9pfs: fix information leak in xattr read
[qemu.git] / hw / 9pfs / 9p.c
1 /*
2 * Virtio 9p backend
3 *
4 * Copyright IBM, Corp. 2010
5 *
6 * Authors:
7 * Anthony Liguori <aliguori@us.ibm.com>
8 *
9 * This work is licensed under the terms of the GNU GPL, version 2. See
10 * the COPYING file in the top-level directory.
11 *
12 */
13
14 #include "qemu/osdep.h"
15 #include <glib/gprintf.h>
16 #include "hw/virtio/virtio.h"
17 #include "qapi/error.h"
18 #include "qemu/error-report.h"
19 #include "qemu/iov.h"
20 #include "qemu/sockets.h"
21 #include "virtio-9p.h"
22 #include "fsdev/qemu-fsdev.h"
23 #include "9p-xattr.h"
24 #include "coth.h"
25 #include "trace.h"
26 #include "migration/migration.h"
27
28 int open_fd_hw;
29 int total_open_fd;
30 static int open_fd_rc;
31
32 enum {
33 Oread = 0x00,
34 Owrite = 0x01,
35 Ordwr = 0x02,
36 Oexec = 0x03,
37 Oexcl = 0x04,
38 Otrunc = 0x10,
39 Orexec = 0x20,
40 Orclose = 0x40,
41 Oappend = 0x80,
42 };
43
44 ssize_t pdu_marshal(V9fsPDU *pdu, size_t offset, const char *fmt, ...)
45 {
46 ssize_t ret;
47 va_list ap;
48
49 va_start(ap, fmt);
50 ret = virtio_pdu_vmarshal(pdu, offset, fmt, ap);
51 va_end(ap);
52
53 return ret;
54 }
55
56 ssize_t pdu_unmarshal(V9fsPDU *pdu, size_t offset, const char *fmt, ...)
57 {
58 ssize_t ret;
59 va_list ap;
60
61 va_start(ap, fmt);
62 ret = virtio_pdu_vunmarshal(pdu, offset, fmt, ap);
63 va_end(ap);
64
65 return ret;
66 }
67
68 static void pdu_push_and_notify(V9fsPDU *pdu)
69 {
70 virtio_9p_push_and_notify(pdu);
71 }
72
73 static int omode_to_uflags(int8_t mode)
74 {
75 int ret = 0;
76
77 switch (mode & 3) {
78 case Oread:
79 ret = O_RDONLY;
80 break;
81 case Ordwr:
82 ret = O_RDWR;
83 break;
84 case Owrite:
85 ret = O_WRONLY;
86 break;
87 case Oexec:
88 ret = O_RDONLY;
89 break;
90 }
91
92 if (mode & Otrunc) {
93 ret |= O_TRUNC;
94 }
95
96 if (mode & Oappend) {
97 ret |= O_APPEND;
98 }
99
100 if (mode & Oexcl) {
101 ret |= O_EXCL;
102 }
103
104 return ret;
105 }
106
107 struct dotl_openflag_map {
108 int dotl_flag;
109 int open_flag;
110 };
111
112 static int dotl_to_open_flags(int flags)
113 {
114 int i;
115 /*
116 * We have same bits for P9_DOTL_READONLY, P9_DOTL_WRONLY
117 * and P9_DOTL_NOACCESS
118 */
119 int oflags = flags & O_ACCMODE;
120
121 struct dotl_openflag_map dotl_oflag_map[] = {
122 { P9_DOTL_CREATE, O_CREAT },
123 { P9_DOTL_EXCL, O_EXCL },
124 { P9_DOTL_NOCTTY , O_NOCTTY },
125 { P9_DOTL_TRUNC, O_TRUNC },
126 { P9_DOTL_APPEND, O_APPEND },
127 { P9_DOTL_NONBLOCK, O_NONBLOCK } ,
128 { P9_DOTL_DSYNC, O_DSYNC },
129 { P9_DOTL_FASYNC, FASYNC },
130 { P9_DOTL_DIRECT, O_DIRECT },
131 { P9_DOTL_LARGEFILE, O_LARGEFILE },
132 { P9_DOTL_DIRECTORY, O_DIRECTORY },
133 { P9_DOTL_NOFOLLOW, O_NOFOLLOW },
134 { P9_DOTL_NOATIME, O_NOATIME },
135 { P9_DOTL_SYNC, O_SYNC },
136 };
137
138 for (i = 0; i < ARRAY_SIZE(dotl_oflag_map); i++) {
139 if (flags & dotl_oflag_map[i].dotl_flag) {
140 oflags |= dotl_oflag_map[i].open_flag;
141 }
142 }
143
144 return oflags;
145 }
146
147 void cred_init(FsCred *credp)
148 {
149 credp->fc_uid = -1;
150 credp->fc_gid = -1;
151 credp->fc_mode = -1;
152 credp->fc_rdev = -1;
153 }
154
155 static int get_dotl_openflags(V9fsState *s, int oflags)
156 {
157 int flags;
158 /*
159 * Filter the client open flags
160 */
161 flags = dotl_to_open_flags(oflags);
162 flags &= ~(O_NOCTTY | O_ASYNC | O_CREAT);
163 /*
164 * Ignore direct disk access hint until the server supports it.
165 */
166 flags &= ~O_DIRECT;
167 return flags;
168 }
169
170 void v9fs_path_init(V9fsPath *path)
171 {
172 path->data = NULL;
173 path->size = 0;
174 }
175
176 void v9fs_path_free(V9fsPath *path)
177 {
178 g_free(path->data);
179 path->data = NULL;
180 path->size = 0;
181 }
182
183
184 void GCC_FMT_ATTR(2, 3)
185 v9fs_path_sprintf(V9fsPath *path, const char *fmt, ...)
186 {
187 va_list ap;
188
189 v9fs_path_free(path);
190
191 va_start(ap, fmt);
192 /* Bump the size for including terminating NULL */
193 path->size = g_vasprintf(&path->data, fmt, ap) + 1;
194 va_end(ap);
195 }
196
197 void v9fs_path_copy(V9fsPath *lhs, V9fsPath *rhs)
198 {
199 v9fs_path_free(lhs);
200 lhs->data = g_malloc(rhs->size);
201 memcpy(lhs->data, rhs->data, rhs->size);
202 lhs->size = rhs->size;
203 }
204
205 int v9fs_name_to_path(V9fsState *s, V9fsPath *dirpath,
206 const char *name, V9fsPath *path)
207 {
208 int err;
209 err = s->ops->name_to_path(&s->ctx, dirpath, name, path);
210 if (err < 0) {
211 err = -errno;
212 }
213 return err;
214 }
215
216 /*
217 * Return TRUE if s1 is an ancestor of s2.
218 *
219 * E.g. "a/b" is an ancestor of "a/b/c" but not of "a/bc/d".
220 * As a special case, We treat s1 as ancestor of s2 if they are same!
221 */
222 static int v9fs_path_is_ancestor(V9fsPath *s1, V9fsPath *s2)
223 {
224 if (!strncmp(s1->data, s2->data, s1->size - 1)) {
225 if (s2->data[s1->size - 1] == '\0' || s2->data[s1->size - 1] == '/') {
226 return 1;
227 }
228 }
229 return 0;
230 }
231
232 static size_t v9fs_string_size(V9fsString *str)
233 {
234 return str->size;
235 }
236
237 /*
238 * returns 0 if fid got re-opened, 1 if not, < 0 on error */
239 static int coroutine_fn v9fs_reopen_fid(V9fsPDU *pdu, V9fsFidState *f)
240 {
241 int err = 1;
242 if (f->fid_type == P9_FID_FILE) {
243 if (f->fs.fd == -1) {
244 do {
245 err = v9fs_co_open(pdu, f, f->open_flags);
246 } while (err == -EINTR && !pdu->cancelled);
247 }
248 } else if (f->fid_type == P9_FID_DIR) {
249 if (f->fs.dir.stream == NULL) {
250 do {
251 err = v9fs_co_opendir(pdu, f);
252 } while (err == -EINTR && !pdu->cancelled);
253 }
254 }
255 return err;
256 }
257
258 static V9fsFidState *coroutine_fn get_fid(V9fsPDU *pdu, int32_t fid)
259 {
260 int err;
261 V9fsFidState *f;
262 V9fsState *s = pdu->s;
263
264 for (f = s->fid_list; f; f = f->next) {
265 BUG_ON(f->clunked);
266 if (f->fid == fid) {
267 /*
268 * Update the fid ref upfront so that
269 * we don't get reclaimed when we yield
270 * in open later.
271 */
272 f->ref++;
273 /*
274 * check whether we need to reopen the
275 * file. We might have closed the fd
276 * while trying to free up some file
277 * descriptors.
278 */
279 err = v9fs_reopen_fid(pdu, f);
280 if (err < 0) {
281 f->ref--;
282 return NULL;
283 }
284 /*
285 * Mark the fid as referenced so that the LRU
286 * reclaim won't close the file descriptor
287 */
288 f->flags |= FID_REFERENCED;
289 return f;
290 }
291 }
292 return NULL;
293 }
294
295 static V9fsFidState *alloc_fid(V9fsState *s, int32_t fid)
296 {
297 V9fsFidState *f;
298
299 for (f = s->fid_list; f; f = f->next) {
300 /* If fid is already there return NULL */
301 BUG_ON(f->clunked);
302 if (f->fid == fid) {
303 return NULL;
304 }
305 }
306 f = g_malloc0(sizeof(V9fsFidState));
307 f->fid = fid;
308 f->fid_type = P9_FID_NONE;
309 f->ref = 1;
310 /*
311 * Mark the fid as referenced so that the LRU
312 * reclaim won't close the file descriptor
313 */
314 f->flags |= FID_REFERENCED;
315 f->next = s->fid_list;
316 s->fid_list = f;
317
318 v9fs_readdir_init(&f->fs.dir);
319 v9fs_readdir_init(&f->fs_reclaim.dir);
320
321 return f;
322 }
323
324 static int coroutine_fn v9fs_xattr_fid_clunk(V9fsPDU *pdu, V9fsFidState *fidp)
325 {
326 int retval = 0;
327
328 if (fidp->fs.xattr.copied_len == -1) {
329 /* getxattr/listxattr fid */
330 goto free_value;
331 }
332 /*
333 * if this is fid for setxattr. clunk should
334 * result in setxattr localcall
335 */
336 if (fidp->fs.xattr.len != fidp->fs.xattr.copied_len) {
337 /* clunk after partial write */
338 retval = -EINVAL;
339 goto free_out;
340 }
341 if (fidp->fs.xattr.len) {
342 retval = v9fs_co_lsetxattr(pdu, &fidp->path, &fidp->fs.xattr.name,
343 fidp->fs.xattr.value,
344 fidp->fs.xattr.len,
345 fidp->fs.xattr.flags);
346 } else {
347 retval = v9fs_co_lremovexattr(pdu, &fidp->path, &fidp->fs.xattr.name);
348 }
349 free_out:
350 v9fs_string_free(&fidp->fs.xattr.name);
351 free_value:
352 g_free(fidp->fs.xattr.value);
353 return retval;
354 }
355
356 static int coroutine_fn free_fid(V9fsPDU *pdu, V9fsFidState *fidp)
357 {
358 int retval = 0;
359
360 if (fidp->fid_type == P9_FID_FILE) {
361 /* If we reclaimed the fd no need to close */
362 if (fidp->fs.fd != -1) {
363 retval = v9fs_co_close(pdu, &fidp->fs);
364 }
365 } else if (fidp->fid_type == P9_FID_DIR) {
366 if (fidp->fs.dir.stream != NULL) {
367 retval = v9fs_co_closedir(pdu, &fidp->fs);
368 }
369 } else if (fidp->fid_type == P9_FID_XATTR) {
370 retval = v9fs_xattr_fid_clunk(pdu, fidp);
371 }
372 v9fs_path_free(&fidp->path);
373 g_free(fidp);
374 return retval;
375 }
376
377 static int coroutine_fn put_fid(V9fsPDU *pdu, V9fsFidState *fidp)
378 {
379 BUG_ON(!fidp->ref);
380 fidp->ref--;
381 /*
382 * Don't free the fid if it is in reclaim list
383 */
384 if (!fidp->ref && fidp->clunked) {
385 if (fidp->fid == pdu->s->root_fid) {
386 /*
387 * if the clunked fid is root fid then we
388 * have unmounted the fs on the client side.
389 * delete the migration blocker. Ideally, this
390 * should be hooked to transport close notification
391 */
392 if (pdu->s->migration_blocker) {
393 migrate_del_blocker(pdu->s->migration_blocker);
394 error_free(pdu->s->migration_blocker);
395 pdu->s->migration_blocker = NULL;
396 }
397 }
398 return free_fid(pdu, fidp);
399 }
400 return 0;
401 }
402
403 static V9fsFidState *clunk_fid(V9fsState *s, int32_t fid)
404 {
405 V9fsFidState **fidpp, *fidp;
406
407 for (fidpp = &s->fid_list; *fidpp; fidpp = &(*fidpp)->next) {
408 if ((*fidpp)->fid == fid) {
409 break;
410 }
411 }
412 if (*fidpp == NULL) {
413 return NULL;
414 }
415 fidp = *fidpp;
416 *fidpp = fidp->next;
417 fidp->clunked = 1;
418 return fidp;
419 }
420
421 void coroutine_fn v9fs_reclaim_fd(V9fsPDU *pdu)
422 {
423 int reclaim_count = 0;
424 V9fsState *s = pdu->s;
425 V9fsFidState *f, *reclaim_list = NULL;
426
427 for (f = s->fid_list; f; f = f->next) {
428 /*
429 * Unlink fids cannot be reclaimed. Check
430 * for them and skip them. Also skip fids
431 * currently being operated on.
432 */
433 if (f->ref || f->flags & FID_NON_RECLAIMABLE) {
434 continue;
435 }
436 /*
437 * if it is a recently referenced fid
438 * we leave the fid untouched and clear the
439 * reference bit. We come back to it later
440 * in the next iteration. (a simple LRU without
441 * moving list elements around)
442 */
443 if (f->flags & FID_REFERENCED) {
444 f->flags &= ~FID_REFERENCED;
445 continue;
446 }
447 /*
448 * Add fids to reclaim list.
449 */
450 if (f->fid_type == P9_FID_FILE) {
451 if (f->fs.fd != -1) {
452 /*
453 * Up the reference count so that
454 * a clunk request won't free this fid
455 */
456 f->ref++;
457 f->rclm_lst = reclaim_list;
458 reclaim_list = f;
459 f->fs_reclaim.fd = f->fs.fd;
460 f->fs.fd = -1;
461 reclaim_count++;
462 }
463 } else if (f->fid_type == P9_FID_DIR) {
464 if (f->fs.dir.stream != NULL) {
465 /*
466 * Up the reference count so that
467 * a clunk request won't free this fid
468 */
469 f->ref++;
470 f->rclm_lst = reclaim_list;
471 reclaim_list = f;
472 f->fs_reclaim.dir.stream = f->fs.dir.stream;
473 f->fs.dir.stream = NULL;
474 reclaim_count++;
475 }
476 }
477 if (reclaim_count >= open_fd_rc) {
478 break;
479 }
480 }
481 /*
482 * Now close the fid in reclaim list. Free them if they
483 * are already clunked.
