9pfs: fix crash when fsdev is missing
[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 = pdu->s->transport->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 = pdu->s->transport->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 pdu->s->transport->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.xattrwalk_fid) {
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;
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 }
552
553 #define P9_QID_TYPE_DIR 0x80
554 #define P9_QID_TYPE_SYMLINK 0x02
555
556 #define P9_STAT_MODE_DIR 0x80000000
557 #define P9_STAT_MODE_APPEND 0x40000000
558 #define P9_STAT_MODE_EXCL 0x20000000
559 #define P9_STAT_MODE_MOUNT 0x10000000
560 #define P9_STAT_MODE_AUTH 0x08000000
561 #define P9_STAT_MODE_TMP 0x04000000
562 #define P9_STAT_MODE_SYMLINK 0x02000000
563 #define P9_STAT_MODE_LINK 0x01000000
564 #define P9_STAT_MODE_DEVICE 0x00800000
565 #define P9_STAT_MODE_NAMED_PIPE 0x00200000
566 #define P9_STAT_MODE_SOCKET 0x00100000
567 #define P9_STAT_MODE_SETUID 0x00080000
568 #define P9_STAT_MODE_SETGID 0x00040000
569 #define P9_STAT_MODE_SETVTX 0x00010000
570
571 #define P9_STAT_MODE_TYPE_BITS (P9_STAT_MODE_DIR | \
572 P9_STAT_MODE_SYMLINK | \
573 P9_STAT_MODE_LINK | \
574 P9_STAT_MODE_DEVICE | \
575 P9_STAT_MODE_NAMED_PIPE | \
576 P9_STAT_MODE_SOCKET)
577
578 /* This is the algorithm from ufs in spfs */
579 static void stat_to_qid(const struct stat *stbuf, V9fsQID *qidp)
580 {
581 size_t size;
582
583 memset(&qidp->path, 0, sizeof(qidp->path));
584 size = MIN(sizeof(stbuf->st_ino), sizeof(qidp->path));
585 memcpy(&qidp->path, &stbuf->st_ino, size);
586 qidp->version = stbuf->st_mtime ^ (stbuf->st_size << 8);
587 qidp->type = 0;
588 if (S_ISDIR(stbuf->st_mode)) {
589 qidp->type |= P9_QID_TYPE_DIR;
590 }
591 if (S_ISLNK(stbuf->st_mode)) {
592 qidp->type |= P9_QID_TYPE_SYMLINK;
593 }
594 }
595
596 static int coroutine_fn fid_to_qid(V9fsPDU *pdu, V9fsFidState *fidp,
597 V9fsQID *qidp)
598 {
599 struct stat stbuf;
600 int err;
601
602 err = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
603 if (err < 0) {
604 return err;
605 }
606 stat_to_qid(&stbuf, qidp);
607 return 0;
608 }
609
610 V9fsPDU *pdu_alloc(V9fsState *s)
611 {
612 V9fsPDU *pdu = NULL;
613
614 if (!QLIST_EMPTY(&s->free_list)) {
615 pdu = QLIST_FIRST(&s->free_list);
616 QLIST_REMOVE(pdu, next);
617 QLIST_INSERT_HEAD(&s->active_list, pdu, next);
618 }
619 return pdu;
620 }
621
622 void pdu_free(V9fsPDU *pdu)
623 {
624 V9fsState *s = pdu->s;
625
626 g_assert(!pdu->cancelled);
627 QLIST_REMOVE(pdu, next);
628 QLIST_INSERT_HEAD(&s->free_list, pdu, next);
629 }
630
631 /*
632 * We don't do error checking for pdu_marshal/unmarshal here
633 * because we always expect to have enough space to encode
634 * error details
635 */
636 static void coroutine_fn pdu_complete(V9fsPDU *pdu, ssize_t len)
637 {
638 int8_t id = pdu->id + 1; /* Response */
639 V9fsState *s = pdu->s;
640
641 if (len < 0) {
642 int err = -len;
643 len = 7;
644
645 if (s->proto_version != V9FS_PROTO_2000L) {
646 V9fsString str;
647
648 str.data = strerror(err);
649 str.size = strlen(str.data);
650
651 len += pdu_marshal(pdu, len, "s", &str);
652 id = P9_RERROR;
653 }
654
655 len += pdu_marshal(pdu, len, "d", err);
656
657 if (s->proto_version == V9FS_PROTO_2000L) {
658 id = P9_RLERROR;
659 }
660 trace_v9fs_rerror(pdu->tag, pdu->id, err); /* Trace ERROR */
661 }
662
663 /* fill out the header */
664 pdu_marshal(pdu, 0, "dbw", (int32_t)len, id, pdu->tag);
665
666 /* keep these in sync */
667 pdu->size = len;
668 pdu->id = id;
669
670 pdu_push_and_notify(pdu);
671
672 /* Now wakeup anybody waiting in flush for this request */
673 if (!qemu_co_queue_next(&pdu->complete)) {
674 pdu_free(pdu);
675 }
676 }
677
678 static mode_t v9mode_to_mode(uint32_t mode, V9fsString *extension)
679 {
680 mode_t ret;
681
682 ret = mode & 0777;
683 if (mode & P9_STAT_MODE_DIR) {
684 ret |= S_IFDIR;
685 }
686
687 if (mode & P9_STAT_MODE_SYMLINK) {
688 ret |= S_IFLNK;
689 }
690 if (mode & P9_STAT_MODE_SOCKET) {
691 ret |= S_IFSOCK;
692 }
693 if (mode & P9_STAT_MODE_NAMED_PIPE) {
694 ret |= S_IFIFO;
695 }
696 if (mode & P9_STAT_MODE_DEVICE) {
697 if (extension->size && extension->data[0] == 'c') {
698 ret |= S_IFCHR;
699 } else {
700 ret |= S_IFBLK;
701 }
702 }
703
704 if (!(ret&~0777)) {
705 ret |= S_IFREG;
706 }
707
708 if (mode & P9_STAT_MODE_SETUID) {
709 ret |= S_ISUID;
710 }
711 if (mode & P9_STAT_MODE_SETGID) {
712 ret |= S_ISGID;
713 }
714 if (mode & P9_STAT_MODE_SETVTX) {
715 ret |= S_ISVTX;
716 }
717
718 return ret;
719 }
720
721 static int donttouch_stat(V9fsStat *stat)
722 {
723 if (stat->type == -1 &&
724 stat->dev == -1 &&
725 stat->qid.type == -1 &&
726 stat->qid.version == -1 &&
727 stat->qid.path == -1 &&
728 stat->mode == -1 &&
729 stat->atime == -1 &&
730 stat->mtime == -1 &&
731 stat->length == -1 &&
732 !stat->name.size &&
733 !stat->uid.size &&
734 !stat->gid.size &&
735 !stat->muid.size &&
736 stat->n_uid == -1 &&
737 stat->n_gid == -1 &&
738 stat->n_muid == -1) {
739 return 1;
740 }
741
742 return 0;
743 }
744
745 static void v9fs_stat_init(V9fsStat *stat)
746 {
747 v9fs_string_init(&stat->name);
748 v9fs_string_init(&stat->uid);
749 v9fs_string_init(&stat->gid);
750 v9fs_string_init(&stat->muid);
751 v9fs_string_init(&stat->extension);
752 }
753
754 static void v9fs_stat_free(V9fsStat *stat)
755 {
756 v9fs_string_free(&stat->name);
757 v9fs_string_free(&stat->uid);
758 v9fs_string_free(&stat->gid);
759 v9fs_string_free(&stat->muid);
760 v9fs_string_free(&stat->extension);
761 }
762
763 static uint32_t stat_to_v9mode(const struct stat *stbuf)
764 {
765 uint32_t mode;
766
767 mode = stbuf->st_mode & 0777;
768 if (S_ISDIR(stbuf->st_mode)) {
769 mode |= P9_STAT_MODE_DIR;
770 }
771
772 if (S_ISLNK(stbuf->st_mode)) {
773 mode |= P9_STAT_MODE_SYMLINK;
774 }
775
776 if (S_ISSOCK(stbuf->st_mode)) {
777 mode |= P9_STAT_MODE_SOCKET;
778 }
779
780 if (S_ISFIFO(stbuf->st_mode)) {
781 mode |= P9_STAT_MODE_NAMED_PIPE;
782 }
783
784 if (S_ISBLK(stbuf->st_mode) || S_ISCHR(stbuf->st_mode)) {
785 mode |= P9_STAT_MODE_DEVICE;
786 }
787
788 if (stbuf->st_mode & S_ISUID) {
789 mode |= P9_STAT_MODE_SETUID;
790 }
791
792 if (stbuf->st_mode & S_ISGID) {
793 mode |= P9_STAT_MODE_SETGID;
794 }
795
796 if (stbuf->st_mode & S_ISVTX) {
797 mode |= P9_STAT_MODE_SETVTX;
798 }
799
800 return mode;
801 }
802
803 static int coroutine_fn stat_to_v9stat(V9fsPDU *pdu, V9fsPath *name,
804 const struct stat *stbuf,
805 V9fsStat *v9stat)
806 {
807 int err;
808 const char *str;
809
810 memset(v9stat, 0, sizeof(*v9stat));
811
812 stat_to_qid(stbuf, &v9stat->qid);
813 v9stat->mode = stat_to_v9mode(stbuf);
814 v9stat->atime = stbuf->st_atime;
815 v9stat->mtime = stbuf->st_mtime;
816 v9stat->length = stbuf->st_size;
817
818 v9fs_string_free(&v9stat->uid);
819 v9fs_string_free(&v9stat->gid);
820 v9fs_string_free(&v9stat->muid);
821
822 v9stat->n_uid = stbuf->st_uid;
823 v9stat->n_gid = stbuf->st_gid;
824 v9stat->n_muid = 0;
825
826 v9fs_string_free(&v9stat->extension);
827
828 if (v9stat->mode & P9_STAT_MODE_SYMLINK) {
829 err = v9fs_co_readlink(pdu, name, &v9stat->extension);
830 if (err < 0) {
831 return err;
832 }
833 } else if (v9stat->mode & P9_STAT_MODE_DEVICE) {
834 v9fs_string_sprintf(&v9stat->extension, "%c %u %u",
835 S_ISCHR(stbuf->st_mode) ? 'c' : 'b',
836 major(stbuf->st_rdev), minor(stbuf->st_rdev));
837 } else if (S_ISDIR(stbuf->st_mode) || S_ISREG(stbuf->st_mode)) {
838 v9fs_string_sprintf(&v9stat->extension, "%s %lu",
839 "HARDLINKCOUNT", (unsigned long)stbuf->st_nlink);
840 }
841
842 str = strrchr(name->data, '/');
843 if (str) {
844 str += 1;
845 } else {
846 str = name->data;
847 }
848
849 v9fs_string_sprintf(&v9stat->name, "%s", str);
850
851 v9stat->size = 61 +
852 v9fs_string_size(&v9stat->name) +
853 v9fs_string_size(&v9stat->uid) +
854 v9fs_string_size(&v9stat->gid) +
855 v9fs_string_size(&v9stat->muid) +
856 v9fs_string_size(&v9stat->extension);
857 return 0;
858 }
859
860 #define P9_STATS_MODE 0x00000001ULL
861 #define P9_STATS_NLINK 0x00000002ULL
862 #define P9_STATS_UID 0x00000004ULL
863 #define P9_STATS_GID 0x00000008ULL
864 #define P9_STATS_RDEV 0x00000010ULL
865 #define P9_STATS_ATIME 0x00000020ULL
866 #define P9_STATS_MTIME 0x00000040ULL
867 #define P9_STATS_CTIME 0x00000080ULL
868 #define P9_STATS_INO 0x00000100ULL
869 #define P9_STATS_SIZE 0x00000200ULL
870 #define P9_STATS_BLOCKS 0x00000400ULL
871
872 #define P9_STATS_BTIME 0x00000800ULL
873 #define P9_STATS_GEN 0x00001000ULL
874 #define P9_STATS_DATA_VERSION 0x00002000ULL
875
876 #define P9_STATS_BASIC 0x000007ffULL /* Mask for fields up to BLOCKS */
877 #define P9_STATS_ALL 0x00003fffULL /* Mask for All fields above */
878
879
880 static void stat_to_v9stat_dotl(V9fsState *s, const struct stat *stbuf,
881 V9fsStatDotl *v9lstat)
882 {
883 memset(v9lstat, 0, sizeof(*v9lstat));
884
885 v9lstat->st_mode = stbuf->st_mode;
886 v9lstat->st_nlink = stbuf->st_nlink;
887 v9lstat->st_uid = stbuf->st_uid;
888 v9lstat->st_gid = stbuf->st_gid;
889 v9lstat->st_rdev = stbuf->st_rdev;
890 v9lstat->st_size = stbuf->st_size;
891 v9lstat->st_blksize = stbuf->st_blksize;
892 v9lstat->st_blocks = stbuf->st_blocks;
893 v9lstat->st_atime_sec = stbuf->st_atime;
894 v9lstat->st_atime_nsec = stbuf->st_atim.tv_nsec;
895 v9lstat->st_mtime_sec = stbuf->st_mtime;
896 v9lstat->st_mtime_nsec = stbuf->st_mtim.tv_nsec;
897 v9lstat->st_ctime_sec = stbuf->st_ctime;
898 v9lstat->st_ctime_nsec = stbuf->st_ctim.tv_nsec;
899 /* Currently we only support BASIC fields in stat */
900 v9lstat->st_result_mask = P9_STATS_BASIC;
901
902 stat_to_qid(stbuf, &v9lstat->qid);
903 }
904
905 static void print_sg(struct iovec *sg, int cnt)
906 {
907 int i;
908
909 printf("sg[%d]: {", cnt);
910 for (i = 0; i < cnt; i++) {
911 if (i) {
912 printf(", ");
913 }
914 printf("(%p, %zd)", sg[i].iov_base, sg[i].iov_len);
915 }
916 printf("}\n");
917 }
918
919 /* Will call this only for path name based fid */
920 static void v9fs_fix_path(V9fsPath *dst, V9fsPath *src, int len)
921 {
922 V9fsPath str;
923 v9fs_path_init(&str);
924 v9fs_path_copy(&str, dst);
925 v9fs_path_sprintf(dst, "%s%s", src->data, str.data + len);
926 v9fs_path_free(&str);
927 }
928
929 static inline bool is_ro_export(FsContext *ctx)
930 {
931 return ctx->export_flags & V9FS_RDONLY;
932 }
933
934 static void coroutine_fn v9fs_version(void *opaque)
935 {
