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 "hw/virtio/virtio.h"
15 #include "hw/i386/pc.h"
16 #include "qemu/sockets.h"
17 #include "virtio-9p.h"
18 #include "fsdev/qemu-fsdev.h"
19 #include "virtio-9p-xattr.h"
20 #include "virtio-9p-coth.h"
22 #include "migration/migration.h"
26 static int open_fd_rc;
40 static int omode_to_uflags(int8_t mode)
74 struct dotl_openflag_map {
79 static int dotl_to_open_flags(int flags)
83 * We have same bits for P9_DOTL_READONLY, P9_DOTL_WRONLY
84 * and P9_DOTL_NOACCESS
86 int oflags = flags & O_ACCMODE;
88 struct dotl_openflag_map dotl_oflag_map[] = {
89 { P9_DOTL_CREATE, O_CREAT },
90 { P9_DOTL_EXCL, O_EXCL },
91 { P9_DOTL_NOCTTY , O_NOCTTY },
92 { P9_DOTL_TRUNC, O_TRUNC },
93 { P9_DOTL_APPEND, O_APPEND },
94 { P9_DOTL_NONBLOCK, O_NONBLOCK } ,
95 { P9_DOTL_DSYNC, O_DSYNC },
96 { P9_DOTL_FASYNC, FASYNC },
97 { P9_DOTL_DIRECT, O_DIRECT },
98 { P9_DOTL_LARGEFILE, O_LARGEFILE },
99 { P9_DOTL_DIRECTORY, O_DIRECTORY },
100 { P9_DOTL_NOFOLLOW, O_NOFOLLOW },
101 { P9_DOTL_NOATIME, O_NOATIME },
102 { P9_DOTL_SYNC, O_SYNC },
105 for (i = 0; i < ARRAY_SIZE(dotl_oflag_map); i++) {
106 if (flags & dotl_oflag_map[i].dotl_flag) {
107 oflags |= dotl_oflag_map[i].open_flag;
114 void cred_init(FsCred *credp)
122 static int get_dotl_openflags(V9fsState *s, int oflags)
126 * Filter the client open flags
128 flags = dotl_to_open_flags(oflags);
129 flags &= ~(O_NOCTTY | O_ASYNC | O_CREAT);
131 * Ignore direct disk access hint until the server supports it.
137 void v9fs_path_init(V9fsPath *path)
143 void v9fs_path_free(V9fsPath *path)
150 void v9fs_path_copy(V9fsPath *lhs, V9fsPath *rhs)
153 lhs->data = g_malloc(rhs->size);
154 memcpy(lhs->data, rhs->data, rhs->size);
155 lhs->size = rhs->size;
158 int v9fs_name_to_path(V9fsState *s, V9fsPath *dirpath,
159 const char *name, V9fsPath *path)
162 err = s->ops->name_to_path(&s->ctx, dirpath, name, path);
170 * Return TRUE if s1 is an ancestor of s2.
172 * E.g. "a/b" is an ancestor of "a/b/c" but not of "a/bc/d".
173 * As a special case, We treat s1 as ancestor of s2 if they are same!
175 static int v9fs_path_is_ancestor(V9fsPath *s1, V9fsPath *s2)
177 if (!strncmp(s1->data, s2->data, s1->size - 1)) {
178 if (s2->data[s1->size - 1] == '\0' || s2->data[s1->size - 1] == '/') {
185 static size_t v9fs_string_size(V9fsString *str)
191 * returns 0 if fid got re-opened, 1 if not, < 0 on error */
192 static int v9fs_reopen_fid(V9fsPDU *pdu, V9fsFidState *f)
195 if (f->fid_type == P9_FID_FILE) {
196 if (f->fs.fd == -1) {
198 err = v9fs_co_open(pdu, f, f->open_flags);
199 } while (err == -EINTR && !pdu->cancelled);
201 } else if (f->fid_type == P9_FID_DIR) {
202 if (f->fs.dir == NULL) {
204 err = v9fs_co_opendir(pdu, f);
205 } while (err == -EINTR && !pdu->cancelled);
211 static V9fsFidState *get_fid(V9fsPDU *pdu, int32_t fid)
215 V9fsState *s = pdu->s;
217 for (f = s->fid_list; f; f = f->next) {
221 * Update the fid ref upfront so that
222 * we don't get reclaimed when we yield
227 * check whether we need to reopen the
228 * file. We might have closed the fd
229 * while trying to free up some file
232 err = v9fs_reopen_fid(pdu, f);
238 * Mark the fid as referenced so that the LRU
239 * reclaim won't close the file descriptor
241 f->flags |= FID_REFERENCED;
248 static V9fsFidState *alloc_fid(V9fsState *s, int32_t fid)
252 for (f = s->fid_list; f; f = f->next) {
253 /* If fid is already there return NULL */
259 f = g_malloc0(sizeof(V9fsFidState));
261 f->fid_type = P9_FID_NONE;
264 * Mark the fid as referenced so that the LRU
265 * reclaim won't close the file descriptor
267 f->flags |= FID_REFERENCED;
268 f->next = s->fid_list;
274 static int v9fs_xattr_fid_clunk(V9fsPDU *pdu, V9fsFidState *fidp)
278 if (fidp->fs.xattr.copied_len == -1) {
279 /* getxattr/listxattr fid */
283 * if this is fid for setxattr. clunk should
284 * result in setxattr localcall
286 if (fidp->fs.xattr.len != fidp->fs.xattr.copied_len) {
287 /* clunk after partial write */
291 if (fidp->fs.xattr.len) {
292 retval = v9fs_co_lsetxattr(pdu, &fidp->path, &fidp->fs.xattr.name,
293 fidp->fs.xattr.value,
295 fidp->fs.xattr.flags);
297 retval = v9fs_co_lremovexattr(pdu, &fidp->path, &fidp->fs.xattr.name);
300 v9fs_string_free(&fidp->fs.xattr.name);
302 if (fidp->fs.xattr.value) {
303 g_free(fidp->fs.xattr.value);
308 static int free_fid(V9fsPDU *pdu, V9fsFidState *fidp)
312 if (fidp->fid_type == P9_FID_FILE) {
313 /* If we reclaimed the fd no need to close */
314 if (fidp->fs.fd != -1) {
315 retval = v9fs_co_close(pdu, &fidp->fs);
317 } else if (fidp->fid_type == P9_FID_DIR) {
318 if (fidp->fs.dir != NULL) {
319 retval = v9fs_co_closedir(pdu, &fidp->fs);
321 } else if (fidp->fid_type == P9_FID_XATTR) {
322 retval = v9fs_xattr_fid_clunk(pdu, fidp);
324 v9fs_path_free(&fidp->path);
329 static int put_fid(V9fsPDU *pdu, V9fsFidState *fidp)
334 * Don't free the fid if it is in reclaim list
336 if (!fidp->ref && fidp->clunked) {
337 if (fidp->fid == pdu->s->root_fid) {
339 * if the clunked fid is root fid then we
340 * have unmounted the fs on the client side.
341 * delete the migration blocker. Ideally, this
342 * should be hooked to transport close notification
344 if (pdu->s->migration_blocker) {
345 migrate_del_blocker(pdu->s->migration_blocker);
346 error_free(pdu->s->migration_blocker);
347 pdu->s->migration_blocker = NULL;
350 return free_fid(pdu, fidp);
355 static V9fsFidState *clunk_fid(V9fsState *s, int32_t fid)
357 V9fsFidState **fidpp, *fidp;
359 for (fidpp = &s->fid_list; *fidpp; fidpp = &(*fidpp)->next) {
360 if ((*fidpp)->fid == fid) {
364 if (*fidpp == NULL) {
373 void v9fs_reclaim_fd(V9fsPDU *pdu)
375 int reclaim_count = 0;
376 V9fsState *s = pdu->s;
377 V9fsFidState *f, *reclaim_list = NULL;
379 for (f = s->fid_list; f; f = f->next) {
381 * Unlink fids cannot be reclaimed. Check
382 * for them and skip them. Also skip fids
383 * currently being operated on.
385 if (f->ref || f->flags & FID_NON_RECLAIMABLE) {
389 * if it is a recently referenced fid
390 * we leave the fid untouched and clear the
391 * reference bit. We come back to it later
392 * in the next iteration. (a simple LRU without
393 * moving list elements around)
395 if (f->flags & FID_REFERENCED) {
396 f->flags &= ~FID_REFERENCED;
400 * Add fids to reclaim list.
402 if (f->fid_type == P9_FID_FILE) {
403 if (f->fs.fd != -1) {
405 * Up the reference count so that
406 * a clunk request won't free this fid
409 f->rclm_lst = reclaim_list;
411 f->fs_reclaim.fd = f->fs.fd;
415 } else if (f->fid_type == P9_FID_DIR) {
416 if (f->fs.dir != NULL) {
418 * Up the reference count so that
419 * a clunk request won't free this fid
422 f->rclm_lst = reclaim_list;
424 f->fs_reclaim.dir = f->fs.dir;
429 if (reclaim_count >= open_fd_rc) {
434 * Now close the fid in reclaim list. Free them if they
435 * are already clunked.
437 while (reclaim_list) {
439 reclaim_list = f->rclm_lst;
440 if (f->fid_type == P9_FID_FILE) {
441 v9fs_co_close(pdu, &f->fs_reclaim);
442 } else if (f->fid_type == P9_FID_DIR) {
443 v9fs_co_closedir(pdu, &f->fs_reclaim);
447 * Now drop the fid reference, free it
454 static int v9fs_mark_fids_unreclaim(V9fsPDU *pdu, V9fsPath *path)
457 V9fsState *s = pdu->s;
458 V9fsFidState *fidp, head_fid;
460 head_fid.next = s->fid_list;
461 for (fidp = s->fid_list; fidp; fidp = fidp->next) {
462 if (fidp->path.size != path->size) {
465 if (!memcmp(fidp->path.data, path->data, path->size)) {
466 /* Mark the fid non reclaimable. */
467 fidp->flags |= FID_NON_RECLAIMABLE;
469 /* reopen the file/dir if already closed */
470 err = v9fs_reopen_fid(pdu, fidp);
475 * Go back to head of fid list because
476 * the list could have got updated when
477 * switched to the worker thread
487 static void virtfs_reset(V9fsPDU *pdu)
489 V9fsState *s = pdu->s;
490 V9fsFidState *fidp = NULL;
493 while (s->fid_list) {
495 s->fid_list = fidp->next;
504 /* One or more unclunked fids found... */
505 error_report("9pfs:%s: One or more uncluncked fids "
506 "found during reset", __func__);
510 #define P9_QID_TYPE_DIR 0x80
511 #define P9_QID_TYPE_SYMLINK 0x02
513 #define P9_STAT_MODE_DIR 0x80000000
514 #define P9_STAT_MODE_APPEND 0x40000000
515 #define P9_STAT_MODE_EXCL 0x20000000
516 #define P9_STAT_MODE_MOUNT 0x10000000
517 #define P9_STAT_MODE_AUTH 0x08000000
518 #define P9_STAT_MODE_TMP 0x04000000
519 #define P9_STAT_MODE_SYMLINK 0x02000000
520 #define P9_STAT_MODE_LINK 0x01000000
521 #define P9_STAT_MODE_DEVICE 0x00800000
522 #define P9_STAT_MODE_NAMED_PIPE 0x00200000
523 #define P9_STAT_MODE_SOCKET 0x00100000
524 #define P9_STAT_MODE_SETUID 0x00080000
525 #define P9_STAT_MODE_SETGID 0x00040000
526 #define P9_STAT_MODE_SETVTX 0x00010000
528 #define P9_STAT_MODE_TYPE_BITS (P9_STAT_MODE_DIR | \
529 P9_STAT_MODE_SYMLINK | \
530 P9_STAT_MODE_LINK | \
531 P9_STAT_MODE_DEVICE | \
532 P9_STAT_MODE_NAMED_PIPE | \
535 /* This is the algorithm from ufs in spfs */
536 static void stat_to_qid(const struct stat *stbuf, V9fsQID *qidp)
540 memset(&qidp->path, 0, sizeof(qidp->path));
541 size = MIN(sizeof(stbuf->st_ino), sizeof(qidp->path));
542 memcpy(&qidp->path, &stbuf->st_ino, size);
543 qidp->version = stbuf->st_mtime ^ (stbuf->st_size << 8);
545 if (S_ISDIR(stbuf->st_mode)) {
546 qidp->type |= P9_QID_TYPE_DIR;
548 if (S_ISLNK(stbuf->st_mode)) {
549 qidp->type |= P9_QID_TYPE_SYMLINK;
553 static int fid_to_qid(V9fsPDU *pdu, V9fsFidState *fidp, V9fsQID *qidp)
558 err = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
562 stat_to_qid(&stbuf, qidp);
566 static V9fsPDU *alloc_pdu(V9fsState *s)
570 if (!QLIST_EMPTY(&s->free_list)) {
571 pdu = QLIST_FIRST(&s->free_list);
572 QLIST_REMOVE(pdu, next);
573 QLIST_INSERT_HEAD(&s->active_list, pdu, next);
578 static void free_pdu(V9fsState *s, V9fsPDU *pdu)
582 * Cancelled pdu are added back to the freelist
585 if (!pdu->cancelled) {
586 QLIST_REMOVE(pdu, next);
587 QLIST_INSERT_HEAD(&s->free_list, pdu, next);
593 * We don't do error checking for pdu_marshal/unmarshal here
594 * because we always expect to have enough space to encode
597 static void complete_pdu(V9fsState *s, V9fsPDU *pdu, ssize_t len)
599 int8_t id = pdu->id + 1; /* Response */
605 if (s->proto_version != V9FS_PROTO_2000L) {
608 str.data = strerror(err);
609 str.size = strlen(str.data);
611 len += pdu_marshal(pdu, len, "s", &str);
615 len += pdu_marshal(pdu, len, "d", err);
617 if (s->proto_version == V9FS_PROTO_2000L) {
620 trace_v9fs_rerror(pdu->tag, pdu->id, err); /* Trace ERROR */
623 /* fill out the header */
624 pdu_marshal(pdu, 0, "dbw", (int32_t)len, id, pdu->tag);
626 /* keep these in sync */
630 /* push onto queue and notify */
631 virtqueue_push(s->vq, &pdu->elem, len);
633 /* FIXME: we should batch these completions */
634 virtio_notify(VIRTIO_DEVICE(s), s->vq);
636 /* Now wakeup anybody waiting in flush for this request */
637 qemu_co_queue_next(&pdu->complete);
642 static mode_t v9mode_to_mode(uint32_t mode, V9fsString *extension)
647 if (mode & P9_STAT_MODE_DIR) {
651 if (mode & P9_STAT_MODE_SYMLINK) {
654 if (mode & P9_STAT_MODE_SOCKET) {
657 if (mode & P9_STAT_MODE_NAMED_PIPE) {
660 if (mode & P9_STAT_MODE_DEVICE) {
661 if (extension->size && extension->data[0] == 'c') {
672 if (mode & P9_STAT_MODE_SETUID) {
675 if (mode & P9_STAT_MODE_SETGID) {
678 if (mode & P9_STAT_MODE_SETVTX) {
685 static int donttouch_stat(V9fsStat *stat)
687 if (stat->type == -1 &&
689 stat->qid.type == -1 &&
690 stat->qid.version == -1 &&
691 stat->qid.path == -1 &&
695 stat->length == -1 &&
702 stat->n_muid == -1) {
709 static void v9fs_stat_init(V9fsStat *stat)
711 v9fs_string_init(&stat->name);
712 v9fs_string_init(&stat->uid);
713 v9fs_string_init(&stat->gid);
714 v9fs_string_init(&stat->muid);
715 v9fs_string_init(&stat->extension);
718 static void v9fs_stat_free(V9fsStat *stat)
720 v9fs_string_free(&stat->name);
721 v9fs_string_free(&stat->uid);
722 v9fs_string_free(&stat->gid);
723 v9fs_string_free(&stat->muid);
724 v9fs_string_free(&stat->extension);
727 static uint32_t stat_to_v9mode(const struct stat *stbuf)
731 mode = stbuf->st_mode & 0777;
732 if (S_ISDIR(stbuf->st_mode)) {
733 mode |= P9_STAT_MODE_DIR;
736 if (S_ISLNK(stbuf->st_mode)) {
737 mode |= P9_STAT_MODE_SYMLINK;
740 if (S_ISSOCK(stbuf->st_mode)) {
741 mode |= P9_STAT_MODE_SOCKET;
744 if (S_ISFIFO(stbuf->st_mode)) {
745 mode |= P9_STAT_MODE_NAMED_PIPE;
748 if (S_ISBLK(stbuf->st_mode) || S_ISCHR(stbuf->st_mode)) {
749 mode |= P9_STAT_MODE_DEVICE;
752 if (stbuf->st_mode & S_ISUID) {
753 mode |= P9_STAT_MODE_SETUID;
756 if (stbuf->st_mode & S_ISGID) {
757 mode |= P9_STAT_MODE_SETGID;
760 if (stbuf->st_mode & S_ISVTX) {
761 mode |= P9_STAT_MODE_SETVTX;
767 static int stat_to_v9stat(V9fsPDU *pdu, V9fsPath *name,
768 const struct stat *stbuf,
774 memset(v9stat, 0, sizeof(*v9stat));
776 stat_to_qid(stbuf, &v9stat->qid);
777 v9stat->mode = stat_to_v9mode(stbuf);
778 v9stat->atime = stbuf->st_atime;
779 v9stat->mtime = stbuf->st_mtime;
780 v9stat->length = stbuf->st_size;
782 v9fs_string_null(&v9stat->uid);
783 v9fs_string_null(&v9stat->gid);
784 v9fs_string_null(&v9stat->muid);
786 v9stat->n_uid = stbuf->st_uid;
787 v9stat->n_gid = stbuf->st_gid;
790 v9fs_string_null(&v9stat->extension);
792 if (v9stat->mode & P9_STAT_MODE_SYMLINK) {
793 err = v9fs_co_readlink(pdu, name, &v9stat->extension);
797 } else if (v9stat->mode & P9_STAT_MODE_DEVICE) {
798 v9fs_string_sprintf(&v9stat->extension, "%c %u %u",
799 S_ISCHR(stbuf->st_mode) ? 'c' : 'b',
800 major(stbuf->st_rdev), minor(stbuf->st_rdev));
801 } else if (S_ISDIR(stbuf->st_mode) || S_ISREG(stbuf->st_mode)) {
802 v9fs_string_sprintf(&v9stat->extension, "%s %lu",
803 "HARDLINKCOUNT", (unsigned long)stbuf->st_nlink);
806 str = strrchr(name->data, '/');
813 v9fs_string_sprintf(&v9stat->name, "%s", str);
816 v9fs_string_size(&v9stat->name) +
817 v9fs_string_size(&v9stat->uid) +
818 v9fs_string_size(&v9stat->gid) +
819 v9fs_string_size(&v9stat->muid) +
820 v9fs_string_size(&v9stat->extension);
824 #define P9_STATS_MODE 0x00000001ULL
825 #define P9_STATS_NLINK 0x00000002ULL
826 #define P9_STATS_UID 0x00000004ULL
827 #define P9_STATS_GID 0x00000008ULL
828 #define P9_STATS_RDEV 0x00000010ULL
829 #define P9_STATS_ATIME 0x00000020ULL
830 #define P9_STATS_MTIME 0x00000040ULL
831 #define P9_STATS_CTIME 0x00000080ULL
832 #define P9_STATS_INO 0x00000100ULL
833 #define P9_STATS_SIZE 0x00000200ULL
834 #define P9_STATS_BLOCKS 0x00000400ULL
836 #define P9_STATS_BTIME 0x00000800ULL
837 #define P9_STATS_GEN 0x00001000ULL
838 #define P9_STATS_DATA_VERSION 0x00002000ULL
840 #define P9_STATS_BASIC 0x000007ffULL /* Mask for fields up to BLOCKS */
841 #define P9_STATS_ALL 0x00003fffULL /* Mask for All fields above */
844 static void stat_to_v9stat_dotl(V9fsState *s, const struct stat *stbuf,
845 V9fsStatDotl *v9lstat)
847 memset(v9lstat, 0, sizeof(*v9lstat));
849 v9lstat->st_mode = stbuf->st_mode;
850 v9lstat->st_nlink = stbuf->st_nlink;
851 v9lstat->st_uid = stbuf->st_uid;
852 v9lstat->st_gid = stbuf->st_gid;
853 v9lstat->st_rdev = stbuf->st_rdev;
854 v9lstat->st_size = stbuf->st_size;
855 v9lstat->st_blksize = stbuf->st_blksize;
856 v9lstat->st_blocks = stbuf->st_blocks;
857 v9lstat->st_atime_sec = stbuf->st_atime;
858 v9lstat->st_atime_nsec = stbuf->st_atim.tv_nsec;
859 v9lstat->st_mtime_sec = stbuf->st_mtime;
860 v9lstat->st_mtime_nsec = stbuf->st_mtim.tv_nsec;
861 v9lstat->st_ctime_sec = stbuf->st_ctime;
862 v9lstat->st_ctime_nsec = stbuf->st_ctim.tv_nsec;
863 /* Currently we only support BASIC fields in stat */
864 v9lstat->st_result_mask = P9_STATS_BASIC;
866 stat_to_qid(stbuf, &v9lstat->qid);
869 static void print_sg(struct iovec *sg, int cnt)
873 printf("sg[%d]: {", cnt);
874 for (i = 0; i < cnt; i++) {
878 printf("(%p, %zd)", sg[i].iov_base, sg[i].iov_len);
883 /* Will call this only for path name based fid */
884 static void v9fs_fix_path(V9fsPath *dst, V9fsPath *src, int len)
887 v9fs_path_init(&str);
888 v9fs_path_copy(&str, dst);
889 v9fs_string_sprintf((V9fsString *)dst, "%s%s", src->data, str.data+len);
890 v9fs_path_free(&str);
891 /* +1 to include terminating NULL */
895 static inline bool is_ro_export(FsContext *ctx)
897 return ctx->export_flags & V9FS_RDONLY;
900 static void v9fs_version(void *opaque)
903 V9fsPDU *pdu = opaque;
904 V9fsState *s = pdu->s;
908 v9fs_string_init(&version);
909 err = pdu_unmarshal(pdu, offset, "ds", &s->msize, &version);
914 trace_v9fs_version(pdu->tag, pdu->id, s->msize, version.data);
918 if (!strcmp(version.data, "9P2000.u")) {
919 s->proto_version = V9FS_PROTO_2000U;
920 } else if (!strcmp(version.data, "9P2000.L")) {
921 s->proto_version = V9FS_PROTO_2000L;
923 v9fs_string_sprintf(&version, "unknown");
926 err = pdu_marshal(pdu, offset, "ds", s->msize, &version);
932 trace_v9fs_version_return(pdu->tag, pdu->id, s->msize, version.data);
934 complete_pdu(s, pdu, offset);
935 v9fs_string_free(&version);
938 static void v9fs_attach(void *opaque)
940 V9fsPDU *pdu = opaque;
941 V9fsState *s = pdu->s;
942 int32_t fid, afid, n_uname;
943 V9fsString uname, aname;
949 v9fs_string_init(&uname);
950 v9fs_string_init(&aname);
951 err = pdu_unmarshal(pdu, offset, "ddssd", &fid,
952 &afid, &uname, &aname, &n_uname);
956 trace_v9fs_attach(pdu->tag, pdu->id, fid, afid, uname.data, aname.data);
958 fidp = alloc_fid(s, fid);
964 err = v9fs_co_name_to_path(pdu, NULL, "/", &fidp->path);
970 err = fid_to_qid(pdu, fidp, &qid);
976 err = pdu_marshal(pdu, offset, "Q", &qid);
982 trace_v9fs_attach_return(pdu->tag, pdu->id,
983 qid.type, qid.version, qid.path);
985 * disable migration if we haven't done already.
986 * attach could get called multiple times for the same export.
988 if (!s->migration_blocker) {
990 error_set(&s->migration_blocker, QERR_VIRTFS_FEATURE_BLOCKS_MIGRATION,
991 s->ctx.fs_root ? s->ctx.fs_root : "NULL", s->tag);
992 migrate_add_blocker(s->migration_blocker);
997 complete_pdu(s, pdu, err);
998 v9fs_string_free(&uname);
999 v9fs_string_free(&aname);
1002 static void v9fs_stat(void *opaque)
1010 V9fsPDU *pdu = opaque;
1011 V9fsState *s = pdu->s;
1013 err = pdu_unmarshal(pdu, offset, "d", &fid);
1017 trace_v9fs_stat(pdu->tag, pdu->id, fid);
1019 fidp = get_fid(pdu, fid);
1024 err = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
1028 err = stat_to_v9stat(pdu, &fidp->path, &stbuf, &v9stat);
1032 err = pdu_marshal(pdu, offset, "wS", 0, &v9stat);
1034 v9fs_stat_free(&v9stat);
1037 trace_v9fs_stat_return(pdu->tag, pdu->id, v9stat.mode,
1038 v9stat.atime, v9stat.mtime, v9stat.length);
1040 v9fs_stat_free(&v9stat);
1044 complete_pdu(s, pdu, err);
1047 static void v9fs_getattr(void *opaque)
1054 uint64_t request_mask;
1055 V9fsStatDotl v9stat_dotl;
1056 V9fsPDU *pdu = opaque;
1057 V9fsState *s = pdu->s;
1059 retval = pdu_unmarshal(pdu, offset, "dq", &fid, &request_mask);
1063 trace_v9fs_getattr(pdu->tag, pdu->id, fid, request_mask);
1065 fidp = get_fid(pdu, fid);
1071 * Currently we only support BASIC fields in stat, so there is no
1072 * need to look at request_mask.
1074 retval = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
1078 stat_to_v9stat_dotl(s, &stbuf, &v9stat_dotl);
1080 /* fill st_gen if requested and supported by underlying fs */
1081 if (request_mask & P9_STATS_GEN) {
1082 retval = v9fs_co_st_gen(pdu, &fidp->path, stbuf.st_mode, &v9stat_dotl);
1086 v9stat_dotl.st_result_mask |= P9_STATS_GEN;
1088 retval = pdu_marshal(pdu, offset, "A", &v9stat_dotl);
1093 trace_v9fs_getattr_return(pdu->tag, pdu->id, v9stat_dotl.st_result_mask,
1094 v9stat_dotl.st_mode, v9stat_dotl.st_uid,
1095 v9stat_dotl.st_gid);
1099 complete_pdu(s, pdu, retval);
1102 /* Attribute flags */
1103 #define P9_ATTR_MODE (1 << 0)
1104 #define P9_ATTR_UID (1 << 1)
1105 #define P9_ATTR_GID (1 << 2)
1106 #define P9_ATTR_SIZE (1 << 3)
1107 #define P9_ATTR_ATIME (1 << 4)
1108 #define P9_ATTR_MTIME (1 << 5)
1109 #define P9_ATTR_CTIME (1 << 6)
1110 #define P9_ATTR_ATIME_SET (1 << 7)
1111 #define P9_ATTR_MTIME_SET (1 << 8)
1113 #define P9_ATTR_MASK 127
1115 static void v9fs_setattr(void *opaque)
1122 V9fsPDU *pdu = opaque;
1123 V9fsState *s = pdu->s;
1125 err = pdu_unmarshal(pdu, offset, "dI", &fid, &v9iattr);
1130 fidp = get_fid(pdu, fid);
1135 if (v9iattr.valid & P9_ATTR_MODE) {
1136 err = v9fs_co_chmod(pdu, &fidp->path, v9iattr.mode);
1141 if (v9iattr.valid & (P9_ATTR_ATIME | P9_ATTR_MTIME)) {
1142 struct timespec times[2];
1143 if (v9iattr.valid & P9_ATTR_ATIME) {
1144 if (v9iattr.valid & P9_ATTR_ATIME_SET) {
1145 times[0].tv_sec = v9iattr.atime_sec;
1146 times[0].tv_nsec = v9iattr.atime_nsec;
1148 times[0].tv_nsec = UTIME_NOW;
1151 times[0].tv_nsec = UTIME_OMIT;
1153 if (v9iattr.valid & P9_ATTR_MTIME) {
1154 if (v9iattr.valid & P9_ATTR_MTIME_SET) {
1155 times[1].tv_sec = v9iattr.mtime_sec;
1156 times[1].tv_nsec = v9iattr.mtime_nsec;
1158 times[1].tv_nsec = UTIME_NOW;
1161 times[1].tv_nsec = UTIME_OMIT;
1163 err = v9fs_co_utimensat(pdu, &fidp->path, times);
1169 * If the only valid entry in iattr is ctime we can call
1170 * chown(-1,-1) to update the ctime of the file
1172 if ((v9iattr.valid & (P9_ATTR_UID | P9_ATTR_GID)) ||
1173 ((v9iattr.valid & P9_ATTR_CTIME)
1174 && !((v9iattr.valid & P9_ATTR_MASK) & ~P9_ATTR_CTIME))) {
1175 if (!(v9iattr.valid & P9_ATTR_UID)) {
1178 if (!(v9iattr.valid & P9_ATTR_GID)) {
1181 err = v9fs_co_chown(pdu, &fidp->path, v9iattr.uid,
1187 if (v9iattr.valid & (P9_ATTR_SIZE)) {
1188 err = v9fs_co_truncate(pdu, &fidp->path, v9iattr.size);
1197 complete_pdu(s, pdu, err);
1200 static int v9fs_walk_marshal(V9fsPDU *pdu, uint16_t nwnames, V9fsQID *qids)
1206 err = pdu_marshal(pdu, offset, "w", nwnames);
1211 for (i = 0; i < nwnames; i++) {
1212 err = pdu_marshal(pdu, offset, "Q", &qids[i]);
1221 static void v9fs_walk(void *opaque)
1224 V9fsQID *qids = NULL;
1226 V9fsPath dpath, path;
1230 int32_t fid, newfid;
1231 V9fsString *wnames = NULL;
1233 V9fsFidState *newfidp = NULL;
1234 V9fsPDU *pdu = opaque;
1235 V9fsState *s = pdu->s;
1237 err = pdu_unmarshal(pdu, offset, "ddw", &fid, &newfid, &nwnames);
1239 complete_pdu(s, pdu, err);
1244 trace_v9fs_walk(pdu->tag, pdu->id, fid, newfid, nwnames);
1246 if (nwnames && nwnames <= P9_MAXWELEM) {
1247 wnames = g_malloc0(sizeof(wnames[0]) * nwnames);
1248 qids = g_malloc0(sizeof(qids[0]) * nwnames);
1249 for (i = 0; i < nwnames; i++) {
1250 err = pdu_unmarshal(pdu, offset, "s", &wnames[i]);
1256 } else if (nwnames > P9_MAXWELEM) {
1260 fidp = get_fid(pdu, fid);
1265 v9fs_path_init(&dpath);
1266 v9fs_path_init(&path);
1268 * Both dpath and path initially poin to fidp.
1269 * Needed to handle request with nwnames == 0
1271 v9fs_path_copy(&dpath, &fidp->path);
1272 v9fs_path_copy(&path, &fidp->path);
1273 for (name_idx = 0; name_idx < nwnames; name_idx++) {
1274 err = v9fs_co_name_to_path(pdu, &dpath, wnames[name_idx].data, &path);
1278 err = v9fs_co_lstat(pdu, &path, &stbuf);
1282 stat_to_qid(&stbuf, &qids[name_idx]);
1283 v9fs_path_copy(&dpath, &path);
1285 if (fid == newfid) {
1286 BUG_ON(fidp->fid_type != P9_FID_NONE);
1287 v9fs_path_copy(&fidp->path, &path);
1289 newfidp = alloc_fid(s, newfid);
1290 if (newfidp == NULL) {
1294 newfidp->uid = fidp->uid;
1295 v9fs_path_copy(&newfidp->path, &path);
1297 err = v9fs_walk_marshal(pdu, nwnames, qids);
1298 trace_v9fs_walk_return(pdu->tag, pdu->id, nwnames, qids);
1302 put_fid(pdu, newfidp);
1304 v9fs_path_free(&dpath);
1305 v9fs_path_free(&path);
1307 complete_pdu(s, pdu, err);
1308 if (nwnames && nwnames <= P9_MAXWELEM) {
1309 for (name_idx = 0; name_idx < nwnames; name_idx++) {
1310 v9fs_string_free(&wnames[name_idx]);
1317 static int32_t get_iounit(V9fsPDU *pdu, V9fsPath *path)
1319 struct statfs stbuf;
1321 V9fsState *s = pdu->s;
1324 * iounit should be multiples of f_bsize (host filesystem block size
1325 * and as well as less than (client msize - P9_IOHDRSZ))
1327 if (!v9fs_co_statfs(pdu, path, &stbuf)) {
1328 iounit = stbuf.f_bsize;
1329 iounit *= (s->msize - P9_IOHDRSZ)/stbuf.f_bsize;
1332 iounit = s->msize - P9_IOHDRSZ;
1337 static void v9fs_open(void *opaque)
1348 V9fsPDU *pdu = opaque;
1349 V9fsState *s = pdu->s;
1351 if (s->proto_version == V9FS_PROTO_2000L) {
1352 err = pdu_unmarshal(pdu, offset, "dd", &fid, &mode);
1355 err = pdu_unmarshal(pdu, offset, "db", &fid, &modebyte);
1361 trace_v9fs_open(pdu->tag, pdu->id, fid, mode);
1363 fidp = get_fid(pdu, fid);
1368 BUG_ON(fidp->fid_type != P9_FID_NONE);
1370 err = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
1374 stat_to_qid(&stbuf, &qid);
1375 if (S_ISDIR(stbuf.st_mode)) {
1376 err = v9fs_co_opendir(pdu, fidp);
1380 fidp->fid_type = P9_FID_DIR;
1381 err = pdu_marshal(pdu, offset, "Qd", &qid, 0);
1387 if (s->proto_version == V9FS_PROTO_2000L) {
1388 flags = get_dotl_openflags(s, mode);
1390 flags = omode_to_uflags(mode);
1392 if (is_ro_export(&s->ctx)) {
1393 if (mode & O_WRONLY || mode & O_RDWR ||
1394 mode & O_APPEND || mode & O_TRUNC) {
1399 err = v9fs_co_open(pdu, fidp, flags);
1403 fidp->fid_type = P9_FID_FILE;
1404 fidp->open_flags = flags;
1405 if (flags & O_EXCL) {
1407 * We let the host file system do O_EXCL check
1408 * We should not reclaim such fd
1410 fidp->flags |= FID_NON_RECLAIMABLE;
1412 iounit = get_iounit(pdu, &fidp->path);
1413 err = pdu_marshal(pdu, offset, "Qd", &qid, iounit);
1419 trace_v9fs_open_return(pdu->tag, pdu->id,
1420 qid.type, qid.version, qid.path, iounit);
1424 complete_pdu(s, pdu, err);
1427 static void v9fs_lcreate(void *opaque)
1429 int32_t dfid, flags, mode;
1438 V9fsPDU *pdu = opaque;
1440 v9fs_string_init(&name);
1441 err = pdu_unmarshal(pdu, offset, "dsddd", &dfid,
1442 &name, &flags, &mode, &gid);
1446 trace_v9fs_lcreate(pdu->tag, pdu->id, dfid, flags, mode, gid);
1448 fidp = get_fid(pdu, dfid);
1454 flags = get_dotl_openflags(pdu->s, flags);
1455 err = v9fs_co_open2(pdu, fidp, &name, gid,
1456 flags | O_CREAT, mode, &stbuf);
1460 fidp->fid_type = P9_FID_FILE;
1461 fidp->open_flags = flags;
1462 if (flags & O_EXCL) {
1464 * We let the host file system do O_EXCL check
1465 * We should not reclaim such fd
1467 fidp->flags |= FID_NON_RECLAIMABLE;
1469 iounit = get_iounit(pdu, &fidp->path);
1470 stat_to_qid(&stbuf, &qid);
1471 err = pdu_marshal(pdu, offset, "Qd", &qid, iounit);
1476 trace_v9fs_lcreate_return(pdu->tag, pdu->id,
1477 qid.type, qid.version, qid.path, iounit);
1481 complete_pdu(pdu->s, pdu, err);
1482 v9fs_string_free(&name);
1485 static void v9fs_fsync(void *opaque)
1492 V9fsPDU *pdu = opaque;
1493 V9fsState *s = pdu->s;
1495 err = pdu_unmarshal(pdu, offset, "dd", &fid, &datasync);
1499 trace_v9fs_fsync(pdu->tag, pdu->id, fid, datasync);
1501 fidp = get_fid(pdu, fid);
1506 err = v9fs_co_fsync(pdu, fidp, datasync);
1512 complete_pdu(s, pdu, err);
1515 static void v9fs_clunk(void *opaque)
1521 V9fsPDU *pdu = opaque;
1522 V9fsState *s = pdu->s;
1524 err = pdu_unmarshal(pdu, offset, "d", &fid);
1528 trace_v9fs_clunk(pdu->tag, pdu->id, fid);
1530 fidp = clunk_fid(s, fid);
1536 * Bump the ref so that put_fid will
1540 err = put_fid(pdu, fidp);
1545 complete_pdu(s, pdu, err);
1548 static int v9fs_xattr_read(V9fsState *s, V9fsPDU *pdu, V9fsFidState *fidp,
1549 uint64_t off, uint32_t max_count)
1556 xattr_len = fidp->fs.xattr.len;
1557 read_count = xattr_len - off;
1558 if (read_count > max_count) {
1559 read_count = max_count;
1560 } else if (read_count < 0) {
1562 * read beyond XATTR value
1566 err = pdu_marshal(pdu, offset, "d", read_count);
1571 err = v9fs_pack(pdu->elem.in_sg, pdu->elem.in_num, offset,
1572 ((char *)fidp->fs.xattr.value) + off,
1581 static int v9fs_do_readdir_with_stat(V9fsPDU *pdu,
1582 V9fsFidState *fidp, uint32_t max_count)
1589 off_t saved_dir_pos;
1590 struct dirent *dent, *result;
1592 /* save the directory position */
1593 saved_dir_pos = v9fs_co_telldir(pdu, fidp);
1594 if (saved_dir_pos < 0) {
1595 return saved_dir_pos;
1598 dent = g_malloc(sizeof(struct dirent));
1601 v9fs_path_init(&path);
1602 err = v9fs_co_readdir_r(pdu, fidp, dent, &result);
1603 if (err || !result) {
1606 err = v9fs_co_name_to_path(pdu, &fidp->path, dent->d_name, &path);
1610 err = v9fs_co_lstat(pdu, &path, &stbuf);
1614 err = stat_to_v9stat(pdu, &path, &stbuf, &v9stat);
1618 /* 11 = 7 + 4 (7 = start offset, 4 = space for storing count) */
1619 len = pdu_marshal(pdu, 11 + count, "S", &v9stat);
1620 if ((len != (v9stat.size + 2)) || ((count + len) > max_count)) {
1621 /* Ran out of buffer. Set dir back to old position and return */
1622 v9fs_co_seekdir(pdu, fidp, saved_dir_pos);
1623 v9fs_stat_free(&v9stat);
1624 v9fs_path_free(&path);
1629 v9fs_stat_free(&v9stat);
1630 v9fs_path_free(&path);
1631 saved_dir_pos = dent->d_off;
1635 v9fs_path_free(&path);
1643 * Create a QEMUIOVector for a sub-region of PDU iovecs
1645 * @qiov: uninitialized QEMUIOVector
1646 * @skip: number of bytes to skip from beginning of PDU
1647 * @size: number of bytes to include
1648 * @is_write: true - write, false - read
1650 * The resulting QEMUIOVector has heap-allocated iovecs and must be cleaned up
1651 * with qemu_iovec_destroy().
1653 static void v9fs_init_qiov_from_pdu(QEMUIOVector *qiov, V9fsPDU *pdu,
1654 size_t skip, size_t size,
1662 iov = pdu->elem.out_sg;
1663 niov = pdu->elem.out_num;
1665 iov = pdu->elem.in_sg;
1666 niov = pdu->elem.in_num;
1669 qemu_iovec_init_external(&elem, iov, niov);
1670 qemu_iovec_init(qiov, niov);
1671 qemu_iovec_concat(qiov, &elem, skip, size);
1674 static void v9fs_read(void *opaque)
1683 V9fsPDU *pdu = opaque;
1684 V9fsState *s = pdu->s;
1686 err = pdu_unmarshal(pdu, offset, "dqd", &fid, &off, &max_count);
1690 trace_v9fs_read(pdu->tag, pdu->id, fid, off, max_count);
1692 fidp = get_fid(pdu, fid);
1697 if (fidp->fid_type == P9_FID_DIR) {
1700 v9fs_co_rewinddir(pdu, fidp);
1702 count = v9fs_do_readdir_with_stat(pdu, fidp, max_count);
1707 err = pdu_marshal(pdu, offset, "d", count);
1711 err += offset + count;
1712 } else if (fidp->fid_type == P9_FID_FILE) {
1713 QEMUIOVector qiov_full;
1717 v9fs_init_qiov_from_pdu(&qiov_full, pdu, offset + 4, max_count, false);
1718 qemu_iovec_init(&qiov, qiov_full.niov);
1720 qemu_iovec_reset(&qiov);
1721 qemu_iovec_concat(&qiov, &qiov_full, count, qiov_full.size - count);
1723 print_sg(qiov.iov, qiov.niov);
1725 /* Loop in case of EINTR */
1727 len = v9fs_co_preadv(pdu, fidp, qiov.iov, qiov.niov, off);
1732 } while (len == -EINTR && !pdu->cancelled);
1734 /* IO error return the error */
1738 } while (count < max_count && len > 0);
1739 err = pdu_marshal(pdu, offset, "d", count);
1743 err += offset + count;
1744 qemu_iovec_destroy(&qiov);
1745 qemu_iovec_destroy(&qiov_full);
1746 } else if (fidp->fid_type == P9_FID_XATTR) {
1747 err = v9fs_xattr_read(s, pdu, fidp, off, max_count);
1751 trace_v9fs_read_return(pdu->tag, pdu->id, count, err);
1755 complete_pdu(s, pdu, err);
1758 static size_t v9fs_readdir_data_size(V9fsString *name)
1761 * Size of each dirent on the wire: size of qid (13) + size of offset (8)
1762 * size of type (1) + size of name.size (2) + strlen(name.data)
1764 return 24 + v9fs_string_size(name);
1767 static int v9fs_do_readdir(V9fsPDU *pdu,
1768 V9fsFidState *fidp, int32_t max_count)
1775 off_t saved_dir_pos;
1776 struct dirent *dent, *result;
1778 /* save the directory position */
1779 saved_dir_pos = v9fs_co_telldir(pdu, fidp);
1780 if (saved_dir_pos < 0) {
1781 return saved_dir_pos;
1784 dent = g_malloc(sizeof(struct dirent));
1787 err = v9fs_co_readdir_r(pdu, fidp, dent, &result);
1788 if (err || !result) {
1791 v9fs_string_init(&name);
1792 v9fs_string_sprintf(&name, "%s", dent->d_name);
1793 if ((count + v9fs_readdir_data_size(&name)) > max_count) {
1794 /* Ran out of buffer. Set dir back to old position and return */
1795 v9fs_co_seekdir(pdu, fidp, saved_dir_pos);
1796 v9fs_string_free(&name);
1801 * Fill up just the path field of qid because the client uses
1802 * only that. To fill the entire qid structure we will have
1803 * to stat each dirent found, which is expensive
1805 size = MIN(sizeof(dent->d_ino), sizeof(qid.path));
1806 memcpy(&qid.path, &dent->d_ino, size);
1807 /* Fill the other fields with dummy values */
1811 /* 11 = 7 + 4 (7 = start offset, 4 = space for storing count) */
1812 len = pdu_marshal(pdu, 11 + count, "Qqbs",
1814 dent->d_type, &name);
1816 v9fs_co_seekdir(pdu, fidp, saved_dir_pos);
1817 v9fs_string_free(&name);
1822 v9fs_string_free(&name);
1823 saved_dir_pos = dent->d_off;
1832 static void v9fs_readdir(void *opaque)
1838 uint64_t initial_offset;
1841 V9fsPDU *pdu = opaque;
1842 V9fsState *s = pdu->s;
1844 retval = pdu_unmarshal(pdu, offset, "dqd", &fid,
1845 &initial_offset, &max_count);
1849 trace_v9fs_readdir(pdu->tag, pdu->id, fid, initial_offset, max_count);
1851 fidp = get_fid(pdu, fid);
1856 if (!fidp->fs.dir) {
1860 if (initial_offset == 0) {
1861 v9fs_co_rewinddir(pdu, fidp);
1863 v9fs_co_seekdir(pdu, fidp, initial_offset);
1865 count = v9fs_do_readdir(pdu, fidp, max_count);
1870 retval = pdu_marshal(pdu, offset, "d", count);
1874 retval += count + offset;
1875 trace_v9fs_readdir_return(pdu->tag, pdu->id, count, retval);
1879 complete_pdu(s, pdu, retval);
1882 static int v9fs_xattr_write(V9fsState *s, V9fsPDU *pdu, V9fsFidState *fidp,
1883 uint64_t off, uint32_t count,
1884 struct iovec *sg, int cnt)
1893 xattr_len = fidp->fs.xattr.len;
1894 write_count = xattr_len - off;
1895 if (write_count > count) {
1896 write_count = count;
1897 } else if (write_count < 0) {
1899 * write beyond XATTR value len specified in
1905 err = pdu_marshal(pdu, offset, "d", write_count);
1910 fidp->fs.xattr.copied_len += write_count;
1912 * Now copy the content from sg list
1914 for (i = 0; i < cnt; i++) {
1915 if (write_count > sg[i].iov_len) {
1916 to_copy = sg[i].iov_len;
1918 to_copy = write_count;
1920 memcpy((char *)fidp->fs.xattr.value + off, sg[i].iov_base, to_copy);
1921 /* updating vs->off since we are not using below */
1923 write_count -= to_copy;
1929 static void v9fs_write(void *opaque)
1939 V9fsPDU *pdu = opaque;
1940 V9fsState *s = pdu->s;
1941 QEMUIOVector qiov_full;
1944 err = pdu_unmarshal(pdu, offset, "dqd", &fid, &off, &count);
1946 return complete_pdu(s, pdu, err);
1949 v9fs_init_qiov_from_pdu(&qiov_full, pdu, offset, count, true);
1950 trace_v9fs_write(pdu->tag, pdu->id, fid, off, count, qiov_full.niov);
1952 fidp = get_fid(pdu, fid);
1957 if (fidp->fid_type == P9_FID_FILE) {
1958 if (fidp->fs.fd == -1) {
1962 } else if (fidp->fid_type == P9_FID_XATTR) {
1964 * setxattr operation
1966 err = v9fs_xattr_write(s, pdu, fidp, off, count,
1967 qiov_full.iov, qiov_full.niov);
1973 qemu_iovec_init(&qiov, qiov_full.niov);
1975 qemu_iovec_reset(&qiov);
1976 qemu_iovec_concat(&qiov, &qiov_full, total, qiov_full.size - total);
1978 print_sg(qiov.iov, qiov.niov);
1980 /* Loop in case of EINTR */
1982 len = v9fs_co_pwritev(pdu, fidp, qiov.iov, qiov.niov, off);
1987 } while (len == -EINTR && !pdu->cancelled);
1989 /* IO error return the error */
1993 } while (total < count && len > 0);
1996 err = pdu_marshal(pdu, offset, "d", total);
2001 trace_v9fs_write_return(pdu->tag, pdu->id, total, err);
2003 qemu_iovec_destroy(&qiov);
2007 qemu_iovec_destroy(&qiov_full);
2008 complete_pdu(s, pdu, err);
2011 static void v9fs_create(void *opaque)
2023 V9fsString extension;
2025 V9fsPDU *pdu = opaque;
2027 v9fs_path_init(&path);
2028 v9fs_string_init(&name);
2029 v9fs_string_init(&extension);
2030 err = pdu_unmarshal(pdu, offset, "dsdbs", &fid, &name,
2031 &perm, &mode, &extension);
2035 trace_v9fs_create(pdu->tag, pdu->id, fid, name.data, perm, mode);
2037 fidp = get_fid(pdu, fid);
2042 if (perm & P9_STAT_MODE_DIR) {
2043 err = v9fs_co_mkdir(pdu, fidp, &name, perm & 0777,
2044 fidp->uid, -1, &stbuf);
2048 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2052 v9fs_path_copy(&fidp->path, &path);
2053 err = v9fs_co_opendir(pdu, fidp);
2057 fidp->fid_type = P9_FID_DIR;
2058 } else if (perm & P9_STAT_MODE_SYMLINK) {
2059 err = v9fs_co_symlink(pdu, fidp, &name,
2060 extension.data, -1 , &stbuf);
2064 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2068 v9fs_path_copy(&fidp->path, &path);
2069 } else if (perm & P9_STAT_MODE_LINK) {
2070 int32_t ofid = atoi(extension.data);
2071 V9fsFidState *ofidp = get_fid(pdu, ofid);
2072 if (ofidp == NULL) {
2076 err = v9fs_co_link(pdu, ofidp, fidp, &name);
2077 put_fid(pdu, ofidp);
2081 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2083 fidp->fid_type = P9_FID_NONE;
2086 v9fs_path_copy(&fidp->path, &path);
2087 err = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
2089 fidp->fid_type = P9_FID_NONE;
2092 } else if (perm & P9_STAT_MODE_DEVICE) {
2094 uint32_t major, minor;
2097 if (sscanf(extension.data, "%c %u %u", &ctype, &major, &minor) != 3) {
2114 nmode |= perm & 0777;
2115 err = v9fs_co_mknod(pdu, fidp, &name, fidp->uid, -1,
2116 makedev(major, minor), nmode, &stbuf);
2120 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2124 v9fs_path_copy(&fidp->path, &path);
2125 } else if (perm & P9_STAT_MODE_NAMED_PIPE) {
2126 err = v9fs_co_mknod(pdu, fidp, &name, fidp->uid, -1,
2127 0, S_IFIFO | (perm & 0777), &stbuf);
2131 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2135 v9fs_path_copy(&fidp->path, &path);
2136 } else if (perm & P9_STAT_MODE_SOCKET) {
2137 err = v9fs_co_mknod(pdu, fidp, &name, fidp->uid, -1,
2138 0, S_IFSOCK | (perm & 0777), &stbuf);
2142 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2146 v9fs_path_copy(&fidp->path, &path);
2148 err = v9fs_co_open2(pdu, fidp, &name, -1,
2149 omode_to_uflags(mode)|O_CREAT, perm, &stbuf);
2153 fidp->fid_type = P9_FID_FILE;
2154 fidp->open_flags = omode_to_uflags(mode);
2155 if (fidp->open_flags & O_EXCL) {
2157 * We let the host file system do O_EXCL check
2158 * We should not reclaim such fd
2160 fidp->flags |= FID_NON_RECLAIMABLE;
2163 iounit = get_iounit(pdu, &fidp->path);
2164 stat_to_qid(&stbuf, &qid);
2165 err = pdu_marshal(pdu, offset, "Qd", &qid, iounit);
2170 trace_v9fs_create_return(pdu->tag, pdu->id,
2171 qid.type, qid.version, qid.path, iounit);
2175 complete_pdu(pdu->s, pdu, err);
2176 v9fs_string_free(&name);
2177 v9fs_string_free(&extension);
2178 v9fs_path_free(&path);
2181 static void v9fs_symlink(void *opaque)
2183 V9fsPDU *pdu = opaque;
2186 V9fsFidState *dfidp;
2194 v9fs_string_init(&name);
2195 v9fs_string_init(&symname);
2196 err = pdu_unmarshal(pdu, offset, "dssd", &dfid, &name, &symname, &gid);
2200 trace_v9fs_symlink(pdu->tag, pdu->id, dfid, name.data, symname.data, gid);
2202 dfidp = get_fid(pdu, dfid);
2203 if (dfidp == NULL) {
2207 err = v9fs_co_symlink(pdu, dfidp, &name, symname.data, gid, &stbuf);
2211 stat_to_qid(&stbuf, &qid);
2212 err = pdu_marshal(pdu, offset, "Q", &qid);
2217 trace_v9fs_symlink_return(pdu->tag, pdu->id,
2218 qid.type, qid.version, qid.path);
2220 put_fid(pdu, dfidp);
2222 complete_pdu(pdu->s, pdu, err);
2223 v9fs_string_free(&name);
2224 v9fs_string_free(&symname);
2227 static void v9fs_flush(void *opaque)
2232 V9fsPDU *cancel_pdu;
2233 V9fsPDU *pdu = opaque;
2234 V9fsState *s = pdu->s;
2236 err = pdu_unmarshal(pdu, offset, "w", &tag);
2238 complete_pdu(s, pdu, err);
2241 trace_v9fs_flush(pdu->tag, pdu->id, tag);
2243 QLIST_FOREACH(cancel_pdu, &s->active_list, next) {
2244 if (cancel_pdu->tag == tag) {
2249 cancel_pdu->cancelled = 1;
2251 * Wait for pdu to complete.
2253 qemu_co_queue_wait(&cancel_pdu->complete);
2254 cancel_pdu->cancelled = 0;
2255 free_pdu(pdu->s, cancel_pdu);
2257 complete_pdu(s, pdu, 7);
2260 static void v9fs_link(void *opaque)
2262 V9fsPDU *pdu = opaque;
2263 V9fsState *s = pdu->s;
2264 int32_t dfid, oldfid;
2265 V9fsFidState *dfidp, *oldfidp;
2270 v9fs_string_init(&name);
2271 err = pdu_unmarshal(pdu, offset, "dds", &dfid, &oldfid, &name);
2275 trace_v9fs_link(pdu->tag, pdu->id, dfid, oldfid, name.data);
2277 dfidp = get_fid(pdu, dfid);
2278 if (dfidp == NULL) {
2283 oldfidp = get_fid(pdu, oldfid);
2284 if (oldfidp == NULL) {
2288 err = v9fs_co_link(pdu, oldfidp, dfidp, &name);
2293 put_fid(pdu, dfidp);
2295 v9fs_string_free(&name);
2296 complete_pdu(s, pdu, err);
2299 /* Only works with path name based fid */
2300 static void v9fs_remove(void *opaque)
2306 V9fsPDU *pdu = opaque;
2308 err = pdu_unmarshal(pdu, offset, "d", &fid);
2312 trace_v9fs_remove(pdu->tag, pdu->id, fid);
2314 fidp = get_fid(pdu, fid);
2319 /* if fs driver is not path based, return EOPNOTSUPP */
2320 if (!(pdu->s->ctx.export_flags & V9FS_PATHNAME_FSCONTEXT)) {
2325 * IF the file is unlinked, we cannot reopen
2326 * the file later. So don't reclaim fd
2328 err = v9fs_mark_fids_unreclaim(pdu, &fidp->path);
2332 err = v9fs_co_remove(pdu, &fidp->path);
2337 /* For TREMOVE we need to clunk the fid even on failed remove */
2338 clunk_fid(pdu->s, fidp->fid);
2341 complete_pdu(pdu->s, pdu, err);
2344 static void v9fs_unlinkat(void *opaque)
2348 int32_t dfid, flags;
2351 V9fsFidState *dfidp;
2352 V9fsPDU *pdu = opaque;
2354 v9fs_string_init(&name);
2355 err = pdu_unmarshal(pdu, offset, "dsd", &dfid, &name, &flags);
2359 dfidp = get_fid(pdu, dfid);
2360 if (dfidp == NULL) {
2365 * IF the file is unlinked, we cannot reopen
2366 * the file later. So don't reclaim fd
2368 v9fs_path_init(&path);
2369 err = v9fs_co_name_to_path(pdu, &dfidp->path, name.data, &path);
2373 err = v9fs_mark_fids_unreclaim(pdu, &path);
2377 err = v9fs_co_unlinkat(pdu, &dfidp->path, &name, flags);
2382 put_fid(pdu, dfidp);
2383 v9fs_path_free(&path);
2385 complete_pdu(pdu->s, pdu, err);
2386 v9fs_string_free(&name);
2390 /* Only works with path name based fid */
2391 static int v9fs_complete_rename(V9fsPDU *pdu, V9fsFidState *fidp,
2392 int32_t newdirfid, V9fsString *name)
2397 V9fsFidState *tfidp;
2398 V9fsState *s = pdu->s;
2399 V9fsFidState *dirfidp = NULL;
2400 char *old_name, *new_name;
2402 v9fs_path_init(&new_path);
2403 if (newdirfid != -1) {
2404 dirfidp = get_fid(pdu, newdirfid);
2405 if (dirfidp == NULL) {
2409 BUG_ON(dirfidp->fid_type != P9_FID_NONE);
2410 v9fs_co_name_to_path(pdu, &dirfidp->path, name->data, &new_path);
2412 old_name = fidp->path.data;
2413 end = strrchr(old_name, '/');
2419 new_name = g_malloc0(end - old_name + name->size + 1);
2420 strncat(new_name, old_name, end - old_name);
2421 strncat(new_name + (end - old_name), name->data, name->size);
2422 v9fs_co_name_to_path(pdu, NULL, new_name, &new_path);
2425 err = v9fs_co_rename(pdu, &fidp->path, &new_path);
2430 * Fixup fid's pointing to the old name to
2431 * start pointing to the new name
2433 for (tfidp = s->fid_list; tfidp; tfidp = tfidp->next) {
2434 if (v9fs_path_is_ancestor(&fidp->path, &tfidp->path)) {
2435 /* replace the name */
2436 v9fs_fix_path(&tfidp->path, &new_path, strlen(fidp->path.data));
2441 put_fid(pdu, dirfidp);
2443 v9fs_path_free(&new_path);
2448 /* Only works with path name based fid */
2449 static void v9fs_rename(void *opaque)
2457 V9fsPDU *pdu = opaque;
2458 V9fsState *s = pdu->s;
2460 v9fs_string_init(&name);
2461 err = pdu_unmarshal(pdu, offset, "dds", &fid, &newdirfid, &name);
2465 fidp = get_fid(pdu, fid);
2470 BUG_ON(fidp->fid_type != P9_FID_NONE);
2471 /* if fs driver is not path based, return EOPNOTSUPP */
2472 if (!(pdu->s->ctx.export_flags & V9FS_PATHNAME_FSCONTEXT)) {
2476 v9fs_path_write_lock(s);
2477 err = v9fs_complete_rename(pdu, fidp, newdirfid, &name);
2478 v9fs_path_unlock(s);
2485 complete_pdu(s, pdu, err);
2486 v9fs_string_free(&name);
2489 static void v9fs_fix_fid_paths(V9fsPDU *pdu, V9fsPath *olddir,
2490 V9fsString *old_name, V9fsPath *newdir,
2491 V9fsString *new_name)
2493 V9fsFidState *tfidp;
2494 V9fsPath oldpath, newpath;
2495 V9fsState *s = pdu->s;
2498 v9fs_path_init(&oldpath);
2499 v9fs_path_init(&newpath);
2500 v9fs_co_name_to_path(pdu, olddir, old_name->data, &oldpath);
2501 v9fs_co_name_to_path(pdu, newdir, new_name->data, &newpath);
2504 * Fixup fid's pointing to the old name to
2505 * start pointing to the new name
2507 for (tfidp = s->fid_list; tfidp; tfidp = tfidp->next) {
2508 if (v9fs_path_is_ancestor(&oldpath, &tfidp->path)) {
2509 /* replace the name */
2510 v9fs_fix_path(&tfidp->path, &newpath, strlen(oldpath.data));
2513 v9fs_path_free(&oldpath);
2514 v9fs_path_free(&newpath);
2517 static int v9fs_complete_renameat(V9fsPDU *pdu, int32_t olddirfid,
2518 V9fsString *old_name, int32_t newdirfid,
2519 V9fsString *new_name)
2522 V9fsState *s = pdu->s;
2523 V9fsFidState *newdirfidp = NULL, *olddirfidp = NULL;
2525 olddirfidp = get_fid(pdu, olddirfid);
2526 if (olddirfidp == NULL) {
2530 if (newdirfid != -1) {
2531 newdirfidp = get_fid(pdu, newdirfid);
2532 if (newdirfidp == NULL) {
2537 newdirfidp = get_fid(pdu, olddirfid);
2540 err = v9fs_co_renameat(pdu, &olddirfidp->path, old_name,
2541 &newdirfidp->path, new_name);
2545 if (s->ctx.export_flags & V9FS_PATHNAME_FSCONTEXT) {
2546 /* Only for path based fid we need to do the below fixup */
2547 v9fs_fix_fid_paths(pdu, &olddirfidp->path, old_name,
2548 &newdirfidp->path, new_name);
2552 put_fid(pdu, olddirfidp);
2555 put_fid(pdu, newdirfidp);
2560 static void v9fs_renameat(void *opaque)
2564 V9fsPDU *pdu = opaque;
2565 V9fsState *s = pdu->s;
2566 int32_t olddirfid, newdirfid;
2567 V9fsString old_name, new_name;
2569 v9fs_string_init(&old_name);
2570 v9fs_string_init(&new_name);
2571 err = pdu_unmarshal(pdu, offset, "dsds", &olddirfid,
2572 &old_name, &newdirfid, &new_name);
2577 v9fs_path_write_lock(s);
2578 err = v9fs_complete_renameat(pdu, olddirfid,
2579 &old_name, newdirfid, &new_name);
2580 v9fs_path_unlock(s);
2586 complete_pdu(s, pdu, err);
2587 v9fs_string_free(&old_name);
2588 v9fs_string_free(&new_name);
2591 static void v9fs_wstat(void *opaque)
2600 V9fsPDU *pdu = opaque;
2601 V9fsState *s = pdu->s;
2603 v9fs_stat_init(&v9stat);
2604 err = pdu_unmarshal(pdu, offset, "dwS", &fid, &unused, &v9stat);
2608 trace_v9fs_wstat(pdu->tag, pdu->id, fid,
2609 v9stat.mode, v9stat.atime, v9stat.mtime);
2611 fidp = get_fid(pdu, fid);
2616 /* do we need to sync the file? */
2617 if (donttouch_stat(&v9stat)) {
2618 err = v9fs_co_fsync(pdu, fidp, 0);
2621 if (v9stat.mode != -1) {
2623 err = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
2627 v9_mode = stat_to_v9mode(&stbuf);
2628 if ((v9stat.mode & P9_STAT_MODE_TYPE_BITS) !=
2629 (v9_mode & P9_STAT_MODE_TYPE_BITS)) {
2630 /* Attempting to change the type */
2634 err = v9fs_co_chmod(pdu, &fidp->path,
2635 v9mode_to_mode(v9stat.mode,
2636 &v9stat.extension));
2641 if (v9stat.mtime != -1 || v9stat.atime != -1) {
2642 struct timespec times[2];
2643 if (v9stat.atime != -1) {
2644 times[0].tv_sec = v9stat.atime;
2645 times[0].tv_nsec = 0;
2647 times[0].tv_nsec = UTIME_OMIT;
2649 if (v9stat.mtime != -1) {
2650 times[1].tv_sec = v9stat.mtime;
2651 times[1].tv_nsec = 0;
2653 times[1].tv_nsec = UTIME_OMIT;
2655 err = v9fs_co_utimensat(pdu, &fidp->path, times);
2660 if (v9stat.n_gid != -1 || v9stat.n_uid != -1) {
2661 err = v9fs_co_chown(pdu, &fidp->path, v9stat.n_uid, v9stat.n_gid);
2666 if (v9stat.name.size != 0) {
2667 err = v9fs_complete_rename(pdu, fidp, -1, &v9stat.name);
2672 if (v9stat.length != -1) {
2673 err = v9fs_co_truncate(pdu, &fidp->path, v9stat.length);
2682 v9fs_stat_free(&v9stat);
2683 complete_pdu(s, pdu, err);
2686 static int v9fs_fill_statfs(V9fsState *s, V9fsPDU *pdu, struct statfs *stbuf)
2698 int32_t bsize_factor;
2701 * compute bsize factor based on host file system block size
2704 bsize_factor = (s->msize - P9_IOHDRSZ)/stbuf->f_bsize;
2705 if (!bsize_factor) {
2708 f_type = stbuf->f_type;
2709 f_bsize = stbuf->f_bsize;
2710 f_bsize *= bsize_factor;
2712 * f_bsize is adjusted(multiplied) by bsize factor, so we need to
2713 * adjust(divide) the number of blocks, free blocks and available
2714 * blocks by bsize factor
2716 f_blocks = stbuf->f_blocks/bsize_factor;
2717 f_bfree = stbuf->f_bfree/bsize_factor;
2718 f_bavail = stbuf->f_bavail/bsize_factor;
2719 f_files = stbuf->f_files;
2720 f_ffree = stbuf->f_ffree;
2721 fsid_val = (unsigned int) stbuf->f_fsid.__val[0] |
2722 (unsigned long long)stbuf->f_fsid.__val[1] << 32;
2723 f_namelen = stbuf->f_namelen;
2725 return pdu_marshal(pdu, offset, "ddqqqqqqd",
2726 f_type, f_bsize, f_blocks, f_bfree,
2727 f_bavail, f_files, f_ffree,
2728 fsid_val, f_namelen);
2731 static void v9fs_statfs(void *opaque)
2737 struct statfs stbuf;
2738 V9fsPDU *pdu = opaque;
2739 V9fsState *s = pdu->s;
2741 retval = pdu_unmarshal(pdu, offset, "d", &fid);
2745 fidp = get_fid(pdu, fid);
2750 retval = v9fs_co_statfs(pdu, &fidp->path, &stbuf);
2754 retval = v9fs_fill_statfs(s, pdu, &stbuf);
2762 complete_pdu(s, pdu, retval);
2765 static void v9fs_mknod(void *opaque)
2778 V9fsPDU *pdu = opaque;
2779 V9fsState *s = pdu->s;
2781 v9fs_string_init(&name);
2782 err = pdu_unmarshal(pdu, offset, "dsdddd", &fid, &name, &mode,
2783 &major, &minor, &gid);
2787 trace_v9fs_mknod(pdu->tag, pdu->id, fid, mode, major, minor);
2789 fidp = get_fid(pdu, fid);
2794 err = v9fs_co_mknod(pdu, fidp, &name, fidp->uid, gid,
2795 makedev(major, minor), mode, &stbuf);
2799 stat_to_qid(&stbuf, &qid);
2800 err = pdu_marshal(pdu, offset, "Q", &qid);
2805 trace_v9fs_mknod_return(pdu->tag, pdu->id,
2806 qid.type, qid.version, qid.path);
2810 complete_pdu(s, pdu, err);
2811 v9fs_string_free(&name);
2815 * Implement posix byte range locking code
2816 * Server side handling of locking code is very simple, because 9p server in
2817 * QEMU can handle only one client. And most of the lock handling
2818 * (like conflict, merging) etc is done by the VFS layer itself, so no need to
2819 * do any thing in * qemu 9p server side lock code path.
2820 * So when a TLOCK request comes, always return success
2822 static void v9fs_lock(void *opaque)
2829 int32_t fid, err = 0;
2830 V9fsPDU *pdu = opaque;
2831 V9fsState *s = pdu->s;
2833 status = P9_LOCK_ERROR;
2834 v9fs_string_init(&flock.client_id);
2835 err = pdu_unmarshal(pdu, offset, "dbdqqds", &fid, &flock.type,
2836 &flock.flags, &flock.start, &flock.length,
2837 &flock.proc_id, &flock.client_id);
2841 trace_v9fs_lock(pdu->tag, pdu->id, fid,
2842 flock.type, flock.start, flock.length);
2845 /* We support only block flag now (that too ignored currently) */
2846 if (flock.flags & ~P9_LOCK_FLAGS_BLOCK) {
2850 fidp = get_fid(pdu, fid);
2855 err = v9fs_co_fstat(pdu, fidp, &stbuf);
2859 status = P9_LOCK_SUCCESS;
2863 err = pdu_marshal(pdu, offset, "b", status);
2867 trace_v9fs_lock_return(pdu->tag, pdu->id, status);
2868 complete_pdu(s, pdu, err);
2869 v9fs_string_free(&flock.client_id);
2873 * When a TGETLOCK request comes, always return success because all lock
2874 * handling is done by client's VFS layer.
2876 static void v9fs_getlock(void *opaque)
2882 int32_t fid, err = 0;
2883 V9fsPDU *pdu = opaque;
2884 V9fsState *s = pdu->s;
2886 v9fs_string_init(&glock.client_id);
2887 err = pdu_unmarshal(pdu, offset, "dbqqds", &fid, &glock.type,
2888 &glock.start, &glock.length, &glock.proc_id,
2893 trace_v9fs_getlock(pdu->tag, pdu->id, fid,
2894 glock.type, glock.start, glock.length);
2896 fidp = get_fid(pdu, fid);
2901 err = v9fs_co_fstat(pdu, fidp, &stbuf);
2905 glock.type = P9_LOCK_TYPE_UNLCK;
2906 err = pdu_marshal(pdu, offset, "bqqds", glock.type,
2907 glock.start, glock.length, glock.proc_id,
2913 trace_v9fs_getlock_return(pdu->tag, pdu->id, glock.type, glock.start,
2914 glock.length, glock.proc_id);
2918 complete_pdu(s, pdu, err);
2919 v9fs_string_free(&glock.client_id);
2922 static void v9fs_mkdir(void *opaque)
2924 V9fsPDU *pdu = opaque;
2935 v9fs_string_init(&name);
2936 err = pdu_unmarshal(pdu, offset, "dsdd", &fid, &name, &mode, &gid);
2940 trace_v9fs_mkdir(pdu->tag, pdu->id, fid, name.data, mode, gid);
2942 fidp = get_fid(pdu, fid);
2947 err = v9fs_co_mkdir(pdu, fidp, &name, mode, fidp->uid, gid, &stbuf);
2951 stat_to_qid(&stbuf, &qid);
2952 err = pdu_marshal(pdu, offset, "Q", &qid);
2957 trace_v9fs_mkdir_return(pdu->tag, pdu->id,
2958 qid.type, qid.version, qid.path, err);
2962 complete_pdu(pdu->s, pdu, err);
2963 v9fs_string_free(&name);
2966 static void v9fs_xattrwalk(void *opaque)
2972 int32_t fid, newfid;
2973 V9fsFidState *file_fidp;
2974 V9fsFidState *xattr_fidp = NULL;
2975 V9fsPDU *pdu = opaque;
2976 V9fsState *s = pdu->s;
2978 v9fs_string_init(&name);
2979 err = pdu_unmarshal(pdu, offset, "dds", &fid, &newfid, &name);
2983 trace_v9fs_xattrwalk(pdu->tag, pdu->id, fid, newfid, name.data);
2985 file_fidp = get_fid(pdu, fid);
2986 if (file_fidp == NULL) {
2990 xattr_fidp = alloc_fid(s, newfid);
2991 if (xattr_fidp == NULL) {
2995 v9fs_path_copy(&xattr_fidp->path, &file_fidp->path);
2996 if (name.data == NULL) {
2998 * listxattr request. Get the size first
3000 size = v9fs_co_llistxattr(pdu, &xattr_fidp->path, NULL, 0);
3003 clunk_fid(s, xattr_fidp->fid);
3007 * Read the xattr value
3009 xattr_fidp->fs.xattr.len = size;
3010 xattr_fidp->fid_type = P9_FID_XATTR;
3011 xattr_fidp->fs.xattr.copied_len = -1;
3013 xattr_fidp->fs.xattr.value = g_malloc(size);
3014 err = v9fs_co_llistxattr(pdu, &xattr_fidp->path,
3015 xattr_fidp->fs.xattr.value,
3016 xattr_fidp->fs.xattr.len);
3018 clunk_fid(s, xattr_fidp->fid);
3022 err = pdu_marshal(pdu, offset, "q", size);
3029 * specific xattr fid. We check for xattr
3030 * presence also collect the xattr size
3032 size = v9fs_co_lgetxattr(pdu, &xattr_fidp->path,
3036 clunk_fid(s, xattr_fidp->fid);
3040 * Read the xattr value
3042 xattr_fidp->fs.xattr.len = size;
3043 xattr_fidp->fid_type = P9_FID_XATTR;
3044 xattr_fidp->fs.xattr.copied_len = -1;
3046 xattr_fidp->fs.xattr.value = g_malloc(size);
3047 err = v9fs_co_lgetxattr(pdu, &xattr_fidp->path,
3048 &name, xattr_fidp->fs.xattr.value,
3049 xattr_fidp->fs.xattr.len);
3051 clunk_fid(s, xattr_fidp->fid);
3055 err = pdu_marshal(pdu, offset, "q", size);
3061 trace_v9fs_xattrwalk_return(pdu->tag, pdu->id, size);
3063 put_fid(pdu, file_fidp);
3065 put_fid(pdu, xattr_fidp);
3068 complete_pdu(s, pdu, err);
3069 v9fs_string_free(&name);
3072 static void v9fs_xattrcreate(void *opaque)
3080 V9fsFidState *file_fidp;
3081 V9fsFidState *xattr_fidp;
3082 V9fsPDU *pdu = opaque;
3083 V9fsState *s = pdu->s;
3085 v9fs_string_init(&name);
3086 err = pdu_unmarshal(pdu, offset, "dsqd", &fid, &name, &size, &flags);
3090 trace_v9fs_xattrcreate(pdu->tag, pdu->id, fid, name.data, size, flags);
3092 file_fidp = get_fid(pdu, fid);
3093 if (file_fidp == NULL) {
3097 /* Make the file fid point to xattr */
3098 xattr_fidp = file_fidp;
3099 xattr_fidp->fid_type = P9_FID_XATTR;
3100 xattr_fidp->fs.xattr.copied_len = 0;
3101 xattr_fidp->fs.xattr.len = size;
3102 xattr_fidp->fs.xattr.flags = flags;
3103 v9fs_string_init(&xattr_fidp->fs.xattr.name);
3104 v9fs_string_copy(&xattr_fidp->fs.xattr.name, &name);
3105 xattr_fidp->fs.xattr.value = g_malloc(size);
3107 put_fid(pdu, file_fidp);
3109 complete_pdu(s, pdu, err);
3110 v9fs_string_free(&name);
3113 static void v9fs_readlink(void *opaque)
3115 V9fsPDU *pdu = opaque;
3122 err = pdu_unmarshal(pdu, offset, "d", &fid);
3126 trace_v9fs_readlink(pdu->tag, pdu->id, fid);
3127 fidp = get_fid(pdu, fid);
3133 v9fs_string_init(&target);
3134 err = v9fs_co_readlink(pdu, &fidp->path, &target);
3138 err = pdu_marshal(pdu, offset, "s", &target);
3140 v9fs_string_free(&target);
3144 trace_v9fs_readlink_return(pdu->tag, pdu->id, target.data);
3145 v9fs_string_free(&target);
3149 complete_pdu(pdu->s, pdu, err);
3152 static CoroutineEntry *pdu_co_handlers[] = {
3153 [P9_TREADDIR] = v9fs_readdir,
3154 [P9_TSTATFS] = v9fs_statfs,
3155 [P9_TGETATTR] = v9fs_getattr,
3156 [P9_TSETATTR] = v9fs_setattr,
3157 [P9_TXATTRWALK] = v9fs_xattrwalk,
3158 [P9_TXATTRCREATE] = v9fs_xattrcreate,
3159 [P9_TMKNOD] = v9fs_mknod,
3160 [P9_TRENAME] = v9fs_rename,
3161 [P9_TLOCK] = v9fs_lock,
3162 [P9_TGETLOCK] = v9fs_getlock,
3163 [P9_TRENAMEAT] = v9fs_renameat,
3164 [P9_TREADLINK] = v9fs_readlink,
3165 [P9_TUNLINKAT] = v9fs_unlinkat,
3166 [P9_TMKDIR] = v9fs_mkdir,
3167 [P9_TVERSION] = v9fs_version,
3168 [P9_TLOPEN] = v9fs_open,
3169 [P9_TATTACH] = v9fs_attach,
3170 [P9_TSTAT] = v9fs_stat,
3171 [P9_TWALK] = v9fs_walk,
3172 [P9_TCLUNK] = v9fs_clunk,
3173 [P9_TFSYNC] = v9fs_fsync,
3174 [P9_TOPEN] = v9fs_open,
3175 [P9_TREAD] = v9fs_read,
3177 [P9_TAUTH] = v9fs_auth,
3179 [P9_TFLUSH] = v9fs_flush,
3180 [P9_TLINK] = v9fs_link,
3181 [P9_TSYMLINK] = v9fs_symlink,
3182 [P9_TCREATE] = v9fs_create,
3183 [P9_TLCREATE] = v9fs_lcreate,
3184 [P9_TWRITE] = v9fs_write,
3185 [P9_TWSTAT] = v9fs_wstat,
3186 [P9_TREMOVE] = v9fs_remove,
3189 static void v9fs_op_not_supp(void *opaque)
3191 V9fsPDU *pdu = opaque;
3192 complete_pdu(pdu->s, pdu, -EOPNOTSUPP);
3195 static void v9fs_fs_ro(void *opaque)
3197 V9fsPDU *pdu = opaque;
3198 complete_pdu(pdu->s, pdu, -EROFS);
3201 static inline bool is_read_only_op(V9fsPDU *pdu)
3228 static void submit_pdu(V9fsState *s, V9fsPDU *pdu)
3231 CoroutineEntry *handler;
3233 if (pdu->id >= ARRAY_SIZE(pdu_co_handlers) ||
3234 (pdu_co_handlers[pdu->id] == NULL)) {
3235 handler = v9fs_op_not_supp;
3237 handler = pdu_co_handlers[pdu->id];
3240 if (is_ro_export(&s->ctx) && !is_read_only_op(pdu)) {
3241 handler = v9fs_fs_ro;
3243 co = qemu_coroutine_create(handler);
3244 qemu_coroutine_enter(co, pdu);
3247 void handle_9p_output(VirtIODevice *vdev, VirtQueue *vq)
3249 V9fsState *s = (V9fsState *)vdev;
3253 while ((pdu = alloc_pdu(s)) &&
3254 (len = virtqueue_pop(vq, &pdu->elem)) != 0) {
3257 BUG_ON(pdu->elem.out_num == 0 || pdu->elem.in_num == 0);
3258 BUG_ON(pdu->elem.out_sg[0].iov_len < 7);
3260 ptr = pdu->elem.out_sg[0].iov_base;
3262 pdu->size = le32_to_cpu(*(uint32_t *)ptr);
3264 pdu->tag = le16_to_cpu(*(uint16_t *)(ptr + 5));
3265 qemu_co_queue_init(&pdu->complete);
3271 static void __attribute__((__constructor__)) virtio_9p_set_fd_limit(void)
3274 if (getrlimit(RLIMIT_NOFILE, &rlim) < 0) {
3275 fprintf(stderr, "Failed to get the resource limit\n");
3278 open_fd_hw = rlim.rlim_cur - MIN(400, rlim.rlim_cur/3);
3279 open_fd_rc = rlim.rlim_cur/2;