4 * Copyright IBM, Corp. 2010
7 * Anthony Liguori <aliguori@us.ibm.com>
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 */
988 error_set(&s
->migration_blocker
, QERR_VIRTFS_FEATURE_BLOCKS_MIGRATION
,
989 s
->ctx
.fs_root
, s
->tag
);
990 migrate_add_blocker(s
->migration_blocker
);
994 complete_pdu(s
, pdu
, err
);
995 v9fs_string_free(&uname
);
996 v9fs_string_free(&aname
);
999 static void v9fs_stat(void *opaque
)
1007 V9fsPDU
*pdu
= opaque
;
1008 V9fsState
*s
= pdu
->s
;
1010 err
= pdu_unmarshal(pdu
, offset
, "d", &fid
);
1014 trace_v9fs_stat(pdu
->tag
, pdu
->id
, fid
);
1016 fidp
= get_fid(pdu
, fid
);
1021 err
= v9fs_co_lstat(pdu
, &fidp
->path
, &stbuf
);
1025 err
= stat_to_v9stat(pdu
, &fidp
->path
, &stbuf
, &v9stat
);
1029 err
= pdu_marshal(pdu
, offset
, "wS", 0, &v9stat
);
1031 v9fs_stat_free(&v9stat
);
1034 trace_v9fs_stat_return(pdu
->tag
, pdu
->id
, v9stat
.mode
,
1035 v9stat
.atime
, v9stat
.mtime
, v9stat
.length
);
1037 v9fs_stat_free(&v9stat
);
1041 complete_pdu(s
, pdu
, err
);
1044 static void v9fs_getattr(void *opaque
)
1051 uint64_t request_mask
;
1052 V9fsStatDotl v9stat_dotl
;
1053 V9fsPDU
*pdu
= opaque
;
1054 V9fsState
*s
= pdu
->s
;
1056 retval
= pdu_unmarshal(pdu
, offset
, "dq", &fid
, &request_mask
);
1060 trace_v9fs_getattr(pdu
->tag
, pdu
->id
, fid
, request_mask
);
1062 fidp
= get_fid(pdu
, fid
);
1068 * Currently we only support BASIC fields in stat, so there is no
1069 * need to look at request_mask.
1071 retval
= v9fs_co_lstat(pdu
, &fidp
->path
, &stbuf
);
1075 stat_to_v9stat_dotl(s
, &stbuf
, &v9stat_dotl
);
1077 /* fill st_gen if requested and supported by underlying fs */
1078 if (request_mask
& P9_STATS_GEN
) {
1079 retval
= v9fs_co_st_gen(pdu
, &fidp
->path
, stbuf
.st_mode
, &v9stat_dotl
);
1083 v9stat_dotl
.st_result_mask
|= P9_STATS_GEN
;
1085 retval
= pdu_marshal(pdu
, offset
, "A", &v9stat_dotl
);
1090 trace_v9fs_getattr_return(pdu
->tag
, pdu
->id
, v9stat_dotl
.st_result_mask
,
1091 v9stat_dotl
.st_mode
, v9stat_dotl
.st_uid
,
1092 v9stat_dotl
.st_gid
);
1096 complete_pdu(s
, pdu
, retval
);
1099 /* Attribute flags */
1100 #define P9_ATTR_MODE (1 << 0)
1101 #define P9_ATTR_UID (1 << 1)
1102 #define P9_ATTR_GID (1 << 2)
1103 #define P9_ATTR_SIZE (1 << 3)
1104 #define P9_ATTR_ATIME (1 << 4)
1105 #define P9_ATTR_MTIME (1 << 5)
1106 #define P9_ATTR_CTIME (1 << 6)
1107 #define P9_ATTR_ATIME_SET (1 << 7)
1108 #define P9_ATTR_MTIME_SET (1 << 8)
1110 #define P9_ATTR_MASK 127
1112 static void v9fs_setattr(void *opaque
)
1119 V9fsPDU
*pdu
= opaque
;
1120 V9fsState
*s
= pdu