936 ssize_t err;
937 V9fsPDU *pdu = opaque;
938 V9fsState *s = pdu->s;
939 V9fsString version;
940 size_t offset = 7;
941
942 v9fs_string_init(&version);
943 err = pdu_unmarshal(pdu, offset, "ds", &s->msize, &version);
944 if (err < 0) {
945 offset = err;
946 goto out;
947 }
948 trace_v9fs_version(pdu->tag, pdu->id, s->msize, version.data);
949
950 virtfs_reset(pdu);
951
952 if (!strcmp(version.data, "9P2000.u")) {
953 s->proto_version = V9FS_PROTO_2000U;
954 } else if (!strcmp(version.data, "9P2000.L")) {
955 s->proto_version = V9FS_PROTO_2000L;
956 } else {
957 v9fs_string_sprintf(&version, "unknown");
958 }
959
960 err = pdu_marshal(pdu, offset, "ds", s->msize, &version);
961 if (err < 0) {
962 offset = err;
963 goto out;
964 }
965 offset += err;
966 trace_v9fs_version_return(pdu->tag, pdu->id, s->msize, version.data);
967 out:
968 pdu_complete(pdu, offset);
969 v9fs_string_free(&version);
970 }
971
972 static void coroutine_fn v9fs_attach(void *opaque)
973 {
974 V9fsPDU *pdu = opaque;
975 V9fsState *s = pdu->s;
976 int32_t fid, afid, n_uname;
977 V9fsString uname, aname;
978 V9fsFidState *fidp;
979 size_t offset = 7;
980 V9fsQID qid;
981 ssize_t err;
982
983 v9fs_string_init(&uname);
984 v9fs_string_init(&aname);
985 err = pdu_unmarshal(pdu, offset, "ddssd", &fid,
986 &afid, &uname, &aname, &n_uname);
987 if (err < 0) {
988 goto out_nofid;
989 }
990 trace_v9fs_attach(pdu->tag, pdu->id, fid, afid, uname.data, aname.data);
991
992 fidp = alloc_fid(s, fid);
993 if (fidp == NULL) {
994 err = -EINVAL;
995 goto out_nofid;
996 }
997 fidp->uid = n_uname;
998 err = v9fs_co_name_to_path(pdu, NULL, "/", &fidp->path);
999 if (err < 0) {
1000 err = -EINVAL;
1001 clunk_fid(s, fid);
1002 goto out;
1003 }
1004 err = fid_to_qid(pdu, fidp, &qid);
1005 if (err < 0) {
1006 err = -EINVAL;
1007 clunk_fid(s, fid);
1008 goto out;
1009 }
1010 err = pdu_marshal(pdu, offset, "Q", &qid);
1011 if (err < 0) {
1012 clunk_fid(s, fid);
1013 goto out;
1014 }
1015 err += offset;
1016 memcpy(&s->root_qid, &qid, sizeof(qid));
1017 trace_v9fs_attach_return(pdu->tag, pdu->id,
1018 qid.type, qid.version, qid.path);
1019 /*
1020 * disable migration if we haven't done already.
1021 * attach could get called multiple times for the same export.
1022 */
1023 if (!s->migration_blocker) {
1024 s->root_fid = fid;
1025 error_setg(&s->migration_blocker,
1026 "Migration is disabled when VirtFS export path '%s' is mounted in the guest using mount_tag '%s'",
1027 s->ctx.fs_root ? s->ctx.fs_root : "NULL", s->tag);
1028 migrate_add_blocker(s->migration_blocker);
1029 }
1030 out:
1031 put_fid(pdu, fidp);
1032 out_nofid:
1033 pdu_complete(pdu, err);
1034 v9fs_string_free(&uname);
1035 v9fs_string_free(&aname);
1036 }
1037
1038 static void coroutine_fn v9fs_stat(void *opaque)
1039 {
1040 int32_t fid;
1041 V9fsStat v9stat;
1042 ssize_t err = 0;
1043 size_t offset = 7;
1044 struct stat stbuf;
1045 V9fsFidState *fidp;
1046 V9fsPDU *pdu = opaque;
1047
1048 err = pdu_unmarshal(pdu, offset, "d", &fid);
1049 if (err < 0) {
1050 goto out_nofid;
1051 }
1052 trace_v9fs_stat(pdu->tag, pdu->id, fid);
1053
1054 fidp = get_fid(pdu, fid);
1055 if (fidp == NULL) {
1056 err = -ENOENT;
1057 goto out_nofid;
1058 }
1059 err = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
1060 if (err < 0) {
1061 goto out;
1062 }
1063 err = stat_to_v9stat(pdu, &fidp->path, &stbuf, &v9stat);
1064 if (err < 0) {
1065 goto out;
1066 }
1067 err = pdu_marshal(pdu, offset, "wS", 0, &v9stat);
1068 if (err < 0) {
1069 v9fs_stat_free(&v9stat);
1070 goto out;
1071 }
1072 trace_v9fs_stat_return(pdu->tag, pdu->id, v9stat.mode,
1073 v9stat.atime, v9stat.mtime, v9stat.length);
1074 err += offset;
1075 v9fs_stat_free(&v9stat);
1076 out:
1077 put_fid(pdu, fidp);
1078 out_nofid:
1079 pdu_complete(pdu, err);
1080 }
1081
1082 static void coroutine_fn v9fs_getattr(void *opaque)
1083 {
1084 int32_t fid;
1085 size_t offset = 7;
1086 ssize_t retval = 0;
1087 struct stat stbuf;
1088 V9fsFidState *fidp;
1089 uint64_t request_mask;
1090 V9fsStatDotl v9stat_dotl;
1091 V9fsPDU *pdu = opaque;
1092 V9fsState *s = pdu->s;
1093
1094 retval = pdu_unmarshal(pdu, offset, "dq", &fid, &request_mask);
1095 if (retval < 0) {
1096 goto out_nofid;
1097 }
1098 trace_v9fs_getattr(pdu->tag, pdu->id, fid, request_mask);
1099
1100 fidp = get_fid(pdu, fid);
1101 if (fidp == NULL) {
1102 retval = -ENOENT;
1103 goto out_nofid;
1104 }
1105 /*
1106 * Currently we only support BASIC fields in stat, so there is no
1107 * need to look at request_mask.
1108 */
1109 retval = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
1110 if (retval < 0) {
1111 goto out;
1112 }
1113 stat_to_v9stat_dotl(s, &stbuf, &v9stat_dotl);
1114
1115 /* fill st_gen if requested and supported by underlying fs */
1116 if (request_mask & P9_STATS_GEN) {
1117 retval = v9fs_co_st_gen(pdu, &fidp->path, stbuf.st_mode, &v9stat_dotl);
1118 switch (retval) {
1119 case 0:
1120 /* we have valid st_gen: update result mask */
1121 v9stat_dotl.st_result_mask |= P9_STATS_GEN;
1122 break;
1123 case -EINTR:
1124 /* request cancelled, e.g. by Tflush */
1125 goto out;
1126 default:
1127 /* failed to get st_gen: not fatal, ignore */
1128 break;
1129 }
1130 }
1131 retval = pdu_marshal(pdu, offset, "A", &v9stat_dotl);
1132 if (retval < 0) {
1133 goto out;
1134 }
1135 retval += offset;
1136 trace_v9fs_getattr_return(pdu->tag, pdu->id, v9stat_dotl.st_result_mask,
1137 v9stat_dotl.st_mode, v9stat_dotl.st_uid,
1138 v9stat_dotl.st_gid);
1139 out:
1140 put_fid(pdu, fidp);
1141 out_nofid:
1142 pdu_complete(pdu, retval);
1143 }
1144
1145 /* Attribute flags */
1146 #define P9_ATTR_MODE (1 << 0)
1147 #define P9_ATTR_UID (1 << 1)
1148 #define P9_ATTR_GID (1 << 2)
1149 #define P9_ATTR_SIZE (1 << 3)
1150 #define P9_ATTR_ATIME (1 << 4)
1151 #define P9_ATTR_MTIME (1 << 5)
1152 #define P9_ATTR_CTIME (1 << 6)
1153 #define P9_ATTR_ATIME_SET (1 << 7)
1154 #define P9_ATTR_MTIME_SET (1 << 8)
1155
1156 #define P9_ATTR_MASK 127
1157
1158 static void coroutine_fn v9fs_setattr(void *opaque)
1159 {
1160 int err = 0;
1161 int32_t fid;
1162 V9fsFidState *fidp;
1163 size_t offset = 7;
1164 V9fsIattr v9iattr;
1165 V9fsPDU *pdu = opaque;
1166
1167 err = pdu_unmarshal(pdu, offset, "dI", &fid, &v9iattr);
1168 if (err < 0) {
1169 goto out_nofid;
1170 }
1171
1172 fidp = get_fid(pdu, fid);
1173 if (fidp == NULL) {
1174 err = -EINVAL;
1175 goto out_nofid;
1176 }
1177 if (v9iattr.valid & P9_ATTR_MODE) {
1178 err = v9fs_co_chmod(pdu, &fidp->path, v9iattr.mode);
1179 if (err < 0) {
1180 goto out;
1181 }
1182 }
1183 if (v9iattr.valid & (P9_ATTR_ATIME | P9_ATTR_MTIME)) {
1184 struct timespec times[2];
1185 if (v9iattr.valid & P9_ATTR_ATIME) {
1186 if (v9iattr.valid & P9_ATTR_ATIME_SET) {
1187 times[0].tv_sec = v9iattr.atime_sec;
1188 times[0].tv_nsec = v9iattr.atime_nsec;
1189 } else {
1190 times[0].tv_nsec = UTIME_NOW;
1191 }
1192 } else {
1193 times[0].tv_nsec = UTIME_OMIT;
1194 }
1195 if (v9iattr.valid & P9_ATTR_MTIME) {
1196 if (v9iattr.valid & P9_ATTR_MTIME_SET) {
1197 times[1].tv_sec = v9iattr.mtime_sec;
1198 times[1].tv_nsec = v9iattr.mtime_nsec;
1199 } else {
1200 times[1].tv_nsec = UTIME_NOW;
1201 }
1202 } else {
1203 times[1].tv_nsec = UTIME_OMIT;
1204 }
1205 err = v9fs_co_utimensat(pdu, &fidp->path, times);
1206 if (err < 0) {
1207 goto out;
1208 }
1209 }
1210 /*
1211 * If the only valid entry in iattr is ctime we can call
1212 * chown(-1,-1) to update the ctime of the file
1213 */
1214 if ((v9iattr.valid & (P9_ATTR_UID | P9_ATTR_GID)) ||
1215 ((v9iattr.valid & P9_ATTR_CTIME)
1216 && !((v9iattr.valid & P9_ATTR_MASK) & ~P9_ATTR_CTIME))) {
1217 if (!(v9iattr.valid & P9_ATTR_UID)) {
1218 v9iattr.uid = -1;
1219 }
1220 if (!(v9iattr.valid & P9_ATTR_GID)) {
1221 v9iattr.gid = -1;
1222 }
1223 err = v9fs_co_chown(pdu, &fidp->path, v9iattr.uid,
1224 v9iattr.gid);
1225 if (err < 0) {
1226 goto out;
1227 }
1228 }
1229 if (v9iattr.valid & (P9_ATTR_SIZE)) {
1230 err = v9fs_co_truncate(pdu, &fidp->path, v9iattr.size);
1231 if (err < 0) {
1232 goto out;
1233 }
1234 }
1235 err = offset;
1236 out:
1237 put_fid(pdu, fidp);
1238 out_nofid:
1239 pdu_complete(pdu, err);
1240 }
1241
1242 static int v9fs_walk_marshal(V9fsPDU *pdu, uint16_t nwnames, V9fsQID *qids)
1243 {
1244 int i;
1245 ssize_t err;
1246 size_t offset = 7;
1247
1248 err = pdu_marshal(pdu, offset, "w", nwnames);
1249 if (err < 0) {
1250 return err;
1251 }
1252 offset += err;
1253 for (i = 0; i < nwnames; i++) {
1254 err = pdu_marshal(pdu, offset, "Q", &qids[i]);
1255 if (err < 0) {
1256 return err;
1257 }
1258 offset += err;
1259 }
1260 return offset;
1261 }
1262
1263 static bool name_is_illegal(const char *name)
1264 {
1265 return !*name || strchr(name, '/') != NULL;
1266 }
1267
1268 static bool not_same_qid(const V9fsQID *qid1, const V9fsQID *qid2)
1269 {
1270 return
1271 qid1->type != qid2->type ||
1272 qid1->version != qid2->version ||
1273 qid1->path != qid2->path;
1274 }
1275
1276 static void coroutine_fn v9fs_walk(void *opaque)
1277 {
1278 int name_idx;
1279 V9fsQID *qids = NULL;
1280 int i, err = 0;
1281 V9fsPath dpath, path;
1282 uint16_t nwnames;
1283 struct stat stbuf;
1284 size_t offset = 7;
1285 int32_t fid, newfid;
1286 V9fsString *wnames = NULL;
1287 V9fsFidState *fidp;
1288 V9fsFidState *newfidp = NULL;
1289 V9fsPDU *pdu = opaque;
1290 V9fsState *s = pdu->s;
1291 V9fsQID qid;
1292
1293 err = pdu_unmarshal(pdu, offset, "ddw", &fid, &newfid, &nwnames);
1294 if (err < 0) {
1295 pdu_complete(pdu, err);
1296 return ;
1297 }
1298 offset += err;
1299
1300 trace_v9fs_walk(pdu->tag, pdu->id, fid, newfid, nwnames);
1301
1302 if (nwnames && nwnames <= P9_MAXWELEM) {
1303 wnames = g_malloc0(sizeof(wnames[0]) * nwnames);
1304 qids = g_malloc0(sizeof(qids[0]) * nwnames);
1305 for (i = 0; i < nwnames; i++) {
1306 err = pdu_unmarshal(pdu, offset, "s", &wnames[i]);
1307 if (err < 0) {
1308 goto out_nofid;
1309 }
1310 if (name_is_illegal(wnames[i].data)) {
1311 err = -ENOENT;
1312 goto out_nofid;
1313 }
1314 offset += err;
1315 }
1316 } else if (nwnames > P9_MAXWELEM) {
1317 err = -EINVAL;
1318 goto out_nofid;
1319 }
1320 fidp = get_fid(pdu, fid);
1321 if (fidp == NULL) {
1322 err = -ENOENT;
1323 goto out_nofid;
1324 }
1325
1326 v9fs_path_init(&dpath);
1327 v9fs_path_init(&path);
1328
1329 err = fid_to_qid(pdu, fidp, &qid);
1330 if (err < 0) {
1331 goto out;
1332 }
1333
1334 /*
1335 * Both dpath and path initially poin to fidp.