484 */
485 while (reclaim_list) {
486 f = reclaim_list;
487 reclaim_list = f->rclm_lst;
488 if (f->fid_type == P9_FID_FILE) {
489 v9fs_co_close(pdu, &f->fs_reclaim);
490 } else if (f->fid_type == P9_FID_DIR) {
491 v9fs_co_closedir(pdu, &f->fs_reclaim);
492 }
493 f->rclm_lst = NULL;
494 /*
495 * Now drop the fid reference, free it
496 * if clunked.
497 */
498 put_fid(pdu, f);
499 }
500 }
501
502 static int coroutine_fn v9fs_mark_fids_unreclaim(V9fsPDU *pdu, V9fsPath *path)
503 {
504 int err;
505 V9fsState *s = pdu->s;
506 V9fsFidState *fidp, head_fid;
507
508 head_fid.next = s->fid_list;
509 for (fidp = s->fid_list; fidp; fidp = fidp->next) {
510 if (fidp->path.size != path->size) {
511 continue;
512 }
513 if (!memcmp(fidp->path.data, path->data, path->size)) {
514 /* Mark the fid non reclaimable. */
515 fidp->flags |= FID_NON_RECLAIMABLE;
516
517 /* reopen the file/dir if already closed */
518 err = v9fs_reopen_fid(pdu, fidp);
519 if (err < 0) {
520 return -1;
521 }
522 /*
523 * Go back to head of fid list because
524 * the list could have got updated when
525 * switched to the worker thread
526 */
527 if (err == 0) {
528 fidp = &head_fid;
529 }
530 }
531 }
532 return 0;
533 }
534
535 static void coroutine_fn virtfs_reset(V9fsPDU *pdu)
536 {
537 V9fsState *s = pdu->s;
538 V9fsFidState *fidp = NULL;
539
540 /* Free all fids */
541 while (s->fid_list) {
542 fidp = s->fid_list;
543 s->fid_list = fidp->next;
544
545 if (fidp->ref) {
546 fidp->clunked = 1;
547 } else {
548 free_fid(pdu, fidp);
549 }
550 }
551 if (fidp) {
552 /* One or more unclunked fids found... */
553 error_report("9pfs:%s: One or more uncluncked fids "
554 "found during reset", __func__);
555 }
556 }
557
558 #define P9_QID_TYPE_DIR 0x80
559 #define P9_QID_TYPE_SYMLINK 0x02
560
561 #define P9_STAT_MODE_DIR 0x80000000
562 #define P9_STAT_MODE_APPEND 0x40000000
563 #define P9_STAT_MODE_EXCL 0x20000000
564 #define P9_STAT_MODE_MOUNT 0x10000000
565 #define P9_STAT_MODE_AUTH 0x08000000
566 #define P9_STAT_MODE_TMP 0x04000000
567 #define P9_STAT_MODE_SYMLINK 0x02000000
568 #define P9_STAT_MODE_LINK 0x01000000
569 #define P9_STAT_MODE_DEVICE 0x00800000
570 #define P9_STAT_MODE_NAMED_PIPE 0x00200000
571 #define P9_STAT_MODE_SOCKET 0x00100000
572 #define P9_STAT_MODE_SETUID 0x00080000
573 #define P9_STAT_MODE_SETGID 0x00040000
574 #define P9_STAT_MODE_SETVTX 0x00010000
575
576 #define P9_STAT_MODE_TYPE_BITS (P9_STAT_MODE_DIR | \
577 P9_STAT_MODE_SYMLINK | \
578 P9_STAT_MODE_LINK | \
579 P9_STAT_MODE_DEVICE | \
580 P9_STAT_MODE_NAMED_PIPE | \
581 P9_STAT_MODE_SOCKET)
582
583 /* This is the algorithm from ufs in spfs */
584 static void stat_to_qid(const struct stat *stbuf, V9fsQID *qidp)
585 {
586 size_t size;
587
588 memset(&qidp->path, 0, sizeof(qidp->path));
589 size = MIN(sizeof(stbuf->st_ino), sizeof(qidp->path));
590 memcpy(&qidp->path, &stbuf->st_ino, size);
591 qidp->version = stbuf->st_mtime ^ (stbuf->st_size << 8);
592 qidp->type = 0;
593 if (S_ISDIR(stbuf->st_mode)) {
594 qidp->type |= P9_QID_TYPE_DIR;
595 }
596 if (S_ISLNK(stbuf->st_mode)) {
597 qidp->type |= P9_QID_TYPE_SYMLINK;
598 }
599 }
600
601 static int coroutine_fn fid_to_qid(V9fsPDU *pdu, V9fsFidState *fidp,
602 V9fsQID *qidp)
603 {
604 struct stat stbuf;
605 int err;
606
607 err = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
608 if (err < 0) {
609 return err;
610 }
611 stat_to_qid(&stbuf, qidp);
612 return 0;
613 }
614
615 V9fsPDU *pdu_alloc(V9fsState *s)
616 {
617 V9fsPDU *pdu = NULL;
618
619 if (!QLIST_EMPTY(&s->free_list)) {
620 pdu = QLIST_FIRST(&s->free_list);
621 QLIST_REMOVE(pdu, next);
622 QLIST_INSERT_HEAD(&s->active_list, pdu, next);
623 }
624 return pdu;
625 }
626
627 void pdu_free(V9fsPDU *pdu)
628 {
629 V9fsState *s = pdu->s;
630
631 g_assert(!pdu->cancelled);
632 QLIST_REMOVE(pdu, next);
633 QLIST_INSERT_HEAD(&s->free_list, pdu, next);
634 }
635
636 /*
637 * We don't do error checking for pdu_marshal/unmarshal here
638 * because we always expect to have enough space to encode
639 * error details
640 */
641 static void coroutine_fn pdu_complete(V9fsPDU *pdu, ssize_t len)
642 {
643 int8_t id = pdu->id + 1; /* Response */
644 V9fsState *s = pdu->s;
645
646 if (len < 0) {
647 int err = -len;
648 len = 7;
649
650 if (s->proto_version != V9FS_PROTO_2000L) {
651 V9fsString str;
652
653 str.data = strerror(err);
654 str.size = strlen(str.data);
655
656 len += pdu_marshal(pdu, len, "s", &str);
657 id = P9_RERROR;
658 }
659
660 len += pdu_marshal(pdu, len, "d", err);
661
662 if (s->proto_version == V9FS_PROTO_2000L) {
663 id = P9_RLERROR;
664 }
665 trace_v9fs_rerror(pdu->tag, pdu->id, err); /* Trace ERROR */
666 }
667
668 /* fill out the header */
669 pdu_marshal(pdu, 0, "dbw", (int32_t)len, id, pdu->tag);
670
671 /* keep these in sync */
672 pdu->size = len;
673 pdu->id = id;
674
675 pdu_push_and_notify(pdu);
676
677 /* Now wakeup anybody waiting in flush for this request */
678 if (!qemu_co_queue_next(&pdu->complete)) {
679 pdu_free(pdu);
680 }
681 }
682
683 static mode_t v9mode_to_mode(uint32_t mode, V9fsString *extension)
684 {
685 mode_t ret;
686
687 ret = mode & 0777;
688 if (mode & P9_STAT_MODE_DIR) {
689 ret |= S_IFDIR;
690 }
691
692 if (mode & P9_STAT_MODE_SYMLINK) {
693 ret |= S_IFLNK;
694 }
695 if (mode & P9_STAT_MODE_SOCKET) {
696 ret |= S_IFSOCK;
697 }
698 if (mode & P9_STAT_MODE_NAMED_PIPE) {
699 ret |= S_IFIFO;
700 }
701 if (mode & P9_STAT_MODE_DEVICE) {
702 if (extension->size && extension->data[0] == 'c') {
703 ret |= S_IFCHR;
704 } else {
705 ret |= S_IFBLK;
706 }
707 }
708
709 if (!(ret&~0777)) {
710 ret |= S_IFREG;
711 }
712
713 if (mode & P9_STAT_MODE_SETUID) {
714 ret |= S_ISUID;
715 }
716 if (mode & P9_STAT_MODE_SETGID) {
717 ret |= S_ISGID;
718 }
719 if (mode & P9_STAT_MODE_SETVTX) {
720 ret |= S_ISVTX;
721 }
722
723 return ret;
724 }
725
726 static int donttouch_stat(V9fsStat *stat)
727 {
728 if (stat->type == -1 &&
729 stat->dev == -1 &&
730 stat->qid.type == -1 &&
731 stat->qid.version == -1 &&
732 stat->qid.path == -1 &&
733 stat->mode == -1 &&
734 stat->atime == -1 &&
735 stat->mtime == -1 &&
736 stat->length == -1 &&
737 !stat->name.size &&
738 !stat->uid.size &&
739 !stat->gid.size &&
740 !stat->muid.size &&
741 stat->n_uid == -1 &&
742 stat->n_gid == -1 &&
743 stat->n_muid == -1) {
744 return 1;
745 }
746
747 return 0;
748 }
749
750 static void v9fs_stat_init(V9fsStat *stat)
751 {
752 v9fs_string_init(&stat->name);
753 v9fs_string_init(&stat->uid);
754 v9fs_string_init(&stat->gid);
755 v9fs_string_init(&stat->muid);
756 v9fs_string_init(&stat->extension);
757 }
758
759 static void v9fs_stat_free(V9fsStat *stat)
760 {
761 v9fs_string_free(&stat->name);
762 v9fs_string_free(&stat->uid);
763 v9fs_string_free(&stat->gid);
764 v9fs_string_free(&stat->muid);
765 v9fs_string_free(&stat->extension);
766 }
767
768 static uint32_t stat_to_v9mode(const struct stat *stbuf)
769 {
770 uint32_t mode;
771
772 mode = stbuf->st_mode & 0777;
773 if (S_ISDIR(stbuf->st_mode)) {
774 mode |= P9_STAT_MODE_DIR;
775 }
776
777 if (S_ISLNK(stbuf->st_mode)) {
778 mode |= P9_STAT_MODE_SYMLINK;
779 }
780
781 if (S_ISSOCK(stbuf->st_mode)) {
782 mode |= P9_STAT_MODE_SOCKET;
783 }
784
785 if (S_ISFIFO(stbuf->st_mode)) {
786 mode |= P9_STAT_MODE_NAMED_PIPE;
787 }
788
789 if (S_ISBLK(stbuf->st_mode) || S_ISCHR(stbuf->st_mode)) {
790 mode |= P9_STAT_MODE_DEVICE;
791 }
792
793 if (stbuf->st_mode & S_ISUID) {
794 mode |= P9_STAT_MODE_SETUID;
795 }
796
797 if (stbuf->st_mode & S_ISGID) {
798 mode |= P9_STAT_MODE_SETGID;
799 }
800
801 if (stbuf->st_mode & S_ISVTX) {
802 mode |= P9_STAT_MODE_SETVTX;
803 }
804
805 return mode;
806 }
807
808 static int coroutine_fn stat_to_v9stat(V9fsPDU *pdu, V9fsPath *name,
809 const struct stat *stbuf,
810 V9fsStat *v9stat)
811 {
812 int err;
813 const char *str;
814
815 memset(v9stat, 0, sizeof(*v9stat));
816
817 stat_to_qid(stbuf, &v9stat->qid);
818 v9stat->mode = stat_to_v9mode(stbuf);
819 v9stat->atime = stbuf->st_atime;
820 v9stat->mtime = stbuf->st_mtime;
821 v9stat->length = stbuf->st_size;
822
823 v9fs_string_free(&v9stat->uid);
824 v9fs_string_free(&v9stat->gid);
825 v9fs_string_free(&v9stat->muid);
826
827 v9stat->n_uid = stbuf->st_uid;
828 v9stat->n_gid = stbuf->st_gid;
829 v9stat->n_muid = 0;
830
831 v9fs_string_free(&v9stat->extension);
832
833 if (v9stat->mode & P9_STAT_MODE_SYMLINK) {
834 err = v9fs_co_readlink(pdu, name, &v9stat->extension);
835 if (err < 0) {
836 return err;
837 }
838 } else if (v9stat->mode & P9_STAT_MODE_DEVICE) {
839 v9fs_string_sprintf(&v9stat->extension, "%c %u %u",
840 S_ISCHR(stbuf->st_mode) ? 'c' : 'b',
841 major(stbuf->st_rdev), minor(stbuf->st_rdev));
842 } else if (S_ISDIR(stbuf->st_mode) || S_ISREG(stbuf->st_mode)) {
843 v9fs_string_sprintf(&v9stat->extension, "%s %lu",
844 "HARDLINKCOUNT", (unsigned long)stbuf->st_nlink);
845 }
846
847 str = strrchr(name->data, '/');
848 if (str) {
849 str += 1;
850 } else {
851 str = name->data;
852 }
853
854 v9fs_string_sprintf(&v9stat->name, "%s", str);
855
856 v9stat->size = 61 +
857 v9fs_string_size(&v9stat->name) +
858 v9fs_string_size(&v9stat->uid) +
859 v9fs_string_size(&v9stat->gid) +
860 v9fs_string_size(&v9stat->muid) +
861 v9fs_string_size(&v9stat->extension);
862 return 0;
863 }
864
865 #define P9_STATS_MODE 0x00000001ULL
866 #define P9_STATS_NLINK 0x00000002ULL
867 #define P9_STATS_UID 0x00000004ULL
868 #define P9_STATS_GID 0x00000008ULL
869 #define P9_STATS_RDEV 0x00000010ULL
870 #define P9_STATS_ATIME 0x00000020ULL
871 #define P9_STATS_MTIME 0x00000040ULL
872 #define P9_STATS_CTIME 0x00000080ULL
873 #define P9_STATS_INO 0x00000100ULL
874 #define P9_STATS_SIZE 0x00000200ULL
875 #define P9_STATS_BLOCKS 0x00000400ULL
876
877 #define P9_STATS_BTIME 0x00000800ULL
878 #define P9_STATS_GEN 0x00001000ULL
879 #define P9_STATS_DATA_VERSION 0x00002000ULL
880
881 #define P9_STATS_BASIC 0x000007ffULL /* Mask for fields up to BLOCKS */
882 #define P9_STATS_ALL 0x00003fffULL /* Mask for All fields above */
883
884
885 static void stat_to_v9stat_dotl(V9fsState *s, const struct stat *stbuf,
886 V9fsStatDotl *v9lstat)
887 {
888 memset(v9lstat, 0, sizeof(*v9lstat));
889
890 v9lstat->st_mode = stbuf->st_mode;
891 v9lstat->st_nlink = stbuf->st_nlink;
892 v9lstat->st_uid = stbuf->st_uid;
893 v9lstat->st_gid = stbuf->st_gid;
894 v9lstat->st_rdev = stbuf->st_rdev;
895 v9lstat->st_size = stbuf->st_size;
896 v9lstat->st_blksize = stbuf->st_blksize;
897 v9lstat->st_blocks = stbuf->st_blocks;
898 v9lstat->st_atime_sec = stbuf->st_atime;
899 v9lstat->st_atime_nsec = stbuf->st_atim.tv_nsec;
900 v9lstat->st_mtime_sec = stbuf->st_mtime;
901 v9lstat->st_mtime_nsec = stbuf->st_mtim.tv_nsec;
902 v9lstat->st_ctime_sec = stbuf->st_ctime;
903 v9lstat->st_ctime_nsec = stbuf->st_ctim.tv_nsec;
904 /* Currently we only support BASIC fields in stat */
905 v9lstat->st_result_mask = P9_STATS_BASIC;
906
907 stat_to_qid(stbuf, &v9lstat->qid);
908 }
909
910 static void print_sg(struct iovec *sg, int cnt)
911 {
912 int i;
913
914 printf("sg[%d]: {", cnt);
915 for (i = 0; i < cnt; i++) {
916 if (i) {
917 printf(", ");
918 }
919 printf("(%p, %zd)", sg[i].iov_base, sg[i].iov_len);
920 }
921 printf("}\n");
922 }
923
924 /* Will call this only for path name based fid */
925 static void v9fs_fix_path(V9fsPath *dst, V9fsPath *src, int len)
926 {
927 V9fsPath str;
928 v9fs_path_init(&str);
929 v9fs_path_copy(&str, dst);
930 v9fs_path_sprintf(dst, "%s%s", src->data, str.data + len);
931 v9fs_path_free(&str);
932 }
933
934 static inline bool is_ro_export(FsContext *ctx)
935 {
936 return ctx->export_flags & V9FS_RDONLY;
937 }
938
939 static void coroutine_fn v9fs_version(void *opaque)
940 {
941 ssize_t err;
942 V9fsPDU *pdu = opaque;
943 V9fsState *s = pdu->s;
944 V9fsString version;
945 size_t offset = 7;
946
947 v9fs_string_init(&version);
948 err = pdu_unmarshal(pdu, offset, "ds", &s->msize, &version);
949 if (err < 0) {
950 offset = err;
951 goto out;
952 }
953 trace_v9fs_version(pdu->tag, pdu->id, s->msize, version.data);
954
955 virtfs_reset(pdu);
956
957 if (!strcmp(version.data, "9P2000.u")) {
958 s->proto_version = V9FS_PROTO_2000U;
959 } else if (!strcmp(version.data, "9P2000.L")) {
960 s->proto_version = V9FS_PROTO_2000L;
961 } else {
962 v9fs_string_sprintf(&version, "unknown");
963 }
964
965 err = pdu_marshal(pdu, offset, "ds", s->msize, &version);
966 if (err < 0) {
967 offset = err;
968 goto out;
969 }
970 offset += err;
971 trace_v9fs_version_return(pdu->tag, pdu->id, s->msize, version.data);
972 out:
973 pdu_complete(pdu, offset);
974 v9fs_string_free(&version);
975 }
976
977 static void coroutine_fn v9fs_attach(void *opaque)
978 {
979 V9fsPDU *pdu = opaque;
980 V9fsState *s = pdu->s;
981 int32_t fid, afid, n_uname;
982 V9fsString uname, aname;
983 V9fsFidState *fidp;
984 size_t offset = 7;
985 V9fsQID qid;
986 ssize_t err;
987
988 v9fs_string_init(&uname);
989 v9fs_string_init(&aname);
990 err = pdu_unmarshal(pdu, offset, "ddssd", &fid,
991 &afid, &uname, &aname, &n_uname);
992 if (err < 0) {
993 goto out_nofid;
994 }
995 trace_v9fs_attach(pdu->tag, pdu->id, fid, afid, uname.data, aname.data);
996
997 fidp = alloc_fid(s, fid);
998 if (fidp == NULL) {
999 err = -EINVAL;
1000 goto out_nofid;
1001 }
1002 fidp->uid = n_uname;
1003 err = v9fs_co_name_to_path(pdu, NULL, "/", &fidp->path);
1004 if (err < 0) {
1005 err = -EINVAL;
1006 clunk_fid(s, fid);
1007 goto out;
1008 }
1009 err = fid_to_qid(pdu, fidp, &qid);
1010 if (err < 0) {
1011 err = -EINVAL;
1012 clunk_fid(s, fid);
1013 goto out;
1014 }
1015 err = pdu_marshal(pdu, offset, "Q", &qid);
1016 if (err < 0) {
1017 clunk_fid(s, fid);
1018 goto out;
1019 }
1020 err += offset;
1021 memcpy(&s->root_qid, &qid, sizeof(qid));
1022 trace_v9fs_attach_return(pdu->tag, pdu->id,
1023 qid.type, qid.version, qid.path);
1024 /*
1025 * disable migration if we haven't done already.
