 | } | ||
650 | deflatePrime(&strm, (8 - log->back) & 7, *buf); | ||
651 | |||
652 | /* compress, finishing with a partial non-last empty static block */ | ||
653 | strm.next_in = data; | ||
654 | max = (((uint)0 - 1) >> 1) + 1; /* in case int smaller than size_t */ | ||
655 | do { | ||
656 | strm.avail_in = len > max ? max : (uint)len; | ||
657 | len -= strm.avail_in; | ||
658 | do { | ||
659 | strm.avail_out = DICT; | ||
660 | strm.next_out = buf; | ||
661 | deflate(&strm, len ? Z_NO_FLUSH : Z_PARTIAL_FLUSH); | ||
662 | got = DICT - strm.avail_out; | ||
663 | if (got && write(log->fd, buf, got) != got) { | ||
664 | deflateEnd(&strm); | ||
665 | return -1; | ||
666 | } | ||
667 | log_touch(log); | ||
668 | } while (strm.avail_out == 0); | ||
669 | } while (len); | ||
670 | deflateEnd(&strm); | ||
671 | BAIL(5); | ||
672 | |||
673 | /* find start of empty static block -- scanning backwards the first one | ||
674 | bit is the second bit of the block, if the last byte is zero, then | ||
675 | we know the byte before that has a one in the top bit, since an | ||
676 | empty static block is ten bits long */ | ||
677 | if ((log->first = lseek(log->fd, -1, SEEK_CUR)) < 0 || | ||
678 | read(log->fd, buf, 1) != 1) | ||
679 | return -1; | ||
680 | log->first++; | ||
681 | if (*buf) { | ||
682 | log->back = 1; | ||
683 | while ((*buf & ((uint)1 << (8 - log->back++))) == 0) | ||
684 | ; /* guaranteed to terminate, since *buf != 0 */ | ||
685 | } | ||
686 | else | ||
687 | log->back = 10; | ||
688 | |||
689 | /* update compressed crc and length */ | ||
690 | log->ccrc = log->tcrc; | ||
691 | log->clen = log->tlen; | ||
692 | } | ||
693 | else { | ||
694 | /* no data to compress -- fix up existing gzip stream */ | ||
695 | log->tcrc = log->ccrc; | ||
696 | log->tlen = log->clen; | ||
113 | } | 697 | } |
698 | |||
699 | /* complete and truncate gzip stream */ | ||
700 | log->last = log->first; | ||
701 | log->stored = 0; | ||
702 | PUT4(buf, log->tcrc); | ||
703 | PUT4(buf + 4, log->tlen); | ||
704 | if (log_last(log, 1) || write(log->fd, buf, 8) != 8 || | ||
705 | (end = lseek(log->fd, 0, SEEK_CUR)) < 0 || ftruncate(log->fd, end)) | ||
706 | return -1; | ||
707 | BAIL(6); | ||
708 | |||
709 | /* mark as being in the replace operation */ | ||
710 | if (log_mark(log, REPLACE_OP)) | ||
711 | return -1; | ||
712 | |||
713 | /* execute the replace operation and mark the file as done */ | ||
714 | return log_replace(log); | ||
114 | } | 715 | } |
115 | 716 | ||
116 | /* initial, empty gzip file for appending */ | 717 | /* log a repair record to the .repairs file */ |
117 | local char empty_gz[] = { | 718 | local void log_log(struct log *log, int op, char *record) |
118 | 0x1f, 0x8b, /* magic gzip id */ | 719 | { |
119 | 8, /* compression method is deflate */ | 720 | time_t now; |
120 | 4, /* there is an extra field */ | 721 | FILE *rec; |
121 | 0, 0, 0, 0, /* no modification time provided */ | ||
122 | 0, 0xff, /* no extra flags, no OS */ | ||
123 | 20, 0, 'a', 'p', 16, 0, /* extra field with "ap" subfield */ | ||
124 | 32, 0, 0, 0, 0, 0, 0, 0, /* offset of uncompressed data */ | ||
125 | 32, 0, 0, 0, 0, 0, 0, 0, /* offset of last block */ | ||
126 | 1, 0, 0, 0xff, 0xff, /* empty stored block (last) */ | ||
127 | 0, 0, 0, 0, /* crc */ | ||
128 | 0, 0, 0, 0 /* uncompressed length */ | ||
129 | }; | ||