1336 * Needed to handle request with nwnames == 0
1337 */
1338 v9fs_path_copy(&dpath, &fidp->path);
1339 v9fs_path_copy(&path, &fidp->path);
1340 for (name_idx = 0; name_idx < nwnames; name_idx++) {
1341 if (not_same_qid(&pdu->s->root_qid, &qid) ||
1342 strcmp("..", wnames[name_idx].data)) {
1343 err = v9fs_co_name_to_path(pdu, &dpath, wnames[name_idx].data,
1344 &path);
1345 if (err < 0) {
1346 goto out;
1347 }
1348
1349 err = v9fs_co_lstat(pdu, &path, &stbuf);
1350 if (err < 0) {
1351 goto out;
1352 }
1353 stat_to_qid(&stbuf, &qid);
1354 v9fs_path_copy(&dpath, &path);
1355 }
1356 memcpy(&qids[name_idx], &qid, sizeof(qid));
1357 }
1358 if (fid == newfid) {
1359 if (fidp->fid_type != P9_FID_NONE) {
1360 err = -EINVAL;
1361 goto out;
1362 }
1363 v9fs_path_copy(&fidp->path, &path);
1364 } else {
1365 newfidp = alloc_fid(s, newfid);
1366 if (newfidp == NULL) {
1367 err = -EINVAL;
1368 goto out;
1369 }
1370 newfidp->uid = fidp->uid;
1371 v9fs_path_copy(&newfidp->path, &path);
1372 }
1373 err = v9fs_walk_marshal(pdu, nwnames, qids);
1374 trace_v9fs_walk_return(pdu->tag, pdu->id, nwnames, qids);
1375 out:
1376 put_fid(pdu, fidp);
1377 if (newfidp) {
1378 put_fid(pdu, newfidp);
1379 }
1380 v9fs_path_free(&dpath);
1381 v9fs_path_free(&path);
1382 out_nofid:
1383 pdu_complete(pdu, err);
1384 if (nwnames && nwnames <= P9_MAXWELEM) {
1385 for (name_idx = 0; name_idx < nwnames; name_idx++) {
1386 v9fs_string_free(&wnames[name_idx]);
1387 }
1388 g_free(wnames);
1389 g_free(qids);
1390 }
1391 }
1392
1393 static int32_t coroutine_fn get_iounit(V9fsPDU *pdu, V9fsPath *path)
1394 {
1395 struct statfs stbuf;
1396 int32_t iounit = 0;
1397 V9fsState *s = pdu->s;
1398
1399 /*
1400 * iounit should be multiples of f_bsize (host filesystem block size
1401 * and as well as less than (client msize - P9_IOHDRSZ))
1402 */
1403 if (!v9fs_co_statfs(pdu, path, &stbuf)) {
1404 iounit = stbuf.f_bsize;
1405 iounit *= (s->msize - P9_IOHDRSZ)/stbuf.f_bsize;
1406 }
1407 if (!iounit) {
1408 iounit = s->msize - P9_IOHDRSZ;
1409 }
1410 return iounit;
1411 }
1412
1413 static void coroutine_fn v9fs_open(void *opaque)
1414 {
1415 int flags;
1416 int32_t fid;
1417 int32_t mode;
1418 V9fsQID qid;
1419 int iounit = 0;
1420 ssize_t err = 0;
1421 size_t offset = 7;
1422 struct stat stbuf;
1423 V9fsFidState *fidp;
1424 V9fsPDU *pdu = opaque;
1425 V9fsState *s = pdu->s;
1426
1427 if (s->proto_version == V9FS_PROTO_2000L) {
1428 err = pdu_unmarshal(pdu, offset, "dd", &fid, &mode);
1429 } else {
1430 uint8_t modebyte;
1431 err = pdu_unmarshal(pdu, offset, "db", &fid, &modebyte);
1432 mode = modebyte;
1433 }
1434 if (err < 0) {
1435 goto out_nofid;
1436 }
1437 trace_v9fs_open(pdu->tag, pdu->id, fid, mode);
1438
1439 fidp = get_fid(pdu, fid);
1440 if (fidp == NULL) {
1441 err = -ENOENT;
1442 goto out_nofid;
1443 }
1444 if (fidp->fid_type != P9_FID_NONE) {
1445 err = -EINVAL;
1446 goto out;
1447 }
1448
1449 err = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
1450 if (err < 0) {
1451 goto out;
1452 }
1453 stat_to_qid(&stbuf, &qid);
1454 if (S_ISDIR(stbuf.st_mode)) {
1455 err = v9fs_co_opendir(pdu, fidp);
1456 if (err < 0) {
1457 goto out;
1458 }
1459 fidp->fid_type = P9_FID_DIR;
1460 err = pdu_marshal(pdu, offset, "Qd", &qid, 0);
1461 if (err < 0) {
1462 goto out;
1463 }
1464 err += offset;
1465 } else {
1466 if (s->proto_version == V9FS_PROTO_2000L) {
1467 flags = get_dotl_openflags(s, mode);
1468 } else {
1469 flags = omode_to_uflags(mode);
1470 }
1471 if (is_ro_export(&s->ctx)) {
1472 if (mode & O_WRONLY || mode & O_RDWR ||
1473 mode & O_APPEND || mode & O_TRUNC) {
1474 err = -EROFS;
1475 goto out;
1476 }
1477 }
1478 err = v9fs_co_open(pdu, fidp, flags);
1479 if (err < 0) {
1480 goto out;
1481 }
1482 fidp->fid_type = P9_FID_FILE;
1483 fidp->open_flags = flags;
1484 if (flags & O_EXCL) {
1485 /*
1486 * We let the host file system do O_EXCL check
1487 * We should not reclaim such fd
1488 */
1489 fidp->flags |= FID_NON_RECLAIMABLE;
1490 }
1491 iounit = get_iounit(pdu, &fidp->path);
1492 err = pdu_marshal(pdu, offset, "Qd", &qid, iounit);
1493 if (err < 0) {
1494 goto out;
1495 }
1496 err += offset;
1497 }
1498 trace_v9fs_open_return(pdu->tag, pdu->id,
1499 qid.type, qid.version, qid.path, iounit);
1500 out:
1501 put_fid(pdu, fidp);
1502 out_nofid:
1503 pdu_complete(pdu, err);
1504 }
1505
1506 static void coroutine_fn v9fs_lcreate(void *opaque)
1507 {
1508 int32_t dfid, flags, mode;
1509 gid_t gid;
1510 ssize_t err = 0;
1511 ssize_t offset = 7;
1512 V9fsString name;
1513 V9fsFidState *fidp;
1514 struct stat stbuf;
1515 V9fsQID qid;
1516 int32_t iounit;
1517 V9fsPDU *pdu = opaque;
1518
1519 v9fs_string_init(&name);
1520 err = pdu_unmarshal(pdu, offset, "dsddd", &dfid,
1521 &name, &flags, &mode, &gid);
1522 if (err < 0) {
1523 goto out_nofid;
1524 }
1525 trace_v9fs_lcreate(pdu->tag, pdu->id, dfid, flags, mode, gid);
1526
1527 if (name_is_illegal(name.data)) {
1528 err = -ENOENT;
1529 goto out_nofid;
1530 }
1531
1532 if (!strcmp(".", name.data) || !strcmp("..", name.data)) {
1533 err = -EEXIST;
1534 goto out_nofid;
1535 }
1536
1537 fidp = get_fid(pdu, dfid);
1538 if (fidp == NULL) {
1539 err = -ENOENT;
1540 goto out_nofid;
1541 }
1542
1543 flags = get_dotl_openflags(pdu->s, flags);
1544 err = v9fs_co_open2(pdu, fidp, &name, gid,
1545 flags | O_CREAT, mode, &stbuf);
1546 if (err < 0) {
1547 goto out;
1548 }
1549 fidp->fid_type = P9_FID_FILE;
1550 fidp->open_flags = flags;
1551 if (flags & O_EXCL) {
1552 /*
1553 * We let the host file system do O_EXCL check
1554 * We should not reclaim such fd
1555 */
1556 fidp->flags |= FID_NON_RECLAIMABLE;
1557 }
1558 iounit = get_iounit(pdu, &fidp->path);
1559 stat_to_qid(&stbuf, &qid);
1560 err = pdu_marshal(pdu, offset, "Qd", &qid, iounit);
1561 if (err < 0) {
1562 goto out;
1563 }
1564 err += offset;
1565 trace_v9fs_lcreate_return(pdu->tag, pdu->id,
1566 qid.type, qid.version, qid.path, iounit);
1567 out:
1568 put_fid(pdu, fidp);
1569 out_nofid:
1570 pdu_complete(pdu, err);
1571 v9fs_string_free(&name);
1572 }
1573
1574 static void v9fs_fsync(void *opaque)
1575 {
1576 int err;
1577 int32_t fid;
1578 int datasync;
1579 size_t offset = 7;
1580 V9fsFidState *fidp;
1581 V9fsPDU *pdu = opaque;
1582
1583 err = pdu_unmarshal(pdu, offset, "dd", &fid, &datasync);
1584 if (err < 0) {
1585 goto out_nofid;
1586 }
1587 trace_v9fs_fsync(pdu->tag, pdu->id, fid, datasync);
1588
1589 fidp = get_fid(pdu, fid);
1590 if (fidp == NULL) {
1591 err = -ENOENT;
1592 goto out_nofid;
1593 }
1594 err = v9fs_co_fsync(pdu, fidp, datasync);
1595 if (!err) {
1596 err = offset;
1597 }
1598 put_fid(pdu, fidp);
1599 out_nofid:
1600 pdu_complete(pdu, err);
1601 }
1602
1603 static void coroutine_fn v9fs_clunk(void *opaque)
1604 {
1605 int err;
1606 int32_t fid;
1607 size_t offset = 7;
1608 V9fsFidState *fidp;
1609 V9fsPDU *pdu = opaque;
1610 V9fsState *s = pdu->s;
1611
1612 err = pdu_unmarshal(pdu, offset, "d", &fid);
1613 if (err < 0) {
1614 goto out_nofid;
1615 }
1616 trace_v9fs_clunk(pdu->tag, pdu->id, fid);
1617
1618 fidp = clunk_fid(s, fid);
1619 if (fidp == NULL) {
1620 err = -ENOENT;
1621 goto out_nofid;
1622 }
1623 /*
1624 * Bump the ref so that put_fid will
1625 * free the fid.