1026 * attach could get called multiple times for the same export.
1027 */
1028 if (!s->migration_blocker) {
1029 s->root_fid = fid;
1030 error_setg(&s->migration_blocker,
1031 "Migration is disabled when VirtFS export path '%s' is mounted in the guest using mount_tag '%s'",
1032 s->ctx.fs_root ? s->ctx.fs_root : "NULL", s->tag);
1033 migrate_add_blocker(s->migration_blocker);
1034 }
1035 out:
1036 put_fid(pdu, fidp);
1037 out_nofid:
1038 pdu_complete(pdu, err);
1039 v9fs_string_free(&uname);
1040 v9fs_string_free(&aname);
1041 }
1042
1043 static void coroutine_fn v9fs_stat(void *opaque)
1044 {
1045 int32_t fid;
1046 V9fsStat v9stat;
1047 ssize_t err = 0;
1048 size_t offset = 7;
1049 struct stat stbuf;
1050 V9fsFidState *fidp;
1051 V9fsPDU *pdu = opaque;
1052
1053 err = pdu_unmarshal(pdu, offset, "d", &fid);
1054 if (err < 0) {
1055 goto out_nofid;
1056 }
1057 trace_v9fs_stat(pdu->tag, pdu->id, fid);
1058
1059 fidp = get_fid(pdu, fid);
1060 if (fidp == NULL) {
1061 err = -ENOENT;
1062 goto out_nofid;
1063 }
1064 err = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
1065 if (err < 0) {
1066 goto out;
1067 }
1068 err = stat_to_v9stat(pdu, &fidp->path, &stbuf, &v9stat);
1069 if (err < 0) {
1070 goto out;
1071 }
1072 err = pdu_marshal(pdu, offset, "wS", 0, &v9stat);
1073 if (err < 0) {
1074 v9fs_stat_free(&v9stat);
1075 goto out;
1076 }
1077 trace_v9fs_stat_return(pdu->tag, pdu->id, v9stat.mode,
1078 v9stat.atime, v9stat.mtime, v9stat.length);
1079 err += offset;
1080 v9fs_stat_free(&v9stat);
1081 out:
1082 put_fid(pdu, fidp);
1083 out_nofid:
1084 pdu_complete(pdu, err);
1085 }
1086
1087 static void coroutine_fn v9fs_getattr(void *opaque)
1088 {
1089 int32_t fid;
1090 size_t offset = 7;
1091 ssize_t retval = 0;
1092 struct stat stbuf;
1093 V9fsFidState *fidp;
1094 uint64_t request_mask;
1095 V9fsStatDotl v9stat_dotl;
1096 V9fsPDU *pdu = opaque;
1097 V9fsState *s = pdu->s;
1098
1099 retval = pdu_unmarshal(pdu, offset, "dq", &fid, &request_mask);
1100 if (retval < 0) {
1101 goto out_nofid;
1102 }
1103 trace_v9fs_getattr(pdu->tag, pdu->id, fid, request_mask);
1104
1105 fidp = get_fid(pdu, fid);
1106 if (fidp == NULL) {
1107 retval = -ENOENT;
1108 goto out_nofid;
1109 }
1110 /*
1111 * Currently we only support BASIC fields in stat, so there is no
1112 * need to look at request_mask.
1113 */
1114 retval = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
1115 if (retval < 0) {
1116 goto out;
1117 }
1118 stat_to_v9stat_dotl(s, &stbuf, &v9stat_dotl);
1119
1120 /* fill st_gen if requested and supported by underlying fs */
1121 if (request_mask & P9_STATS_GEN) {
1122 retval = v9fs_co_st_gen(pdu, &fidp->path, stbuf.st_mode, &v9stat_dotl);
1123 switch (retval) {
1124 case 0:
1125 /* we have valid st_gen: update result mask */
1126 v9stat_dotl.st_result_mask |= P9_STATS_GEN;
1127 break;
1128 case -EINTR:
1129 /* request cancelled, e.g. by Tflush */
1130 goto out;
1131 default:
1132 /* failed to get st_gen: not fatal, ignore */
1133 break;
1134 }
1135 }
1136 retval = pdu_marshal(pdu, offset, "A", &v9stat_dotl);
1137 if (retval < 0) {
1138 goto out;
1139 }
1140 retval += offset;
1141 trace_v9fs_getattr_return(pdu->tag, pdu->id, v9stat_dotl.st_result_mask,
1142 v9stat_dotl.st_mode, v9stat_dotl.st_uid,
1143 v9stat_dotl.st_gid);
1144 out:
1145 put_fid(pdu, fidp);
1146 out_nofid:
1147 pdu_complete(pdu, retval);
1148 }
1149
1150 /* Attribute flags */
1151 #define P9_ATTR_MODE (1 << 0)
1152 #define P9_ATTR_UID (1 << 1)
1153 #define P9_ATTR_GID (1 << 2)
1154 #define P9_ATTR_SIZE (1 << 3)
1155 #define P9_ATTR_ATIME (1 << 4)
1156 #define P9_ATTR_MTIME (1 << 5)
1157 #define P9_ATTR_CTIME (1 << 6)
1158 #define P9_ATTR_ATIME_SET (1 << 7)
1159 #define P9_ATTR_MTIME_SET (1 << 8)
1160
1161 #define P9_ATTR_MASK 127
1162
1163 static void coroutine_fn v9fs_setattr(void *opaque)
1164 {
1165 int err = 0;
1166 int32_t fid;
1167 V9fsFidState *fidp;
1168 size_t offset = 7;
1169 V9fsIattr v9iattr;
1170 V9fsPDU *pdu = opaque;
1171
1172 err = pdu_unmarshal(pdu, offset, "dI", &fid, &v9iattr);
1173 if (err < 0) {
1174 goto out_nofid;
1175 }
1176
1177 fidp = get_fid(pdu, fid);
1178 if (fidp == NULL) {
1179 err = -EINVAL;
1180 goto out_nofid;
1181 }
1182 if (v9iattr.valid & P9_ATTR_MODE) {
1183 err = v9fs_co_chmod(pdu, &fidp->path, v9iattr.mode);
1184 if (err < 0) {
1185 goto out;
1186 }
1187 }
1188 if (v9iattr.valid & (P9_ATTR_ATIME | P9_ATTR_MTIME)) {
1189 struct timespec times[2];
1190 if (v9iattr.valid & P9_ATTR_ATIME) {
1191 if (v9iattr.valid & P9_ATTR_ATIME_SET) {
1192 times[0].tv_sec = v9iattr.atime_sec;
1193 times[0].tv_nsec = v9iattr.atime_nsec;
1194 } else {
1195 times[0].tv_nsec = UTIME_NOW;
1196 }
1197 } else {
1198 times[0].tv_nsec = UTIME_OMIT;
1199 }
1200 if (v9iattr.valid & P9_ATTR_MTIME) {
1201 if (v9iattr.valid & P9_ATTR_MTIME_SET) {
1202 times[1].tv_sec = v9iattr.mtime_sec;
1203 times[1].tv_nsec = v9iattr.mtime_nsec;
1204 } else {
1205 times[1].tv_nsec = UTIME_NOW;
1206 }
1207 } else {
1208 times[1].tv_nsec = UTIME_OMIT;
1209 }
1210 err = v9fs_co_utimensat(pdu, &fidp->path, times);
1211 if (err < 0) {
1212 goto out;
1213 }
1214 }
1215 /*
1216 * If the only valid entry in iattr is ctime we can call
1217 * chown(-1,-1) to update the ctime of the file
1218 */
1219 if ((v9iattr.valid & (P9_ATTR_UID | P9_ATTR_GID)) ||
1220 ((v9iattr.valid & P9_ATTR_CTIME)
1221 && !((v9iattr.valid & P9_ATTR_MASK) & ~P9_ATTR_CTIME))) {
1222 if (!(v9iattr.valid & P9_ATTR_UID)) {
1223 v9iattr.uid = -1;
1224 }
1225 if (!(v9iattr.valid & P9_ATTR_GID)) {
1226 v9iattr.gid = -1;
1227 }
1228 err = v9fs_co_chown(pdu, &fidp->path, v9iattr.uid,
1229 v9iattr.gid);
1230 if (err < 0) {
1231 goto out;
1232 }
1233 }
1234 if (v9iattr.valid & (P9_ATTR_SIZE)) {
1235 err = v9fs_co_truncate(pdu, &fidp->path, v9iattr.size);
1236 if (err < 0) {
1237 goto out;
1238 }
1239 }
1240 err = offset;
1241 out:
1242 put_fid(pdu, fidp);
1243 out_nofid:
1244 pdu_complete(pdu, err);
1245 }
1246
1247 static int v9fs_walk_marshal(V9fsPDU *pdu, uint16_t nwnames, V9fsQID *qids)
1248 {
1249 int i;
1250 ssize_t err;
1251 size_t offset = 7;
1252
1253 err = pdu_marshal(pdu, offset, "w", nwnames);
1254 if (err < 0) {
1255 return err;
1256 }
1257 offset += err;
1258 for (i = 0; i < nwnames; i++) {
1259 err = pdu_marshal(pdu, offset, "Q", &qids[i]);
1260 if (err < 0) {
1261 return err;
1262 }
1263 offset += err;
1264 }
1265 return offset;
1266 }
1267
1268 static bool name_is_illegal(const char *name)
1269 {
1270 return !*name || strchr(name, '/') != NULL;
1271 }
1272
1273 static bool not_same_qid(const V9fsQID *qid1, const V9fsQID *qid2)
1274 {
1275 return
1276 qid1->type != qid2->type ||
1277 qid1->version != qid2->version ||
1278 qid1->path != qid2->path;
1279 }
1280
1281 static void coroutine_fn v9fs_walk(void *opaque)
1282 {
1283 int name_idx;
1284 V9fsQID *qids = NULL;
1285 int i, err = 0;
1286 V9fsPath dpath, path;
1287 uint16_t nwnames;
1288 struct stat stbuf;
1289 size_t offset = 7;
1290 int32_t fid, newfid;
1291 V9fsString *wnames = NULL;
1292 V9fsFidState *fidp;
1293 V9fsFidState *newfidp = NULL;
1294 V9fsPDU *pdu = opaque;
1295 V9fsState *s = pdu->s;
1296 V9fsQID qid;
1297
1298 err = pdu_unmarshal(pdu, offset, "ddw", &fid, &newfid, &nwnames);
1299 if (err < 0) {
1300 pdu_complete(pdu, err);
1301 return ;
1302 }
1303 offset += err;
1304
1305 trace_v9fs_walk(pdu->tag, pdu->id, fid, newfid, nwnames);
1306
1307 if (nwnames && nwnames <= P9_MAXWELEM) {
1308 wnames = g_malloc0(sizeof(wnames[0]) * nwnames);
1309 qids = g_malloc0(sizeof(qids[0]) * nwnames);
1310 for (i = 0; i < nwnames; i++) {
1311 err = pdu_unmarshal(pdu, offset, "s", &wnames[i]);
1312 if (err < 0) {
1313 goto out_nofid;
1314 }
1315 if (name_is_illegal(wnames[i].data)) {
1316 err = -ENOENT;
1317 goto out_nofid;
1318 }
1319 offset += err;
1320 }
1321 } else if (nwnames > P9_MAXWELEM) {
1322 err = -EINVAL;
1323 goto out_nofid;
1324 }
1325 fidp = get_fid(pdu, fid);
1326 if (fidp == NULL) {
1327 err = -ENOENT;
1328 goto out_nofid;
1329 }
1330
1331 v9fs_path_init(&dpath);
1332 v9fs_path_init(&path);
1333
1334 err = fid_to_qid(pdu, fidp, &qid);
1335 if (err < 0) {
1336 goto out;
1337 }
1338
1339 /*
1340 * Both dpath and path initially poin to fidp.
