summaryrefslogtreecommitdiff
path: root/usr.bin/rsync/uploader.c
blob: 38d5d24fcc7bc69fca78ef8dbc22262a32960c66 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
/*	$Id: uploader.c,v 1.23 2019/08/26 22:22:14 benno Exp $ */
/*
 * Copyright (c) 2019 Kristaps Dzonsons <kristaps@bsd.lv>
 * Copyright (c) 2019 Florian Obser <florian@openbsd.org>
 *
 * Permission to use, copy, modify, and distribute this software for any
 * purpose with or without fee is hereby granted, provided that the above
 * copyright notice and this permission notice appear in all copies.
 *
 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
 */
#include <sys/mman.h>
#include <sys/stat.h>

#include <assert.h>
#include <errno.h>
#include <fcntl.h>
#include <inttypes.h>
#include <math.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <time.h>
#include <unistd.h>

#include "extern.h"

enum	uploadst {
	UPLOAD_FIND_NEXT = 0, /* find next to upload to sender */
	UPLOAD_WRITE_LOCAL, /* wait to write to sender */
	UPLOAD_READ_LOCAL, /* wait to read from local file */
	UPLOAD_FINISHED /* nothing more to do in phase */
};

/*
 * Used to keep track of data flowing from the receiver to the sender.
 * This is managed by the receiver process.
 */
struct	upload {
	enum uploadst	    state;
	char		   *buf; /* if not NULL, pending upload */
	size_t		    bufsz; /* size of buf */
	size_t		    bufmax; /* maximum size of buf */
	size_t		    bufpos; /* position in buf */
	size_t		    idx; /* current transfer index */
	mode_t		    oumask; /* umask for creating files */
	char		   *root; /* destination directory path */
	int		    rootfd; /* destination directory */
	size_t		    csumlen; /* checksum length */
	int		    fdout; /* write descriptor to sender */
	const struct flist *fl; /* file list */
	size_t		    flsz; /* size of file list */
	int		   *newdir; /* non-zero if mkdir'd */
};

/*
 * Log a directory by emitting the file and a trailing slash, just to
 * show the operator that we're a directory.
 */
static void
log_dir(struct sess *sess, const struct flist *f)
{
	size_t	 sz;

	if (sess->opts->server)
		return;
	sz = strlen(f->path);
	assert(sz > 0);
	LOG1("%s%s", f->path, (f->path[sz - 1] == '/') ? "" : "/");
}

/*
 * Log a link by emitting the file and the target, just to show the
 * operator that we're a link.
 */
static void
log_link(struct sess *sess, const struct flist *f)
{

	if (!sess->opts->server)
		LOG1("%s -> %s", f->path, f->link);
}

/*
 * Simply log the filename.
 */
static void
log_file(struct sess *sess, const struct flist *f)
{

	if (!sess->opts->server)
		LOG1("%s", f->path);
}

/*
 * Prepare the overall block set's metadata.
 * We always have at least one block.
 * The block size is an important part of the algorithm.
 * I use the same heuristic as the reference rsync, but implemented in a
 * bit more of a straightforward way.
 * In general, the individual block length is the rounded square root of
 * the total file size.
 * The minimum block length is 700.
 */
static void
init_blkset(struct blkset *p, off_t sz)
{
	double	 v;

	if (sz >= (BLOCK_SIZE_MIN * BLOCK_SIZE_MIN)) {
		/* Simple rounded-up integer square root. */

		v = sqrt(sz);
		p->len = ceil(v);

		/*
		 * Always be a multiple of eight.
		 * There's no reason to do this, but rsync does.
		 */

		if ((p->len % 8) > 0)
			p->len += 8 - (p->len % 8);
	} else
		p->len = BLOCK_SIZE_MIN;

	p->size = sz;
	if ((p->blksz = sz / p->len) == 0)
		p->rem = sz;
	else
		p->rem = sz % p->len;

