Merge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux
[linux-2.6.git] / fs / btrfs / send.c
blobd3f3b43cae0bdef23c889f6c5939c49aea13eda6
1 /*
2 * Copyright (C) 2012 Alexander Block. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/bsearch.h>
20 #include <linux/fs.h>
21 #include <linux/file.h>
22 #include <linux/sort.h>
23 #include <linux/mount.h>
24 #include <linux/xattr.h>
25 #include <linux/posix_acl_xattr.h>
26 #include <linux/radix-tree.h>
27 #include <linux/crc32c.h>
28 #include <linux/vmalloc.h>
30 #include "send.h"
31 #include "backref.h"
32 #include "locking.h"
33 #include "disk-io.h"
34 #include "btrfs_inode.h"
35 #include "transaction.h"
37 static int g_verbose = 0;
39 #define verbose_printk(...) if (g_verbose) printk(__VA_ARGS__)
42 * A fs_path is a helper to dynamically build path names with unknown size.
43 * It reallocates the internal buffer on demand.
44 * It allows fast adding of path elements on the right side (normal path) and
45 * fast adding to the left side (reversed path). A reversed path can also be
46 * unreversed if needed.
48 struct fs_path {
49 union {
50 struct {
51 char *start;
52 char *end;
53 char *prepared;
55 char *buf;
56 int buf_len;
57 int reversed:1;
58 int virtual_mem:1;
59 char inline_buf[];
61 char pad[PAGE_SIZE];
64 #define FS_PATH_INLINE_SIZE \
65 (sizeof(struct fs_path) - offsetof(struct fs_path, inline_buf))
68 /* reused for each extent */
69 struct clone_root {
70 struct btrfs_root *root;
71 u64 ino;
72 u64 offset;
74 u64 found_refs;
77 #define SEND_CTX_MAX_NAME_CACHE_SIZE 128
78 #define SEND_CTX_NAME_CACHE_CLEAN_SIZE (SEND_CTX_MAX_NAME_CACHE_SIZE * 2)
80 struct send_ctx {
81 struct file *send_filp;
82 loff_t send_off;
83 char *send_buf;
84 u32 send_size;
85 u32 send_max_size;
86 u64 total_send_size;
87 u64 cmd_send_size[BTRFS_SEND_C_MAX + 1];
88 u64 flags; /* 'flags' member of btrfs_ioctl_send_args is u64 */
90 struct vfsmount *mnt;
92 struct btrfs_root *send_root;
93 struct btrfs_root *parent_root;
94 struct clone_root *clone_roots;
95 int clone_roots_cnt;
97 /* current state of the compare_tree call */
98 struct btrfs_path *left_path;
99 struct btrfs_path *right_path;
100 struct btrfs_key *cmp_key;
103 * infos of the currently processed inode. In case of deleted inodes,
104 * these are the values from the deleted inode.
106 u64 cur_ino;
107 u64 cur_inode_gen;
108 int cur_inode_new;
109 int cur_inode_new_gen;
110 int cur_inode_deleted;
111 u64 cur_inode_size;
112 u64 cur_inode_mode;
114 u64 send_progress;
116 struct list_head new_refs;
117 struct list_head deleted_refs;
119 struct radix_tree_root name_cache;
120 struct list_head name_cache_list;
121 int name_cache_size;
123 struct file *cur_inode_filp;
124 char *read_buf;
127 struct name_cache_entry {
128 struct list_head list;
130 * radix_tree has only 32bit entries but we need to handle 64bit inums.
131 * We use the lower 32bit of the 64bit inum to store it in the tree. If
132 * more then one inum would fall into the same entry, we use radix_list
133 * to store the additional entries. radix_list is also used to store
134 * entries where two entries have the same inum but different
135 * generations.
137 struct list_head radix_list;
138 u64 ino;
139 u64 gen;
140 u64 parent_ino;
141 u64 parent_gen;
142 int ret;
143 int need_later_update;
144 int name_len;
145 char name[];
148 static void fs_path_reset(struct fs_path *p)
150 if (p->reversed) {
151 p->start = p->buf + p->buf_len - 1;
152 p->end = p->start;
153 *p->start = 0;
154 } else {
155 p->start = p->buf;
156 p->end = p->start;
157 *p->start = 0;
161 static struct fs_path *fs_path_alloc(void)
163 struct fs_path *p;
165 p = kmalloc(sizeof(*p), GFP_NOFS);
166 if (!p)
167 return NULL;
168 p->reversed = 0;
169 p->virtual_mem = 0;
170 p->buf = p->inline_buf;
171 p->buf_len = FS_PATH_INLINE_SIZE;
172 fs_path_reset(p);
173 return p;
176 static struct fs_path *fs_path_alloc_reversed(void)
178 struct fs_path *p;
180 p = fs_path_alloc();
181 if (!p)
182 return NULL;
183 p->reversed = 1;
184 fs_path_reset(p);
185 return p;
188 static void fs_path_free(struct fs_path *p)
190 if (!p)
191 return;
192 if (p->buf != p->inline_buf) {
193 if (p->virtual_mem)
194 vfree(p->buf);
195 else
196 kfree(p->buf);
198 kfree(p);
201 static int fs_path_len(struct fs_path *p)
203 return p->end - p->start;
206 static int fs_path_ensure_buf(struct fs_path *p, int len)
208 char *tmp_buf;
209 int path_len;
210 int old_buf_len;
212 len++;
214 if (p->buf_len >= len)
215 return 0;
217 path_len = p->end - p->start;
218 old_buf_len = p->buf_len;
219 len = PAGE_ALIGN(len);
221 if (p->buf == p->inline_buf) {
222 tmp_buf = kmalloc(len, GFP_NOFS);
223 if (!tmp_buf) {
224 tmp_buf = vmalloc(len);
225 if (!tmp_buf)
226 return -ENOMEM;
227 p->virtual_mem = 1;
229 memcpy(tmp_buf, p->buf, p->buf_len);
230 p->buf = tmp_buf;
231 p->buf_len = len;
232 } else {
233 if (p->virtual_mem) {
234 tmp_buf = vmalloc(len);
235 if (!tmp_buf)
236 return -ENOMEM;
237 memcpy(tmp_buf, p->buf, p->buf_len);
238 vfree(p->buf);
239 } else {
240 tmp_buf = krealloc(p->buf, len, GFP_NOFS);
241 if (!tmp_buf) {
242 tmp_buf = vmalloc(len);
243 if (!tmp_buf)
244 return -ENOMEM;
245 memcpy(tmp_buf, p->buf, p->buf_len);
246 kfree(p->buf);
247 p->virtual_mem = 1;
250 p->buf = tmp_buf;
251 p->buf_len = len;
253 if (p->reversed) {
254 tmp_buf = p->buf + old_buf_len - path_len - 1;
255 p->end = p->buf + p->buf_len - 1;
256 p->start = p->end - path_len;
257 memmove(p->start, tmp_buf, path_len + 1);
258 } else {
259 p->start = p->buf;
260 p->end = p->start + path_len;
262 return 0;
265 static int fs_path_prepare_for_add(struct fs_path *p, int name_len)
267 int ret;
268 int new_len;
270 new_len = p->end - p->start + name_len;
271 if (p->start != p->end)
272 new_len++;
273 ret = fs_path_ensure_buf(p, new_len);
274 if (ret < 0)
275 goto out;
277 if (p->reversed) {
278 if (p->start != p->end)
279 *--p->start = '/';
280 p->start -= name_len;
281 p->prepared = p->start;
282 } else {
283 if (p->start != p->end)
284 *p->end++ = '/';
285 p->prepared = p->end;
286 p->end += name_len;
287 *p->end = 0;
290 out:
291 return ret;
294 static int fs_path_add(struct fs_path *p, const char *name, int name_len)
296 int ret;
298 ret = fs_path_prepare_for_add(p, name_len);
299 if (ret < 0)
300 goto out;
301 memcpy(p->prepared, name, name_len);
302 p->prepared = NULL;
304 out:
305 return ret;
308 static int fs_path_add_path(struct fs_path *p, struct fs_path *p2)
310 int ret;
312 ret = fs_path_prepare_for_add(p, p2->end - p2->start);
313 if (ret < 0)
314 goto out;
315 memcpy(p->prepared, p2->start, p2->end - p2->start);
316 p->prepared = NULL;
318 out:
319 return ret;
322 static int fs_path_add_from_extent_buffer(struct fs_path *p,
323 struct extent_buffer *eb,
324 unsigned long off, int len)
326 int ret;
328 ret = fs_path_prepare_for_add(p, len);
329 if (ret < 0)
330 goto out;
332 read_extent_buffer(eb, p->prepared, off, len);
333 p->prepared = NULL;
335 out:
336 return ret;
339 #if 0
340 static void fs_path_remove(struct fs_path *p)
342 BUG_ON(p->reversed);
343 while (p->start != p->end && *p->end != '/')
344 p->end--;
345 *p->end = 0;
347 #endif
349 static int fs_path_copy(struct fs_path *p, struct fs_path *from)
351 int ret;
353 p->reversed = from->reversed;
354 fs_path_reset(p);
356 ret = fs_path_add_path(p, from);
358 return ret;
362 static void fs_path_unreverse(struct fs_path *p)
364 char *tmp;
365 int len;
367 if (!p->reversed)
368 return;
370 tmp = p->start;
371 len = p->end - p->start;
372 p->start = p->buf;
373 p->end = p->start + len;
374 memmove(p->start, tmp, len + 1);
375 p->reversed = 0;
378 static struct btrfs_path *alloc_path_for_send(void)
380 struct btrfs_path *path;
382 path = btrfs_alloc_path();
383 if (!path)
384 return NULL;
385 path->search_commit_root = 1;
386 path->skip_locking = 1;
387 return path;
390 static int write_buf(struct file *filp, const void *buf, u32 len, loff_t *off)
392 int ret;
393 mm_segment_t old_fs;
394 u32 pos = 0;
396 old_fs = get_fs();
397 set_fs(KERNEL_DS);
399 while (pos < len) {
400 ret = vfs_write(filp, (char *)buf + pos, len - pos, off);
401 /* TODO handle that correctly */
402 /*if (ret == -ERESTARTSYS) {
403 continue;
405 if (ret < 0)
406 goto out;
407 if (ret == 0) {
408 ret = -EIO;
409 goto out;
411 pos += ret;
414 ret = 0;
416 out:
417 set_fs(old_fs);
418 return ret;
421 static int tlv_put(struct send_ctx *sctx, u16 attr, const void *data, int len)
423 struct btrfs_tlv_header *hdr;
424 int total_len = sizeof(*hdr) + len;
425 int left = sctx->send_max_size - sctx->send_size;
427 if (unlikely(left < total_len))
428 return -EOVERFLOW;
430 hdr = (struct btrfs_tlv_header *) (sctx->send_buf + sctx->send_size);
431 hdr->tlv_type = cpu_to_le16(attr);
432 hdr->tlv_len = cpu_to_le16(len);
433 memcpy(hdr + 1, data, len);
434 sctx->send_size += total_len;
436 return 0;
439 #if 0
440 static int tlv_put_u8(struct send_ctx *sctx, u16 attr, u8 value)
442 return tlv_put(sctx, attr, &value, sizeof(value));
445 static int tlv_put_u16(struct send_ctx *sctx, u16 attr, u16 value)
447 __le16 tmp = cpu_to_le16(value);
448 return tlv_put(sctx, attr, &tmp, sizeof(tmp));
451 static int tlv_put_u32(struct send_ctx *sctx, u16 attr, u32 value)
453 __le32 tmp = cpu_to_le32(value);
454 return tlv_put(sctx, attr, &tmp, sizeof(tmp));
456 #endif
458 static int tlv_put_u64(struct send_ctx *sctx, u16 attr, u64 value)
460 __le64 tmp = cpu_to_le64(value);
461 return tlv_put(sctx, attr, &tmp, sizeof(tmp));
464 static int tlv_put_string(struct send_ctx *sctx, u16 attr,
465 const char *str, int len)
467 if (len == -1)
468 len = strlen(str);
469 return tlv_put(sctx, attr, str, len);
472 static int tlv_put_uuid(struct send_ctx *sctx, u16 attr,
473 const u8 *uuid)
475 return tlv_put(sctx, attr, uuid, BTRFS_UUID_SIZE);
478 #if 0
479 static int tlv_put_timespec(struct send_ctx *sctx, u16 attr,
480 struct timespec *ts)
482 struct btrfs_timespec bts;
483 bts.sec = cpu_to_le64(ts->tv_sec);
484 bts.nsec = cpu_to_le32(ts->tv_nsec);
485 return tlv_put(sctx, attr, &bts, sizeof(bts));
487 #endif
489 static int tlv_put_btrfs_timespec(struct send_ctx *sctx, u16 attr,
490 struct extent_buffer *eb,
491 struct btrfs_timespec *ts)
493 struct btrfs_timespec bts;
494 read_extent_buffer(eb, &bts, (unsigned long)ts, sizeof(bts));
495 return tlv_put(sctx, attr, &bts, sizeof(bts));
499 #define TLV_PUT(sctx, attrtype, attrlen, data) \
500 do { \
501 ret = tlv_put(sctx, attrtype, attrlen, data); \
502 if (ret < 0) \
503 goto tlv_put_failure; \
504 } while (0)
506 #define TLV_PUT_INT(sctx, attrtype, bits, value) \
507 do { \
508 ret = tlv_put_u##bits(sctx, attrtype, value); \
509 if (ret < 0) \
510 goto tlv_put_failure; \
511 } while (0)
513 #define TLV_PUT_U8(sctx, attrtype, data) TLV_PUT_INT(sctx, attrtype, 8, data)
514 #define TLV_PUT_U16(sctx, attrtype, data) TLV_PUT_INT(sctx, attrtype, 16, data)
515 #define TLV_PUT_U32(sctx, attrtype, data) TLV_PUT_INT(sctx, attrtype, 32, data)
516 #define TLV_PUT_U64(sctx, attrtype, data) TLV_PUT_INT(sctx, attrtype, 64, data)
517 #define TLV_PUT_STRING(sctx, attrtype, str, len) \
518 do { \
519 ret = tlv_put_string(sctx, attrtype, str, len); \
520 if (ret < 0) \
521 goto tlv_put_failure; \
522 } while (0)
523 #define TLV_PUT_PATH(sctx, attrtype, p) \
524 do { \
525 ret = tlv_put_string(sctx, attrtype, p->start, \
526 p->end - p->start); \
527 if (ret < 0) \
528 goto tlv_put_failure; \
529 } while(0)
530 #define TLV_PUT_UUID(sctx, attrtype, uuid) \
531 do { \
532 ret = tlv_put_uuid(sctx, attrtype, uuid); \
533 if (ret < 0) \
534 goto tlv_put_failure; \
535 } while (0)
536 #define TLV_PUT_TIMESPEC(sctx, attrtype, ts) \
537 do { \
538 ret = tlv_put_timespec(sctx, attrtype, ts); \
539 if (ret < 0) \
540 goto tlv_put_failure; \
541 } while (0)
542 #define TLV_PUT_BTRFS_TIMESPEC(sctx, attrtype, eb, ts) \
543 do { \
544 ret = tlv_put_btrfs_timespec(sctx, attrtype, eb, ts); \
545 if (ret < 0) \
546 goto tlv_put_failure; \
547 } while (0)
549 static int send_header(struct send_ctx *sctx)
551 struct btrfs_stream_header hdr;
553 strcpy(hdr.magic, BTRFS_SEND_STREAM_MAGIC);
554 hdr.version = cpu_to_le32(BTRFS_SEND_STREAM_VERSION);
556 return write_buf(sctx->send_filp, &hdr, sizeof(hdr),
557 &sctx->send_off);
561 * For each command/item we want to send to userspace, we call this function.
563 static int begin_cmd(struct send_ctx *sctx, int cmd)
565 struct btrfs_cmd_header *hdr;
567 if (!sctx->send_buf) {
568 WARN_ON(1);
569 return -EINVAL;
572 BUG_ON(sctx->send_size);
574 sctx->send_size += sizeof(*hdr);
575 hdr = (struct btrfs_cmd_header *)sctx->send_buf;
576 hdr->cmd = cpu_to_le16(cmd);
578 return 0;
581 static int send_cmd(struct send_ctx *sctx)
583 int ret;
584 struct btrfs_cmd_header *hdr;
585 u32 crc;
587 hdr = (struct btrfs_cmd_header *)sctx->send_buf;
588 hdr->len = cpu_to_le32(sctx->send_size - sizeof(*hdr));
589 hdr->crc = 0;
591 crc = crc32c(0, (unsigned char *)sctx->send_buf, sctx->send_size);
592 hdr->crc = cpu_to_le32(crc);
594 ret = write_buf(sctx->send_filp, sctx->send_buf, sctx->send_size,
595 &sctx->send_off);
597 sctx->total_send_size += sctx->send_size;
598 sctx->cmd_send_size[le16_to_cpu(hdr->cmd)] += sctx->send_size;
599 sctx->send_size = 0;
601 return ret;
605 * Sends a move instruction to user space
607 static int send_rename(struct send_ctx *sctx,
608 struct fs_path *from, struct fs_path *to)
610 int ret;
612 verbose_printk("btrfs: send_rename %s -> %s\n", from->start, to->start);
614 ret = begin_cmd(sctx, BTRFS_SEND_C_RENAME);
615 if (ret < 0)
616 goto out;
618 TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, from);
619 TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH_TO, to);
621 ret = send_cmd(sctx);
623 tlv_put_failure:
624 out:
625 return ret;
629 * Sends a link instruction to user space
631 static int send_link(struct send_ctx *sctx,
632 struct fs_path *path, struct fs_path *lnk)
634 int ret;
636 verbose_printk("btrfs: send_link %s -> %s\n", path->start, lnk->start);
638 ret = begin_cmd(sctx, BTRFS_SEND_C_LINK);
639 if (ret < 0)
640 goto out;
642 TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, path);
643 TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH_LINK, lnk);
645 ret = send_cmd(sctx);
647 tlv_put_failure:
648 out:
649 return ret;
653 * Sends an unlink instruction to user space
655 static int send_unlink(struct send_ctx *sctx, struct fs_path *path)
657 int ret;
659 verbose_printk("btrfs: send_unlink %s\n", path->start);
661 ret = begin_cmd(sctx, BTRFS_SEND_C_UNLINK);
662 if (ret < 0)
663 goto out;
665 TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, path);
667 ret = send_cmd(sctx);
669 tlv_put_failure:
670 out:
671 return ret;
675 * Sends a rmdir instruction to user space
677 static int send_rmdir(struct send_ctx *sctx, struct fs_path *path)
679 int ret;
681 verbose_printk("btrfs: send_rmdir %s\n", path->start);
683 ret = begin_cmd(sctx, BTRFS_SEND_C_RMDIR);
684 if (ret < 0)
685 goto out;
687 TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, path);
689 ret = send_cmd(sctx);
691 tlv_put_failure:
692 out:
693 return ret;
697 * Helper function to retrieve some fields from an inode item.