1341 * Needed to handle request with nwnames == 0
1342 */
1343 v9fs_path_copy(&dpath, &fidp->path);
1344 v9fs_path_copy(&path, &fidp->path);
1345 for (name_idx = 0; name_idx < nwnames; name_idx++) {
1346 if (not_same_qid(&pdu->s->root_qid, &qid) ||
1347 strcmp("..", wnames[name_idx].data)) {
1348 err = v9fs_co_name_to_path(pdu, &dpath, wnames[name_idx].data,
1349 &path);
1350 if (err < 0) {
1351 goto out;
1352 }
1353
1354 err = v9fs_co_lstat(pdu, &path, &stbuf);
1355 if (err < 0) {
1356 goto out;
1357 }
1358 stat_to_qid(&stbuf, &qid);
1359 v9fs_path_copy(&dpath, &path);
1360 }
1361 memcpy(&qids[name_idx], &qid, sizeof(qid));
1362 }
1363 if (fid == newfid) {
1364 BUG_ON(fidp->fid_type != P9_FID_NONE);
1365 v9fs_path_copy(&fidp->path, &path);
1366 } else {
1367 newfidp = alloc_fid(s, newfid);
1368 if (newfidp == NULL) {
1369 err = -EINVAL;
1370 goto out;
1371 }
1372 newfidp->uid = fidp->uid;
1373 v9fs_path_copy(&newfidp->path, &path);
1374 }
1375 err = v9fs_walk_marshal(pdu, nwnames, qids);
1376 trace_v9fs_walk_return(pdu->tag, pdu->id, nwnames, qids);
1377 out:
1378 put_fid(pdu, fidp);
1379 if (newfidp) {
1380 put_fid(pdu, newfidp);
1381 }
1382 v9fs_path_free(&dpath);
1383 v9fs_path_free(&path);
1384 out_nofid:
1385 pdu_complete(pdu, err);
1386 if (nwnames && nwnames <= P9_MAXWELEM) {
1387 for (name_idx = 0; name_idx < nwnames; name_idx++) {
1388 v9fs_string_free(&wnames[name_idx]);
1389 }
1390 g_free(wnames);
1391 g_free(qids);
1392 }
1393 }
1394
1395 static int32_t coroutine_fn get_iounit(V9fsPDU *pdu, V9fsPath *path)
1396 {
1397 struct statfs stbuf;
1398 int32_t iounit = 0;
1399 V9fsState *s = pdu->s;
1400
1401 /*
1402 * iounit should be multiples of f_bsize (host filesystem block size
1403 * and as well as less than (client msize - P9_IOHDRSZ))
1404 */
1405 if (!v9fs_co_statfs(pdu, path, &stbuf)) {
1406 iounit = stbuf.f_bsize;
1407 iounit *= (s->msize - P9_IOHDRSZ)/stbuf.f_bsize;
1408 }
1409 if (!iounit) {
1410 iounit = s->msize - P9_IOHDRSZ;
1411 }
1412 return iounit;
1413 }
1414
1415 static void coroutine_fn v9fs_open(void *opaque)
1416 {
1417 int flags;
1418 int32_t fid;
1419 int32_t mode;
1420 V9fsQID qid;
1421 int iounit = 0;
1422 ssize_t err = 0;
1423 size_t offset = 7;
1424 struct stat stbuf;
1425 V9fsFidState *fidp;
1426 V9fsPDU *pdu = opaque;
1427 V9fsState *s = pdu->s;
1428
1429 if (s->proto_version == V9FS_PROTO_2000L) {
1430 err = pdu_unmarshal(pdu, offset, "dd", &fid, &mode);
1431 } else {
1432 uint8_t modebyte;
1433 err = pdu_unmarshal(pdu, offset, "db", &fid, &modebyte);
1434 mode = modebyte;
1435 }
1436 if (err < 0) {
1437 goto out_nofid;
1438 }
1439 trace_v9fs_open(pdu->tag, pdu->id, fid, mode);
1440
1441 fidp = get_fid(pdu, fid);
1442 if (fidp == NULL) {
1443 err = -ENOENT;
1444 goto out_nofid;
1445 }
1446 BUG_ON(fidp->fid_type != P9_FID_NONE);
1447
1448 err = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
1449 if (err < 0) {
1450 goto out;
1451 }
1452 stat_to_qid(&stbuf, &qid);
1453 if (S_ISDIR(stbuf.st_mode)) {
1454 err = v9fs_co_opendir(pdu, fidp);
1455 if (err < 0) {
1456 goto out;
1457 }
1458 fidp->fid_type = P9_FID_DIR;
1459 err = pdu_marshal(pdu, offset, "Qd", &qid, 0);
1460 if (err < 0) {
1461 goto out;
1462 }
1463 err += offset;
1464 } else {
1465 if (s->proto_version == V9FS_PROTO_2000L) {
1466 flags = get_dotl_openflags(s, mode);
1467 } else {
1468 flags = omode_to_uflags(mode);
1469 }
1470 if (is_ro_export(&s->ctx)) {
1471 if (mode & O_WRONLY || mode & O_RDWR ||
1472 mode & O_APPEND || mode & O_TRUNC) {
1473 err = -EROFS;
1474 goto out;
1475 }
1476 }
1477 err = v9fs_co_open(pdu, fidp, flags);
1478 if (err < 0) {
1479 goto out;
1480 }
1481 fidp->fid_type = P9_FID_FILE;
1482 fidp->open_flags = flags;
1483 if (flags & O_EXCL) {
1484 /*
1485 * We let the host file system do O_EXCL check
1486 * We should not reclaim such fd
1487 */
1488 fidp->flags |= FID_NON_RECLAIMABLE;
1489 }
1490 iounit = get_iounit(pdu, &fidp->path);
1491 err = pdu_marshal(pdu, offset, "Qd", &qid, iounit);
1492 if (err < 0) {
1493 goto out;
1494 }
1495 err += offset;
1496 }
1497 trace_v9fs_open_return(pdu->tag, pdu->id,
1498 qid.type, qid.version, qid.path, iounit);
1499 out:
1500 put_fid(pdu, fidp);
1501 out_nofid:
1502 pdu_complete(pdu, err);
1503 }
1504
1505 static void coroutine_fn v9fs_lcreate(void *opaque)
1506 {
1507 int32_t dfid, flags, mode;
1508 gid_t gid;
1509 ssize_t err = 0;
1510 ssize_t offset = 7;
1511 V9fsString name;
1512 V9fsFidState *fidp;
1513 struct stat stbuf;
1514 V9fsQID qid;
1515 int32_t iounit;
1516 V9fsPDU *pdu = opaque;
1517
1518 v9fs_string_init(&name);
1519 err = pdu_unmarshal(pdu, offset, "dsddd", &dfid,
1520 &name, &flags, &mode, &gid);
1521 if (err < 0) {
1522 goto out_nofid;
1523 }
1524 trace_v9fs_lcreate(pdu->tag, pdu->id, dfid, flags, mode, gid);
1525
1526 if (name_is_illegal(name.data)) {
1527 err = -ENOENT;
1528 goto out_nofid;
1529 }
1530
1531 if (!strcmp(".", name.data) || !strcmp("..", name.data)) {
1532 err = -EEXIST;
1533 goto out_nofid;
1534 }
1535
1536 fidp = get_fid(pdu, dfid);
1537 if (fidp == NULL) {
1538 err = -ENOENT;
1539 goto out_nofid;
1540 }
1541
1542 flags = get_dotl_openflags(pdu->s, flags);
1543 err = v9fs_co_open2(pdu, fidp, &name, gid,
1544 flags | O_CREAT, mode, &stbuf);
1545 if (err < 0) {
1546 goto out;
1547 }
1548 fidp->fid_type = P9_FID_FILE;
1549 fidp->open_flags = flags;
1550 if (flags & O_EXCL) {
1551 /*
1552 * We let the host file system do O_EXCL check
1553 * We should not reclaim such fd
1554 */
1555 fidp->flags |= FID_NON_RECLAIMABLE;
1556 }
1557 iounit = get_iounit(pdu, &fidp->path);
1558 stat_to_qid(&stbuf, &qid);
1559 err = pdu_marshal(pdu, offset, "Qd", &qid, iounit);
1560 if (err < 0) {
1561 goto out;
1562 }
1563 err += offset;
1564 trace_v9fs_lcreate_return(pdu->tag, pdu->id,
1565 qid.type, qid.version, qid.path, iounit);
1566 out:
1567 put_fid(pdu, fidp);
1568 out_nofid:
1569 pdu_complete(pdu, err);
1570 v9fs_string_free(&name);
1571 }
1572
1573 static void v9fs_fsync(void *opaque)
1574 {
1575 int err;
1576 int32_t fid;
1577 int datasync;
1578 size_t offset = 7;
1579 V9fsFidState *fidp;
1580 V9fsPDU *pdu = opaque;
1581
1582 err = pdu_unmarshal(pdu, offset, "dd", &fid, &datasync);
1583 if (err < 0) {
1584 goto out_nofid;
1585 }
1586 trace_v9fs_fsync(pdu->tag, pdu->id, fid, datasync);
1587
1588 fidp = get_fid(pdu, fid);
1589 if (fidp == NULL) {
1590 err = -ENOENT;
1591 goto out_nofid;
1592 }
1593 err = v9fs_co_fsync(pdu, fidp, datasync);
1594 if (!err) {
1595 err = offset;
1596 }
1597 put_fid(pdu, fidp);
1598 out_nofid:
1599 pdu_complete(pdu, err);
1600 }
1601
1602 static void coroutine_fn v9fs_clunk(void *opaque)
1603 {
1604 int err;
1605 int32_t fid;
1606 size_t offset = 7;
1607 V9fsFidState *fidp;
1608 V9fsPDU *pdu = opaque;
1609 V9fsState *s = pdu->s;
1610
1611 err = pdu_unmarshal(pdu, offset, "d", &fid);
1612 if (err < 0) {
1613 goto out_nofid;
1614 }
1615 trace_v9fs_clunk(pdu->tag, pdu->id, fid);
1616
1617 fidp = clunk_fid(s, fid);
1618 if (fidp == NULL) {
1619 err = -ENOENT;
1620 goto out_nofid;
1621 }
1622 /*
1623 * Bump the ref so that put_fid will
1624 * free the fid.
1625 */
1626 fidp->ref++;
1627 err = put_fid(pdu, fidp);
1628 if (!err) {
1629 err = offset;
1630 }
1631 out_nofid:
1632 pdu_complete(pdu, err);
1633 }
1634
1635 static int v9fs_xattr_read(V9fsState *s, V9fsPDU *pdu, V9fsFidState *fidp,
1636 uint64_t off, uint32_t max_count)
1637 {
1638 ssize_t err;
1639 size_t offset = 7;
1640 int read_count;
1641 int64_t xattr_len;
1642 V9fsVirtioState *v = container_of(s, V9fsVirtioState, state);
1643 VirtQueueElement *elem = v->elems[pdu->idx];
1644
1645 xattr_len = fidp->fs.xattr.len;
1646 read_count = xattr_len - off;
1647 if (read_count > max_count) {
1648 read_count = max_count;
1649 } else if (read_count < 0) {
1650 /*
1651 * read beyond XATTR value
1652 */
1653 read_count = 0;
1654 }
1655 err = pdu_marshal(pdu, offset, "d", read_count);
1656 if (err < 0) {
1657 return err;
1658 }
1659 offset += err;
1660
1661 err = v9fs_pack(elem->in_sg, elem->in_num, offset,
1662 ((char *)fidp->fs.xattr.value) + off,
1663 read_count);
1664 if (err < 0) {
1665 return err;
1666 }
1667 offset += err;
1668 return offset;
1669 }
1670
1671 static int coroutine_fn v9fs_do_readdir_with_stat(V9fsPDU *pdu,
1672 V9fsFidState *fidp,
1673 uint32_t max_count)
1674 {
1675 V9fsPath path;
1676 V9fsStat v9stat;
1677 int len, err = 0;
1678 int32_t count = 0;
1679 struct stat stbuf;
1680 off_t saved_dir_pos;
1681 struct dirent *dent;
1682
1683 /* save the directory position */
1684 saved_dir_pos = v9fs_co_telldir(pdu, fidp);
1685 if (saved_dir_pos < 0) {
1686 return saved_dir_pos;
1687 }
1688
1689 while (1) {
1690 v9fs_path_init(&path);
1691
1692 v9fs_readdir_lock(&fidp->fs.dir);
1693
1694 err = v9fs_co_readdir(pdu, fidp, &dent);
1695 if (err || !dent) {
1696 break;
1697 }
1698 err = v9fs_co_name_to_path(pdu, &fidp->path, dent->d_name, &path);
1699 if (err < 0) {
1700 break;
1701 }
1702 err = v9fs_co_lstat(pdu, &path, &stbuf);
1703 if (err < 0) {
1704 break;
1705 }
1706 err = stat_to_v9stat(pdu, &path, &stbuf, &v9stat);
1707 if (err < 0) {
1708 break;
1709 }
1710 /* 11 = 7 + 4 (7 = start offset, 4 = space for storing count) */
1711 len = pdu_marshal(pdu, 11 + count, "S", &v9stat);
1712
1713 v9fs_readdir_unlock(&fidp->fs.dir);
1714
1715 if ((len != (v9stat.size + 2)) || ((count + len) > max_count)) {
1716 /* Ran out of buffer. Set dir back to old position and return */
1717 v9fs_co_seekdir(pdu, fidp, saved_dir_pos);
1718 v9fs_stat_free(&v9stat);
1719 v9fs_path_free(&path);
1720 return count;
1721 }
1722 count += len;
1723 v9fs_stat_free(&v9stat);
1724 v9fs_path_free(&path);
1725 saved_dir_pos = dent->d_off;
1726 }
1727
1728 v9fs_readdir_unlock(&fidp->fs.dir);
1729
1730 v9fs_path_free(&path);
1731 if (err < 0) {
1732 return err;
1733 }
1734 return count;
1735 }
1736
1737 /*
1738 * Create a QEMUIOVector for a sub-region of PDU iovecs
1739 *
1740 * @qiov: uninitialized QEMUIOVector
1741 * @skip: number of bytes to skip from beginning of PDU
1742 * @size: number of bytes to include
1743 * @is_write: true - write, false - read
1744 *
1745 * The resulting QEMUIOVector has heap-allocated iovecs and must be cleaned up
1746 * with qemu_iovec_destroy().