1626 */
1627 fidp->ref++;
1628 err = put_fid(pdu, fidp);
1629 if (!err) {
1630 err = offset;
1631 }
1632 out_nofid:
1633 pdu_complete(pdu, err);
1634 }
1635
1636 /*
1637 * Create a QEMUIOVector for a sub-region of PDU iovecs
1638 *
1639 * @qiov: uninitialized QEMUIOVector
1640 * @skip: number of bytes to skip from beginning of PDU
1641 * @size: number of bytes to include
1642 * @is_write: true - write, false - read
1643 *
1644 * The resulting QEMUIOVector has heap-allocated iovecs and must be cleaned up
1645 * with qemu_iovec_destroy().
1646 */
1647 static void v9fs_init_qiov_from_pdu(QEMUIOVector *qiov, V9fsPDU *pdu,
1648 size_t skip, size_t size,
1649 bool is_write)
1650 {
1651 QEMUIOVector elem;
1652 struct iovec *iov;
1653 unsigned int niov;
1654
1655 if (is_write) {
1656 pdu->s->transport->init_out_iov_from_pdu(pdu, &iov, &niov);
1657 } else {
1658 pdu->s->transport->init_in_iov_from_pdu(pdu, &iov, &niov, size);
1659 }
1660
1661 qemu_iovec_init_external(&elem, iov, niov);
1662 qemu_iovec_init(qiov, niov);
1663 qemu_iovec_concat(qiov, &elem, skip, size);
1664 }
1665
1666 static int v9fs_xattr_read(V9fsState *s, V9fsPDU *pdu, V9fsFidState *fidp,
1667 uint64_t off, uint32_t max_count)
1668 {
1669 ssize_t err;
1670 size_t offset = 7;
1671 uint64_t read_count;
1672 QEMUIOVector qiov_full;
1673
1674 if (fidp->fs.xattr.len < off) {
1675 read_count = 0;
1676 } else {
1677 read_count = fidp->fs.xattr.len - off;
1678 }
1679 if (read_count > max_count) {
1680 read_count = max_count;
1681 }
1682 err = pdu_marshal(pdu, offset, "d", read_count);
1683 if (err < 0) {
1684 return err;
1685 }
1686 offset += err;
1687
1688 v9fs_init_qiov_from_pdu(&qiov_full, pdu, 0, read_count, false);
1689 err = v9fs_pack(qiov_full.iov, qiov_full.niov, offset,
1690 ((char *)fidp->fs.xattr.value) + off,
1691 read_count);
1692 qemu_iovec_destroy(&qiov_full);
1693 if (err < 0) {
1694 return err;
1695 }
1696 offset += err;
1697 return offset;
1698 }
1699
1700 static int coroutine_fn v9fs_do_readdir_with_stat(V9fsPDU *pdu,
1701 V9fsFidState *fidp,
1702 uint32_t max_count)
1703 {
1704 V9fsPath path;
1705 V9fsStat v9stat;
1706 int len, err = 0;
1707 int32_t count = 0;
1708 struct stat stbuf;
1709 off_t saved_dir_pos;
1710 struct dirent *dent;
1711
1712 /* save the directory position */
1713 saved_dir_pos = v9fs_co_telldir(pdu, fidp);
1714 if (saved_dir_pos < 0) {
1715 return saved_dir_pos;
1716 }
1717
1718 while (1) {
1719 v9fs_path_init(&path);
1720
1721 v9fs_readdir_lock(&fidp->fs.dir);
1722
1723 err = v9fs_co_readdir(pdu, fidp, &dent);
1724 if (err || !dent) {
1725 break;
1726 }
1727 err = v9fs_co_name_to_path(pdu, &fidp->path, dent->d_name, &path);
1728 if (err < 0) {
1729 break;
1730 }
1731 err = v9fs_co_lstat(pdu, &path, &stbuf);
1732 if (err < 0) {
1733 break;
1734 }
1735 err = stat_to_v9stat(pdu, &path, &stbuf, &v9stat);
1736 if (err < 0) {
1737 break;
1738 }
1739 /* 11 = 7 + 4 (7 = start offset, 4 = space for storing count) */
1740 len = pdu_marshal(pdu, 11 + count, "S", &v9stat);
1741
1742 v9fs_readdir_unlock(&fidp->fs.dir);
1743
1744 if ((len != (v9stat.size + 2)) || ((count + len) > max_count)) {
1745 /* Ran out of buffer. Set dir back to old position and return */
1746 v9fs_co_seekdir(pdu, fidp, saved_dir_pos);
1747 v9fs_stat_free(&v9stat);
1748 v9fs_path_free(&path);
1749 return count;
1750 }
1751 count += len;
1752 v9fs_stat_free(&v9stat);
1753 v9fs_path_free(&path);
1754 saved_dir_pos = dent->d_off;
1755 }
1756
1757 v9fs_readdir_unlock(&fidp->fs.dir);
1758
1759 v9fs_path_free(&path);
1760 if (err < 0) {
1761 return err;
1762 }
1763 return count;
1764 }
1765
1766 static void coroutine_fn v9fs_read(void *opaque)
1767 {
1768 int32_t fid;
1769 uint64_t off;
1770 ssize_t err = 0;
1771 int32_t count = 0;
1772 size_t offset = 7;
1773 uint32_t max_count;
1774 V9fsFidState *fidp;
1775 V9fsPDU *pdu = opaque;
1776 V9fsState *s = pdu->s;
1777
1778 err = pdu_unmarshal(pdu, offset, "dqd", &fid, &off, &max_count);
1779 if (err < 0) {
1780 goto out_nofid;
1781 }
1782 trace_v9fs_read(pdu->tag, pdu->id, fid, off, max_count);
1783
1784 fidp = get_fid(pdu, fid);
1785 if (fidp == NULL) {
1786 err = -EINVAL;
1787 goto out_nofid;
1788 }
1789 if (fidp->fid_type == P9_FID_DIR) {
1790
1791 if (off == 0) {
1792 v9fs_co_rewinddir(pdu, fidp);
1793 }
1794 count = v9fs_do_readdir_with_stat(pdu, fidp, max_count);
1795 if (count < 0) {
1796 err = count;
1797 goto out;
1798 }
1799 err = pdu_marshal(pdu, offset, "d", count);
1800 if (err < 0) {
1801 goto out;
1802 }
1803 err += offset + count;
1804 } else if (fidp->fid_type == P9_FID_FILE) {
1805 QEMUIOVector qiov_full;
1806 QEMUIOVector qiov;
1807 int32_t len;
1808
1809 v9fs_init_qiov_from_pdu(&qiov_full, pdu, offset + 4, max_count, false);
1810 qemu_iovec_init(&qiov, qiov_full.niov);
1811 do {
1812 qemu_iovec_reset(&qiov);
1813 qemu_iovec_concat(&qiov, &qiov_full, count, qiov_full.size - count);
1814 if (0) {
1815 print_sg(qiov.iov, qiov.niov);
1816 }
1817 /* Loop in case of EINTR */
1818 do {
1819 len = v9fs_co_preadv(pdu, fidp, qiov.iov, qiov.niov, off);
1820 if (len >= 0) {
1821 off += len;
1822 count += len;
1823 }
1824 } while (len == -EINTR && !pdu->cancelled);
1825 if (len < 0) {
1826 /* IO error return the error */
1827 err = len;
1828 goto out_free_iovec;
1829 }
1830 } while (count < max_count && len > 0);
1831 err = pdu_marshal(pdu, offset, "d", count);
1832 if (err < 0) {
1833 goto out_free_iovec;
1834 }
1835 err += offset + count;
1836 out_free_iovec:
1837 qemu_iovec_destroy(&qiov);
1838 qemu_iovec_destroy(&qiov_full);
1839 } else if (fidp->fid_type == P9_FID_XATTR) {
1840 err = v9fs_xattr_read(s, pdu, fidp, off, max_count);
1841 } else {
1842 err = -EINVAL;
1843 }
1844 trace_v9fs_read_return(pdu->tag, pdu->id, count, err);
1845 out:
1846 put_fid(pdu, fidp);
1847 out_nofid:
1848 pdu_complete(pdu, err);
1849 }
1850
1851 static size_t v9fs_readdir_data_size(V9fsString *name)
1852 {
1853 /*
1854 * Size of each dirent on the wire: size of qid (13) + size of offset (8)
1855 * size of type (1) + size of name.size (2) + strlen(name.data)
1856 */
1857 return 24 + v9fs_string_size(name);
1858 }
1859
1860 static int coroutine_fn v9fs_do_readdir(V9fsPDU *pdu, V9fsFidState *fidp,
1861 int32_t max_count)
1862 {
1863 size_t size;
1864 V9fsQID qid;
1865 V9fsString name;
1866 int len, err = 0;
1867 int32_t count = 0;
1868 off_t saved_dir_pos;
1869 struct dirent *dent;
1870
1871 /* save the directory position */
1872 saved_dir_pos = v9fs_co_telldir(pdu, fidp);
1873 if (saved_dir_pos < 0) {
1874 return saved_dir_pos;
1875 }
1876
1877 while (1) {
1878 v9fs_readdir_lock(&fidp->fs.dir);
1879
1880 err = v9fs_co_readdir(pdu, fidp, &dent);
1881 if (err || !dent) {
1882 break;
1883 }
1884 v9fs_string_init(&name);
1885 v9fs_string_sprintf(&name, "%s", dent->d_name);
1886 if ((count + v9fs_readdir_data_size(&name)) > max_count) {
1887 v9fs_readdir_unlock(&fidp->fs.dir);
1888
1889 /* Ran out of buffer. Set dir back to old position and return */
1890 v9fs_co_seekdir(pdu, fidp, saved_dir_pos);
1891 v9fs_string_free(&name);
1892 return count;
1893 }
1894 /*
1895 * Fill up just the path field of qid because the client uses
1896 * only that. To fill the entire qid structure we will have
1897 * to stat each dirent found, which is expensive
1898 */
1899 size = MIN(sizeof(dent->d_ino), sizeof(qid.path));
1900 memcpy(&qid.path, &dent->d_ino, size);
1901 /* Fill the other fields with dummy values */
1902 qid.type = 0;
1903 qid.version = 0;
1904
1905 /* 11 = 7 + 4 (7 = start offset, 4 = space for storing count) */
1906 len = pdu_marshal(pdu, 11 + count, "Qqbs",
1907 &qid, dent->d_off,
1908 dent->d_type, &name);
1909
1910 v9fs_readdir_unlock(&fidp->fs.dir);
1911
1912 if (len < 0) {
1913 v9fs_co_seekdir(pdu, fidp, saved_dir_pos);
1914 v9fs_string_free(&name);
1915 return len;
1916 }
1917 count += len;
1918 v9fs_string_free(&name);
1919 saved_dir_pos = dent->d_off;
1920 }
1921
1922 v9fs_readdir_unlock(&fidp->fs.dir);
1923
1924 if (err < 0) {
1925 return err;
1926 }
1927 return count;
1928 }
1929
1930 static void coroutine_fn v9fs_readdir(void *opaque)
1931 {
1932 int32_t fid;
1933 V9fsFidState *fidp;
1934 ssize_t retval = 0;
1935 size_t offset = 7;
1936 uint64_t initial_offset;
1937 int32_t count;
1938 uint32_t max_count;
1939 V9fsPDU *pdu = opaque;
1940
1941 retval = pdu_unmarshal(pdu, offset, "dqd", &fid,
1942 &initial_offset, &max_count);
1943 if (retval < 0) {
1944 goto out_nofid;
1945 }
1946 trace_v9fs_readdir(pdu->tag, pdu->id, fid, initial_offset, max_count);
1947
1948 fidp = get_fid(pdu, fid);
1949 if (fidp == NULL) {
1950 retval = -EINVAL;
1951 goto out_nofid;
1952 }
1953 if (!fidp->fs.dir.stream) {
1954 retval = -EINVAL;
1955 goto out;
1956 }
1957 if (initial_offset == 0) {
1958 v9fs_co_rewinddir(pdu, fidp);
1959 } else {
1960 v9fs_co_seekdir(pdu, fidp, initial_offset);
1961 }
1962 count = v9fs_do_readdir(pdu, fidp, max_count);
1963 if (count < 0) {
1964 retval = count;
1965 goto out;
1966 }
1967 retval = pdu_marshal(pdu, offset, "d", count);
1968 if (retval < 0) {
1969 goto out;
1970 }
1971 retval += count + offset;
1972 trace_v9fs_readdir_return(pdu->tag, pdu->id, count, retval);
1973 out:
1974 put_fid(pdu, fidp);
1975 out_nofid:
1976 pdu_complete(pdu, retval);
1977 }
1978
1979 static int v9fs_xattr_write(V9fsState *s, V9fsPDU *pdu, V9fsFidState *fidp,
1980 uint64_t off, uint32_t count,
1981 struct iovec *sg, int cnt)
1982 {
1983 int i, to_copy;
1984 ssize_t err = 0;
1985 uint64_t write_count;
1986 size_t offset = 7;