699 static int get_inode_info(struct btrfs_root *root,
700 u64 ino, u64 *size, u64 *gen,
701 u64 *mode, u64 *uid, u64 *gid,
702 u64 *rdev)
704 int ret;
705 struct btrfs_inode_item *ii;
706 struct btrfs_key key;
707 struct btrfs_path *path;
709 path = alloc_path_for_send();
710 if (!path)
711 return -ENOMEM;
713 key.objectid = ino;
714 key.type = BTRFS_INODE_ITEM_KEY;
715 key.offset = 0;
716 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
717 if (ret < 0)
718 goto out;
719 if (ret) {
720 ret = -ENOENT;
721 goto out;
724 ii = btrfs_item_ptr(path->nodes[0], path->slots[0],
725 struct btrfs_inode_item);
726 if (size)
727 *size = btrfs_inode_size(path->nodes[0], ii);
728 if (gen)
729 *gen = btrfs_inode_generation(path->nodes[0], ii);
730 if (mode)
731 *mode = btrfs_inode_mode(path->nodes[0], ii);
732 if (uid)
733 *uid = btrfs_inode_uid(path->nodes[0], ii);
734 if (gid)
735 *gid = btrfs_inode_gid(path->nodes[0], ii);
736 if (rdev)
737 *rdev = btrfs_inode_rdev(path->nodes[0], ii);
739 out:
740 btrfs_free_path(path);
741 return ret;
744 typedef int (*iterate_inode_ref_t)(int num, u64 dir, int index,
745 struct fs_path *p,
746 void *ctx);
749 * Helper function to iterate the entries in ONE btrfs_inode_ref or
750 * btrfs_inode_extref.
751 * The iterate callback may return a non zero value to stop iteration. This can
752 * be a negative value for error codes or 1 to simply stop it.
754 * path must point to the INODE_REF or INODE_EXTREF when called.
756 static int iterate_inode_ref(struct btrfs_root *root, struct btrfs_path *path,
757 struct btrfs_key *found_key, int resolve,
758 iterate_inode_ref_t iterate, void *ctx)
760 struct extent_buffer *eb = path->nodes[0];
761 struct btrfs_item *item;
762 struct btrfs_inode_ref *iref;
763 struct btrfs_inode_extref *extref;
764 struct btrfs_path *tmp_path;
765 struct fs_path *p;
766 u32 cur = 0;
767 u32 total;
768 int slot = path->slots[0];
769 u32 name_len;
770 char *start;
771 int ret = 0;
772 int num = 0;
773 int index;
774 u64 dir;
775 unsigned long name_off;
776 unsigned long elem_size;
777 unsigned long ptr;
779 p = fs_path_alloc_reversed();
780 if (!p)
781 return -ENOMEM;
783 tmp_path = alloc_path_for_send();
784 if (!tmp_path) {
785 fs_path_free(p);
786 return -ENOMEM;
790 if (found_key->type == BTRFS_INODE_REF_KEY) {
791 ptr = (unsigned long)btrfs_item_ptr(eb, slot,
792 struct btrfs_inode_ref);
793 item = btrfs_item_nr(eb, slot);
794 total = btrfs_item_size(eb, item);
795 elem_size = sizeof(*iref);
796 } else {
797 ptr = btrfs_item_ptr_offset(eb, slot);
798 total = btrfs_item_size_nr(eb, slot);
799 elem_size = sizeof(*extref);
802 while (cur < total) {
803 fs_path_reset(p);
805 if (found_key->type == BTRFS_INODE_REF_KEY) {
806 iref = (struct btrfs_inode_ref *)(ptr + cur);
807 name_len = btrfs_inode_ref_name_len(eb, iref);
808 name_off = (unsigned long)(iref + 1);
809 index = btrfs_inode_ref_index(eb, iref);
810 dir = found_key->offset;
811 } else {
812 extref = (struct btrfs_inode_extref *)(ptr + cur);
813 name_len = btrfs_inode_extref_name_len(eb, extref);
814 name_off = (unsigned long)&extref->name;
815 index = btrfs_inode_extref_index(eb, extref);
816 dir = btrfs_inode_extref_parent(eb, extref);
819 if (resolve) {
820 start = btrfs_ref_to_path(root, tmp_path, name_len,
821 name_off, eb, dir,
822 p->buf, p->buf_len);
823 if (IS_ERR(start)) {
824 ret = PTR_ERR(start);
825 goto out;
827 if (start < p->buf) {
828 /* overflow , try again with larger buffer */
829 ret = fs_path_ensure_buf(p,
830 p->buf_len + p->buf - start);
831 if (ret < 0)
832 goto out;
833 start = btrfs_ref_to_path(root, tmp_path,
834 name_len, name_off,
835 eb, dir,
836 p->buf, p->buf_len);
837 if (IS_ERR(start)) {
838 ret = PTR_ERR(start);
839 goto out;
841 BUG_ON(start < p->buf);
843 p->start = start;
844 } else {
845 ret = fs_path_add_from_extent_buffer(p, eb, name_off,
846 name_len);
847 if (ret < 0)
848 goto out;
851 cur += elem_size + name_len;
852 ret = iterate(num, dir, index, p, ctx);
853 if (ret)
854 goto out;
855 num++;
858 out:
859 btrfs_free_path(tmp_path);
860 fs_path_free(p);
861 return ret;
864 typedef int (*iterate_dir_item_t)(int num, struct btrfs_key *di_key,
865 const char *name, int name_len,
866 const char *data, int data_len,
867 u8 type, void *ctx);
870 * Helper function to iterate the entries in ONE btrfs_dir_item.
871 * The iterate callback may return a non zero value to stop iteration. This can
872 * be a negative value for error codes or 1 to simply stop it.
874 * path must point to the dir item when called.
876 static int iterate_dir_item(struct btrfs_root *root, struct btrfs_path *path,
877 struct btrfs_key *found_key,
878 iterate_dir_item_t iterate, void *ctx)
880 int ret = 0;
881 struct extent_buffer *eb;
882 struct btrfs_item *item;
883 struct btrfs_dir_item *di;
884 struct btrfs_key di_key;
885 char *buf = NULL;
886 char *buf2 = NULL;
887 int buf_len;
888 int buf_virtual = 0;
889 u32 name_len;
890 u32 data_len;
891 u32 cur;
892 u32 len;
893 u32 total;
894 int slot;
895 int num;
896 u8 type;
898 buf_len = PAGE_SIZE;
899 buf = kmalloc(buf_len, GFP_NOFS);
900 if (!buf) {
901 ret = -ENOMEM;
902 goto out;
905 eb = path->nodes[0];
906 slot = path->slots[0];
907 item = btrfs_item_nr(eb, slot);
908 di = btrfs_item_ptr(eb, slot, struct btrfs_dir_item);
909 cur = 0;
910 len = 0;
911 total = btrfs_item_size(eb, item);
913 num = 0;
914 while (cur < total) {
915 name_len = btrfs_dir_name_len(eb, di);
916 data_len = btrfs_dir_data_len(eb, di);
917 type = btrfs_dir_type(eb, di);
918 btrfs_dir_item_key_to_cpu(eb, di, &di_key);
920 if (name_len + data_len > buf_len) {
921 buf_len = PAGE_ALIGN(name_len + data_len);
922 if (buf_virtual) {
923 buf2 = vmalloc(buf_len);
924 if (!buf2) {
925 ret = -ENOMEM;
926 goto out;
928 vfree(buf);
929 } else {
930 buf2 = krealloc(buf, buf_len, GFP_NOFS);
931 if (!buf2) {
932 buf2 = vmalloc(buf_len);
933 if (!buf2) {
934 ret = -ENOMEM;
935 goto out;
937 kfree(buf);
938 buf_virtual = 1;
942 buf = buf2;
943 buf2 = NULL;
946 read_extent_buffer(eb, buf, (unsigned long)(di + 1),
947 name_len + data_len);
949 len = sizeof(*di) + name_len + data_len;
950 di = (struct btrfs_dir_item *)((char *)di + len);
951 cur += len;
953 ret = iterate(num, &di_key, buf, name_len, buf + name_len,
954 data_len, type, ctx);
955 if (ret < 0)
956 goto out;
957 if (ret) {
958 ret = 0;
959 goto out;
962 num++;
965 out:
966 if (buf_virtual)
967 vfree(buf);
968 else
969 kfree(buf);
970 return ret;
973 static int __copy_first_ref(int num, u64 dir, int index,
974 struct fs_path *p, void *ctx)
976 int ret;
977 struct fs_path *pt = ctx;
979 ret = fs_path_copy(pt, p);
980 if (ret < 0)
981 return ret;
983 /* we want the first only */
984 return 1;
988 * Retrieve the first path of an inode. If an inode has more then one
989 * ref/hardlink, this is ignored.
991 static int get_inode_path(struct btrfs_root *root,
992 u64 ino, struct fs_path *path)
994 int ret;
995 struct btrfs_key key, found_key;
996 struct btrfs_path *p;
998 p = alloc_path_for_send();
999 if (!p)
1000 return -ENOMEM;
1002 fs_path_reset(path);
1004 key.objectid = ino;
1005 key.type = BTRFS_INODE_REF_KEY;
1006 key.offset = 0;
1008 ret = btrfs_search_slot_for_read(root, &key, p, 1, 0);
1009 if (ret < 0)
1010 goto out;
1011 if (ret) {
1012 ret = 1;
1013 goto out;
1015 btrfs_item_key_to_cpu(p->nodes[0], &found_key, p->slots[0]);
1016 if (found_key.objectid != ino ||
1017 (found_key.type != BTRFS_INODE_REF_KEY &&
1018 found_key.type != BTRFS_INODE_EXTREF_KEY)) {
1019 ret = -ENOENT;
1020 goto out;
1023 ret = iterate_inode_ref(root, p, &found_key, 1,
1024 __copy_first_ref, path);
1025 if (ret < 0)
1026 goto out;
1027 ret = 0;
1029 out:
1030 btrfs_free_path(p);
1031 return ret;
1034 struct backref_ctx {
1035 struct send_ctx *sctx;
1037 /* number of total found references */
1038 u64 found;
1041 * used for clones found in send_root. clones found behind cur_objectid
1042 * and cur_offset are not considered as allowed clones.
1044 u64 cur_objectid;
1045 u64 cur_offset;
1047 /* may be truncated in case it's the last extent in a file */
1048 u64 extent_len;
1050 /* Just to check for bugs in backref resolving */
1051 int found_itself;
1054 static int __clone_root_cmp_bsearch(const void *key, const void *elt)
1056 u64 root = (u64)(uintptr_t)key;
1057 struct clone_root *cr = (struct clone_root *)elt;
1059 if (root < cr->root->objectid)
1060 return -1;
1061 if (root > cr->root->objectid)
1062 return 1;
1063 return 0;
1066 static int __clone_root_cmp_sort(const void *e1, const void *e2)
1068 struct clone_root *cr1 = (struct clone_root *)e1;
1069 struct clone_root *cr2 = (struct clone_root *)e2;
1071 if (cr1->root->objectid < cr2->root->objectid)
1072 return -1;
1073 if (cr1->root->objectid > cr2->root->objectid)
1074 return 1;
1075 return 0;
1079 * Called for every backref that is found for the current extent.
1080 * Results are collected in sctx->clone_roots->ino/offset/found_refs
1082 static int __iterate_backrefs(u64 ino, u64 offset, u64 root, void *ctx_)
1084 struct backref_ctx *bctx = ctx_;
1085 struct clone_root *found;
1086 int ret;
1087 u64 i_size;
1089 /* First check if the root is in the list of accepted clone sources */
1090 found = bsearch((void *)(uintptr_t)root, bctx->sctx->clone_roots,
1091 bctx->sctx->clone_roots_cnt,
1092 sizeof(struct clone_root),
1093 __clone_root_cmp_bsearch);
1094 if (!found)
1095 return 0;
1097 if (found->root == bctx->sctx->send_root &&
1098 ino == bctx->cur_objectid &&
1099 offset == bctx->cur_offset) {
1100 bctx->found_itself = 1;
1104 * There are inodes that have extents that lie behind its i_size. Don't
1105 * accept clones from these extents.
1107 ret = get_inode_info(found->root, ino, &i_size, NULL, NULL, NULL, NULL,
1108 NULL);
1109 if (ret < 0)
1110 return ret;
1112 if (offset + bctx->extent_len > i_size)
1113 return 0;
1116 * Make sure we don't consider clones from send_root that are
1117 * behind the current inode/offset.
1119 if (found->root == bctx->sctx->send_root) {
1121 * TODO for the moment we don't accept clones from the inode
1122 * that is currently send. We may change this when
1123 * BTRFS_IOC_CLONE_RANGE supports cloning from and to the same
1124 * file.
1126 if (ino >= bctx->cur_objectid)
1127 return 0;
1128 #if 0
1129 if (ino > bctx->cur_objectid)
1130 return 0;
1131 if (offset + bctx->extent_len > bctx->cur_offset)
1132 return 0;
1133 #endif
1136 bctx->found++;
1137 found->found_refs++;
1138 if (ino < found->ino) {
1139 found->ino = ino;
1140 found->offset = offset;
1141 } else if (found->ino == ino) {
1143 * same extent found more then once in the same file.
1145 if (found->offset > offset + bctx->extent_len)
1146 found->offset = offset;
1149 return 0;
1153 * Given an inode, offset and extent item, it finds a good clone for a clone
1154 * instruction. Returns -ENOENT when none could be found. The function makes
1155 * sure that the returned clone is usable at the point where sending is at the
1156 * moment. This means, that no clones are accepted which lie behind the current
1157 * inode+offset.
1159 * path must point to the extent item when called.
1161 static int find_extent_clone(struct send_ctx *sctx,
1162 struct btrfs_path *path,
1163 u64 ino, u64 data_offset,
1164 u64 ino_size,
1165 struct clone_root **found)
1167 int ret;
1168 int extent_type;
1169 u64 logical;
1170 u64 disk_byte;
1171 u64 num_bytes;
1172 u64 extent_item_pos;
1173 u64 flags = 0;
1174 struct btrfs_file_extent_item *fi;
1175 struct extent_buffer *eb = path->nodes[0];
1176 struct backref_ctx *backref_ctx = NULL;
1177 struct clone_root *cur_clone_root;
1178 struct btrfs_key found_key;
1179 struct btrfs_path *tmp_path;
1180 int compressed;
1181 u32 i;
1183 tmp_path = alloc_path_for_send();
1184 if (!tmp_path)
1185 return -ENOMEM;
1187 backref_ctx = kmalloc(sizeof(*backref_ctx), GFP_NOFS);
1188 if (!backref_ctx) {
1189 ret = -ENOMEM;
1190 goto out;
1193 if (data_offset >= ino_size) {
1195 * There may be extents that lie behind the file's size.
1196 * I at least had this in combination with snapshotting while
1197 * writing large files.
1199 ret = 0;
1200 goto out;
1203 fi = btrfs_item_ptr(eb, path->slots[0],
1204 struct btrfs_file_extent_item);
1205 extent_type = btrfs_file_extent_type(eb, fi);
1206 if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
1207 ret = -ENOENT;
1208 goto out;
1210 compressed = btrfs_file_extent_compression(eb, fi);
1212 num_bytes = btrfs_file_extent_num_bytes(eb, fi);
1213 disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
1214 if (disk_byte == 0) {
1215 ret = -ENOENT;
1216 goto out;
1218 logical = disk_byte + btrfs_file_extent_offset(eb, fi);
1220 ret = extent_from_logical(sctx->send_root->fs_info, disk_byte, tmp_path,
1221 &found_key, &flags);
1222 btrfs_release_path(tmp_path);
1224 if (ret < 0)
1225 goto out;
1226 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
1227 ret = -EIO;
1228 goto out;
1232 * Setup the clone roots.
1234 for (i = 0; i < sctx->clone_roots_cnt; i++) {
1235 cur_clone_root = sctx->clone_roots + i;
1236 cur_clone_root->ino = (u64)-1;
1237 cur_clone_root->offset = 0;
1238 cur_clone_root->found_refs = 0;
1241 backref_ctx->sctx = sctx;
1242 backref_ctx->found = 0;
1243 backref_ctx->cur_objectid = ino;
1244 backref_ctx->cur_offset = data_offset;
1245 backref_ctx->found_itself = 0;
1246 backref_ctx->extent_len = num_bytes;
1249 * The last extent of a file may be too large due to page alignment.
1250 * We need to adjust extent_len in this case so that the checks in
1251 * __iterate_backrefs work.
1253 if (data_offset + num_bytes >= ino_size)
1254 backref_ctx->extent_len = ino_size - data_offset;
1257 * Now collect all backrefs.