1747 */
1748 static void v9fs_init_qiov_from_pdu(QEMUIOVector *qiov, V9fsPDU *pdu,
1749 size_t skip, size_t size,
1750 bool is_write)
1751 {
1752 QEMUIOVector elem;
1753 struct iovec *iov;
1754 unsigned int niov;
1755
1756 virtio_init_iov_from_pdu(pdu, &iov, &niov, is_write);
1757
1758 qemu_iovec_init_external(&elem, iov, niov);
1759 qemu_iovec_init(qiov, niov);
1760 qemu_iovec_concat(qiov, &elem, skip, size);
1761 }
1762
1763 static void coroutine_fn v9fs_read(void *opaque)
1764 {
1765 int32_t fid;
1766 uint64_t off;
1767 ssize_t err = 0;
1768 int32_t count = 0;
1769 size_t offset = 7;
1770 uint32_t max_count;
1771 V9fsFidState *fidp;
1772 V9fsPDU *pdu = opaque;
1773 V9fsState *s = pdu->s;
1774
1775 err = pdu_unmarshal(pdu, offset, "dqd", &fid, &off, &max_count);
1776 if (err < 0) {
1777 goto out_nofid;
1778 }
1779 trace_v9fs_read(pdu->tag, pdu->id, fid, off, max_count);
1780
1781 fidp = get_fid(pdu, fid);
1782 if (fidp == NULL) {
1783 err = -EINVAL;
1784 goto out_nofid;
1785 }
1786 if (fidp->fid_type == P9_FID_DIR) {
1787
1788 if (off == 0) {
1789 v9fs_co_rewinddir(pdu, fidp);
1790 }
1791 count = v9fs_do_readdir_with_stat(pdu, fidp, max_count);
1792 if (count < 0) {
1793 err = count;
1794 goto out;
1795 }
1796 err = pdu_marshal(pdu, offset, "d", count);
1797 if (err < 0) {
1798 goto out;
1799 }
1800 err += offset + count;
1801 } else if (fidp->fid_type == P9_FID_FILE) {
1802 QEMUIOVector qiov_full;
1803 QEMUIOVector qiov;
1804 int32_t len;
1805
1806 v9fs_init_qiov_from_pdu(&qiov_full, pdu, offset + 4, max_count, false);
1807 qemu_iovec_init(&qiov, qiov_full.niov);
1808 do {
1809 qemu_iovec_reset(&qiov);
1810 qemu_iovec_concat(&qiov, &qiov_full, count, qiov_full.size - count);
1811 if (0) {
1812 print_sg(qiov.iov, qiov.niov);
1813 }
1814 /* Loop in case of EINTR */
1815 do {
1816 len = v9fs_co_preadv(pdu, fidp, qiov.iov, qiov.niov, off);
1817 if (len >= 0) {
1818 off += len;
1819 count += len;
1820 }
1821 } while (len == -EINTR && !pdu->cancelled);
1822 if (len < 0) {
1823 /* IO error return the error */
1824 err = len;
1825 goto out_free_iovec;
1826 }
1827 } while (count < max_count && len > 0);
1828 err = pdu_marshal(pdu, offset, "d", count);
1829 if (err < 0) {
1830 goto out_free_iovec;
1831 }
1832 err += offset + count;
1833 out_free_iovec:
1834 qemu_iovec_destroy(&qiov);
1835 qemu_iovec_destroy(&qiov_full);
1836 } else if (fidp->fid_type == P9_FID_XATTR) {
1837 err = v9fs_xattr_read(s, pdu, fidp, off, max_count);
1838 } else {
1839 err = -EINVAL;
1840 }
1841 trace_v9fs_read_return(pdu->tag, pdu->id, count, err);
1842 out:
1843 put_fid(pdu, fidp);
1844 out_nofid:
1845 pdu_complete(pdu, err);
1846 }
1847
1848 static size_t v9fs_readdir_data_size(V9fsString *name)
1849 {
1850 /*
1851 * Size of each dirent on the wire: size of qid (13) + size of offset (8)
1852 * size of type (1) + size of name.size (2) + strlen(name.data)
1853 */
1854 return 24 + v9fs_string_size(name);
1855 }
1856
1857 static int coroutine_fn v9fs_do_readdir(V9fsPDU *pdu, V9fsFidState *fidp,
1858 int32_t max_count)
1859 {
1860 size_t size;
1861 V9fsQID qid;
1862 V9fsString name;
1863 int len, err = 0;
1864 int32_t count = 0;
1865 off_t saved_dir_pos;
1866 struct dirent *dent;
1867
1868 /* save the directory position */
1869 saved_dir_pos = v9fs_co_telldir(pdu, fidp);
1870 if (saved_dir_pos < 0) {
1871 return saved_dir_pos;
1872 }
1873
1874 while (1) {
1875 v9fs_readdir_lock(&fidp->fs.dir);
1876
1877 err = v9fs_co_readdir(pdu, fidp, &dent);
1878 if (err || !dent) {
1879 break;
1880 }
1881 v9fs_string_init(&name);
1882 v9fs_string_sprintf(&name, "%s", dent->d_name);
1883 if ((count + v9fs_readdir_data_size(&name)) > max_count) {
1884 v9fs_readdir_unlock(&fidp->fs.dir);
1885
1886 /* Ran out of buffer. Set dir back to old position and return */
1887 v9fs_co_seekdir(pdu, fidp, saved_dir_pos);
1888 v9fs_string_free(&name);
1889 return count;
1890 }
1891 /*
1892 * Fill up just the path field of qid because the client uses
1893 * only that. To fill the entire qid structure we will have
1894 * to stat each dirent found, which is expensive
1895 */
1896 size = MIN(sizeof(dent->d_ino), sizeof(qid.path));
1897 memcpy(&qid.path, &dent->d_ino, size);
1898 /* Fill the other fields with dummy values */
1899 qid.type = 0;
1900 qid.version = 0;
1901
1902 /* 11 = 7 + 4 (7 = start offset, 4 = space for storing count) */
1903 len = pdu_marshal(pdu, 11 + count, "Qqbs",
1904 &qid, dent->d_off,
1905 dent->d_type, &name);
1906
1907 v9fs_readdir_unlock(&fidp->fs.dir);
1908
1909 if (len < 0) {
1910 v9fs_co_seekdir(pdu, fidp, saved_dir_pos);
1911 v9fs_string_free(&name);
1912 return len;
1913 }
1914 count += len;
1915 v9fs_string_free(&name);
1916 saved_dir_pos = dent->d_off;
1917 }
1918
1919 v9fs_readdir_unlock(&fidp->fs.dir);
1920
1921 if (err < 0) {
1922 return err;
1923 }
1924 return count;
1925 }
1926
1927 static void coroutine_fn v9fs_readdir(void *opaque)
1928 {
1929 int32_t fid;
1930 V9fsFidState *fidp;
1931 ssize_t retval = 0;
1932 size_t offset = 7;
1933 uint64_t initial_offset;
1934 int32_t count;
1935 uint32_t max_count;
1936 V9fsPDU *pdu = opaque;
1937
1938 retval = pdu_unmarshal(pdu, offset, "dqd", &fid,
1939 &initial_offset, &max_count);
1940 if (retval < 0) {
1941 goto out_nofid;
1942 }
1943 trace_v9fs_readdir(pdu->tag, pdu->id, fid, initial_offset, max_count);
1944
1945 fidp = get_fid(pdu, fid);
1946 if (fidp == NULL) {
1947 retval = -EINVAL;
1948 goto out_nofid;
1949 }
1950 if (!fidp->fs.dir.stream) {
1951 retval = -EINVAL;
1952 goto out;
1953 }
1954 if (initial_offset == 0) {
1955 v9fs_co_rewinddir(pdu, fidp);
1956 } else {
1957 v9fs_co_seekdir(pdu, fidp, initial_offset);
1958 }
1959 count = v9fs_do_readdir(pdu, fidp, max_count);
1960 if (count < 0) {
1961 retval = count;
1962 goto out;
1963 }
1964 retval = pdu_marshal(pdu, offset, "d", count);
1965 if (retval < 0) {
1966 goto out;
1967 }
1968 retval += count + offset;
1969 trace_v9fs_readdir_return(pdu->tag, pdu->id, count, retval);
1970 out:
1971 put_fid(pdu, fidp);
1972 out_nofid:
1973 pdu_complete(pdu, retval);
1974 }
1975
1976 static int v9fs_xattr_write(V9fsState *s, V9fsPDU *pdu, V9fsFidState *fidp,
1977 uint64_t off, uint32_t count,
1978 struct iovec *sg, int cnt)
1979 {
1980 int i, to_copy;
1981 ssize_t err = 0;
1982 int write_count;
1983 int64_t xattr_len;
1984 size_t offset = 7;
1985
1986
1987 xattr_len = fidp->fs.xattr.len;
1988 write_count = xattr_len - off;
1989 if (write_count > count) {
1990 write_count = count;
1991 } else if (write_count < 0) {
1992 /*
1993 * write beyond XATTR value len specified in
1994 * xattrcreate
1995 */
1996 err = -ENOSPC;
1997 goto out;
1998 }
1999 err = pdu_marshal(pdu, offset, "d", write_count);
2000 if (err < 0) {
2001 return err;
2002 }
2003 err += offset;
2004 fidp->fs.xattr.copied_len += write_count;
2005 /*
2006 * Now copy the content from sg list
2007 */
2008 for (i = 0; i < cnt; i++) {
2009 if (write_count > sg[i].iov_len) {
2010 to_copy = sg[i].iov_len;
2011 } else {
2012 to_copy = write_count;
2013 }
2014 memcpy((char *)fidp->fs.xattr.value + off, sg[i].iov_base, to_copy);
2015 /* updating vs->off since we are not using below */
2016 off += to_copy;
2017 write_count -= to_copy;
2018 }
2019 out:
2020 return err;
2021 }
2022
2023 static void coroutine_fn v9fs_write(void *opaque)
2024 {
2025 ssize_t err;
2026 int32_t fid;
2027 uint64_t off;
2028 uint32_t count;
2029 int32_t len = 0;
2030 int32_t total = 0;
2031 size_t offset = 7;
2032 V9fsFidState *fidp;
2033 V9fsPDU *pdu = opaque;
2034 V9fsState *s = pdu->s;
2035 QEMUIOVector qiov_full;
2036 QEMUIOVector qiov;
2037
2038 err = pdu_unmarshal(pdu, offset, "dqd", &fid, &off, &count);
2039 if (err < 0) {
2040 pdu_complete(pdu, err);
2041 return;
2042 }
2043 offset += err;
2044 v9fs_init_qiov_from_pdu(&qiov_full, pdu, offset, count, true);
2045 trace_v9fs_write(pdu->tag, pdu->id, fid, off, count, qiov_full.niov);
2046
2047 fidp = get_fid(pdu, fid);
2048 if (fidp == NULL) {
2049 err = -EINVAL;
2050 goto out_nofid;
2051 }
2052 if (fidp->fid_type == P9_FID_FILE) {
2053 if (fidp->fs.fd == -1) {
2054 err = -EINVAL;
2055 goto out;
2056 }
2057 } else if (fidp->fid_type == P9_FID_XATTR) {
2058 /*
2059 * setxattr operation
2060 */
2061 err = v9fs_xattr_write(s, pdu, fidp, off, count,
2062 qiov_full.iov, qiov_full.niov);
2063 goto out;
2064 } else {
2065 err = -EINVAL;
2066 goto out;
2067 }
2068 qemu_iovec_init(&qiov, qiov_full.niov);
2069 do {
2070 qemu_iovec_reset(&qiov);
2071 qemu_iovec_concat(&qiov, &qiov_full, total, qiov_full.size - total);
2072 if (0) {
2073 print_sg(qiov.iov, qiov.niov);
2074 }
2075 /* Loop in case of EINTR */
2076 do {
2077 len = v9fs_co_pwritev(pdu, fidp, qiov.iov, qiov.niov, off);
2078 if (len >= 0) {
2079 off += len;
2080 total += len;
2081 }
2082 } while (len == -EINTR && !pdu->cancelled);
2083 if (len < 0) {
2084 /* IO error return the error */
2085 err = len;
2086 goto out_qiov;
2087 }
2088 } while (total < count && len > 0);
2089
2090 offset = 7;
2091 err = pdu_marshal(pdu, offset, "d", total);
2092 if (err < 0) {
2093 goto out;
2094 }
2095 err += offset;
2096 trace_v9fs_write_return(pdu->tag, pdu->id, total, err);
2097 out_qiov:
2098 qemu_iovec_destroy(&qiov);
2099 out:
2100 put_fid(pdu, fidp);
2101 out_nofid:
2102 qemu_iovec_destroy(&qiov_full);
2103 pdu_complete(pdu, err);
2104 }
2105
2106 static void coroutine_fn v9fs_create(void *opaque)
2107 {
2108 int32_t fid;
2109 int err = 0;
2110 size_t offset = 7;
2111 V9fsFidState *fidp;
2112 V9fsQID qid;
2113 int32_t perm;
2114 int8_t mode;
2115 V9fsPath path;
2116 struct stat stbuf;
2117 V9fsString name;
2118 V9fsString extension;
2119 int iounit;
2120 V9fsPDU *pdu = opaque;
2121
2122 v9fs_path_init(&path);
2123 v9fs_string_init(&name);
2124 v9fs_string_init(&extension);
2125 err = pdu_unmarshal(pdu, offset, "dsdbs", &fid, &name,
2126 &perm, &mode, &extension);
2127 if (err < 0) {
2128 goto out_nofid;
2129 }
2130 trace_v9fs_create(pdu->tag, pdu->id, fid, name.data, perm, mode);
2131
2132 if (name_is_illegal(name.data)) {
2133 err = -ENOENT;
2134 goto out_nofid;
2135 }
2136
2137 if (!strcmp(".", name.data) || !strcmp("..", name.data)) {
2138 err = -EEXIST;
2139 goto out_nofid;
2140 }
2141
2142 fidp = get_fid(pdu, fid);
2143 if (fidp == NULL) {
2144 err = -EINVAL;
2145 goto out_nofid;
2146 }
2147 if (perm & P9_STAT_MODE_DIR) {
2148 err = v9fs_co_mkdir(pdu, fidp, &name, perm & 0777,
2149 fidp->uid, -1, &stbuf);
2150 if (err < 0) {
2151 goto out;
2152 }
2153 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2154 if (err < 0) {
2155 goto out;
2156 }
2157 v9fs_path_copy(&fidp->path, &path);
2158 err = v9fs_co_opendir(pdu, fidp);
2159 if (err < 0) {
2160 goto out;
2161 }
2162 fidp->fid_type = P9_FID_DIR;
2163 } else if (perm & P9_STAT_MODE_SYMLINK) {
2164 err = v9fs_co_symlink(pdu, fidp, &name,
2165 extension.data, -1 , &stbuf);
2166 if (err < 0) {
2167 goto out;
2168 }
2169 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2170 if (err < 0) {
2171 goto out;
2172 }
2173 v9fs_path_copy(&fidp->path, &path);
2174 } else if (perm & P9_STAT_MODE_LINK) {
2175 int32_t ofid = atoi(extension.data);
2176 V9fsFidState *ofidp = get_fid(pdu, ofid);