1987
1988
1989 if (fidp->fs.xattr.len < off) {
1990 err = -ENOSPC;
1991 goto out;
1992 }
1993 write_count = fidp->fs.xattr.len - off;
1994 if (write_count > count) {
1995 write_count = count;
1996 }
1997 err = pdu_marshal(pdu, offset, "d", write_count);
1998 if (err < 0) {
1999 return err;
2000 }
2001 err += offset;
2002 fidp->fs.xattr.copied_len += write_count;
2003 /*
2004 * Now copy the content from sg list
2005 */
2006 for (i = 0; i < cnt; i++) {
2007 if (write_count > sg[i].iov_len) {
2008 to_copy = sg[i].iov_len;
2009 } else {
2010 to_copy = write_count;
2011 }
2012 memcpy((char *)fidp->fs.xattr.value + off, sg[i].iov_base, to_copy);
2013 /* updating vs->off since we are not using below */
2014 off += to_copy;
2015 write_count -= to_copy;
2016 }
2017 out:
2018 return err;
2019 }
2020
2021 static void coroutine_fn v9fs_write(void *opaque)
2022 {
2023 ssize_t err;
2024 int32_t fid;
2025 uint64_t off;
2026 uint32_t count;
2027 int32_t len = 0;
2028 int32_t total = 0;
2029 size_t offset = 7;
2030 V9fsFidState *fidp;
2031 V9fsPDU *pdu = opaque;
2032 V9fsState *s = pdu->s;
2033 QEMUIOVector qiov_full;
2034 QEMUIOVector qiov;
2035
2036 err = pdu_unmarshal(pdu, offset, "dqd", &fid, &off, &count);
2037 if (err < 0) {
2038 pdu_complete(pdu, err);
2039 return;
2040 }
2041 offset += err;
2042 v9fs_init_qiov_from_pdu(&qiov_full, pdu, offset, count, true);
2043 trace_v9fs_write(pdu->tag, pdu->id, fid, off, count, qiov_full.niov);
2044
2045 fidp = get_fid(pdu, fid);
2046 if (fidp == NULL) {
2047 err = -EINVAL;
2048 goto out_nofid;
2049 }
2050 if (fidp->fid_type == P9_FID_FILE) {
2051 if (fidp->fs.fd == -1) {
2052 err = -EINVAL;
2053 goto out;
2054 }
2055 } else if (fidp->fid_type == P9_FID_XATTR) {
2056 /*
2057 * setxattr operation
2058 */
2059 err = v9fs_xattr_write(s, pdu, fidp, off, count,
2060 qiov_full.iov, qiov_full.niov);
2061 goto out;
2062 } else {
2063 err = -EINVAL;
2064 goto out;
2065 }
2066 qemu_iovec_init(&qiov, qiov_full.niov);
2067 do {
2068 qemu_iovec_reset(&qiov);
2069 qemu_iovec_concat(&qiov, &qiov_full, total, qiov_full.size - total);
2070 if (0) {
2071 print_sg(qiov.iov, qiov.niov);
2072 }
2073 /* Loop in case of EINTR */
2074 do {
2075 len = v9fs_co_pwritev(pdu, fidp, qiov.iov, qiov.niov, off);
2076 if (len >= 0) {
2077 off += len;
2078 total += len;
2079 }
2080 } while (len == -EINTR && !pdu->cancelled);
2081 if (len < 0) {
2082 /* IO error return the error */
2083 err = len;
2084 goto out_qiov;
2085 }
2086 } while (total < count && len > 0);
2087
2088 offset = 7;
2089 err = pdu_marshal(pdu, offset, "d", total);
2090 if (err < 0) {
2091 goto out_qiov;
2092 }
2093 err += offset;
2094 trace_v9fs_write_return(pdu->tag, pdu->id, total, err);
2095 out_qiov:
2096 qemu_iovec_destroy(&qiov);
2097 out:
2098 put_fid(pdu, fidp);
2099 out_nofid:
2100 qemu_iovec_destroy(&qiov_full);
2101 pdu_complete(pdu, err);
2102 }
2103
2104 static void coroutine_fn v9fs_create(void *opaque)
2105 {
2106 int32_t fid;
2107 int err = 0;
2108 size_t offset = 7;
2109 V9fsFidState *fidp;
2110 V9fsQID qid;
2111 int32_t perm;
2112 int8_t mode;
2113 V9fsPath path;
2114 struct stat stbuf;
2115 V9fsString name;
2116 V9fsString extension;
2117 int iounit;
2118 V9fsPDU *pdu = opaque;
2119
2120 v9fs_path_init(&path);
2121 v9fs_string_init(&name);
2122 v9fs_string_init(&extension);
2123 err = pdu_unmarshal(pdu, offset, "dsdbs", &fid, &name,
2124 &perm, &mode, &extension);
2125 if (err < 0) {
2126 goto out_nofid;
2127 }
2128 trace_v9fs_create(pdu->tag, pdu->id, fid, name.data, perm, mode);
2129
2130 if (name_is_illegal(name.data)) {
2131 err = -ENOENT;
2132 goto out_nofid;
2133 }
2134
2135 if (!strcmp(".", name.data) || !strcmp("..", name.data)) {
2136 err = -EEXIST;
2137 goto out_nofid;
2138 }
2139
2140 fidp = get_fid(pdu, fid);
2141 if (fidp == NULL) {
2142 err = -EINVAL;
2143 goto out_nofid;
2144 }
2145 if (perm & P9_STAT_MODE_DIR) {
2146 err = v9fs_co_mkdir(pdu, fidp, &name, perm & 0777,
2147 fidp->uid, -1, &stbuf);
2148 if (err < 0) {
2149 goto out;
2150 }
2151 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2152 if (err < 0) {
2153 goto out;
2154 }
2155 v9fs_path_copy(&fidp->path, &path);
2156 err = v9fs_co_opendir(pdu, fidp);
2157 if (err < 0) {
2158 goto out;
2159 }
2160 fidp->fid_type = P9_FID_DIR;
2161 } else if (perm & P9_STAT_MODE_SYMLINK) {
2162 err = v9fs_co_symlink(pdu, fidp, &name,
2163 extension.data, -1 , &stbuf);
2164 if (err < 0) {
2165 goto out;
2166 }
2167 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2168 if (err < 0) {
2169 goto out;
2170 }
2171 v9fs_path_copy(&fidp->path, &path);
2172 } else if (perm & P9_STAT_MODE_LINK) {
2173 int32_t ofid = atoi(extension.data);
2174 V9fsFidState *ofidp = get_fid(pdu, ofid);
2175 if (ofidp == NULL) {
2176 err = -EINVAL;
2177 goto out;
2178 }
2179 err = v9fs_co_link(pdu, ofidp, fidp, &name);
2180 put_fid(pdu, ofidp);
2181 if (err < 0) {
2182 goto out;
2183 }
2184 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2185 if (err < 0) {
2186 fidp->fid_type = P9_FID_NONE;
2187 goto out;
2188 }
2189 v9fs_path_copy(&fidp->path, &path);
2190 err = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
2191 if (err < 0) {
2192 fidp->fid_type = P9_FID_NONE;
2193 goto out;
2194 }
2195 } else if (perm & P9_STAT_MODE_DEVICE) {
2196 char ctype;
2197 uint32_t major, minor;
2198 mode_t nmode = 0;
2199
2200 if (sscanf(extension.data, "%c %u %u", &ctype, &major, &minor) != 3) {
2201 err = -errno;
2202 goto out;
2203 }
2204
2205 switch (ctype) {
2206 case 'c':
2207 nmode = S_IFCHR;
2208 break;
2209 case 'b':
2210 nmode = S_IFBLK;
2211 break;
2212 default:
2213 err = -EIO;
2214 goto out;
2215 }
2216
2217 nmode |= perm & 0777;
2218 err = v9fs_co_mknod(pdu, fidp, &name, fidp->uid, -1,
2219 makedev(major, minor), nmode, &stbuf);
2220 if (err < 0) {
2221 goto out;
2222 }
2223 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2224 if (err < 0) {
2225 goto out;
2226 }
2227 v9fs_path_copy(&fidp->path, &path);
2228 } else if (perm & P9_STAT_MODE_NAMED_PIPE) {
2229 err = v9fs_co_mknod(pdu, fidp, &name, fidp->uid, -1,
2230 0, S_IFIFO | (perm & 0777), &stbuf);
2231 if (err < 0) {
2232 goto out;
2233 }
2234 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2235 if (err < 0) {
2236 goto out;
2237 }
2238 v9fs_path_copy(&fidp->path, &path);
2239 } else if (perm & P9_STAT_MODE_SOCKET) {
2240 err = v9fs_co_mknod(pdu, fidp, &name, fidp->uid, -1,
2241 0, S_IFSOCK | (perm & 0777), &stbuf);
2242 if (err < 0) {
2243 goto out;
2244 }
2245 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2246 if (err < 0) {
2247 goto out;
2248 }
2249 v9fs_path_copy(&fidp->path, &path);
2250 } else {
2251 err = v9fs_co_open2(pdu, fidp, &name, -1,
2252 omode_to_uflags(mode)|O_CREAT, perm, &stbuf);
2253 if (err < 0) {
2254 goto out;
2255 }
2256 fidp->fid_type = P9_FID_FILE;
2257 fidp->open_flags = omode_to_uflags(mode);
2258 if (fidp->open_flags & O_EXCL) {
2259 /*
2260 * We let the host file system do O_EXCL check
2261 * We should not reclaim such fd
2262 */
2263 fidp->flags |= FID_NON_RECLAIMABLE;
2264 }
2265 }
2266 iounit = get_iounit(pdu, &fidp->path);
2267 stat_to_qid(&stbuf, &qid);
2268 err = pdu_marshal(pdu, offset, "Qd", &qid, iounit);
2269 if (err < 0) {
2270 goto out;
2271 }
2272 err += offset;
2273 trace_v9fs_create_return(pdu->tag, pdu->id,
2274 qid.type, qid.version, qid.path, iounit);
2275 out:
2276 put_fid(pdu, fidp);
2277 out_nofid:
2278 pdu_complete(pdu, err);
2279 v9fs_string_free(&name);
2280 v9fs_string_free(&extension);
2281 v9fs_path_free(&path);
2282 }
2283
2284 static void coroutine_fn v9fs_symlink(void *opaque)
2285 {
2286 V9fsPDU *pdu = opaque;
2287 V9fsString name;
2288 V9fsString symname;
2289 V9fsFidState *dfidp;
2290 V9fsQID qid;
2291 struct stat stbuf;
2292 int32_t dfid;
2293 int err = 0;
2294 gid_t gid;
2295 size_t offset = 7;
2296
2297 v9fs_string_init(&name);
2298 v9fs_string_init(&symname);
2299 err = pdu_unmarshal(pdu, offset, "dssd", &dfid, &name, &symname, &gid);
2300 if (err < 0) {
2301 goto out_nofid;
2302 }
2303 trace_v9fs_symlink(pdu->tag, pdu->id, dfid, name.data, symname.data, gid);
2304
2305 if (name_is_illegal(name.data)) {
2306 err = -ENOENT;
2307 goto out_nofid;
2308 }
2309
2310 if (!strcmp(".", name.data) || !strcmp("..", name.data)) {
2311 err = -EEXIST;
2312 goto out_nofid;
2313 }
2314
2315 dfidp = get_fid(pdu, dfid);
2316 if (dfidp == NULL) {
2317 err = -EINVAL;
2318 goto out_nofid;
2319 }
2320 err = v9fs_co_symlink(pdu, dfidp, &name, symname.data, gid, &stbuf);
2321 if (err < 0) {
2322 goto out;
2323 }
2324 stat_to_qid(&stbuf, &qid);
2325 err = pdu_marshal(pdu, offset, "Q", &qid);
2326 if (err < 0) {
2327 goto out;
2328 }
2329 err += offset;
2330 trace_v9fs_symlink_return(pdu->tag, pdu->id,
2331 qid.type, qid.version, qid.path);
2332 out:
2333 put_fid(pdu, dfidp);
2334 out_nofid:
2335 pdu_complete(pdu, err);
2336 v9fs_string_free(&name);
2337 v9fs_string_free(&symname);
2338 }
2339
2340 static void v9fs_flush(void *opaque)
2341 {
2342 ssize_t err;
2343 int16_t tag;
2344 size_t offset = 7;
2345 V9fsPDU *cancel_pdu;
2346 V9fsPDU *pdu = opaque;
2347 V9fsState *s = pdu->s;
2348
2349 err = pdu_unmarshal(pdu, offset, "w", &tag);
2350 if (err < 0) {
2351 pdu_complete(pdu, err);
2352 return;
2353 }
2354 trace_v9fs_flush(pdu->tag, pdu->id, tag);
2355
2356 QLIST_FOREACH(cancel_pdu, &s->active_list, next) {
2357 if (cancel_pdu->tag == tag) {
2358 break;
2359 }
2360 }
2361 if (cancel_pdu) {
2362 cancel_pdu->cancelled = 1;
2363 /*
2364 * Wait for pdu to complete.