	/* If we have a remainder, then we need an extra block. */

	if (p->rem)
		p->blksz++;
}

/*
 * For each block, prepare the block's metadata.
 * We use the mapped "map" file to set our checksums.
 */
static void
init_blk(struct blk *p, const struct blkset *set, off_t offs,
	size_t idx, const void *map, const struct sess *sess)
{

	assert(map != MAP_FAILED);

	/* Block length inherits for all but the last. */

	p->idx = idx;
	p->len = idx < set->blksz - 1 ? set->len : set->rem;
	p->offs = offs;

	p->chksum_short = hash_fast(map, p->len);
	hash_slow(map, p->len, p->chksum_long, sess);
}

/*
 * Handle a symbolic link.
 * If we encounter directories existing in the symbolic link's place,
 * then try to unlink the directory.
 * Otherwise, simply overwrite with the symbolic link by renaming.
 * Return <0 on failure 0 on success.
 */
static int
pre_link(struct upload *p, struct sess *sess)
{
	struct stat		 st;
	const struct flist	*f;
	int			 rc, newlink = 0, updatelink = 0;
	char			*b, *temp = NULL;

	f = &p->fl[p->idx];
	assert(S_ISLNK(f->st.mode));

	if (!sess->opts->preserve_links) {
		WARNX("%s: ignoring symlink", f->path);
		return 0;
	} else if (sess->opts->dry_run) {
		log_link(sess, f);
		return 0;
	}

	/*
	 * See if the symlink already exists.
	 * If it's a directory, then try to unlink the directory prior
	 * to overwriting with a symbolic link.
	 * If it's a non-directory, we just overwrite it.
	 */

	assert(p->rootfd != -1);
	rc = fstatat(p->rootfd, f->path, &st, AT_SYMLINK_NOFOLLOW);
	if (rc != -1 && !S_ISLNK(st.st_mode)) {
		if (S_ISDIR(st.st_mode) &&
		    unlinkat(p->rootfd, f->path, AT_REMOVEDIR) == -1) {
			ERR("%s: unlinkat", f->path);
			return -1;
		}
		rc = -1;
	} else if (rc == -1 && errno != ENOENT) {
		ERR("%s: fstatat", f->path);
		return -1;
	}

	/*
	 * If the symbolic link already exists, then make sure that it
	 * points to the correct place.
	 */

	if (rc != -1) {
		b = symlinkat_read(p->rootfd, f->path);
		if (b == NULL) {
			ERRX1("symlinkat_read");
			return -1;
		}
		if (strcmp(f->link, b)) {
			free(b);
			b = NULL;
			LOG3("%s: updating symlink: %s", f->path, f->link);
			updatelink = 1;
		}
		free(b);
		b = NULL;
	}

	/*
	 * Create the temporary file as a symbolic link, then rename the
	 * temporary file as the real one, overwriting anything there.
	 */

	if (rc == -1 || updatelink) {
		LOG3("%s: creating symlink: %s", f->path, f->link);
		if (mktemplate(&temp, f->path, sess->opts->recursive) == -1) {
			ERRX1("mktemplate");
			return -1;
		}
		if (mkstemplinkat(f->link, p->rootfd, temp) == NULL) {
			ERR("mkstemplinkat");
			free(temp);
			return -1;
		}
		newlink = 1;
	}

	rsync_set_metadata_at(sess, newlink,
		p->rootfd, f, newlink ? temp : f->path);

	if (newlink) {
		if (renameat(p->rootfd, temp, p->rootfd, f->path) == -1) {
			ERR("%s: renameat %s", temp, f->path);
			(void)unlinkat(p->rootfd, temp, 0);
			free(temp);
			return -1;
		}
		free(temp);
	}

