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.h"
16 #include "qemu_socket.h"
17 #include "hw/virtio-pci.h"
18 #include "virtio-9p.h"
19 #include "fsdev/qemu-fsdev.h"
20 #include "virtio-9p-xattr.h"
21 #include "virtio-9p-coth.h"
23 #include "migration.h"
27 static int open_fd_rc;
41 static int omode_to_uflags(int8_t mode)
75 struct dotl_openflag_map {
80 static int dotl_to_open_flags(int flags)
84 * We have same bits for P9_DOTL_READONLY, P9_DOTL_WRONLY
85 * and P9_DOTL_NOACCESS
87 int oflags = flags & O_ACCMODE;
89 struct dotl_openflag_map dotl_oflag_map[] = {
90 { P9_DOTL_CREATE, O_CREAT },
91 { P9_DOTL_EXCL, O_EXCL },
92 { P9_DOTL_NOCTTY , O_NOCTTY },
93 { P9_DOTL_TRUNC, O_TRUNC },
94 { P9_DOTL_APPEND, O_APPEND },
95 { P9_DOTL_NONBLOCK, O_NONBLOCK } ,
96 { P9_DOTL_DSYNC, O_DSYNC },
97 { P9_DOTL_FASYNC, FASYNC },
98 { P9_DOTL_DIRECT, O_DIRECT },
99 { P9_DOTL_LARGEFILE, O_LARGEFILE },
100 { P9_DOTL_DIRECTORY, O_DIRECTORY },
101 { P9_DOTL_NOFOLLOW, O_NOFOLLOW },
102 { P9_DOTL_NOATIME, O_NOATIME },
103 { P9_DOTL_SYNC, O_SYNC },
106 for (i = 0; i < ARRAY_SIZE(dotl_oflag_map); i++) {
107 if (flags & dotl_oflag_map[i].dotl_flag) {
108 oflags |= dotl_oflag_map[i].open_flag;
115 void cred_init(FsCred *credp)
123 static int get_dotl_openflags(V9fsState *s, int oflags)
127 * Filter the client open flags
129 flags = dotl_to_open_flags(oflags);
130 flags &= ~(O_NOCTTY | O_ASYNC | O_CREAT);
132 * Ignore direct disk access hint until the server supports it.
138 void v9fs_path_init(V9fsPath *path)
144 void v9fs_path_free(V9fsPath *path)
151 void v9fs_path_copy(V9fsPath *lhs, V9fsPath *rhs)
154 lhs->data = g_malloc(rhs->size);
155 memcpy(lhs->data, rhs->data, rhs->size);
156 lhs->size = rhs->size;
159 int v9fs_name_to_path(V9fsState *s, V9fsPath *dirpath,
160 const char *name, V9fsPath *path)
163 err = s->ops->name_to_path(&s->ctx, dirpath, name, path);
171 * Return TRUE if s1 is an ancestor of s2.
173 * E.g. "a/b" is an ancestor of "a/b/c" but not of "a/bc/d".
174 * As a special case, We treat s1 as ancestor of s2 if they are same!
176 static int v9fs_path_is_ancestor(V9fsPath *s1, V9fsPath *s2)
178 if (!strncmp(s1->data, s2->data, s1->size - 1)) {
179 if (s2->data[s1->size - 1] == '\0' || s2->data[s1->size - 1] == '/') {
186 static size_t v9fs_string_size(V9fsString *str)
192 * returns 0 if fid got re-opened, 1 if not, < 0 on error */
193 static int v9fs_reopen_fid(V9fsPDU *pdu, V9fsFidState *f)
196 if (f->fid_type == P9_FID_FILE) {
197 if (f->fs.fd == -1) {
199 err = v9fs_co_open(pdu, f, f->open_flags);
200 } while (err == -EINTR && !pdu->cancelled);
202 } else if (f->fid_type == P9_FID_DIR) {
203 if (f->fs.dir == NULL) {
205 err = v9fs_co_opendir(pdu, f);
206 } while (err == -EINTR && !pdu->cancelled);
212 static V9fsFidState *get_fid(V9fsPDU *pdu, int32_t fid)
216 V9fsState *s = pdu->s;
218 for (f = s->fid_list; f; f = f->next) {
222 * Update the fid ref upfront so that
223 * we don't get reclaimed when we yield
228 * check whether we need to reopen the
229 * file. We might have closed the fd
230 * while trying to free up some file
233 err = v9fs_reopen_fid(pdu, f);
239 * Mark the fid as referenced so that the LRU
240 * reclaim won't close the file descriptor
242 f->flags |= FID_REFERENCED;
249 static V9fsFidState *alloc_fid(V9fsState *s, int32_t fid)
253 for (f = s->fid_list; f; f = f->next) {
254 /* If fid is already there return NULL */
260 f = g_malloc0(sizeof(V9fsFidState));
262 f->fid_type = P9_FID_NONE;
265 * Mark the fid as referenced so that the LRU
266 * reclaim won't close the file descriptor
268 f->flags |= FID_REFERENCED;
269 f->next = s->fid_list;
275 static int v9fs_xattr_fid_clunk(V9fsPDU *pdu, V9fsFidState *fidp)
279 if (fidp->fs.xattr.copied_len == -1) {
280 /* getxattr/listxattr fid */
284 * if this is fid for setxattr. clunk should
285 * result in setxattr localcall
287 if (fidp->fs.xattr.len != fidp->fs.xattr.copied_len) {
288 /* clunk after partial write */
292 if (fidp->fs.xattr.len) {
293 retval = v9fs_co_lsetxattr(pdu, &fidp->path, &fidp->fs.xattr.name,
294 fidp->fs.xattr.value,
296 fidp->fs.xattr.flags);
298 retval = v9fs_co_lremovexattr(pdu, &fidp->path, &fidp->fs.xattr.name);
301 v9fs_string_free(&fidp->fs.xattr.name);
303 if (fidp->fs.xattr.value) {
304 g_free(fidp->fs.xattr.value);
309 static int free_fid(V9fsPDU *pdu, V9fsFidState *fidp)
313 if (fidp->fid_type == P9_FID_FILE) {
314 /* If we reclaimed the fd no need to close */
315 if (fidp->fs.fd != -1) {
316 retval = v9fs_co_close(pdu, &fidp->fs);
318 } else if (fidp->fid_type == P9_FID_DIR) {
319 if (fidp->fs.dir != NULL) {
320 retval = v9fs_co_closedir(pdu, &fidp->fs);
322 } else if (fidp->fid_type == P9_FID_XATTR) {
323 retval = v9fs_xattr_fid_clunk(pdu, fidp);
325 v9fs_path_free(&fidp->path);
330 static void put_fid(V9fsPDU *pdu, V9fsFidState *fidp)
335 * Don't free the fid if it is in reclaim list
337 if (!fidp->ref && fidp->clunked) {
338 if (fidp->fid == pdu->s->root_fid) {
340 * if the clunked fid is root fid then we
341 * have unmounted the fs on the client side.
342 * delete the migration blocker. Ideally, this
343 * should be hooked to transport close notification
345 if (pdu->s->migration_blocker) {
346 migrate_del_blocker(pdu->s->migration_blocker);
347 error_free(pdu->s->migration_blocker);
348 pdu->s->migration_blocker = NULL;
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__);
511 #define P9_QID_TYPE_DIR 0x80
512 #define P9_QID_TYPE_SYMLINK 0x02
514 #define P9_STAT_MODE_DIR 0x80000000
515 #define P9_STAT_MODE_APPEND 0x40000000
516 #define P9_STAT_MODE_EXCL 0x20000000
517 #define P9_STAT_MODE_MOUNT 0x10000000
518 #define P9_STAT_MODE_AUTH 0x08000000
519 #define P9_STAT_MODE_TMP 0x04000000
520 #define P9_STAT_MODE_SYMLINK 0x02000000
521 #define P9_STAT_MODE_LINK 0x01000000
522 #define P9_STAT_MODE_DEVICE 0x00800000
523 #define P9_STAT_MODE_NAMED_PIPE 0x00200000
524 #define P9_STAT_MODE_SOCKET 0x00100000
525 #define P9_STAT_MODE_SETUID 0x00080000
526 #define P9_STAT_MODE_SETGID 0x00040000
527 #define P9_STAT_MODE_SETVTX 0x00010000
529 #define P9_STAT_MODE_TYPE_BITS (P9_STAT_MODE_DIR | \
530 P9_STAT_MODE_SYMLINK | \
531 P9_STAT_MODE_LINK | \
532 P9_STAT_MODE_DEVICE | \
533 P9_STAT_MODE_NAMED_PIPE | \
536 /* This is the algorithm from ufs in spfs */
537 static void stat_to_qid(const struct stat *stbuf, V9fsQID *qidp)
541 memset(&qidp->path, 0, sizeof(qidp->path));
542 size = MIN(sizeof(stbuf->st_ino), sizeof(qidp->path));
543 memcpy(&qidp->path, &stbuf->st_ino, size);
544 qidp->version = stbuf->st_mtime ^ (stbuf->st_size << 8);
546 if (S_ISDIR(stbuf->st_mode)) {
547 qidp->type |= P9_QID_TYPE_DIR;
549 if (S_ISLNK(stbuf->st_mode)) {
550 qidp->type |= P9_QID_TYPE_SYMLINK;
554 static int fid_to_qid(V9fsPDU *pdu, V9fsFidState *fidp, V9fsQID *qidp)
559 err = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
563 stat_to_qid(&stbuf, qidp);
567 static V9fsPDU *alloc_pdu(V9fsState *s)
571 if (!QLIST_EMPTY(&s->free_list)) {
572 pdu = QLIST_FIRST(&s->free_list);
573 QLIST_REMOVE(pdu, next);
574 QLIST_INSERT_HEAD(&s->active_list, pdu, next);
579 static void free_pdu(V9fsState *s, V9fsPDU *pdu)
583 * Cancelled pdu are added back to the freelist
586 if (!pdu->cancelled) {
587 QLIST_REMOVE(pdu, next);
588 QLIST_INSERT_HEAD(&s->free_list, pdu, next);
594 * We don't do error checking for pdu_marshal/unmarshal here
595 * because we always expect to have enough space to encode
598 static void complete_pdu(V9fsState *s, V9fsPDU *pdu, ssize_t len)
600 int8_t id = pdu->id + 1; /* Response */
606 if (s->proto_version != V9FS_PROTO_2000L) {
609 str.data = strerror(err);
610 str.size = strlen(str.data);
612 len += pdu_marshal(pdu, len, "s", &str);
616 len += pdu_marshal(pdu, len, "d", err);
618 if (s->proto_version == V9FS_PROTO_2000L) {
621 trace_v9fs_rerror(pdu->tag, pdu->id, err); /* Trace ERROR */
624 /* fill out the header */
625 pdu_marshal(pdu, 0, "dbw", (int32_t)len, id, pdu->tag);
627 /* keep these in sync */
631 /* push onto queue and notify */
632 virtqueue_push(s->vq, &pdu->elem, len);
634 /* FIXME: we should batch these completions */
635 virtio_notify(&s->vdev, s->vq);
637 /* Now wakeup anybody waiting in flush for this request */
638 qemu_co_queue_next(&pdu->complete);
643 static mode_t v9mode_to_mode(uint32_t mode, V9fsString *extension)
648 if (mode & P9_STAT_MODE_DIR) {
652 if (mode & P9_STAT_MODE_SYMLINK) {
655 if (mode & P9_STAT_MODE_SOCKET) {
658 if (mode & P9_STAT_MODE_NAMED_PIPE) {
661 if (mode & P9_STAT_MODE_DEVICE) {
662 if (extension && extension->data[0] == 'c') {
673 if (mode & P9_STAT_MODE_SETUID) {
676 if (mode & P9_STAT_MODE_SETGID) {
679 if (mode & P9_STAT_MODE_SETVTX) {
686 static int donttouch_stat(V9fsStat *stat)
688 if (stat->type == -1 &&
690 stat->qid.type == -1 &&
691 stat->qid.version == -1 &&
692 stat->qid.path == -1 &&
696 stat->length == -1 &&
703 stat->n_muid == -1) {
710 static void v9fs_stat_init(V9fsStat *stat)
712 v9fs_string_init(&stat->name);
713 v9fs_string_init(&stat->uid);
714 v9fs_string_init(&stat->gid);
715 v9fs_string_init(&stat->muid);
716 v9fs_string_init(&stat->extension);
719 static void v9fs_stat_free(V9fsStat *stat)
721 v9fs_string_free(&stat->name);
722 v9fs_string_free(&stat->uid);
723 v9fs_string_free(&stat->gid);
724 v9fs_string_free(&stat->muid);
725 v9fs_string_free(&stat->extension);
728 static uint32_t stat_to_v9mode(const struct stat *stbuf)
732 mode = stbuf->st_mode & 0777;
733 if (S_ISDIR(stbuf->st_mode)) {
734 mode |= P9_STAT_MODE_DIR;
737 if (S_ISLNK(stbuf->st_mode)) {
738 mode |= P9_STAT_MODE_SYMLINK;
741 if (S_ISSOCK(stbuf->st_mode)) {
742 mode |= P9_STAT_MODE_SOCKET;
745 if (S_ISFIFO(stbuf->st_mode)) {
746 mode |= P9_STAT_MODE_NAMED_PIPE;
749 if (S_ISBLK(stbuf->st_mode) || S_ISCHR(stbuf->st_mode)) {
750 mode |= P9_STAT_MODE_DEVICE;
753 if (stbuf->st_mode & S_ISUID) {
754 mode |= P9_STAT_MODE_SETUID;
757 if (stbuf->st_mode & S_ISGID) {
758 mode |= P9_STAT_MODE_SETGID;
761 if (stbuf->st_mode & S_ISVTX) {
762 mode |= P9_STAT_MODE_SETVTX;
768 static int stat_to_v9stat(V9fsPDU *pdu, V9fsPath *name,
769 const struct stat *stbuf,
775 memset(v9stat, 0, sizeof(*v9stat));
777 stat_to_qid(stbuf, &v9stat->qid);
778 v9stat->mode = stat_to_v9mode(stbuf);
779 v9stat->atime = stbuf->st_atime;
780 v9stat->mtime = stbuf->st_mtime;
781 v9stat->length = stbuf->st_size;
783 v9fs_string_null(&v9stat->uid);
784 v9fs_string_null(&v9stat->gid);
785 v9fs_string_null(&v9stat->muid);
787 v9stat->n_uid = stbuf->st_uid;
788 v9stat->n_gid = stbuf->st_gid;
791 v9fs_string_null(&v9stat->extension);
793 if (v9stat->mode & P9_STAT_MODE_SYMLINK) {
794 err = v9fs_co_readlink(pdu, name, &v9stat->extension);
798 } else if (v9stat->mode & P9_STAT_MODE_DEVICE) {
799 v9fs_string_sprintf(&v9stat->extension, "%c %u %u",
800 S_ISCHR(stbuf->st_mode) ? 'c' : 'b',
801 major(stbuf->st_rdev), minor(stbuf->st_rdev));
802 } else if (S_ISDIR(stbuf->st_mode) || S_ISREG(stbuf->st_mode)) {
803 v9fs_string_sprintf(&v9stat->extension, "%s %lu",
804 "HARDLINKCOUNT", (unsigned long)stbuf->st_nlink);
807 str = strrchr(name->data, '/');
814 v9fs_string_sprintf(&v9stat->name, "%s", str);
817 v9fs_string_size(&v9stat->name) +
818 v9fs_string_size(&v9stat->uid) +
819 v9fs_string_size(&v9stat->gid) +
820 v9fs_string_size(&v9stat->muid) +
821 v9fs_string_size(&v9stat->extension);
825 #define P9_STATS_MODE 0x00000001ULL
826 #define P9_STATS_NLINK 0x00000002ULL
827 #define P9_STATS_UID 0x00000004ULL
828 #define P9_STATS_GID 0x00000008ULL
829 #define P9_STATS_RDEV 0x00000010ULL
830 #define P9_STATS_ATIME 0x00000020ULL
831 #define P9_STATS_MTIME 0x00000040ULL
832 #define P9_STATS_CTIME 0x00000080ULL
833 #define P9_STATS_INO 0x00000100ULL
834 #define P9_STATS_SIZE 0x00000200ULL
835 #define P9_STATS_BLOCKS 0x00000400ULL
837 #define P9_STATS_BTIME 0x00000800ULL
838 #define P9_STATS_GEN 0x00001000ULL
839 #define P9_STATS_DATA_VERSION 0x00002000ULL
841 #define P9_STATS_BASIC 0x000007ffULL /* Mask for fields up to BLOCKS */
842 #define P9_STATS_ALL 0x00003fffULL /* Mask for All fields above */
845 static void stat_to_v9stat_dotl(V9fsState *s, const struct stat *stbuf,
846 V9fsStatDotl *v9lstat)
848 memset(v9lstat, 0, sizeof(*v9lstat));
850 v9lstat->st_mode = stbuf->st_mode;
851 v9lstat->st_nlink = stbuf->st_nlink;
852 v9lstat->st_uid = stbuf->st_uid;
853 v9lstat->st_gid = stbuf->st_gid;
854 v9lstat->st_rdev = stbuf->st_rdev;
855 v9lstat->st_size = stbuf->st_size;
856 v9lstat->st_blksize = stbuf->st_blksize;
857 v9lstat->st_blocks = stbuf->st_blocks;
858 v9lstat->st_atime_sec = stbuf->st_atime;
859 v9lstat->st_atime_nsec = stbuf->st_atim.tv_nsec;
860 v9lstat->st_mtime_sec = stbuf->st_mtime;
861 v9lstat->st_mtime_nsec = stbuf->st_mtim.tv_nsec;
862 v9lstat->st_ctime_sec = stbuf->st_ctime;
863 v9lstat->st_ctime_nsec = stbuf->st_ctim.tv_nsec;
864 /* Currently we only support BASIC fields in stat */
865 v9lstat->st_result_mask = P9_STATS_BASIC;
867 stat_to_qid(stbuf, &v9lstat->qid);
870 static void print_sg(struct iovec *sg, int cnt)
874 printf("sg[%d]: {", cnt);
875 for (i = 0; i < cnt; i++) {
879 printf("(%p, %zd)", sg[i].iov_base, sg[i].iov_len);
884 /* Will call this only for path name based fid */
885 static void v9fs_fix_path(V9fsPath *dst, V9fsPath *src, int len)
888 v9fs_path_init(&str);
889 v9fs_path_copy(&str, dst);
890 v9fs_string_sprintf((V9fsString *)dst, "%s%s", src->data, str.data+len);
891 v9fs_path_free(&str);
892 /* +1 to include terminating NULL */
896 static inline bool is_ro_export(FsContext *ctx)
898 return ctx->export_flags & V9FS_RDONLY;
901 static void v9fs_version(void *opaque)
904 V9fsPDU *pdu = opaque;
905 V9fsState *s = pdu->s;
909 v9fs_string_init(&version);
910 err = pdu_unmarshal(pdu, offset, "ds", &s->msize, &version);
915 trace_v9fs_version(pdu->tag, pdu->id, s->msize, version.data);
919 if (!strcmp(version.data, "9P2000.u")) {
920 s->proto_version = V9FS_PROTO_2000U;
921 } else if (!strcmp(version.data, "9P2000.L")) {
922 s->proto_version = V9FS_PROTO_2000L;
924 v9fs_string_sprintf(&version, "unknown");
927 err = pdu_marshal(pdu, offset, "ds", s->msize, &version);
933 trace_v9fs_version_return(pdu->tag, pdu->id, s->msize, version.data);
935 complete_pdu(s, pdu, offset);
936 v9fs_string_free(&version);
940 static void v9fs_attach(void *opaque)
942 V9fsPDU *pdu = opaque;
943 V9fsState *s = pdu->s;
944 int32_t fid, afid, n_uname;
945 V9fsString uname, aname;
951 v9fs_string_init(&uname);
952 v9fs_string_init(&aname);
953 err = pdu_unmarshal(pdu, offset, "ddssd", &fid,
954 &afid, &uname, &aname, &n_uname);
958 trace_v9fs_attach(pdu->tag, pdu->id, fid, afid, uname.data, aname.data);
960 fidp = alloc_fid(s, fid);
966 err = v9fs_co_name_to_path(pdu, NULL, "/", &fidp->path);
972 err = fid_to_qid(pdu, fidp, &qid);
978 err = pdu_marshal(pdu, offset, "Q", &qid);
984 trace_v9fs_attach_return(pdu->tag, pdu->id,
985 qid.type, qid.version, qid.path);
987 * disable migration if we haven't done already.
988 * attach could get called multiple times for the same export.
990 if (!s->migration_blocker) {
992 error_set(&s->migration_blocker, QERR_VIRTFS_FEATURE_BLOCKS_MIGRATION,
993 s->ctx.fs_root ? s->ctx.fs_root : "NULL", s->tag);
994 migrate_add_blocker(s->migration_blocker);
999 complete_pdu(s, pdu, err);
1000 v9fs_string_free(&uname);
1001 v9fs_string_free(&aname);
1004 static void v9fs_stat(void *opaque)
1012 V9fsPDU *pdu = opaque;
1013 V9fsState *s = pdu->s;
1015 err = pdu_unmarshal(pdu, offset, "d", &fid);
1019 trace_v9fs_stat(pdu->tag, pdu->id, fid);
1021 fidp = get_fid(pdu, fid);
1026 err = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
1030 err = stat_to_v9stat(pdu, &fidp->path, &stbuf, &v9stat);
1034 err = pdu_marshal(pdu, offset, "wS", 0, &v9stat);
1036 v9fs_stat_free(&v9stat);
1039 trace_v9fs_stat_return(pdu->tag, pdu->id, v9stat.mode,
1040 v9stat.atime, v9stat.mtime, v9stat.length);
1042 v9fs_stat_free(&v9stat);
1046 complete_pdu(s, pdu, err);
1049 static void v9fs_getattr(void *opaque)
1056 uint64_t request_mask;
1057 V9fsStatDotl v9stat_dotl;
1058 V9fsPDU *pdu = opaque;
1059 V9fsState *s = pdu->s;
1061 retval = pdu_unmarshal(pdu, offset, "dq", &fid, &request_mask);
1065 trace_v9fs_getattr(pdu->tag, pdu->id, fid, request_mask);
1067 fidp = get_fid(pdu, fid);
1073 * Currently we only support BASIC fields in stat, so there is no
1074 * need to look at request_mask.
1076 retval = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
1080 stat_to_v9stat_dotl(s, &stbuf, &v9stat_dotl);
1082 /* fill st_gen if requested and supported by underlying fs */
1083 if (request_mask & P9_STATS_GEN) {
1084 retval = v9fs_co_st_gen(pdu, &fidp->path, stbuf.st_mode, &v9stat_dotl);
1088 v9stat_dotl.st_result_mask |= P9_STATS_GEN;
1090 retval = pdu_marshal(pdu, offset, "A", &v9stat_dotl);
1095 trace_v9fs_getattr_return(pdu->tag, pdu->id, v9stat_dotl.st_result_mask,
1096 v9stat_dotl.st_mode, v9stat_dotl.st_uid,
1097 v9stat_dotl.st_gid);
1101 complete_pdu(s, pdu, retval);
1104 /* Attribute flags */
1105 #define P9_ATTR_MODE (1 << 0)
1106 #define P9_ATTR_UID (1 << 1)
1107 #define P9_ATTR_GID (1 << 2)
1108 #define P9_ATTR_SIZE (1 << 3)
1109 #define P9_ATTR_ATIME (1 << 4)
1110 #define P9_ATTR_MTIME (1 << 5)
1111 #define P9_ATTR_CTIME (1 << 6)
1112 #define P9_ATTR_ATIME_SET (1 << 7)
1113 #define P9_ATTR_MTIME_SET (1 << 8)
1115 #define P9_ATTR_MASK 127
1117 static void v9fs_setattr(void *opaque)
1124 V9fsPDU *pdu = opaque;
1125 V9fsState *s = pdu->s;
1127 err = pdu_unmarshal(pdu, offset, "dI", &fid, &v9iattr);
1132 fidp = get_fid(pdu, fid);
1137 if (v9iattr.valid & P9_ATTR_MODE) {
1138 err = v9fs_co_chmod(pdu, &fidp->path, v9iattr.mode);
1143 if (v9iattr.valid & (P9_ATTR_ATIME | P9_ATTR_MTIME)) {
1144 struct timespec times[2];
1145 if (v9iattr.valid & P9_ATTR_ATIME) {
1146 if (v9iattr.valid & P9_ATTR_ATIME_SET) {
1147 times[0].tv_sec = v9iattr.atime_sec;
1148 times[0].tv_nsec = v9iattr.atime_nsec;
1150 times[0].tv_nsec = UTIME_NOW;
1153 times[0].tv_nsec = UTIME_OMIT;
1155 if (v9iattr.valid & P9_ATTR_MTIME) {
1156 if (v9iattr.valid & P9_ATTR_MTIME_SET) {
1157 times[1].tv_sec = v9iattr.mtime_sec;
1158 times[1].tv_nsec = v9iattr.mtime_nsec;
1160 times[1].tv_nsec = UTIME_NOW;
1163 times[1].tv_nsec = UTIME_OMIT;
1165 err = v9fs_co_utimensat(pdu, &fidp->path, times);
1171 * If the only valid entry in iattr is ctime we can call
1172 * chown(-1,-1) to update the ctime of the file
1174 if ((v9iattr.valid & (P9_ATTR_UID | P9_ATTR_GID)) ||
1175 ((v9iattr.valid & P9_ATTR_CTIME)
1176 && !((v9iattr.valid & P9_ATTR_MASK) & ~P9_ATTR_CTIME))) {
1177 if (!(v9iattr.valid & P9_ATTR_UID)) {
1180 if (!(v9iattr.valid & P9_ATTR_GID)) {
1183 err = v9fs_co_chown(pdu, &fidp->path, v9iattr.uid,
1189 if (v9iattr.valid & (P9_ATTR_SIZE)) {
1190 err = v9fs_co_truncate(pdu, &fidp->path, v9iattr.size);
1199 complete_pdu(s, pdu, err);
1202 static int v9fs_walk_marshal(V9fsPDU *pdu, uint16_t nwnames, V9fsQID *qids)
1208 err = pdu_marshal(pdu, offset, "w", nwnames);
1213 for (i = 0; i < nwnames; i++) {
1214 err = pdu_marshal(pdu, offset, "Q", &qids[i]);
1223 static void v9fs_walk(void *opaque)
1226 V9fsQID *qids = NULL;
1228 V9fsPath dpath, path;
1232 int32_t fid, newfid;
1233 V9fsString *wnames = NULL;
1235 V9fsFidState *newfidp = NULL;
1236 V9fsPDU *pdu = opaque;
1237 V9fsState *s = pdu->s;
1239 err = pdu_unmarshal(pdu, offset, "ddw", &fid, &newfid, &nwnames);
1241 complete_pdu(s, pdu, err);
1246 trace_v9fs_walk(pdu->tag, pdu->id, fid, newfid, nwnames);
1248 if (nwnames && nwnames <= P9_MAXWELEM) {
1249 wnames = g_malloc0(sizeof(wnames[0]) * nwnames);
1250 qids = g_malloc0(sizeof(qids[0]) * nwnames);
1251 for (i = 0; i < nwnames; i++) {
1252 err = pdu_unmarshal(pdu, offset, "s", &wnames[i]);
1258 } else if (nwnames > P9_MAXWELEM) {
1262 fidp = get_fid(pdu, fid);
1267 v9fs_path_init(&dpath);
1268 v9fs_path_init(&path);
1270 * Both dpath and path initially poin to fidp.
1271 * Needed to handle request with nwnames == 0
1273 v9fs_path_copy(&dpath, &fidp->path);
1274 v9fs_path_copy(&path, &fidp->path);
1275 for (name_idx = 0; name_idx < nwnames; name_idx++) {
1276 err = v9fs_co_name_to_path(pdu, &dpath, wnames[name_idx].data, &path);
1280 err = v9fs_co_lstat(pdu, &path, &stbuf);
1284 stat_to_qid(&stbuf, &qids[name_idx]);
1285 v9fs_path_copy(&dpath, &path);
1287 if (fid == newfid) {
1288 BUG_ON(fidp->fid_type != P9_FID_NONE);
1289 v9fs_path_copy(&fidp->path, &path);
1291 newfidp = alloc_fid(s, newfid);
1292 if (newfidp == NULL) {
1296 newfidp->uid = fidp->uid;
1297 v9fs_path_copy(&newfidp->path, &path);
1299 err = v9fs_walk_marshal(pdu, nwnames, qids);
1300 trace_v9fs_walk_return(pdu->tag, pdu->id, nwnames, qids);
1304 put_fid(pdu, newfidp);
1306 v9fs_path_free(&dpath);
1307 v9fs_path_free(&path);
1309 complete_pdu(s, pdu, err);
1310 if (nwnames && nwnames <= P9_MAXWELEM) {
1311 for (name_idx = 0; name_idx < nwnames; name_idx++) {
1312 v9fs_string_free(&wnames[name_idx]);
1320 static int32_t get_iounit(V9fsPDU *pdu, V9fsPath *path)
1322 struct statfs stbuf;
1324 V9fsState *s = pdu->s;
1327 * iounit should be multiples of f_bsize (host filesystem block size
1328 * and as well as less than (client msize - P9_IOHDRSZ))
1330 if (!v9fs_co_statfs(pdu, path, &stbuf)) {
1331 iounit = stbuf.f_bsize;
1332 iounit *= (s->msize - P9_IOHDRSZ)/stbuf.f_bsize;
1335 iounit = s->msize - P9_IOHDRSZ;
1340 static void v9fs_open(void *opaque)
1351 V9fsPDU *pdu = opaque;
1352 V9fsState *s = pdu->s;
1354 if (s->proto_version == V9FS_PROTO_2000L) {
1355 err = pdu_unmarshal(pdu, offset, "dd", &fid, &mode);
1358 err = pdu_unmarshal(pdu, offset, "db", &fid, &modebyte);
1364 trace_v9fs_open(pdu->tag, pdu->id, fid, mode);
1366 fidp = get_fid(pdu, fid);
1371 BUG_ON(fidp->fid_type != P9_FID_NONE);
1373 err = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
1377 stat_to_qid(&stbuf, &qid);
1378 if (S_ISDIR(stbuf.st_mode)) {
1379 err = v9fs_co_opendir(pdu, fidp);
1383 fidp->fid_type = P9_FID_DIR;
1384 err = pdu_marshal(pdu, offset, "Qd", &qid, 0);
1390 if (s->proto_version == V9FS_PROTO_2000L) {
1391 flags = get_dotl_openflags(s, mode);
1393 flags = omode_to_uflags(mode);
1395 if (is_ro_export(&s->ctx)) {
1396 if (mode & O_WRONLY || mode & O_RDWR ||
1397 mode & O_APPEND || mode & O_TRUNC) {
1402 err = v9fs_co_open(pdu, fidp, flags);
1406 fidp->fid_type = P9_FID_FILE;
1407 fidp->open_flags = flags;
1408 if (flags & O_EXCL) {
1410 * We let the host file system do O_EXCL check
1411 * We should not reclaim such fd
1413 fidp->flags |= FID_NON_RECLAIMABLE;
1415 iounit = get_iounit(pdu, &fidp->path);
1416 err = pdu_marshal(pdu, offset, "Qd", &qid, iounit);
1422 trace_v9fs_open_return(pdu->tag, pdu->id,
1423 qid.type, qid.version, qid.path, iounit);
1427 complete_pdu(s, pdu, err);
1430 static void v9fs_lcreate(void *opaque)
1432 int32_t dfid, flags, mode;
1441 V9fsPDU *pdu = opaque;
1443 v9fs_string_init(&name);
1444 err = pdu_unmarshal(pdu, offset, "dsddd", &dfid,
1445 &name, &flags, &mode, &gid);
1449 trace_v9fs_lcreate(pdu->tag, pdu->id, dfid, flags, mode, gid);
1451 fidp = get_fid(pdu, dfid);
1457 flags = get_dotl_openflags(pdu->s, flags);
1458 err = v9fs_co_open2(pdu, fidp, &name, gid,
1459 flags | O_CREAT, mode, &stbuf);
1463 fidp->fid_type = P9_FID_FILE;
1464 fidp->open_flags = flags;
1465 if (flags & O_EXCL) {
1467 * We let the host file system do O_EXCL check
1468 * We should not reclaim such fd
1470 fidp->flags |= FID_NON_RECLAIMABLE;
1472 iounit = get_iounit(pdu, &fidp->path);
1473 stat_to_qid(&stbuf, &qid);
1474 err = pdu_marshal(pdu, offset, "Qd", &qid, iounit);
1479 trace_v9fs_lcreate_return(pdu->tag, pdu->id,
1480 qid.type, qid.version, qid.path, iounit);
1484 complete_pdu(pdu->s, pdu, err);
1485 v9fs_string_free(&name);
1488 static void v9fs_fsync(void *opaque)
1495 V9fsPDU *pdu = opaque;
1496 V9fsState *s = pdu->s;
1498 err = pdu_unmarshal(pdu, offset, "dd", &fid, &datasync);
1502 trace_v9fs_fsync(pdu->tag, pdu->id, fid, datasync);
1504 fidp = get_fid(pdu, fid);
1509 err = v9fs_co_fsync(pdu, fidp, datasync);
1515 complete_pdu(s, pdu, err);
1518 static void v9fs_clunk(void *opaque)
1524 V9fsPDU *pdu = opaque;
1525 V9fsState *s = pdu->s;
1527 err = pdu_unmarshal(pdu, offset, "d", &fid);
1531 trace_v9fs_clunk(pdu->tag, pdu->id, fid);
1533 fidp = clunk_fid(s, fid);
1539 * Bump the ref so that put_fid will
1547 complete_pdu(s, pdu, err);
1550 static int v9fs_xattr_read(V9fsState *s, V9fsPDU *pdu, V9fsFidState *fidp,
1551 uint64_t off, uint32_t max_count)
1558 xattr_len = fidp->fs.xattr.len;
1559 read_count = xattr_len - off;
1560 if (read_count > max_count) {
1561 read_count = max_count;
1562 } else if (read_count < 0) {
1564 * read beyond XATTR value
1568 err = pdu_marshal(pdu, offset, "d", read_count);
1573 err = v9fs_pack(pdu->elem.in_sg, pdu->elem.in_num, offset,
1574 ((char *)fidp->fs.xattr.value) + off,
1583 static int v9fs_do_readdir_with_stat(V9fsPDU *pdu,
1584 V9fsFidState *fidp, uint32_t max_count)
1591 off_t saved_dir_pos;
1592 struct dirent *dent, *result;
1594 /* save the directory position */
1595 saved_dir_pos = v9fs_co_telldir(pdu, fidp);
1596 if (saved_dir_pos < 0) {
1597 return saved_dir_pos;
1600 dent = g_malloc(sizeof(struct dirent));
1603 v9fs_path_init(&path);
1604 err = v9fs_co_readdir_r(pdu, fidp, dent, &result);
1605 if (err || !result) {
1608 err = v9fs_co_name_to_path(pdu, &fidp->path, dent->d_name, &path);
1612 err = v9fs_co_lstat(pdu, &path, &stbuf);
1616 err = stat_to_v9stat(pdu, &path, &stbuf, &v9stat);
1620 /* 11 = 7 + 4 (7 = start offset, 4 = space for storing count) */
1621 len = pdu_marshal(pdu, 11 + count, "S", &v9stat);
1622 if ((len != (v9stat.size + 2)) || ((count + len) > max_count)) {
1623 /* Ran out of buffer. Set dir back to old position and return */
1624 v9fs_co_seekdir(pdu, fidp, saved_dir_pos);
1625 v9fs_stat_free(&v9stat);
1626 v9fs_path_free(&path);
1631 v9fs_stat_free(&v9stat);
1632 v9fs_path_free(&path);
1633 saved_dir_pos = dent->d_off;
1637 v9fs_path_free(&path);
1645 * Create a QEMUIOVector for a sub-region of PDU iovecs
1647 * @qiov: uninitialized QEMUIOVector
1648 * @skip: number of bytes to skip from beginning of PDU
1649 * @size: number of bytes to include
1650 * @is_write: true - write, false - read
1652 * The resulting QEMUIOVector has heap-allocated iovecs and must be cleaned up
1653 * with qemu_iovec_destroy().
1655 static void v9fs_init_qiov_from_pdu(QEMUIOVector *qiov, V9fsPDU *pdu,
1656 size_t skip, size_t size,
1664 iov = pdu->elem.out_sg;
1665 niov = pdu->elem.out_num;
1667 iov = pdu->elem.in_sg;
1668 niov = pdu->elem.in_num;
1671 qemu_iovec_init_external(&elem, iov, niov);
1672 qemu_iovec_init(qiov, niov);
1673 qemu_iovec_concat(qiov, &elem, skip, size);
1676 static void v9fs_read(void *opaque)
1685 V9fsPDU *pdu = opaque;
1686 V9fsState *s = pdu->s;
1688 err = pdu_unmarshal(pdu, offset, "dqd", &fid, &off, &max_count);
1692 trace_v9fs_read(pdu->tag, pdu->id, fid, off, max_count);
1694 fidp = get_fid(pdu, fid);
1699 if (fidp->fid_type == P9_FID_DIR) {
1702 v9fs_co_rewinddir(pdu, fidp);
1704 count = v9fs_do_readdir_with_stat(pdu, fidp, max_count);
1709 err = pdu_marshal(pdu, offset, "d", count);
1713 err += offset + count;
1714 } else if (fidp->fid_type == P9_FID_FILE) {
1715 QEMUIOVector qiov_full;
1719 v9fs_init_qiov_from_pdu(&qiov_full, pdu, offset + 4, max_count, false);
1720 qemu_iovec_init(&qiov, qiov_full.niov);
1722 qemu_iovec_reset(&qiov);
1723 qemu_iovec_concat(&qiov, &qiov_full, count, qiov_full.size - count);
1725 print_sg(qiov.iov, qiov.niov);
1727 /* Loop in case of EINTR */
1729 len = v9fs_co_preadv(pdu, fidp, qiov.iov, qiov.niov, off);
1734 } while (len == -EINTR && !pdu->cancelled);
1736 /* IO error return the error */
1740 } while (count < max_count && len > 0);
1741 err = pdu_marshal(pdu, offset, "d", count);
1745 err += offset + count;
1746 qemu_iovec_destroy(&qiov);
1747 qemu_iovec_destroy(&qiov_full);
1748 } else if (fidp->fid_type == P9_FID_XATTR) {
1749 err = v9fs_xattr_read(s, pdu, fidp, off, max_count);
1753 trace_v9fs_read_return(pdu->tag, pdu->id, count, err);
1757 complete_pdu(s, pdu, err);
1760 static size_t v9fs_readdir_data_size(V9fsString *name)
1763 * Size of each dirent on the wire: size of qid (13) + size of offset (8)
1764 * size of type (1) + size of name.size (2) + strlen(name.data)
1766 return 24 + v9fs_string_size(name);
1769 static int v9fs_do_readdir(V9fsPDU *pdu,
1770 V9fsFidState *fidp, int32_t max_count)
1777 off_t saved_dir_pos;
1778 struct dirent *dent, *result;
1780 /* save the directory position */
1781 saved_dir_pos = v9fs_co_telldir(pdu, fidp);
1782 if (saved_dir_pos < 0) {
1783 return saved_dir_pos;
1786 dent = g_malloc(sizeof(struct dirent));
1789 err = v9fs_co_readdir_r(pdu, fidp, dent, &result);
1790 if (err || !result) {
1793 v9fs_string_init(&name);
1794 v9fs_string_sprintf(&name, "%s", dent->d_name);
1795 if ((count + v9fs_readdir_data_size(&name)) > max_count) {
1796 /* Ran out of buffer. Set dir back to old position and return */
1797 v9fs_co_seekdir(pdu, fidp, saved_dir_pos);
1798 v9fs_string_free(&name);
1803 * Fill up just the path field of qid because the client uses
1804 * only that. To fill the entire qid structure we will have
1805 * to stat each dirent found, which is expensive
1807 size = MIN(sizeof(dent->d_ino), sizeof(qid.path));
1808 memcpy(&qid.path, &dent->d_ino, size);
1809 /* Fill the other fields with dummy values */
1813 /* 11 = 7 + 4 (7 = start offset, 4 = space for storing count) */
1814 len = pdu_marshal(pdu, 11 + count, "Qqbs",
1816 dent->d_type, &name);
1818 v9fs_co_seekdir(pdu, fidp, saved_dir_pos);
1819 v9fs_string_free(&name);
1824 v9fs_string_free(&name);
1825 saved_dir_pos = dent->d_off;
1834 static void v9fs_readdir(void *opaque)
1840 uint64_t initial_offset;
1843 V9fsPDU *pdu = opaque;
1844 V9fsState *s = pdu->s;
1846 retval = pdu_unmarshal(pdu, offset, "dqd", &fid,
1847 &initial_offset, &max_count);
1851 trace_v9fs_readdir(pdu->tag, pdu->id, fid, initial_offset, max_count);
1853 fidp = get_fid(pdu, fid);
1858 if (!fidp->fs.dir) {
1862 if (initial_offset == 0) {
1863 v9fs_co_rewinddir(pdu, fidp);
1865 v9fs_co_seekdir(pdu, fidp, initial_offset);
1867 count = v9fs_do_readdir(pdu, fidp, max_count);
1872 retval = pdu_marshal(pdu, offset, "d", count);
1876 retval += count + offset;
1877 trace_v9fs_readdir_return(pdu->tag, pdu->id, count, retval);
1881 complete_pdu(s, pdu, retval);
1884 static int v9fs_xattr_write(V9fsState *s, V9fsPDU *pdu, V9fsFidState *fidp,
1885 uint64_t off, uint32_t count,
1886 struct iovec *sg, int cnt)
1895 xattr_len = fidp->fs.xattr.len;
1896 write_count = xattr_len - off;
1897 if (write_count > count) {
1898 write_count = count;
1899 } else if (write_count < 0) {
1901 * write beyond XATTR value len specified in
1907 err = pdu_marshal(pdu, offset, "d", write_count);
1912 fidp->fs.xattr.copied_len += write_count;
1914 * Now copy the content from sg list
1916 for (i = 0; i < cnt; i++) {
1917 if (write_count > sg[i].iov_len) {
1918 to_copy = sg[i].iov_len;
1920 to_copy = write_count;
1922 memcpy((char *)fidp->fs.xattr.value + off, sg[i].iov_base, to_copy);
1923 /* updating vs->off since we are not using below */
1925 write_count -= to_copy;
1931 static void v9fs_write(void *opaque)
1941 V9fsPDU *pdu = opaque;
1942 V9fsState *s = pdu->s;
1943 QEMUIOVector qiov_full;
1946 err = pdu_unmarshal(pdu, offset, "dqd", &fid, &off, &count);
1948 return complete_pdu(s, pdu, err);
1951 v9fs_init_qiov_from_pdu(&qiov_full, pdu, offset, count, true);
1952 trace_v9fs_write(pdu->tag, pdu->id, fid, off, count, qiov_full.niov);
1954 fidp = get_fid(pdu, fid);
1959 if (fidp->fid_type == P9_FID_FILE) {
1960 if (fidp->fs.fd == -1) {
1964 } else if (fidp->fid_type == P9_FID_XATTR) {
1966 * setxattr operation
1968 err = v9fs_xattr_write(s, pdu, fidp, off, count,
1969 qiov_full.iov, qiov_full.niov);
1975 qemu_iovec_init(&qiov, qiov_full.niov);
1977 qemu_iovec_reset(&qiov);
1978 qemu_iovec_concat(&qiov, &qiov_full, total, qiov_full.size - total);
1980 print_sg(qiov.iov, qiov.niov);
1982 /* Loop in case of EINTR */
1984 len = v9fs_co_pwritev(pdu, fidp, qiov.iov, qiov.niov, off);
1989 } while (len == -EINTR && !pdu->cancelled);
1991 /* IO error return the error */
1995 } while (total < count && len > 0);
1998 err = pdu_marshal(pdu, offset, "d", total);
2003 trace_v9fs_write_return(pdu->tag, pdu->id, total, err);
2005 qemu_iovec_destroy(&qiov);
2009 qemu_iovec_destroy(&qiov_full);
2010 complete_pdu(s, pdu, err);
2013 static void v9fs_create(void *opaque)
2025 V9fsString extension;
2027 V9fsPDU *pdu = opaque;
2029 v9fs_path_init(&path);
2030 v9fs_string_init(&name);
2031 v9fs_string_init(&extension);
2032 err = pdu_unmarshal(pdu, offset, "dsdbs", &fid, &name,
2033 &perm, &mode, &extension);
2037 trace_v9fs_create(pdu->tag, pdu->id, fid, name.data, perm, mode);
2039 fidp = get_fid(pdu, fid);
2044 if (perm & P9_STAT_MODE_DIR) {
2045 err = v9fs_co_mkdir(pdu, fidp, &name, perm & 0777,
2046 fidp->uid, -1, &stbuf);
2050 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2054 v9fs_path_copy(&fidp->path, &path);
2055 err = v9fs_co_opendir(pdu, fidp);
2059 fidp->fid_type = P9_FID_DIR;
2060 } else if (perm & P9_STAT_MODE_SYMLINK) {
2061 err = v9fs_co_symlink(pdu, fidp, &name,
2062 extension.data, -1 , &stbuf);
2066 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2070 v9fs_path_copy(&fidp->path, &path);
2071 } else if (perm & P9_STAT_MODE_LINK) {
2072 int32_t ofid = atoi(extension.data);
2073 V9fsFidState *ofidp = get_fid(pdu, ofid);
2074 if (ofidp == NULL) {
2078 err = v9fs_co_link(pdu, ofidp, fidp, &name);
2079 put_fid(pdu, ofidp);
2083 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2085 fidp->fid_type = P9_FID_NONE;
2088 v9fs_path_copy(&fidp->path, &path);
2089 err = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
2091 fidp->fid_type = P9_FID_NONE;
2094 } else if (perm & P9_STAT_MODE_DEVICE) {
2096 uint32_t major, minor;
2099 if (sscanf(extension.data, "%c %u %u", &ctype, &major, &minor) != 3) {
2116 nmode |= perm & 0777;
2117 err = v9fs_co_mknod(pdu, fidp, &name, fidp->uid, -1,
2118 makedev(major, minor), nmode, &stbuf);
2122 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2126 v9fs_path_copy(&fidp->path, &path);
2127 } else if (perm & P9_STAT_MODE_NAMED_PIPE) {
2128 err = v9fs_co_mknod(pdu, fidp, &name, fidp->uid, -1,
2129 0, S_IFIFO | (perm & 0777), &stbuf);
2133 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2137 v9fs_path_copy(&fidp->path, &path);
2138 } else if (perm & P9_STAT_MODE_SOCKET) {
2139 err = v9fs_co_mknod(pdu, fidp, &name, fidp->uid, -1,
2140 0, S_IFSOCK | (perm & 0777), &stbuf);
2144 err = v9fs_co_name_to_path(pdu, &fidp->path, name.data, &path);
2148 v9fs_path_copy(&fidp->path, &path);
2150 err = v9fs_co_open2(pdu, fidp, &name, -1,
2151 omode_to_uflags(mode)|O_CREAT, perm, &stbuf);
2155 fidp->fid_type = P9_FID_FILE;
2156 fidp->open_flags = omode_to_uflags(mode);
2157 if (fidp->open_flags & O_EXCL) {
2159 * We let the host file system do O_EXCL check
2160 * We should not reclaim such fd
2162 fidp->flags |= FID_NON_RECLAIMABLE;
2165 iounit = get_iounit(pdu, &fidp->path);
2166 stat_to_qid(&stbuf, &qid);
2167 err = pdu_marshal(pdu, offset, "Qd", &qid, iounit);
2172 trace_v9fs_create_return(pdu->tag, pdu->id,
2173 qid.type, qid.version, qid.path, iounit);
2177 complete_pdu(pdu->s, pdu, err);
2178 v9fs_string_free(&name);
2179 v9fs_string_free(&extension);
2180 v9fs_path_free(&path);
2183 static void v9fs_symlink(void *opaque)
2185 V9fsPDU *pdu = opaque;
2188 V9fsFidState *dfidp;
2196 v9fs_string_init(&name);
2197 v9fs_string_init(&symname);
2198 err = pdu_unmarshal(pdu, offset, "dssd", &dfid, &name, &symname, &gid);
2202 trace_v9fs_symlink(pdu->tag, pdu->id, dfid, name.data, symname.data, gid);
2204 dfidp = get_fid(pdu, dfid);
2205 if (dfidp == NULL) {
2209 err = v9fs_co_symlink(pdu, dfidp, &name, symname.data, gid, &stbuf);
2213 stat_to_qid(&stbuf, &qid);
2214 err = pdu_marshal(pdu, offset, "Q", &qid);
2219 trace_v9fs_symlink_return(pdu->tag, pdu->id,
2220 qid.type, qid.version, qid.path);
2222 put_fid(pdu, dfidp);
2224 complete_pdu(pdu->s, pdu, err);
2225 v9fs_string_free(&name);
2226 v9fs_string_free(&symname);
2229 static void v9fs_flush(void *opaque)
2234 V9fsPDU *cancel_pdu;
2235 V9fsPDU *pdu = opaque;
2236 V9fsState *s = pdu->s;
2238 err = pdu_unmarshal(pdu, offset, "w", &tag);
2240 complete_pdu(s, pdu, err);
2243 trace_v9fs_flush(pdu->tag, pdu->id, tag);
2245 QLIST_FOREACH(cancel_pdu, &s->active_list, next) {
2246 if (cancel_pdu->tag == tag) {
2251 cancel_pdu->cancelled = 1;
2253 * Wait for pdu to complete.
2255 qemu_co_queue_wait(&cancel_pdu->complete);
2256 cancel_pdu->cancelled = 0;
2257 free_pdu(pdu->s, cancel_pdu);
2259 complete_pdu(s, pdu, 7);
2263 static void v9fs_link(void *opaque)
2265 V9fsPDU *pdu = opaque;
2266 V9fsState *s = pdu->s;
2267 int32_t dfid, oldfid;
2268 V9fsFidState *dfidp, *oldfidp;
2273 v9fs_string_init(&name);
2274 err = pdu_unmarshal(pdu, offset, "dds", &dfid, &oldfid, &name);
2278 trace_v9fs_link(pdu->tag, pdu->id, dfid, oldfid, name.data);
2280 dfidp = get_fid(pdu, dfid);
2281 if (dfidp == NULL) {
2286 oldfidp = get_fid(pdu, oldfid);
2287 if (oldfidp == NULL) {
2291 err = v9fs_co_link(pdu, oldfidp, dfidp, &name);
2296 put_fid(pdu, dfidp);
2298 v9fs_string_free(&name);
2299 complete_pdu(s, pdu, err);
2302 /* Only works with path name based fid */
2303 static void v9fs_remove(void *opaque)
2309 V9fsPDU *pdu = opaque;
2311 err = pdu_unmarshal(pdu, offset, "d", &fid);
2315 trace_v9fs_remove(pdu->tag, pdu->id, fid);
2317 fidp = get_fid(pdu, fid);
2322 /* if fs driver is not path based, return EOPNOTSUPP */
2323 if (!(pdu->s->ctx.export_flags & V9FS_PATHNAME_FSCONTEXT)) {
2328 * IF the file is unlinked, we cannot reopen
2329 * the file later. So don't reclaim fd
2331 err = v9fs_mark_fids_unreclaim(pdu, &fidp->path);
2335 err = v9fs_co_remove(pdu, &fidp->path);
2340 /* For TREMOVE we need to clunk the fid even on failed remove */
2341 clunk_fid(pdu->s, fidp->fid);
2344 complete_pdu(pdu->s, pdu, err);
2347 static void v9fs_unlinkat(void *opaque)
2351 int32_t dfid, flags;
2354 V9fsFidState *dfidp;
2355 V9fsPDU *pdu = opaque;
2357 v9fs_string_init(&name);
2358 err = pdu_unmarshal(pdu, offset, "dsd", &dfid, &name, &flags);
2362 dfidp = get_fid(pdu, dfid);
2363 if (dfidp == NULL) {
2368 * IF the file is unlinked, we cannot reopen
2369 * the file later. So don't reclaim fd
2371 v9fs_path_init(&path);
2372 err = v9fs_co_name_to_path(pdu, &dfidp->path, name.data, &path);
2376 err = v9fs_mark_fids_unreclaim(pdu, &path);
2380 err = v9fs_co_unlinkat(pdu, &dfidp->path, &name, flags);
2385 put_fid(pdu, dfidp);
2386 v9fs_path_free(&path);
2388 complete_pdu(pdu->s, pdu, err);
2389 v9fs_string_free(&name);
2393 /* Only works with path name based fid */
2394 static int v9fs_complete_rename(V9fsPDU *pdu, V9fsFidState *fidp,
2395 int32_t newdirfid, V9fsString *name)
2400 V9fsFidState *tfidp;
2401 V9fsState *s = pdu->s;
2402 V9fsFidState *dirfidp = NULL;
2403 char *old_name, *new_name;
2405 v9fs_path_init(&new_path);
2406 if (newdirfid != -1) {
2407 dirfidp = get_fid(pdu, newdirfid);
2408 if (dirfidp == NULL) {
2412 BUG_ON(dirfidp->fid_type != P9_FID_NONE);
2413 v9fs_co_name_to_path(pdu, &dirfidp->path, name->data, &new_path);
2415 old_name = fidp->path.data;
2416 end = strrchr(old_name, '/');
2422 new_name = g_malloc0(end - old_name + name->size + 1);
2423 strncat(new_name, old_name, end - old_name);
2424 strncat(new_name + (end - old_name), name->data, name->size);
2425 v9fs_co_name_to_path(pdu, NULL, new_name, &new_path);
2428 err = v9fs_co_rename(pdu, &fidp->path, &new_path);
2433 * Fixup fid's pointing to the old name to
2434 * start pointing to the new name
2436 for (tfidp = s->fid_list; tfidp; tfidp = tfidp->next) {
2437 if (v9fs_path_is_ancestor(&fidp->path, &tfidp->path)) {
2438 /* replace the name */
2439 v9fs_fix_path(&tfidp->path, &new_path, strlen(fidp->path.data));
2444 put_fid(pdu, dirfidp);
2446 v9fs_path_free(&new_path);
2451 /* Only works with path name based fid */
2452 static void v9fs_rename(void *opaque)
2460 V9fsPDU *pdu = opaque;
2461 V9fsState *s = pdu->s;
2463 v9fs_string_init(&name);
2464 err = pdu_unmarshal(pdu, offset, "dds", &fid, &newdirfid, &name);
2468 fidp = get_fid(pdu, fid);
2473 BUG_ON(fidp->fid_type != P9_FID_NONE);
2474 /* if fs driver is not path based, return EOPNOTSUPP */
2475 if (!(pdu->s->ctx.export_flags & V9FS_PATHNAME_FSCONTEXT)) {
2479 v9fs_path_write_lock(s);
2480 err = v9fs_complete_rename(pdu, fidp, newdirfid, &name);
2481 v9fs_path_unlock(s);
2488 complete_pdu(s, pdu, err);
2489 v9fs_string_free(&name);
2492 static void v9fs_fix_fid_paths(V9fsPDU *pdu, V9fsPath *olddir,
2493 V9fsString *old_name, V9fsPath *newdir,
2494 V9fsString *new_name)
2496 V9fsFidState *tfidp;
2497 V9fsPath oldpath, newpath;
2498 V9fsState *s = pdu->s;
2501 v9fs_path_init(&oldpath);
2502 v9fs_path_init(&newpath);
2503 v9fs_co_name_to_path(pdu, olddir, old_name->data, &oldpath);
2504 v9fs_co_name_to_path(pdu, newdir, new_name->data, &newpath);
2507 * Fixup fid's pointing to the old name to
2508 * start pointing to the new name
2510 for (tfidp = s->fid_list; tfidp; tfidp = tfidp->next) {
2511 if (v9fs_path_is_ancestor(&oldpath, &tfidp->path)) {
2512 /* replace the name */
2513 v9fs_fix_path(&tfidp->path, &newpath, strlen(oldpath.data));
2516 v9fs_path_free(&oldpath);
2517 v9fs_path_free(&newpath);
2520 static int v9fs_complete_renameat(V9fsPDU *pdu, int32_t olddirfid,
2521 V9fsString *old_name, int32_t newdirfid,
2522 V9fsString *new_name)
2525 V9fsState *s = pdu->s;
2526 V9fsFidState *newdirfidp = NULL, *olddirfidp = NULL;
2528 olddirfidp = get_fid(pdu, olddirfid);
2529 if (olddirfidp == NULL) {
2533 if (newdirfid != -1) {
2534 newdirfidp = get_fid(pdu, newdirfid);
2535 if (newdirfidp == NULL) {
2540 newdirfidp = get_fid(pdu, olddirfid);
2543 err = v9fs_co_renameat(pdu, &olddirfidp->path, old_name,
2544 &newdirfidp->path, new_name);
2548 if (s->ctx.export_flags & V9FS_PATHNAME_FSCONTEXT) {
2549 /* Only for path based fid we need to do the below fixup */
2550 v9fs_fix_fid_paths(pdu, &olddirfidp->path, old_name,
2551 &newdirfidp->path, new_name);
2555 put_fid(pdu, olddirfidp);
2558 put_fid(pdu, newdirfidp);
2563 static void v9fs_renameat(void *opaque)
2567 V9fsPDU *pdu = opaque;
2568 V9fsState *s = pdu->s;
2569 int32_t olddirfid, newdirfid;
2570 V9fsString old_name, new_name;
2572 v9fs_string_init(&old_name);
2573 v9fs_string_init(&new_name);
2574 err = pdu_unmarshal(pdu, offset, "dsds", &olddirfid,
2575 &old_name, &newdirfid, &new_name);
2580 v9fs_path_write_lock(s);
2581 err = v9fs_complete_renameat(pdu, olddirfid,
2582 &old_name, newdirfid, &new_name);
2583 v9fs_path_unlock(s);
2589 complete_pdu(s, pdu, err);
2590 v9fs_string_free(&old_name);
2591 v9fs_string_free(&new_name);
2594 static void v9fs_wstat(void *opaque)
2603 V9fsPDU *pdu = opaque;
2604 V9fsState *s = pdu->s;
2606 v9fs_stat_init(&v9stat);
2607 err = pdu_unmarshal(pdu, offset, "dwS", &fid, &unused, &v9stat);
2611 trace_v9fs_wstat(pdu->tag, pdu->id, fid,
2612 v9stat.mode, v9stat.atime, v9stat.mtime);
2614 fidp = get_fid(pdu, fid);
2619 /* do we need to sync the file? */
2620 if (donttouch_stat(&v9stat)) {
2621 err = v9fs_co_fsync(pdu, fidp, 0);
2624 if (v9stat.mode != -1) {
2626 err = v9fs_co_lstat(pdu, &fidp->path, &stbuf);
2630 v9_mode = stat_to_v9mode(&stbuf);
2631 if ((v9stat.mode & P9_STAT_MODE_TYPE_BITS) !=
2632 (v9_mode & P9_STAT_MODE_TYPE_BITS)) {
2633 /* Attempting to change the type */
2637 err = v9fs_co_chmod(pdu, &fidp->path,
2638 v9mode_to_mode(v9stat.mode,
2639 &v9stat.extension));
2644 if (v9stat.mtime != -1 || v9stat.atime != -1) {
2645 struct timespec times[2];
2646 if (v9stat.atime != -1) {
2647 times[0].tv_sec = v9stat.atime;
2648 times[0].tv_nsec = 0;
2650 times[0].tv_nsec = UTIME_OMIT;
2652 if (v9stat.mtime != -1) {
2653 times[1].tv_sec = v9stat.mtime;
2654 times[1].tv_nsec = 0;
2656 times[1].tv_nsec = UTIME_OMIT;
2658 err = v9fs_co_utimensat(pdu, &fidp->path, times);
2663 if (v9stat.n_gid != -1 || v9stat.n_uid != -1) {
2664 err = v9fs_co_chown(pdu, &fidp->path, v9stat.n_uid, v9stat.n_gid);
2669 if (v9stat.name.size != 0) {
2670 err = v9fs_complete_rename(pdu, fidp, -1, &v9stat.name);
2675 if (v9stat.length != -1) {
2676 err = v9fs_co_truncate(pdu, &fidp->path, v9stat.length);
2685 v9fs_stat_free(&v9stat);
2686 complete_pdu(s, pdu, err);
2689 static int v9fs_fill_statfs(V9fsState *s, V9fsPDU *pdu, struct statfs *stbuf)
2701 int32_t bsize_factor;
2704 * compute bsize factor based on host file system block size
2707 bsize_factor = (s->msize - P9_IOHDRSZ)/stbuf->f_bsize;
2708 if (!bsize_factor) {
2711 f_type = stbuf->f_type;
2712 f_bsize = stbuf->f_bsize;
2713 f_bsize *= bsize_factor;
2715 * f_bsize is adjusted(multiplied) by bsize factor, so we need to
2716 * adjust(divide) the number of blocks, free blocks and available
2717 * blocks by bsize factor
2719 f_blocks = stbuf->f_blocks/bsize_factor;
2720 f_bfree = stbuf->f_bfree/bsize_factor;
2721 f_bavail = stbuf->f_bavail/bsize_factor;
2722 f_files = stbuf->f_files;
2723 f_ffree = stbuf->f_ffree;
2724 fsid_val = (unsigned int) stbuf->f_fsid.__val[0] |
2725 (unsigned long long)stbuf->f_fsid.__val[1] << 32;
2726 f_namelen = stbuf->f_namelen;
2728 return pdu_marshal(pdu, offset, "ddqqqqqqd",
2729 f_type, f_bsize, f_blocks, f_bfree,
2730 f_bavail, f_files, f_ffree,
2731 fsid_val, f_namelen);
2734 static void v9fs_statfs(void *opaque)
2740 struct statfs stbuf;
2741 V9fsPDU *pdu = opaque;
2742 V9fsState *s = pdu->s;
2744 retval = pdu_unmarshal(pdu, offset, "d", &fid);
2748 fidp = get_fid(pdu, fid);
2753 retval = v9fs_co_statfs(pdu, &fidp->path, &stbuf);
2757 retval = v9fs_fill_statfs(s, pdu, &stbuf);
2765 complete_pdu(s, pdu, retval);
2769 static void v9fs_mknod(void *opaque)
2782 V9fsPDU *pdu = opaque;
2783 V9fsState *s = pdu->s;
2785 v9fs_string_init(&name);
2786 err = pdu_unmarshal(pdu, offset, "dsdddd", &fid, &name, &mode,
2787 &major, &minor, &gid);
2791 trace_v9fs_mknod(pdu->tag, pdu->id, fid, mode, major, minor);
2793 fidp = get_fid(pdu, fid);
2798 err = v9fs_co_mknod(pdu, fidp, &name, fidp->uid, gid,
2799 makedev(major, minor), mode, &stbuf);
2803 stat_to_qid(&stbuf, &qid);
2804 err = pdu_marshal(pdu, offset, "Q", &qid);
2809 trace_v9fs_mknod_return(pdu->tag, pdu->id,
2810 qid.type, qid.version, qid.path);
2814 complete_pdu(s, pdu, err);
2815 v9fs_string_free(&name);
2819 * Implement posix byte range locking code
2820 * Server side handling of locking code is very simple, because 9p server in
2821 * QEMU can handle only one client. And most of the lock handling
2822 * (like conflict, merging) etc is done by the VFS layer itself, so no need to
2823 * do any thing in * qemu 9p server side lock code path.
2824 * So when a TLOCK request comes, always return success
2826 static void v9fs_lock(void *opaque)
2833 int32_t fid, err = 0;
2834 V9fsPDU *pdu = opaque;
2835 V9fsState *s = pdu->s;
2837 status = P9_LOCK_ERROR;
2838 v9fs_string_init(&flock.client_id);
2839 err = pdu_unmarshal(pdu, offset, "dbdqqds", &fid, &flock.type,
2840 &flock.flags, &flock.start, &flock.length,
2841 &flock.proc_id, &flock.client_id);
2845 trace_v9fs_lock(pdu->tag, pdu->id, fid,
2846 flock.type, flock.start, flock.length);
2849 /* We support only block flag now (that too ignored currently) */
2850 if (flock.flags & ~P9_LOCK_FLAGS_BLOCK) {
2854 fidp = get_fid(pdu, fid);
2859 err = v9fs_co_fstat(pdu, fidp, &stbuf);
2863 status = P9_LOCK_SUCCESS;
2867 err = pdu_marshal(pdu, offset, "b", status);
2871 trace_v9fs_lock_return(pdu->tag, pdu->id, status);
2872 complete_pdu(s, pdu, err);
2873 v9fs_string_free(&flock.client_id);
2877 * When a TGETLOCK request comes, always return success because all lock
2878 * handling is done by client's VFS layer.
2880 static void v9fs_getlock(void *opaque)
2886 int32_t fid, err = 0;
2887 V9fsPDU *pdu = opaque;
2888 V9fsState *s = pdu->s;
2890 v9fs_string_init(&glock.client_id);
2891 err = pdu_unmarshal(pdu, offset, "dbqqds", &fid, &glock.type,
2892 &glock.start, &glock.length, &glock.proc_id,
2897 trace_v9fs_getlock(pdu->tag, pdu->id, fid,
2898 glock.type, glock.start, glock.length);
2900 fidp = get_fid(pdu, fid);
2905 err = v9fs_co_fstat(pdu, fidp, &stbuf);
2909 glock.type = P9_LOCK_TYPE_UNLCK;
2910 err = pdu_marshal(pdu, offset, "bqqds", glock.type,
2911 glock.start, glock.length, glock.proc_id,
2917 trace_v9fs_getlock_return(pdu->tag, pdu->id, glock.type, glock.start,
2918 glock.length, glock.proc_id);
2922 complete_pdu(s, pdu, err);
2923 v9fs_string_free(&glock.client_id);
2926 static void v9fs_mkdir(void *opaque)
2928 V9fsPDU *pdu = opaque;
2939 v9fs_string_init(&name);
2940 err = pdu_unmarshal(pdu, offset, "dsdd", &fid, &name, &mode, &gid);
2944 trace_v9fs_mkdir(pdu->tag, pdu->id, fid, name.data, mode, gid);
2946 fidp = get_fid(pdu, fid);
2951 err = v9fs_co_mkdir(pdu, fidp, &name, mode, fidp->uid, gid, &stbuf);
2955 stat_to_qid(&stbuf, &qid);
2956 err = pdu_marshal(pdu, offset, "Q", &qid);
2961 trace_v9fs_mkdir_return(pdu->tag, pdu->id,
2962 qid.type, qid.version, qid.path, err);
2966 complete_pdu(pdu->s, pdu, err);
2967 v9fs_string_free(&name);
2970 static void v9fs_xattrwalk(void *opaque)
2976 int32_t fid, newfid;
2977 V9fsFidState *file_fidp;
2978 V9fsFidState *xattr_fidp = NULL;
2979 V9fsPDU *pdu = opaque;
2980 V9fsState *s = pdu->s;
2982 v9fs_string_init(&name);
2983 err = pdu_unmarshal(pdu, offset, "dds", &fid, &newfid, &name);
2987 trace_v9fs_xattrwalk(pdu->tag, pdu->id, fid, newfid, name.data);
2989 file_fidp = get_fid(pdu, fid);
2990 if (file_fidp == NULL) {
2994 xattr_fidp = alloc_fid(s, newfid);
2995 if (xattr_fidp == NULL) {
2999 v9fs_path_copy(&xattr_fidp->path, &file_fidp->path);
3000 if (name.data == NULL) {
3002 * listxattr request. Get the size first
3004 size = v9fs_co_llistxattr(pdu, &xattr_fidp->path, NULL, 0);
3007 clunk_fid(s, xattr_fidp->fid);
3011 * Read the xattr value
3013 xattr_fidp->fs.xattr.len = size;
3014 xattr_fidp->fid_type = P9_FID_XATTR;
3015 xattr_fidp->fs.xattr.copied_len = -1;
3017 xattr_fidp->fs.xattr.value = g_malloc(size);
3018 err = v9fs_co_llistxattr(pdu, &xattr_fidp->path,
3019 xattr_fidp->fs.xattr.value,
3020 xattr_fidp->fs.xattr.len);
3022 clunk_fid(s, xattr_fidp->fid);
3026 err = pdu_marshal(pdu, offset, "q", size);
3033 * specific xattr fid. We check for xattr
3034 * presence also collect the xattr size
3036 size = v9fs_co_lgetxattr(pdu, &xattr_fidp->path,
3040 clunk_fid(s, xattr_fidp->fid);
3044 * Read the xattr value
3046 xattr_fidp->fs.xattr.len = size;
3047 xattr_fidp->fid_type = P9_FID_XATTR;
3048 xattr_fidp->fs.xattr.copied_len = -1;
3050 xattr_fidp->fs.xattr.value = g_malloc(size);
3051 err = v9fs_co_lgetxattr(pdu, &xattr_fidp->path,
3052 &name, xattr_fidp->fs.xattr.value,
3053 xattr_fidp->fs.xattr.len);
3055 clunk_fid(s, xattr_fidp->fid);
3059 err = pdu_marshal(pdu, offset, "q", size);
3065 trace_v9fs_xattrwalk_return(pdu->tag, pdu->id, size);
3067 put_fid(pdu, file_fidp);
3069 put_fid(pdu, xattr_fidp);
3072 complete_pdu(s, pdu, err);
3073 v9fs_string_free(&name);
3076 static void v9fs_xattrcreate(void *opaque)
3084 V9fsFidState *file_fidp;
3085 V9fsFidState *xattr_fidp;
3086 V9fsPDU *pdu = opaque;
3087 V9fsState *s = pdu->s;
3089 v9fs_string_init(&name);
3090 err = pdu_unmarshal(pdu, offset, "dsqd", &fid, &name, &size, &flags);
3094 trace_v9fs_xattrcreate(pdu->tag, pdu->id, fid, name.data, size, flags);
3096 file_fidp = get_fid(pdu, fid);
3097 if (file_fidp == NULL) {
3101 /* Make the file fid point to xattr */
3102 xattr_fidp = file_fidp;
3103 xattr_fidp->fid_type = P9_FID_XATTR;
3104 xattr_fidp->fs.xattr.copied_len = 0;
3105 xattr_fidp->fs.xattr.len = size;
3106 xattr_fidp->fs.xattr.flags = flags;
3107 v9fs_string_init(&xattr_fidp->fs.xattr.name);
3108 v9fs_string_copy(&xattr_fidp->fs.xattr.name, &name);
3110 xattr_fidp->fs.xattr.value = g_malloc(size);
3112 xattr_fidp->fs.xattr.value = NULL;
3115 put_fid(pdu, file_fidp);
3117 complete_pdu(s, pdu, err);
3118 v9fs_string_free(&name);
3121 static void v9fs_readlink(void *opaque)
3123 V9fsPDU *pdu = opaque;
3130 err = pdu_unmarshal(pdu, offset, "d", &fid);
3134 trace_v9fs_readlink(pdu->tag, pdu->id, fid);
3135 fidp = get_fid(pdu, fid);
3141 v9fs_string_init(&target);
3142 err = v9fs_co_readlink(pdu, &fidp->path, &target);
3146 err = pdu_marshal(pdu, offset, "s", &target);
3148 v9fs_string_free(&target);
3152 trace_v9fs_readlink_return(pdu->tag, pdu->id, target.data);
3153 v9fs_string_free(&target);
3157 complete_pdu(pdu->s, pdu, err);
3160 static CoroutineEntry *pdu_co_handlers[] = {
3161 [P9_TREADDIR] = v9fs_readdir,
3162 [P9_TSTATFS] = v9fs_statfs,
3163 [P9_TGETATTR] = v9fs_getattr,
3164 [P9_TSETATTR] = v9fs_setattr,
3165 [P9_TXATTRWALK] = v9fs_xattrwalk,
3166 [P9_TXATTRCREATE] = v9fs_xattrcreate,
3167 [P9_TMKNOD] = v9fs_mknod,
3168 [P9_TRENAME] = v9fs_rename,
3169 [P9_TLOCK] = v9fs_lock,
3170 [P9_TGETLOCK] = v9fs_getlock,
3171 [P9_TRENAMEAT] = v9fs_renameat,
3172 [P9_TREADLINK] = v9fs_readlink,
3173 [P9_TUNLINKAT] = v9fs_unlinkat,
3174 [P9_TMKDIR] = v9fs_mkdir,
3175 [P9_TVERSION] = v9fs_version,
3176 [P9_TLOPEN] = v9fs_open,
3177 [P9_TATTACH] = v9fs_attach,
3178 [P9_TSTAT] = v9fs_stat,
3179 [P9_TWALK] = v9fs_walk,
3180 [P9_TCLUNK] = v9fs_clunk,
3181 [P9_TFSYNC] = v9fs_fsync,
3182 [P9_TOPEN] = v9fs_open,
3183 [P9_TREAD] = v9fs_read,
3185 [P9_TAUTH] = v9fs_auth,
3187 [P9_TFLUSH] = v9fs_flush,
3188 [P9_TLINK] = v9fs_link,
3189 [P9_TSYMLINK] = v9fs_symlink,
3190 [P9_TCREATE] = v9fs_create,
3191 [P9_TLCREATE] = v9fs_lcreate,
3192 [P9_TWRITE] = v9fs_write,
3193 [P9_TWSTAT] = v9fs_wstat,
3194 [P9_TREMOVE] = v9fs_remove,
3197 static void v9fs_op_not_supp(void *opaque)
3199 V9fsPDU *pdu = opaque;
3200 complete_pdu(pdu->s, pdu, -EOPNOTSUPP);
3203 static void v9fs_fs_ro(void *opaque)
3205 V9fsPDU *pdu = opaque;
3206 complete_pdu(pdu->s, pdu, -EROFS);
3209 static inline bool is_read_only_op(V9fsPDU *pdu)
3236 static void submit_pdu(V9fsState *s, V9fsPDU *pdu)
3239 CoroutineEntry *handler;
3241 if (pdu->id >= ARRAY_SIZE(pdu_co_handlers) ||
3242 (pdu_co_handlers[pdu->id] == NULL)) {
3243 handler = v9fs_op_not_supp;
3245 handler = pdu_co_handlers[pdu->id];
3248 if (is_ro_export(&s->ctx) && !is_read_only_op(pdu)) {
3249 handler = v9fs_fs_ro;
3251 co = qemu_coroutine_create(handler);
3252 qemu_coroutine_enter(co, pdu);
3255 void handle_9p_output(VirtIODevice *vdev, VirtQueue *vq)
3257 V9fsState *s = (V9fsState *)vdev;
3261 while ((pdu = alloc_pdu(s)) &&
3262 (len = virtqueue_pop(vq, &pdu->elem)) != 0) {
3265 BUG_ON(pdu->elem.out_num == 0 || pdu->elem.in_num == 0);
3266 BUG_ON(pdu->elem.out_sg[0].iov_len < 7);
3268 ptr = pdu->elem.out_sg[0].iov_base;
3270 pdu->size = le32_to_cpu(*(uint32_t *)ptr);
3272 pdu->tag = le16_to_cpu(*(uint16_t *)(ptr + 5));
3273 qemu_co_queue_init(&pdu->complete);
3279 void virtio_9p_set_fd_limit(void)
3282 if (getrlimit(RLIMIT_NOFILE, &rlim) < 0) {
3283 fprintf(stderr, "Failed to get the resource limit\n");
3286 open_fd_hw = rlim.rlim_cur - MIN(400, rlim.rlim_cur/3);
3287 open_fd_rc = rlim.rlim_cur/2;