2365 */
2366 qemu_co_queue_wait(&cancel_pdu->complete);
2367 cancel_pdu->cancelled = 0;
2368 pdu_free(cancel_pdu);
2369 }
2370 pdu_complete(pdu, 7);
2371 }
2372
2373 static void coroutine_fn v9fs_link(void *opaque)
2374 {
2375 V9fsPDU *pdu = opaque;
2376 int32_t dfid, oldfid;
2377 V9fsFidState *dfidp, *oldfidp;
2378 V9fsString name;
2379 size_t offset = 7;
2380 int err = 0;
2381
2382 v9fs_string_init(&name);
2383 err = pdu_unmarshal(pdu, offset, "dds", &dfid, &oldfid, &name);
2384 if (err < 0) {
2385 goto out_nofid;
2386 }
2387 trace_v9fs_link(pdu->tag, pdu->id, dfid, oldfid, name.data);
2388
2389 if (name_is_illegal(name.data)) {
2390 err = -ENOENT;
2391 goto out_nofid;
2392 }
2393
2394 if (!strcmp(".", name.data) || !strcmp("..", name.data)) {
2395 err = -EEXIST;
2396 goto out_nofid;
2397 }
2398
2399 dfidp = get_fid(pdu, dfid);
2400 if (dfidp == NULL) {
2401 err = -ENOENT;
2402 goto out_nofid;
2403 }
2404
2405 oldfidp = get_fid(pdu, oldfid);
2406 if (oldfidp == NULL) {
2407 err = -ENOENT;
2408 goto out;
2409 }
2410 err = v9fs_co_link(pdu, oldfidp, dfidp, &name);
2411 if (!err) {
2412 err = offset;
2413 }
2414 put_fid(pdu, oldfidp);
2415 out:
2416 put_fid(pdu, dfidp);
2417 out_nofid:
2418 v9fs_string_free(&name);
2419 pdu_complete(pdu, err);
2420 }
2421
2422 /* Only works with path name based fid */
2423 static void coroutine_fn v9fs_remove(void *opaque)
2424 {
2425 int32_t fid;
2426 int err = 0;
2427 size_t offset = 7;
2428 V9fsFidState *fidp;
2429 V9fsPDU *pdu = opaque;
2430
2431 err = pdu_unmarshal(pdu, offset, "d", &fid);
2432 if (err < 0) {
2433 goto out_nofid;
2434 }
2435 trace_v9fs_remove(pdu->tag, pdu->id, fid);
2436
2437 fidp = get_fid(pdu, fid);
2438 if (fidp == NULL) {
2439 err = -EINVAL;
2440 goto out_nofid;
2441 }
2442 /* if fs driver is not path based, return EOPNOTSUPP */
2443 if (!(pdu->s->ctx.export_flags & V9FS_PATHNAME_FSCONTEXT)) {
2444 err = -EOPNOTSUPP;
2445 goto out_err;
2446 }
2447 /*
2448 * IF the file is unlinked, we cannot reopen
2449 * the file later. So don't reclaim fd
2450 */
2451 err = v9fs_mark_fids_unreclaim(pdu, &fidp->path);
2452 if (err < 0) {
2453 goto out_err;
2454 }
2455 err = v9fs_co_remove(pdu, &fidp->path);
2456 if (!err) {
2457 err = offset;
2458 }
2459 out_err:
2460 /* For TREMOVE we need to clunk the fid even on failed remove */
2461 clunk_fid(pdu->s, fidp->fid);
2462 put_fid(pdu, fidp);
2463 out_nofid:
2464 pdu_complete(pdu, err);
2465 }
2466
2467 static void coroutine_fn v9fs_unlinkat(void *opaque)
2468 {
2469 int err = 0;
2470 V9fsString name;
2471 int32_t dfid, flags;
2472 size_t offset = 7;
2473 V9fsPath path;
2474 V9fsFidState *dfidp;
2475 V9fsPDU *pdu = opaque;
2476
2477 v9fs_string_init(&name);
2478 err = pdu_unmarshal(pdu, offset, "dsd", &dfid, &name, &flags);
2479 if (err < 0) {
2480 goto out_nofid;
2481 }
2482
2483 if (name_is_illegal(name.data)) {
2484 err = -ENOENT;
2485 goto out_nofid;
2486 }
2487
2488 if (!strcmp(".", name.data)) {
2489 err = -EINVAL;
2490 goto out_nofid;
2491 }
2492
2493 if (!strcmp("..", name.data)) {
2494 err = -ENOTEMPTY;
2495 goto out_nofid;
2496 }
2497
2498 dfidp = get_fid(pdu, dfid);
2499 if (dfidp == NULL) {
2500 err = -EINVAL;
2501 goto out_nofid;
2502 }
2503 /*
2504 * IF the file is unlinked, we cannot reopen
2505 * the file later. So don't reclaim fd
2506 */
2507 v9fs_path_init(&path);
2508 err = v9fs_co_name_to_path(pdu, &dfidp->path, name.data, &path);
2509 if (err < 0) {
2510 goto out_err;
2511 }
2512 err = v9fs_mark_fids_unreclaim(pdu, &path);
2513 if (err < 0) {
2514 goto out_err;
2515 }
2516 err = v9fs_co_unlinkat(pdu, &dfidp->path, &name, flags);
2517 if (!err) {
2518 err = offset;
2519 }
2520 out_err:
2521 put_fid(pdu, dfidp);
2522 v9fs_path_free(&path);
2523 out_nofid:
2524 pdu_complete(pdu, err);
2525 v9fs_string_free(&name);
2526 }
2527
2528
2529 /* Only works with path name based fid */
2530 static int coroutine_fn v9fs_complete_rename(V9fsPDU *pdu, V9fsFidState *fidp,
2531 int32_t newdirfid,
2532 V9fsString *name)
2533 {
2534 char *end;
2535 int err = 0;
2536 V9fsPath new_path;
2537 V9fsFidState *tfidp;
2538 V9fsState *s = pdu->s;
2539 V9fsFidState *dirfidp = NULL;
2540 char *old_name, *new_name;
2541
2542 v9fs_path_init(&new_path);
2543 if (newdirfid != -1) {
2544 dirfidp = get_fid(pdu, newdirfid);
2545 if (dirfidp == NULL) {
2546 err = -ENOENT;
2547 goto out_nofid;
2548 }
2549 if (fidp->fid_type != P9_FID_NONE) {
2550 err = -EINVAL;
2551 goto out;
2552 }
2553 v9fs_co_name_to_path(pdu, &dirfidp->path, name->data, &new_path);
2554 } else {
2555 old_name = fidp->path.data;
2556 end = strrchr(old_name, '/');
2557 if (end) {
2558 end++;
2559 } else {
2560 end = old_name;
2561 }
2562 new_name = g_malloc0(end - old_name + name->size + 1);
2563 strncat(new_name, old_name, end - old_name);
2564 strncat(new_name + (end - old_name), name->data, name->size);
2565 v9fs_co_name_to_path(pdu, NULL, new_name, &new_path);
2566 g_free(new_name);
2567 }
2568 err = v9fs_co_rename(pdu, &fidp->path, &new_path);
2569 if (err < 0) {
2570 goto out;
2571 }
2572 /*
2573 * Fixup fid's pointing to the old name to
2574 * start pointing to the new name
2575 */
2576 for (tfidp = s->fid_list; tfidp; tfidp = tfidp->next) {
2577 if (v9fs_path_is_ancestor(&fidp->path, &tfidp->path)) {
2578 /* replace the name */
2579 v9fs_fix_path(&tfidp->path, &new_path, strlen(fidp->path.data));
2580 }
2581 }
2582 out:
2583 if (dirfidp) {
2584 put_fid(pdu, dirfidp);
2585 }
2586 v9fs_path_free(&new_path);
2587 out_nofid:
2588 return err;
2589 }
2590
2591 /* Only works with path name based fid */
2592 static void coroutine_fn v9fs_rename(void *opaque)
2593 {
2594 int32_t fid;
2595 ssize_t err = 0;
2596 size_t offset = 7;
2597 V9fsString name;
2598 int32_t newdirfid;
2599 V9fsFidState *fidp;
2600 V9fsPDU *pdu = opaque;
2601 V9fsState *s = pdu->s;
2602
2603 v9fs_string_init(&name);
2604 err = pdu_unmarshal(pdu, offset, "dds", &fid, &newdirfid, &name);
2605 if (err < 0) {
2606 goto out_nofid;
2607 }
2608
2609 if (name_is_illegal(name.data)) {
2610 err = -ENOENT;
2611 goto out_nofid;
2612 }
2613
2614 if (!strcmp(".", name.data) || !strcmp("..", name.data)) {
2615 err = -EISDIR;
2616 goto out_nofid;
2617 }
2618
2619 fidp = get_fid(pdu, fid);
2620 if (fidp == NULL) {
2621 err = -ENOENT;
2622 goto out_nofid;
2623 }
2624 if (fidp->fid_type != P9_FID_NONE) {
2625 err = -EINVAL;
2626 goto out;
2627 }
2628 /* if fs driver is not path based, return EOPNOTSUPP */
2629 if (!(pdu->s->ctx.export_flags & V9FS_PATHNAME_FSCONTEXT)) {
2630 err = -EOPNOTSUPP;
2631 goto out;
2632 }
2633 v9fs_path_write_lock(s);
2634 err = v9fs_complete_rename(pdu, fidp, newdirfid, &name);
2635 v9fs_path_unlock(s);
2636 if (!err) {
2637 err = offset;
2638 }
2639 out:
2640 put_fid(pdu, fidp);
2641 out_nofid:
2642 pdu_complete(pdu, err);
2643 v9fs_string_free(&name);
2644 }
2645
2646 static void coroutine_fn v9fs_fix_fid_paths(V9fsPDU *pdu, V9fsPath *olddir,
2647 V9fsString *old_name,
2648 V9fsPath *newdir,
2649 V9fsString *new_name)
2650 {
2651 V9fsFidState *tfidp;
2652 V9fsPath oldpath, newpath;
2653 V9fsState *s = pdu->s;
2654
2655
2656 v9fs_path_init(&oldpath);
2657 v9fs_path_init(&newpath);
2658 v9fs_co_name_to_path(pdu, olddir, old_name->data, &oldpath);
2659 v9fs_co_name_to_path(pdu, newdir, new_name->data, &newpath);
2660
2661 /*
2662 * Fixup fid's pointing to the old name to
2663 * start pointing to the new name
2664 */
2665 for (tfidp = s->fid_list; tfidp; tfidp = tfidp->next) {
2666 if (v9fs_path_is_ancestor(&oldpath, &tfidp->path)) {
2667 /* replace the name */
2668 v9fs_fix_path(&tfidp->path, &newpath, strlen(oldpath.data));
2669 }
2670 }
2671 v9fs_path_free(&oldpath);
2672 v9fs_path_free(&newpath);
2673 }
2674
2675 static int coroutine_fn v9fs_complete_renameat(V9fsPDU *pdu, int32_t olddirfid,
2676 V9fsString *old_name,
2677 int32_t newdirfid,
2678 V9fsString *new_name)
2679 {
2680 int err = 0;
2681 V9fsState *s = pdu->s;
2682 V9fsFidState *newdirfidp = NULL, *olddirfidp = NULL;
2683
2684 olddirfidp = get_fid(pdu, olddirfid);
2685 if (olddirfidp == NULL) {
2686 err = -ENOENT;
2687 goto out;
2688 }
2689 if (newdirfid != -1) {
2690 newdirfidp = get_fid(pdu, newdirfid);
2691 if (newdirfidp == NULL) {
2692 err = -ENOENT;
2693 goto out;
2694 }
2695 } else {
2696 newdirfidp = get_fid(pdu, olddirfid);
2697 }
2698
2699 err = v9fs_co_renameat(pdu, &olddirfidp->path, old_name,
2700 &newdirfidp->path, new_name);
2701 if (err < 0) {
2702 goto out;
2703 }
2704 if (s->ctx.export_flags & V9FS_PATHNAME_FSCONTEXT) {
2705 /* Only for path based fid we need to do the below fixup */
2706 v9fs_fix_fid_paths(pdu, &olddirfidp->path, old_name,
2707 &newdirfidp->path, new_name);
2708 }
2709 out:
2710 if (olddirfidp) {
2711 put_fid(pdu, olddirfidp);
2712 }
2713 if (newdirfidp) {
2714 put_fid(pdu, newdirfidp);
2715 }
2716 return err;