1259 if (compressed == BTRFS_COMPRESS_NONE)
1260 extent_item_pos = logical - found_key.objectid;
1261 else
1262 extent_item_pos = 0;
1264 extent_item_pos = logical - found_key.objectid;
1265 ret = iterate_extent_inodes(sctx->send_root->fs_info,
1266 found_key.objectid, extent_item_pos, 1,
1267 __iterate_backrefs, backref_ctx);
1269 if (ret < 0)
1270 goto out;
1272 if (!backref_ctx->found_itself) {
1273 /* found a bug in backref code? */
1274 ret = -EIO;
1275 printk(KERN_ERR "btrfs: ERROR did not find backref in "
1276 "send_root. inode=%llu, offset=%llu, "
1277 "disk_byte=%llu found extent=%llu\n",
1278 ino, data_offset, disk_byte, found_key.objectid);
1279 goto out;
1282 verbose_printk(KERN_DEBUG "btrfs: find_extent_clone: data_offset=%llu, "
1283 "ino=%llu, "
1284 "num_bytes=%llu, logical=%llu\n",
1285 data_offset, ino, num_bytes, logical);
1287 if (!backref_ctx->found)
1288 verbose_printk("btrfs: no clones found\n");
1290 cur_clone_root = NULL;
1291 for (i = 0; i < sctx->clone_roots_cnt; i++) {
1292 if (sctx->clone_roots[i].found_refs) {
1293 if (!cur_clone_root)
1294 cur_clone_root = sctx->clone_roots + i;
1295 else if (sctx->clone_roots[i].root == sctx->send_root)
1296 /* prefer clones from send_root over others */
1297 cur_clone_root = sctx->clone_roots + i;
1302 if (cur_clone_root) {
1303 *found = cur_clone_root;
1304 ret = 0;
1305 } else {
1306 ret = -ENOENT;
1309 out:
1310 btrfs_free_path(tmp_path);
1311 kfree(backref_ctx);
1312 return ret;
1315 static int read_symlink(struct btrfs_root *root,
1316 u64 ino,
1317 struct fs_path *dest)
1319 int ret;
1320 struct btrfs_path *path;
1321 struct btrfs_key key;
1322 struct btrfs_file_extent_item *ei;
1323 u8 type;
1324 u8 compression;
1325 unsigned long off;
1326 int len;
1328 path = alloc_path_for_send();
1329 if (!path)
1330 return -ENOMEM;
1332 key.objectid = ino;
1333 key.type = BTRFS_EXTENT_DATA_KEY;
1334 key.offset = 0;
1335 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1336 if (ret < 0)
1337 goto out;
1338 BUG_ON(ret);
1340 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
1341 struct btrfs_file_extent_item);
1342 type = btrfs_file_extent_type(path->nodes[0], ei);
1343 compression = btrfs_file_extent_compression(path->nodes[0], ei);
1344 BUG_ON(type != BTRFS_FILE_EXTENT_INLINE);
1345 BUG_ON(compression);
1347 off = btrfs_file_extent_inline_start(ei);
1348 len = btrfs_file_extent_inline_len(path->nodes[0], ei);
1350 ret = fs_path_add_from_extent_buffer(dest, path->nodes[0], off, len);
1352 out:
1353 btrfs_free_path(path);
1354 return ret;
1358 * Helper function to generate a file name that is unique in the root of
1359 * send_root and parent_root. This is used to generate names for orphan inodes.
1361 static int gen_unique_name(struct send_ctx *sctx,
1362 u64 ino, u64 gen,
1363 struct fs_path *dest)
1365 int ret = 0;
1366 struct btrfs_path *path;
1367 struct btrfs_dir_item *di;
1368 char tmp[64];
1369 int len;
1370 u64 idx = 0;
1372 path = alloc_path_for_send();
1373 if (!path)
1374 return -ENOMEM;
1376 while (1) {
1377 len = snprintf(tmp, sizeof(tmp) - 1, "o%llu-%llu-%llu",
1378 ino, gen, idx);
1379 if (len >= sizeof(tmp)) {
1380 /* should really not happen */
1381 ret = -EOVERFLOW;
1382 goto out;
1385 di = btrfs_lookup_dir_item(NULL, sctx->send_root,
1386 path, BTRFS_FIRST_FREE_OBJECTID,
1387 tmp, strlen(tmp), 0);
1388 btrfs_release_path(path);
1389 if (IS_ERR(di)) {
1390 ret = PTR_ERR(di);
1391 goto out;
1393 if (di) {
1394 /* not unique, try again */
1395 idx++;
1396 continue;
1399 if (!sctx->parent_root) {
1400 /* unique */
1401 ret = 0;
1402 break;
1405 di = btrfs_lookup_dir_item(NULL, sctx->parent_root,
1406 path, BTRFS_FIRST_FREE_OBJECTID,
1407 tmp, strlen(tmp), 0);
1408 btrfs_release_path(path);
1409 if (IS_ERR(di)) {
1410 ret = PTR_ERR(di);
1411 goto out;
1413 if (di) {
1414 /* not unique, try again */
1415 idx++;
1416 continue;
1418 /* unique */
1419 break;
1422 ret = fs_path_add(dest, tmp, strlen(tmp));
1424 out:
1425 btrfs_free_path(path);
1426 return ret;
1429 enum inode_state {
1430 inode_state_no_change,
1431 inode_state_will_create,
1432 inode_state_did_create,
1433 inode_state_will_delete,
1434 inode_state_did_delete,
1437 static int get_cur_inode_state(struct send_ctx *sctx, u64 ino, u64 gen)
1439 int ret;
1440 int left_ret;
1441 int right_ret;
1442 u64 left_gen;
1443 u64 right_gen;
1445 ret = get_inode_info(sctx->send_root, ino, NULL, &left_gen, NULL, NULL,
1446 NULL, NULL);
1447 if (ret < 0 && ret != -ENOENT)
1448 goto out;
1449 left_ret = ret;
1451 if (!sctx->parent_root) {
1452 right_ret = -ENOENT;
1453 } else {
1454 ret = get_inode_info(sctx->parent_root, ino, NULL, &right_gen,
1455 NULL, NULL, NULL, NULL);
1456 if (ret < 0 && ret != -ENOENT)
1457 goto out;
1458 right_ret = ret;
1461 if (!left_ret && !right_ret) {
1462 if (left_gen == gen && right_gen == gen) {
1463 ret = inode_state_no_change;
1464 } else if (left_gen == gen) {
1465 if (ino < sctx->send_progress)
1466 ret = inode_state_did_create;
1467 else
1468 ret = inode_state_will_create;
1469 } else if (right_gen == gen) {
1470 if (ino < sctx->send_progress)
1471 ret = inode_state_did_delete;
1472 else
1473 ret = inode_state_will_delete;
1474 } else {
1475 ret = -ENOENT;
1477 } else if (!left_ret) {
1478 if (left_gen == gen) {
1479 if (ino < sctx->send_progress)
1480 ret = inode_state_did_create;
1481 else
1482 ret = inode_state_will_create;
1483 } else {
1484 ret = -ENOENT;
1486 } else if (!right_ret) {
1487 if (right_gen == gen) {
1488 if (ino < sctx->send_progress)
1489 ret = inode_state_did_delete;
1490 else
1491 ret = inode_state_will_delete;
1492 } else {
1493 ret = -ENOENT;
1495 } else {
1496 ret = -ENOENT;
1499 out:
1500 return ret;
1503 static int is_inode_existent(struct send_ctx *sctx, u64 ino, u64 gen)
1505 int ret;
1507 ret = get_cur_inode_state(sctx, ino, gen);
1508 if (ret < 0)
1509 goto out;
1511 if (ret == inode_state_no_change ||
1512 ret == inode_state_did_create ||
1513 ret == inode_state_will_delete)
1514 ret = 1;
1515 else
1516 ret = 0;
1518 out:
1519 return ret;
1523 * Helper function to lookup a dir item in a dir.
1525 static int lookup_dir_item_inode(struct btrfs_root *root,
1526 u64 dir, const char *name, int name_len,
1527 u64 *found_inode,
1528 u8 *found_type)
1530 int ret = 0;
1531 struct btrfs_dir_item *di;
1532 struct btrfs_key key;
1533 struct btrfs_path *path;
1535 path = alloc_path_for_send();
1536 if (!path)
1537 return -ENOMEM;
1539 di = btrfs_lookup_dir_item(NULL, root, path,
1540 dir, name, name_len, 0);
1541 if (!di) {
1542 ret = -ENOENT;
1543 goto out;
1545 if (IS_ERR(di)) {
1546 ret = PTR_ERR(di);
1547 goto out;
1549 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1550 *found_inode = key.objectid;
1551 *found_type = btrfs_dir_type(path->nodes[0], di);
1553 out:
1554 btrfs_free_path(path);
1555 return ret;
1559 * Looks up the first btrfs_inode_ref of a given ino. It returns the parent dir,
1560 * generation of the parent dir and the name of the dir entry.
1562 static int get_first_ref(struct btrfs_root *root, u64 ino,
1563 u64 *dir, u64 *dir_gen, struct fs_path *name)
1565 int ret;
1566 struct btrfs_key key;
1567 struct btrfs_key found_key;
1568 struct btrfs_path *path;
1569 int len;
1570 u64 parent_dir;
1572 path = alloc_path_for_send();
1573 if (!path)
1574 return -ENOMEM;
1576 key.objectid = ino;
1577 key.type = BTRFS_INODE_REF_KEY;
1578 key.offset = 0;
1580 ret = btrfs_search_slot_for_read(root, &key, path, 1, 0);
1581 if (ret < 0)
1582 goto out;
1583 if (!ret)
1584 btrfs_item_key_to_cpu(path->nodes[0], &found_key,
1585 path->slots[0]);
1586 if (ret || found_key.objectid != ino ||
1587 (found_key.type != BTRFS_INODE_REF_KEY &&
1588 found_key.type != BTRFS_INODE_EXTREF_KEY)) {
1589 ret = -ENOENT;
1590 goto out;
1593 if (key.type == BTRFS_INODE_REF_KEY) {
1594 struct btrfs_inode_ref *iref;
1595 iref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1596 struct btrfs_inode_ref);
1597 len = btrfs_inode_ref_name_len(path->nodes[0], iref);
1598 ret = fs_path_add_from_extent_buffer(name, path->nodes[0],
1599 (unsigned long)(iref + 1),
1600 len);
1601 parent_dir = found_key.offset;
1602 } else {
1603 struct btrfs_inode_extref *extref;
1604 extref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1605 struct btrfs_inode_extref);
1606 len = btrfs_inode_extref_name_len(path->nodes[0], extref);
1607 ret = fs_path_add_from_extent_buffer(name, path->nodes[0],
1608 (unsigned long)&extref->name, len);
1609 parent_dir = btrfs_inode_extref_parent(path->nodes[0], extref);
1611 if (ret < 0)
1612 goto out;
1613 btrfs_release_path(path);
1615 ret = get_inode_info(root, parent_dir, NULL, dir_gen, NULL, NULL,
1616 NULL, NULL);
1617 if (ret < 0)
1618 goto out;
1620 *dir = parent_dir;
1622 out:
1623 btrfs_free_path(path);
1624 return ret;
1627 static int is_first_ref(struct btrfs_root *root,
1628 u64 ino, u64 dir,
1629 const char *name, int name_len)
1631 int ret;
1632 struct fs_path *tmp_name;
1633 u64 tmp_dir;
1634 u64 tmp_dir_gen;
1636 tmp_name = fs_path_alloc();
1637 if (!tmp_name)
1638 return -ENOMEM;
1640 ret = get_first_ref(root, ino, &tmp_dir, &tmp_dir_gen, tmp_name);
1641 if (ret < 0)
1642 goto out;
1644 if (dir != tmp_dir || name_len != fs_path_len(tmp_name)) {
1645 ret = 0;
1646 goto out;
1649 ret = !memcmp(tmp_name->start, name, name_len);
1651 out:
1652 fs_path_free(tmp_name);
1653 return ret;
1657 * Used by process_recorded_refs to determine if a new ref would overwrite an
1658 * already existing ref. In case it detects an overwrite, it returns the
1659 * inode/gen in who_ino/who_gen.
1660 * When an overwrite is detected, process_recorded_refs does proper orphanizing
1661 * to make sure later references to the overwritten inode are possible.
1662 * Orphanizing is however only required for the first ref of an inode.
1663 * process_recorded_refs does an additional is_first_ref check to see if
1664 * orphanizing is really required.
1666 static int will_overwrite_ref(struct send_ctx *sctx, u64 dir, u64 dir_gen,
1667 const char *name, int name_len,
1668 u64 *who_ino, u64 *who_gen)
1670 int ret = 0;
1671 u64 other_inode = 0;
1672 u8 other_type = 0;
1674 if (!sctx->parent_root)
1675 goto out;
1677 ret = is_inode_existent(sctx, dir, dir_gen);
1678 if (ret <= 0)
1679 goto out;
1681 ret = lookup_dir_item_inode(sctx->parent_root, dir, name, name_len,
1682 &other_inode, &other_type);
1683 if (ret < 0 && ret != -ENOENT)
1684 goto out;
1685 if (ret) {
1686 ret = 0;
1687 goto out;
1691 * Check if the overwritten ref was already processed. If yes, the ref
1692 * was already unlinked/moved, so we can safely assume that we will not
1693 * overwrite anything at this point in time.
1695 if (other_inode > sctx->send_progress) {
1696 ret = get_inode_info(sctx->parent_root, other_inode, NULL,
1697 who_gen, NULL, NULL, NULL, NULL);
1698 if (ret < 0)
1699 goto out;
1701 ret = 1;
1702 *who_ino = other_inode;
1703 } else {
1704 ret = 0;
1707 out:
1708 return ret;
1712 * Checks if the ref was overwritten by an already processed inode. This is
1713 * used by __get_cur_name_and_parent to find out if the ref was orphanized and
1714 * thus the orphan name needs be used.
1715 * process_recorded_refs also uses it to avoid unlinking of refs that were
1716 * overwritten.
1718 static int did_overwrite_ref(struct send_ctx *sctx,
1719 u64 dir, u64 dir_gen,
1720 u64 ino, u64 ino_gen,
1721 const char *name, int name_len)
1723 int ret = 0;
1724 u64 gen;
1725 u64 ow_inode;
1726 u8 other_type;
1728 if (!sctx->parent_root)
1729 goto out;
1731 ret = is_inode_existent(sctx, dir, dir_gen);
1732 if (ret <= 0)
1733 goto out;
1735 /* check if the ref was overwritten by another ref */
1736 ret = lookup_dir_item_inode(sctx->send_root, dir, name, name_len,
1737 &ow_inode, &other_type);
1738 if (ret < 0 && ret != -ENOENT)
1739 goto out;
1740 if (ret) {
1741 /* was never and will never be overwritten */
1742 ret = 0;
1743 goto out;
1746 ret = get_inode_info(sctx->send_root, ow_inode, NULL, &gen, NULL, NULL,
1747 NULL, NULL);
1748 if (ret < 0)
1749 goto out;
1751 if (ow_inode == ino && gen == ino_gen) {
1752 ret = 0;
1753 goto out;
1756 /* we know that it is or will be overwritten. check this now */
1757 if (ow_inode < sctx->send_progress)
1758 ret = 1;
1759 else
1760 ret = 0;
1762 out:
1763 return ret;
1767 * Same as did_overwrite_ref, but also checks if it is the first ref of an inode
1768 * that got overwritten. This is used by process_recorded_refs to determine
1769 * if it has to use the path as returned by get_cur_path or the orphan name.
1771 static int did_overwrite_first_ref(struct send_ctx *sctx, u64 ino, u64 gen)
1773 int ret = 0;
1774 struct fs_path *name = NULL;
1775 u64 dir;
1776 u64 dir_gen;
1778 if (!sctx->parent_root)
1779 goto out;
1781 name = fs_path_alloc();
1782 if (!name)
1783 return -ENOMEM;
1785 ret = get_first_ref(sctx->parent_root, ino, &dir, &dir_gen, name);
1786 if (ret < 0)
1787 goto out;
1789 ret = did_overwrite_ref(sctx, dir, dir_gen, ino, gen,
1790 name->start, fs_path_len(name));
1792 out:
1793 fs_path_free(name);
1794 return ret;
1798 * Insert a name cache entry. On 32bit kernels the radix tree index is 32bit,
1799 * so we need to do some special handling in case we have clashes. This function
1800 * takes care of this with the help of name_cache_entry::radix_list.
1801 * In case of error, nce is kfreed.
1803 static int name_cache_insert(struct send_ctx *sctx,
1804 struct name_cache_entry *nce)
1806 int ret = 0;
1807 struct list_head *nce_head;
1809 nce_head = radix_tree_lookup(&sctx->name_cache,
1810 (unsigned long)nce->ino);
1811 if (!nce_head) {
1812 nce_head = kmalloc(sizeof(*nce_head), GFP_NOFS);
1813 if (!nce_head) {
1814 kfree(nce);
1815 return -ENOMEM;
1817 INIT_LIST_HEAD(nce_head);
1819 ret = radix_tree_insert(&sctx->name_cache, nce->ino, nce_head);
1820 if (ret < 0) {
1821 kfree(nce_head);
1822 kfree(nce);
1823 return ret;
1826 list_add_tail(&nce->radix_list, nce_head);
1827 list_add_tail(&nce->list, &sctx->name_cache_list);
1828 sctx->name_cache_size++;
1830 return ret;
1833 static void name_cache_delete(struct send_ctx *sctx,
1834 struct name_cache_entry *nce)
1836 struct list_head *nce_head;
1838 nce_head = radix_tree_lookup(&sctx->name_cache,
1839 (unsigned long)nce->ino);
1840 BUG_ON(!nce_head);
1842 list_del(&nce->radix_list);
1843 list_del(&nce->list);
1844 sctx->name_cache_size--;
1846 if (list_empty(nce_head)) {
1847 radix_tree_delete(&sctx->name_cache, (unsigned long)nce->ino);
1848 kfree(nce_head);
1852 static struct name_cache_entry *name_cache_search(struct send_ctx *sctx,
1853 u64 ino, u64 gen)
1855 struct list_head *nce_head;
1856 struct name_cache_entry *cur;
1858 nce_head = radix_tree_lookup(&sctx->name_cache, (unsigned long)ino);
1859 if (!nce_head)
1860 return NULL;
1862 list_for_each_entry(cur, nce_head, radix_list) {
1863 if (cur->ino == ino && cur->gen == gen)
1864 return cur;
1866 return NULL;
1870 * Removes the entry from the list and adds it back to the end. This marks the
1871 * entry as recently used so that name_cache_clean_unused does not remove it.
1873 static void name_cache_used(struct send_ctx *sctx, struct name_cache_entry *nce)
1875 list_del(&nce->list);
1876 list_add_tail(&nce->list, &sctx->name_cache_list);
1880 * Remove some entries from the beginning of name_cache_list.