2177 if (ofidp == NULL) {
2178 err = -EINVAL;
2179 goto out;
2180 }
2181 err = v9fs_co_link(pdu, ofidp, fidp, &name);
2182 put_fid(pdu, ofidp);
2183 if (err < 0) {
2184 goto out;
2185 }
2186 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2187 if (err < 0) {
2188 fidp->fid_type = P9_FID_NONE;
2189 goto out;
2190 }
2191 v9fs_path_copy(&fidp->path, &path);
2192 err = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
2193 if (err < 0) {
2194 fidp->fid_type = P9_FID_NONE;
2195 goto out;
2196 }
2197 } else if (perm & P9_STAT_MODE_DEVICE) {
2198 char ctype;
2199 uint32_t major, minor;
2200 mode_t nmode = 0;
2201
2202 if (sscanf(extension.data, "%c %u %u", &ctype, &major, &minor) != 3) {
2203 err = -errno;
2204 goto out;
2205 }
2206
2207 switch (ctype) {
2208 case 'c':
2209 nmode = S_IFCHR;
2210 break;
2211 case 'b':
2212 nmode = S_IFBLK;
2213 break;
2214 default:
2215 err = -EIO;
2216 goto out;
2217 }
2218
2219 nmode |= perm & 0777;
2220 err = v9fs_co_mknod(pdu, fidp, &name, fidp->uid, -1,
2221 makedev(major, minor), nmode, &stbuf);
2222 if (err < 0) {
2223 goto out;
2224 }
2225 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2226 if (err < 0) {
2227 goto out;
2228 }
2229 v9fs_path_copy(&fidp->path, &path);
2230 } else if (perm & P9_STAT_MODE_NAMED_PIPE) {
2231 err = v9fs_co_mknod(pdu, fidp, &name, fidp->uid, -1,
2232 0, S_IFIFO | (perm & 0777), &stbuf);
2233 if (err < 0) {
2234 goto out;
2235 }
2236 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2237 if (err < 0) {
2238 goto out;
2239 }
2240 v9fs_path_copy(&fidp->path, &path);
2241 } else if (perm & P9_STAT_MODE_SOCKET) {
2242 err = v9fs_co_mknod(pdu, fidp, &name, fidp->uid, -1,
2243 0, S_IFSOCK | (perm & 0777), &stbuf);
2244 if (err < 0) {
2245 goto out;
2246 }
2247 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2248 if (err < 0) {
2249 goto out;
2250 }
2251 v9fs_path_copy(&fidp->path, &path);
2252 } else {
2253 err = v9fs_co_open2(pdu, fidp, &name, -1,
2254 omode_to_uflags(mode)|O_CREAT, perm, &stbuf);
2255 if (err < 0) {
2256 goto out;
2257 }
2258 fidp->fid_type = P9_FID_FILE;
2259 fidp->open_flags = omode_to_uflags(mode);
2260 if (fidp->open_flags & O_EXCL) {
2261 /*
2262 * We let the host file system do O_EXCL check
2263 * We should not reclaim such fd
2264 */
2265 fidp->flags |= FID_NON_RECLAIMABLE;
2266 }
2267 }
2268 iounit = get_iounit(pdu, &fidp->path);
2269 stat_to_qid(&stbuf, &qid);
2270 err = pdu_marshal(pdu, offset, "Qd", &qid, iounit);
2271 if (err < 0) {
2272 goto out;
2273 }
2274 err += offset;
2275 trace_v9fs_create_return(pdu->tag, pdu->id,
2276 qid.type, qid.version, qid.path, iounit);
2277 out:
2278 put_fid(pdu, fidp);
2279 out_nofid:
2280 pdu_complete(pdu, err);
2281 v9fs_string_free(&name);
2282 v9fs_string_free(&extension);
2283 v9fs_path_free(&path);
2284 }
2285
2286 static void coroutine_fn v9fs_symlink(void *opaque)
2287 {
2288 V9fsPDU *pdu = opaque;
2289 V9fsString name;
2290 V9fsString symname;
2291 V9fsFidState *dfidp;
2292 V9fsQID qid;
2293 struct stat stbuf;
2294 int32_t dfid;
2295 int err = 0;
2296 gid_t gid;
2297 size_t offset = 7;
2298
2299 v9fs_string_init(&name);
2300 v9fs_string_init(&symname);
2301 err = pdu_unmarshal(pdu, offset, "dssd", &dfid, &name, &symname, &gid);
2302 if (err < 0) {
2303 goto out_nofid;
2304 }
2305 trace_v9fs_symlink(pdu->tag, pdu->id, dfid, name.data, symname.data, gid);
2306
2307 if (name_is_illegal(name.data)) {
2308 err = -ENOENT;
2309 goto out_nofid;
2310 }
2311
2312 if (!strcmp(".", name.data) || !strcmp("..", name.data)) {
2313 err = -EEXIST;
2314 goto out_nofid;
2315 }
2316
2317 dfidp = get_fid(pdu, dfid);
2318 if (dfidp == NULL) {
2319 err = -EINVAL;
2320 goto out_nofid;
2321 }
2322 err = v9fs_co_symlink(pdu, dfidp, &name, symname.data, gid, &stbuf);
2323 if (err < 0) {
2324 goto out;
2325 }
2326 stat_to_qid(&stbuf, &qid);
2327 err = pdu_marshal(pdu, offset, "Q", &qid);
2328 if (err < 0) {
2329 goto out;
2330 }
2331 err += offset;
2332 trace_v9fs_symlink_return(pdu->tag, pdu->id,
2333 qid.type, qid.version, qid.path);
2334 out:
2335 put_fid(pdu, dfidp);
2336 out_nofid:
2337 pdu_complete(pdu, err);
2338 v9fs_string_free(&name);
2339 v9fs_string_free(&symname);
2340 }
2341
2342 static void v9fs_flush(void *opaque)
2343 {
2344 ssize_t err;
2345 int16_t tag;
2346 size_t offset = 7;
2347 V9fsPDU *cancel_pdu;
2348 V9fsPDU *pdu = opaque;
2349 V9fsState *s = pdu->s;
2350
2351 err = pdu_unmarshal(pdu, offset, "w", &tag);
2352 if (err < 0) {
2353 pdu_complete(pdu, err);
2354 return;
2355 }
2356 trace_v9fs_flush(pdu->tag, pdu->id, tag);
2357
2358 QLIST_FOREACH(cancel_pdu, &s->active_list, next) {
2359 if (cancel_pdu->tag == tag) {
2360 break;
2361 }
2362 }
2363 if (cancel_pdu) {
2364 cancel_pdu->cancelled = 1;
2365 /*
2366 * Wait for pdu to complete.
2367 */
2368 qemu_co_queue_wait(&cancel_pdu->complete);
2369 cancel_pdu->cancelled = 0;
2370 pdu_free(cancel_pdu);
2371 }
2372 pdu_complete(pdu, 7);
2373 }
2374
2375 static void coroutine_fn v9fs_link(void *opaque)
2376 {
2377 V9fsPDU *pdu = opaque;
2378 int32_t dfid, oldfid;
2379 V9fsFidState *dfidp, *oldfidp;
2380 V9fsString name;
2381 size_t offset = 7;
2382 int err = 0;
2383
2384 v9fs_string_init(&name);
2385 err = pdu_unmarshal(pdu, offset, "dds", &dfid, &oldfid, &name);
2386 if (err < 0) {
2387 goto out_nofid;
2388 }
2389 trace_v9fs_link(pdu->tag, pdu->id, dfid, oldfid, name.data);
2390
2391 if (name_is_illegal(name.data)) {
2392 err = -ENOENT;
2393 goto out_nofid;
2394 }
2395
2396 if (!strcmp(".", name.data) || !strcmp("..", name.data)) {
2397 err = -EEXIST;
2398 goto out_nofid;
2399 }
2400
2401 dfidp = get_fid(pdu, dfid);
2402 if (dfidp == NULL) {
2403 err = -ENOENT;
2404 goto out_nofid;
2405 }
2406
2407 oldfidp = get_fid(pdu, oldfid);
2408 if (oldfidp == NULL) {
2409 err = -ENOENT;
2410 goto out;
2411 }
2412 err = v9fs_co_link(pdu, oldfidp, dfidp, &name);
2413 if (!err) {
2414 err = offset;
2415 }
2416 out:
2417 put_fid(pdu, dfidp);
2418 out_nofid:
2419 v9fs_string_free(&name);
2420 pdu_complete(pdu, err);
2421 }
2422
2423 /* Only works with path name based fid */
2424 static void coroutine_fn v9fs_remove(void *opaque)
2425 {
2426 int32_t fid;
2427 int err = 0;
2428 size_t offset = 7;
2429 V9fsFidState *fidp;
2430 V9fsPDU *pdu = opaque;
2431
2432 err = pdu_unmarshal(pdu, offset, "d", &fid);
2433 if (err < 0) {
2434 goto out_nofid;
2435 }
2436 trace_v9fs_remove(pdu->tag, pdu->id, fid);
2437
2438 fidp = get_fid(pdu, fid);
2439 if (fidp == NULL) {
2440 err = -EINVAL;
2441 goto out_nofid;
2442 }
2443 /* if fs driver is not path based, return EOPNOTSUPP */
2444 if (!(pdu->s->ctx.export_flags & V9FS_PATHNAME_FSCONTEXT)) {
2445 err = -EOPNOTSUPP;
2446 goto out_err;
2447 }
2448 /*
2449 * IF the file is unlinked, we cannot reopen
2450 * the file later. So don't reclaim fd
2451 */
2452 err = v9fs_mark_fids_unreclaim(pdu, &fidp->path);
2453 if (err < 0) {
2454 goto out_err;
2455 }
2456 err = v9fs_co_remove(pdu, &fidp->path);
2457 if (!err) {
2458 err = offset;
2459 }
2460 out_err:
2461 /* For TREMOVE we need to clunk the fid even on failed remove */
2462 clunk_fid(pdu->s, fidp->fid);
2463 put_fid(pdu, fidp);
2464 out_nofid:
2465 pdu_complete(pdu, err);
2466 }
2467
2468 static void coroutine_fn v9fs_unlinkat(void *opaque)
2469 {
2470 int err = 0;
2471 V9fsString name;
2472 int32_t dfid, flags;
2473 size_t offset = 7;
2474 V9fsPath path;
2475 V9fsFidState *dfidp;
2476 V9fsPDU *pdu = opaque;
2477
2478 v9fs_string_init(&name);
2479 err = pdu_unmarshal(pdu, offset, "dsd", &dfid, &name, &flags);
2480 if (err < 0) {
2481 goto out_nofid;
2482 }
2483
2484 if (name_is_illegal(name.data)) {
2485 err = -ENOENT;
2486 goto out_nofid;
2487 }
2488
2489 if (!strcmp(".", name.data)) {
2490 err = -EINVAL;
2491 goto out_nofid;
2492 }
2493
2494 if (!strcmp("..", name.data)) {
2495 err = -ENOTEMPTY;
2496 goto out_nofid;
2497 }
2498
2499 dfidp = get_fid(pdu, dfid);
2500 if (dfidp == NULL) {
2501 err = -EINVAL;
2502 goto out_nofid;
2503 }
2504 /*
2505 * IF the file is unlinked, we cannot reopen
2506 * the file later. So don't reclaim fd
2507 */
2508 v9fs_path_init(&path);
2509 err = v9fs_co_name_to_path(pdu, &dfidp->path, name.data, &path);
2510 if (err < 0) {
2511 goto out_err;
2512 }
2513 err = v9fs_mark_fids_unreclaim(pdu, &path);
2514 if (err < 0) {
2515 goto out_err;
2516 }
2517 err = v9fs_co_unlinkat(pdu, &dfidp->path, &name, flags);
2518 if (!err) {
2519 err = offset;
2520 }
2521 out_err:
2522 put_fid(pdu, dfidp);
2523 v9fs_path_free(&path);
2524 out_nofid:
2525 pdu_complete(pdu, err);
2526 v9fs_string_free(&name);
2527 }
2528
2529
2530 /* Only works with path name based fid */
2531 static int coroutine_fn v9fs_complete_rename(V9fsPDU *pdu, V9fsFidState *fidp,
2532 int32_t newdirfid,
2533 V9fsString *name)
2534 {
2535 char *end;
2536 int err = 0;
2537 V9fsPath new_path;
2538 V9fsFidState *tfidp;
2539 V9fsState *s = pdu->s;
2540 V9fsFidState *dirfidp = NULL;
2541 char *old_name, *new_name;
2542
2543 v9fs_path_init(&new_path);
2544 if (newdirfid != -1) {
2545 dirfidp = get_fid(pdu, newdirfid);
2546 if (dirfidp == NULL) {
2547 err = -ENOENT;
2548 goto out_nofid;
2549 }
2550 BUG_ON(dirfidp->fid_type != P9_FID_NONE);
2551 v9fs_co_name_to_path(pdu, &dirfidp->path, name->data, &new_path);
2552 } else {
2553 old_name = fidp->path.data;
2554 end = strrchr(old_name, '/');
2555 if (end) {
2556 end++;
2557 } else {
2558 end = old_name;
2559 }
2560 new_name = g_malloc0(end - old_name + name->size + 1);
2561 strncat(new_name, old_name, end - old_name);
2562 strncat(new_name + (end - old_name), name->data, name->size);
2563 v9fs_co_name_to_path(pdu, NULL, new_name, &new_path);
2564 g_free(new_name);
2565 }
2566 err = v9fs_co_rename(pdu, &fidp->path, &new_path);
2567 if (err < 0) {
2568 goto out;
2569 }
2570 /*
2571 * Fixup fid's pointing to the old name to
2572 * start pointing to the new name
2573 */
2574 for (tfidp = s->fid_list; tfidp; tfidp = tfidp->next) {
2575 if (v9fs_path_is_ancestor(&fidp->path, &tfidp->path)) {
2576 /* replace the name */
2577 v9fs_fix_path(&tfidp->path, &new_path, strlen(fidp->path.data));
2578 }
2579 }
2580 out:
2581 if (dirfidp) {
2582 put_fid(pdu, dirfidp);
2583 }
2584 v9fs_path_free(&new_path);
2585 out_nofid:
2586 return err;
2587 }
2588
2589 /* Only works with path name based fid */
2590 static void coroutine_fn v9fs_rename(void *opaque)
2591 {
2592 int32_t fid;
2593 ssize_t err = 0;
2594 size_t offset = 7;
2595 V9fsString name;
2596 int32_t newdirfid;
2597 V9fsFidState *fidp;
2598 V9fsPDU *pdu = opaque;
2599 V9fsState *s = pdu->s;
2600
2601 v9fs_string_init(&name);
2602 err = pdu_unmarshal(pdu, offset, "dds", &fid, &newdirfid, &name);
2603 if (err < 0) {
2604 goto out_nofid;
2605 }
2606
2607 if (name_is_illegal(name.data)) {
2608 err = -ENOENT;
2609 goto out_nofid;
2610 }
2611
2612 if (!strcmp(".", name.data) || !strcmp("..", name.data)) {
2613 err = -EISDIR;
2614 goto out_nofid;
2615 }
2616
2617 fidp = get_fid(pdu, fid);
2618 if (fidp == NULL) {
2619 err = -ENOENT;
2620 goto out_nofid;
2621 }
2622 BUG_ON(fidp->fid_type != P9_FID_NONE);
2623 /* if fs driver is not path based, return EOPNOTSUPP */
2624 if (!(pdu->s->ctx.export_flags & V9FS_PATHNAME_FSCONTEXT)) {