	log_link(sess, f);
	return 0;
}

/*
 * See pre_link(), but for devices.
 * FIXME: this is very similar to the other pre_xxx() functions.
 * Return <0 on failure 0 on success.
 */
static int
pre_dev(struct upload *p, struct sess *sess)
{
	struct stat		 st;
	const struct flist	*f;
	int			 rc, newdev = 0, updatedev = 0;
	char			*temp = NULL;

	f = &p->fl[p->idx];
	assert(S_ISBLK(f->st.mode) || S_ISCHR(f->st.mode));

	if (!sess->opts->devices || getuid() != 0) {
		WARNX("skipping non-regular file %s", f->path);
		return 0;
	} else if (sess->opts->dry_run) {
		log_file(sess, f);
		return 0;
	}

	/*
	 * See if the dev already exists.
	 * If a non-device exists in its place, we'll replace that.
	 * If it replaces a directory, remove the directory first.
	 */

	assert(p->rootfd != -1);
	rc = fstatat(p->rootfd, f->path, &st, AT_SYMLINK_NOFOLLOW);

	if (rc != -1 && !(S_ISBLK(st.st_mode) || S_ISCHR(st.st_mode))) {
		if (S_ISDIR(st.st_mode) &&
		    unlinkat(p->rootfd, f->path, AT_REMOVEDIR) == -1) {
			ERR("%s: unlinkat", f->path);
			return -1;
		}
		rc = -1;
	} else if (rc == -1 && errno != ENOENT) {
		ERR("%s: fstatat", f->path);
		return -1;
	}

	/* Make sure existing device is of the correct type. */

	if (rc != -1) {
		if ((f->st.mode & (S_IFCHR|S_IFBLK)) !=
		    (st.st_mode & (S_IFCHR|S_IFBLK)) ||
		    f->st.rdev != st.st_rdev) {
			LOG3("%s: updating device", f->path);
			updatedev = 1;
		}
	}

	if (rc == -1 || updatedev) {
		newdev = 1;
		if (mktemplate(&temp, f->path, sess->opts->recursive) == -1) {
			ERRX1("mktemplate");
			return -1;
		}
		if (mkstempnodat(p->rootfd, temp,
		    f->st.mode & (S_IFCHR|S_IFBLK), f->st.rdev) == NULL) {
			ERR("mkstempnodat");
			free(temp);
			return -1;
		}
	}

	rsync_set_metadata_at(sess, newdev,
	    p->rootfd, f, newdev ? temp : f->path);

	if (newdev) {
		if (renameat(p->rootfd, temp, p->rootfd, f->path) == -1) {
			ERR("%s: renameat %s", temp, f->path);
			(void)unlinkat(p->rootfd, temp, 0);
			free(temp);
			return -1;
		}
		free(temp);
	}

	log_file(sess, f);
	return 0;
}

/*
 * See pre_link(), but for FIFOs.
 * FIXME: this is very similar to the other pre_xxx() functions.
 * Return <0 on failure 0 on success.
 */
static int
pre_fifo(struct upload *p, struct sess *sess)
{
	struct stat		 st;
	const struct flist	*f;
	int			 rc, newfifo = 0;
	char			*temp = NULL;

	f = &p->fl[p->idx];
	assert(S_ISFIFO(f->st.mode));

	if (!sess->opts->specials) {
		WARNX("skipping non-regular file %s", f->path);
		return 0;
	} else if (sess->opts->dry_run) {
		log_file(sess, f);
		return 0;
	}

