]>
Commit | Line | Data |
---|---|---|
c988ea91 CH |
1 | /* |
2 | * Copyright (c) 2000-2002 Silicon Graphics, Inc. | |
3 | * All Rights Reserved. | |
4 | * | |
5 | * This program is free software; you can redistribute it and/or | |
6 | * modify it under the terms of the GNU General Public License as | |
7 | * published by the Free Software Foundation. | |
8 | * | |
9 | * This program is distributed in the hope that it would be useful, | |
10 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
11 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
12 | * GNU General Public License for more details. | |
13 | * | |
14 | * You should have received a copy of the GNU General Public License | |
15 | * along with this program; if not, write the Free Software Foundation, | |
16 | * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
17 | */ | |
18 | ||
19 | #include <xfs/xfs.h> | |
20 | #include <xfs/jdm.h> | |
21 | #include <xfs/xfs_dfrag.h> | |
22 | ||
23 | #include <fcntl.h> | |
24 | #include <errno.h> | |
25 | #include <malloc.h> | |
26 | #include <mntent.h> | |
27 | #include <syslog.h> | |
28 | #include <signal.h> | |
29 | #include <sys/ioctl.h> | |
30 | #include <sys/wait.h> | |
31 | #include <sys/vfs.h> | |
32 | #include <sys/statvfs.h> | |
33 | #include <sys/xattr.h> | |
34 | ||
35 | ||
36 | #ifndef XFS_XFLAG_NODEFRAG | |
37 | #define XFS_XFLAG_NODEFRAG 0x00002000 /* src dependancy, remove later */ | |
38 | #endif | |
39 | ||
40 | #define _PATH_FSRLAST "/var/tmp/.fsrlast_xfs" | |
41 | ||
42 | char *progname; | |
43 | ||
44 | int vflag; | |
45 | int gflag; | |
46 | static int Mflag; | |
47 | /* static int nflag; */ | |
48 | int dflag = 0; | |
49 | /* static int sflag; */ | |
50 | int argv_blksz_dio; | |
51 | extern int max_ext_size; | |
52 | static int npasses = 10; | |
53 | static int startpass = 0; | |
54 | ||
55 | struct getbmap *outmap = NULL; | |
56 | int outmap_size = 0; | |
57 | int RealUid; | |
58 | int tmp_agi; | |
59 | static __int64_t minimumfree = 2048; | |
60 | ||
61 | #define MNTTYPE_XFS "xfs" | |
62 | ||
63 | #define SMBUFSZ 1024 | |
64 | #define ROOT 0 | |
65 | #define NULLFD -1 | |
66 | #define GRABSZ 64 | |
67 | #define TARGETRANGE 10 | |
68 | #define V_NONE 0 | |
69 | #define V_OVERVIEW 1 | |
70 | #define V_ALL 2 | |
71 | #define BUFFER_SIZE (1<<16) | |
72 | #define BUFFER_MAX (1<<24) | |
73 | #define min(x, y) ((x) < (y) ? (x) : (y)) | |
74 | ||
75 | static time_t howlong = 7200; /* default seconds of reorganizing */ | |
76 | static char *leftofffile = _PATH_FSRLAST; /* where we left off last */ | |
77 | static char *mtab = MOUNTED; | |
78 | static time_t endtime; | |
79 | static time_t starttime; | |
80 | static xfs_ino_t leftoffino = 0; | |
81 | static int pagesize; | |
82 | ||
83 | void usage(int ret); | |
84 | static int fsrfile(char *fname, xfs_ino_t ino); | |
85 | static int fsrfile_common( char *fname, char *tname, char *mnt, | |
86 | int fd, xfs_bstat_t *statp); | |
87 | static int packfile(char *fname, char *tname, int fd, | |
88 | xfs_bstat_t *statp, struct fsxattr *fsxp); | |
89 | static void fsrdir(char *dirname); | |
90 | static int fsrfs(char *mntdir, xfs_ino_t ino, int targetrange); | |
91 | static void initallfs(char *mtab); | |
92 | static void fsrallfs(int howlong, char *leftofffile); | |
93 | static void fsrall_cleanup(int timeout); | |
94 | static int getnextents(int); | |
95 | int xfsrtextsize(int fd); | |
96 | int xfs_getrt(int fd, struct statvfs64 *sfbp); | |
97 | char * gettmpname(char *fname); | |
98 | char * getparent(char *fname); | |
99 | int fsrprintf(const char *fmt, ...); | |
100 | int read_fd_bmap(int, xfs_bstat_t *, int *); | |
101 | int cmp(const void *, const void *); | |
102 | static void tmp_init(char *mnt); | |
103 | static char * tmp_next(char *mnt); | |
104 | static void tmp_close(char *mnt); | |
105 | int xfs_getgeom(int , xfs_fsop_geom_v1_t * ); | |
106 | static int getmntany(FILE *, struct mntent *, struct mntent *, struct stat64 *); | |
107 | ||
108 | xfs_fsop_geom_v1_t fsgeom; /* geometry of active mounted system */ | |
109 | ||
110 | #define NMOUNT 64 | |
111 | static int numfs; | |
112 | ||
113 | typedef struct fsdesc { | |
114 | char *dev; | |
115 | char *mnt; | |
116 | int npass; | |
117 | } fsdesc_t; | |
118 | ||
119 | fsdesc_t *fs, *fsbase, *fsend; | |
120 | int fsbufsize = 10; /* A starting value */ | |
121 | int nfrags = 0; /* Debug option: Coerse into specific number | |
122 | * of extents */ | |
123 | int openopts = O_CREAT|O_EXCL|O_RDWR|O_DIRECT; | |
124 | ||
125 | int | |
126 | xfs_fsgeometry(int fd, xfs_fsop_geom_v1_t *geom) | |
127 | { | |
128 | return ioctl(fd, XFS_IOC_FSGEOMETRY_V1, geom); | |
129 | } | |
130 | ||
131 | int | |
132 | xfs_bulkstat_single(int fd, xfs_ino_t *lastip, xfs_bstat_t *ubuffer) | |
133 | { | |
134 | xfs_fsop_bulkreq_t bulkreq; | |
135 | ||
136 | bulkreq.lastip = lastip; | |
137 | bulkreq.icount = 1; | |
138 | bulkreq.ubuffer = ubuffer; | |
139 | bulkreq.ocount = NULL; | |
140 | return ioctl(fd, XFS_IOC_FSBULKSTAT_SINGLE, &bulkreq); | |
141 | } | |
142 | ||
143 | int | |
144 | xfs_bulkstat(int fd, xfs_ino_t *lastip, int icount, | |
145 | xfs_bstat_t *ubuffer, __s32 *ocount) | |
146 | { | |
147 | xfs_fsop_bulkreq_t bulkreq; | |
148 | ||
149 | bulkreq.lastip = lastip; | |
150 | bulkreq.icount = icount; | |
151 | bulkreq.ubuffer = ubuffer; | |
152 | bulkreq.ocount = ocount; | |
153 | return ioctl(fd, XFS_IOC_FSBULKSTAT, &bulkreq); | |
154 | } | |
155 | ||
156 | int | |
157 | xfs_swapext(int fd, xfs_swapext_t *sx) | |
158 | { | |
159 | return ioctl(fd, XFS_IOC_SWAPEXT, sx); | |
160 | } | |
161 | ||
162 | int | |
163 | xfs_fscounts(int fd, xfs_fsop_counts_t *counts) | |
164 | { | |
165 | return ioctl(fd, XFS_IOC_FSCOUNTS, counts); | |
166 | } | |
167 | ||
168 | void | |
169 | aborter(int unused) | |
170 | { | |
171 | fsrall_cleanup(1); | |
172 | exit(1); | |
173 | } | |
174 | ||
175 | int | |
176 | main(int argc, char **argv) | |
177 | { | |
178 | struct stat64 sb, sb2; | |
179 | char *argname; | |
180 | char *cp; | |
181 | int c; | |
182 | struct mntent mntpref; | |
183 | register struct mntent *mntp; | |
184 | struct mntent ment; | |
185 | register FILE *mtabp; | |
186 | ||
187 | setlinebuf(stdout); | |
188 | progname = basename(argv[0]); | |