1882 static void name_cache_clean_unused(struct send_ctx *sctx)
1884 struct name_cache_entry *nce;
1886 if (sctx->name_cache_size < SEND_CTX_NAME_CACHE_CLEAN_SIZE)
1887 return;
1889 while (sctx->name_cache_size > SEND_CTX_MAX_NAME_CACHE_SIZE) {
1890 nce = list_entry(sctx->name_cache_list.next,
1891 struct name_cache_entry, list);
1892 name_cache_delete(sctx, nce);
1893 kfree(nce);
1897 static void name_cache_free(struct send_ctx *sctx)
1899 struct name_cache_entry *nce;
1901 while (!list_empty(&sctx->name_cache_list)) {
1902 nce = list_entry(sctx->name_cache_list.next,
1903 struct name_cache_entry, list);
1904 name_cache_delete(sctx, nce);
1905 kfree(nce);
1910 * Used by get_cur_path for each ref up to the root.
1911 * Returns 0 if it succeeded.
1912 * Returns 1 if the inode is not existent or got overwritten. In that case, the
1913 * name is an orphan name. This instructs get_cur_path to stop iterating. If 1
1914 * is returned, parent_ino/parent_gen are not guaranteed to be valid.
1915 * Returns <0 in case of error.
1917 static int __get_cur_name_and_parent(struct send_ctx *sctx,
1918 u64 ino, u64 gen,
1919 u64 *parent_ino,
1920 u64 *parent_gen,
1921 struct fs_path *dest)
1923 int ret;
1924 int nce_ret;
1925 struct btrfs_path *path = NULL;
1926 struct name_cache_entry *nce = NULL;
1929 * First check if we already did a call to this function with the same
1930 * ino/gen. If yes, check if the cache entry is still up-to-date. If yes
1931 * return the cached result.
1933 nce = name_cache_search(sctx, ino, gen);
1934 if (nce) {
1935 if (ino < sctx->send_progress && nce->need_later_update) {
1936 name_cache_delete(sctx, nce);
1937 kfree(nce);
1938 nce = NULL;
1939 } else {
1940 name_cache_used(sctx, nce);
1941 *parent_ino = nce->parent_ino;
1942 *parent_gen = nce->parent_gen;
1943 ret = fs_path_add(dest, nce->name, nce->name_len);
1944 if (ret < 0)
1945 goto out;
1946 ret = nce->ret;
1947 goto out;
1951 path = alloc_path_for_send();
1952 if (!path)
1953 return -ENOMEM;
1956 * If the inode is not existent yet, add the orphan name and return 1.
1957 * This should only happen for the parent dir that we determine in
1958 * __record_new_ref
1960 ret = is_inode_existent(sctx, ino, gen);
1961 if (ret < 0)
1962 goto out;
1964 if (!ret) {
1965 ret = gen_unique_name(sctx, ino, gen, dest);
1966 if (ret < 0)
1967 goto out;
1968 ret = 1;
1969 goto out_cache;
1973 * Depending on whether the inode was already processed or not, use
1974 * send_root or parent_root for ref lookup.
1976 if (ino < sctx->send_progress)
1977 ret = get_first_ref(sctx->send_root, ino,
1978 parent_ino, parent_gen, dest);
1979 else
1980 ret = get_first_ref(sctx->parent_root, ino,
1981 parent_ino, parent_gen, dest);
1982 if (ret < 0)
1983 goto out;
1986 * Check if the ref was overwritten by an inode's ref that was processed
1987 * earlier. If yes, treat as orphan and return 1.
1989 ret = did_overwrite_ref(sctx, *parent_ino, *parent_gen, ino, gen,
1990 dest->start, dest->end - dest->start);
1991 if (ret < 0)
1992 goto out;
1993 if (ret) {
1994 fs_path_reset(dest);
1995 ret = gen_unique_name(sctx, ino, gen, dest);
1996 if (ret < 0)
1997 goto out;
1998 ret = 1;
2001 out_cache:
2003 * Store the result of the lookup in the name cache.
2005 nce = kmalloc(sizeof(*nce) + fs_path_len(dest) + 1, GFP_NOFS);
2006 if (!nce) {
2007 ret = -ENOMEM;
2008 goto out;
2011 nce->ino = ino;
2012 nce->gen = gen;
2013 nce->parent_ino = *parent_ino;
2014 nce->parent_gen = *parent_gen;
2015 nce->name_len = fs_path_len(dest);
2016 nce->ret = ret;
2017 strcpy(nce->name, dest->start);
2019 if (ino < sctx->send_progress)
2020 nce->need_later_update = 0;
2021 else
2022 nce->need_later_update = 1;
2024 nce_ret = name_cache_insert(sctx, nce);
2025 if (nce_ret < 0)
2026 ret = nce_ret;
2027 name_cache_clean_unused(sctx);
2029 out:
2030 btrfs_free_path(path);
2031 return ret;
2035 * Magic happens here. This function returns the first ref to an inode as it
2036 * would look like while receiving the stream at this point in time.
2037 * We walk the path up to the root. For every inode in between, we check if it
2038 * was already processed/sent. If yes, we continue with the parent as found
2039 * in send_root. If not, we continue with the parent as found in parent_root.
2040 * If we encounter an inode that was deleted at this point in time, we use the
2041 * inodes "orphan" name instead of the real name and stop. Same with new inodes
2042 * that were not created yet and overwritten inodes/refs.
2044 * When do we have have orphan inodes:
2045 * 1. When an inode is freshly created and thus no valid refs are available yet
2046 * 2. When a directory lost all it's refs (deleted) but still has dir items
2047 * inside which were not processed yet (pending for move/delete). If anyone
2048 * tried to get the path to the dir items, it would get a path inside that
2049 * orphan directory.
2050 * 3. When an inode is moved around or gets new links, it may overwrite the ref
2051 * of an unprocessed inode. If in that case the first ref would be
2052 * overwritten, the overwritten inode gets "orphanized". Later when we
2053 * process this overwritten inode, it is restored at a new place by moving
2054 * the orphan inode.
2056 * sctx->send_progress tells this function at which point in time receiving
2057 * would be.
2059 static int get_cur_path(struct send_ctx *sctx, u64 ino, u64 gen,
2060 struct fs_path *dest)
2062 int ret = 0;
2063 struct fs_path *name = NULL;
2064 u64 parent_inode = 0;
2065 u64 parent_gen = 0;
2066 int stop = 0;
2068 name = fs_path_alloc();
2069 if (!name) {
2070 ret = -ENOMEM;
2071 goto out;
2074 dest->reversed = 1;
2075 fs_path_reset(dest);
2077 while (!stop && ino != BTRFS_FIRST_FREE_OBJECTID) {
2078 fs_path_reset(name);
2080 ret = __get_cur_name_and_parent(sctx, ino, gen,
2081 &parent_inode, &parent_gen, name);
2082 if (ret < 0)
2083 goto out;
2084 if (ret)
2085 stop = 1;
2087 ret = fs_path_add_path(dest, name);
2088 if (ret < 0)
2089 goto out;
2091 ino = parent_inode;
2092 gen = parent_gen;
2095 out:
2096 fs_path_free(name);
2097 if (!ret)
2098 fs_path_unreverse(dest);
2099 return ret;
2103 * Called for regular files when sending extents data. Opens a struct file
2104 * to read from the file.
2106 static int open_cur_inode_file(struct send_ctx *sctx)
2108 int ret = 0;
2109 struct btrfs_key key;
2110 struct path path;
2111 struct inode *inode;
2112 struct dentry *dentry;
2113 struct file *filp;
2114 int new = 0;
2116 if (sctx->cur_inode_filp)
2117 goto out;
2119 key.objectid = sctx->cur_ino;
2120 key.type = BTRFS_INODE_ITEM_KEY;
2121 key.offset = 0;
2123 inode = btrfs_iget(sctx->send_root->fs_info->sb, &key, sctx->send_root,
2124 &new);
2125 if (IS_ERR(inode)) {
2126 ret = PTR_ERR(inode);
2127 goto out;
2130 dentry = d_obtain_alias(inode);
2131 inode = NULL;
2132 if (IS_ERR(dentry)) {
2133 ret = PTR_ERR(dentry);
2134 goto out;
2137 path.mnt = sctx->mnt;
2138 path.dentry = dentry;
2139 filp = dentry_open(&path, O_RDONLY | O_LARGEFILE, current_cred());
2140 dput(dentry);
2141 dentry = NULL;
2142 if (IS_ERR(filp)) {
2143 ret = PTR_ERR(filp);
2144 goto out;
2146 sctx->cur_inode_filp = filp;
2148 out:
2150 * no xxxput required here as every vfs op
2151 * does it by itself on failure
2153 return ret;
2157 * Closes the struct file that was created in open_cur_inode_file
2159 static int close_cur_inode_file(struct send_ctx *sctx)
2161 int ret = 0;
2163 if (!sctx->cur_inode_filp)
2164 goto out;
2166 ret = filp_close(sctx->cur_inode_filp, NULL);
2167 sctx->cur_inode_filp = NULL;
2169 out:
2170 return ret;
2174 * Sends a BTRFS_SEND_C_SUBVOL command/item to userspace
2176 static int send_subvol_begin(struct send_ctx *sctx)
2178 int ret;
2179 struct btrfs_root *send_root = sctx->send_root;
2180 struct btrfs_root *parent_root = sctx->parent_root;
2181 struct btrfs_path *path;
2182 struct btrfs_key key;
2183 struct btrfs_root_ref *ref;
2184 struct extent_buffer *leaf;
2185 char *name = NULL;
2186 int namelen;
2188 path = alloc_path_for_send();
2189 if (!path)
2190 return -ENOMEM;
2192 name = kmalloc(BTRFS_PATH_NAME_MAX, GFP_NOFS);
2193 if (!name) {
2194 btrfs_free_path(path);
2195 return -ENOMEM;
2198 key.objectid = send_root->objectid;
2199 key.type = BTRFS_ROOT_BACKREF_KEY;
2200 key.offset = 0;
2202 ret = btrfs_search_slot_for_read(send_root->fs_info->tree_root,
2203 &key, path, 1, 0);
2204 if (ret < 0)
2205 goto out;
2206 if (ret) {
2207 ret = -ENOENT;
2208 goto out;
2211 leaf = path->nodes[0];
2212 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
2213 if (key.type != BTRFS_ROOT_BACKREF_KEY ||
2214 key.objectid != send_root->objectid) {
2215 ret = -ENOENT;
2216 goto out;
2218 ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_root_ref);
2219 namelen = btrfs_root_ref_name_len(leaf, ref);
2220 read_extent_buffer(leaf, name, (unsigned long)(ref + 1), namelen);
2221 btrfs_release_path(path);
2223 if (parent_root) {
2224 ret = begin_cmd(sctx, BTRFS_SEND_C_SNAPSHOT);
2225 if (ret < 0)
2226 goto out;
2227 } else {
2228 ret = begin_cmd(sctx, BTRFS_SEND_C_SUBVOL);
2229 if (ret < 0)
2230 goto out;
2233 TLV_PUT_STRING(sctx, BTRFS_SEND_A_PATH, name, namelen);
2234 TLV_PUT_UUID(sctx, BTRFS_SEND_A_UUID,
2235 sctx->send_root->root_item.uuid);
2236 TLV_PUT_U64(sctx, BTRFS_SEND_A_CTRANSID,
2237 sctx->send_root->root_item.ctransid);
2238 if (parent_root) {
2239 TLV_PUT_UUID(sctx, BTRFS_SEND_A_CLONE_UUID,
2240 sctx->parent_root->root_item.uuid);
2241 TLV_PUT_U64(sctx, BTRFS_SEND_A_CLONE_CTRANSID,
2242 sctx->parent_root->root_item.ctransid);
2245 ret = send_cmd(sctx);
2247 tlv_put_failure:
2248 out:
2249 btrfs_free_path(path);
2250 kfree(name);
2251 return ret;
2254 static int send_truncate(struct send_ctx *sctx, u64 ino, u64 gen, u64 size)
2256 int ret = 0;
2257 struct fs_path *p;
2259 verbose_printk("btrfs: send_truncate %llu size=%llu\n", ino, size);
2261 p = fs_path_alloc();
2262 if (!p)
2263 return -ENOMEM;
2265 ret = begin_cmd(sctx, BTRFS_SEND_C_TRUNCATE);
2266 if (ret < 0)
2267 goto out;
2269 ret = get_cur_path(sctx, ino, gen, p);
2270 if (ret < 0)
2271 goto out;
2272 TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, p);
2273 TLV_PUT_U64(sctx, BTRFS_SEND_A_SIZE, size);
2275 ret = send_cmd(sctx);
2277 tlv_put_failure:
2278 out:
2279 fs_path_free(p);
2280 return ret;
2283 static int send_chmod(struct send_ctx *sctx, u64 ino, u64 gen, u64 mode)
2285 int ret = 0;
2286 struct fs_path *p;
2288 verbose_printk("btrfs: send_chmod %llu mode=%llu\n", ino, mode);
2290 p = fs_path_alloc();
2291 if (!p)
2292 return -ENOMEM;
2294 ret = begin_cmd(sctx, BTRFS_SEND_C_CHMOD);
2295 if (ret < 0)
2296 goto out;
2298 ret = get_cur_path(sctx, ino, gen, p);
2299 if (ret < 0)
2300 goto out;
2301 TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, p);
2302 TLV_PUT_U64(sctx, BTRFS_SEND_A_MODE, mode & 07777);
2304 ret = send_cmd(sctx);
2306 tlv_put_failure:
2307 out:
2308 fs_path_free(p);
2309 return ret;
2312 static int send_chown(struct send_ctx *sctx, u64 ino, u64 gen, u64 uid, u64 gid)
2314 int ret = 0;
2315 struct fs_path *p;
2317 verbose_printk("btrfs: send_chown %llu uid=%llu, gid=%llu\n", ino, uid, gid);
2319 p = fs_path_alloc();
2320 if (!p)
2321 return -ENOMEM;
2323 ret = begin_cmd(sctx, BTRFS_SEND_C_CHOWN);
2324 if (ret < 0)
2325 goto out;
2327 ret = get_cur_path(sctx, ino, gen, p);
2328 if (ret < 0)
2329 goto out;
2330 TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, p);
2331 TLV_PUT_U64(sctx, BTRFS_SEND_A_UID, uid);
2332 TLV_PUT_U64(sctx, BTRFS_SEND_A_GID, gid);
2334 ret = send_cmd(sctx);
2336 tlv_put_failure:
2337 out:
2338 fs_path_free(p);
2339 return ret;
2342 static int send_utimes(struct send_ctx *sctx, u64 ino, u64 gen)
2344 int ret = 0;
2345 struct fs_path *p = NULL;
2346 struct btrfs_inode_item *ii;
2347 struct btrfs_path *path = NULL;
2348 struct extent_buffer *eb;
2349 struct btrfs_key key;
2350 int slot;
2352 verbose_printk("btrfs: send_utimes %llu\n", ino);
2354 p = fs_path_alloc();
2355 if (!p)
2356 return -ENOMEM;
2358 path = alloc_path_for_send();
2359 if (!path) {
2360 ret = -ENOMEM;
2361 goto out;
2364 key.objectid = ino;
2365 key.type = BTRFS_INODE_ITEM_KEY;
2366 key.offset = 0;
2367 ret = btrfs_search_slot(NULL, sctx->send_root, &key, path, 0, 0);
2368 if (ret < 0)
2369 goto out;
2371 eb = path->nodes[0];
2372 slot = path->slots[0];
2373 ii = btrfs_item_ptr(eb, slot, struct btrfs_inode_item);
2375 ret = begin_cmd(sctx, BTRFS_SEND_C_UTIMES);
2376 if (ret < 0)
2377 goto out;
2379 ret = get_cur_path(sctx, ino, gen, p);
2380 if (ret < 0)
2381 goto out;
2382 TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, p);
2383 TLV_PUT_BTRFS_TIMESPEC(sctx, BTRFS_SEND_A_ATIME, eb,
2384 btrfs_inode_atime(ii));
2385 TLV_PUT_BTRFS_TIMESPEC(sctx, BTRFS_SEND_A_MTIME, eb,
2386 btrfs_inode_mtime(ii));
2387 TLV_PUT_BTRFS_TIMESPEC(sctx, BTRFS_SEND_A_CTIME, eb,
2388 btrfs_inode_ctime(ii));
2389 /* TODO Add otime support when the otime patches get into upstream */
2391 ret = send_cmd(sctx);
2393 tlv_put_failure:
2394 out:
2395 fs_path_free(p);
2396 btrfs_free_path(path);
2397 return ret;
2401 * Sends a BTRFS_SEND_C_MKXXX or SYMLINK command to user space. We don't have
2402 * a valid path yet because we did not process the refs yet. So, the inode
2403 * is created as orphan.
2405 static int send_create_inode(struct send_ctx *sctx, u64 ino)
2407 int ret = 0;
2408 struct fs_path *p;
2409 int cmd;
2410 u64 gen;
2411 u64 mode;
2412 u64 rdev;
2414 verbose_printk("btrfs: send_create_inode %llu\n", ino);
2416 p = fs_path_alloc();
2417 if (!p)
2418 return -ENOMEM;
2420 ret = get_inode_info(sctx->send_root, ino, NULL, &gen, &mode, NULL,
2421 NULL, &rdev);
2422 if (ret < 0)
2423 goto out;
2425 if (S_ISREG(mode)) {
2426 cmd = BTRFS_SEND_C_MKFILE;
2427 } else if (S_ISDIR(mode)) {
2428 cmd = BTRFS_SEND_C_MKDIR;
2429 } else if (S_ISLNK(mode)) {
2430 cmd = BTRFS_SEND_C_SYMLINK;
2431 } else if (S_ISCHR(mode) || S_ISBLK(mode)) {
2432 cmd = BTRFS_SEND_C_MKNOD;
2433 } else if (S_ISFIFO(mode)) {
2434 cmd = BTRFS_SEND_C_MKFIFO;
2435 } else if (S_ISSOCK(mode)) {
2436 cmd = BTRFS_SEND_C_MKSOCK;
2437 } else {
2438 printk(KERN_WARNING "btrfs: unexpected inode type %o",
2439 (int)(mode & S_IFMT));
2440 ret = -ENOTSUPP;
2441 goto out;
2444 ret = begin_cmd(sctx, cmd);
2445 if (ret < 0)
2446 goto out;
2448 ret = gen_unique_name(sctx, ino, gen, p);
2449 if (ret < 0)
2450 goto out;
2452 TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, p);
2453 TLV_PUT_U64(sctx, BTRFS_SEND_A_INO, ino);
2455 if (S_ISLNK(mode)) {
2456 fs_path_reset(p);
2457 ret = read_symlink(sctx->send_root, ino, p);
2458 if (ret < 0)
2459 goto out;
2460 TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH_LINK, p);
2461 } else if (S_ISCHR(mode) || S_ISBLK(mode) ||
2462 S_ISFIFO(mode) || S_ISSOCK(mode)) {
2463 TLV_PUT_U64(sctx, BTRFS_SEND_A_RDEV, new_encode_dev(rdev));
2464 TLV_PUT_U64(sctx, BTRFS_SEND_A_MODE, mode);
2467 ret = send_cmd(sctx);
2468 if (ret < 0)
2469 goto out;
2472 tlv_put_failure:
2473 out:
2474 fs_path_free(p);
2475 return ret;
2479 * We need some special handling for inodes that get processed before the parent
2480 * directory got created. See process_recorded_refs for details.