2625 err = -EOPNOTSUPP;
2626 goto out;
2627 }
2628 v9fs_path_write_lock(s);
2629 err = v9fs_complete_rename(pdu, fidp, newdirfid, &name);
2630 v9fs_path_unlock(s);
2631 if (!err) {
2632 err = offset;
2633 }
2634 out:
2635 put_fid(pdu, fidp);
2636 out_nofid:
2637 pdu_complete(pdu, err);
2638 v9fs_string_free(&name);
2639 }
2640
2641 static void coroutine_fn v9fs_fix_fid_paths(V9fsPDU *pdu, V9fsPath *olddir,
2642 V9fsString *old_name,
2643 V9fsPath *newdir,
2644 V9fsString *new_name)
2645 {
2646 V9fsFidState *tfidp;
2647 V9fsPath oldpath, newpath;
2648 V9fsState *s = pdu->s;
2649
2650
2651 v9fs_path_init(&oldpath);
2652 v9fs_path_init(&newpath);
2653 v9fs_co_name_to_path(pdu, olddir, old_name->data, &oldpath);
2654 v9fs_co_name_to_path(pdu, newdir, new_name->data, &newpath);
2655
2656 /*
2657 * Fixup fid's pointing to the old name to
2658 * start pointing to the new name
2659 */
2660 for (tfidp = s->fid_list; tfidp; tfidp = tfidp->next) {
2661 if (v9fs_path_is_ancestor(&oldpath, &tfidp->path)) {
2662 /* replace the name */
2663 v9fs_fix_path(&tfidp->path, &newpath, strlen(oldpath.data));
2664 }
2665 }
2666 v9fs_path_free(&oldpath);
2667 v9fs_path_free(&newpath);
2668 }
2669
2670 static int coroutine_fn v9fs_complete_renameat(V9fsPDU *pdu, int32_t olddirfid,
2671 V9fsString *old_name,
2672 int32_t newdirfid,
2673 V9fsString *new_name)
2674 {
2675 int err = 0;
2676 V9fsState *s = pdu->s;
2677 V9fsFidState *newdirfidp = NULL, *olddirfidp = NULL;
2678
2679 olddirfidp = get_fid(pdu, olddirfid);
2680 if (olddirfidp == NULL) {
2681 err = -ENOENT;
2682 goto out;
2683 }
2684 if (newdirfid != -1) {
2685 newdirfidp = get_fid(pdu, newdirfid);
2686 if (newdirfidp == NULL) {
2687 err = -ENOENT;
2688 goto out;
2689 }
2690 } else {
2691 newdirfidp = get_fid(pdu, olddirfid);
2692 }
2693
2694 err = v9fs_co_renameat(pdu, &olddirfidp->path, old_name,
2695 &newdirfidp->path, new_name);
2696 if (err < 0) {
2697 goto out;
2698 }
2699 if (s->ctx.export_flags & V9FS_PATHNAME_FSCONTEXT) {
2700 /* Only for path based fid we need to do the below fixup */
2701 v9fs_fix_fid_paths(pdu, &olddirfidp->path, old_name,
2702 &newdirfidp->path, new_name);
2703 }
2704 out:
2705 if (olddirfidp) {
2706 put_fid(pdu, olddirfidp);
2707 }
2708 if (newdirfidp) {
2709 put_fid(pdu, newdirfidp);
2710 }
2711 return err;
2712 }
2713
2714 static void coroutine_fn v9fs_renameat(void *opaque)
2715 {
2716 ssize_t err = 0;
2717 size_t offset = 7;
2718 V9fsPDU *pdu = opaque;
2719 V9fsState *s = pdu->s;
2720 int32_t olddirfid, newdirfid;
2721 V9fsString old_name, new_name;
2722
2723 v9fs_string_init(&old_name);
2724 v9fs_string_init(&new_name);
2725 err = pdu_unmarshal(pdu, offset, "dsds", &olddirfid,
2726 &old_name, &newdirfid, &new_name);
2727 if (err < 0) {
2728 goto out_err;
2729 }
2730
2731 if (name_is_illegal(old_name.data) || name_is_illegal(new_name.data)) {
2732 err = -ENOENT;
2733 goto out_err;
2734 }
2735
2736 if (!strcmp(".", old_name.data) || !strcmp("..", old_name.data) ||
2737 !strcmp(".", new_name.data) || !strcmp("..", new_name.data)) {
2738 err = -EISDIR;
2739 goto out_err;
2740 }
2741
2742 v9fs_path_write_lock(s);
2743 err = v9fs_complete_renameat(pdu, olddirfid,
2744 &old_name, newdirfid, &new_name);
2745 v9fs_path_unlock(s);
2746 if (!err) {
2747 err = offset;
2748 }
2749
2750 out_err:
2751 pdu_complete(pdu, err);
2752 v9fs_string_free(&old_name);
2753 v9fs_string_free(&new_name);
2754 }
2755
2756 static void coroutine_fn v9fs_wstat(void *opaque)
2757 {
2758 int32_t fid;
2759 int err = 0;
2760 int16_t unused;
2761 V9fsStat v9stat;
2762 size_t offset = 7;
2763 struct stat stbuf;
2764 V9fsFidState *fidp;
2765 V9fsPDU *pdu = opaque;
2766
2767 v9fs_stat_init(&v9stat);
2768 err = pdu_unmarshal(pdu, offset, "dwS", &fid, &unused, &v9stat);
2769 if (err < 0) {
2770 goto out_nofid;
2771 }
2772 trace_v9fs_wstat(pdu->tag, pdu->id, fid,
2773 v9stat.mode, v9stat.atime, v9stat.mtime);
2774
2775 fidp = get_fid(pdu, fid);
2776 if (fidp == NULL) {
2777 err = -EINVAL;
2778 goto out_nofid;
2779 }
2780 /* do we need to sync the file? */
2781 if (donttouch_stat(&v9stat)) {
2782 err = v9fs_co_fsync(pdu, fidp, 0);
2783 goto out;
2784 }
2785 if (v9stat.mode != -1) {
2786 uint32_t v9_mode;
2787 err = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
2788 if (err < 0) {
2789 goto out;
2790 }
2791 v9_mode = stat_to_v9mode(&stbuf);
2792 if ((v9stat.mode & P9_STAT_MODE_TYPE_BITS) !=
2793 (v9_mode & P9_STAT_MODE_TYPE_BITS)) {
2794 /* Attempting to change the type */
2795 err = -EIO;
2796 goto out;
2797 }
2798 err = v9fs_co_chmod(pdu, &fidp->path,
2799 v9mode_to_mode(v9stat.mode,
2800 &v9stat.extension));
2801 if (err < 0) {
2802 goto out;
2803 }
2804 }
2805 if (v9stat.mtime != -1 || v9stat.atime != -1) {
2806 struct timespec times[2];
2807 if (v9stat.atime != -1) {
2808 times[0].tv_sec = v9stat.atime;
2809 times[0].tv_nsec = 0;
2810 } else {
2811 times[0].tv_nsec = UTIME_OMIT;
2812 }
2813 if (v9stat.mtime != -1) {
2814 times[1].tv_sec = v9stat.mtime;
2815 times[1].tv_nsec = 0;
2816 } else {
2817 times[1].tv_nsec = UTIME_OMIT;
2818 }
2819 err = v9fs_co_utimensat(pdu, &fidp->path, times);
2820 if (err < 0) {
2821 goto out;
2822 }
2823 }
2824 if (v9stat.n_gid != -1 || v9stat.n_uid != -1) {
2825 err = v9fs_co_chown(pdu, &fidp->path, v9stat.n_uid, v9stat.n_gid);
2826 if (err < 0) {
2827 goto out;
2828 }
2829 }
2830 if (v9stat.name.size != 0) {
2831 err = v9fs_complete_rename(pdu, fidp, -1, &v9stat.name);
2832 if (err < 0) {
2833 goto out;
2834 }
2835 }
2836 if (v9stat.length != -1) {
2837 err = v9fs_co_truncate(pdu, &fidp->path, v9stat.length);
2838 if (err < 0) {
2839 goto out;
2840 }
2841 }
2842 err = offset;
2843 out:
2844 put_fid(pdu, fidp);
2845 out_nofid:
2846 v9fs_stat_free(&v9stat);
2847 pdu_complete(pdu, err);
2848 }
2849
2850 static int v9fs_fill_statfs(V9fsState *s, V9fsPDU *pdu, struct statfs *stbuf)
2851 {
2852 uint32_t f_type;
2853 uint32_t f_bsize;
2854 uint64_t f_blocks;
2855 uint64_t f_bfree;
2856 uint64_t f_bavail;
2857 uint64_t f_files;
2858 uint64_t f_ffree;
2859 uint64_t fsid_val;
2860 uint32_t f_namelen;
2861 size_t offset = 7;
2862 int32_t bsize_factor;
2863
2864 /*
2865 * compute bsize factor based on host file system block size
2866 * and client msize
2867 */
2868 bsize_factor = (s->msize - P9_IOHDRSZ)/stbuf->f_bsize;
2869 if (!bsize_factor) {
2870 bsize_factor = 1;
2871 }
2872 f_type = stbuf->f_type;
2873 f_bsize = stbuf->f_bsize;
2874 f_bsize *= bsize_factor;
2875 /*
2876 * f_bsize is adjusted(multiplied) by bsize factor, so we need to
2877 * adjust(divide) the number of blocks, free blocks and available
2878 * blocks by bsize factor
2879 */
2880 f_blocks = stbuf->f_blocks/bsize_factor;
2881 f_bfree = stbuf->f_bfree/bsize_factor;
2882 f_bavail = stbuf->f_bavail/bsize_factor;
2883 f_files = stbuf->f_files;
2884 f_ffree = stbuf->f_ffree;
2885 fsid_val = (unsigned int) stbuf->f_fsid.__val[0] |
2886 (unsigned long long)stbuf->f_fsid.__val[1] << 32;
2887 f_namelen = stbuf->f_namelen;
2888
2889 return pdu_marshal(pdu, offset, "ddqqqqqqd",
2890 f_type, f_bsize, f_blocks, f_bfree,
2891 f_bavail, f_files, f_ffree,
2892 fsid_val, f_namelen);
2893 }
2894
2895 static void coroutine_fn v9fs_statfs(void *opaque)
2896 {
2897 int32_t fid;
2898 ssize_t retval = 0;
2899 size_t offset = 7;
2900 V9fsFidState *fidp;
2901 struct statfs stbuf;
2902 V9fsPDU *pdu = opaque;
2903 V9fsState *s = pdu->s;
2904
2905 retval = pdu_unmarshal(pdu, offset, "d", &fid);
2906 if (retval < 0) {
2907 goto out_nofid;
2908 }
2909 fidp = get_fid(pdu, fid);
2910 if (fidp == NULL) {
2911 retval = -ENOENT;
2912 goto out_nofid;
2913 }
2914 retval = v9fs_co_statfs(pdu, &fidp->path, &stbuf);
2915 if (retval < 0) {
2916 goto out;
2917 }
2918 retval = v9fs_fill_statfs(s, pdu, &stbuf);
2919 if (retval < 0) {
2920 goto out;
2921 }
2922 retval += offset;
2923 out:
2924 put_fid(pdu, fidp);
2925 out_nofid:
2926 pdu_complete(pdu, retval);
2927 }
2928
2929 static void coroutine_fn v9fs_mknod(void *opaque)
2930 {
2931
2932 int mode;
2933 gid_t gid;
2934 int32_t fid;
2935 V9fsQID qid;
2936 int err = 0;
2937 int major, minor;
2938 size_t offset = 7;
2939 V9fsString name;
2940 struct stat stbuf;
2941 V9fsFidState *fidp;
2942 V9fsPDU *pdu = opaque;
2943
2944 v9fs_string_init(&name);
2945 err = pdu_unmarshal(pdu, offset, "dsdddd", &fid, &name, &mode,
2946 &major, &minor, &gid);
2947 if (err < 0) {
2948 goto out_nofid;
2949 }
2950 trace_v9fs_mknod(pdu->tag, pdu->id, fid, mode, major, minor);
2951
2952 if (name_is_illegal(name.data)) {
2953 err = -ENOENT;
2954 goto out_nofid;
2955 }
2956
2957 if (!strcmp(".", name.data) || !strcmp("..", name.data)) {
2958 err = -EEXIST;
2959 goto out_nofid;
2960 }
2961
2962 fidp = get_fid(pdu, fid);
2963 if (fidp == NULL) {
2964 err = -ENOENT;
2965 goto out_nofid;
2966 }
2967 err = v9fs_co_mknod(pdu, fidp, &name, fidp->uid, gid,
2968 makedev(major, minor), mode, &stbuf);
2969 if (err < 0) {
2970 goto out;
2971 }
2972 stat_to_qid(&stbuf, &qid);
2973 err = pdu_marshal(pdu, offset, "Q", &qid);
2974 if (err < 0) {
2975 goto out;
2976 }
2977 err += offset;
2978 trace_v9fs_mknod_return(pdu->tag, pdu->id,
2979 qid.type, qid.version, qid.path);
2980 out:
2981 put_fid(pdu, fidp);
2982 out_nofid:
2983 pdu_complete(pdu, err);
2984 v9fs_string_free(&name);
2985 }
2986
2987 /*
2988 * Implement posix byte range locking code
2989 * Server side handling of locking code is very simple, because 9p server in
2990 * QEMU can handle only one client. And most of the lock handling
2991 * (like conflict, merging) etc is done by the VFS layer itself, so no need to
2992 * do any thing in * qemu 9p server side lock code path.
2993 * So when a TLOCK request comes, always return success
2994 */
2995 static void coroutine_fn v9fs_lock(void *opaque)
2996 {
2997 int8_t status;
2998 V9fsFlock flock;
2999 size_t offset = 7;
3000 struct stat stbuf;
3001 V9fsFidState *fidp;
3002 int32_t fid, err = 0;
3003 V9fsPDU *pdu = opaque;
3004
3005 status = P9_LOCK_ERROR;
3006 v9fs_string_init(&flock.client_id);
3007 err = pdu_unmarshal(pdu, offset, "dbdqqds", &fid, &flock.type,
3008 &flock.flags, &flock.start, &flock.length,
3009 &flock.proc_id, &flock.client_id);
3010 if (err < 0) {
3011 goto out_nofid;
3012 }
3013 trace_v9fs_lock(pdu->tag, pdu->id, fid,
3014 flock.type, flock.start, flock.length);
3015
3016
3017 /* We support only block flag now (that too ignored currently) */
3018 if (flock.flags & ~P9_LOCK_FLAGS_BLOCK) {
3019 err = -EINVAL;
3020 goto out_nofid;
3021 }
3022 fidp = get_fid(pdu, fid);
3023 if (fidp == NULL) {
3024 err = -ENOENT;
3025 goto out_nofid;
3026 }
3027 err = v9fs_co_fstat(pdu, fidp, &stbuf);
3028 if (err < 0) {
3029 goto out;
3030 }
3031 status = P9_LOCK_SUCCESS;
3032 out:
3033 put_fid(pdu, fidp);
3034 out_nofid:
3035 err = pdu_marshal(pdu, offset, "b", status);
3036 if (err > 0) {
3037 err += offset;
3038 }
3039 trace_v9fs_lock_return(pdu->tag, pdu->id, status);
3040 pdu_complete(pdu, err);
3041 v9fs_string_free(&flock.client_id);
3042 }
3043
3044 /*
3045 * When a TGETLOCK request comes, always return success because all lock
3046 * handling is done by client's VFS layer.