189 | ||
190 | setlocale(LC_ALL, ""); | |
191 | bindtextdomain(PACKAGE, LOCALEDIR); | |
192 | textdomain(PACKAGE); | |
193 | ||
194 | gflag = ! isatty(0); | |
195 | ||
196 | while ((c = getopt(argc, argv, "C:p:e:MgsdnvTt:f:m:b:N:FV")) != -1 ) | |
197 | switch (c) { | |
198 | case 'M': | |
199 | Mflag = 1; | |
200 | break; | |
201 | case 'g': | |
202 | gflag = 1; | |
203 | break; | |
204 | case 'n': | |
205 | /* nflag = 1; */ | |
206 | break; | |
207 | case 'v': | |
208 | ++vflag; | |
209 | break; | |
210 | case 'd': | |
211 | dflag = 1; | |
212 | break; | |
213 | case 's': /* frag stats only */ | |
214 | /* sflag = 1; */ | |
215 | fprintf(stderr, | |
216 | _("%s: Stats not yet supported for XFS\n"), | |
217 | progname); | |
218 | usage(1); | |
219 | break; | |
220 | case 't': | |
221 | howlong = atoi(optarg); | |
222 | break; | |
223 | case 'f': | |
224 | leftofffile = optarg; | |
225 | break; | |
226 | case 'm': | |
227 | mtab = optarg; | |
228 | break; | |
229 | case 'b': | |
230 | argv_blksz_dio = atoi(optarg); | |
231 | break; | |
232 | case 'p': | |
233 | npasses = atoi(optarg); | |
234 | break; | |
235 | case 'C': | |
236 | /* Testing opt: coerses frag count in result */ | |
237 | if (getenv("FSRXFSTEST") != NULL) { | |
238 | nfrags = atoi(optarg); | |
239 | openopts |= O_SYNC; | |
240 | } | |
241 | break; | |
242 | case 'V': | |
243 | printf(_("%s version %s\n"), progname, VERSION); | |
244 | exit(0); | |
245 | default: | |
246 | usage(1); | |
247 | } | |
248 | if (vflag) | |
249 | setbuf(stdout, NULL); | |
250 | ||
251 | starttime = time(0); | |
252 | ||
253 | /* Save the caller's real uid */ | |
254 | RealUid = getuid(); | |
255 | ||
256 | pagesize = getpagesize(); | |
257 | ||
258 | if (optind < argc) { | |
259 | for (; optind < argc; optind++) { | |
260 | argname = argv[optind]; | |
261 | mntp = NULL; | |
262 | if (lstat64(argname, &sb) < 0) { | |
263 | fprintf(stderr, | |
264 | _("%s: could not stat: %s: %s\n"), | |
265 | progname, argname, strerror(errno)); | |
266 | continue; | |
267 | } | |
268 | if (S_ISLNK(sb.st_mode) && stat64(argname, &sb2) == 0 && | |
269 | (S_ISBLK(sb2.st_mode) || S_ISCHR(sb2.st_mode))) | |
270 | sb = sb2; | |
271 | if (S_ISBLK(sb.st_mode) || (S_ISDIR(sb.st_mode))) { | |
272 | if ((mtabp = setmntent(mtab, "r")) == NULL) { | |
273 | fprintf(stderr, | |
274 | _("%s: cannot read %s\n"), | |
275 | progname, mtab); | |
276 | exit(1); | |
277 | } | |
278 | bzero(&mntpref, sizeof(mntpref)); | |
279 | if (S_ISDIR(sb.st_mode)) | |
280 | mntpref.mnt_dir = argname; | |
281 | else | |
282 | mntpref.mnt_fsname = argname; | |
283 | ||
284 | if (getmntany(mtabp, &ment, &mntpref, &sb) && | |
285 | strcmp(ment.mnt_type, MNTTYPE_XFS) == 0) { | |
286 | mntp = &ment; | |
287 | if (S_ISBLK(sb.st_mode)) { | |
288 | cp = mntp->mnt_dir; | |
289 | if (cp == NULL || | |
290 | stat64(cp, &sb2) < 0) { | |
291 | fprintf(stderr, _( | |
292 | "%s: could not stat: %s: %s\n"), | |
293 | progname, argname, | |
294 | strerror(errno)); | |
295 | continue; | |
296 | } | |
297 | sb = sb2; | |
298 | argname = cp; | |
299 | } | |
300 | } | |
301 | } | |
302 | if (mntp != NULL) { | |
303 | fsrfs(mntp->mnt_dir, 0, 100); | |
304 | } else if (S_ISCHR(sb.st_mode)) { | |
305 | fprintf(stderr, _( | |
306 | "%s: char special not supported: %s\n"), | |
307 | progname, argname); | |
308 | exit(1); | |
309 | } else if (S_ISDIR(sb.st_mode) || S_ISREG(sb.st_mode)) { | |
310 | if (!platform_test_xfs_path(argname)) { | |
311 | fprintf(stderr, _( | |
312 | "%s: cannot defragment: %s: Not XFS\n"), | |
313 | progname, argname); | |
314 | continue; | |
315 | } | |
316 | if (S_ISDIR(sb.st_mode)) | |
317 | fsrdir(argname); | |
318 | else | |
319 | fsrfile(argname, sb.st_ino); | |
320 | } else { | |
321 | printf( | |
322 | _("%s: not fsys dev, dir, or reg file, ignoring\n"), | |
323 | argname); | |
324 | } | |
325 | } | |
326 | } else { | |
327 | initallfs(mtab); | |
328 | fsrallfs(howlong, leftofffile); | |
329 | } | |
330 | return 0; | |
331 | } | |
332 | ||
333 | void | |
334 | usage(int ret) | |
335 | { | |
336 | fprintf(stderr, _( | |
337 | "Usage: %s [-d] [-v] [-n] [-s] [-g] [-t time] [-p passes] [-f leftf] [-m mtab]\n" | |
338 | " %s [-d] [-v] [-n] [-s] [-g] xfsdev | dir | file ...\n\n" | |
339 | "Options:\n" | |
340 | " -n Do nothing, only interesting with -v. Not\n" | |
341 | " effective with in mtab mode.\n" | |
342 | " -s Print statistics only.\n" | |
343 | " -g Print to syslog (default if stdout not a tty).\n" | |
344 | " -t time How long to run in seconds.\n" | |
345 | " -p passes Number of passes before terminating global re-org.\n" | |
346 | " -f leftoff Use this instead of %s.\n" | |
347 | " -m mtab Use something other than /etc/mtab.\n" | |
348 | " -d Debug, print even more.\n" | |
349 | " -v Verbose, more -v's more verbose.\n" | |
350 | ), progname, progname, _PATH_FSRLAST); | |
351 | exit(ret); | |
352 | } | |
353 | ||
354 | /* | |
355 | * initallfs -- read the mount table and set up an internal form | |
356 | */ | |
357 | static void | |
358 | initallfs(char *mtab) | |
359 | { | |
360 | FILE *fp; | |
361 | struct mntent *mp; | |
362 | int mi; | |
363 | char *cp; | |
364 | struct stat64 sb; | |
365 | ||
366 | fp = setmntent(mtab, "r"); | |
367 | if (fp == NULL) { | |
368 | fsrprintf(_("could not open mtab file: %s\n"), mtab); | |
369 | exit(1); | |
370 | } | |
371 | ||
372 | /* malloc a number of descriptors, increased later if needed */ | |
373 | if (!(fsbase = (fsdesc_t *)malloc(fsbufsize * sizeof(fsdesc_t)))) { | |
374 | fsrprintf(_("out of memory: %s\n"), strerror(errno)); | |
375 | exit(1); | |
376 | } | |
377 | fsend = (fsbase + fsbufsize - 1); | |
378 | ||
379 | /* find all rw xfs file systems */ | |
380 | mi = 0; | |
381 | fs = fsbase; | |
382 | while ((mp = getmntent(fp))) { | |
383 | int rw = 0; | |
384 | ||
385 | if (strcmp(mp->mnt_type, MNTTYPE_XFS ) != 0 || | |
386 | stat64(mp->mnt_fsname, &sb) == -1 || | |
387 | !S_ISBLK(sb.st_mode)) | |
388 | continue; | |
389 | ||
390 | cp = strtok(mp->mnt_opts,","); | |
391 | do { | |
392 | if (strcmp("rw", cp) == 0) | |
393 | rw++; | |
394 | } while ((cp = strtok(NULL, ",")) != NULL); | |
395 | if (rw == 0) { | |
396 | if (dflag) | |
397 | fsrprintf(_("Skipping %s: not mounted rw\n"), | |