2481 * This function does the check if we already created the dir out of order.
2483 static int did_create_dir(struct send_ctx *sctx, u64 dir)
2485 int ret = 0;
2486 struct btrfs_path *path = NULL;
2487 struct btrfs_key key;
2488 struct btrfs_key found_key;
2489 struct btrfs_key di_key;
2490 struct extent_buffer *eb;
2491 struct btrfs_dir_item *di;
2492 int slot;
2494 path = alloc_path_for_send();
2495 if (!path) {
2496 ret = -ENOMEM;
2497 goto out;
2500 key.objectid = dir;
2501 key.type = BTRFS_DIR_INDEX_KEY;
2502 key.offset = 0;
2503 while (1) {
2504 ret = btrfs_search_slot_for_read(sctx->send_root, &key, path,
2505 1, 0);
2506 if (ret < 0)
2507 goto out;
2508 if (!ret) {
2509 eb = path->nodes[0];
2510 slot = path->slots[0];
2511 btrfs_item_key_to_cpu(eb, &found_key, slot);
2513 if (ret || found_key.objectid != key.objectid ||
2514 found_key.type != key.type) {
2515 ret = 0;
2516 goto out;
2519 di = btrfs_item_ptr(eb, slot, struct btrfs_dir_item);
2520 btrfs_dir_item_key_to_cpu(eb, di, &di_key);
2522 if (di_key.objectid < sctx->send_progress) {
2523 ret = 1;
2524 goto out;
2527 key.offset = found_key.offset + 1;
2528 btrfs_release_path(path);
2531 out:
2532 btrfs_free_path(path);
2533 return ret;
2537 * Only creates the inode if it is:
2538 * 1. Not a directory
2539 * 2. Or a directory which was not created already due to out of order
2540 * directories. See did_create_dir and process_recorded_refs for details.
2542 static int send_create_inode_if_needed(struct send_ctx *sctx)
2544 int ret;
2546 if (S_ISDIR(sctx->cur_inode_mode)) {
2547 ret = did_create_dir(sctx, sctx->cur_ino);
2548 if (ret < 0)
2549 goto out;
2550 if (ret) {
2551 ret = 0;
2552 goto out;
2556 ret = send_create_inode(sctx, sctx->cur_ino);
2557 if (ret < 0)
2558 goto out;
2560 out:
2561 return ret;
2564 struct recorded_ref {
2565 struct list_head list;
2566 char *dir_path;
2567 char *name;
2568 struct fs_path *full_path;
2569 u64 dir;
2570 u64 dir_gen;
2571 int dir_path_len;
2572 int name_len;
2576 * We need to process new refs before deleted refs, but compare_tree gives us
2577 * everything mixed. So we first record all refs and later process them.
2578 * This function is a helper to record one ref.
2580 static int record_ref(struct list_head *head, u64 dir,
2581 u64 dir_gen, struct fs_path *path)
2583 struct recorded_ref *ref;
2584 char *tmp;
2586 ref = kmalloc(sizeof(*ref), GFP_NOFS);
2587 if (!ref)
2588 return -ENOMEM;
2590 ref->dir = dir;
2591 ref->dir_gen = dir_gen;
2592 ref->full_path = path;
2594 tmp = strrchr(ref->full_path->start, '/');
2595 if (!tmp) {
2596 ref->name_len = ref->full_path->end - ref->full_path->start;
2597 ref->name = ref->full_path->start;
2598 ref->dir_path_len = 0;
2599 ref->dir_path = ref->full_path->start;
2600 } else {
2601 tmp++;
2602 ref->name_len = ref->full_path->end - tmp;
2603 ref->name = tmp;
2604 ref->dir_path = ref->full_path->start;
2605 ref->dir_path_len = ref->full_path->end -
2606 ref->full_path->start - 1 - ref->name_len;
2609 list_add_tail(&ref->list, head);
2610 return 0;
2613 static void __free_recorded_refs(struct list_head *head)
2615 struct recorded_ref *cur;
2617 while (!list_empty(head)) {
2618 cur = list_entry(head->next, struct recorded_ref, list);
2619 fs_path_free(cur->full_path);
2620 list_del(&cur->list);
2621 kfree(cur);
2625 static void free_recorded_refs(struct send_ctx *sctx)
2627 __free_recorded_refs(&sctx->new_refs);
2628 __free_recorded_refs(&sctx->deleted_refs);
2632 * Renames/moves a file/dir to its orphan name. Used when the first
2633 * ref of an unprocessed inode gets overwritten and for all non empty
2634 * directories.
2636 static int orphanize_inode(struct send_ctx *sctx, u64 ino, u64 gen,
2637 struct fs_path *path)
2639 int ret;
2640 struct fs_path *orphan;
2642 orphan = fs_path_alloc();
2643 if (!orphan)
2644 return -ENOMEM;
2646 ret = gen_unique_name(sctx, ino, gen, orphan);
2647 if (ret < 0)
2648 goto out;
2650 ret = send_rename(sctx, path, orphan);
2652 out:
2653 fs_path_free(orphan);
2654 return ret;
2658 * Returns 1 if a directory can be removed at this point in time.
2659 * We check this by iterating all dir items and checking if the inode behind
2660 * the dir item was already processed.
2662 static int can_rmdir(struct send_ctx *sctx, u64 dir, u64 send_progress)
2664 int ret = 0;
2665 struct btrfs_root *root = sctx->parent_root;
2666 struct btrfs_path *path;
2667 struct btrfs_key key;
2668 struct btrfs_key found_key;
2669 struct btrfs_key loc;
2670 struct btrfs_dir_item *di;
2673 * Don't try to rmdir the top/root subvolume dir.
2675 if (dir == BTRFS_FIRST_FREE_OBJECTID)
2676 return 0;
2678 path = alloc_path_for_send();
2679 if (!path)
2680 return -ENOMEM;
2682 key.objectid = dir;
2683 key.type = BTRFS_DIR_INDEX_KEY;
2684 key.offset = 0;
2686 while (1) {
2687 ret = btrfs_search_slot_for_read(root, &key, path, 1, 0);
2688 if (ret < 0)
2689 goto out;
2690 if (!ret) {
2691 btrfs_item_key_to_cpu(path->nodes[0], &found_key,
2692 path->slots[0]);
2694 if (ret || found_key.objectid != key.objectid ||
2695 found_key.type != key.type) {
2696 break;
2699 di = btrfs_item_ptr(path->nodes[0], path->slots[0],
2700 struct btrfs_dir_item);
2701 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &loc);
2703 if (loc.objectid > send_progress) {
2704 ret = 0;
2705 goto out;
2708 btrfs_release_path(path);
2709 key.offset = found_key.offset + 1;
2712 ret = 1;
2714 out:
2715 btrfs_free_path(path);
2716 return ret;
2720 * This does all the move/link/unlink/rmdir magic.
2722 static int process_recorded_refs(struct send_ctx *sctx)
2724 int ret = 0;
2725 struct recorded_ref *cur;
2726 struct recorded_ref *cur2;
2727 struct ulist *check_dirs = NULL;
2728 struct ulist_iterator uit;
2729 struct ulist_node *un;
2730 struct fs_path *valid_path = NULL;
2731 u64 ow_inode = 0;
2732 u64 ow_gen;
2733 int did_overwrite = 0;
2734 int is_orphan = 0;
2736 verbose_printk("btrfs: process_recorded_refs %llu\n", sctx->cur_ino);
2739 * This should never happen as the root dir always has the same ref
2740 * which is always '..'
2742 BUG_ON(sctx->cur_ino <= BTRFS_FIRST_FREE_OBJECTID);
2744 valid_path = fs_path_alloc();
2745 if (!valid_path) {
2746 ret = -ENOMEM;
2747 goto out;
2750 check_dirs = ulist_alloc(GFP_NOFS);
2751 if (!check_dirs) {
2752 ret = -ENOMEM;
2753 goto out;
2757 * First, check if the first ref of the current inode was overwritten
2758 * before. If yes, we know that the current inode was already orphanized
2759 * and thus use the orphan name. If not, we can use get_cur_path to
2760 * get the path of the first ref as it would like while receiving at
2761 * this point in time.
2762 * New inodes are always orphan at the beginning, so force to use the
2763 * orphan name in this case.
2764 * The first ref is stored in valid_path and will be updated if it
2765 * gets moved around.
2767 if (!sctx->cur_inode_new) {
2768 ret = did_overwrite_first_ref(sctx, sctx->cur_ino,
2769 sctx->cur_inode_gen);
2770 if (ret < 0)
2771 goto out;
2772 if (ret)
2773 did_overwrite = 1;
2775 if (sctx->cur_inode_new || did_overwrite) {
2776 ret = gen_unique_name(sctx, sctx->cur_ino,
2777 sctx->cur_inode_gen, valid_path);
2778 if (ret < 0)
2779 goto out;
2780 is_orphan = 1;
2781 } else {
2782 ret = get_cur_path(sctx, sctx->cur_ino, sctx->cur_inode_gen,
2783 valid_path);
2784 if (ret < 0)
2785 goto out;
2788 list_for_each_entry(cur, &sctx->new_refs, list) {
2790 * We may have refs where the parent directory does not exist
2791 * yet. This happens if the parent directories inum is higher
2792 * the the current inum. To handle this case, we create the
2793 * parent directory out of order. But we need to check if this
2794 * did already happen before due to other refs in the same dir.
2796 ret = get_cur_inode_state(sctx, cur->dir, cur->dir_gen);
2797 if (ret < 0)
2798 goto out;
2799 if (ret == inode_state_will_create) {
2800 ret = 0;
2802 * First check if any of the current inodes refs did
2803 * already create the dir.
2805 list_for_each_entry(cur2, &sctx->new_refs, list) {
2806 if (cur == cur2)
2807 break;
2808 if (cur2->dir == cur->dir) {
2809 ret = 1;
2810 break;
2815 * If that did not happen, check if a previous inode
2816 * did already create the dir.
2818 if (!ret)
2819 ret = did_create_dir(sctx, cur->dir);
2820 if (ret < 0)
2821 goto out;
2822 if (!ret) {
2823 ret = send_create_inode(sctx, cur->dir);
2824 if (ret < 0)
2825 goto out;
2830 * Check if this new ref would overwrite the first ref of
2831 * another unprocessed inode. If yes, orphanize the
2832 * overwritten inode. If we find an overwritten ref that is
2833 * not the first ref, simply unlink it.
2835 ret = will_overwrite_ref(sctx, cur->dir, cur->dir_gen,
2836 cur->name, cur->name_len,
2837 &ow_inode, &ow_gen);
2838 if (ret < 0)
2839 goto out;
2840 if (ret) {
2841 ret = is_first_ref(sctx->parent_root,
2842 ow_inode, cur->dir, cur->name,
2843 cur->name_len);
2844 if (ret < 0)
2845 goto out;
2846 if (ret) {
2847 ret = orphanize_inode(sctx, ow_inode, ow_gen,
2848 cur->full_path);
2849 if (ret < 0)
2850 goto out;
2851 } else {
2852 ret = send_unlink(sctx, cur->full_path);
2853 if (ret < 0)
2854 goto out;
2859 * link/move the ref to the new place. If we have an orphan
2860 * inode, move it and update valid_path. If not, link or move
2861 * it depending on the inode mode.
2863 if (is_orphan) {
2864 ret = send_rename(sctx, valid_path, cur->full_path);
2865 if (ret < 0)
2866 goto out;
2867 is_orphan = 0;
2868 ret = fs_path_copy(valid_path, cur->full_path);
2869 if (ret < 0)
2870 goto out;
2871 } else {
2872 if (S_ISDIR(sctx->cur_inode_mode)) {
2874 * Dirs can't be linked, so move it. For moved
2875 * dirs, we always have one new and one deleted
2876 * ref. The deleted ref is ignored later.
2878 ret = send_rename(sctx, valid_path,
2879 cur->full_path);
2880 if (ret < 0)
2881 goto out;
2882 ret = fs_path_copy(valid_path, cur->full_path);
2883 if (ret < 0)
2884 goto out;
2885 } else {
2886 ret = send_link(sctx, cur->full_path,
2887 valid_path);
2888 if (ret < 0)
2889 goto out;
2892 ret = ulist_add(check_dirs, cur->dir, cur->dir_gen,
2893 GFP_NOFS);
2894 if (ret < 0)
2895 goto out;
2898 if (S_ISDIR(sctx->cur_inode_mode) && sctx->cur_inode_deleted) {
2900 * Check if we can already rmdir the directory. If not,
2901 * orphanize it. For every dir item inside that gets deleted
2902 * later, we do this check again and rmdir it then if possible.
2903 * See the use of check_dirs for more details.
2905 ret = can_rmdir(sctx, sctx->cur_ino, sctx->cur_ino);
2906 if (ret < 0)
2907 goto out;
2908 if (ret) {
2909 ret = send_rmdir(sctx, valid_path);
2910 if (ret < 0)
2911 goto out;
2912 } else if (!is_orphan) {
2913 ret = orphanize_inode(sctx, sctx->cur_ino,
2914 sctx->cur_inode_gen, valid_path);
2915 if (ret < 0)
2916 goto out;
2917 is_orphan = 1;
2920 list_for_each_entry(cur, &sctx->deleted_refs, list) {
2921 ret = ulist_add(check_dirs, cur->dir, cur->dir_gen,
2922 GFP_NOFS);
2923 if (ret < 0)
2924 goto out;
2926 } else if (S_ISDIR(sctx->cur_inode_mode) &&
2927 !list_empty(&sctx->deleted_refs)) {
2929 * We have a moved dir. Add the old parent to check_dirs
2931 cur = list_entry(sctx->deleted_refs.next, struct recorded_ref,
2932 list);
2933 ret = ulist_add(check_dirs, cur->dir, cur->dir_gen,
2934 GFP_NOFS);
2935 if (ret < 0)
2936 goto out;
2937 } else if (!S_ISDIR(sctx->cur_inode_mode)) {
2939 * We have a non dir inode. Go through all deleted refs and
2940 * unlink them if they were not already overwritten by other
2941 * inodes.
2943 list_for_each_entry(cur, &sctx->deleted_refs, list) {
2944 ret = did_overwrite_ref(sctx, cur->dir, cur->dir_gen,
2945 sctx->cur_ino, sctx->cur_inode_gen,
2946 cur->name, cur->name_len);
2947 if (ret < 0)
2948 goto out;
2949 if (!ret) {
2950 ret = send_unlink(sctx, cur->full_path);
2951 if (ret < 0)
2952 goto out;
2954 ret = ulist_add(check_dirs, cur->dir, cur->dir_gen,
2955 GFP_NOFS);
2956 if (ret < 0)
2957 goto out;
2961 * If the inode is still orphan, unlink the orphan. This may
2962 * happen when a previous inode did overwrite the first ref
2963 * of this inode and no new refs were added for the current
2964 * inode. Unlinking does not mean that the inode is deleted in
2965 * all cases. There may still be links to this inode in other
2966 * places.
2968 if (is_orphan) {
2969 ret = send_unlink(sctx, valid_path);
2970 if (ret < 0)
2971 goto out;
2976 * We did collect all parent dirs where cur_inode was once located. We
2977 * now go through all these dirs and check if they are pending for
2978 * deletion and if it's finally possible to perform the rmdir now.
2979 * We also update the inode stats of the parent dirs here.
2981 ULIST_ITER_INIT(&uit);
2982 while ((un = ulist_next(check_dirs, &uit))) {
2984 * In case we had refs into dirs that were not processed yet,
2985 * we don't need to do the utime and rmdir logic for these dirs.
2986 * The dir will be processed later.