->s
;
1122 err
= pdu_unmarshal(pdu
, offset
, "dI", &fid
, &v9iattr
);
1127 fidp
= get_fid(pdu
, fid
);
1132 if (v9iattr
.valid
& P9_ATTR_MODE
) {
1133 err
= v9fs_co_chmod(pdu
, &fidp
->path
, v9iattr
.mode
);
1138 if (v9iattr
.valid
& (P9_ATTR_ATIME
| P9_ATTR_MTIME
)) {
1139 struct timespec times
[2];
1140 if (v9iattr
.valid
& P9_ATTR_ATIME
) {
1141 if (v9iattr
.valid
& P9_ATTR_ATIME_SET
) {
1142 times
[0].tv_sec
= v9iattr
.atime_sec
;
1143 times
[0].tv_nsec
= v9iattr
.atime_nsec
;
1145 times
[0].tv_nsec
= UTIME_NOW
;
1148 times
[0].tv_nsec
= UTIME_OMIT
;
1150 if (v9iattr
.valid
& P9_ATTR_MTIME
) {
1151 if (v9iattr
.valid
& P9_ATTR_MTIME_SET
) {
1152 times
[1].tv_sec
= v9iattr
.mtime_sec
;
1153 times
[1].tv_nsec
= v9iattr
.mtime_nsec
;
1155 times
[1].tv_nsec
= UTIME_NOW
;
1158 times
[1].tv_nsec
= UTIME_OMIT
;
1160 err
= v9fs_co_utimensat(pdu
, &fidp
->path
, times
);
1166 * If the only valid entry in iattr is ctime we can call
1167 * chown(-1,-1) to update the ctime of the file
1169 if ((v9iattr
.valid
& (P9_ATTR_UID
| P9_ATTR_GID
)) ||
1170 ((v9iattr
.valid
& P9_ATTR_CTIME
)
1171 && !((v9iattr
.valid
& P9_ATTR_MASK
) & ~P9_ATTR_CTIME
))) {
1172 if (!(v9iattr
.valid
& P9_ATTR_UID
)) {
1175 if (!(v9iattr
.valid
& P9_ATTR_GID
)) {
1178 err
= v9fs_co_chown(pdu
, &fidp
->path
, v9iattr
.uid
,
1184 if (v9iattr
.valid
& (P9_ATTR_SIZE
)) {
1185 err
= v9fs_co_truncate(pdu
, &fidp
->path
, v9iattr
.size
);
1194 complete_pdu(s
, pdu
, err
);
1197 static int v9fs_walk_marshal(V9fsPDU
*pdu
, uint16_t nwnames
, V9fsQID
*qids
)
1203 err
= pdu_marshal(pdu
, offset
, "w", nwnames
);
1208 for (i
= 0; i
< nwnames
; i
++) {
1209 err
= pdu_marshal(pdu
, offset
, "Q", &qids
[i
]);
1218 static void v9fs_walk(void *opaque
)
1221 V9fsQID
*qids
= NULL
;
1223 V9fsPath dpath
, path
;
1227 int32_t fid
, newfid
;
1228 V9fsString
*wnames
= NULL
;
1230 V9fsFidState
*newfidp
= NULL
;
1231 V9fsPDU
*pdu
= opaque
;
1232 V9fsState
*s
= pdu
->s
;
1234 err
= pdu_unmarshal(pdu
, offset
, "ddw", &fid
, &newfid
, &nwnames
);
1236 complete_pdu(s
, pdu
, err
);
1241 trace_v9fs_walk(pdu
->tag
, pdu
->id
, fid
, newfid
, nwnames
);
1243 if (nwnames
&& nwnames
<= P9_MAXWELEM
) {
1244 wnames
= g_malloc0(sizeof(wnames
[0]) * nwnames
);
1245 qids
= g_malloc0(sizeof(qids
[0]) * nwnames
);
1246 for (i
= 0; i
< nwnames
; i
++) {
1247 err
= pdu_unmarshal(pdu
, offset
, "s", &wnames
[i
]);
1253 } else if (nwnames
> P9_MAXWELEM
) {
1257 fidp
= get_fid(pdu
, fid
);
1262 v9fs_path_init(&dpath
);
1263 v9fs_path_init(&path
);
1265 * Both dpath and path initially poin to fidp.