3047 */
3048 static void coroutine_fn v9fs_getlock(void *opaque)
3049 {
3050 size_t offset = 7;
3051 struct stat stbuf;
3052 V9fsFidState *fidp;
3053 V9fsGetlock glock;
3054 int32_t fid, err = 0;
3055 V9fsPDU *pdu = opaque;
3056
3057 v9fs_string_init(&glock.client_id);
3058 err = pdu_unmarshal(pdu, offset, "dbqqds", &fid, &glock.type,
3059 &glock.start, &glock.length, &glock.proc_id,
3060 &glock.client_id);
3061 if (err < 0) {
3062 goto out_nofid;
3063 }
3064 trace_v9fs_getlock(pdu->tag, pdu->id, fid,
3065 glock.type, glock.start, glock.length);
3066
3067 fidp = get_fid(pdu, fid);
3068 if (fidp == NULL) {
3069 err = -ENOENT;
3070 goto out_nofid;
3071 }
3072 err = v9fs_co_fstat(pdu, fidp, &stbuf);
3073 if (err < 0) {
3074 goto out;
3075 }
3076 glock.type = P9_LOCK_TYPE_UNLCK;
3077 err = pdu_marshal(pdu, offset, "bqqds", glock.type,
3078 glock.start, glock.length, glock.proc_id,
3079 &glock.client_id);
3080 if (err < 0) {
3081 goto out;
3082 }
3083 err += offset;
3084 trace_v9fs_getlock_return(pdu->tag, pdu->id, glock.type, glock.start,
3085 glock.length, glock.proc_id);
3086 out:
3087 put_fid(pdu, fidp);
3088 out_nofid:
3089 pdu_complete(pdu, err);
3090 v9fs_string_free(&glock.client_id);
3091 }
3092
3093 static void coroutine_fn v9fs_mkdir(void *opaque)
3094 {
3095 V9fsPDU *pdu = opaque;
3096 size_t offset = 7;
3097 int32_t fid;
3098 struct stat stbuf;
3099 V9fsQID qid;
3100 V9fsString name;
3101 V9fsFidState *fidp;
3102 gid_t gid;
3103 int mode;
3104 int err = 0;
3105
3106 v9fs_string_init(&name);
3107 err = pdu_unmarshal(pdu, offset, "dsdd", &fid, &name, &mode, &gid);
3108 if (err < 0) {
3109 goto out_nofid;
3110 }
3111 trace_v9fs_mkdir(pdu->tag, pdu->id, fid, name.data, mode, gid);
3112
3113 if (name_is_illegal(name.data)) {
3114 err = -ENOENT;
3115 goto out_nofid;
3116 }
3117
3118 if (!strcmp(".", name.data) || !strcmp("..", name.data)) {
3119 err = -EEXIST;
3120 goto out_nofid;
3121 }
3122
3123 fidp = get_fid(pdu, fid);
3124 if (fidp == NULL) {
3125 err = -ENOENT;
3126 goto out_nofid;
3127 }
3128 err = v9fs_co_mkdir(pdu, fidp, &name, mode, fidp->uid, gid, &stbuf);
3129 if (err < 0) {
3130 goto out;
3131 }
3132 stat_to_qid(&stbuf, &qid);
3133 err = pdu_marshal(pdu, offset, "Q", &qid);
3134 if (err < 0) {
3135 goto out;
3136 }
3137 err += offset;
3138 trace_v9fs_mkdir_return(pdu->tag, pdu->id,
3139 qid.type, qid.version, qid.path, err);
3140 out:
3141 put_fid(pdu, fidp);
3142 out_nofid:
3143 pdu_complete(pdu, err);
3144 v9fs_string_free(&name);
3145 }
3146
3147 static void coroutine_fn v9fs_xattrwalk(void *opaque)
3148 {
3149 int64_t size;
3150 V9fsString name;
3151 ssize_t err = 0;
3152 size_t offset = 7;
3153 int32_t fid, newfid;
3154 V9fsFidState *file_fidp;
3155 V9fsFidState *xattr_fidp = NULL;
3156 V9fsPDU *pdu = opaque;
3157 V9fsState *s = pdu->s;
3158
3159 v9fs_string_init(&name);
3160 err = pdu_unmarshal(pdu, offset, "dds", &fid, &newfid, &name);
3161 if (err < 0) {
3162 goto out_nofid;
3163 }
3164 trace_v9fs_xattrwalk(pdu->tag, pdu->id, fid, newfid, name.data);
3165
3166 file_fidp = get_fid(pdu, fid);
3167 if (file_fidp == NULL) {
3168 err = -ENOENT;
3169 goto out_nofid;
3170 }
3171 xattr_fidp = alloc_fid(s, newfid);
3172 if (xattr_fidp == NULL) {
3173 err = -EINVAL;
3174 goto out;
3175 }
3176 v9fs_path_copy(&xattr_fidp->path, &file_fidp->path);
3177 if (!v9fs_string_size(&name)) {
3178 /*
3179 * listxattr request. Get the size first
3180 */
3181 size = v9fs_co_llistxattr(pdu, &xattr_fidp->path, NULL, 0);
3182 if (size < 0) {
3183 err = size;
3184 clunk_fid(s, xattr_fidp->fid);
3185 goto out;
3186 }
3187 /*
3188 * Read the xattr value
3189 */
3190 xattr_fidp->fs.xattr.len = size;
3191 xattr_fidp->fid_type = P9_FID_XATTR;
3192 xattr_fidp->fs.xattr.copied_len = -1;
3193 if (size) {
3194 xattr_fidp->fs.xattr.value = g_malloc(size);
3195 err = v9fs_co_llistxattr(pdu, &xattr_fidp->path,
3196 xattr_fidp->fs.xattr.value,
3197 xattr_fidp->fs.xattr.len);
3198 if (err < 0) {
3199 clunk_fid(s, xattr_fidp->fid);
3200 goto out;
3201 }
3202 }
3203 err = pdu_marshal(pdu, offset, "q", size);
3204 if (err < 0) {
3205 goto out;
3206 }
3207 err += offset;
3208 } else {
3209 /*
3210 * specific xattr fid. We check for xattr
3211 * presence also collect the xattr size
3212 */
3213 size = v9fs_co_lgetxattr(pdu, &xattr_fidp->path,
3214 &name, NULL, 0);
3215 if (size < 0) {
3216 err = size;
3217 clunk_fid(s, xattr_fidp->fid);
3218 goto out;
3219 }
3220 /*
3221 * Read the xattr value
3222 */
3223 xattr_fidp->fs.xattr.len = size;
3224 xattr_fidp->fid_type = P9_FID_XATTR;
3225 xattr_fidp->fs.xattr.copied_len = -1;
3226 if (size) {
3227 xattr_fidp->fs.xattr.value = g_malloc(size);
3228 err = v9fs_co_lgetxattr(pdu, &xattr_fidp->path,
3229 &name, xattr_fidp->fs.xattr.value,
3230 xattr_fidp->fs.xattr.len);
3231 if (err < 0) {
3232 clunk_fid(s, xattr_fidp->fid);
3233 goto out;
3234 }
3235 }
3236 err = pdu_marshal(pdu, offset, "q", size);
3237 if (err < 0) {
3238 goto out;
3239 }
3240 err += offset;
3241 }
3242 trace_v9fs_xattrwalk_return(pdu->tag, pdu->id, size);
3243 out:
3244 put_fid(pdu, file_fidp);
3245 if (xattr_fidp) {
3246 put_fid(pdu, xattr_fidp);
3247 }
3248 out_nofid:
3249 pdu_complete(pdu, err);
3250 v9fs_string_free(&name);
3251 }
3252
3253 static void coroutine_fn v9fs_xattrcreate(void *opaque)
3254 {
3255 int flags;
3256 int32_t fid;
3257 int64_t size;
3258 ssize_t err = 0;
3259 V9fsString name;
3260 size_t offset = 7;
3261 V9fsFidState *file_fidp;
3262 V9fsFidState *xattr_fidp;
3263 V9fsPDU *pdu = opaque;
3264
3265 v9fs_string_init(&name);
3266 err = pdu_unmarshal(pdu, offset, "dsqd", &fid, &name, &size, &flags);
3267 if (err < 0) {
3268 goto out_nofid;
3269 }
3270 trace_v9fs_xattrcreate(pdu->tag, pdu->id, fid, name.data, size, flags);
3271
3272 file_fidp = get_fid(pdu, fid);
3273 if (file_fidp == NULL) {
3274 err = -EINVAL;
3275 goto out_nofid;
3276 }
3277 /* Make the file fid point to xattr */
3278 xattr_fidp = file_fidp;
3279 xattr_fidp->fid_type = P9_FID_XATTR;
3280 xattr_fidp->fs.xattr.copied_len = 0;
3281 xattr_fidp->fs.xattr.len = size;
3282 xattr_fidp->fs.xattr.flags = flags;
3283 v9fs_string_init(&xattr_fidp->fs.xattr.name);
3284 v9fs_string_copy(&xattr_fidp->fs.xattr.name, &name);
3285 xattr_fidp->fs.xattr.value = g_malloc0(size);
3286 err = offset;
3287 put_fid(pdu, file_fidp);
3288 out_nofid:
3289 pdu_complete(pdu, err);
3290 v9fs_string_free(&name);
3291 }
3292
3293 static void coroutine_fn v9fs_readlink(void *opaque)
3294 {
3295 V9fsPDU *pdu = opaque;
3296 size_t offset = 7;
3297 V9fsString target;
3298 int32_t fid;
3299 int err = 0;
3300 V9fsFidState *fidp;
3301
3302 err = pdu_unmarshal(pdu, offset, "d", &fid);
3303 if (err < 0) {
3304 goto out_nofid;
3305 }
3306 trace_v9fs_readlink(pdu->tag, pdu->id, fid);
3307 fidp = get_fid(pdu, fid);
3308 if (fidp == NULL) {
3309 err = -ENOENT;
3310 goto out_nofid;
3311 }
3312
3313 v9fs_string_init(&target);
3314 err = v9fs_co_readlink(pdu, &fidp->path, &target);
3315 if (err < 0) {
3316 goto out;
3317 }
3318 err = pdu_marshal(pdu, offset, "s", &target);
3319 if (err < 0) {
3320 v9fs_string_free(&target);
3321 goto out;
3322 }
3323 err += offset;
3324 trace_v9fs_readlink_return(pdu->tag, pdu->id, target.data);
3325 v9fs_string_free(&target);
3326 out:
3327 put_fid(pdu, fidp);
3328 out_nofid:
3329 pdu_complete(pdu, err);
3330 }
3331
3332 static CoroutineEntry *pdu_co_handlers[] = {
3333 [P9_TREADDIR] = v9fs_readdir,
3334 [P9_TSTATFS] = v9fs_statfs,
3335 [P9_TGETATTR] = v9fs_getattr,
3336 [P9_TSETATTR] = v9fs_setattr,
3337 [P9_TXATTRWALK] = v9fs_xattrwalk,
3338 [P9_TXATTRCREATE] = v9fs_xattrcreate,
3339 [P9_TMKNOD] = v9fs_mknod,
3340 [P9_TRENAME] = v9fs_rename,
3341 [P9_TLOCK] = v9fs_lock,
3342 [P9_TGETLOCK] = v9fs_getlock,
3343 [P9_TRENAMEAT] = v9fs_renameat,
3344 [P9_TREADLINK] = v9fs_readlink,
3345 [P9_TUNLINKAT] = v9fs_unlinkat,
3346 [P9_TMKDIR] = v9fs_mkdir,
3347 [P9_TVERSION] = v9fs_version,
3348 [P9_TLOPEN] = v9fs_open,
3349 [P9_TATTACH] = v9fs_attach,
3350 [P9_TSTAT] = v9fs_stat,
3351 [P9_TWALK] = v9fs_walk,
3352 [P9_TCLUNK] = v9fs_clunk,
3353 [P9_TFSYNC] = v9fs_fsync,
3354 [P9_TOPEN] = v9fs_open,
3355 [P9_TREAD] = v9fs_read,
3356 #if 0
3357 [P9_TAUTH] = v9fs_auth,
3358 #endif
3359 [P9_TFLUSH] = v9fs_flush,
3360 [P9_TLINK] = v9fs_link,
3361 [P9_TSYMLINK] = v9fs_symlink,
3362 [P9_TCREATE] = v9fs_create,
3363 [P9_TLCREATE] = v9fs_lcreate,
3364 [P9_TWRITE] = v9fs_write,
3365 [P9_TWSTAT] = v9fs_wstat,
3366 [P9_TREMOVE] = v9fs_remove,
3367 };
3368
3369 static void coroutine_fn v9fs_op_not_supp(void *opaque)
3370 {
3371 V9fsPDU *pdu = opaque;
3372 pdu_complete(pdu, -EOPNOTSUPP);
3373 }
3374
3375 static void coroutine_fn v9fs_fs_ro(void *opaque)
3376 {
3377 V9fsPDU *pdu = opaque;
3378 pdu_complete(pdu, -EROFS);
3379 }
3380
3381 static inline bool is_read_only_op(V9fsPDU *pdu)
3382 {
3383 switch (pdu->id) {
3384 case P9_TREADDIR:
3385 case P9_TSTATFS:
3386 case P9_TGETATTR:
3387 case P9_TXATTRWALK:
3388 case P9_TLOCK:
3389 case P9_TGETLOCK:
3390 case P9_TREADLINK:
3391 case P9_TVERSION:
3392 case P9_TLOPEN:
3393 case P9_TATTACH:
3394 case P9_TSTAT:
3395 case P9_TWALK:
3396 case P9_TCLUNK:
3397 case P9_TFSYNC:
3398 case P9_TOPEN:
3399 case P9_TREAD:
3400 case P9_TAUTH:
3401 case P9_TFLUSH:
3402 return 1;
3403 default:
3404 return 0;
3405 }
3406 }
3407
3408 void pdu_submit(V9fsPDU *pdu)
3409 {
3410 Coroutine *co;
3411 CoroutineEntry *handler;
3412 V9fsState *s = pdu->s;
3413
3414 if (pdu->id >= ARRAY_SIZE(pdu_co_handlers) ||
3415 (pdu_co_handlers[pdu->id] == NULL)) {