2988 if (un->val > sctx->cur_ino)
2989 continue;
2991 ret = get_cur_inode_state(sctx, un->val, un->aux);
2992 if (ret < 0)
2993 goto out;
2995 if (ret == inode_state_did_create ||
2996 ret == inode_state_no_change) {
2997 /* TODO delayed utimes */
2998 ret = send_utimes(sctx, un->val, un->aux);
2999 if (ret < 0)
3000 goto out;
3001 } else if (ret == inode_state_did_delete) {
3002 ret = can_rmdir(sctx, un->val, sctx->cur_ino);
3003 if (ret < 0)
3004 goto out;
3005 if (ret) {
3006 ret = get_cur_path(sctx, un->val, un->aux,
3007 valid_path);
3008 if (ret < 0)
3009 goto out;
3010 ret = send_rmdir(sctx, valid_path);
3011 if (ret < 0)
3012 goto out;
3017 ret = 0;
3019 out:
3020 free_recorded_refs(sctx);
3021 ulist_free(check_dirs);
3022 fs_path_free(valid_path);
3023 return ret;
3026 static int __record_new_ref(int num, u64 dir, int index,
3027 struct fs_path *name,
3028 void *ctx)
3030 int ret = 0;
3031 struct send_ctx *sctx = ctx;
3032 struct fs_path *p;
3033 u64 gen;
3035 p = fs_path_alloc();
3036 if (!p)
3037 return -ENOMEM;
3039 ret = get_inode_info(sctx->send_root, dir, NULL, &gen, NULL, NULL,
3040 NULL, NULL);
3041 if (ret < 0)
3042 goto out;
3044 ret = get_cur_path(sctx, dir, gen, p);
3045 if (ret < 0)
3046 goto out;
3047 ret = fs_path_add_path(p, name);
3048 if (ret < 0)
3049 goto out;
3051 ret = record_ref(&sctx->new_refs, dir, gen, p);
3053 out:
3054 if (ret)
3055 fs_path_free(p);
3056 return ret;
3059 static int __record_deleted_ref(int num, u64 dir, int index,
3060 struct fs_path *name,
3061 void *ctx)
3063 int ret = 0;
3064 struct send_ctx *sctx = ctx;
3065 struct fs_path *p;
3066 u64 gen;
3068 p = fs_path_alloc();
3069 if (!p)
3070 return -ENOMEM;
3072 ret = get_inode_info(sctx->parent_root, dir, NULL, &gen, NULL, NULL,
3073 NULL, NULL);
3074 if (ret < 0)
3075 goto out;
3077 ret = get_cur_path(sctx, dir, gen, p);
3078 if (ret < 0)
3079 goto out;
3080 ret = fs_path_add_path(p, name);
3081 if (ret < 0)
3082 goto out;
3084 ret = record_ref(&sctx->deleted_refs, dir, gen, p);
3086 out:
3087 if (ret)
3088 fs_path_free(p);
3089 return ret;
3092 static int record_new_ref(struct send_ctx *sctx)
3094 int ret;
3096 ret = iterate_inode_ref(sctx->send_root, sctx->left_path,
3097 sctx->cmp_key, 0, __record_new_ref, sctx);
3098 if (ret < 0)
3099 goto out;
3100 ret = 0;
3102 out:
3103 return ret;
3106 static int record_deleted_ref(struct send_ctx *sctx)
3108 int ret;
3110 ret = iterate_inode_ref(sctx->parent_root, sctx->right_path,
3111 sctx->cmp_key, 0, __record_deleted_ref, sctx);
3112 if (ret < 0)
3113 goto out;
3114 ret = 0;
3116 out:
3117 return ret;
3120 struct find_ref_ctx {
3121 u64 dir;
3122 struct fs_path *name;
3123 int found_idx;
3126 static int __find_iref(int num, u64 dir, int index,
3127 struct fs_path *name,
3128 void *ctx_)
3130 struct find_ref_ctx *ctx = ctx_;
3132 if (dir == ctx->dir && fs_path_len(name) == fs_path_len(ctx->name) &&
3133 strncmp(name->start, ctx->name->start, fs_path_len(name)) == 0) {
3134 ctx->found_idx = num;
3135 return 1;
3137 return 0;
3140 static int find_iref(struct btrfs_root *root,
3141 struct btrfs_path *path,
3142 struct btrfs_key *key,
3143 u64 dir, struct fs_path *name)
3145 int ret;
3146 struct find_ref_ctx ctx;
3148 ctx.dir = dir;
3149 ctx.name = name;
3150 ctx.found_idx = -1;
3152 ret = iterate_inode_ref(root, path, key, 0, __find_iref, &ctx);
3153 if (ret < 0)
3154 return ret;
3156 if (ctx.found_idx == -1)
3157 return -ENOENT;
3159 return ctx.found_idx;
3162 static int __record_changed_new_ref(int num, u64 dir, int index,
3163 struct fs_path *name,
3164 void *ctx)
3166 int ret;
3167 struct send_ctx *sctx = ctx;
3169 ret = find_iref(sctx->parent_root, sctx->right_path,
3170 sctx->cmp_key, dir, name);
3171 if (ret == -ENOENT)
3172 ret = __record_new_ref(num, dir, index, name, sctx);
3173 else if (ret > 0)
3174 ret = 0;
3176 return ret;
3179 static int __record_changed_deleted_ref(int num, u64 dir, int index,
3180 struct fs_path *name,
3181 void *ctx)
3183 int ret;
3184 struct send_ctx *sctx = ctx;
3186 ret = find_iref(sctx->send_root, sctx->left_path, sctx->cmp_key,
3187 dir, name);
3188 if (ret == -ENOENT)
3189 ret = __record_deleted_ref(num, dir, index, name, sctx);
3190 else if (ret > 0)
3191 ret = 0;
3193 return ret;
3196 static int record_changed_ref(struct send_ctx *sctx)
3198 int ret = 0;
3200 ret = iterate_inode_ref(sctx->send_root, sctx->left_path,
3201 sctx->cmp_key, 0, __record_changed_new_ref, sctx);
3202 if (ret < 0)
3203 goto out;
3204 ret = iterate_inode_ref(sctx->parent_root, sctx->right_path,
3205 sctx->cmp_key, 0, __record_changed_deleted_ref, sctx);
3206 if (ret < 0)
3207 goto out;
3208 ret = 0;
3210 out:
3211 return ret;
3215 * Record and process all refs at once. Needed when an inode changes the
3216 * generation number, which means that it was deleted and recreated.
3218 static int process_all_refs(struct send_ctx *sctx,
3219 enum btrfs_compare_tree_result cmd)
3221 int ret;
3222 struct btrfs_root *root;
3223 struct btrfs_path *path;
3224 struct btrfs_key key;
3225 struct btrfs_key found_key;
3226 struct extent_buffer *eb;
3227 int slot;
3228 iterate_inode_ref_t cb;
3230 path = alloc_path_for_send();
3231 if (!path)
3232 return -ENOMEM;
3234 if (cmd == BTRFS_COMPARE_TREE_NEW) {
3235 root = sctx->send_root;
3236 cb = __record_new_ref;
3237 } else if (cmd == BTRFS_COMPARE_TREE_DELETED) {
3238 root = sctx->parent_root;
3239 cb = __record_deleted_ref;
3240 } else {
3241 BUG();
3244 key.objectid = sctx->cmp_key->objectid;
3245 key.type = BTRFS_INODE_REF_KEY;
3246 key.offset = 0;
3247 while (1) {
3248 ret = btrfs_search_slot_for_read(root, &key, path, 1, 0);
3249 if (ret < 0)
3250 goto out;
3251 if (ret)
3252 break;
3254 eb = path->nodes[0];
3255 slot = path->slots[0];
3256 btrfs_item_key_to_cpu(eb, &found_key, slot);
3258 if (found_key.objectid != key.objectid ||
3259 (found_key.type != BTRFS_INODE_REF_KEY &&
3260 found_key.type != BTRFS_INODE_EXTREF_KEY))
3261 break;
3263 ret = iterate_inode_ref(root, path, &found_key, 0, cb, sctx);
3264 btrfs_release_path(path);
3265 if (ret < 0)
3266 goto out;
3268 key.offset = found_key.offset + 1;
3270 btrfs_release_path(path);
3272 ret = process_recorded_refs(sctx);
3274 out:
3275 btrfs_free_path(path);
3276 return ret;
3279 static int send_set_xattr(struct send_ctx *sctx,
3280 struct fs_path *path,
3281 const char *name, int name_len,
3282 const char *data, int data_len)
3284 int ret = 0;
3286 ret = begin_cmd(sctx, BTRFS_SEND_C_SET_XATTR);
3287 if (ret < 0)
3288 goto out;
3290 TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, path);
3291 TLV_PUT_STRING(sctx, BTRFS_SEND_A_XATTR_NAME, name, name_len);
3292 TLV_PUT(sctx, BTRFS_SEND_A_XATTR_DATA, data, data_len);
3294 ret = send_cmd(sctx);
3296 tlv_put_failure:
3297 out:
3298 return ret;
3301 static int send_remove_xattr(struct send_ctx *sctx,
3302 struct fs_path *path,
3303 const char *name, int name_len)
3305 int ret = 0;
3307 ret = begin_cmd(sctx, BTRFS_SEND_C_REMOVE_XATTR);
3308 if (ret < 0)
3309 goto out;
3311 TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, path);
3312 TLV_PUT_STRING(sctx, BTRFS_SEND_A_XATTR_NAME, name, name_len);
3314 ret = send_cmd(sctx);
3316 tlv_put_failure:
3317 out:
3318 return ret;
3321 static int __process_new_xattr(int num, struct btrfs_key *di_key,
3322 const char *name, int name_len,
3323 const char *data, int data_len,
3324 u8 type, void *ctx)
3326 int ret;
3327 struct send_ctx *sctx = ctx;
3328 struct fs_path *p;
3329 posix_acl_xattr_header dummy_acl;
3331 p = fs_path_alloc();
3332 if (!p)
3333 return -ENOMEM;
3336 * This hack is needed because empty acl's are stored as zero byte
3337 * data in xattrs. Problem with that is, that receiving these zero byte
3338 * acl's will fail later. To fix this, we send a dummy acl list that
3339 * only contains the version number and no entries.
3341 if (!strncmp(name, XATTR_NAME_POSIX_ACL_ACCESS, name_len) ||
3342 !strncmp(name, XATTR_NAME_POSIX_ACL_DEFAULT, name_len)) {
3343 if (data_len == 0) {
3344 dummy_acl.a_version =
3345 cpu_to_le32(POSIX_ACL_XATTR_VERSION);
3346 data = (char *)&dummy_acl;
3347 data_len = sizeof(dummy_acl);
3351 ret = get_cur_path(sctx, sctx->cur_ino, sctx->cur_inode_gen, p);
3352 if (ret < 0)
3353 goto out;
3355 ret = send_set_xattr(sctx, p, name, name_len, data, data_len);
3357 out:
3358 fs_path_free(p);
3359 return ret;
3362 static int __process_deleted_xattr(int num, struct btrfs_key *di_key,
3363 const char *name, int name_len,
3364 const char *data, int data_len,
3365 u8 type, void *ctx)
3367 int ret;
3368 struct send_ctx *sctx = ctx;
3369 struct fs_path *p;
3371 p = fs_path_alloc();
3372 if (!p)
3373 return -ENOMEM;
3375 ret = get_cur_path(sctx, sctx->cur_ino, sctx->cur_inode_gen, p);
3376 if (ret < 0)
3377 goto out;
3379 ret = send_remove_xattr(sctx, p, name, name_len);
3381 out:
3382 fs_path_free(p);
3383 return ret;
3386 static int process_new_xattr(struct send_ctx *sctx)
3388 int ret = 0;
3390 ret = iterate_dir_item(sctx->send_root, sctx->left_path,
3391 sctx->cmp_key, __process_new_xattr, sctx);
3393 return ret;
3396 static int process_deleted_xattr(struct send_ctx *sctx)
3398 int ret;
3400 ret = iterate_dir_item(sctx->parent_root, sctx->right_path,
3401 sctx->cmp_key, __process_deleted_xattr, sctx);
3403 return ret;
3406 struct find_xattr_ctx {
3407 const char *name;
3408 int name_len;
3409 int found_idx;
3410 char *found_data;
3411 int found_data_len;
3414 static int __find_xattr(int num, struct btrfs_key *di_key,
3415 const char *name, int name_len,
3416 const char *data, int data_len,
3417 u8 type, void *vctx)
3419 struct find_xattr_ctx *ctx = vctx;
3421 if (name_len == ctx->name_len &&
3422 strncmp(name, ctx->name, name_len) == 0) {
3423 ctx->found_idx = num;
3424 ctx->found_data_len = data_len;
3425 ctx->found_data = kmemdup(data, data_len, GFP_NOFS);
3426 if (!ctx->found_data)
3427 return -ENOMEM;
3428 return 1;
3430 return 0;
3433 static int find_xattr(struct btrfs_root *root,
3434 struct btrfs_path *path,
3435 struct btrfs_key *key,
3436 const char *name, int name_len,
3437 char **data, int *data_len)
3439 int ret;
3440 struct find_xattr_ctx ctx;
3442 ctx.name = name;
3443 ctx.name_len = name_len;
3444 ctx.found_idx = -1;
3445 ctx.found_data = NULL;
3446 ctx.found_data_len = 0;
3448 ret = iterate_dir_item(root, path, key, __find_xattr, &ctx);
3449 if (ret < 0)
3450 return ret;
3452 if (ctx.found_idx == -1)
3453 return -ENOENT;
3454 if (data) {
3455 *data = ctx.found_data;
3456 *data_len = ctx.found_data_len;
3457 } else {
3458 kfree(ctx.found_data);
3460 return ctx.found_idx;
3464 static int __process_changed_new_xattr(int num, struct btrfs_key *di_key,
3465 const char *name, int name_len,
3466 const char *data, int data_len,
3467 u8 type, void *ctx)
3469 int ret;
3470 struct send_ctx *sctx = ctx;
3471 char *found_data = NULL;
3472 int found_data_len = 0;
3474 ret = find_xattr(sctx->parent_root, sctx->right_path,
3475 sctx->cmp_key, name, name_len, &found_data,
3476 &found_data_len);
3477 if (ret == -ENOENT) {
3478 ret = __process_new_xattr(num, di_key, name, name_len, data,
3479 data_len, type, ctx);
3480 } else if (ret >= 0) {
3481 if (data_len != found_data_len ||
3482 memcmp(data, found_data, data_len)) {
3483 ret = __process_new_xattr(num, di_key, name, name_len,
3484 data, data_len, type, ctx);
3485 } else {
3486 ret = 0;
3490 kfree(found_data);
3491 return ret;
3494 static int __process_changed_deleted_xattr(int num, struct btrfs_key *di_key,
3495 const char *name, int name_len,
3496 const char *data, int data_len,
3497 u8 type, void *ctx)
3499 int ret;
3500 struct send_ctx *sctx = ctx;
3502 ret = find_xattr(sctx->send_root, sctx->left_path, sctx->cmp_key,
3503 name, name_len, NULL, NULL);
3504 if (ret == -ENOENT)
3505 ret = __process_deleted_xattr(num, di_key, name, name_len, data,
3506 data_len, type, ctx);
3507 else if (ret >= 0)
3508 ret = 0;
3510 return ret;
3513 static int process_changed_xattr(struct send_ctx *sctx)
3515 int ret = 0;
3517 ret = iterate_dir_item(sctx->send_root, sctx->left_path,
3518 sctx->cmp_key, __process_changed_new_xattr, sctx);
3519 if (ret < 0)
3520 goto out;
3521 ret = iterate_dir_item(sctx->parent_root, sctx->right_path,
3522 sctx->cmp_key, __process_changed_deleted_xattr, sctx);
3524 out:
3525 return ret;
3528 static int process_all_new_xattrs(struct send_ctx *sctx)
3530 int ret;
3531 struct btrfs_root *root;
3532 struct btrfs_path *path;
3533 struct btrfs_key key;
3534 struct btrfs_key found_key;
3535 struct extent_buffer *eb;
3536 int slot;
3538 path = alloc_path_for_send();
3539 if (!path)
3540 return -ENOMEM;
3542 root = sctx->send_root;
3544 key.objectid = sctx->cmp_key->objectid;
3545 key.type = BTRFS_XATTR_ITEM_KEY;
3546 key.offset = 0;
3547 while (1) {
3548 ret = btrfs_search_slot_for_read(root, &key, path, 1, 0);
3549 if (ret < 0)
3550 goto out;
3551 if (ret) {
3552 ret = 0;
3553 goto out;
3556 eb = path->nodes[0];
3557 slot = path->slots[0];
3558 btrfs_item_key_to_cpu(eb, &found_key, slot);
3560 if (found_key.objectid != key.objectid ||
3561 found_key.type != key.type) {
3562 ret = 0;
3563 goto out;
3566 ret = iterate_dir_item(root, path, &found_key,
3567 __process_new_xattr, sctx);
3568 if (ret < 0)
3569 goto out;
3571 btrfs_release_path(path);
3572 key.offset = found_key.offset + 1;
3575 out:
3576 btrfs_free_path(path);
3577 return ret;
3581 * Read some bytes from the current inode/file and send a write command to
3582 * user space.
3584 static int send_write(struct send_ctx *sctx, u64 offset, u32 len)
3586 int ret = 0;
3587 struct fs_path *p;
3588 loff_t pos = offset;
3589 int num_read = 0;
3590 mm_segment_t old_fs;
3592 p = fs_path_alloc();
3593 if (!p)
3594 return -ENOMEM;
3597 * vfs normally only accepts user space buffers for security reasons.
3598 * we only read from the file and also only provide the read_buf buffer
3599 * to vfs. As this buffer does not come from a user space call, it's
3600 * ok to temporary allow kernel space buffers.
3602 old_fs = get_fs();
3603 set_fs(KERNEL_DS);
3605 verbose_printk("btrfs: send_write offset=%llu, len=%d\n", offset, len);
3607 ret = open_cur_inode_file(sctx);
3608 if (ret < 0)
3609 goto out;
3611 ret = vfs_read(sctx->cur_inode_filp, sctx->read_buf, len, &pos);
3612 if (ret < 0)
3613 goto out;
3614 num_read = ret;
3615 if (!num_read)
3616 goto out;
3618 ret = begin_cmd(sctx, BTRFS_SEND_C_WRITE);
3619 if (ret < 0)
3620 goto out;
3622 ret = get_cur_path(sctx, sctx->cur_ino, sctx->cur_inode_gen, p);
3623 if (ret < 0)
3624 goto out;
3626 TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, p);
3627 TLV_PUT_U64(sctx, BTRFS_SEND_A_FILE_OFFSET, offset);
3628 TLV_PUT(sctx, BTRFS_SEND_A_DATA, sctx->read_buf, num_read);
3630 ret = send_cmd(sctx);
3632 tlv_put_failure:
3633 out:
3634 fs_path_free(p);
3635 set_fs(old_fs);
3636 if (ret < 0)
3637 return ret;
3638 return num_read;
3642 * Send a clone command to user space.