1266 * Needed to handle request with nwnames == 0
1268 v9fs_path_copy(&dpath
, &fidp
->path
);
1269 v9fs_path_copy(&path
, &fidp
->path
);
1270 for (name_idx
= 0; name_idx
< nwnames
; name_idx
++) {
1271 err
= v9fs_co_name_to_path(pdu
, &dpath
, wnames
[name_idx
].data
, &path
);
1275 err
= v9fs_co_lstat(pdu
, &path
, &stbuf
);
1279 stat_to_qid(&stbuf
, &qids
[name_idx
]);
1280 v9fs_path_copy(&dpath
, &path
);
1282 if (fid
== newfid
) {
1283 BUG_ON(fidp
->fid_type
!= P9_FID_NONE
);
1284 v9fs_path_copy(&fidp
->path
, &path
);
1286 newfidp
= alloc_fid(s
, newfid
);
1287 if (newfidp
== NULL
) {
1291 newfidp
->uid
= fidp
->uid
;
1292 v9fs_path_copy(&newfidp
->path
, &path
);
1294 err
= v9fs_walk_marshal(pdu
, nwnames
, qids
);
1295 trace_v9fs_walk_return(pdu
->tag
, pdu
->id
, nwnames
, qids
);
1299 put_fid(pdu
, newfidp
);
1301 v9fs_path_free(&dpath
);
1302 v9fs_path_free(&path
);
1304 complete_pdu(s
, pdu
, err
);
1305 if (nwnames
&& nwnames
<= P9_MAXWELEM
) {
1306 for (name_idx
= 0; name_idx
< nwnames
; name_idx
++) {
1307 v9fs_string_free(&wnames
[name_idx
]);
1315 static int32_t get_iounit(V9fsPDU
*pdu
, V9fsPath
*path
)
1317 struct statfs stbuf
;
1319 V9fsState
*s
= pdu
->s
;
1322 * iounit should be multiples of f_bsize (host filesystem block size
1323 * and as well as less than (client msize - P9_IOHDRSZ))
1325 if (!v9fs_co_statfs(pdu
, path
, &stbuf
)) {
1326 iounit
= stbuf
.f_bsize
;
1327 iounit
*= (s
->msize
- P9_IOHDRSZ
)/stbuf
.f_bsize
;
1330 iounit
= s
->msize
- P9_IOHDRSZ
;
1335 static void v9fs_open(void *opaque
)
1346 V9fsPDU
*pdu
= opaque
;
1347 V9fsState
*s
= pdu
->s
;
1349 if (s
->proto_version
== V9FS_PROTO_2000L
) {
1350 err
= pdu_unmarshal(pdu
, offset
, "dd", &fid
, &mode
);
1352 err
= pdu_unmarshal(pdu
, offset
, "db", &fid
, &mode
);
1357 trace_v9fs_open(pdu
->tag
, pdu
->id
, fid
, mode
);
1359 fidp
= get_fid(pdu
, fid
);
1364 BUG_ON(fidp
->fid_type
!= P9_FID_NONE
);
1366 err
= v9fs_co_lstat(pdu
, &fidp
->path
, &stbuf
);
1370 stat_to_qid(&stbuf
, &qid
);
1371 if (S_ISDIR(stbuf
.st_mode
)) {
1372 err
= v9fs_co_opendir(pdu
, fidp
);
1376 fidp
->fid_type
= P9_FID_DIR
;
1377 err
= pdu_marshal(pdu
, offset
, "Qd", &qid
, 0);
1383 if (s
->proto_version
== V9FS_PROTO_2000L
) {
1384 flags
= get_dotl_openflags(s
, mode
);
1386 flags
= omode_to_uflags(mode
);
1388 if (is_ro_export(&s
->ctx
)) {
1389 if (mode
& O_WRONLY
|| mode
& O_RDWR
||
1390 mode
& O_APPEND
|| mode
& O_TRUNC
) {
1396 err
= v9fs_co_open(pdu
, fidp
, flags
);
1400 fidp
->fid_type
= P9_FID_FILE
;
1401 fidp
->open_flags
= flags
;
1402 if (flags
& O_EXCL
) {
1404 * We let the host file system do O_EXCL check
1405 * We should not reclaim such fd
1407 fidp
->flags
|= FID_NON_RECLAIMABLE
;
1409 iounit
= get_iounit(pdu
, &fidp
->path
);
1410 err
= pdu_marshal(pdu
, offset
, "Qd", &qid
, iounit