398 | mp->mnt_fsname); | |
399 | continue; | |
400 | } | |
401 | ||
402 | if (mi == fsbufsize) { | |
403 | fsbufsize += NMOUNT; | |
404 | if ((fsbase = (fsdesc_t *)realloc((char *)fsbase, | |
405 | fsbufsize * sizeof(fsdesc_t))) == NULL) { | |
406 | fsrprintf(_("out of memory: %s\n"), | |
407 | strerror(errno)); | |
408 | exit(1); | |
409 | } | |
410 | if (!fsbase) { | |
411 | fsrprintf(_("out of memory on realloc: %s\n"), | |
412 | strerror(errno)); | |
413 | exit(1); | |
414 | } | |
415 | fs = (fsbase + mi); /* Needed ? */ | |
416 | } | |
417 | ||
418 | fs->dev = strdup(mp->mnt_fsname); | |
419 | fs->mnt = strdup(mp->mnt_dir); | |
420 | ||
421 | if (fs->mnt == NULL || fs->mnt == NULL) { | |
422 | fsrprintf(_("strdup(%s) failed\n"), mp->mnt_fsname); | |
423 | exit(1); | |
424 | } | |
425 | mi++; | |
426 | fs++; | |
427 | } | |
428 | numfs = mi; | |
429 | fsend = (fsbase + numfs); | |
430 | endmntent(fp); | |
431 | if (numfs == 0) { | |
432 | fsrprintf(_("no rw xfs file systems in mtab: %s\n"), mtab); | |
433 | exit(0); | |
434 | } | |
435 | if (vflag || dflag) { | |
436 | fsrprintf(_("Found %d mounted, writable, XFS filesystems\n"), | |
437 | numfs); | |
438 | if (dflag) | |
439 | for (fs = fsbase; fs < fsend; fs++) | |
440 | fsrprintf("\t%-30.30s%-30.30s\n", fs->dev, fs->mnt); | |
441 | } | |
442 | } | |
443 | ||
444 | static void | |
445 | fsrallfs(int howlong, char *leftofffile) | |
446 | { | |
447 | int fd; | |
448 | int error; | |
449 | int found = 0; | |
450 | char *fsname; | |
451 | char buf[SMBUFSZ]; | |
452 | int mdonly = Mflag; | |
453 | char *ptr; | |
454 | xfs_ino_t startino = 0; | |
455 | fsdesc_t *fsp; | |
456 | struct stat64 sb, sb2; | |
457 | ||
458 | fsrprintf("xfs_fsr -m %s -t %d -f %s ...\n", mtab, howlong, leftofffile); | |
459 | ||
460 | endtime = starttime + howlong; | |
461 | fs = fsbase; | |
462 | ||
463 | /* where'd we leave off last time? */ | |
464 | if (lstat64(leftofffile, &sb) == 0) { | |
465 | if ( (fd = open(leftofffile, O_RDONLY)) == -1 ) { | |
466 | fsrprintf(_("%s: open failed\n"), leftofffile); | |
467 | } | |
468 | else if ( fstat64(fd, &sb2) == 0) { | |
469 | /* | |
470 | * Verify that lstat & fstat point to the | |
471 | * same regular file (no links/no quick spoofs) | |
472 | */ | |
473 | if ( (sb.st_dev != sb2.st_dev) || | |
474 | (sb.st_ino != sb2.st_ino) || | |
475 | ((sb.st_mode & S_IFMT) != S_IFREG) || | |
476 | ((sb2.st_mode & S_IFMT) != S_IFREG) || | |
477 | (sb2.st_uid != ROOT) || | |
478 | (sb2.st_nlink != 1) | |
479 | ) | |
480 | { | |
481 | fsrprintf(_("Can't use %s: mode=0%o own=%d" | |
482 | " nlink=%d\n"), | |
483 | leftofffile, sb.st_mode, | |
484 | sb.st_uid, sb.st_nlink); | |
485 | close(fd); | |
486 | fd = NULLFD; | |
487 | } | |
488 | } | |
489 | else { | |
490 | close(fd); | |
491 | fd = NULLFD; | |
492 | } | |
493 | } | |
494 | else { | |
495 | fd = NULLFD; | |
496 | } | |
497 | ||
498 | if (fd != NULLFD) { | |
499 | if (read(fd, buf, SMBUFSZ) == -1) { | |
500 | fs = fsbase; | |
501 | fsrprintf(_("could not read %s, starting with %s\n"), | |
502 | leftofffile, *fs->dev); | |
503 | } else { | |
504 | for (fs = fsbase; fs < fsend; fs++) { | |
505 | fsname = fs->dev; | |
506 | if ((strncmp(buf,fsname,strlen(fsname)) == 0) | |
507 | && buf[strlen(fsname)] == ' ') { | |
508 | found = 1; | |
509 | break; | |
510 | } | |
511 | } | |
512 | if (! found) | |
513 | fs = fsbase; | |
514 | ||
515 | ptr = strchr(buf, ' '); | |
516 | if (ptr) { | |
517 | startpass = atoi(++ptr); | |
518 | ptr = strchr(ptr, ' '); | |
519 | if (ptr) { | |
520 | startino = strtoull(++ptr, NULL, 10); | |
521 | } | |
522 | } | |
523 | if (startpass < 0) | |
524 | startpass = 0; | |
525 | ||
526 | /* Init pass counts */ | |
527 | for (fsp = fsbase; fsp < fs; fsp++) { | |
528 | fsp->npass = startpass + 1; | |
529 | } | |
530 | for (fsp = fs; fsp <= fsend; fsp++) { | |
531 | fsp->npass = startpass; | |
532 | } | |
533 | } | |
534 | close(fd); | |
535 | } | |
536 | ||
537 | if (vflag) { | |
538 | fsrprintf(_("START: pass=%d ino=%llu %s %s\n"), | |
539 | fs->npass, (unsigned long long)startino, | |
540 | fs->dev, fs->mnt); | |
541 | } | |
542 | ||
543 | signal(SIGABRT, aborter); | |
544 | signal(SIGHUP, aborter); | |
545 | signal(SIGINT, aborter); | |
546 | signal(SIGQUIT, aborter); | |
547 | signal(SIGTERM, aborter); | |
548 | ||
549 | /* reorg for 'howlong' -- checked in 'fsrfs' */ | |
550 | while (endtime > time(0)) { | |
551 | pid_t pid; | |
552 | if (fs == fsend) | |
553 | fs = fsbase; | |
554 | if (fs->npass == npasses) { | |
555 | fsrprintf(_("Completed all %d passes\n"), npasses); | |
556 | break; | |
557 | } | |
558 | if (npasses > 1 && !fs->npass) | |
559 | Mflag = 1; | |
560 | else | |
561 | Mflag = mdonly; | |
562 | pid = fork(); | |
563 | switch(pid) { | |
564 | case -1: | |
565 | fsrprintf(_("couldn't fork sub process:")); | |
566 | exit(1); | |
567 | break; | |
568 | case 0: | |
569 | error = fsrfs(fs->mnt, startino, TARGETRANGE); | |
570 | exit (error); | |
571 | break; | |
572 | default: | |
573 | wait(&error); | |
574 | close(fd); | |
575 | if (WIFEXITED(error) && WEXITSTATUS(error) == 1) { | |
576 | /* child timed out & did fsrall_cleanup */ | |
577 | exit(0); | |
578 | } | |
579 | break; | |
580 | } | |
581 | startino = 0; /* reset after the first time through */ | |
582 | fs->npass++; | |
583 | fs++; | |
584 | } | |
585 | fsrall_cleanup(endtime <= time(0)); | |
586 | } | |
587 | ||
588 | /* | |
589 | * fsrall_cleanup -- close files, print next starting location, etc. | |
590 | */ | |
591 | static void | |
592 | fsrall_cleanup(int timeout) | |
593 | { | |
594 | int fd; | |
595 | int ret; | |
596 | char buf[SMBUFSZ]; | |
597 | ||
598 | /* record where we left off */ | |
599 | unlink(leftofffile); | |
600 | fd = open(leftofffile, O_WRONLY|O_CREAT|O_EXCL, 0644); | |
601 | if (fd == -1) | |
602 | fsrprintf(_("open(%s) failed: %s\n"), | |
603 | leftofffile, strerror(errno)); | |
604 | else { | |
605 | if (timeout) { | |
606 | ret = sprintf(buf, "%s %d %llu\n", fs->dev, | |
607 | fs->npass, (unsigned long long)leftoffino); | |
608 | if (write(fd, buf, ret) < strlen(buf)) | |
609 | fsrprintf(_("write(%s) failed: %s\n"), | |
610 | leftofffile, strerror(errno)); | |
611 | close(fd); | |
612 | } | |
613 | } | |
614 | ||
615 | if (timeout) | |
616 | fsrprintf(_("%s startpass %d, endpass %d, time %d seconds\n"), | |
617 | progname, startpass, fs->npass, | |
618 | time(0) - endtime + howlong); | |
619 | } | |
620 | ||
621 | /* | |
622 | * fsrfs -- reorganize a file system | |
623 | */ | |
624 | static int | |
625 | fsrfs(char *mntdir, xfs_ino_t startino, int targetrange) | |
626 | { | |
627 | ||
628 | int fsfd, fd; | |
629 | int count = 0; | |
630 | int ret; | |
631 | __s32 buflenout; | |
632 | xfs_bstat_t buf[GRABSZ]; | |
633 | char fname[64]; | |
634 | char *tname; | |
635 | jdm_fshandle_t *fshandlep; | |
636 | xfs_ino_t lastino = startino; | |
637 | ||
638 | fsrprintf(_("%s start inode=%llu\n"), mntdir, | |
639 | (unsigned long long)startino); | |
640 | ||
641 | fshandlep = jdm_getfshandle( mntdir ); | |
642 | if ( ! fshandlep ) { | |
643 | fsrprintf(_("unable to get handle: %s: %s\n"), | |
644 | mntdir, strerror( errno )); | |
645 | return -1; | |
646 | } | |
647 | ||
648 | if ((fsfd = open(mntdir, O_RDONLY)) < 0) { | |
649 | fsrprintf(_("unable to open: %s: %s\n"), | |
650 | mntdir, strerror( errno )); | |
651 | return -1; | |
652 | } | |
653 | ||
654 | if (xfs_getgeom(fsfd, &fsgeom) < 0 ) { | |
655 | fsrprintf(_("Skipping %s: could not get XFS geometry\n"), | |
656 | mntdir); | |
657 | return -1; | |
658 | } | |
659 | ||
660 | tmp_init(mntdir); | |
661 | ||
662 | while ((ret = xfs_bulkstat(fsfd, | |
663 | &lastino, GRABSZ, &buf[0], &buflenout) == 0)) { | |
664 | xfs_bstat_t *p; | |
665 | xfs_bstat_t *endp; | |
666 | ||
667 | if (buflenout == 0) | |
668 | goto out0; | |
669 | ||
670 | /* Each loop through, defrag targetrange percent of the files */ | |
671 | count = (buflenout * targetrange) / 100; | |
672 | ||
673 | qsort((char *)buf, buflenout, sizeof(struct xfs_bstat), cmp); | |
674 | ||
675 | for (p = buf, endp = (buf + buflenout); p < endp ; p++) { | |
676 | /* Do some obvious checks now */ | |
677 | if (((p->bs_mode & S_IFMT) != S_IFREG) || | |
678 | (p->bs_extents < 2)) | |
679 | continue; | |
680 | ||
681 | if ((fd = jdm_open(fshandlep, p, O_RDWR)) < 0) { | |
682 | /* This probably means the file was | |
683 | * removed while in progress of handling | |
684 | * it. Just quietly ignore this file. | |
685 | */ | |
686 | if (dflag) | |
687 | fsrprintf(_("could not open: " | |
688 | "inode %llu\n"), p->bs_ino); | |
689 | continue; | |
690 | } | |
691 | ||
692 | /* Don't know the pathname, so make up something */ | |
693 | sprintf(fname, "ino=%lld", (long long)p->bs_ino); | |
694 | ||
695 | /* Get a tmp file name */ | |
696 | tname = tmp_next(mntdir); | |
697 | ||
698 | ret = fsrfile_common(fname, tname, mntdir, fd, p); | |
699 | ||
700 | leftoffino = p->bs_ino; | |
701 | ||
702 | close(fd); | |
703 | ||
704 | if (ret == 0) { | |
705 | if (--count <= 0) | |
706 | break; | |
707 | } | |
708 | } | |
709 | if (endtime && endtime < time(0)) { | |
710 | tmp_close(mntdir); | |
711 | close(fsfd); | |
712 | fsrall_cleanup(1); | |
713 | exit(1); | |
714 | } | |
715 | } | |
716 | if (ret < 0) | |
717 | fsrprintf(_("%s: xfs_bulkstat: %s\n"), progname, strerror(errno)); | |
718 | out0: | |
719 | tmp_close(mntdir); | |
720 | close(fsfd); | |
721 | return 0; | |
722 | } | |
723 | ||
724 | /* | |
725 | * To compare bstat structs for qsort. | |
726 | */ | |
727 | int | |
728 | cmp(const void *s1, const void *s2) | |
729 | { | |
730 | return( ((xfs_bstat_t *)s2)->bs_extents - | |
731 | ((xfs_bstat_t *)s1)->bs_extents); | |
732 | ||
733 | } | |
734 | ||
735 | /* | |
736 | * reorganize by directory hierarchy. | |
737 | * Stay in dev (a restriction based on structure of this program -- either | |
738 | * call efs_{n,u}mount() around each file, something smarter or this) | |
739 | */ | |
740 | static void | |
741 | fsrdir(char *dirname) | |
742 | { | |
743 | fsrprintf(_("%s: Directory defragmentation not supported\n"), dirname); | |
744 | } | |
745 | ||
746 | /* | |
747 | * Sets up the defragmentation of a file based on the | |
748 | * filepath. It collects the bstat information, does | |
749 | * an open on the file and passes this all to fsrfile_common. | |
750 | */ | |
751 | static int | |
752 | fsrfile(char *fname, xfs_ino_t ino) | |
753 | { | |
754 | xfs_bstat_t statbuf; | |
755 | jdm_fshandle_t *fshandlep; | |
756 | int fd, fsfd; | |
757 | int error = 0; | |
758 | char *tname; | |
759 | ||
760 | fshandlep = jdm_getfshandle(getparent (fname) ); | |
761 | if (! fshandlep) { | |
762 | fsrprintf(_("unable to construct sys handle for %s: %s\n"), | |
763 | fname, strerror(errno)); | |
764 | return -1; | |
765 | } | |
766 | ||
767 | /* | |
768 | * Need to open something on the same filesystem as the | |
769 | * file. Open the parent. | |
770 | */ | |
771 | fsfd = open(getparent(fname), O_RDONLY); | |
772 | if (fsfd < 0) { | |
773 | fsrprintf(_("unable to open sys handle for %s: %s\n"), | |
774 | fname, strerror(errno)); | |
775 | return -1; | |
776 | } | |
777 | ||
778 | if ((xfs_bulkstat_single(fsfd, &ino, &statbuf)) < 0) { | |
779 | fsrprintf(_("unable to get bstat on %s: %s\n"), | |
780 | fname, strerror(errno)); | |
781 | close(fsfd); | |
782 | return -1; | |
783 | } | |
784 | ||
785 | fd = jdm_open( fshandlep, &statbuf, O_RDWR); | |
786 | if (fd < 0) { | |
787 | fsrprintf(_("unable to open handle %s: %s\n"), | |
788 | fname, strerror(errno)); | |
789 | close(fsfd); | |
790 | return -1; | |
791 | } | |
792 | ||
793 | /* Get the fs geometry */ | |
794 | if (xfs_getgeom(fsfd, &fsgeom) < 0 ) { | |
795 | fsrprintf(_("Unable to get geom on fs for: %s\n"), fname); | |
796 | close(fsfd); | |
797 | return -1; | |
798 | } | |
799 | ||
800 | close(fsfd); | |
801 | ||
802 | tname = gettmpname(fname); | |
803 | ||
804 | if (tname) | |
805 | error = fsrfile_common(fname, tname, NULL, fd, &statbuf); | |
806 | ||
807 | close(fd); | |
808 | ||
809 | return error; | |
810 | } | |
811 | ||
812 | ||
813 | /* | |
814 | * This is the common defrag code for either a full fs | |
815 | * defragmentation or a single file. Check as much as | |
816 | * possible with the file, fork a process to setuid to the | |
817 | * target file owner's uid and defragment the file. | |
818 | * This is done so the new extents created in a tmp file are | |
819 | * reflected in the owners' quota without having to do any | |
820 | * special code in the kernel. When the existing extents | |
821 | * are removed, the quotas will be correct. It's ugly but | |
822 | * it saves us from doing some quota re-construction in | |
823 | * the extent swap. The price is that the defragmentation | |
824 | * will fail if the owner of the target file is already at | |
825 | * their quota limit. | |
826 | */ | |
827 | static int | |
828 | fsrfile_common( | |
829 | char *fname, | |
830 | char *tname, | |
831 | char *fsname, | |
832 | int fd, | |
833 | xfs_bstat_t *statp) | |
834 | { | |
835 | int error; | |
836 | struct statvfs64 vfss; | |
837 | struct fsxattr fsx; | |
838 | unsigned long bsize; | |
839 | ||
840 | if (vflag) | |
841 | fsrprintf("%s\n", fname); | |
842 | ||
843 | if (fsync(fd) < 0) { | |
844 | fsrprintf(_("sync failed: %s: %s\n"), fname, strerror(errno)); | |
845 | return -1; | |
846 | } | |
847 | ||
848 | if (statp->bs_size == 0) { | |
849 | if (vflag) | |
850 | fsrprintf(_("%s: zero size, ignoring\n"), fname); | |
851 | return(0); | |
852 | } | |
853 | ||
854 | /* Check if a mandatory lock is set on the file to try and | |
855 | * avoid blocking indefinitely on the reads later. Note that | |
856 | * someone could still set a mandatory lock after this check | |
857 | * but before all reads have completed to block fsr reads. | |
858 | * This change just closes the window a bit. | |
859 | */ | |
860 | if ( (statp->bs_mode & S_ISGID) && ( ! (statp->bs_mode&S_IXGRP) ) ) { | |
861 | struct flock fl; | |
862 | ||
863 | fl.l_type = F_RDLCK; | |
864 | fl.l_whence = SEEK_SET; | |
865 | fl.l_start = (off_t)0; | |
866 | fl.l_len = 0; | |
867 | if ((fcntl(fd, F_GETLK, &fl)) < 0 ) { | |
868 | if (vflag) | |
869 | fsrprintf(_("locking check failed: %s\n"), | |
870 | fname); | |
871 | return(-1); | |
872 | } | |
873 | if (fl.l_type != F_UNLCK) { | |
874 | /* Mandatory lock is set */ | |
875 | if (vflag) | |
876 | fsrprintf(_("mandatory lock: %s: ignoring\n"), | |
877 | fname); | |
878 | return(-1); | |
879 | } | |
880 | } | |
881 | ||
882 | /* | |
883 | * Check if there is room to copy the file. | |
884 | * | |
885 | * Note that xfs_bstat.bs_blksize returns the filesystem blocksize, | |
886 | * not the optimal I/O size as struct stat. | |
887 | */ | |
888 | if (statvfs64(fsname ? fsname : fname, &vfss) < 0) { | |
889 | fsrprintf(_("unable to get fs stat on %s: %s\n"), | |
890 | fname, strerror(errno)); | |
891 | return -1; | |
892 | } | |
893 | bsize = vfss.f_frsize ? vfss.f_frsize : vfss.f_bsize; | |
894 | if (statp->bs_blksize * statp->bs_blocks > | |
895 | vfss.f_bfree * bsize - minimumfree) { | |
896 | fsrprintf(_("insufficient freespace for: %s: " | |
897 | "size=%lld: ignoring\n"), fname, | |
898 | statp->bs_blksize * statp->bs_blocks); | |
899 | return 1; | |
900 | } | |
901 | ||
902 | if ((ioctl(fd, XFS_IOC_FSGETXATTR, &fsx)) < 0) { | |
903 | fsrprintf(_("failed to get inode attrs: %s\n"), fname); | |
904 | return(-1); | |
905 | } | |
906 | if (fsx.fsx_xflags & (XFS_XFLAG_IMMUTABLE|XFS_XFLAG_APPEND)) { | |
907 | if (vflag) | |
908 | fsrprintf(_("%s: immutable/append, ignoring\n"), fname); | |
909 | return(0); | |
910 | } | |
911 | if (fsx.fsx_xflags & XFS_XFLAG_NODEFRAG) { | |
912 | if (vflag) | |
913 | fsrprintf(_("%s: marked as don't defrag, ignoring\n"), | |
914 | fname); | |
915 | return(0); | |
916 | } | |
917 | if (fsx.fsx_xflags & XFS_XFLAG_REALTIME) { | |
918 | if (xfs_getrt(fd, &vfss) < 0) { | |
919 | fsrprintf(_("cannot get realtime geometry for: %s\n"), | |
920 | fname); | |
921 | return(-1); | |
922 | } | |
923 | if (statp->bs_size > ((vfss.f_bfree * bsize) - minimumfree)) { | |
924 | fsrprintf(_("low on realtime free space: %s: " | |
925 | "ignoring file\n"), fname); | |
926 | return(-1); | |
927 | } | |
928 | } | |
929 | ||
930 | if ((RealUid != ROOT) && (RealUid != statp->bs_uid)) { | |
931 | fsrprintf(_("cannot open: %s: Permission denied\n"), fname); | |
932 | return -1; | |
933 | } | |
934 | ||
935 | /* | |
936 | * Previously the code forked here, & the child changed it's uid to | |
937 | * that of the file's owner and then called packfile(), to keep | |
938 | * quota counts correct. (defragged files could use fewer blocks). | |
939 | * | |
940 | * Instead, just fchown() the temp file to the uid,gid of the | |
941 | * file we're defragging, in packfile(). | |
942 | */ | |
943 | ||
944 | if ((error = packfile(fname, tname, fd, statp, &fsx))) | |
945 | return error; | |
946 | return -1; /* no error */ | |
947 | } | |
948 | ||
949 | ||
950 | /* | |
951 | * Do the defragmentation of a single file. | |
952 | * We already are pretty sure we can and want to | |
953 | * defragment the file. Create the tmp file, copy | |
954 | * the data (maintaining holes) and call the kernel | |
955 | * extent swap routinte. | |
956 | */ | |
957 | static int | |
958 | packfile(char *fname, char *tname, int fd, | |
959 | xfs_bstat_t *statp, struct fsxattr *fsxp) | |
960 | { | |
961 | int tfd; | |
962 | int srval; | |
963 | int nextents, extent, cur_nextents, new_nextents; | |
964 | unsigned blksz_dio; | |
965 | unsigned dio_min; | |
966 | struct dioattr dio; | |
967 | static xfs_swapext_t sx; | |
968 | struct xfs_flock64 space; | |
969 | off64_t cnt, pos; | |
970 | void *fbuf; | |
971 | int ct, wc, wc_b4; | |
972 | char ffname[SMBUFSZ]; | |
973 | int ffd = -1; | |
974 | ||
975 | /* | |
976 | * Work out the extent map - nextents will be set to the | |
977 | * minimum number of extents needed for the file (taking | |
978 | * into account holes), cur_nextents is the current number | |
979 | * of extents. | |
980 | */ | |
981 | nextents = read_fd_bmap(fd, statp, &cur_nextents); | |
982 | ||
983 | if (cur_nextents == 1 || cur_nextents <= nextents) { | |
984 | if (vflag) | |
985 | fsrprintf(_("%s already fully defragmented.\n"), fname); | |
986 | return 1; /* indicates no change/no error */ | |
987 | } | |
988 | ||
989 | if (dflag) | |
990 | fsrprintf(_("%s extents=%d can_save=%d tmp=%s\n"), | |
991 | fname, cur_nextents, (cur_nextents - nextents), | |
992 | tname); | |
993 | ||
994 | if ((tfd = open(tname, openopts, 0666)) < 0) { | |
995 | if (vflag) | |
996 | fsrprintf(_("could not open tmp file: %s: %s\n"), | |
997 | tname, strerror(errno)); | |
998 | return -1; | |
999 | } | |
1000 | unlink(tname); | |
1001 | ||
1002 | /* Setup extended attributes */ | |
1003 | if (statp->bs_xflags & XFS_XFLAG_HASATTR) { | |
1004 | if (fsetxattr(tfd, "user.X", "X", 1, XATTR_CREATE) != 0) { | |
1005 | fsrprintf(_("could not set ATTR on tmp: %s:\n"), tname); | |
1006 | close(tfd); | |
1007 | return -1; | |
1008 | } | |
1009 | if (dflag) | |
1010 | fsrprintf(_("%s set temp attr\n"), tname); | |
1011 | } | |
1012 | ||
1013 | /* Setup extended inode flags, project identifier, etc */ | |
1014 | if (fsxp->fsx_xflags || fsxp->fsx_projid) { | |
1015 | if (ioctl(tfd, XFS_IOC_FSSETXATTR, fsxp) < 0) { | |
1016 | fsrprintf(_("could not set inode attrs on tmp: %s\n"), | |
1017 | tname); | |
1018 | close(tfd); | |
1019 | return -1; | |
1020 | } | |
1021 | } | |
1022 | ||
1023 | if ((ioctl(tfd, XFS_IOC_DIOINFO, &dio)) < 0 ) { | |
1024 | fsrprintf(_("could not get DirectIO info on tmp: %s\n"), tname); | |
1025 | close(tfd); | |
1026 | return -1; | |
1027 | } | |
1028 | ||
1029 | dio_min = dio.d_miniosz; | |
1030 | if (statp->bs_size <= dio_min) { | |
1031 | blksz_dio = dio_min; | |
1032 | } else { | |
1033 | blksz_dio = min(dio.d_maxiosz, BUFFER_MAX - pagesize); | |
1034 | if (argv_blksz_dio != 0) | |
1035 | blksz_dio = min(argv_blksz_dio, blksz_dio); | |
1036 | blksz_dio = (min(statp->bs_size, blksz_dio) / dio_min) * dio_min; | |
1037 | } | |
1038 | ||
1039 | if (dflag) { | |
1040 | fsrprintf(_("DEBUG: " | |
1041 | "fsize=%lld blsz_dio=%d d_min=%d d_max=%d pgsz=%d\n"), | |
1042 | statp->bs_size, blksz_dio, dio.d_miniosz, | |
1043 | dio.d_maxiosz, pagesize); | |
1044 | } | |
1045 | ||
1046 | if (!(fbuf = (char *)memalign(dio.d_mem, blksz_dio))) { | |
1047 | fsrprintf(_("could not allocate buf: %s\n"), tname); | |
1048 | close(tfd); | |
1049 | return -1; | |
1050 | } | |
1051 | ||
1052 | if (nfrags) { | |
1053 | /* Create new tmp file in same AG as first */ | |
1054 | sprintf(ffname, "%s.frag", tname); | |
1055 | ||
1056 | /* Open the new file for sync writes */ | |
1057 | if ((ffd = open(ffname, openopts, 0666)) < 0) { | |
1058 | fsrprintf(_("could not open fragfile: %s : %s\n"), | |
1059 | ffname, strerror(errno)); | |
1060 | close(tfd); | |
1061 | free(fbuf); | |
1062 | return -1; | |
1063 | } | |
1064 | unlink(ffname); | |
1065 | } | |
1066 | ||
1067 | /* Loop through block map allocating new extents */ | |
1068 | for (extent = 0; extent < nextents; extent++) { | |
1069 | pos = outmap[extent].bmv_offset; | |
1070 | if (outmap[extent].bmv_block == -1) { | |
1071 | space.l_whence = SEEK_SET; | |
1072 | space.l_start = pos; | |
1073 | space.l_len = outmap[extent].bmv_length; | |
1074 | if (ioctl(tfd, XFS_IOC_UNRESVSP64, &space) < 0) { | |
1075 | fsrprintf(_("could not trunc tmp %s\n"), | |
1076 | tname); | |
1077 | } | |
1078 | lseek64(tfd, outmap[extent].bmv_length, SEEK_CUR); | |
1079 | continue; | |
1080 | } else if (outmap[extent].bmv_length == 0) { | |
1081 | /* to catch holes at the beginning of the file */ | |
1082 | continue; | |
1083 | } | |
1084 | if (! nfrags) { | |
1085 | space.l_whence = SEEK_CUR; | |
1086 | space.l_start = 0; | |
1087 | space.l_len = outmap[extent].bmv_length; | |
1088 | ||
1089 | if (ioctl(tfd, XFS_IOC_RESVSP64, &space) < 0) { | |
1090 | fsrprintf(_("could not pre-allocate tmp space:" | |
1091 | " %s\n"), tname); | |
1092 | close(tfd); | |
1093 | free(fbuf); | |
1094 | return -1; | |
1095 | } | |
1096 | lseek64(tfd, outmap[extent].bmv_length, SEEK_CUR); | |
1097 | } | |
1098 | } /* end of space allocation loop */ | |
1099 | ||
1100 | if (lseek64(tfd, 0, SEEK_SET)) { | |
1101 | fsrprintf(_("Couldn't rewind on temporary file\n")); | |
1102 | close(tfd); | |
1103 | free(fbuf); | |
1104 | return -1; | |
1105 | } | |
1106 | ||
1107 | /* Check if the temporary file has fewer extents */ | |
1108 | new_nextents = getnextents(tfd); | |
1109 | if (dflag) | |
1110 | fsrprintf(_("Temporary file has %d extents (%d in original)\n"), new_nextents, cur_nextents); | |
1111 | if (cur_nextents <= new_nextents) { | |
1112 | if (vflag) | |
1113 | fsrprintf(_("No improvement will be made (skipping): %s\n"), fname); | |
1114 | free(fbuf); | |
1115 | close(tfd); | |
1116 | return 1; /* no change/no error */ | |
1117 | } | |
1118 | ||
1119 | /* Loop through block map copying the file. */ | |
1120 | for (extent = 0; extent < nextents; extent++) { | |
1121 | pos = outmap[extent].bmv_offset; | |
1122 | if (outmap[extent].bmv_block == -1) { | |
1123 | lseek64(tfd, outmap[extent].bmv_length, SEEK_CUR); | |
1124 | lseek64(fd, outmap[extent].bmv_length, SEEK_CUR); | |
1125 | continue; | |
1126 | } else if (outmap[extent].bmv_length == 0) { | |
1127 | /* to catch holes at the beginning of the file */ | |
1128 | continue; | |
1129 | } | |
1130 | for (cnt = outmap[extent].bmv_length; cnt > 0; | |
1131 | cnt -= ct, pos += ct) { | |
1132 | if (nfrags && --nfrags) { | |
1133 | ct = min(cnt, dio_min); | |
1134 | } else if (cnt % dio_min == 0) { | |
1135 | ct = min(cnt, blksz_dio); | |
1136 | } else { | |
1137 | ct = min(cnt + dio_min - (cnt % dio_min), | |
1138 | blksz_dio); | |
1139 | } | |
1140 | ct = read(fd, fbuf, ct); | |
1141 | if (ct == 0) { | |
1142 | /* EOF, stop trying to read */ | |
1143 | extent = nextents; | |
1144 | break; | |
1145 | } | |
1146 | /* Ensure we do direct I/O to correct block | |
1147 | * boundaries. | |
1148 | */ | |
1149 | if (ct % dio_min != 0) { | |
1150 | wc = ct + dio_min - (ct % dio_min); | |
1151 | } else { | |
1152 | wc = ct; | |
1153 | } | |
1154 | wc_b4 = wc; | |
1155 | if (ct < 0 || ((wc = write(tfd, fbuf, wc)) != wc_b4)) { | |
1156 | if (ct < 0) | |
1157 | fsrprintf(_("bad read of %d bytes " | |
1158 | "from %s: %s\n"), wc_b4, | |
1159 | fname, strerror(errno)); | |
1160 | else if (wc < 0) | |
1161 | fsrprintf(_("bad write of %d bytes " | |
1162 | "to %s: %s\n"), wc_b4, | |
1163 | tname, strerror(errno)); | |
1164 | else { | |
1165 | /* | |
1166 | * Might be out of space | |
1167 | * | |
1168 | * Try to finish write | |
1169 | */ | |
1170 | int resid = ct-wc; | |
1171 | ||
1172 | if ((wc = write(tfd, ((char *)fbuf)+wc, | |
1173 | resid)) == resid) { | |
1174 | /* worked on second attempt? */ | |
1175 | continue; | |
1176 | } | |
1177 | else if (wc < 0) { | |
1178 | fsrprintf(_("bad write2 of %d " | |
1179 | "bytes to %s: %s\n"), | |
1180 | resid, tname, | |
1181 | strerror(errno)); | |
1182 | } else { | |
1183 | fsrprintf(_("bad copy to %s\n"), | |
1184 | tname); | |
1185 | } | |
1186 | } | |
1187 | free(fbuf); | |
1188 | close(tfd); | |
1189 | return -1; | |
1190 | } | |
1191 | if (nfrags) { | |
1192 | /* Do a matching write to the tmp file */ | |
9a5c6ff4 | 1193 | wc = wc_b4; |
c988ea91 CH |
1194 | if (((wc = write(ffd, fbuf, wc)) != wc_b4)) { |
1195 | fsrprintf(_("bad write of %d bytes " | |
1196 | "to %s: %s\n"), | |
1197 | wc_b4, ffname, strerror(errno)); | |
1198 | } | |
1199 | } | |
1200 | } | |
1201 | } | |
1202 | ftruncate64(tfd, statp->bs_size); | |
1203 | if (ffd > 0) close(ffd); | |
1204 | fsync(tfd); | |
1205 | ||
1206 | free(fbuf); | |
1207 | ||
1208 | sx.sx_stat = *statp; /* struct copy */ | |
1209 | sx.sx_version = XFS_SX_VERSION; | |
1210 | sx.sx_fdtarget = fd; | |
1211 | sx.sx_fdtmp = tfd; | |
1212 | sx.sx_offset = 0; | |
1213 | sx.sx_length = statp->bs_size; | |
1214 | ||
1215 | /* switch to the owner's id, to keep quota in line */ | |
1216 | if (fchown(tfd, statp->bs_uid, statp->bs_gid) < 0) { | |
1217 | if (vflag) | |
1218 | fsrprintf(_("failed to fchown tmpfile %s: %s\n"), | |
1219 | tname, strerror(errno)); | |
1220 | close(tfd); | |
1221 | return -1; | |
1222 | } | |
1223 | ||
1224 | /* Swap the extents */ | |
1225 | srval = xfs_swapext(fd, &sx); | |
1226 | if (srval < 0) { | |
1227 | if (errno == ENOTSUP) { | |
1228 | if (vflag || dflag) | |
1229 | fsrprintf(_("%s: file type not supported\n"), fname); | |
1230 | } else if (errno == EFAULT) { | |
1231 | /* The file has changed since we started the copy */ | |
1232 | if (vflag || dflag) | |
1233 | fsrprintf(_("%s: file modified defrag aborted\n"), | |
1234 | fname); | |
1235 | } else if (errno == EBUSY) { | |
1236 | /* Timestamp has changed or mmap'ed file */ | |
1237 | if (vflag || dflag) | |
1238 | fsrprintf(_("%s: file busy\n"), fname); | |
1239 | } else { | |
1240 | fsrprintf(_("XFS_IOC_SWAPEXT failed: %s: %s\n"), | |
1241 | fname, strerror(errno)); | |
1242 | } | |
1243 | close(tfd); | |
1244 | return -1; | |
1245 | } | |
1246 | ||
1247 | /* Report progress */ | |
1248 | if (vflag) | |
1249 | fsrprintf(_("extents before:%d after:%d %s %s\n"), | |
1250 | cur_nextents, new_nextents, | |
1251 | (new_nextents <= nextents ? "DONE" : " " ), | |
1252 | fname); | |
1253 | close(tfd); | |
1254 | return 0; | |
1255 | } | |
1256 | ||
1257 | char * | |
1258 | gettmpname(char *fname) | |
1259 | { | |
1260 | static char buf[PATH_MAX+1]; | |
1261 | char sbuf[SMBUFSZ]; | |
1262 | char *ptr; | |
1263 | ||
1264 | sprintf(sbuf, "/.fsr%d", getpid()); | |
1265 | ||
1266 | strcpy(buf, fname); | |
1267 | ptr = strrchr(buf, '/'); | |
1268 | if (ptr) { | |
1269 | *ptr = '\0'; | |
1270 | } else { | |
1271 | strcpy(buf, "."); | |
1272 | } | |
1273 | ||
1274 | if ((strlen(buf) + strlen (sbuf)) > PATH_MAX) { | |
1275 | fsrprintf(_("tmp file name too long: %s\n"), fname); | |
1276 | return(NULL); | |
1277 | } | |
1278 | ||
1279 | strcat(buf, sbuf); | |
1280 | ||
1281 | return(buf); | |
1282 | } | |
1283 | ||
1284 | char * | |
1285 | getparent(char *fname) | |
1286 | { | |
1287 | static char buf[PATH_MAX+1]; | |
1288 | char *ptr; | |
1289 | ||
1290 | strcpy(buf, fname); | |
1291 | ptr = strrchr(buf, '/'); | |
1292 | if (ptr) { | |
1293 | if (ptr == &buf[0]) | |
1294 | ++ptr; | |
1295 | *ptr = '\0'; | |
1296 | } else { | |
1297 | strcpy(buf, "."); | |
1298 | } | |
1299 | ||
1300 | return(buf); | |
1301 | } | |
1302 | ||
1303 | /* | |
1304 | * Read in block map of the input file, coalesce contiguous | |
1305 | * extents into a single range, keep all holes. Convert from 512 byte | |
1306 | * blocks to bytes. | |
1307 | * | |
1308 | * This code was borrowed from mv.c with some minor mods. | |
1309 | */ | |
1310 | #define MAPSIZE 128 | |
1311 | #define OUTMAP_SIZE_INCREMENT MAPSIZE | |
1312 | ||
1313 | int read_fd_bmap(int fd, xfs_bstat_t *sin, int *cur_nextents) | |
1314 | { | |
1315 | int i, cnt; | |
1316 | struct getbmap map[MAPSIZE]; | |
1317 | ||
1318 | #define BUMP_CNT \ | |
1319 | if (++cnt >= outmap_size) { \ | |
1320 | outmap_size += OUTMAP_SIZE_INCREMENT; \ | |
1321 | outmap = (struct getbmap *)realloc(outmap, \ | |
1322 | outmap_size*sizeof(*outmap)); \ | |
1323 | if (outmap == NULL) { \ | |
1324 | fsrprintf(_("realloc failed: %s\n"), \ | |
1325 | strerror(errno)); \ | |
1326 | exit(1); \ | |
1327 | } \ | |
1328 | } | |
1329 | ||
1330 | /* Initialize the outmap array. It always grows - never shrinks. | |
1331 | * Left-over memory allocation is saved for the next files. | |
1332 | */ | |
1333 | if (outmap_size == 0) { | |
1334 | outmap_size = OUTMAP_SIZE_INCREMENT; /* Initial size */ | |
1335 | outmap = (struct getbmap *)malloc(outmap_size*sizeof(*outmap)); | |
1336 | if (!outmap) { | |
1337 | fsrprintf(_("malloc failed: %s\n"), | |
1338 | strerror(errno)); | |
1339 | exit(1); | |
1340 | } | |
1341 | } | |
1342 | ||
1343 | outmap[0].bmv_block = 0; | |
1344 | outmap[0].bmv_offset = 0; | |
1345 | outmap[0].bmv_length = sin->bs_size; | |
1346 | ||
1347 | /* | |
1348 | * If a non regular file is involved then forget holes | |
1349 | */ | |
1350 | ||
1351 | if (!S_ISREG(sin->bs_mode)) | |
1352 | return(1); | |
1353 | ||
1354 | outmap[0].bmv_length = 0; | |
1355 | ||
1356 | map[0].bmv_offset = 0; | |
1357 | map[0].bmv_block = 0; | |
1358 | map[0].bmv_entries = 0; | |
1359 | map[0].bmv_count = MAPSIZE; | |
1360 | map[0].bmv_length = -1; | |
1361 | ||
1362 | cnt = 0; | |
1363 | *cur_nextents = 0; | |
1364 | ||
1365 | do { | |
1366 | if (ioctl(fd, XFS_IOC_GETBMAP, map) < 0) { | |
1367 | fsrprintf(_("failed reading extents: inode %llu"), | |
1368 | (unsigned long long)sin->bs_ino); | |
1369 | exit(1); | |
1370 | } | |
1371 | ||
1372 | /* Concatenate extents together and replicate holes into | |
1373 | * the output map. | |
1374 | */ | |
1375 | *cur_nextents += map[0].bmv_entries; | |
1376 | for (i = 0; i < map[0].bmv_entries; i++) { | |
1377 | if (map[i + 1].bmv_block == -1) { | |
1378 | BUMP_CNT; | |
1379 | outmap[cnt] = map[i+1]; | |
1380 | } else if (outmap[cnt].bmv_block == -1) { | |
1381 | BUMP_CNT; | |
1382 | outmap[cnt] = map[i+1]; | |
1383 | } else { | |
1384 | outmap[cnt].bmv_length += map[i + 1].bmv_length; | |
1385 | } | |
1386 | } | |
1387 | } while (map[0].bmv_entries == (MAPSIZE-1)); | |
1388 | for (i = 0; i <= cnt; i++) { | |
1389 | outmap[i].bmv_offset = BBTOB(outmap[i].bmv_offset); | |
1390 | outmap[i].bmv_length = BBTOB(outmap[i].bmv_length); | |
1391 | } | |
1392 | ||
1393 | outmap[cnt].bmv_length = sin->bs_size - outmap[cnt].bmv_offset; | |
1394 | ||
1395 | return(cnt+1); | |
1396 | } | |
1397 | ||
1398 | /* | |
1399 | * Read the block map and return the number of extents. | |
1400 | */ | |
1401 | int | |
1402 | getnextents(int fd) | |
1403 | { | |
1404 | int nextents; | |
1405 | struct getbmap map[MAPSIZE]; | |
1406 | ||
1407 | map[0].bmv_offset = 0; | |
1408 | map[0].bmv_block = 0; | |
1409 | map[0].bmv_entries = 0; | |
1410 | map[0].bmv_count = MAPSIZE; | |
1411 | map[0].bmv_length = -1; | |
1412 | ||
1413 | nextents = 0; | |
1414 | ||
1415 | do { | |
1416 | if (ioctl(fd,XFS_IOC_GETBMAP, map) < 0) { | |
1417 | fsrprintf(_("failed reading extents")); | |
1418 | exit(1); | |
1419 | } | |
1420 | ||
1421 | nextents += map[0].bmv_entries; | |
1422 | } while (map[0].bmv_entries == (MAPSIZE-1)); | |
1423 | ||
1424 | return(nextents); | |
1425 | } | |
1426 | ||
1427 | /* | |
1428 | * Get the fs geometry | |
1429 | */ | |
1430 | int | |
1431 | xfs_getgeom(int fd, xfs_fsop_geom_v1_t * fsgeom) | |
1432 | { | |
1433 | if (xfs_fsgeometry(fd, fsgeom) < 0) { | |
1434 | return -1; | |
1435 | } | |
1436 | return 0; | |
1437 | } | |
1438 | ||
1439 | /* | |
1440 | * Get xfs realtime space information | |
1441 | */ | |
1442 | int | |
1443 | xfs_getrt(int fd, struct statvfs64 *sfbp) | |
1444 | { | |
1445 | unsigned long bsize; | |
1446 | unsigned long factor; | |
1447 | xfs_fsop_counts_t cnt; | |
1448 | ||
1449 | if (!fsgeom.rtblocks) | |
1450 | return -1; | |
1451 | ||
1452 | if (xfs_fscounts(fd, &cnt) < 0) { | |
1453 | close(fd); | |
1454 | return -1; | |
1455 | } | |
1456 | bsize = (sfbp->f_frsize ? sfbp->f_frsize : sfbp->f_bsize); | |
1457 | factor = fsgeom.blocksize / bsize; /* currently this is == 1 */ | |
1458 | sfbp->f_bfree = (cnt.freertx * fsgeom.rtextsize) * factor; | |
1459 | return 0; | |
1460 | } | |
1461 | ||
1462 | int | |
1463 | fsrprintf(const char *fmt, ...) | |
1464 | { | |
1465 | va_list ap; | |
1466 | ||
1467 | va_start(ap, fmt); | |
1468 | if (gflag) { | |
1469 | static int didopenlog; | |
1470 | if (!didopenlog) { | |
1471 | openlog("fsr", LOG_PID, LOG_USER); | |
1472 | didopenlog = 1; | |
1473 | } | |
1474 | vsyslog(LOG_INFO, fmt, ap); | |
1475 | } else | |
1476 | vprintf(fmt, ap); | |
1477 | va_end(ap); | |
1478 | return 0; | |
1479 | } | |
1480 | ||
1481 | /* | |
1482 | * emulate getmntany | |
1483 | */ | |
1484 | static int | |
1485 | getmntany(FILE *fp, struct mntent *mp, struct mntent *mpref, struct stat64 *s) | |
1486 | { | |
1487 | struct mntent *t; | |
1488 | struct stat64 ms; | |
1489 | ||
1490 | while ((t = getmntent(fp))) { | |
1491 | if (mpref->mnt_fsname) { /* device */ | |
1492 | if (stat64(t->mnt_fsname, &ms) < 0) | |
1493 | continue; | |
1494 | if (s->st_rdev != ms.st_rdev) | |
1495 | continue; | |
1496 | } | |
1497 | if (mpref->mnt_dir) { /* mount point */ | |
1498 | if (stat64(t->mnt_dir, &ms) < 0) | |
1499 | continue; | |
1500 | if (s->st_ino != ms.st_ino || s->st_dev != ms.st_dev) | |
1501 | continue; | |
1502 | } | |
1503 | *mp = *t; | |
1504 | break; | |
1505 | } | |
1506 | return (t != NULL); | |
1507 | } | |
1508 | ||
1509 | ||
1510 | /* | |
1511 | * Initialize a directory for tmp file use. This is used | |
1512 | * by the full filesystem defragmentation when we're walking | |
1513 | * the inodes and do not know the path for the individual | |
1514 | * files. Multiple directories are used to spread out the | |
1515 | * tmp data around to different ag's (since file data is | |
1516 | * usually allocated to the same ag as the directory and | |
1517 | * directories allocated round robin from the same | |
1518 | * parent directory). | |
1519 | */ | |
1520 | static void | |
1521 | tmp_init(char *mnt) | |
1522 | { | |
1523 | int i; | |
1524 | static char buf[SMBUFSZ]; | |
1525 | mode_t mask; | |
1526 | ||
1527 | tmp_agi = 0; | |
1528 | sprintf(buf, "%s/.fsr", mnt); | |
1529 | ||
1530 | mask = umask(0); | |
1531 | if (mkdir(buf, 0700) < 0) { | |
1532 | if (errno == EEXIST) { | |
1533 | if (dflag) | |
1534 | fsrprintf(_("tmpdir already exists: %s\n"), | |
1535 | buf); | |
1536 | } else { | |
1537 | fsrprintf(_("could not create tmpdir: %s: %s\n"), | |
1538 | buf, strerror(errno)); | |
1539 | exit(-1); | |
1540 | } | |
1541 | } | |
1542 | for (i=0; i < fsgeom.agcount; i++) { | |
1543 | sprintf(buf, "%s/.fsr/ag%d", mnt, i); | |
1544 | if (mkdir(buf, 0777) < 0) { | |
1545 | if (errno == EEXIST) { | |
1546 | if (dflag) | |
1547 | fsrprintf( | |
1548 | _("tmpdir already exists: %s\n"), buf); | |
1549 | } else { | |
1550 | fsrprintf(_("cannot create tmpdir: %s: %s\n"), | |
1551 | buf, strerror(errno)); | |
1552 | exit(-1); | |
1553 | } | |
1554 | } | |
1555 | } | |
1556 | (void)umask(mask); | |
1557 | return; | |
1558 | } | |
1559 | ||
1560 | static char * | |
1561 | tmp_next(char *mnt) | |
1562 | { | |
1563 | static char buf[SMBUFSZ]; | |
1564 | ||
1565 | sprintf(buf, "%s/.fsr/ag%d/tmp%d", | |
1566 | ( (strcmp(mnt, "/") == 0) ? "" : mnt), | |
1567 | tmp_agi, | |
1568 | getpid()); | |
1569 | ||
1570 | if (++tmp_agi == fsgeom.agcount) | |
1571 | tmp_agi = 0; | |
1572 | ||
1573 | return(buf); | |
1574 | } | |
1575 | ||
1576 | static void | |
1577 | tmp_close(char *mnt) | |
1578 | { | |
1579 | static char buf[SMBUFSZ]; | |
1580 | int i; | |
1581 | ||
1582 | /* No data is ever actually written so we can just do rmdir's */ | |
1583 | for (i=0; i < fsgeom.agcount; i++) { | |
1584 | sprintf(buf, "%s/.fsr/ag%d", mnt, i); | |
1585 | if (rmdir(buf) < 0) { | |
1586 | if (errno != ENOENT) { | |
1587 | fsrprintf( | |
1588 | _("could not remove tmpdir: %s: %s\n"), | |
1589 | buf, strerror(errno)); | |
1590 | } | |
1591 | } | |
1592 | } | |
1593 | sprintf(buf, "%s/.fsr", mnt); | |
1594 | if (rmdir(buf) < 0) { | |
1595 | if (errno != ENOENT) { | |
1596 | fsrprintf(_("could not remove tmpdir: %s: %s\n"), | |
1597 | buf, strerror(errno)); | |
1598 | } | |
1599 | } | |
1600 | } |