3644 static int send_clone(struct send_ctx *sctx,
3645 u64 offset, u32 len,
3646 struct clone_root *clone_root)
3648 int ret = 0;
3649 struct fs_path *p;
3650 u64 gen;
3652 verbose_printk("btrfs: send_clone offset=%llu, len=%d, clone_root=%llu, "
3653 "clone_inode=%llu, clone_offset=%llu\n", offset, len,
3654 clone_root->root->objectid, clone_root->ino,
3655 clone_root->offset);
3657 p = fs_path_alloc();
3658 if (!p)
3659 return -ENOMEM;
3661 ret = begin_cmd(sctx, BTRFS_SEND_C_CLONE);
3662 if (ret < 0)
3663 goto out;
3665 ret = get_cur_path(sctx, sctx->cur_ino, sctx->cur_inode_gen, p);
3666 if (ret < 0)
3667 goto out;
3669 TLV_PUT_U64(sctx, BTRFS_SEND_A_FILE_OFFSET, offset);
3670 TLV_PUT_U64(sctx, BTRFS_SEND_A_CLONE_LEN, len);
3671 TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, p);
3673 if (clone_root->root == sctx->send_root) {
3674 ret = get_inode_info(sctx->send_root, clone_root->ino, NULL,
3675 &gen, NULL, NULL, NULL, NULL);
3676 if (ret < 0)
3677 goto out;
3678 ret = get_cur_path(sctx, clone_root->ino, gen, p);
3679 } else {
3680 ret = get_inode_path(clone_root->root, clone_root->ino, p);
3682 if (ret < 0)
3683 goto out;
3685 TLV_PUT_UUID(sctx, BTRFS_SEND_A_CLONE_UUID,
3686 clone_root->root->root_item.uuid);
3687 TLV_PUT_U64(sctx, BTRFS_SEND_A_CLONE_CTRANSID,
3688 clone_root->root->root_item.ctransid);
3689 TLV_PUT_PATH(sctx, BTRFS_SEND_A_CLONE_PATH, p);
3690 TLV_PUT_U64(sctx, BTRFS_SEND_A_CLONE_OFFSET,
3691 clone_root->offset);
3693 ret = send_cmd(sctx);
3695 tlv_put_failure:
3696 out:
3697 fs_path_free(p);
3698 return ret;
3702 * Send an update extent command to user space.
3704 static int send_update_extent(struct send_ctx *sctx,
3705 u64 offset, u32 len)
3707 int ret = 0;
3708 struct fs_path *p;
3710 p = fs_path_alloc();
3711 if (!p)
3712 return -ENOMEM;
3714 ret = begin_cmd(sctx, BTRFS_SEND_C_UPDATE_EXTENT);
3715 if (ret < 0)
3716 goto out;
3718 ret = get_cur_path(sctx, sctx->cur_ino, sctx->cur_inode_gen, p);
3719 if (ret < 0)
3720 goto out;
3722 TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, p);
3723 TLV_PUT_U64(sctx, BTRFS_SEND_A_FILE_OFFSET, offset);
3724 TLV_PUT_U64(sctx, BTRFS_SEND_A_SIZE, len);
3726 ret = send_cmd(sctx);
3728 tlv_put_failure:
3729 out:
3730 fs_path_free(p);
3731 return ret;
3734 static int send_write_or_clone(struct send_ctx *sctx,
3735 struct btrfs_path *path,
3736 struct btrfs_key *key,
3737 struct clone_root *clone_root)
3739 int ret = 0;
3740 struct btrfs_file_extent_item *ei;
3741 u64 offset = key->offset;
3742 u64 pos = 0;
3743 u64 len;
3744 u32 l;
3745 u8 type;
3747 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3748 struct btrfs_file_extent_item);
3749 type = btrfs_file_extent_type(path->nodes[0], ei);
3750 if (type == BTRFS_FILE_EXTENT_INLINE) {
3751 len = btrfs_file_extent_inline_len(path->nodes[0], ei);
3753 * it is possible the inline item won't cover the whole page,
3754 * but there may be items after this page. Make
3755 * sure to send the whole thing
3757 len = PAGE_CACHE_ALIGN(len);
3758 } else {
3759 len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3762 if (offset + len > sctx->cur_inode_size)
3763 len = sctx->cur_inode_size - offset;
3764 if (len == 0) {
3765 ret = 0;
3766 goto out;
3769 if (clone_root) {
3770 ret = send_clone(sctx, offset, len, clone_root);
3771 } else if (sctx->flags & BTRFS_SEND_FLAG_NO_FILE_DATA) {
3772 ret = send_update_extent(sctx, offset, len);
3773 } else {
3774 while (pos < len) {
3775 l = len - pos;
3776 if (l > BTRFS_SEND_READ_SIZE)
3777 l = BTRFS_SEND_READ_SIZE;
3778 ret = send_write(sctx, pos + offset, l);
3779 if (ret < 0)
3780 goto out;
3781 if (!ret)
3782 break;
3783 pos += ret;
3785 ret = 0;
3787 out:
3788 return ret;
3791 static int is_extent_unchanged(struct send_ctx *sctx,
3792 struct btrfs_path *left_path,
3793 struct btrfs_key *ekey)
3795 int ret = 0;
3796 struct btrfs_key key;
3797 struct btrfs_path *path = NULL;
3798 struct extent_buffer *eb;
3799 int slot;
3800 struct btrfs_key found_key;
3801 struct btrfs_file_extent_item *ei;
3802 u64 left_disknr;
3803 u64 right_disknr;
3804 u64 left_offset;
3805 u64 right_offset;
3806 u64 left_offset_fixed;
3807 u64 left_len;
3808 u64 right_len;
3809 u64 left_gen;
3810 u64 right_gen;
3811 u8 left_type;
3812 u8 right_type;
3814 path = alloc_path_for_send();
3815 if (!path)
3816 return -ENOMEM;
3818 eb = left_path->nodes[0];
3819 slot = left_path->slots[0];
3820 ei = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
3821 left_type = btrfs_file_extent_type(eb, ei);
3823 if (left_type != BTRFS_FILE_EXTENT_REG) {
3824 ret = 0;
3825 goto out;
3827 left_disknr = btrfs_file_extent_disk_bytenr(eb, ei);
3828 left_len = btrfs_file_extent_num_bytes(eb, ei);
3829 left_offset = btrfs_file_extent_offset(eb, ei);
3830 left_gen = btrfs_file_extent_generation(eb, ei);
3833 * Following comments will refer to these graphics. L is the left
3834 * extents which we are checking at the moment. 1-8 are the right
3835 * extents that we iterate.
3837 * |-----L-----|
3838 * |-1-|-2a-|-3-|-4-|-5-|-6-|
3840 * |-----L-----|
3841 * |--1--|-2b-|...(same as above)
3843 * Alternative situation. Happens on files where extents got split.
3844 * |-----L-----|
3845 * |-----------7-----------|-6-|
3847 * Alternative situation. Happens on files which got larger.
3848 * |-----L-----|
3849 * |-8-|
3850 * Nothing follows after 8.
3853 key.objectid = ekey->objectid;
3854 key.type = BTRFS_EXTENT_DATA_KEY;
3855 key.offset = ekey->offset;
3856 ret = btrfs_search_slot_for_read(sctx->parent_root, &key, path, 0, 0);
3857 if (ret < 0)
3858 goto out;
3859 if (ret) {
3860 ret = 0;
3861 goto out;
3865 * Handle special case where the right side has no extents at all.
3867 eb = path->nodes[0];
3868 slot = path->slots[0];
3869 btrfs_item_key_to_cpu(eb, &found_key, slot);
3870 if (found_key.objectid != key.objectid ||
3871 found_key.type != key.type) {
3872 ret = 0;
3873 goto out;
3877 * We're now on 2a, 2b or 7.
3879 key = found_key;
3880 while (key.offset < ekey->offset + left_len) {
3881 ei = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
3882 right_type = btrfs_file_extent_type(eb, ei);
3883 right_disknr = btrfs_file_extent_disk_bytenr(eb, ei);
3884 right_len = btrfs_file_extent_num_bytes(eb, ei);
3885 right_offset = btrfs_file_extent_offset(eb, ei);
3886 right_gen = btrfs_file_extent_generation(eb, ei);
3888 if (right_type != BTRFS_FILE_EXTENT_REG) {
3889 ret = 0;
3890 goto out;
3894 * Are we at extent 8? If yes, we know the extent is changed.
3895 * This may only happen on the first iteration.
3897 if (found_key.offset + right_len <= ekey->offset) {
3898 ret = 0;
3899 goto out;
3902 left_offset_fixed = left_offset;
3903 if (key.offset < ekey->offset) {
3904 /* Fix the right offset for 2a and 7. */
3905 right_offset += ekey->offset - key.offset;
3906 } else {
3907 /* Fix the left offset for all behind 2a and 2b */
3908 left_offset_fixed += key.offset - ekey->offset;
3912 * Check if we have the same extent.
3914 if (left_disknr != right_disknr ||
3915 left_offset_fixed != right_offset ||
3916 left_gen != right_gen) {
3917 ret = 0;
3918 goto out;
3922 * Go to the next extent.
3924 ret = btrfs_next_item(sctx->parent_root, path);
3925 if (ret < 0)
3926 goto out;
3927 if (!ret) {
3928 eb = path->nodes[0];
3929 slot = path->slots[0];
3930 btrfs_item_key_to_cpu(eb, &found_key, slot);
3932 if (ret || found_key.objectid != key.objectid ||
3933 found_key.type != key.type) {
3934 key.offset += right_len;
3935 break;
3937 if (found_key.offset != key.offset + right_len) {
3938 ret = 0;
3939 goto out;
3941 key = found_key;
3945 * We're now behind the left extent (treat as unchanged) or at the end
3946 * of the right side (treat as changed).
3948 if (key.offset >= ekey->offset + left_len)
3949 ret = 1;
3950 else
3951 ret = 0;
3954 out:
3955 btrfs_free_path(path);
3956 return ret;
3959 static int process_extent(struct send_ctx *sctx,
3960 struct btrfs_path *path,
3961 struct btrfs_key *key)
3963 int ret = 0;
3964 struct clone_root *found_clone = NULL;
3966 if (S_ISLNK(sctx->cur_inode_mode))
3967 return 0;
3969 if (sctx->parent_root && !sctx->cur_inode_new) {
3970 ret = is_extent_unchanged(sctx, path, key);
3971 if (ret < 0)
3972 goto out;
3973 if (ret) {
3974 ret = 0;
3975 goto out;
3979 ret = find_extent_clone(sctx, path, key->objectid, key->offset,
3980 sctx->cur_inode_size, &found_clone);
3981 if (ret != -ENOENT && ret < 0)
3982 goto out;
3984 ret = send_write_or_clone(sctx, path, key, found_clone);
3986 out:
3987 return ret;
3990 static int process_all_extents(struct send_ctx *sctx)
3992 int ret;
3993 struct btrfs_root *root;
3994 struct btrfs_path *path;
3995 struct btrfs_key key;
3996 struct btrfs_key found_key;
3997 struct extent_buffer *eb;
3998 int slot;
4000 root = sctx->send_root;
4001 path = alloc_path_for_send();
4002 if (!path)
4003 return -ENOMEM;
4005 key.objectid = sctx->cmp_key->objectid;
4006 key.type = BTRFS_EXTENT_DATA_KEY;
4007 key.offset = 0;
4008 while (1) {
4009 ret = btrfs_search_slot_for_read(root, &key, path, 1, 0);
4010 if (ret < 0)
4011 goto out;
4012 if (ret) {
4013 ret = 0;
4014 goto out;
4017 eb = path->nodes[0];
4018 slot = path->slots[0];
4019 btrfs_item_key_to_cpu(eb, &found_key, slot);
4021 if (found_key.objectid != key.objectid ||
4022 found_key.type != key.type) {
4023 ret = 0;
4024 goto out;
4027 ret = process_extent(sctx, path, &found_key);
4028 if (ret < 0)
4029 goto out;
4031 btrfs_release_path(path);
4032 key.offset = found_key.offset + 1;
4035 out:
4036 btrfs_free_path(path);
4037 return ret;
4040 static int process_recorded_refs_if_needed(struct send_ctx *sctx, int at_end)
4042 int ret = 0;
4044 if (sctx->cur_ino == 0)
4045 goto out;
4046 if (!at_end && sctx->cur_ino == sctx->cmp_key->objectid &&
4047 sctx->cmp_key->type <= BTRFS_INODE_EXTREF_KEY)
4048 goto out;
4049 if (list_empty(&sctx->new_refs) && list_empty(&sctx->deleted_refs))
4050 goto out;
4052 ret = process_recorded_refs(sctx);
4053 if (ret < 0)
4054 goto out;
4057 * We have processed the refs and thus need to advance send_progress.
4058 * Now, calls to get_cur_xxx will take the updated refs of the current
4059 * inode into account.
4061 sctx->send_progress = sctx->cur_ino + 1;
4063 out:
4064 return ret;
4067 static int finish_inode_if_needed(struct send_ctx *sctx, int at_end)
4069 int ret = 0;
4070 u64 left_mode;
4071 u64 left_uid;
4072 u64 left_gid;
4073 u64 right_mode;
4074 u64 right_uid;
4075 u64 right_gid;
4076 int need_chmod = 0;
4077 int need_chown = 0;
4079 ret = process_recorded_refs_if_needed(sctx, at_end);
4080 if (ret < 0)
4081 goto out;
4083 if (sctx->cur_ino == 0 || sctx->cur_inode_deleted)
4084 goto out;
4085 if (!at_end && sctx->cmp_key->objectid == sctx->cur_ino)
4086 goto out;
4088 ret = get_inode_info(sctx->send_root, sctx->cur_ino, NULL, NULL,
4089 &left_mode, &left_uid, &left_gid, NULL);
4090 if (ret < 0)
4091 goto out;
4093 if (!sctx->parent_root || sctx->cur_inode_new) {
4094 need_chown = 1;
4095 if (!S_ISLNK(sctx->cur_inode_mode))
4096 need_chmod = 1;
4097 } else {
4098 ret = get_inode_info(sctx->parent_root, sctx->cur_ino,
4099 NULL, NULL, &right_mode, &right_uid,
4100 &right_gid, NULL);
4101 if (ret < 0)
4102 goto out;
4104 if (left_uid != right_uid || left_gid != right_gid)
4105 need_chown = 1;
4106 if (!S_ISLNK(sctx->cur_inode_mode) && left_mode != right_mode)
4107 need_chmod = 1;
4110 if (S_ISREG(sctx->cur_inode_mode)) {
4111 ret = send_truncate(sctx, sctx->cur_ino, sctx->cur_inode_gen,
4112 sctx->cur_inode_size);
4113 if (ret < 0)
4114 goto out;
4117 if (need_chown) {
4118 ret = send_chown(sctx, sctx->cur_ino, sctx->cur_inode_gen,
4119 left_uid, left_gid);
4120 if (ret < 0)
4121 goto out;
4123 if (need_chmod) {
4124 ret = send_chmod(sctx, sctx->cur_ino, sctx->cur_inode_gen,
4125 left_mode);
4126 if (ret < 0)
4127 goto out;
4131 * Need to send that every time, no matter if it actually changed
4132 * between the two trees as we have done changes to the inode before.
4134 ret = send_utimes(sctx, sctx->cur_ino, sctx->cur_inode_gen);
4135 if (ret < 0)
4136 goto out;
4138 out:
4139 return ret;
4142 static int changed_inode(struct send_ctx *sctx,
4143 enum btrfs_compare_tree_result result)
4145 int ret = 0;
4146 struct btrfs_key *key = sctx->cmp_key;
4147 struct btrfs_inode_item *left_ii = NULL;
4148 struct btrfs_inode_item *right_ii = NULL;
4149 u64 left_gen = 0;
4150 u64 right_gen = 0;
4152 ret = close_cur_inode_file(sctx);
4153 if (ret < 0)
4154 goto out;
4156 sctx->cur_ino = key->objectid;
4157 sctx->cur_inode_new_gen = 0;
4160 * Set send_progress to current inode. This will tell all get_cur_xxx
4161 * functions that the current inode's refs are not updated yet. Later,
4162 * when process_recorded_refs is finished, it is set to cur_ino + 1.
4164 sctx->send_progress = sctx->cur_ino;
4166 if (result == BTRFS_COMPARE_TREE_NEW ||
4167 result == BTRFS_COMPARE_TREE_CHANGED) {
4168 left_ii = btrfs_item_ptr(sctx->left_path->nodes[0],
4169 sctx->left_path->slots[0],
4170 struct btrfs_inode_item);
4171 left_gen = btrfs_inode_generation(sctx->left_path->nodes[0],
4172 left_ii);
4173 } else {
4174 right_ii = btrfs_item_ptr(sctx->right_path->nodes[0],
4175 sctx->right_path->slots[0],
4176 struct btrfs_inode_item);
4177 right_gen = btrfs_inode_generation(sctx->right_path->nodes[0],
4178 right_ii);
4180 if (result == BTRFS_COMPARE_TREE_CHANGED) {
4181 right_ii = btrfs_item_ptr(sctx->right_path->nodes[0],
4182 sctx->right_path->slots[0],
4183 struct btrfs_inode_item);
4185 right_gen = btrfs_inode_generation(sctx->right_path->nodes[0],
4186 right_ii);
4189 * The cur_ino = root dir case is special here. We can't treat
4190 * the inode as deleted+reused because it would generate a
4191 * stream that tries to delete/mkdir the root dir.
4193 if (left_gen != right_gen &&
4194 sctx->cur_ino != BTRFS_FIRST_FREE_OBJECTID)
4195 sctx->cur_inode_new_gen = 1;
4198 if (result == BTRFS_COMPARE_TREE_NEW) {
4199 sctx->cur_inode_gen = left_gen;
4200 sctx->cur_inode_new = 1;
4201 sctx->cur_inode_deleted = 0;
4202 sctx->cur_inode_size = btrfs_inode_size(
4203 sctx->left_path->nodes[0], left_ii);
4204 sctx->cur_inode_mode = btrfs_inode_mode(
4205 sctx->left_path->nodes[0], left_ii);
4206 if (sctx->cur_ino != BTRFS_FIRST_FREE_OBJECTID)
4207 ret = send_create_inode_if_needed(sctx);
4208 } else if (result == BTRFS_COMPARE_TREE_DELETED) {
4209 sctx->cur_inode_gen = right_gen;
4210 sctx->cur_inode_new = 0;
4211 sctx->cur_inode_deleted = 1;
4212 sctx->cur_inode_size = btrfs_inode_size(
4213 sctx->right_path->nodes[0], right_ii);
4214 sctx->cur_inode_mode = btrfs_inode_mode(
4215 sctx->right_path->nodes[0], right_ii);
4216 } else if (result == BTRFS_COMPARE_TREE_CHANGED) {
4218 * We need to do some special handling in case the inode was
4219 * reported as changed with a changed generation number. This
4220 * means that the original inode was deleted and new inode
4221 * reused the same inum. So we have to treat the old inode as
4222 * deleted and the new one as new.