2717 }
2718
2719 static void coroutine_fn v9fs_renameat(void *opaque)
2720 {
2721 ssize_t err = 0;
2722 size_t offset = 7;
2723 V9fsPDU *pdu = opaque;
2724 V9fsState *s = pdu->s;
2725 int32_t olddirfid, newdirfid;
2726 V9fsString old_name, new_name;
2727
2728 v9fs_string_init(&old_name);
2729 v9fs_string_init(&new_name);
2730 err = pdu_unmarshal(pdu, offset, "dsds", &olddirfid,
2731 &old_name, &newdirfid, &new_name);
2732 if (err < 0) {
2733 goto out_err;
2734 }
2735
2736 if (name_is_illegal(old_name.data) || name_is_illegal(new_name.data)) {
2737 err = -ENOENT;
2738 goto out_err;
2739 }
2740
2741 if (!strcmp(".", old_name.data) || !strcmp("..", old_name.data) ||
2742 !strcmp(".", new_name.data) || !strcmp("..", new_name.data)) {
2743 err = -EISDIR;
2744 goto out_err;
2745 }
2746
2747 v9fs_path_write_lock(s);
2748 err = v9fs_complete_renameat(pdu, olddirfid,
2749 &old_name, newdirfid, &new_name);
2750 v9fs_path_unlock(s);
2751 if (!err) {
2752 err = offset;
2753 }
2754
2755 out_err:
2756 pdu_complete(pdu, err);
2757 v9fs_string_free(&old_name);
2758 v9fs_string_free(&new_name);
2759 }
2760
2761 static void coroutine_fn v9fs_wstat(void *opaque)
2762 {
2763 int32_t fid;
2764 int err = 0;
2765 int16_t unused;
2766 V9fsStat v9stat;
2767 size_t offset = 7;
2768 struct stat stbuf;
2769 V9fsFidState *fidp;
2770 V9fsPDU *pdu = opaque;
2771
2772 v9fs_stat_init(&v9stat);
2773 err = pdu_unmarshal(pdu, offset, "dwS", &fid, &unused, &v9stat);
2774 if (err < 0) {
2775 goto out_nofid;
2776 }
2777 trace_v9fs_wstat(pdu->tag, pdu->id, fid,
2778 v9stat.mode, v9stat.atime, v9stat.mtime);
2779
2780 fidp = get_fid(pdu, fid);
2781 if (fidp == NULL) {
2782 err = -EINVAL;
2783 goto out_nofid;
2784 }
2785 /* do we need to sync the file? */
2786 if (donttouch_stat(&v9stat)) {
2787 err = v9fs_co_fsync(pdu, fidp, 0);
2788 goto out;
2789 }
2790 if (v9stat.mode != -1) {
2791 uint32_t v9_mode;
2792 err = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
2793 if (err < 0) {
2794 goto out;
2795 }
2796 v9_mode = stat_to_v9mode(&stbuf);
2797 if ((v9stat.mode & P9_STAT_MODE_TYPE_BITS) !=
2798 (v9_mode & P9_STAT_MODE_TYPE_BITS)) {
2799 /* Attempting to change the type */
2800 err = -EIO;
2801 goto out;
2802 }
2803 err = v9fs_co_chmod(pdu, &fidp->path,
2804 v9mode_to_mode(v9stat.mode,
2805 &v9stat.extension));
2806 if (err < 0) {
2807 goto out;
2808 }
2809 }
2810 if (v9stat.mtime != -1 || v9stat.atime != -1) {
2811 struct timespec times[2];
2812 if (v9stat.atime != -1) {
2813 times[0].tv_sec = v9stat.atime;
2814 times[0].tv_nsec = 0;
2815 } else {
2816 times[0].tv_nsec = UTIME_OMIT;
2817 }
2818 if (v9stat.mtime != -1) {
2819 times[1].tv_sec = v9stat.mtime;
2820 times[1].tv_nsec = 0;
2821 } else {
2822 times[1].tv_nsec = UTIME_OMIT;
2823 }
2824 err = v9fs_co_utimensat(pdu, &fidp->path, times);
2825 if (err < 0) {
2826 goto out;
2827 }
2828 }
2829 if (v9stat.n_gid != -1 || v9stat.n_uid != -1) {
2830 err = v9fs_co_chown(pdu, &fidp->path, v9stat.n_uid, v9stat.n_gid);
2831 if (err < 0) {
2832 goto out;
2833 }
2834 }
2835 if (v9stat.name.size != 0) {
2836 err = v9fs_complete_rename(pdu, fidp, -1, &v9stat.name);
2837 if (err < 0) {
2838 goto out;
2839 }
2840 }
2841 if (v9stat.length != -1) {
2842 err = v9fs_co_truncate(pdu, &fidp->path, v9stat.length);
2843 if (err < 0) {
2844 goto out;
2845 }
2846 }
2847 err = offset;
2848 out:
2849 put_fid(pdu, fidp);
2850 out_nofid:
2851 v9fs_stat_free(&v9stat);
2852 pdu_complete(pdu, err);
2853 }
2854
2855 static int v9fs_fill_statfs(V9fsState *s, V9fsPDU *pdu, struct statfs *stbuf)
2856 {
2857 uint32_t f_type;
2858 uint32_t f_bsize;
2859 uint64_t f_blocks;
2860 uint64_t f_bfree;
2861 uint64_t f_bavail;
2862 uint64_t f_files;
2863 uint64_t f_ffree;
2864 uint64_t fsid_val;
2865 uint32_t f_namelen;
2866 size_t offset = 7;
2867 int32_t bsize_factor;
2868
2869 /*
2870 * compute bsize factor based on host file system block size
2871 * and client msize
2872 */
2873 bsize_factor = (s->msize - P9_IOHDRSZ)/stbuf->f_bsize;
2874 if (!bsize_factor) {
2875 bsize_factor = 1;
2876 }
2877 f_type = stbuf->f_type;
2878 f_bsize = stbuf->f_bsize;
2879 f_bsize *= bsize_factor;
2880 /*
2881 * f_bsize is adjusted(multiplied) by bsize factor, so we need to
2882 * adjust(divide) the number of blocks, free blocks and available
2883 * blocks by bsize factor
2884 */
2885 f_blocks = stbuf->f_blocks/bsize_factor;
2886 f_bfree = stbuf->f_bfree/bsize_factor;
2887 f_bavail = stbuf->f_bavail/bsize_factor;
2888 f_files = stbuf->f_files;
2889 f_ffree = stbuf->f_ffree;
2890 fsid_val = (unsigned int) stbuf->f_fsid.__val[0] |
2891 (unsigned long long)stbuf->f_fsid.__val[1] << 32;
2892 f_namelen = stbuf->f_namelen;
2893
2894 return pdu_marshal(pdu, offset, "ddqqqqqqd",
2895 f_type, f_bsize, f_blocks, f_bfree,
2896 f_bavail, f_files, f_ffree,
2897 fsid_val, f_namelen);
2898 }
2899
2900 static void coroutine_fn v9fs_statfs(void *opaque)
2901 {
2902 int32_t fid;
2903 ssize_t retval = 0;
2904 size_t offset = 7;
2905 V9fsFidState *fidp;
2906 struct statfs stbuf;
2907 V9fsPDU *pdu = opaque;
2908 V9fsState *s = pdu->s;
2909
2910 retval = pdu_unmarshal(pdu, offset, "d", &fid);
2911 if (retval < 0) {
2912 goto out_nofid;
2913 }
2914 fidp = get_fid(pdu, fid);
2915 if (fidp == NULL) {
2916 retval = -ENOENT;
2917 goto out_nofid;
2918 }
2919 retval = v9fs_co_statfs(pdu, &fidp->path, &stbuf);
2920 if (retval < 0) {
2921 goto out;
2922 }
2923 retval = v9fs_fill_statfs(s, pdu, &stbuf);
2924 if (retval < 0) {
2925 goto out;
2926 }
2927 retval += offset;
2928 out:
2929 put_fid(pdu, fidp);
2930 out_nofid:
2931 pdu_complete(pdu, retval);
2932 }
2933
2934 static void coroutine_fn v9fs_mknod(void *opaque)
2935 {
2936
2937 int mode;
2938 gid_t gid;
2939 int32_t fid;
2940 V9fsQID qid;
2941 int err = 0;
2942 int major, minor;
2943 size_t offset = 7;
2944 V9fsString name;
2945 struct stat stbuf;
2946 V9fsFidState *fidp;
2947 V9fsPDU *pdu = opaque;
2948
2949 v9fs_string_init(&name);
2950 err = pdu_unmarshal(pdu, offset, "dsdddd", &fid, &name, &mode,
2951 &major, &minor, &gid);
2952 if (err < 0) {
2953 goto out_nofid;
2954 }
2955 trace_v9fs_mknod(pdu->tag, pdu->id, fid, mode, major, minor);
2956
2957 if (name_is_illegal(name.data)) {
2958 err = -ENOENT;
2959 goto out_nofid;
2960 }
2961
2962 if (!strcmp(".", name.data) || !strcmp("..", name.data)) {
2963 err = -EEXIST;
2964 goto out_nofid;
2965 }
2966
2967 fidp = get_fid(pdu, fid);
2968 if (fidp == NULL) {
2969 err = -ENOENT;
2970 goto out_nofid;
2971 }
2972 err = v9fs_co_mknod(pdu, fidp, &name, fidp->uid, gid,
2973 makedev(major, minor), mode, &stbuf);
2974 if (err < 0) {
2975 goto out;
2976 }
2977 stat_to_qid(&stbuf, &qid);
2978 err = pdu_marshal(pdu, offset, "Q", &qid);
2979 if (err < 0) {
2980 goto out;
2981 }
2982 err += offset;
2983 trace_v9fs_mknod_return(pdu->tag, pdu->id,
2984 qid.type, qid.version, qid.path);
2985 out:
2986 put_fid(pdu, fidp);
2987 out_nofid:
2988 pdu_complete(pdu, err);
2989 v9fs_string_free(&name);
2990 }
2991
2992 /*
2993 * Implement posix byte range locking code
2994 * Server side handling of locking code is very simple, because 9p server in
2995 * QEMU can handle only one client. And most of the lock handling
2996 * (like conflict, merging) etc is done by the VFS layer itself, so no need to
2997 * do any thing in * qemu 9p server side lock code path.
2998 * So when a TLOCK request comes, always return success
2999 */
3000 static void coroutine_fn v9fs_lock(void *opaque)
3001 {
3002 int8_t status;
3003 V9fsFlock flock;
3004 size_t offset = 7;
3005 struct stat stbuf;
3006 V9fsFidState *fidp;
3007 int32_t fid, err = 0;
3008 V9fsPDU *pdu = opaque;
3009
3010 status = P9_LOCK_ERROR;
3011 v9fs_string_init(&flock.client_id);
3012 err = pdu_unmarshal(pdu, offset, "dbdqqds", &fid, &flock.type,
3013 &flock.flags, &flock.start, &flock.length,
3014 &flock.proc_id, &flock.client_id);
3015 if (err < 0) {
3016 goto out_nofid;
3017 }
3018 trace_v9fs_lock(pdu->tag, pdu->id, fid,
3019 flock.type, flock.start, flock.length);
3020
3021
3022 /* We support only block flag now (that too ignored currently) */
3023 if (flock.flags & ~P9_LOCK_FLAGS_BLOCK) {
3024 err = -EINVAL;
3025 goto out_nofid;
3026 }
3027 fidp = get_fid(pdu, fid);
3028 if (fidp == NULL) {
3029 err = -ENOENT;
3030 goto out_nofid;
3031 }
3032 err = v9fs_co_fstat(pdu, fidp, &stbuf);
3033 if (err < 0) {
3034 goto out;
3035 }
3036 status = P9_LOCK_SUCCESS;
3037 out:
3038 put_fid(pdu, fidp);
3039 out_nofid:
3040 err = pdu_marshal(pdu, offset, "b", status);
3041 if (err > 0) {
3042 err += offset;
3043 }
3044 trace_v9fs_lock_return(pdu->tag, pdu->id, status);
3045 pdu_complete(pdu, err);
3046 v9fs_string_free(&flock.client_id);
3047 }
3048
3049 /*
3050 * When a TGETLOCK request comes, always return success because all lock
3051 * handling is done by client's VFS layer.