);
1416 trace_v9fs_open_return(pdu
->tag
, pdu
->id
,
1417 qid
.type
, qid
.version
, qid
.path
, iounit
);
1421 complete_pdu(s
, pdu
, err
);
1424 static void v9fs_lcreate(void *opaque
)
1426 int32_t dfid
, flags
, mode
;
1435 V9fsPDU
*pdu
= opaque
;
1437 v9fs_string_init(&name
);
1438 err
= pdu_unmarshal(pdu
, offset
, "dsddd", &dfid
,
1439 &name
, &flags
, &mode
, &gid
);
1443 trace_v9fs_lcreate(pdu
->tag
, pdu
->id
, dfid
, flags
, mode
, gid
);
1445 fidp
= get_fid(pdu
, dfid
);
1451 flags
= get_dotl_openflags(pdu
->s
, flags
);
1452 err
= v9fs_co_open2(pdu
, fidp
, &name
, gid
,
1453 flags
| O_CREAT
, mode
, &stbuf
);
1457 fidp
->fid_type
= P9_FID_FILE
;
1458 fidp
->open_flags
= flags
;
1459 if (flags
& O_EXCL
) {
1461 * We let the host file system do O_EXCL check
1462 * We should not reclaim such fd
1464 fidp
->flags
|= FID_NON_RECLAIMABLE
;
1466 iounit
= get_iounit(pdu
, &fidp
->path
);
1467 stat_to_qid(&stbuf
, &qid
);
1468 err
= pdu_marshal(pdu
, offset
, "Qd", &qid
, iounit
);
1473 trace_v9fs_lcreate_return(pdu
->tag
, pdu
->id
,
1474 qid
.type
, qid
.version
, qid
.path
, iounit
);
1478 complete_pdu(pdu
->s
, pdu
, err
);
1479 v9fs_string_free(&name
);
1482 static void v9fs_fsync(void *opaque
)
1489 V9fsPDU
*pdu
= opaque
;
1490 V9fsState
*s
= pdu
->s
;
1492 err
= pdu_unmarshal(pdu
, offset
, "dd", &fid
, &datasync
);
1496 trace_v9fs_fsync(pdu
->tag
, pdu
->id
, fid
, datasync
);
1498 fidp
= get_fid(pdu
, fid
);
1503 err
= v9fs_co_fsync(pdu
, fidp
, datasync
);
1509 complete_pdu(s
, pdu
, err
);
1512 static void v9fs_clunk(void *opaque
)
1518 V9fsPDU
*pdu
= opaque
;
1519 V9fsState
*s
= pdu
->s
;
1521 err
= pdu_unmarshal(pdu
, offset
, "d", &fid
);
1525 trace_v9fs_clunk(pdu
->tag
, pdu
->id
, fid
);
1527 fidp
= clunk_fid(s
, fid
);
1533 * Bump the ref so that put_fid will
1541 complete_pdu(s
, pdu
, err
);
1544 static int v9fs_xattr_read(V9fsState
*s
, V9fsPDU
*pdu
, V9fsFidState
*fidp
,
1545 uint64_t off
, uint32_t max_count
)
1552 xattr_len
= fidp
->fs
.xattr
.len
;
1553 read_count
= xattr_len
- off
;
1554 if (read_count
> max_count
) {
1555 read_count
= max_count
;
1556 } else if (read_count
< 0) {
1558 * read beyond XATTR value
1562 err
= pdu_marshal(pdu
, offset
, "d", read_count
);
1567 err
= v9fs_pack(pdu
->elem
.in_sg
, pdu
->elem
.in_num
, offset
,
1568 ((char *)fidp
->fs
.xattr
.value
) + off
,
1577 static int v9fs_do_readdir_with_stat(V9fsPDU
*pdu
,
1578 V9fsFidState
*fidp
, uint32_t max_count
)
1585 off_t saved_dir_pos
;
1586 struct dirent
*dent
, *result
;
1588 /* save the directory position */
1589 saved_dir_pos
= v9fs_co_telldir(pdu
, fidp
);
1590 if (saved_dir_pos
< 0) {
1591 return saved_dir_pos
;
1594 dent
= g_malloc(sizeof(struct dirent
));
1597 v9fs_path_init(&path
);
1598 err
= v9fs_co_readdir_r(pdu
, fidp
, dent
, &result
);
1599 if (err
|| !result
) {
1602 err
= v9fs_co_name_to_path(pdu
, &fidp
->path
, dent
->d_name
, &path
);
1606 err
= v9fs_co_lstat(pdu
, &path
, &stbuf
);
1610 err
= stat_to_v9stat(pdu
, &path
, &stbuf
, &v9stat
);
1614 /* 11 = 7 + 4 (7 = start offset, 4 = space for storing count) */
1615 len
= pdu_marshal(pdu
, 11 + count
, "S", &v9stat
);
1616 if ((len
!= (v9stat
.size
+ 2)) || ((count
+ len
) > max_count
)) {
1617 /* Ran out of buffer. Set dir back to old position and return */
1618 v9fs_co_seekdir(pdu
, fidp
, saved_dir_pos
);
1619 v9fs_stat_free(&v9stat
);
1620 v9fs_path_free(&path
);
1625 v9fs_stat_free(&v9stat
);
1626 v9fs_path_free(&path
);
1627 saved_dir_pos
= dent
->d_off
;
1631 v9fs_path_free(&path
);
1639 * Create a QEMUIOVector for a sub-region of PDU iovecs
1641 * @qiov: uninitialized QEMUIOVector
1642 * @skip: number of bytes to skip from beginning of PDU
1643 * @size: number of bytes to include
1644 * @is_write: true - write, false - read
1646 * The resulting QEMUIOVector has heap-allocated iovecs and must be cleaned up
1647 * with qemu_iovec_destroy().
1649 static void v9fs_init_qiov_from_pdu(QEMUIOVector
*qiov
, V9fsPDU
*pdu
,
1650 uint64_t skip
, size_t size
,
1658 iov
= pdu
->elem
.out_sg
;
1659 niov
= pdu
->elem
.out_num
;
1661 iov
= pdu
->elem
.in_sg
;
1662 niov
= pdu
->elem
.in_num
;
1665 qemu_iovec_init_external(&elem
, iov
, niov
);
1666 qemu_iovec_init(qiov
, niov
);
1667 qemu_iovec_copy(qiov
, &elem
, skip
, size
);
1670 static void v9fs_read(void *opaque
)
1679 V9fsPDU
*pdu
= opaque
;
1680 V9fsState
*s
= pdu
->s
;
1682 err
= pdu_unmarshal(pdu
, offset
, "dqd", &fid
, &off
, &max_count
);
1686 trace_v9fs_read(pdu
->tag
, pdu
->id
, fid
, off
, max_count
);
1688 fidp
= get_fid(pdu
, fid
);
1693 if (fidp
->fid_type
== P9_FID_DIR
) {
1696 v9fs_co_rewinddir(pdu
, fidp
);
1698 count
= v9fs_do_readdir_with_stat(pdu
, fidp
, max_count
);
1703 err
= pdu_marshal(pdu
, offset
, "d", count
);
1707 err
+= offset
+ count
;
1708 } else if (fidp
->fid_type
== P9_FID_FILE
) {
1709 QEMUIOVector qiov_full
;
1713 v9fs_init_qiov_from_pdu(&qiov_full
, pdu
, offset
+ 4, max_count
, false);
1714 qemu_iovec_init(&qiov
, qiov_full
.niov
);
1716 qemu_iovec_reset(&qiov
);
1717 qemu_iovec_copy(&qiov
, &qiov_full
, count
, qiov_full
.size
- count
);
1719 print_sg(qiov
.iov
, qiov
.niov
);
1721 /* Loop in case of EINTR */
1723 len
= v9fs_co_preadv(pdu
, fidp
, qiov
.iov
, qiov
.niov
, off
);
1728 } while (len
== -EINTR
&& !pdu
->cancelled
);
1730 /* IO error return the error */
1734 } while (count
< max_count
&& len
> 0);
1735 err
= pdu_marshal(pdu
, offset
, "d", count
);
1739 err
+= offset
+ count
;
1740 qemu_iovec_destroy(&qiov
);
1741 qemu_iovec_destroy(&qiov_full
);
1742 } else if (fidp
->fid_type
== P9_FID_XATTR
) {
1743 err
= v9fs_xattr_read(s
, pdu
, fidp
, off
, max_count
);
1747 trace_v9fs_read_return(pdu
->tag
, pdu
->id
, count
, err
);
1751 complete_pdu(s
, pdu
, err
);
1754 static size_t v9fs_readdir_data_size(V9fsString
*name
)
1757 * Size of each dirent on the wire: size of qid (13) + size of offset (8)
1758 * size of type (1) + size of name.size (2) + strlen(name.data)
1760 return 24 + v9fs_string_size(name
);
1763 static int v9fs_do_readdir(V9fsPDU
*pdu
,
1764 V9fsFidState
*fidp
, int32_t max_count
)
1771 off_t saved_dir_pos
;
1772 struct dirent
*dent
, *result
;
1774 /* save the directory position */
1775 saved_dir_pos
= v9fs_co_telldir(pdu
, fidp
);
1776 if (saved_dir_pos
< 0) {
1777 return saved_dir_pos
;
1780 dent
= g_malloc(sizeof(struct dirent
));
1783 err
= v9fs_co_readdir_r(pdu
, fidp
, dent
, &result
);
1784 if (err
|| !result
) {
1787 v9fs_string_init(&name
);
1788 v9fs_string_sprintf(&name
, "%s", dent
->d_name
);
1789 if ((count
+ v9fs_readdir_data_size(&name
)) > max_count
) {
1790 /* Ran out of buffer. Set dir back to old position and return */
1791 v9fs_co_seekdir(pdu
, fidp
, saved_dir_pos
);
1792 v9fs_string_free(&name
);
1797 * Fill up just the path field of qid because the client uses
1798 * only that. To fill the entire qid structure we will have
1799 * to stat each dirent found, which is expensive
1801 size
= MIN(sizeof(dent
->d_ino
), sizeof(qid
.path
));
1802 memcpy(&qid
.path
, &dent
->d_ino
, size
);
1803 /* Fill the other fields with dummy values */
1807 /* 11 = 7 + 4 (7 = start offset, 4 = space for storing count) */
1808 len
= pdu_marshal(pdu
, 11 + count
, "Qqbs",
1810 dent
->d_type
, &name
);
1812 v9fs_co_seekdir(pdu
, fidp
, saved_dir_pos
);
1813 v9fs_string_free(&name
);
1818 v9fs_string_free(&name
);
1819 saved_dir_pos
= dent
->d_off
;
1828 static void v9fs_readdir(void *opaque
)
1834 uint64_t initial_offset
;
1837 V9fsPDU
*pdu
= opaque
;
1838 V9fsState
*s
= pdu
->s
;
1840 retval
= pdu_unmarshal(pdu
, offset
, "dqd", &fid
,
1841 &initial_offset
, &max_count
);
1845 trace_v9fs_readdir(pdu
->tag
, pdu
->id
, fid
, initial_offset
, max_count
);
1847 fidp
= get_fid(pdu
, fid
);
1852 if (!fidp
->fs
.dir
) {
1856 if (initial_offset
== 0) {
1857 v9fs_co_rewinddir(pdu
, fidp
);
1859 v9fs_co_seekdir(pdu
, fidp
, initial_offset
);
1861 count
= v9fs_do_readdir(pdu
, fidp
, max_count
);
1866 retval
= pdu_marshal(pdu
, offset
, "d", count
);
1870 retval
+= count
+ offset
;
1871 trace_v9fs_readdir_return(pdu
->tag
, pdu
->id
, count
, retval
);
1875 complete_pdu(s
, pdu
, retval
);
1878 static int v9fs_xattr_write(V9fsState
*s
, V9fsPDU
*pdu
, V9fsFidState
*fidp
,
1879 uint64_t off
, uint32_t count
,
1880 struct iovec
*sg
, int cnt
)
1889 xattr_len
= fidp
->fs
.xattr
.len
;
1890 write_count
= xattr_len
- off
;
1891 if (write_count
> count
) {
1892 write_count
= count
;
1893 } else if (write_count