4224 if (sctx->cur_inode_new_gen) {
4226 * First, process the inode as if it was deleted.
4228 sctx->cur_inode_gen = right_gen;
4229 sctx->cur_inode_new = 0;
4230 sctx->cur_inode_deleted = 1;
4231 sctx->cur_inode_size = btrfs_inode_size(
4232 sctx->right_path->nodes[0], right_ii);
4233 sctx->cur_inode_mode = btrfs_inode_mode(
4234 sctx->right_path->nodes[0], right_ii);
4235 ret = process_all_refs(sctx,
4236 BTRFS_COMPARE_TREE_DELETED);
4237 if (ret < 0)
4238 goto out;
4241 * Now process the inode as if it was new.
4243 sctx->cur_inode_gen = left_gen;
4244 sctx->cur_inode_new = 1;
4245 sctx->cur_inode_deleted = 0;
4246 sctx->cur_inode_size = btrfs_inode_size(
4247 sctx->left_path->nodes[0], left_ii);
4248 sctx->cur_inode_mode = btrfs_inode_mode(
4249 sctx->left_path->nodes[0], left_ii);
4250 ret = send_create_inode_if_needed(sctx);
4251 if (ret < 0)
4252 goto out;
4254 ret = process_all_refs(sctx, BTRFS_COMPARE_TREE_NEW);
4255 if (ret < 0)
4256 goto out;
4258 * Advance send_progress now as we did not get into
4259 * process_recorded_refs_if_needed in the new_gen case.
4261 sctx->send_progress = sctx->cur_ino + 1;
4264 * Now process all extents and xattrs of the inode as if
4265 * they were all new.
4267 ret = process_all_extents(sctx);
4268 if (ret < 0)
4269 goto out;
4270 ret = process_all_new_xattrs(sctx);
4271 if (ret < 0)
4272 goto out;
4273 } else {
4274 sctx->cur_inode_gen = left_gen;
4275 sctx->cur_inode_new = 0;
4276 sctx->cur_inode_new_gen = 0;
4277 sctx->cur_inode_deleted = 0;
4278 sctx->cur_inode_size = btrfs_inode_size(
4279 sctx->left_path->nodes[0], left_ii);
4280 sctx->cur_inode_mode = btrfs_inode_mode(
4281 sctx->left_path->nodes[0], left_ii);
4285 out:
4286 return ret;
4290 * We have to process new refs before deleted refs, but compare_trees gives us
4291 * the new and deleted refs mixed. To fix this, we record the new/deleted refs
4292 * first and later process them in process_recorded_refs.
4293 * For the cur_inode_new_gen case, we skip recording completely because
4294 * changed_inode did already initiate processing of refs. The reason for this is
4295 * that in this case, compare_tree actually compares the refs of 2 different
4296 * inodes. To fix this, process_all_refs is used in changed_inode to handle all
4297 * refs of the right tree as deleted and all refs of the left tree as new.
4299 static int changed_ref(struct send_ctx *sctx,
4300 enum btrfs_compare_tree_result result)
4302 int ret = 0;
4304 BUG_ON(sctx->cur_ino != sctx->cmp_key->objectid);
4306 if (!sctx->cur_inode_new_gen &&
4307 sctx->cur_ino != BTRFS_FIRST_FREE_OBJECTID) {
4308 if (result == BTRFS_COMPARE_TREE_NEW)
4309 ret = record_new_ref(sctx);
4310 else if (result == BTRFS_COMPARE_TREE_DELETED)
4311 ret = record_deleted_ref(sctx);
4312 else if (result == BTRFS_COMPARE_TREE_CHANGED)
4313 ret = record_changed_ref(sctx);
4316 return ret;
4320 * Process new/deleted/changed xattrs. We skip processing in the
4321 * cur_inode_new_gen case because changed_inode did already initiate processing
4322 * of xattrs. The reason is the same as in changed_ref
4324 static int changed_xattr(struct send_ctx *sctx,
4325 enum btrfs_compare_tree_result result)
4327 int ret = 0;
4329 BUG_ON(sctx->cur_ino != sctx->cmp_key->objectid);
4331 if (!sctx->cur_inode_new_gen && !sctx->cur_inode_deleted) {
4332 if (result == BTRFS_COMPARE_TREE_NEW)
4333 ret = process_new_xattr(sctx);
4334 else if (result == BTRFS_COMPARE_TREE_DELETED)
4335 ret = process_deleted_xattr(sctx);
4336 else if (result == BTRFS_COMPARE_TREE_CHANGED)
4337 ret = process_changed_xattr(sctx);
4340 return ret;
4344 * Process new/deleted/changed extents. We skip processing in the
4345 * cur_inode_new_gen case because changed_inode did already initiate processing
4346 * of extents. The reason is the same as in changed_ref
4348 static int changed_extent(struct send_ctx *sctx,
4349 enum btrfs_compare_tree_result result)
4351 int ret = 0;
4353 BUG_ON(sctx->cur_ino != sctx->cmp_key->objectid);
4355 if (!sctx->cur_inode_new_gen && !sctx->cur_inode_deleted) {
4356 if (result != BTRFS_COMPARE_TREE_DELETED)
4357 ret = process_extent(sctx, sctx->left_path,
4358 sctx->cmp_key);
4361 return ret;
4365 * Updates compare related fields in sctx and simply forwards to the actual
4366 * changed_xxx functions.
4368 static int changed_cb(struct btrfs_root *left_root,
4369 struct btrfs_root *right_root,
4370 struct btrfs_path *left_path,
4371 struct btrfs_path *right_path,
4372 struct btrfs_key *key,
4373 enum btrfs_compare_tree_result result,
4374 void *ctx)
4376 int ret = 0;
4377 struct send_ctx *sctx = ctx;
4379 sctx->left_path = left_path;
4380 sctx->right_path = right_path;
4381 sctx->cmp_key = key;
4383 ret = finish_inode_if_needed(sctx, 0);
4384 if (ret < 0)
4385 goto out;
4387 /* Ignore non-FS objects */
4388 if (key->objectid == BTRFS_FREE_INO_OBJECTID ||
4389 key->objectid == BTRFS_FREE_SPACE_OBJECTID)
4390 goto out;
4392 if (key->type == BTRFS_INODE_ITEM_KEY)
4393 ret = changed_inode(sctx, result);
4394 else if (key->type == BTRFS_INODE_REF_KEY ||
4395 key->type == BTRFS_INODE_EXTREF_KEY)
4396 ret = changed_ref(sctx, result);
4397 else if (key->type == BTRFS_XATTR_ITEM_KEY)
4398 ret = changed_xattr(sctx, result);
4399 else if (key->type == BTRFS_EXTENT_DATA_KEY)
4400 ret = changed_extent(sctx, result);
4402 out:
4403 return ret;
4406 static int full_send_tree(struct send_ctx *sctx)
4408 int ret;
4409 struct btrfs_trans_handle *trans = NULL;
4410 struct btrfs_root *send_root = sctx->send_root;
4411 struct btrfs_key key;
4412 struct btrfs_key found_key;
4413 struct btrfs_path *path;
4414 struct extent_buffer *eb;
4415 int slot;
4416 u64 start_ctransid;
4417 u64 ctransid;
4419 path = alloc_path_for_send();
4420 if (!path)
4421 return -ENOMEM;
4423 spin_lock(&send_root->root_item_lock);
4424 start_ctransid = btrfs_root_ctransid(&send_root->root_item);
4425 spin_unlock(&send_root->root_item_lock);
4427 key.objectid = BTRFS_FIRST_FREE_OBJECTID;
4428 key.type = BTRFS_INODE_ITEM_KEY;
4429 key.offset = 0;
4431 join_trans:
4433 * We need to make sure the transaction does not get committed
4434 * while we do anything on commit roots. Join a transaction to prevent
4435 * this.
4437 trans = btrfs_join_transaction(send_root);
4438 if (IS_ERR(trans)) {
4439 ret = PTR_ERR(trans);
4440 trans = NULL;
4441 goto out;
4445 * Make sure the tree has not changed after re-joining. We detect this
4446 * by comparing start_ctransid and ctransid. They should always match.
4448 spin_lock(&send_root->root_item_lock);
4449 ctransid = btrfs_root_ctransid(&send_root->root_item);
4450 spin_unlock(&send_root->root_item_lock);
4452 if (ctransid != start_ctransid) {
4453 WARN(1, KERN_WARNING "btrfs: the root that you're trying to "
4454 "send was modified in between. This is "
4455 "probably a bug.\n");
4456 ret = -EIO;
4457 goto out;
4460 ret = btrfs_search_slot_for_read(send_root, &key, path, 1, 0);
4461 if (ret < 0)
4462 goto out;
4463 if (ret)
4464 goto out_finish;
4466 while (1) {
4468 * When someone want to commit while we iterate, end the
4469 * joined transaction and rejoin.
4471 if (btrfs_should_end_transaction(trans, send_root)) {
4472 ret = btrfs_end_transaction(trans, send_root);
4473 trans = NULL;
4474 if (ret < 0)
4475 goto out;
4476 btrfs_release_path(path);
4477 goto join_trans;
4480 eb = path->nodes[0];
4481 slot = path->slots[0];
4482 btrfs_item_key_to_cpu(eb, &found_key, slot);
4484 ret = changed_cb(send_root, NULL, path, NULL,
4485 &found_key, BTRFS_COMPARE_TREE_NEW, sctx);
4486 if (ret < 0)
4487 goto out;
4489 key.objectid = found_key.objectid;
4490 key.type = found_key.type;
4491 key.offset = found_key.offset + 1;
4493 ret = btrfs_next_item(send_root, path);
4494 if (ret < 0)
4495 goto out;
4496 if (ret) {
4497 ret = 0;
4498 break;
4502 out_finish:
4503 ret = finish_inode_if_needed(sctx, 1);
4505 out:
4506 btrfs_free_path(path);
4507 if (trans) {
4508 if (!ret)
4509 ret = btrfs_end_transaction(trans, send_root);
4510 else
4511 btrfs_end_transaction(trans, send_root);
4513 return ret;
4516 static int send_subvol(struct send_ctx *sctx)
4518 int ret;
4520 if (!(sctx->flags & BTRFS_SEND_FLAG_OMIT_STREAM_HEADER)) {
4521 ret = send_header(sctx);
4522 if (ret < 0)
4523 goto out;
4526 ret = send_subvol_begin(sctx);
4527 if (ret < 0)
4528 goto out;
4530 if (sctx->parent_root) {
4531 ret = btrfs_compare_trees(sctx->send_root, sctx->parent_root,
4532 changed_cb, sctx);
4533 if (ret < 0)
4534 goto out;
4535 ret = finish_inode_if_needed(sctx, 1);
4536 if (ret < 0)
4537 goto out;
4538 } else {
4539 ret = full_send_tree(sctx);
4540 if (ret < 0)
4541 goto out;
4544 out:
4545 if (!ret)
4546 ret = close_cur_inode_file(sctx);
4547 else
4548 close_cur_inode_file(sctx);
4550 free_recorded_refs(sctx);
4551 return ret;
4554 long btrfs_ioctl_send(struct file *mnt_file, void __user *arg_)
4556 int ret = 0;
4557 struct btrfs_root *send_root;
4558 struct btrfs_root *clone_root;
4559 struct btrfs_fs_info *fs_info;
4560 struct btrfs_ioctl_send_args *arg = NULL;
4561 struct btrfs_key key;
4562 struct send_ctx *sctx = NULL;
4563 u32 i;
4564 u64 *clone_sources_tmp = NULL;
4566 if (!capable(CAP_SYS_ADMIN))
4567 return -EPERM;
4569 send_root = BTRFS_I(file_inode(mnt_file))->root;
4570 fs_info = send_root->fs_info;
4573 * This is done when we lookup the root, it should already be complete
4574 * by the time we get here.
4576 WARN_ON(send_root->orphan_cleanup_state != ORPHAN_CLEANUP_DONE);
4579 * If we just created this root we need to make sure that the orphan
4580 * cleanup has been done and committed since we search the commit root,
4581 * so check its commit root transid with our otransid and if they match
4582 * commit the transaction to make sure everything is updated.
4584 down_read(&send_root->fs_info->extent_commit_sem);
4585 if (btrfs_header_generation(send_root->commit_root) ==
4586 btrfs_root_otransid(&send_root->root_item)) {
4587 struct btrfs_trans_handle *trans;
4589 up_read(&send_root->fs_info->extent_commit_sem);
4591 trans = btrfs_attach_transaction_barrier(send_root);
4592 if (IS_ERR(trans)) {
4593 if (PTR_ERR(trans) != -ENOENT) {
4594 ret = PTR_ERR(trans);
4595 goto out;
4597 /* ENOENT means theres no transaction */
4598 } else {
4599 ret = btrfs_commit_transaction(trans, send_root);
4600 if (ret)
4601 goto out;
4603 } else {
4604 up_read(&send_root->fs_info->extent_commit_sem);
4607 arg = memdup_user(arg_, sizeof(*arg));
4608 if (IS_ERR(arg)) {
4609 ret = PTR_ERR(arg);
4610 arg = NULL;
4611 goto out;
4614 if (!access_ok(VERIFY_READ, arg->clone_sources,
4615 sizeof(*arg->clone_sources *
4616 arg->clone_sources_count))) {
4617 ret = -EFAULT;
4618 goto out;
4621 if (arg->flags & ~BTRFS_SEND_FLAG_MASK) {
4622 ret = -EINVAL;
4623 goto out;
4626 sctx = kzalloc(sizeof(struct send_ctx), GFP_NOFS);
4627 if (!sctx) {
4628 ret = -ENOMEM;
4629 goto out;
4632 INIT_LIST_HEAD(&sctx->new_refs);
4633 INIT_LIST_HEAD(&sctx->deleted_refs);
4634 INIT_RADIX_TREE(&sctx->name_cache, GFP_NOFS);
4635 INIT_LIST_HEAD(&sctx->name_cache_list);
4637 sctx->flags = arg->flags;
4639 sctx->send_filp = fget(arg->send_fd);
4640 if (!sctx->send_filp) {
4641 ret = -EBADF;
4642 goto out;
4645 sctx->mnt = mnt_file->f_path.mnt;
4647 sctx->send_root = send_root;
4648 sctx->clone_roots_cnt = arg->clone_sources_count;
4650 sctx->send_max_size = BTRFS_SEND_BUF_SIZE;
4651 sctx->send_buf = vmalloc(sctx->send_max_size);
4652 if (!sctx->send_buf) {
4653 ret = -ENOMEM;
4654 goto out;
4657 sctx->read_buf = vmalloc(BTRFS_SEND_READ_SIZE);
4658 if (!sctx->read_buf) {
4659 ret = -ENOMEM;
4660 goto out;
4663 sctx->clone_roots = vzalloc(sizeof(struct clone_root) *
4664 (arg->clone_sources_count + 1));
4665 if (!sctx->clone_roots) {
4666 ret = -ENOMEM;
4667 goto out;
4670 if (arg->clone_sources_count) {
4671 clone_sources_tmp = vmalloc(arg->clone_sources_count *
4672 sizeof(*arg->clone_sources));
4673 if (!clone_sources_tmp) {
4674 ret = -ENOMEM;
4675 goto out;
4678 ret = copy_from_user(clone_sources_tmp, arg->clone_sources,
4679 arg->clone_sources_count *
4680 sizeof(*arg->clone_sources));
4681 if (ret) {
4682 ret = -EFAULT;
4683 goto out;
4686 for (i = 0; i < arg->clone_sources_count; i++) {
4687 key.objectid = clone_sources_tmp[i];
4688 key.type = BTRFS_ROOT_ITEM_KEY;
4689 key.offset = (u64)-1;
4690 clone_root = btrfs_read_fs_root_no_name(fs_info, &key);
4691 if (IS_ERR(clone_root)) {
4692 ret = PTR_ERR(clone_root);
4693 goto out;
4695 sctx->clone_roots[i].root = clone_root;
4697 vfree(clone_sources_tmp);
4698 clone_sources_tmp = NULL;
4701 if (arg->parent_root) {
4702 key.objectid = arg->parent_root;
4703 key.type = BTRFS_ROOT_ITEM_KEY;
4704 key.offset = (u64)-1;
4705 sctx->parent_root = btrfs_read_fs_root_no_name(fs_info, &key);
4706 if (IS_ERR(sctx->parent_root)) {
4707 ret = PTR_ERR(sctx->parent_root);
4708 goto out;
4713 * Clones from send_root are allowed, but only if the clone source
4714 * is behind the current send position. This is checked while searching
4715 * for possible clone sources.
4717 sctx->clone_roots[sctx->clone_roots_cnt++].root = sctx->send_root;
4719 /* We do a bsearch later */
4720 sort(sctx->clone_roots, sctx->clone_roots_cnt,
4721 sizeof(*sctx->clone_roots), __clone_root_cmp_sort,
4722 NULL);
4724 ret = send_subvol(sctx);
4725 if (ret < 0)
4726 goto out;
4728 if (!(sctx->flags & BTRFS_SEND_FLAG_OMIT_END_CMD)) {
4729 ret = begin_cmd(sctx, BTRFS_SEND_C_END);
4730 if (ret < 0)
4731 goto out;
4732 ret = send_cmd(sctx);
4733 if (ret < 0)
4734 goto out;
4737 out:
4738 kfree(arg);
4739 vfree(clone_sources_tmp);
4741 if (sctx) {
4742 if (sctx->send_filp)
4743 fput(sctx->send_filp);
4745 vfree(sctx->clone_roots);
4746 vfree(sctx->send_buf);
4747 vfree(sctx->read_buf);
4749 name_cache_free(sctx);
4751 kfree(sctx);
4754 return ret;