130 | 722 | ||
131 | /* initialize a log object with locking */ | 723 | now = time(NULL); |
132 | void *gzlog_open(char *path) | 724 | strcpy(log->end, ".repairs"); |
725 | rec = fopen(log->path, "a"); | ||
726 | if (rec == NULL) | ||
727 | return; | ||
728 | fprintf(rec, "%.24s %s recovery: %s\n", ctime(&now), op == APPEND_OP ? | ||
729 | "append" : (op == COMPRESS_OP ? "compress" : "replace"), record); | ||
730 | fclose(rec); | ||
731 | return; | ||
732 | } | ||
733 | |||
734 | /* Recover the interrupted operation op. First read foo.add for recovering an | ||
735 | append or compress operation. Return -1 if there was an error reading or | ||
736 | writing foo.gz or reading an existing foo.add, or -2 if there was a memory | ||
737 | allocation failure. */ | ||
738 | local int log_recover(struct log *log, int op) | ||
133 | { | 739 | { |
134 | unsigned xlen; | 740 | int fd, ret = 0; |
135 | unsigned char temp[20]; | 741 | unsigned char *data = NULL; |
136 | unsigned sub_len; | 742 | size_t len = 0; |
137 | int good; | 743 | struct stat st; |
138 | gz_log *log; | ||
139 | |||
140 | /* allocate log structure */ | ||
141 | log = malloc(sizeof(gz_log)); | ||
142 | if (log == NULL) | ||
143 | return NULL; | ||
144 | log->id = GZLOGID; | ||
145 | 744 | ||
146 | /* open file, creating it if necessary, and locking it */ | 745 | /* log recovery */ |
147 | log->fd = open(path, O_RDWR | O_CREAT, 0600); | 746 | log_log(log, op, "start"); |
148 | if (log->fd < 0) { | 747 | |
149 | free(log); | 748 | /* load foo.add file if expected and present */ |
150 | return NULL; | 749 | if (op == APPEND_OP || op == COMPRESS_OP) { |
750 | strcpy(log->end, ".add"); | ||
751 | if (stat(log->path, &st) == 0 && st.st_size) { | ||
752 | len = (size_t)(st.st_size); | ||
753 | if (len != st.st_size || (data = malloc(st.st_size)) == NULL) { | ||
754 | log_log(log, op, "allocation failure"); | ||
755 | return -2; | ||
756 | } | ||
757 | if ((fd = open(log->path, O_RDONLY, 0)) < 0) { | ||
758 | log_log(log, op, ".add file read failure"); | ||
759 | return -1; | ||
760 | } | ||
761 | ret = read(fd, data, len) != len; | ||
762 | close(fd); | ||
763 | if (ret) { | ||
764 | log_log(log, op, ".add file read failure"); | ||
765 | return -1; | ||
766 | } | ||
767 | log_log(log, op, "loaded .add file"); | ||
768 | } | ||
769 | else | ||
770 | log_log(log, op, "missing .add file!"); | ||
771 | } | ||
772 | |||
773 | /* recover the interrupted operation */ | ||
774 | switch (op) { | ||
775 | case APPEND_OP: | ||
776 | ret = log_append(log, data, len); | ||
777 | break; | ||
778 | case COMPRESS_OP: | ||
779 | ret = log_compress(log, data, len); | ||
780 | break; | ||
781 | case REPLACE_OP: | ||
782 | ret = log_replace(log); | ||
151 | } | 783 | } |
152 | if (lock(log->fd)) { | 784 | |
785 | /* log status */ | ||
786 | log_log(log, op, ret ? "failure" : "complete"); | ||
787 | |||
788 | /* clean up */ | ||
789 | if (data != NULL) | ||
790 | free(data); | ||
791 | return ret; | ||
792 | } | ||
793 | |||
794 | /* Close the foo.gz file (if open) and release the lock. */ | ||
795 | local void log_close(struct log *log) | ||
796 | { | ||
797 | if (log->fd >= 0) | ||
153 | close(log->fd); | 798 | close(log->fd); |
154 | free(log); | 799 | log->fd = -1; |
155 | return NULL; | 800 | log_unlock(log); |
801 | } | ||
802 | |||