	/*
	 * See if the fifo already exists.
	 * If it exists as a non-FIFO, unlink it (if a directory) then
	 * mark it from replacement.
	 */

	assert(p->rootfd != -1);
	rc = fstatat(p->rootfd, f->path, &st, AT_SYMLINK_NOFOLLOW);

	if (rc != -1 && !S_ISFIFO(st.st_mode)) {
		if (S_ISDIR(st.st_mode) &&
		    unlinkat(p->rootfd, f->path, AT_REMOVEDIR) == -1) {
			ERR("%s: unlinkat", f->path);
			return -1;
		}
		rc = -1;
	} else if (rc == -1 && errno != ENOENT) {
		ERR("%s: fstatat", f->path);
		return -1;
	}

	if (rc == -1) {
		newfifo = 1;
		if (mktemplate(&temp, f->path, sess->opts->recursive) == -1) {
			ERRX1("mktemplate");
			return -1;
		}
		if (mkstempfifoat(p->rootfd, temp) == NULL) {
			ERR("mkstempfifoat");
			free(temp);
			return -1;
		}
	}

	rsync_set_metadata_at(sess, newfifo,
		p->rootfd, f, newfifo ? temp : f->path);

	if (newfifo) {
		if (renameat(p->rootfd, temp, p->rootfd, f->path) == -1) {
			ERR("%s: renameat %s", temp, f->path);
			(void)unlinkat(p->rootfd, temp, 0);
			free(temp);
			return -1;
		}
		free(temp);
	}

	log_file(sess, f);
	return 0;
}

/*
 * See pre_link(), but for socket files.
 * FIXME: this is very similar to the other pre_xxx() functions.
 * Return <0 on failure 0 on success.
 */
static int
pre_sock(struct upload *p, struct sess *sess)
{
	struct stat		 st;
	const struct flist	*f;
	int			 rc, newsock = 0;
	char			*temp = NULL;

	f = &p->fl[p->idx];
	assert(S_ISSOCK(f->st.mode));

	if (!sess->opts->specials) {
		WARNX("skipping non-regular file %s", f->path);
		return 0;
	} else if (sess->opts->dry_run) {
		log_file(sess, f);
		return 0;
	}

	/*
	 * See if the fifo already exists.
	 * If it exists as a non-FIFO, unlink it (if a directory) then
	 * mark it from replacement.
	 */

	assert(p->rootfd != -1);
	rc = fstatat(p->rootfd, f->path, &st, AT_SYMLINK_NOFOLLOW);

	if (rc != -1 && !S_ISSOCK(st.st_mode)) {
		if (S_ISDIR(st.st_mode) &&
		    unlinkat(p->rootfd, f->path, AT_REMOVEDIR) == -1) {
			ERR("%s: unlinkat", f->path);
			return -1;
		}
		rc = -1;
	} else if (rc == -1 && errno != ENOENT) {
		ERR("%s: fstatat", f->path);
		return -1;
	}

	if (rc == -1) {
		newsock = 1;
		if (mktemplate(&temp, f->path, sess->opts->recursive) == -1) {
			ERRX1("mktemplate");
			return -1;
		}
		if (mkstempsock(p->root, temp) == NULL) {
			ERR("mkstempsock");
			free(temp);
			return -1;
		}
	}

	rsync_set_metadata_at(sess, newsock,
		p->rootfd, f, newsock ? temp : f->path);

	if (newsock) {
		if (renameat(p->rootfd, temp, p->rootfd, f->path) == -1) {
			ERR("%s: renameat %s", temp, f->path);
			(void)unlinkat(p->rootfd, temp, 0);
			free(temp);
			return -1;
		}
		free(temp);
	}

	log_file(sess, f);
	return 0;
}

/*
 * If not found, create the destination directory in prefix order.
 * Create directories using the existing umask.
 * Return <0 on failure 0 on success.
 */
static int
pre_dir(const struct upload *p, struct sess *sess)
{
	struct stat	 st;
	int		 rc;
	const struct flist *f;

	f = &p->fl[p->idx];
	assert(S_ISDIR(f->st.mode));

	if (!sess->opts->recursive) {
		WARNX("%s: ignoring directory", f->path);
		return 0;
	} else if (sess->opts->dry_run) {
		log_dir(sess, f);
		return 0;
	}

	assert(p->rootfd != -1);
	rc = fstatat(p->rootfd, f->path, &st, AT_SYMLINK_NOFOLLOW);

	if (rc == -1 && errno != ENOENT) {
		ERR("%s: fstatat", f->path);
		return -1;
	} else if (rc != -1 && !S_ISDIR(st.st_mode)) {
		ERRX("%s: not a directory", f->path);
		return -1;
	} else if (rc != -1) {
		/*
		 * FIXME: we should fchmod the permissions here as well,
		 * as we may locally have shut down writing into the
		 * directory and that doesn't work.
		 */
		LOG3("%s: updating directory", f->path);
		return 0;
	}