3052 */
3053 static void coroutine_fn v9fs_getlock(void *opaque)
3054 {
3055 size_t offset = 7;
3056 struct stat stbuf;
3057 V9fsFidState *fidp;
3058 V9fsGetlock glock;
3059 int32_t fid, err = 0;
3060 V9fsPDU *pdu = opaque;
3061
3062 v9fs_string_init(&glock.client_id);
3063 err = pdu_unmarshal(pdu, offset, "dbqqds", &fid, &glock.type,
3064 &glock.start, &glock.length, &glock.proc_id,
3065 &glock.client_id);
3066 if (err < 0) {
3067 goto out_nofid;
3068 }
3069 trace_v9fs_getlock(pdu->tag, pdu->id, fid,
3070 glock.type, glock.start, glock.length);
3071
3072 fidp = get_fid(pdu, fid);
3073 if (fidp == NULL) {
3074 err = -ENOENT;
3075 goto out_nofid;
3076 }
3077 err = v9fs_co_fstat(pdu, fidp, &stbuf);
3078 if (err < 0) {
3079 goto out;
3080 }
3081 glock.type = P9_LOCK_TYPE_UNLCK;
3082 err = pdu_marshal(pdu, offset, "bqqds", glock.type,
3083 glock.start, glock.length, glock.proc_id,
3084 &glock.client_id);
3085 if (err < 0) {
3086 goto out;
3087 }
3088 err += offset;
3089 trace_v9fs_getlock_return(pdu->tag, pdu->id, glock.type, glock.start,
3090 glock.length, glock.proc_id);
3091 out:
3092 put_fid(pdu, fidp);
3093 out_nofid:
3094 pdu_complete(pdu, err);
3095 v9fs_string_free(&glock.client_id);
3096 }
3097
3098 static void coroutine_fn v9fs_mkdir(void *opaque)
3099 {
3100 V9fsPDU *pdu = opaque;
3101 size_t offset = 7;
3102 int32_t fid;
3103 struct stat stbuf;
3104 V9fsQID qid;
3105 V9fsString name;
3106 V9fsFidState *fidp;
3107 gid_t gid;
3108 int mode;
3109 int err = 0;
3110
3111 v9fs_string_init(&name);
3112 err = pdu_unmarshal(pdu, offset, "dsdd", &fid, &name, &mode, &gid);
3113 if (err < 0) {
3114 goto out_nofid;
3115 }
3116 trace_v9fs_mkdir(pdu->tag, pdu->id, fid, name.data, mode, gid);
3117
3118 if (name_is_illegal(name.data)) {
3119 err = -ENOENT;
3120 goto out_nofid;
3121 }
3122
3123 if (!strcmp(".", name.data) || !strcmp("..", name.data)) {
3124 err = -EEXIST;
3125 goto out_nofid;
3126 }
3127
3128 fidp = get_fid(pdu, fid);
3129 if (fidp == NULL) {
3130 err = -ENOENT;
3131 goto out_nofid;
3132 }
3133 err = v9fs_co_mkdir(pdu, fidp, &name, mode, fidp->uid, gid, &stbuf);
3134 if (err < 0) {
3135 goto out;
3136 }
3137 stat_to_qid(&stbuf, &qid);
3138 err = pdu_marshal(pdu, offset, "Q", &qid);
3139 if (err < 0) {
3140 goto out;
3141 }
3142 err += offset;
3143 trace_v9fs_mkdir_return(pdu->tag, pdu->id,
3144 qid.type, qid.version, qid.path, err);
3145 out:
3146 put_fid(pdu, fidp);
3147 out_nofid:
3148 pdu_complete(pdu, err);
3149 v9fs_string_free(&name);
3150 }
3151
3152 static void coroutine_fn v9fs_xattrwalk(void *opaque)
3153 {
3154 int64_t size;
3155 V9fsString name;
3156 ssize_t err = 0;
3157 size_t offset = 7;
3158 int32_t fid, newfid;
3159 V9fsFidState *file_fidp;
3160 V9fsFidState *xattr_fidp = NULL;
3161 V9fsPDU *pdu = opaque;
3162 V9fsState *s = pdu->s;
3163
3164 v9fs_string_init(&name);
3165 err = pdu_unmarshal(pdu, offset, "dds", &fid, &newfid, &name);
3166 if (err < 0) {
3167 goto out_nofid;
3168 }
3169 trace_v9fs_xattrwalk(pdu->tag, pdu->id, fid, newfid, name.data);
3170
3171 file_fidp = get_fid(pdu, fid);
3172 if (file_fidp == NULL) {
3173 err = -ENOENT;
3174 goto out_nofid;
3175 }
3176 xattr_fidp = alloc_fid(s, newfid);
3177 if (xattr_fidp == NULL) {
3178 err = -EINVAL;
3179 goto out;
3180 }
3181 v9fs_path_copy(&xattr_fidp->path, &file_fidp->path);
3182 if (!v9fs_string_size(&name)) {
3183 /*
3184 * listxattr request. Get the size first
3185 */
3186 size = v9fs_co_llistxattr(pdu, &xattr_fidp->path, NULL, 0);
3187 if (size < 0) {
3188 err = size;
3189 clunk_fid(s, xattr_fidp->fid);
3190 goto out;
3191 }
3192 /*
3193 * Read the xattr value
3194 */
3195 xattr_fidp->fs.xattr.len = size;
3196 xattr_fidp->fid_type = P9_FID_XATTR;
3197 xattr_fidp->fs.xattr.xattrwalk_fid = true;
3198 if (size) {
3199 xattr_fidp->fs.xattr.value = g_malloc(size);
3200 err = v9fs_co_llistxattr(pdu, &xattr_fidp->path,
3201 xattr_fidp->fs.xattr.value,
3202 xattr_fidp->fs.xattr.len);
3203 if (err < 0) {
3204 clunk_fid(s, xattr_fidp->fid);
3205 goto out;
3206 }
3207 }
3208 err = pdu_marshal(pdu, offset, "q", size);
3209 if (err < 0) {
3210 goto out;
3211 }
3212 err += offset;
3213 } else {
3214 /*
3215 * specific xattr fid. We check for xattr
3216 * presence also collect the xattr size
3217 */
3218 size = v9fs_co_lgetxattr(pdu, &xattr_fidp->path,
3219 &name, NULL, 0);
3220 if (size < 0) {
3221 err = size;
3222 clunk_fid(s, xattr_fidp->fid);
3223 goto out;
3224 }
3225 /*
3226 * Read the xattr value
3227 */
3228 xattr_fidp->fs.xattr.len = size;
3229 xattr_fidp->fid_type = P9_FID_XATTR;
3230 xattr_fidp->fs.xattr.xattrwalk_fid = true;
3231 if (size) {
3232 xattr_fidp->fs.xattr.value = g_malloc(size);
3233 err = v9fs_co_lgetxattr(pdu, &xattr_fidp->path,
3234 &name, xattr_fidp->fs.xattr.value,
3235 xattr_fidp->fs.xattr.len);
3236 if (err < 0) {
3237 clunk_fid(s, xattr_fidp->fid);
3238 goto out;
3239 }
3240 }
3241 err = pdu_marshal(pdu, offset, "q", size);
3242 if (err < 0) {
3243 goto out;
3244 }
3245 err += offset;
3246 }
3247 trace_v9fs_xattrwalk_return(pdu->tag, pdu->id, size);
3248 out:
3249 put_fid(pdu, file_fidp);
3250 if (xattr_fidp) {
3251 put_fid(pdu, xattr_fidp);
3252 }
3253 out_nofid:
3254 pdu_complete(pdu, err);
3255 v9fs_string_free(&name);
3256 }
3257
3258 static void coroutine_fn v9fs_xattrcreate(void *opaque)
3259 {
3260 int flags;
3261 int32_t fid;
3262 uint64_t size;
3263 ssize_t err = 0;
3264 V9fsString name;
3265 size_t offset = 7;
3266 V9fsFidState *file_fidp;
3267 V9fsFidState *xattr_fidp;
3268 V9fsPDU *pdu = opaque;
3269
3270 v9fs_string_init(&name);
3271 err = pdu_unmarshal(pdu, offset, "dsqd", &fid, &name, &size, &flags);
3272 if (err < 0) {
3273 goto out_nofid;
3274 }
3275 trace_v9fs_xattrcreate(pdu->tag, pdu->id, fid, name.data, size, flags);
3276
3277 if (size > XATTR_SIZE_MAX) {
3278 err = -E2BIG;
3279 goto out_nofid;
3280 }
3281
3282 file_fidp = get_fid(pdu, fid);
3283 if (file_fidp == NULL) {
3284 err = -EINVAL;
3285 goto out_nofid;
3286 }
3287 if (file_fidp->fid_type != P9_FID_NONE) {
3288 err = -EINVAL;
3289 goto out_put_fid;
3290 }
3291
3292 /* Make the file fid point to xattr */
3293 xattr_fidp = file_fidp;
3294 xattr_fidp->fid_type = P9_FID_XATTR;
3295 xattr_fidp->fs.xattr.copied_len = 0;
3296 xattr_fidp->fs.xattr.xattrwalk_fid = false;
3297 xattr_fidp->fs.xattr.len = size;
3298 xattr_fidp->fs.xattr.flags = flags;
3299 v9fs_string_init(&xattr_fidp->fs.xattr.name);
3300 v9fs_string_copy(&xattr_fidp->fs.xattr.name, &name);
3301 xattr_fidp->fs.xattr.value = g_malloc0(size);
3302 err = offset;
3303 out_put_fid:
3304 put_fid(pdu, file_fidp);
3305 out_nofid:
3306 pdu_complete(pdu, err);
3307 v9fs_string_free(&name);
3308 }
3309
3310 static void coroutine_fn v9fs_readlink(void *opaque)
3311 {
3312 V9fsPDU *pdu = opaque;
3313 size_t offset = 7;
3314 V9fsString target;
3315 int32_t fid;
3316 int err = 0;
3317 V9fsFidState *fidp;
3318
3319 err = pdu_unmarshal(pdu, offset, "d", &fid);
3320 if (err < 0) {
3321 goto out_nofid;
3322 }
3323 trace_v9fs_readlink(pdu->tag, pdu->id, fid);
3324 fidp = get_fid(pdu, fid);
3325 if (fidp == NULL) {
3326 err = -ENOENT;
3327 goto out_nofid;
3328 }
3329
3330 v9fs_string_init(&target);
3331 err = v9fs_co_readlink(pdu, &fidp->path, &target);
3332 if (err < 0) {
3333 goto out;
3334 }
3335 err = pdu_marshal(pdu, offset, "s", &target);
3336 if (err < 0) {
3337 v9fs_string_free(&target);
3338 goto out;
3339 }
3340 err += offset;
3341 trace_v9fs_readlink_return(pdu->tag, pdu->id, target.data);
3342 v9fs_string_free(&target);
3343 out:
3344 put_fid(pdu, fidp);
3345 out_nofid:
3346 pdu_complete(pdu, err);
3347 }
3348
3349 static CoroutineEntry *pdu_co_handlers[] = {
3350 [P9_TREADDIR] = v9fs_readdir,
3351 [P9_TSTATFS] = v9fs_statfs,
3352 [P9_TGETATTR] = v9fs_getattr,
3353 [P9_TSETATTR] = v9fs_setattr,
3354 [P9_TXATTRWALK] = v9fs_xattrwalk,
3355 [P9_TXATTRCREATE] = v9fs_xattrcreate,
3356 [P9_TMKNOD] = v9fs_mknod,
3357 [P9_TRENAME] = v9fs_rename,
3358 [P9_TLOCK] = v9fs_lock,
3359 [P9_TGETLOCK] = v9fs_getlock,
3360 [P9_TRENAMEAT] = v9fs_renameat,
3361 [P9_TREADLINK] = v9fs_readlink,
3362 [P9_TUNLINKAT] = v9fs_unlinkat,
3363 [P9_TMKDIR] = v9fs_mkdir,
3364 [P9_TVERSION] = v9fs_version,
3365 [P9_TLOPEN] = v9fs_open,
3366 [P9_TATTACH] = v9fs_attach,
3367 [P9_TSTAT] = v9fs_stat,
3368 [P9_TWALK] = v9fs_walk,
3369 [P9_TCLUNK] = v9fs_clunk,
3370 [P9_TFSYNC] = v9fs_fsync,
3371 [P9_TOPEN] = v9fs_open,
3372 [P9_TREAD] = v9fs_read,
3373 #if 0
3374 [P9_TAUTH] = v9fs_auth,
3375 #endif
3376 [P9_TFLUSH] = v9fs_flush,
3377 [P9_TLINK] = v9fs_link,
3378 [P9_TSYMLINK] = v9fs_symlink,
3379 [P9_TCREATE] = v9fs_create,
3380 [P9_TLCREATE] = v9fs_lcreate,
3381 [P9_TWRITE] = v9fs_write,
3382 [P9_TWSTAT] = v9fs_wstat,
3383 [P9_TREMOVE] = v9fs_remove,
3384 };
3385
3386 static void coroutine_fn v9fs_op_not_supp(void *opaque)
3387 {
3388 V9fsPDU *pdu = opaque;
3389 pdu_complete(pdu, -EOPNOTSUPP);
3390 }
3391
3392 static void coroutine_fn v9fs_fs_ro(void *opaque)
3393 {
3394 V9fsPDU *pdu = opaque;
3395 pdu_complete(pdu, -EROFS);
3396 }
3397
3398 static inline bool is_read_only_op(V9fsPDU *pdu)
3399 {
3400 switch (pdu->id) {
3401 case P9_TREADDIR:
3402 case P9_TSTATFS:
3403 case P9_TGETATTR:
3404 case P9_TXATTRWALK:
3405 case P9_TLOCK:
3406 case P9_TGETLOCK:
3407 case P9_TREADLINK:
3408 case P9_TVERSION:
3409 case P9_TLOPEN:
3410 case P9_TATTACH:
3411 case P9_TSTAT:
3412 case P9_TWALK:
3413 case P9_TCLUNK:
3414 case P9_TFSYNC: