4 * Copyright IBM, Corp. 2010
9 * This work is licensed under the terms of the GNU GPL, version 2. See
10 * the COPYING file in the top-level directory.
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"
20 #include "qemu/sockets.h"
21 #include "virtio-9p.h"
22 #include "fsdev/qemu-fsdev.h"
26 #include "migration/migration.h"
30 static int open_fd_rc;
44 ssize_t pdu_marshal(V9fsPDU *pdu, size_t offset, const char *fmt, ...)
50 ret = virtio_pdu_vmarshal(pdu, offset, fmt, ap);
56 ssize_t pdu_unmarshal(V9fsPDU *pdu, size_t offset, const char *fmt, ...)
62 ret = virtio_pdu_vunmarshal(pdu, offset, fmt, ap);
68 static void pdu_push_and_notify(V9fsPDU *pdu)
70 virtio_9p_push_and_notify(pdu);
73 static int omode_to_uflags(int8_t mode)
107 struct dotl_openflag_map {
112 static int dotl_to_open_flags(int flags)
116 * We have same bits for P9_DOTL_READONLY, P9_DOTL_WRONLY
117 * and P9_DOTL_NOACCESS
119 int oflags = flags & O_ACCMODE;
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 },
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;
147 void cred_init(FsCred *credp)
155 static int get_dotl_openflags(V9fsState *s, int oflags)
159 * Filter the client open flags
161 flags = dotl_to_open_flags(oflags);
162 flags &= ~(O_NOCTTY | O_ASYNC | O_CREAT);
164 * Ignore direct disk access hint until the server supports it.
170 void v9fs_path_init(V9fsPath *path)
176 void v9fs_path_free(V9fsPath *path)
184 void GCC_FMT_ATTR(2, 3)
185 v9fs_path_sprintf(V9fsPath *path, const char *fmt, ...)
189 v9fs_path_free(path);
192 /* Bump the size for including terminating NULL */
193 path->size = g_vasprintf(&path->data, fmt, ap) + 1;
197 void v9fs_path_copy(V9fsPath *lhs, V9fsPath *rhs)
200 lhs->data = g_malloc(rhs->size);
201 memcpy(lhs->data, rhs->data, rhs->size);
202 lhs->size = rhs->size;
205 int v9fs_name_to_path(V9fsState *s, V9fsPath *dirpath,
206 const char *name, V9fsPath *path)
209 err = s->ops->name_to_path(&s->ctx, dirpath, name, path);
217 * Return TRUE if s1 is an ancestor of s2.
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!
222 static int v9fs_path_is_ancestor(V9fsPath *s1, V9fsPath *s2)
224 if (!strncmp(s1->data, s2->data, s1->size - 1)) {
225 if (s2->data[s1->size - 1] == '\0' || s2->data[s1->size - 1] == '/') {
232 static size_t v9fs_string_size(V9fsString *str)
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)
242 if (f->fid_type == P9_FID_FILE) {
243 if (f->fs.fd == -1) {
245 err = v9fs_co_open(pdu, f, f->open_flags);
246 } while (err == -EINTR && !pdu->cancelled);
248 } else if (f->fid_type == P9_FID_DIR) {
249 if (f->fs.dir.stream == NULL) {
251 err = v9fs_co_opendir(pdu, f);
252 } while (err == -EINTR && !pdu->cancelled);
258 static V9fsFidState *coroutine_fn get_fid(V9fsPDU *pdu, int32_t fid)
262 V9fsState *s = pdu->s;
264 for (f = s->fid_list; f; f = f->next) {
268 * Update the fid ref upfront so that
269 * we don't get reclaimed when we yield
274 * check whether we need to reopen the
275 * file. We might have closed the fd
276 * while trying to free up some file
279 err = v9fs_reopen_fid(pdu, f);
285 * Mark the fid as referenced so that the LRU
286 * reclaim won't close the file descriptor
288 f->flags |= FID_REFERENCED;
295 static V9fsFidState *alloc_fid(V9fsState *s, int32_t fid)
299 for (f = s->fid_list; f; f = f->next) {
300 /* If fid is already there return NULL */
306 f = g_malloc0(sizeof(V9fsFidState));
308 f->fid_type = P9_FID_NONE;
311 * Mark the fid as referenced so that the LRU
312 * reclaim won't close the file descriptor
314 f->flags |= FID_REFERENCED;
315 f->next = s->fid_list;
318 v9fs_readdir_init(&f->fs.dir);
319 v9fs_readdir_init(&f->fs_reclaim.dir);
324 static int coroutine_fn v9fs_xattr_fid_clunk(V9fsPDU *pdu, V9fsFidState *fidp)
328 if (fidp->fs.xattr.xattrwalk_fid) {
329 /* getxattr/listxattr fid */
333 * if this is fid for setxattr. clunk should
334 * result in setxattr localcall
336 if (fidp->fs.xattr.len != fidp->fs.xattr.copied_len) {
337 /* clunk after partial write */
341 if (fidp->fs.xattr.len) {
342 retval = v9fs_co_lsetxattr(pdu, &fidp->path, &fidp->fs.xattr.name,
343 fidp->fs.xattr.value,
345 fidp->fs.xattr.flags);
347 retval = v9fs_co_lremovexattr(pdu, &fidp->path, &fidp->fs.xattr.name);
350 v9fs_string_free(&fidp->fs.xattr.name);
352 g_free(fidp->fs.xattr.value);
356 static int coroutine_fn free_fid(V9fsPDU *pdu, V9fsFidState *fidp)
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);
365 } else if (fidp->fid_type == P9_FID_DIR) {
366 if (fidp->fs.dir.stream != NULL) {
367 retval = v9fs_co_closedir(pdu, &fidp->fs);
369 } else if (fidp->fid_type == P9_FID_XATTR) {
370 retval = v9fs_xattr_fid_clunk(pdu, fidp);
372 v9fs_path_free(&fidp->path);
377 static int coroutine_fn put_fid(V9fsPDU *pdu, V9fsFidState *fidp)
382 * Don't free the fid if it is in reclaim list
384 if (!fidp->ref && fidp->clunked) {
385 if (fidp->fid == pdu->s->root_fid) {
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
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;
398 return free_fid(pdu, fidp);
403 static V9fsFidState *clunk_fid(V9fsState *s, int32_t fid)
405 V9fsFidState **fidpp, *fidp;
407 for (fidpp = &s->fid_list; *fidpp; fidpp = &(*fidpp)->next) {
408 if ((*fidpp)->fid == fid) {
412 if (*fidpp == NULL) {
421 void coroutine_fn v9fs_reclaim_fd(V9fsPDU *pdu)
423 int reclaim_count = 0;
424 V9fsState *s = pdu->s;
425 V9fsFidState *f, *reclaim_list = NULL;
427 for (f = s->fid_list; f; f = f->next) {
429 * Unlink fids cannot be reclaimed. Check
430 * for them and skip them. Also skip fids
431 * currently being operated on.
433 if (f->ref || f->flags & FID_NON_RECLAIMABLE) {
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)
443 if (f->flags & FID_REFERENCED) {
444 f->flags &= ~FID_REFERENCED;
448 * Add fids to reclaim list.
450 if (f->fid_type == P9_FID_FILE) {
451 if (f->fs.fd != -1) {
453 * Up the reference count so that
454 * a clunk request won't free this fid
457 f->rclm_lst = reclaim_list;
459 f->fs_reclaim.fd = f->fs.fd;
463 } else if (f->fid_type == P9_FID_DIR) {
464 if (f->fs.dir.stream != NULL) {
466 * Up the reference count so that
467 * a clunk request won't free this fid
470 f->rclm_lst = reclaim_list;
472 f->fs_reclaim.dir.stream = f->fs.dir.stream;
473 f->fs.dir.stream = NULL;
477 if (reclaim_count >= open_fd_rc) {
482 * Now close the fid in reclaim list. Free them if they
483 * are already clunked.
485 while (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);
495 * Now drop the fid reference, free it
502 static int coroutine_fn v9fs_mark_fids_unreclaim(V9fsPDU *pdu, V9fsPath *path)
505 V9fsState *s = pdu->s;
506 V9fsFidState *fidp, head_fid;
508 head_fid.next = s->fid_list;
509 for (fidp = s->fid_list; fidp; fidp = fidp->next) {
510 if (fidp->path.size != path->size) {
513 if (!memcmp(fidp->path.data, path->data, path->size)) {
514 /* Mark the fid non reclaimable. */
515 fidp->flags |= FID_NON_RECLAIMABLE;
517 /* reopen the file/dir if already closed */
518 err = v9fs_reopen_fid(pdu, fidp);
523 * Go back to head of fid list because
524 * the list could have got updated when
525 * switched to the worker thread
535 static void coroutine_fn virtfs_reset(V9fsPDU *pdu)
537 V9fsState *s = pdu->s;
541 while (s->fid_list) {
543 s->fid_list = fidp->next;
553 #define P9_QID_TYPE_DIR 0x80
554 #define P9_QID_TYPE_SYMLINK 0x02
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
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 | \
578 /* This is the algorithm from ufs in spfs */
579 static void stat_to_qid(const struct stat *stbuf, V9fsQID *qidp)
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);
588 if (S_ISDIR(stbuf->st_mode)) {
589 qidp->type |= P9_QID_TYPE_DIR;
591 if (S_ISLNK(stbuf->st_mode)) {
592 qidp->type |= P9_QID_TYPE_SYMLINK;
596 static int coroutine_fn fid_to_qid(V9fsPDU *pdu, V9fsFidState *fidp,
602 err = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
606 stat_to_qid(&stbuf, qidp);
610 V9fsPDU *pdu_alloc(V9fsState *s)
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);
622 void pdu_free(V9fsPDU *pdu)
624 V9fsState *s = pdu->s;
626 g_assert(!pdu->cancelled);
627 QLIST_REMOVE(pdu, next);
628 QLIST_INSERT_HEAD(&s->free_list, pdu, next);
632 * We don't do error checking for pdu_marshal/unmarshal here
633 * because we always expect to have enough space to encode
636 static void coroutine_fn pdu_complete(V9fsPDU *pdu, ssize_t len)
638 int8_t id = pdu->id + 1; /* Response */
639 V9fsState *s = pdu->s;
645 if (s->proto_version != V9FS_PROTO_2000L) {
648 str.data = strerror(err);
649 str.size = strlen(str.data);
651 len += pdu_marshal(pdu, len, "s", &str);
655 len += pdu_marshal(pdu, len, "d", err);
657 if (s->proto_version == V9FS_PROTO_2000L) {
660 trace_v9fs_rerror(pdu->tag, pdu->id, err); /* Trace ERROR */
663 /* fill out the header */
664 pdu_marshal(pdu, 0, "dbw", (int32_t)len, id, pdu->tag);
666 /* keep these in sync */
670 pdu_push_and_notify(pdu);
672 /* Now wakeup anybody waiting in flush for this request */
673 if (!qemu_co_queue_next(&pdu->complete)) {
678 static mode_t v9mode_to_mode(uint32_t mode, V9fsString *extension)
683 if (mode & P9_STAT_MODE_DIR) {
687 if (mode & P9_STAT_MODE_SYMLINK) {
690 if (mode & P9_STAT_MODE_SOCKET) {
693 if (mode & P9_STAT_MODE_NAMED_PIPE) {
696 if (mode & P9_STAT_MODE_DEVICE) {
697 if (extension->size && extension->data[0] == 'c') {
708 if (mode & P9_STAT_MODE_SETUID) {
711 if (mode & P9_STAT_MODE_SETGID) {
714 if (mode & P9_STAT_MODE_SETVTX) {
721 static int donttouch_stat(V9fsStat *stat)
723 if (stat->type == -1 &&
725 stat->qid.type == -1 &&
726 stat->qid.version == -1 &&
727 stat->qid.path == -1 &&
731 stat->length == -1 &&
738 stat->n_muid == -1) {
745 static void v9fs_stat_init(V9fsStat *stat)
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);
754 static void v9fs_stat_free(V9fsStat *stat)
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);
763 static uint32_t stat_to_v9mode(const struct stat *stbuf)
767 mode = stbuf->st_mode & 0777;
768 if (S_ISDIR(stbuf->st_mode)) {
769 mode |= P9_STAT_MODE_DIR;
772 if (S_ISLNK(stbuf->st_mode)) {
773 mode |= P9_STAT_MODE_SYMLINK;
776 if (S_ISSOCK(stbuf->st_mode)) {
777 mode |= P9_STAT_MODE_SOCKET;
780 if (S_ISFIFO(stbuf->st_mode)) {
781 mode |= P9_STAT_MODE_NAMED_PIPE;
784 if (S_ISBLK(stbuf->st_mode) || S_ISCHR(stbuf->st_mode)) {
785 mode |= P9_STAT_MODE_DEVICE;
788 if (stbuf->st_mode & S_ISUID) {
789 mode |= P9_STAT_MODE_SETUID;
792 if (stbuf->st_mode & S_ISGID) {
793 mode |= P9_STAT_MODE_SETGID;
796 if (stbuf->st_mode & S_ISVTX) {
797 mode |= P9_STAT_MODE_SETVTX;
803 static int coroutine_fn stat_to_v9stat(V9fsPDU *pdu, V9fsPath *name,
804 const struct stat *stbuf,
810 memset(v9stat, 0, sizeof(*v9stat));
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;
818 v9fs_string_free(&v9stat->uid);
819 v9fs_string_free(&v9stat->gid);
820 v9fs_string_free(&v9stat->muid);
822 v9stat->n_uid = stbuf->st_uid;
823 v9stat->n_gid = stbuf->st_gid;
826 v9fs_string_free(&v9stat->extension);
828 if (v9stat->mode & P9_STAT_MODE_SYMLINK) {
829 err = v9fs_co_readlink(pdu, name, &v9stat->extension);
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);
842 str = strrchr(name->data, '/');
849 v9fs_string_sprintf(&v9stat->name, "%s", str);
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);
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
872 #define P9_STATS_BTIME 0x00000800ULL
873 #define P9_STATS_GEN 0x00001000ULL
874 #define P9_STATS_DATA_VERSION 0x00002000ULL
876 #define P9_STATS_BASIC 0x000007ffULL /* Mask for fields up to BLOCKS */
877 #define P9_STATS_ALL 0x00003fffULL /* Mask for All fields above */
880 static void stat_to_v9stat_dotl(V9fsState *s, const struct stat *stbuf,
881 V9fsStatDotl *v9lstat)
883 memset(v9lstat, 0, sizeof(*v9lstat));
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;
902 stat_to_qid(stbuf, &v9lstat->qid);
905 static void print_sg(struct iovec *sg, int cnt)
909 printf("sg[%d]: {", cnt);
910 for (i = 0; i < cnt; i++) {
914 printf("(%p, %zd)", sg[i].iov_base, sg[i].iov_len);
919 /* Will call this only for path name based fid */
920 static void v9fs_fix_path(V9fsPath *dst, V9fsPath *src, int len)
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);
929 static inline bool is_ro_export(FsContext *ctx)
931 return ctx->export_flags & V9FS_RDONLY;
934 static void coroutine_fn v9fs_version(void *opaque)
937 V9fsPDU *pdu = opaque;
938 V9fsState *s = pdu->s;
942 v9fs_string_init(&version);
943 err = pdu_unmarshal(pdu, offset, "ds", &s->msize, &version);
948 trace_v9fs_version(pdu->tag, pdu->id, s->msize, version.data);
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;
957 v9fs_string_sprintf(&version, "unknown");
960 err = pdu_marshal(pdu, offset, "ds", s->msize, &version);
966 trace_v9fs_version_return(pdu->tag, pdu->id, s->msize, version.data);
968 pdu_complete(pdu, offset);
969 v9fs_string_free(&version);
972 static void coroutine_fn v9fs_attach(void *opaque)
974 V9fsPDU *pdu = opaque;
975 V9fsState *s = pdu->s;
976 int32_t fid, afid, n_uname;
977 V9fsString uname, aname;
983 v9fs_string_init(&uname);
984 v9fs_string_init(&aname);
985 err = pdu_unmarshal(pdu, offset, "ddssd", &fid,
986 &afid, &uname, &aname, &n_uname);
990 trace_v9fs_attach(pdu->tag, pdu->id, fid, afid, uname.data, aname.data);
992 fidp = alloc_fid(s, fid);
998 err = v9fs_co_name_to_path(pdu, NULL, "/", &fidp->path);
1004 err = fid_to_qid(pdu, fidp, &qid);
1010 err = pdu_marshal(pdu, offset, "Q", &qid);
1016 memcpy(&s->root_qid, &qid, sizeof(qid));
1017 trace_v9fs_attach_return(pdu->tag, pdu->id,
1018 qid.type, qid.version, qid.path);
1020 * disable migration if we haven't done already.
1021 * attach could get called multiple times for the same export.
1023 if (!s->migration_blocker) {
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);
1033 pdu_complete(pdu, err);
1034 v9fs_string_free(&uname);
1035 v9fs_string_free(&aname);
1038 static void coroutine_fn v9fs_stat(void *opaque)
1046 V9fsPDU *pdu = opaque;
1048 err = pdu_unmarshal(pdu, offset, "d", &fid);
1052 trace_v9fs_stat(pdu->tag, pdu->id, fid);
1054 fidp = get_fid(pdu, fid);
1059 err = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
1063 err = stat_to_v9stat(pdu, &fidp->path, &stbuf, &v9stat);
1067 err = pdu_marshal(pdu, offset, "wS", 0, &v9stat);
1069 v9fs_stat_free(&v9stat);
1072 trace_v9fs_stat_return(pdu->tag, pdu->id, v9stat.mode,
1073 v9stat.atime, v9stat.mtime, v9stat.length);
1075 v9fs_stat_free(&v9stat);
1079 pdu_complete(pdu, err);
1082 static void coroutine_fn v9fs_getattr(void *opaque)
1089 uint64_t request_mask;
1090 V9fsStatDotl v9stat_dotl;
1091 V9fsPDU *pdu = opaque;
1092 V9fsState *s = pdu->s;
1094 retval = pdu_unmarshal(pdu, offset, "dq", &fid, &request_mask);
1098 trace_v9fs_getattr(pdu->tag, pdu->id, fid, request_mask);
1100 fidp = get_fid(pdu, fid);
1106 * Currently we only support BASIC fields in stat, so there is no
1107 * need to look at request_mask.
1109 retval = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
1113 stat_to_v9stat_dotl(s, &stbuf, &v9stat_dotl);
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);
1120 /* we have valid st_gen: update result mask */
1121 v9stat_dotl.st_result_mask |= P9_STATS_GEN;
1124 /* request cancelled, e.g. by Tflush */
1127 /* failed to get st_gen: not fatal, ignore */
1131 retval = pdu_marshal(pdu, offset, "A", &v9stat_dotl);
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);
1142 pdu_complete(pdu, retval);
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)
1156 #define P9_ATTR_MASK 127
1158 static void coroutine_fn v9fs_setattr(void *opaque)
1165 V9fsPDU *pdu = opaque;
1167 err = pdu_unmarshal(pdu, offset, "dI", &fid, &v9iattr);
1172 fidp = get_fid(pdu, fid);
1177 if (v9iattr.valid & P9_ATTR_MODE) {
1178 err = v9fs_co_chmod(pdu, &fidp->path, v9iattr.mode);
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;
1190 times[0].tv_nsec = UTIME_NOW;
1193 times[0].tv_nsec = UTIME_OMIT;
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;
1200 times[1].tv_nsec = UTIME_NOW;
1203 times[1].tv_nsec = UTIME_OMIT;
1205 err = v9fs_co_utimensat(pdu, &fidp->path, times);
1211 * If the only valid entry in iattr is ctime we can call
1212 * chown(-1,-1) to update the ctime of the file
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)) {
1220 if (!(v9iattr.valid & P9_ATTR_GID)) {
1223 err = v9fs_co_chown(pdu, &fidp->path, v9iattr.uid,
1229 if (v9iattr.valid & (P9_ATTR_SIZE)) {
1230 err = v9fs_co_truncate(pdu, &fidp->path, v9iattr.size);
1239 pdu_complete(pdu, err);
1242 static int v9fs_walk_marshal(V9fsPDU *pdu, uint16_t nwnames, V9fsQID *qids)
1248 err = pdu_marshal(pdu, offset, "w", nwnames);
1253 for (i = 0; i < nwnames; i++) {
1254 err = pdu_marshal(pdu, offset, "Q", &qids[i]);
1263 static bool name_is_illegal(const char *name)
1265 return !*name || strchr(name, '/') != NULL;
1268 static bool not_same_qid(const V9fsQID *qid1, const V9fsQID *qid2)
1271 qid1->type != qid2->type ||
1272 qid1->version != qid2->version ||
1273 qid1->path != qid2->path;
1276 static void coroutine_fn v9fs_walk(void *opaque)
1279 V9fsQID *qids = NULL;
1281 V9fsPath dpath, path;
1285 int32_t fid, newfid;
1286 V9fsString *wnames = NULL;
1288 V9fsFidState *newfidp = NULL;
1289 V9fsPDU *pdu = opaque;
1290 V9fsState *s = pdu->s;
1293 err = pdu_unmarshal(pdu, offset, "ddw", &fid, &newfid, &nwnames);
1295 pdu_complete(pdu, err);
1300 trace_v9fs_walk(pdu->tag, pdu->id, fid, newfid, nwnames);
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]);
1310 if (name_is_illegal(wnames[i].data)) {
1316 } else if (nwnames > P9_MAXWELEM) {
1320 fidp = get_fid(pdu, fid);
1326 v9fs_path_init(&dpath);
1327 v9fs_path_init(&path);
1329 err = fid_to_qid(pdu, fidp, &qid);
1335 * Both dpath and path initially poin to fidp.
1336 * Needed to handle request with nwnames == 0
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,
1349 err = v9fs_co_lstat(pdu, &path, &stbuf);
1353 stat_to_qid(&stbuf, &qid);
1354 v9fs_path_copy(&dpath, &path);
1356 memcpy(&qids[name_idx], &qid, sizeof(qid));
1358 if (fid == newfid) {
1359 if (fidp->fid_type != P9_FID_NONE) {
1363 v9fs_path_copy(&fidp->path, &path);
1365 newfidp = alloc_fid(s, newfid);
1366 if (newfidp == NULL) {
1370 newfidp->uid = fidp->uid;
1371 v9fs_path_copy(&newfidp->path, &path);
1373 err = v9fs_walk_marshal(pdu, nwnames, qids);
1374 trace_v9fs_walk_return(pdu->tag, pdu->id, nwnames, qids);
1378 put_fid(pdu, newfidp);
1380 v9fs_path_free(&dpath);
1381 v9fs_path_free(&path);
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]);
1393 static int32_t coroutine_fn get_iounit(V9fsPDU *pdu, V9fsPath *path)
1395 struct statfs stbuf;
1397 V9fsState *s = pdu->s;
1400 * iounit should be multiples of f_bsize (host filesystem block size
1401 * and as well as less than (client msize - P9_IOHDRSZ))
1403 if (!v9fs_co_statfs(pdu, path, &stbuf)) {
1404 iounit = stbuf.f_bsize;
1405 iounit *= (s->msize - P9_IOHDRSZ)/stbuf.f_bsize;
1408 iounit = s->msize - P9_IOHDRSZ;
1413 static void coroutine_fn v9fs_open(void *opaque)
1424 V9fsPDU *pdu = opaque;
1425 V9fsState *s = pdu->s;
1427 if (s->proto_version == V9FS_PROTO_2000L) {
1428 err = pdu_unmarshal(pdu, offset, "dd", &fid, &mode);
1431 err = pdu_unmarshal(pdu, offset, "db", &fid, &modebyte);
1437 trace_v9fs_open(pdu->tag, pdu->id, fid, mode);
1439 fidp = get_fid(pdu, fid);
1444 if (fidp->fid_type != P9_FID_NONE) {
1449 err = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
1453 stat_to_qid(&stbuf, &qid);
1454 if (S_ISDIR(stbuf.st_mode)) {
1455 err = v9fs_co_opendir(pdu, fidp);
1459 fidp->fid_type = P9_FID_DIR;
1460 err = pdu_marshal(pdu, offset, "Qd", &qid, 0);
1466 if (s->proto_version == V9FS_PROTO_2000L) {
1467 flags = get_dotl_openflags(s, mode);
1469 flags = omode_to_uflags(mode);
1471 if (is_ro_export(&s->ctx)) {
1472 if (mode & O_WRONLY || mode & O_RDWR ||
1473 mode & O_APPEND || mode & O_TRUNC) {
1478 err = v9fs_co_open(pdu, fidp, flags);
1482 fidp->fid_type = P9_FID_FILE;
1483 fidp->open_flags = flags;
1484 if (flags & O_EXCL) {
1486 * We let the host file system do O_EXCL check
1487 * We should not reclaim such fd
1489 fidp->flags |= FID_NON_RECLAIMABLE;
1491 iounit = get_iounit(pdu, &fidp->path);
1492 err = pdu_marshal(pdu, offset, "Qd", &qid, iounit);
1498 trace_v9fs_open_return(pdu->tag, pdu->id,
1499 qid.type, qid.version, qid.path, iounit);
1503 pdu_complete(pdu, err);
1506 static void coroutine_fn v9fs_lcreate(void *opaque)
1508 int32_t dfid, flags, mode;
1517 V9fsPDU *pdu = opaque;
1519 v9fs_string_init(&name);
1520 err = pdu_unmarshal(pdu, offset, "dsddd", &dfid,
1521 &name, &flags, &mode, &gid);
1525 trace_v9fs_lcreate(pdu->tag, pdu->id, dfid, flags, mode, gid);
1527 if (name_is_illegal(name.data)) {
1532 if (!strcmp(".", name.data) || !strcmp("..", name.data)) {
1537 fidp = get_fid(pdu, dfid);
1543 flags = get_dotl_openflags(pdu->s, flags);
1544 err = v9fs_co_open2(pdu, fidp, &name, gid,
1545 flags | O_CREAT, mode, &stbuf);
1549 fidp->fid_type = P9_FID_FILE;
1550 fidp->open_flags = flags;
1551 if (flags & O_EXCL) {
1553 * We let the host file system do O_EXCL check
1554 * We should not reclaim such fd
1556 fidp->flags |= FID_NON_RECLAIMABLE;
1558 iounit = get_iounit(pdu, &fidp->path);
1559 stat_to_qid(&stbuf, &qid);
1560 err = pdu_marshal(pdu, offset, "Qd", &qid, iounit);
1565 trace_v9fs_lcreate_return(pdu->tag, pdu->id,
1566 qid.type, qid.version, qid.path, iounit);
1570 pdu_complete(pdu, err);
1571 v9fs_string_free(&name);
1574 static void v9fs_fsync(void *opaque)
1581 V9fsPDU *pdu = opaque;
1583 err = pdu_unmarshal(pdu, offset, "dd", &fid, &datasync);
1587 trace_v9fs_fsync(pdu->tag, pdu->id, fid, datasync);
1589 fidp = get_fid(pdu, fid);
1594 err = v9fs_co_fsync(pdu, fidp, datasync);
1600 pdu_complete(pdu, err);
1603 static void coroutine_fn v9fs_clunk(void *opaque)
1609 V9fsPDU *pdu = opaque;
1610 V9fsState *s = pdu->s;
1612 err = pdu_unmarshal(pdu, offset, "d", &fid);
1616 trace_v9fs_clunk(pdu->tag, pdu->id, fid);
1618 fidp = clunk_fid(s, fid);
1624 * Bump the ref so that put_fid will
1628 err = put_fid(pdu, fidp);
1633 pdu_complete(pdu, err);
1636 static int v9fs_xattr_read(V9fsState *s, V9fsPDU *pdu, V9fsFidState *fidp,
1637 uint64_t off, uint32_t max_count)
1641 uint64_t read_count;
1642 V9fsVirtioState *v = container_of(s, V9fsVirtioState, state);
1643 VirtQueueElement *elem = v->elems[pdu->idx];
1645 if (fidp->fs.xattr.len < off) {
1648 read_count = fidp->fs.xattr.len - off;
1650 if (read_count > max_count) {
1651 read_count = max_count;
1653 err = pdu_marshal(pdu, offset, "d", read_count);
1659 err = v9fs_pack(elem->in_sg, elem->in_num, offset,
1660 ((char *)fidp->fs.xattr.value) + off,
1669 static int coroutine_fn v9fs_do_readdir_with_stat(V9fsPDU *pdu,
1678 off_t saved_dir_pos;
1679 struct dirent *dent;
1681 /* save the directory position */
1682 saved_dir_pos = v9fs_co_telldir(pdu, fidp);
1683 if (saved_dir_pos < 0) {
1684 return saved_dir_pos;
1688 v9fs_path_init(&path);
1690 v9fs_readdir_lock(&fidp->fs.dir);
1692 err = v9fs_co_readdir(pdu, fidp, &dent);
1696 err = v9fs_co_name_to_path(pdu, &fidp->path, dent->d_name, &path);
1700 err = v9fs_co_lstat(pdu, &path, &stbuf);
1704 err = stat_to_v9stat(pdu, &path, &stbuf, &v9stat);
1708 /* 11 = 7 + 4 (7 = start offset, 4 = space for storing count) */
1709 len = pdu_marshal(pdu, 11 + count, "S", &v9stat);
1711 v9fs_readdir_unlock(&fidp->fs.dir);
1713 if ((len != (v9stat.size + 2)) || ((count + len) > max_count)) {
1714 /* Ran out of buffer. Set dir back to old position and return */
1715 v9fs_co_seekdir(pdu, fidp, saved_dir_pos);
1716 v9fs_stat_free(&v9stat);
1717 v9fs_path_free(&path);
1721 v9fs_stat_free(&v9stat);
1722 v9fs_path_free(&path);
1723 saved_dir_pos = dent->d_off;
1726 v9fs_readdir_unlock(&fidp->fs.dir);
1728 v9fs_path_free(&path);
1736 * Create a QEMUIOVector for a sub-region of PDU iovecs
1738 * @qiov: uninitialized QEMUIOVector
1739 * @skip: number of bytes to skip from beginning of PDU
1740 * @size: number of bytes to include
1741 * @is_write: true - write, false - read
1743 * The resulting QEMUIOVector has heap-allocated iovecs and must be cleaned up
1744 * with qemu_iovec_destroy().
1746 static void v9fs_init_qiov_from_pdu(QEMUIOVector *qiov, V9fsPDU *pdu,
1747 size_t skip, size_t size,
1754 virtio_init_iov_from_pdu(pdu, &iov, &niov, is_write);
1756 qemu_iovec_init_external(&elem, iov, niov);
1757 qemu_iovec_init(qiov, niov);
1758 qemu_iovec_concat(qiov, &elem, skip, size);
1761 static void coroutine_fn v9fs_read(void *opaque)
1770 V9fsPDU *pdu = opaque;
1771 V9fsState *s = pdu->s;
1773 err = pdu_unmarshal(pdu, offset, "dqd", &fid, &off, &max_count);
1777 trace_v9fs_read(pdu->tag, pdu->id, fid, off, max_count);
1779 fidp = get_fid(pdu, fid);
1784 if (fidp->fid_type == P9_FID_DIR) {
1787 v9fs_co_rewinddir(pdu, fidp);
1789 count = v9fs_do_readdir_with_stat(pdu, fidp, max_count);
1794 err = pdu_marshal(pdu, offset, "d", count);
1798 err += offset + count;
1799 } else if (fidp->fid_type == P9_FID_FILE) {
1800 QEMUIOVector qiov_full;
1804 v9fs_init_qiov_from_pdu(&qiov_full, pdu, offset + 4, max_count, false);
1805 qemu_iovec_init(&qiov, qiov_full.niov);
1807 qemu_iovec_reset(&qiov);
1808 qemu_iovec_concat(&qiov, &qiov_full, count, qiov_full.size - count);
1810 print_sg(qiov.iov, qiov.niov);
1812 /* Loop in case of EINTR */
1814 len = v9fs_co_preadv(pdu, fidp, qiov.iov, qiov.niov, off);
1819 } while (len == -EINTR && !pdu->cancelled);
1821 /* IO error return the error */
1823 goto out_free_iovec;
1825 } while (count < max_count && len > 0);
1826 err = pdu_marshal(pdu, offset, "d", count);
1828 goto out_free_iovec;
1830 err += offset + count;
1832 qemu_iovec_destroy(&qiov);
1833 qemu_iovec_destroy(&qiov_full);
1834 } else if (fidp->fid_type == P9_FID_XATTR) {
1835 err = v9fs_xattr_read(s, pdu, fidp, off, max_count);
1839 trace_v9fs_read_return(pdu->tag, pdu->id, count, err);
1843 pdu_complete(pdu, err);
1846 static size_t v9fs_readdir_data_size(V9fsString *name)
1849 * Size of each dirent on the wire: size of qid (13) + size of offset (8)
1850 * size of type (1) + size of name.size (2) + strlen(name.data)
1852 return 24 + v9fs_string_size(name);
1855 static int coroutine_fn v9fs_do_readdir(V9fsPDU *pdu, V9fsFidState *fidp,
1863 off_t saved_dir_pos;
1864 struct dirent *dent;
1866 /* save the directory position */
1867 saved_dir_pos = v9fs_co_telldir(pdu, fidp);
1868 if (saved_dir_pos < 0) {
1869 return saved_dir_pos;
1873 v9fs_readdir_lock(&fidp->fs.dir);
1875 err = v9fs_co_readdir(pdu, fidp, &dent);
1879 v9fs_string_init(&name);
1880 v9fs_string_sprintf(&name, "%s", dent->d_name);
1881 if ((count + v9fs_readdir_data_size(&name)) > max_count) {
1882 v9fs_readdir_unlock(&fidp->fs.dir);
1884 /* Ran out of buffer. Set dir back to old position and return */
1885 v9fs_co_seekdir(pdu, fidp, saved_dir_pos);
1886 v9fs_string_free(&name);
1890 * Fill up just the path field of qid because the client uses
1891 * only that. To fill the entire qid structure we will have
1892 * to stat each dirent found, which is expensive
1894 size = MIN(sizeof(dent->d_ino), sizeof(qid.path));
1895 memcpy(&qid.path, &dent->d_ino, size);
1896 /* Fill the other fields with dummy values */
1900 /* 11 = 7 + 4 (7 = start offset, 4 = space for storing count) */
1901 len = pdu_marshal(pdu, 11 + count, "Qqbs",
1903 dent->d_type, &name);
1905 v9fs_readdir_unlock(&fidp->fs.dir);
1908 v9fs_co_seekdir(pdu, fidp, saved_dir_pos);
1909 v9fs_string_free(&name);
1913 v9fs_string_free(&name);
1914 saved_dir_pos = dent->d_off;
1917 v9fs_readdir_unlock(&fidp->fs.dir);
1925 static void coroutine_fn v9fs_readdir(void *opaque)
1931 uint64_t initial_offset;
1934 V9fsPDU *pdu = opaque;
1936 retval = pdu_unmarshal(pdu, offset, "dqd", &fid,
1937 &initial_offset, &max_count);
1941 trace_v9fs_readdir(pdu->tag, pdu->id, fid, initial_offset, max_count);
1943 fidp = get_fid(pdu, fid);
1948 if (!fidp->fs.dir.stream) {
1952 if (initial_offset == 0) {
1953 v9fs_co_rewinddir(pdu, fidp);
1955 v9fs_co_seekdir(pdu, fidp, initial_offset);
1957 count = v9fs_do_readdir(pdu, fidp, max_count);
1962 retval = pdu_marshal(pdu, offset, "d", count);
1966 retval += count + offset;
1967 trace_v9fs_readdir_return(pdu->tag, pdu->id, count, retval);
1971 pdu_complete(pdu, retval);
1974 static int v9fs_xattr_write(V9fsState *s, V9fsPDU *pdu, V9fsFidState *fidp,
1975 uint64_t off, uint32_t count,
1976 struct iovec *sg, int cnt)
1980 uint64_t write_count;
1984 if (fidp->fs.xattr.len < off) {
1988 write_count = fidp->fs.xattr.len - off;
1989 if (write_count > count) {
1990 write_count = count;
1992 err = pdu_marshal(pdu, offset, "d", write_count);
1997 fidp->fs.xattr.copied_len += write_count;
1999 * Now copy the content from sg list
2001 for (i = 0; i < cnt; i++) {
2002 if (write_count > sg[i].iov_len) {
2003 to_copy = sg[i].iov_len;
2005 to_copy = write_count;
2007 memcpy((char *)fidp->fs.xattr.value + off, sg[i].iov_base, to_copy);
2008 /* updating vs->off since we are not using below */
2010 write_count -= to_copy;
2016 static void coroutine_fn v9fs_write(void *opaque)
2026 V9fsPDU *pdu = opaque;
2027 V9fsState *s = pdu->s;
2028 QEMUIOVector qiov_full;
2031 err = pdu_unmarshal(pdu, offset, "dqd", &fid, &off, &count);
2033 pdu_complete(pdu, err);
2037 v9fs_init_qiov_from_pdu(&qiov_full, pdu, offset, count, true);
2038 trace_v9fs_write(pdu->tag, pdu->id, fid, off, count, qiov_full.niov);
2040 fidp = get_fid(pdu, fid);
2045 if (fidp->fid_type == P9_FID_FILE) {
2046 if (fidp->fs.fd == -1) {
2050 } else if (fidp->fid_type == P9_FID_XATTR) {
2052 * setxattr operation
2054 err = v9fs_xattr_write(s, pdu, fidp, off, count,
2055 qiov_full.iov, qiov_full.niov);
2061 qemu_iovec_init(&qiov, qiov_full.niov);
2063 qemu_iovec_reset(&qiov);
2064 qemu_iovec_concat(&qiov, &qiov_full, total, qiov_full.size - total);
2066 print_sg(qiov.iov, qiov.niov);
2068 /* Loop in case of EINTR */
2070 len = v9fs_co_pwritev(pdu, fidp, qiov.iov, qiov.niov, off);
2075 } while (len == -EINTR && !pdu->cancelled);
2077 /* IO error return the error */
2081 } while (total < count && len > 0);
2084 err = pdu_marshal(pdu, offset, "d", total);
2089 trace_v9fs_write_return(pdu->tag, pdu->id, total, err);
2091 qemu_iovec_destroy(&qiov);
2095 qemu_iovec_destroy(&qiov_full);
2096 pdu_complete(pdu, err);
2099 static void coroutine_fn v9fs_create(void *opaque)
2111 V9fsString extension;
2113 V9fsPDU *pdu = opaque;
2115 v9fs_path_init(&path);
2116 v9fs_string_init(&name);
2117 v9fs_string_init(&extension);
2118 err = pdu_unmarshal(pdu, offset, "dsdbs", &fid, &name,
2119 &perm, &mode, &extension);
2123 trace_v9fs_create(pdu->tag, pdu->id, fid, name.data, perm, mode);
2125 if (name_is_illegal(name.data)) {
2130 if (!strcmp(".", name.data) || !strcmp("..", name.data)) {
2135 fidp = get_fid(pdu, fid);
2140 if (perm & P9_STAT_MODE_DIR) {
2141 err = v9fs_co_mkdir(pdu, fidp, &name, perm & 0777,
2142 fidp->uid, -1, &stbuf);
2146 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2150 v9fs_path_copy(&fidp->path, &path);
2151 err = v9fs_co_opendir(pdu, fidp);
2155 fidp->fid_type = P9_FID_DIR;
2156 } else if (perm & P9_STAT_MODE_SYMLINK) {
2157 err = v9fs_co_symlink(pdu, fidp, &name,
2158 extension.data, -1 , &stbuf);
2162 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2166 v9fs_path_copy(&fidp->path, &path);
2167 } else if (perm & P9_STAT_MODE_LINK) {
2168 int32_t ofid = atoi(extension.data);
2169 V9fsFidState *ofidp = get_fid(pdu, ofid);
2170 if (ofidp == NULL) {
2174 err = v9fs_co_link(pdu, ofidp, fidp, &name);
2175 put_fid(pdu, ofidp);
2179 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2181 fidp->fid_type = P9_FID_NONE;
2184 v9fs_path_copy(&fidp->path, &path);
2185 err = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
2187 fidp->fid_type = P9_FID_NONE;
2190 } else if (perm & P9_STAT_MODE_DEVICE) {
2192 uint32_t major, minor;
2195 if (sscanf(extension.data, "%c %u %u", &ctype, &major, &minor) != 3) {
2212 nmode |= perm & 0777;
2213 err = v9fs_co_mknod(pdu, fidp, &name, fidp->uid, -1,
2214 makedev(major, minor), nmode, &stbuf);
2218 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2222 v9fs_path_copy(&fidp->path, &path);
2223 } else if (perm & P9_STAT_MODE_NAMED_PIPE) {
2224 err = v9fs_co_mknod(pdu, fidp, &name, fidp->uid, -1,
2225 0, S_IFIFO | (perm & 0777), &stbuf);
2229 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2233 v9fs_path_copy(&fidp->path, &path);
2234 } else if (perm & P9_STAT_MODE_SOCKET) {
2235 err = v9fs_co_mknod(pdu, fidp, &name, fidp->uid, -1,
2236 0, S_IFSOCK | (perm & 0777), &stbuf);
2240 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2244 v9fs_path_copy(&fidp->path, &path);
2246 err = v9fs_co_open2(pdu, fidp, &name, -1,
2247 omode_to_uflags(mode)|O_CREAT, perm, &stbuf);
2251 fidp->fid_type = P9_FID_FILE;
2252 fidp->open_flags = omode_to_uflags(mode);
2253 if (fidp->open_flags & O_EXCL) {
2255 * We let the host file system do O_EXCL check
2256 * We should not reclaim such fd
2258 fidp->flags |= FID_NON_RECLAIMABLE;
2261 iounit = get_iounit(pdu, &fidp->path);
2262 stat_to_qid(&stbuf, &qid);
2263 err = pdu_marshal(pdu, offset, "Qd", &qid, iounit);
2268 trace_v9fs_create_return(pdu->tag, pdu->id,
2269 qid.type, qid.version, qid.path, iounit);
2273 pdu_complete(pdu, err);
2274 v9fs_string_free(&name);
2275 v9fs_string_free(&extension);
2276 v9fs_path_free(&path);
2279 static void coroutine_fn v9fs_symlink(void *opaque)
2281 V9fsPDU *pdu = opaque;
2284 V9fsFidState *dfidp;
2292 v9fs_string_init(&name);
2293 v9fs_string_init(&symname);
2294 err = pdu_unmarshal(pdu, offset, "dssd", &dfid, &name, &symname, &gid);
2298 trace_v9fs_symlink(pdu->tag, pdu->id, dfid, name.data, symname.data, gid);
2300 if (name_is_illegal(name.data)) {
2305 if (!strcmp(".", name.data) || !strcmp("..", name.data)) {
2310 dfidp = get_fid(pdu, dfid);
2311 if (dfidp == NULL) {
2315 err = v9fs_co_symlink(pdu, dfidp, &name, symname.data, gid, &stbuf);
2319 stat_to_qid(&stbuf, &qid);
2320 err = pdu_marshal(pdu, offset, "Q", &qid);
2325 trace_v9fs_symlink_return(pdu->tag, pdu->id,
2326 qid.type, qid.version, qid.path);
2328 put_fid(pdu, dfidp);
2330 pdu_complete(pdu, err);
2331 v9fs_string_free(&name);
2332 v9fs_string_free(&symname);
2335 static void v9fs_flush(void *opaque)
2340 V9fsPDU *cancel_pdu;
2341 V9fsPDU *pdu = opaque;
2342 V9fsState *s = pdu->s;
2344 err = pdu_unmarshal(pdu, offset, "w", &tag);
2346 pdu_complete(pdu, err);
2349 trace_v9fs_flush(pdu->tag, pdu->id, tag);
2351 QLIST_FOREACH(cancel_pdu, &s->active_list, next) {
2352 if (cancel_pdu->tag == tag) {
2357 cancel_pdu->cancelled = 1;
2359 * Wait for pdu to complete.
2361 qemu_co_queue_wait(&cancel_pdu->complete);
2362 cancel_pdu->cancelled = 0;
2363 pdu_free(cancel_pdu);
2365 pdu_complete(pdu, 7);
2368 static void coroutine_fn v9fs_link(void *opaque)
2370 V9fsPDU *pdu = opaque;
2371 int32_t dfid, oldfid;
2372 V9fsFidState *dfidp, *oldfidp;
2377 v9fs_string_init(&name);
2378 err = pdu_unmarshal(pdu, offset, "dds", &dfid, &oldfid, &name);
2382 trace_v9fs_link(pdu->tag, pdu->id, dfid, oldfid, name.data);
2384 if (name_is_illegal(name.data)) {
2389 if (!strcmp(".", name.data) || !strcmp("..", name.data)) {
2394 dfidp = get_fid(pdu, dfid);
2395 if (dfidp == NULL) {
2400 oldfidp = get_fid(pdu, oldfid);
2401 if (oldfidp == NULL) {
2405 err = v9fs_co_link(pdu, oldfidp, dfidp, &name);
2409 put_fid(pdu, oldfidp);
2411 put_fid(pdu, dfidp);
2413 v9fs_string_free(&name);
2414 pdu_complete(pdu, err);
2417 /* Only works with path name based fid */
2418 static void coroutine_fn v9fs_remove(void *opaque)
2424 V9fsPDU *pdu = opaque;
2426 err = pdu_unmarshal(pdu, offset, "d", &fid);
2430 trace_v9fs_remove(pdu->tag, pdu->id, fid);
2432 fidp = get_fid(pdu, fid);
2437 /* if fs driver is not path based, return EOPNOTSUPP */
2438 if (!(pdu->s->ctx.export_flags & V9FS_PATHNAME_FSCONTEXT)) {
2443 * IF the file is unlinked, we cannot reopen
2444 * the file later. So don't reclaim fd
2446 err = v9fs_mark_fids_unreclaim(pdu, &fidp->path);
2450 err = v9fs_co_remove(pdu, &fidp->path);
2455 /* For TREMOVE we need to clunk the fid even on failed remove */
2456 clunk_fid(pdu->s, fidp->fid);
2459 pdu_complete(pdu, err);
2462 static void coroutine_fn v9fs_unlinkat(void *opaque)
2466 int32_t dfid, flags;
2469 V9fsFidState *dfidp;
2470 V9fsPDU *pdu = opaque;
2472 v9fs_string_init(&name);
2473 err = pdu_unmarshal(pdu, offset, "dsd", &dfid, &name, &flags);
2478 if (name_is_illegal(name.data)) {
2483 if (!strcmp(".", name.data)) {
2488 if (!strcmp("..", name.data)) {
2493 dfidp = get_fid(pdu, dfid);
2494 if (dfidp == NULL) {
2499 * IF the file is unlinked, we cannot reopen
2500 * the file later. So don't reclaim fd
2502 v9fs_path_init(&path);
2503 err = v9fs_co_name_to_path(pdu, &dfidp->path, name.data, &path);
2507 err = v9fs_mark_fids_unreclaim(pdu, &path);
2511 err = v9fs_co_unlinkat(pdu, &dfidp->path, &name, flags);
2516 put_fid(pdu, dfidp);
2517 v9fs_path_free(&path);
2519 pdu_complete(pdu, err);
2520 v9fs_string_free(&name);
2524 /* Only works with path name based fid */
2525 static int coroutine_fn v9fs_complete_rename(V9fsPDU *pdu, V9fsFidState *fidp,
2532 V9fsFidState *tfidp;
2533 V9fsState *s = pdu->s;
2534 V9fsFidState *dirfidp = NULL;
2535 char *old_name, *new_name;
2537 v9fs_path_init(&new_path);
2538 if (newdirfid != -1) {
2539 dirfidp = get_fid(pdu, newdirfid);
2540 if (dirfidp == NULL) {
2544 if (fidp->fid_type != P9_FID_NONE) {
2548 v9fs_co_name_to_path(pdu, &dirfidp->path, name->data, &new_path);
2550 old_name = fidp->path.data;
2551 end = strrchr(old_name, '/');
2557 new_name = g_malloc0(end - old_name + name->size + 1);
2558 strncat(new_name, old_name, end - old_name);
2559 strncat(new_name + (end - old_name), name->data, name->size);
2560 v9fs_co_name_to_path(pdu, NULL, new_name, &new_path);
2563 err = v9fs_co_rename(pdu, &fidp->path, &new_path);
2568 * Fixup fid's pointing to the old name to
2569 * start pointing to the new name
2571 for (tfidp = s->fid_list; tfidp; tfidp = tfidp->next) {
2572 if (v9fs_path_is_ancestor(&fidp->path, &tfidp->path)) {
2573 /* replace the name */
2574 v9fs_fix_path(&tfidp->path, &new_path, strlen(fidp->path.data));
2579 put_fid(pdu, dirfidp);
2581 v9fs_path_free(&new_path);
2586 /* Only works with path name based fid */
2587 static void coroutine_fn v9fs_rename(void *opaque)
2595 V9fsPDU *pdu = opaque;
2596 V9fsState *s = pdu->s;
2598 v9fs_string_init(&name);
2599 err = pdu_unmarshal(pdu, offset, "dds", &fid, &newdirfid, &name);
2604 if (name_is_illegal(name.data)) {
2609 if (!strcmp(".", name.data) || !strcmp("..", name.data)) {
2614 fidp = get_fid(pdu, fid);
2619 if (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)) {
2628 v9fs_path_write_lock(s);
2629 err = v9fs_complete_rename(pdu, fidp, newdirfid, &name);
2630 v9fs_path_unlock(s);
2637 pdu_complete(pdu, err);
2638 v9fs_string_free(&name);
2641 static void coroutine_fn v9fs_fix_fid_paths(V9fsPDU *pdu, V9fsPath *olddir,
2642 V9fsString *old_name,
2644 V9fsString *new_name)
2646 V9fsFidState *tfidp;
2647 V9fsPath oldpath, newpath;
2648 V9fsState *s = pdu->s;
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);
2657 * Fixup fid's pointing to the old name to
2658 * start pointing to the new name
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));
2666 v9fs_path_free(&oldpath);
2667 v9fs_path_free(&newpath);
2670 static int coroutine_fn v9fs_complete_renameat(V9fsPDU *pdu, int32_t olddirfid,
2671 V9fsString *old_name,
2673 V9fsString *new_name)
2676 V9fsState *s = pdu->s;
2677 V9fsFidState *newdirfidp = NULL, *olddirfidp = NULL;
2679 olddirfidp = get_fid(pdu, olddirfid);
2680 if (olddirfidp == NULL) {
2684 if (newdirfid != -1) {
2685 newdirfidp = get_fid(pdu, newdirfid);
2686 if (newdirfidp == NULL) {
2691 newdirfidp = get_fid(pdu, olddirfid);
2694 err = v9fs_co_renameat(pdu, &olddirfidp->path, old_name,
2695 &newdirfidp->path, new_name);
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);
2706 put_fid(pdu, olddirfidp);
2709 put_fid(pdu, newdirfidp);
2714 static void coroutine_fn v9fs_renameat(void *opaque)
2718 V9fsPDU *pdu = opaque;
2719 V9fsState *s = pdu->s;
2720 int32_t olddirfid, newdirfid;
2721 V9fsString old_name, new_name;
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);
2731 if (name_is_illegal(old_name.data) || name_is_illegal(new_name.data)) {
2736 if (!strcmp(".", old_name.data) || !strcmp("..", old_name.data) ||
2737 !strcmp(".", new_name.data) || !strcmp("..", new_name.data)) {
2742 v9fs_path_write_lock(s);
2743 err = v9fs_complete_renameat(pdu, olddirfid,
2744 &old_name, newdirfid, &new_name);
2745 v9fs_path_unlock(s);
2751 pdu_complete(pdu, err);
2752 v9fs_string_free(&old_name);
2753 v9fs_string_free(&new_name);
2756 static void coroutine_fn v9fs_wstat(void *opaque)
2765 V9fsPDU *pdu = opaque;
2767 v9fs_stat_init(&v9stat);
2768 err = pdu_unmarshal(pdu, offset, "dwS", &fid, &unused, &v9stat);
2772 trace_v9fs_wstat(pdu->tag, pdu->id, fid,
2773 v9stat.mode, v9stat.atime, v9stat.mtime);
2775 fidp = get_fid(pdu, fid);
2780 /* do we need to sync the file? */
2781 if (donttouch_stat(&v9stat)) {
2782 err = v9fs_co_fsync(pdu, fidp, 0);
2785 if (v9stat.mode != -1) {
2787 err = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
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 */
2798 err = v9fs_co_chmod(pdu, &fidp->path,
2799 v9mode_to_mode(v9stat.mode,
2800 &v9stat.extension));
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;
2811 times[0].tv_nsec = UTIME_OMIT;
2813 if (v9stat.mtime != -1) {
2814 times[1].tv_sec = v9stat.mtime;
2815 times[1].tv_nsec = 0;
2817 times[1].tv_nsec = UTIME_OMIT;
2819 err = v9fs_co_utimensat(pdu, &fidp->path, times);
2824 if (v9stat.n_gid != -1 || v9stat.n_uid != -1) {
2825 err = v9fs_co_chown(pdu, &fidp->path, v9stat.n_uid, v9stat.n_gid);
2830 if (v9stat.name.size != 0) {
2831 err = v9fs_complete_rename(pdu, fidp, -1, &v9stat.name);
2836 if (v9stat.length != -1) {
2837 err = v9fs_co_truncate(pdu, &fidp->path, v9stat.length);
2846 v9fs_stat_free(&v9stat);
2847 pdu_complete(pdu, err);
2850 static int v9fs_fill_statfs(V9fsState *s, V9fsPDU *pdu, struct statfs *stbuf)
2862 int32_t bsize_factor;
2865 * compute bsize factor based on host file system block size
2868 bsize_factor = (s->msize - P9_IOHDRSZ)/stbuf->f_bsize;
2869 if (!bsize_factor) {
2872 f_type = stbuf->f_type;
2873 f_bsize = stbuf->f_bsize;
2874 f_bsize *= bsize_factor;
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
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;
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);
2895 static void coroutine_fn v9fs_statfs(void *opaque)
2901 struct statfs stbuf;
2902 V9fsPDU *pdu = opaque;
2903 V9fsState *s = pdu->s;
2905 retval = pdu_unmarshal(pdu, offset, "d", &fid);
2909 fidp = get_fid(pdu, fid);
2914 retval = v9fs_co_statfs(pdu, &fidp->path, &stbuf);
2918 retval = v9fs_fill_statfs(s, pdu, &stbuf);
2926 pdu_complete(pdu, retval);
2929 static void coroutine_fn v9fs_mknod(void *opaque)
2942 V9fsPDU *pdu = opaque;
2944 v9fs_string_init(&name);
2945 err = pdu_unmarshal(pdu, offset, "dsdddd", &fid, &name, &mode,
2946 &major, &minor, &gid);
2950 trace_v9fs_mknod(pdu->tag, pdu->id, fid, mode, major, minor);
2952 if (name_is_illegal(name.data)) {
2957 if (!strcmp(".", name.data) || !strcmp("..", name.data)) {
2962 fidp = get_fid(pdu, fid);
2967 err = v9fs_co_mknod(pdu, fidp, &name, fidp->uid, gid,
2968 makedev(major, minor), mode, &stbuf);
2972 stat_to_qid(&stbuf, &qid);
2973 err = pdu_marshal(pdu, offset, "Q", &qid);
2978 trace_v9fs_mknod_return(pdu->tag, pdu->id,
2979 qid.type, qid.version, qid.path);
2983 pdu_complete(pdu, err);
2984 v9fs_string_free(&name);
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
2995 static void coroutine_fn v9fs_lock(void *opaque)
3002 int32_t fid, err = 0;
3003 V9fsPDU *pdu = opaque;
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);
3013 trace_v9fs_lock(pdu->tag, pdu->id, fid,
3014 flock.type, flock.start, flock.length);
3017 /* We support only block flag now (that too ignored currently) */
3018 if (flock.flags & ~P9_LOCK_FLAGS_BLOCK) {
3022 fidp = get_fid(pdu, fid);
3027 err = v9fs_co_fstat(pdu, fidp, &stbuf);
3031 status = P9_LOCK_SUCCESS;
3035 err = pdu_marshal(pdu, offset, "b", status);
3039 trace_v9fs_lock_return(pdu->tag, pdu->id, status);
3040 pdu_complete(pdu, err);
3041 v9fs_string_free(&flock.client_id);
3045 * When a TGETLOCK request comes, always return success because all lock
3046 * handling is done by client's VFS layer.
3048 static void coroutine_fn v9fs_getlock(void *opaque)
3054 int32_t fid, err = 0;
3055 V9fsPDU *pdu = opaque;
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,
3064 trace_v9fs_getlock(pdu->tag, pdu->id, fid,
3065 glock.type, glock.start, glock.length);
3067 fidp = get_fid(pdu, fid);
3072 err = v9fs_co_fstat(pdu, fidp, &stbuf);
3076 glock.type = P9_LOCK_TYPE_UNLCK;
3077 err = pdu_marshal(pdu, offset, "bqqds", glock.type,
3078 glock.start, glock.length, glock.proc_id,
3084 trace_v9fs_getlock_return(pdu->tag, pdu->id, glock.type, glock.start,
3085 glock.length, glock.proc_id);
3089 pdu_complete(pdu, err);
3090 v9fs_string_free(&glock.client_id);
3093 static void coroutine_fn v9fs_mkdir(void *opaque)
3095 V9fsPDU *pdu = opaque;
3106 v9fs_string_init(&name);
3107 err = pdu_unmarshal(pdu, offset, "dsdd", &fid, &name, &mode, &gid);
3111 trace_v9fs_mkdir(pdu->tag, pdu->id, fid, name.data, mode, gid);
3113 if (name_is_illegal(name.data)) {
3118 if (!strcmp(".", name.data) || !strcmp("..", name.data)) {
3123 fidp = get_fid(pdu, fid);
3128 err = v9fs_co_mkdir(pdu, fidp, &name, mode, fidp->uid, gid, &stbuf);
3132 stat_to_qid(&stbuf, &qid);
3133 err = pdu_marshal(pdu, offset, "Q", &qid);
3138 trace_v9fs_mkdir_return(pdu->tag, pdu->id,
3139 qid.type, qid.version, qid.path, err);
3143 pdu_complete(pdu, err);
3144 v9fs_string_free(&name);
3147 static void coroutine_fn v9fs_xattrwalk(void *opaque)
3153 int32_t fid, newfid;
3154 V9fsFidState *file_fidp;
3155 V9fsFidState *xattr_fidp = NULL;
3156 V9fsPDU *pdu = opaque;
3157 V9fsState *s = pdu->s;
3159 v9fs_string_init(&name);
3160 err = pdu_unmarshal(pdu, offset, "dds", &fid, &newfid, &name);
3164 trace_v9fs_xattrwalk(pdu->tag, pdu->id, fid, newfid, name.data);
3166 file_fidp = get_fid(pdu, fid);
3167 if (file_fidp == NULL) {
3171 xattr_fidp = alloc_fid(s, newfid);
3172 if (xattr_fidp == NULL) {
3176 v9fs_path_copy(&xattr_fidp->path, &file_fidp->path);
3177 if (!v9fs_string_size(&name)) {
3179 * listxattr request. Get the size first
3181 size = v9fs_co_llistxattr(pdu, &xattr_fidp->path, NULL, 0);
3184 clunk_fid(s, xattr_fidp->fid);
3188 * Read the xattr value
3190 xattr_fidp->fs.xattr.len = size;
3191 xattr_fidp->fid_type = P9_FID_XATTR;
3192 xattr_fidp->fs.xattr.xattrwalk_fid = true;
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);
3199 clunk_fid(s, xattr_fidp->fid);
3203 err = pdu_marshal(pdu, offset, "q", size);
3210 * specific xattr fid. We check for xattr
3211 * presence also collect the xattr size
3213 size = v9fs_co_lgetxattr(pdu, &xattr_fidp->path,
3217 clunk_fid(s, xattr_fidp->fid);
3221 * Read the xattr value
3223 xattr_fidp->fs.xattr.len = size;
3224 xattr_fidp->fid_type = P9_FID_XATTR;
3225 xattr_fidp->fs.xattr.xattrwalk_fid = true;
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);
3232 clunk_fid(s, xattr_fidp->fid);
3236 err = pdu_marshal(pdu, offset, "q", size);
3242 trace_v9fs_xattrwalk_return(pdu->tag, pdu->id, size);
3244 put_fid(pdu, file_fidp);
3246 put_fid(pdu, xattr_fidp);
3249 pdu_complete(pdu, err);
3250 v9fs_string_free(&name);
3253 static void coroutine_fn v9fs_xattrcreate(void *opaque)
3261 V9fsFidState *file_fidp;
3262 V9fsFidState *xattr_fidp;
3263 V9fsPDU *pdu = opaque;
3265 v9fs_string_init(&name);
3266 err = pdu_unmarshal(pdu, offset, "dsqd", &fid, &name, &size, &flags);
3270 trace_v9fs_xattrcreate(pdu->tag, pdu->id, fid, name.data, size, flags);
3272 if (size > XATTR_SIZE_MAX) {
3277 file_fidp = get_fid(pdu, fid);
3278 if (file_fidp == NULL) {
3282 if (file_fidp->fid_type != P9_FID_NONE) {
3287 /* Make the file fid point to xattr */
3288 xattr_fidp = file_fidp;
3289 xattr_fidp->fid_type = P9_FID_XATTR;
3290 xattr_fidp->fs.xattr.copied_len = 0;
3291 xattr_fidp->fs.xattr.xattrwalk_fid = false;
3292 xattr_fidp->fs.xattr.len = size;
3293 xattr_fidp->fs.xattr.flags = flags;
3294 v9fs_string_init(&xattr_fidp->fs.xattr.name);
3295 v9fs_string_copy(&xattr_fidp->fs.xattr.name, &name);
3296 xattr_fidp->fs.xattr.value = g_malloc0(size);
3299 put_fid(pdu, file_fidp);
3301 pdu_complete(pdu, err);
3302 v9fs_string_free(&name);
3305 static void coroutine_fn v9fs_readlink(void *opaque)
3307 V9fsPDU *pdu = opaque;
3314 err = pdu_unmarshal(pdu, offset, "d", &fid);
3318 trace_v9fs_readlink(pdu->tag, pdu->id, fid);
3319 fidp = get_fid(pdu, fid);
3325 v9fs_string_init(&target);
3326 err = v9fs_co_readlink(pdu, &fidp->path, &target);
3330 err = pdu_marshal(pdu, offset, "s", &target);
3332 v9fs_string_free(&target);
3336 trace_v9fs_readlink_return(pdu->tag, pdu->id, target.data);
3337 v9fs_string_free(&target);
3341 pdu_complete(pdu, err);
3344 static CoroutineEntry *pdu_co_handlers[] = {
3345 [P9_TREADDIR] = v9fs_readdir,
3346 [P9_TSTATFS] = v9fs_statfs,
3347 [P9_TGETATTR] = v9fs_getattr,
3348 [P9_TSETATTR] = v9fs_setattr,
3349 [P9_TXATTRWALK] = v9fs_xattrwalk,
3350 [P9_TXATTRCREATE] = v9fs_xattrcreate,
3351 [P9_TMKNOD] = v9fs_mknod,
3352 [P9_TRENAME] = v9fs_rename,
3353 [P9_TLOCK] = v9fs_lock,
3354 [P9_TGETLOCK] = v9fs_getlock,
3355 [P9_TRENAMEAT] = v9fs_renameat,
3356 [P9_TREADLINK] = v9fs_readlink,
3357 [P9_TUNLINKAT] = v9fs_unlinkat,
3358 [P9_TMKDIR] = v9fs_mkdir,
3359 [P9_TVERSION] = v9fs_version,
3360 [P9_TLOPEN] = v9fs_open,
3361 [P9_TATTACH] = v9fs_attach,
3362 [P9_TSTAT] = v9fs_stat,
3363 [P9_TWALK] = v9fs_walk,
3364 [P9_TCLUNK] = v9fs_clunk,
3365 [P9_TFSYNC] = v9fs_fsync,
3366 [P9_TOPEN] = v9fs_open,
3367 [P9_TREAD] = v9fs_read,
3369 [P9_TAUTH] = v9fs_auth,
3371 [P9_TFLUSH] = v9fs_flush,
3372 [P9_TLINK] = v9fs_link,
3373 [P9_TSYMLINK] = v9fs_symlink,
3374 [P9_TCREATE] = v9fs_create,
3375 [P9_TLCREATE] = v9fs_lcreate,
3376 [P9_TWRITE] = v9fs_write,
3377 [P9_TWSTAT] = v9fs_wstat,
3378 [P9_TREMOVE] = v9fs_remove,
3381 static void coroutine_fn v9fs_op_not_supp(void *opaque)
3383 V9fsPDU *pdu = opaque;
3384 pdu_complete(pdu, -EOPNOTSUPP);
3387 static void coroutine_fn v9fs_fs_ro(void *opaque)
3389 V9fsPDU *pdu = opaque;
3390 pdu_complete(pdu, -EROFS);
3393 static inline bool is_read_only_op(V9fsPDU *pdu)
3420 void pdu_submit(V9fsPDU *pdu)
3423 CoroutineEntry *handler;
3424 V9fsState *s = pdu->s;
3426 if (pdu->id >= ARRAY_SIZE(pdu_co_handlers) ||
3427 (pdu_co_handlers[pdu->id] == NULL)) {
3428 handler = v9fs_op_not_supp;
3430 handler = pdu_co_handlers[pdu->id];
3433 if (is_ro_export(&s->ctx) && !is_read_only_op(pdu)) {
3434 handler = v9fs_fs_ro;
3436 co = qemu_coroutine_create(handler, pdu);
3437 qemu_coroutine_enter(co);
3440 /* Returns 0 on success, 1 on failure. */
3441 int v9fs_device_realize_common(V9fsState *s, Error **errp)
3443 V9fsVirtioState *v = container_of(s, V9fsVirtioState, state);
3450 /* initialize pdu allocator */
3451 QLIST_INIT(&s->free_list);
3452 QLIST_INIT(&s->active_list);
3453 for (i = 0; i < (MAX_REQ - 1); i++) {
3454 QLIST_INSERT_HEAD(&s->free_list, &v->pdus[i], next);
3459 v9fs_path_init(&path);
3461 fse = get_fsdev_fsentry(s->fsconf.fsdev_id);
3464 /* We don't have a fsdev identified by fsdev_id */
3465 error_setg(errp, "9pfs device couldn't find fsdev with the "
3467 s->fsconf.fsdev_id ? s->fsconf.fsdev_id : "NULL");
3471 if (!s->fsconf.tag) {
3472 /* we haven't specified a mount_tag */
3473 error_setg(errp, "fsdev with id %s needs mount_tag arguments",
3474 s->fsconf.fsdev_id);
3478 s->ctx.export_flags = fse->export_flags;
3479 s->ctx.fs_root = g_strdup(fse->path);
3480 s->ctx.exops.get_st_gen = NULL;
3481 len = strlen(s->fsconf.tag);
3482 if (len > MAX_TAG_LEN - 1) {
3483 error_setg(errp, "mount tag '%s' (%d bytes) is longer than "
3484 "maximum (%d bytes)", s->fsconf.tag, len, MAX_TAG_LEN - 1);
3488 s->tag = g_strdup(s->fsconf.tag);
3494 qemu_co_rwlock_init(&s->rename_lock);
3496 if (s->ops->init(&s->ctx) < 0) {
3497 error_setg(errp, "9pfs Failed to initialize fs-driver with id:%s"
3498 " and export path:%s", s->fsconf.fsdev_id, s->ctx.fs_root);
3503 * Check details of export path, We need to use fs driver
3504 * call back to do that. Since we are in the init path, we don't
3505 * use co-routines here.
3507 if (s->ops->name_to_path(&s->ctx, NULL, "/", &path) < 0) {
3509 "error in converting name to path %s", strerror(errno));
3512 if (s->ops->lstat(&s->ctx, &path, &stat)) {
3513 error_setg(errp, "share path %s does not exist", fse->path);
3515 } else if (!S_ISDIR(stat.st_mode)) {
3516 error_setg(errp, "share path %s is not a directory", fse->path);
3519 v9fs_path_free(&path);
3524 if (s->ops->cleanup && s->ctx.private) {
3525 s->ops->cleanup(&s->ctx);
3528 g_free(s->ctx.fs_root);
3529 v9fs_path_free(&path);
3534 void v9fs_device_unrealize_common(V9fsState *s, Error **errp)
3536 if (s->ops->cleanup) {
3537 s->ops->cleanup(&s->ctx);
3540 g_free(s->ctx.fs_root);
3543 typedef struct VirtfsCoResetData {
3546 } VirtfsCoResetData;
3548 static void coroutine_fn virtfs_co_reset(void *opaque)
3550 VirtfsCoResetData *data = opaque;
3552 virtfs_reset(&data->pdu);
3556 void v9fs_reset(V9fsState *s)
3558 VirtfsCoResetData data = { .pdu = { .s = s }, .done = false };
3561 while (!QLIST_EMPTY(&s->active_list)) {
3562 aio_poll(qemu_get_aio_context(), true);
3565 co = qemu_coroutine_create(virtfs_co_reset, &data);
3566 qemu_coroutine_enter(co);
3568 while (!data.done) {
3569 aio_poll(qemu_get_aio_context(), true);
3573 static void __attribute__((__constructor__)) v9fs_set_fd_limit(void)
3576 if (getrlimit(RLIMIT_NOFILE, &rlim) < 0) {
3577 error_report("Failed to get the resource limit");
3580 open_fd_hw = rlim.rlim_cur - MIN(400, rlim.rlim_cur/3);
3581 open_fd_rc = rlim.rlim_cur/2;