	/*
	 * We want to make the directory with default permissions (using
	 * our old umask, which we've since unset), then adjust
	 * permissions (assuming preserve_perms or new) afterward in
	 * case it's u-w or something.
	 */

	LOG3("%s: creating directory", f->path);
	if (mkdirat(p->rootfd, f->path, 0777 & ~p->oumask) == -1) {
		ERR("%s: mkdirat", f->path);
		return -1;
	}

	p->newdir[p->idx] = 1;
	log_dir(sess, f);
	return 0;
}

/*
 * Process the directory time and mode for "idx" in the file list.
 * Returns zero on failure, non-zero on success.
 */
static int
post_dir(struct sess *sess, const struct upload *u, size_t idx)
{
	struct timespec	 tv[2];
	int		 rc;
	struct stat	 st;
	const struct flist *f;

	f = &u->fl[idx];
	assert(S_ISDIR(f->st.mode));

	/* We already warned about the directory in pre_process_dir(). */

	if (!sess->opts->recursive)
		return 1;
	else if (sess->opts->dry_run)
		return 1;

	if (fstatat(u->rootfd, f->path, &st, AT_SYMLINK_NOFOLLOW) == -1) {
		ERR("%s: fstatat", f->path);
		return 0;
	} else if (!S_ISDIR(st.st_mode)) {
		WARNX("%s: not a directory", f->path);
		return 0;
	}

	/*
	 * Update the modification time if we're a new directory *or* if
	 * we're preserving times and the time has changed.
	 * FIXME: run rsync_set_metadata()?
	 */

	if (u->newdir[idx] ||
	    (sess->opts->preserve_times &&
	     st.st_mtime != f->st.mtime)) {
		tv[0].tv_sec = time(NULL);
		tv[0].tv_nsec = 0;
		tv[1].tv_sec = f->st.mtime;
		tv[1].tv_nsec = 0;
		rc = utimensat(u->rootfd, f->path, tv, 0);
		if (rc == -1) {
			ERR("%s: utimensat", f->path);
			return 0;
		}
		LOG4("%s: updated date", f->path);
	}

	/*
	 * Update the mode if we're a new directory *or* if we're
	 * preserving modes and it has changed.
	 */

	if (u->newdir[idx] ||
	    (sess->opts->preserve_perms && st.st_mode != f->st.mode)) {
		rc = fchmodat(u->rootfd, f->path, f->st.mode, 0);
		if (rc == -1) {
			ERR("%s: fchmodat", f->path);
			return 0;
		}
		LOG4("%s: updated mode", f->path);
	}

	return 1;
}

/*
 * Try to open the file at the current index.
 * If the file does not exist, returns with success.
 * Return <0 on failure, 0 on success w/nothing to be done, >0 on
 * success and the file needs attention.
 */
static int
pre_file(const struct upload *p, int *filefd, struct sess *sess)
{
	const struct flist *f;

	f = &p->fl[p->idx];
	assert(S_ISREG(f->st.mode));

	if (sess->opts->dry_run) {
		log_file(sess, f);
		if (!io_write_int(sess, p->fdout, p->idx)) {
			ERRX1("io_write_int");
			return -1;
		}
		return 0;
	}

	/*
	 * For non dry-run cases, we'll write the acknowledgement later
	 * in the rsync_uploader() function because we need to wait for
	 * the open() call to complete.
	 * If the call to openat() fails with ENOENT, there's a
	 * fast-path between here and the write function, so we won't do
	 * any blocking between now and then.
	 */

	*filefd = openat(p->rootfd, f->path,
		O_RDONLY | O_NOFOLLOW | O_NONBLOCK, 0);
	if (*filefd != -1 || errno == ENOENT)
		return 1;
	ERR("%s: openat", f->path);
	return -1;
}

/*
 * Allocate an uploader object in the correct state to start.
 * Returns NULL on failure or the pointer otherwise.
 * On success, upload_free() must be called with the allocated pointer.
 */
struct upload *
upload_alloc(const char *root, int rootfd, int fdout,
	size_t clen, const struct flist *fl, size_t flsz, mode_t msk)
{
	struct upload	*p;

	if ((p = calloc(1, sizeof(struct upload))) == NULL) {
		ERR("calloc");
		return NULL;
	}

	p->state = UPLOAD_FIND_NEXT;
	p->oumask = msk;
	p->root = strdup(root);
	if (p->root == NULL) {
		ERR("strdup");
		free(p);
		return NULL;
	}
	p->rootfd = rootfd;
	p->csumlen = clen;
	p->fdout = fdout;
	p->fl = fl;
	p->flsz = flsz;
	p->newdir = calloc(flsz, sizeof(int));
	if (p->newdir == NULL) {
		ERR("calloc");
		free(p->root);
		free(p);
		return NULL;
	}
	return p;
}

/*
 * Perform all cleanups and free.
 * Passing a NULL to this function is ok.
 */
void
upload_free(struct upload *p)
{

	if (p == NULL)
		return;
	free(p->root);
	free(p->newdir);
	free(p->buf);
	free(p);
}

/*
 * Iterates through all available files and conditionally gets the file
 * ready for processing to check whether it's up to date.
 * If not up to date or empty, sends file information to the sender.
 * If returns 0, we've processed all files there are to process.
 * If returns >0, we're waiting for POLLIN or POLLOUT data.
 * Otherwise returns <0, which is an error.
 */
int
rsync_uploader(struct upload *u, int *fileinfd,
	struct sess *sess, int *fileoutfd)
{
	struct blkset	    blk;
	struct stat	    st;
	void		   *mbuf, *bufp;
	ssize_t		    msz;
	size_t		    i, pos, sz;
	off_t		    offs;
	int		    c;
	const struct flist *f;

	/* This should never get called. */

	assert(u->state != UPLOAD_FINISHED);

	/*
	 * If we have an upload in progress, then keep writing until the
	 * buffer has been fully written.
	 * We must only have the output file descriptor working and also
	 * have a valid buffer to write.
	 */

	if (u->state == UPLOAD_WRITE_LOCAL) {
		assert(u->buf != NULL);
		assert(*fileoutfd != -1);
		assert(*fileinfd == -1);

		/*
		 * Unfortunately, we need to chunk these: if we're
		 * the server side of things, then we're multiplexing
		 * output and need to wrap this in chunks.
		 * This is a major deficiency of rsync.
		 * FIXME: add a "fast-path" mode that simply dumps out
		 * the buffer non-blocking if we're not mplexing.
		 */

		if (u->bufpos < u->bufsz) {
			sz = MAX_CHUNK < (u->bufsz - u->bufpos) ?
				MAX_CHUNK : (u->bufsz - u->bufpos);
			c = io_write_buf(sess, u->fdout,
				u->buf + u->bufpos, sz);
			if (c == 0) {
				ERRX1("io_write_nonblocking");
				return -1;
			}
			u->bufpos += sz;
			if (u->bufpos < u->bufsz)
				return 1;
		}

		/*
		 * Let the UPLOAD_FIND_NEXT state handle things if we
		 * finish, as we'll need to write a POLLOUT message and
		 * not have a writable descriptor yet.
		 */

		u->state = UPLOAD_FIND_NEXT;
		u->idx++;
		return 1;
	}

	/*
	 * If we invoke the uploader without a file currently open, then
	 * we iterate through til the next available regular file and
	 * start the opening process.
	 * This means we must have the output file descriptor working.
	 */

	if (u->state == UPLOAD_FIND_NEXT) {
		assert(*fileinfd == -1);
		assert(*fileoutfd != -1);

		for ( ; u->idx < u->flsz; u->idx++) {
			if (S_ISDIR(u->fl[u->idx].st.mode))
				c = pre_dir(u, sess);
			else if (S_ISLNK(u->fl[u->idx].st.mode))
				c = pre_link(u, sess);
			else if (S_ISREG(u->fl[u->idx].st.mode))
				c = pre_file(u, fileinfd, sess);
			else if (S_ISBLK(u->fl[u->idx].st.mode) ||
			    S_ISCHR(u->fl[u->idx].st.mode))
				c = pre_dev(u, sess);
			else if (S_ISFIFO(u->fl[u->idx].st.mode))
				c = pre_fifo(u, sess);
			else if (S_ISSOCK(u->fl[u->idx].st.mode))
				c = pre_sock(u, sess);
			else
				c = 0;

			if (c < 0)
				return -1;
			else if (c > 0)
				break;
		}

		/*
		 * Whether we've finished writing files or not, we
		 * disable polling on the output channel.
		 */

		*fileoutfd = -1;
		if (u->idx == u->flsz) {
			assert(*fileinfd == -1);
			if (!io_write_int(sess, u->fdout, -1)) {
				ERRX1("io_write_int");
				return -1;
			}
			u->state = UPLOAD_FINISHED;
			LOG4("uploader: finished");
			return 0;
		}

		/* Go back to the event loop, if necessary. */

		u->state = (*fileinfd == -1) ?
			UPLOAD_WRITE_LOCAL : UPLOAD_READ_LOCAL;
		if (u->state == UPLOAD_READ_LOCAL)
			return 1;
	}

	/*
	 * If an input file is open, stat it and see if it's already up
	 * to date, in which case close it and go to the next one.
	 * Either way, we don't have a write channel open.
	 */

	if (u->state == UPLOAD_READ_LOCAL) {
		assert(*fileinfd != -1);
		assert(*fileoutfd == -1);
		f = &u->fl[u->idx];

		if (fstat(*fileinfd, &st) == -1) {
			ERR("%s: fstat", f->path);
			close(*fileinfd);
			*fileinfd = -1;
			return -1;
		} else if (!S_ISREG(st.st_mode)) {
			ERRX("%s: not regular", f->path);
			close(*fileinfd);
			*fileinfd = -1;
			return -1;
		}

		if (st.st_size == f->st.size &&
		    st.st_mtime == f->st.mtime) {
			LOG3("%s: skipping: up to date", f->path);
			if (!rsync_set_metadata
			    (sess, 0, *fileinfd, f, f->path)) {
				ERRX1("rsync_set_metadata");
				close(*fileinfd);
				*fileinfd = -1;
				return -1;
			}
			close(*fileinfd);
			*fileinfd = -1;
			*fileoutfd = u->fdout;
			u->state = UPLOAD_FIND_NEXT;
			u->idx++;
			return 1;
		}