803 | /* Open foo.gz, verify the header, and load the extra field contents, after | ||
804 | first creating the foo.lock file to gain exclusive access to the foo.* | ||
805 | files. If foo.gz does not exist or is empty, then write the initial header, | ||
806 | extra, and body content of an empty foo.gz log file. If there is an error | ||
807 | creating the lock file due to access restrictions, or an error reading or | ||
808 | writing the foo.gz file, or if the foo.gz file is not a proper log file for | ||
809 | this object (e.g. not a gzip file or does not contain the expected extra | ||
810 | field), then return true. If there is an error, the lock is released. | ||
811 | Otherwise, the lock is left in place. */ | ||
812 | local int log_open(struct log *log) | ||
813 | { | ||
814 | int op; | ||
815 | |||
816 | /* release open file resource if left over -- can occur if lock lost | ||
817 | between gzlog_open() and gzlog_write() */ | ||
818 | if (log->fd >= 0) | ||
819 | close(log->fd); | ||
820 | log->fd = -1; | ||
821 | |||
822 | /* negotiate exclusive access */ | ||
823 | if (log_lock(log) < 0) | ||
824 | return -1; | ||
825 | |||
826 | /* open the log file, foo.gz */ | ||
827 | strcpy(log->end, ".gz"); | ||
828 | log->fd = open(log->path, O_RDWR | O_CREAT, 0644); | ||
829 | if (log->fd < 0) { | ||
830 | log_close(log); | ||
831 | return -1; | ||
156 | } | 832 | } |
157 | 833 | ||
158 | /* if file is empty, write new gzip stream */ | 834 | /* if new, initialize foo.gz with an empty log, delete old dictionary */ |
159 | if (lseek(log->fd, 0, SEEK_END) == 0) { | 835 | if (lseek(log->fd, 0, SEEK_END) == 0) { |
160 | if (write(log->fd, empty_gz, sizeof(empty_gz)) != sizeof(empty_gz)) { | 836 | if (write(log->fd, log_gzhead, HEAD) != HEAD || |
161 | log_clean(log); | 837 | write(log->fd, log_gzext, EXTRA) != EXTRA || |
162 | return NULL; | 838 | write(log->fd, log_gzbody, BODY) != BODY) { |
839 | log_close(log); | ||
840 | return -1; | ||
163 | } | 841 | } |
842 | strcpy(log->end, ".dict"); | ||
843 | unlink(log->path); | ||
164 | } | 844 | } |
165 | 845 | ||
166 | /* check gzip header */ | 846 | /* verify log file and load extra field information */ |
167 | (void)lseek(log->fd, 0, SEEK_SET); | 847 | if ((op = log_head(log)) < 0) { |
168 | if (read(log->fd, temp, 12) != 12 || temp[0] != 0x1f || | 848 | log_close(log); |
169 | temp[1] != 0x8b || temp[2] != 8 || (temp[3] & 4) == 0) { | 849 | return -1; |
170 | log_clean(log); | ||
171 | return NULL; | ||
172 | } | 850 | } |
173 | 851 | ||
174 | /* process extra field to find "ap" sub-field */ | 852 | /* check for interrupted process and if so, recover */ |
175 | xlen = temp[10] + (temp[11] << 8); | 853 | if (op != NO_OP && log_recover(log, op)) { |
176 | good = 0; | 854 | log_close(log); |
177 | while (xlen) { | 855 | return -1; |
178 | if (xlen < 4 || read(log->fd, temp, 4) != 4) | ||
179 | break; | ||
180 | sub_len = temp[2]; | ||
181 | sub_len += temp[3] << 8; | ||
182 | xlen -= 4; | ||
183 | if (memcmp(temp, "ap", 2) == 0 && sub_len == 16) { | ||
184 | good = 1; | ||
185 | break; | ||
186 | } | ||
187 | if (xlen < sub_len) | ||
188 | break; | ||
189 | (void)lseek(log->fd, sub_len, SEEK_CUR); | ||
190 | xlen -= sub_len; | ||
191 | } | 856 | } |
192 | if (!good) { | 857 | |
193 | log_clean(log); | 858 | /* touch the lock file to prevent another process from grabbing it */ |
859 | log_touch(log); | ||
860 | return 0; | ||
861 | } | ||
862 | |||
863 | /* See gzlog.h for the description of the external methods below */ | ||
864 | gzlog *gzlog_open(char *path) | ||
865 | { | ||
866 | size_t n; | ||
867 | struct log *log; | ||
868 | |||
869 | /* check arguments */ | ||
870 | if (path == NULL || *path == 0) | ||
194 | return NULL; | 871 | return NULL; |
195 | } | ||
196 | 872 | ||
197 | /* read in "ap" sub-field */ | 873 | /* allocate and initialize log structure */ |
198 | log->extra = lseek(log->fd, 0, SEEK_CUR); | 874 | log = malloc(sizeof(struct log)); |
199 | if (read(log->fd, temp, 16) != 16) { | 875 | if (log == NULL) |
200 | log_clean(log); | 876 | return NULL; |
877 | strcpy(log->id, LOGID); | ||
878 | log->fd = -1; | ||
879 | |||
880 | /* save path and end of path for name construction */ | ||
881 | n = strlen(path); | ||
882 | log->path = malloc(n + 9); /* allow for ".repairs" */ | ||
883 | if (log->path == NULL) { | ||
884 | free(log); | ||
201 | return NULL; | 885 | return NULL; |
202 | } | 886 | } |
203 | log->mark_off = make_off(temp); | 887 | strcpy(log->path, path); |
204 | log->last_off = make_off(temp + 8); | 888 | log->end = log->path + n; |
205 | 889 | ||
206 | /* get crc, length of gzip file */ | 890 | /* gain exclusive access and verify log file -- may perform a |
207 | (void)lseek(log->fd, log->last_off, SEEK_SET); | 891 | recovery operation if needed */ |
208 | if (read(log->fd, temp, 13) != 13 || | 892 | if (log_open(log)) { |
209 | memcmp(temp, "\001\000\000\377\377", 5) != 0) { | 893 | free(log->path); |
210 | log_clean(log); | 894 | free(log); |
211 | return NULL; | 895 | return NULL; |
212 | } | 896 | } |
213 | log->crc = make_ulg(temp + 5); | ||
214 | log->len = make_ulg(temp + 9); | ||
215 | 897 | ||
216 | /* set up to write over empty last block */ | 898 | /* return pointer to log structure */ |
217 | (void)lseek(log->fd, log->last_off + 5, SEEK_SET); | 899 | return log; |
218 | log->stored = 0; | ||
219 | return (void *)log; | ||
220 | } | 900 | } |
221 | 901 | ||
222 | /* maximum amount to put in a stored block before starting a new one */ | 902 | /* gzlog_compress() return values: |
223 | #define MAX_BLOCK 16384 | 903 | 0: all good |
224 | 904 | -1: file i/o error (usually access issue) | |
225 | /* write a block to a log object */ | 905 | -2: memory allocation failure |
226 | int gzlog_write(void *obj, char *data, size_t len) | 906 | -3: invalid log pointer argument */ |
907 | int gzlog_compress(gzlog *logd) | ||
227 | { | 908 | { |
228 | size_t some; | 909 | int fd, ret; |
229 | unsigned char temp[5]; | 910 | uint block; |
230 | gz_log *log; | 911 | size_t len, next; |
912 | unsigned char *data, buf[5]; | ||
913 | struct log *log = logd; | ||
231 | 914 | ||
232 | /* check object */ | 915 | /* check arguments */ |
233 | log = (gz_log *)obj; | 916 | if (log == NULL || strcmp(log->id, LOGID) || len < 0) |
234 | if (log == NULL || log->id != GZLOGID) | 917 | return -3; |
235 | return 1; | ||
236 | 918 | ||
237 | /* write stored blocks until all of the input is written */ | 919 | /* see if we lost the lock -- if so get it again and reload the extra |
238 | do { | 920 | field information (it probably changed), recover last operation if |
239 | some = MAX_BLOCK - log->stored; | 921 | necessary */ |
240 | if (some > len) | 922 | if (log_check(log) && log_open(log)) |
241 | some = len; | 923 | return -1; |
242 | if (write(log->fd, data, some) != some) | ||
243 | return 1; | ||
244 | log->crc = crc32(log->crc, (unsigned char *)data, some); | ||
245 | log->len += some; | ||
246 | len -= some; | ||
247 | data += some; | ||
248 | log->stored += some; | ||
249 | |||
250 | /* if the stored block is full, end it and start another */ | ||
251 | if (log->stored == MAX_BLOCK) { | ||
252 | (void)lseek(log->fd, log->last_off, SEEK_SET); | ||
253 | temp[0] = 0; | ||
254 | dice_ulg(log->stored + ((unsigned long)(~log->stored) << 16), | ||
255 | temp + 1); | ||
256 | if (write(log->fd, temp, 5) != 5) | ||
257 | return 1; | ||
258 | log->last_off = lseek(log->fd, log->stored, SEEK_CUR); | ||
259 | (void)lseek(log->fd, 5, SEEK_CUR); | ||
260 | log->stored = 0; | ||
261 | } | ||
262 | } while (len); | ||
263 | return 0; | ||
264 | } | ||
265 | 924 | ||
266 | /* recompress the remaining stored deflate data in place */ | 925 | /* create space for uncompressed data */ |
267 | local int recomp(gz_log *log) | 926 | len = ((size_t)(log->last - log->first) & ~(((size_t)1 << 10) - 1)) + |
268 | { | 927 | log->stored; |
269 | z_stream strm; | 928 | if ((data = malloc(len)) == NULL) |
270 | size_t len, max; | 929 | return -2; |
271 | unsigned char *in; | ||
272 | unsigned char *out; | ||
273 | unsigned char temp[16]; | ||
274 | |||
275 | /* allocate space and read it all in (it's around 1 MB) */ | ||
276 | len = log->last_off - log->mark_off; | ||
277 | max = len + (len >> 12) + (len >> 14) + 11; | ||
278 | out = malloc(max); | ||
279 | if (out == NULL) | ||
280 | return 1; | ||
281 | in = malloc(len); | ||
282 | if (in == NULL) { | ||
283 | free(out); | ||
284 | return 1; | ||
285 | } | ||
286 | (void)lseek(log->fd, log->mark_off, SEEK_SET); | ||
287 | if (read(log->fd, in, len) != len) { | ||
288 | free(in); | ||
289 | free(out); | ||
290 | return 1; | ||
291 | } | ||
292 | 930 | ||
293 | /* recompress in memory, decoding stored data as we go */ | 931 | /* do statement here is just a cheap trick for error handling */ |
294 | /* note: this assumes that unsigned is four bytes or more */ | 932 | do { |
295 | /* consider not making that assumption */ | 933 | /* read in the uncompressed data */ |
296 | strm.zalloc = Z_NULL; | 934 | if (lseek(log->fd, log->first - 1, SEEK_SET) < 0) |
297 | strm.zfree = Z_NULL; | ||
298 | strm.opaque = Z_NULL; | ||
299 | if (deflateInit2(&strm, Z_BEST_COMPRESSION, Z_DEFLATED, -15, 8, | ||
300 | Z_DEFAULT_STRATEGY) != Z_OK) { | ||
301 | free(in); | ||
302 | free(out); | ||
303 | return 1; | ||
304 | } | ||
305 | strm.next_in = in; | ||
306 | strm.avail_out = max; | ||
307 | strm.next_out = out; | ||
308 | while (len >= 5) { | ||
309 | if (strm.next_in[0] != 0) | ||
310 | break; | 935 | break; |
311 | strm.avail_in = strm.next_in[1] + (strm.next_in[2] << 8); | 936 | next = 0; |
312 | strm.next_in += 5; | 937 | while (next < len) { |
313 | len -= 5; | 938 | if (read(log->fd, buf, 5) != 5) |
314 | if (strm.avail_in != 0) { | ||
315 | if (len < strm.avail_in) | ||
316 | break; | 939 | break; |
317 | len -= strm.avail_in; | 940 | block = PULL2(buf + 1); |
318 | (void)deflate(&strm, Z_NO_FLUSH); | 941 | if (next + block > len || |
319 | if (strm.avail_in != 0 || strm.avail_out == 0) | 942 | read(log->fd, (char *)data + next, block) != block) |
320 | break; | 943 | break; |
944 | next += block; | ||
321 | } | 945 | } |
322 | } | 946 | if (lseek(log->fd, 0, SEEK_CUR) != log->last + 4 + log->stored) |
323 | (void)deflate(&strm, Z_SYNC_FLUSH); | 947 | break; |
324 | (void)deflateEnd(&strm); | 948 | log_touch(log); |
325 | free(in); | ||
326 | if (len != 0 || strm.avail_out == 0) { | ||
327 | free(out); | ||
328 | return 1; | ||
329 | } | ||
330 | 949 | ||
331 | /* overwrite stored data with compressed data */ | 950 | /* write the uncompressed data to the .add file */ |
332 | (void)lseek(log->fd, log->mark_off, SEEK_SET); | 951 | strcpy(log->end, ".add"); |
333 | len = max - strm.avail_out; | 952 | fd = open(log->path, O_WRONLY | O_CREAT | O_TRUNC, 0644); |
334 | if (write(log->fd, out, len) != len) { | 953 | if (fd < 0) |
335 | free(out); | 954 | break; |
336 | return 1; | 955 | ret = write(fd, data, len) != len; |
337 | } | 956 | if (ret | close(fd)) |
338 | free(out); | 957 | break; |
339 | 958 | log_touch(log); | |
340 | /* write last empty block, crc, and length */ | ||
341 | log->mark_off = log->last_off = lseek(log->fd, 0, SEEK_CUR); | ||
342 | temp[0] = 1; | ||
343 | dice_ulg(0xffffL << 16, temp + 1); | ||
344 | dice_ulg(log->crc, temp + 5); | ||
345 | dice_ulg(log->len, temp + 9); | ||
346 | if (write(log->fd, temp, 13) != 13) | ||
347 | return 1; | ||
348 | 959 | ||
349 | /* truncate file to discard remaining stored data and old trailer */ | 960 | /* write the dictionary for the next compress to the .temp file */ |
350 | ftruncate(log->fd, lseek(log->fd, 0, SEEK_CUR)); | 961 | strcpy(log->end, ".temp"); |
962 | fd = open(log->path, O_WRONLY | O_CREAT | O_TRUNC, 0644); | ||
963 | if (fd < 0) | ||
964 | break; | ||
965 | next = DICT > len ? len : DICT; | ||
966 | ret = write(fd, (char *)data + len - next, next) != next; | ||
967 | if (ret | close(fd)) | ||
968 | break; | ||
969 | log_touch(log); | ||
351 | 970 | ||
352 | /* update extra field to point to new last empty block */ | 971 | /* roll back to compressed data, mark the compress in progress */ |
353 | (void)lseek(log->fd, log->extra, SEEK_SET); | 972 | log->last = log->first; |
354 | dice_off(log->mark_off, temp); | 973 | log->stored = 0; |
355 | dice_off(log->last_off, temp + 8); | 974 | if (log_mark(log, COMPRESS_OP)) |
356 | if (write(log->fd, temp, 16) != 16) | 975 | break; |
357 | return 1; | 976 | BAIL(7); |
358 | return 0; | 977 | |
359 | } | 978 | /* compress and append the data (clears mark) */ |
979 | ret = log_compress(log, data, len); | ||
980 | free(data); | ||
981 | return ret; | ||
982 | } while (0); | ||
360 | 983 | ||
361 | /* maximum accumulation of stored blocks before compressing */ | 984 | /* broke out of do above on i/o error */ |
362 | #define MAX_STORED 1048576 | 985 | free(data); |
986 | return -1; | ||
987 | } | ||
363 | 988 | ||
364 | /* close log object */ | 989 | /* gzlog_write() return values: |
365 | int gzlog_close(void *obj) | 990 | 0: all good |
991 | -1: file i/o error (usually access issue) | ||
992 | -2: memory allocation failure | ||
993 | -3: invalid log pointer argument */ | ||
994 | int gzlog_write(gzlog *logd, void *data, size_t len) | ||
366 | { | 995 | { |
367 | unsigned char temp[8]; | 996 | int fd, ret; |
368 | gz_log *log; | 997 | struct log *log = logd; |
369 | 998 | ||
370 | /* check object */ | 999 | /* check arguments */ |
371 | log = (gz_log *)obj; | 1000 | if (log == NULL || strcmp(log->id, LOGID) || len < 0) |
372 | if (log == NULL || log->id != GZLOGID) | 1001 | return -3; |
373 | return 1; | 1002 | if (data == NULL || len == 0) |
1003 | return 0; | ||
374 | 1004 | ||
375 | /* go to start of most recent block being written */ | 1005 | /* see if we lost the lock -- if so get it again and reload the extra |
376 | (void)lseek(log->fd, log->last_off, SEEK_SET); | 1006 | field information (it probably changed), recover last operation if |
377 | 1007 | necessary */ | |
378 | /* if some stuff was put there, update block */ | 1008 | if (log_check(log) && log_open(log)) |
379 | if (log->stored) { | 1009 | return -1; |
380 | temp[0] = 0; | ||
381 | dice_ulg(log->stored + ((unsigned long)(~log->stored) << 16), | ||
382 | temp + 1); | ||
383 | if (write(log->fd, temp, 5) != 5) | ||
384 | return 1; | ||
385 | log->last_off = lseek(log->fd, log->stored, SEEK_CUR); | ||
386 | } | ||
387 | 1010 | ||
388 | /* write last block (empty) */ | 1011 | /* create and write .add file */ |
389 | if (write(log->fd, "\001\000\000\377\377", 5) != 5) | 1012 | strcpy(log->end, ".add"); |
390 | return 1; | 1013 | fd = open(log->path, O_WRONLY | O_CREAT | O_TRUNC, 0644); |
1014 | if (fd < 0) | ||
1015 | return -1; | ||
1016 | ret = write(fd, data, len) != len; | ||
1017 | if (ret | close(fd)) | ||
1018 | return -1; | ||
1019 | log_touch(log); | ||
391 | 1020 | ||
392 | /* write updated crc and uncompressed length */ | 1021 | /* mark log file with append in progress */ |
393 | dice_ulg(log->crc, temp); | 1022 | if (log_mark(log, APPEND_OP)) |
394 | dice_ulg(log->len, temp + 4); | 1023 | return -1; |
395 | if (write(log->fd, temp, 8) != 8) | 1024 | BAIL(8); |
396 | return 1; | ||
397 | 1025 | ||
398 | /* put offset of that last block in gzip extra block */ | 1026 | /* append data (clears mark) */ |
399 | (void)lseek(log->fd, log->extra + 8, SEEK_SET); | 1027 | if (log_append(log, data, len)) |
400 | dice_off(log->last_off, temp); | 1028 | return -1; |
401 | if (write(log->fd, temp, 8) != 8) | ||
402 | return 1; | ||
403 | 1029 | ||
404 | /* if more than 1 MB stored, then time to compress it */ | 1030 | /* check to see if it's time to compress -- if not, then done */ |
405 | if (log->last_off - log->mark_off > MAX_STORED) { | 1031 | if (((log->last - log->first) >> 10) + (log->stored >> 10) < TRIGGER) |
406 | if (recomp(log)) | 1032 | return 0; |
407 | return 1; | 1033 | |
408 | } | 1034 | /* time to compress */ |
1035 | return gzlog_compress(log); | ||
1036 | } | ||
1037 | |||
1038 | /* gzlog_close() return values: | ||
1039 | 0: ok | ||
1040 | -3: invalid log pointer argument */ | ||
1041 | int gzlog_close(gzlog *logd) | ||
1042 | { | ||
1043 | struct log *log = logd; | ||
409 | 1044 | ||
410 | /* unlock and close file */ | 1045 | /* check arguments */ |
411 | log_clean(log); | 1046 | if (log == NULL || strcmp(log->id, LOGID)) |
1047 | return -3; | ||
1048 | |||
1049 | /* close the log file and release the lock */ | ||
1050 | log_close(log); | ||
1051 | |||
1052 | /* free structure and return */ | ||
1053 | if (log->path != NULL) | ||
1054 | free(log->path); | ||
1055 | strcpy(log->id, "bad"); | ||
1056 | free(log); | ||
412 | return 0; | 1057 | return 0; |
413 | } | 1058 | } |