		/* Fallthrough... */

		u->state = UPLOAD_WRITE_LOCAL;
	}

	/* Initialies our blocks. */

	assert(u->state == UPLOAD_WRITE_LOCAL);
	memset(&blk, 0, sizeof(struct blkset));
	blk.csum = u->csumlen;

	if (*fileinfd != -1 && st.st_size > 0) {
		init_blkset(&blk, st.st_size);
		assert(blk.blksz);

		blk.blks = calloc(blk.blksz, sizeof(struct blk));
		if (blk.blks == NULL) {
			ERR("calloc");
			close(*fileinfd);
			*fileinfd = -1;
			return -1;
		}

		if ((mbuf = calloc(1, blk.len)) == NULL) {
			ERR("calloc");
			close(*fileinfd);
			*fileinfd = -1;
			return -1;
		}

		offs = 0;
		i = 0;
		do {
			msz = pread(*fileinfd, mbuf, blk.len, offs);
			if (msz < 0) {
				ERR("pread");
				close(*fileinfd);
				*fileinfd = -1;
				return -1;
			}
			if ((size_t)msz != blk.len && (size_t)msz != blk.rem) {
				/* short read, try again */
				continue;
			}
			init_blk(&blk.blks[i], &blk, offs, i, mbuf, sess);
			offs += blk.len;
			LOG3(
			    "i=%ld, offs=%lld, msz=%ld, blk.len=%lu, blk.rem=%lu",
			    i, offs, msz, blk.len, blk.rem);
			i++;
		} while (i < blk.blksz);

		close(*fileinfd);
		*fileinfd = -1;
		LOG3("%s: mapped %jd B with %zu blocks",
		    u->fl[u->idx].path, (intmax_t)blk.size,
		    blk.blksz);
	} else {
		if (*fileinfd != -1) {
			close(*fileinfd);
			*fileinfd = -1;
		}
		blk.len = MAX_CHUNK; /* Doesn't matter. */
		LOG3("%s: not mapped", u->fl[u->idx].path);
	}

	assert(*fileinfd == -1);

	/* Make sure the block metadata buffer is big enough. */

	u->bufsz =
	     sizeof(int32_t) + /* identifier */
	     sizeof(int32_t) + /* block count */
	     sizeof(int32_t) + /* block length */
	     sizeof(int32_t) + /* checksum length */
	     sizeof(int32_t) + /* block remainder */
	     blk.blksz *
	     (sizeof(int32_t) + /* short checksum */
	      blk.csum); /* long checksum */

	if (u->bufsz > u->bufmax) {
		if ((bufp = realloc(u->buf, u->bufsz)) == NULL) {
			ERR("realloc");
			return -1;
		}
		u->buf = bufp;
		u->bufmax = u->bufsz;
	}

	u->bufpos = pos = 0;
	io_buffer_int(u->buf, &pos, u->bufsz, u->idx);
	io_buffer_int(u->buf, &pos, u->bufsz, blk.blksz);
	io_buffer_int(u->buf, &pos, u->bufsz, blk.len);
	io_buffer_int(u->buf, &pos, u->bufsz, blk.csum);
	io_buffer_int(u->buf, &pos, u->bufsz, blk.rem);
	for (i = 0; i < blk.blksz; i++) {
		io_buffer_int(u->buf, &pos, u->bufsz,
			blk.blks[i].chksum_short);
		io_buffer_buf(u->buf, &pos, u->bufsz,
			blk.blks[i].chksum_long, blk.csum);
	}
	assert(pos == u->bufsz);

	/* Reenable the output poller and clean up. */

	*fileoutfd = u->fdout;
	free(blk.blks);
	return 1;
}

/*
 * Fix up the directory permissions and times post-order.
 * We can't fix up directory permissions in place because the server may
 * want us to have overly-tight permissions---say, those that don't
 * allow writing into the directory.
 * We also need to do our directory times post-order because making
 * files within the directory will change modification times.
 * Returns zero on failure, non-zero on success.
 */
int
rsync_uploader_tail(struct upload *u, struct sess *sess)
{
	size_t	 i;


	if (!sess->opts->preserve_times &&
	    !sess->opts->preserve_perms)
		return 1;

	LOG2("fixing up directory times and permissions");

	for (i = 0; i < u->flsz; i++)
		if (S_ISDIR(u->fl[i].st.mode))
			if (!post_dir(sess, u, i))
				return 0;

	return 1;
}