< 0) {
1895 * write beyond XATTR value len specified in
1901 err
= pdu_marshal(pdu
, offset
, "d", write_count
);
1906 fidp
->fs
.xattr
.copied_len
+= write_count
;
1908 * Now copy the content from sg list
1910 for (i
= 0; i
< cnt
; i
++) {
1911 if (write_count
> sg
[i
].iov_len
) {
1912 to_copy
= sg
[i
].iov_len
;
1914 to_copy
= write_count
;
1916 memcpy((char *)fidp
->fs
.xattr
.value
+ off
, sg
[i
].iov_base
, to_copy
);
1917 /* updating vs->off since we are not using below */
1919 write_count
-= to_copy
;
1925 static void v9fs_write(void *opaque
)
1935 V9fsPDU
*pdu
= opaque
;
1936 V9fsState
*s
= pdu
->s
;
1937 QEMUIOVector qiov_full
;
1940 err
= pdu_unmarshal(pdu
, offset
, "dqd", &fid
, &off
, &count
);
1942 return complete_pdu(s
, pdu
, err
);
1945 v9fs_init_qiov_from_pdu(&qiov_full
, pdu
, offset
, count
, true);
1946 trace_v9fs_write(pdu
->tag
, pdu
->id
, fid
, off
, count
, qiov_full
.niov
);
1948 fidp
= get_fid(pdu
, fid
);
1953 if (fidp
->fid_type
== P9_FID_FILE
) {
1954 if (fidp
->fs
.fd
== -1) {
1958 } else if (fidp
->fid_type
== P9_FID_XATTR
) {
1960 * setxattr operation
1962 err
= v9fs_xattr_write(s
, pdu
, fidp
, off
, count
,
1963 qiov_full
.iov
, qiov_full
.niov
);
1969 qemu_iovec_init(&qiov
, qiov_full
.niov
);
1971 qemu_iovec_reset(&qiov
);
1972 qemu_iovec_copy(&qiov
, &qiov_full
, total
, qiov_full
.size
- total
);
1974 print_sg(qiov
.iov
, qiov
.niov
);
1976 /* Loop in case of EINTR */
1978 len
= v9fs_co_pwritev(pdu
, fidp
, qiov
.iov
, qiov
.niov
, off
);
1983 } while (len
== -EINTR
&& !pdu
->cancelled
);
1985 /* IO error return the error */
1989 } while (total
< count
&& len
> 0);
1992 err
= pdu_marshal(pdu
, offset
, "d", total
);
1997 trace_v9fs_write_return(pdu
->tag
, pdu
->id
, total
, err
);
1999 qemu_iovec_destroy(&qiov
);
2003 qemu_iovec_destroy(&qiov_full
);
2004 complete_pdu(s
, pdu
, err
);
2007 static void v9fs_create(void *opaque
)
2019 V9fsString extension
;
2021 V9fsPDU
*pdu
= opaque
;
2023 v9fs_path_init(&path
);
2024 v9fs_string_init(&name
);
2025 v9fs_string_init(&extension
);
2026 err
= pdu_unmarshal(pdu
, offset
, "dsdbs", &fid
, &name
,
2027 &perm
, &mode
, &extension
);
2031 trace_v9fs_create(pdu
->tag
, pdu
->id
, fid
, name
.data
, perm
, mode
);
2033 fidp
= get_fid(pdu
, fid
);
2038 if (perm
& P9_STAT_MODE_DIR
) {
2039 err
= v9fs_co_mkdir(pdu
, fidp
, &name
, perm
& 0777,
2040 fidp
->uid
, -1, &stbuf
);
2044 err
= v9fs_co_name_to_path(pdu
, &fidp
->path
, name
.data
, &path
);
2048 v9fs_path_copy(&fidp
->path
, &path
);
2049 err
= v9fs_co_opendir(pdu
, fidp
);
2053 fidp
->fid_type
= P9_FID_DIR
;
2054 } else if (perm
& P9_STAT_MODE_SYMLINK
) {
2055 err
= v9fs_co_symlink(pdu
, fidp
, &name
,
2056 extension
.data
, -1 , &stbuf
);
2060 err
= v9fs_co_name_to_path(pdu
, &fidp
->path
, name
.data
, &path
);
2064 v9fs_path_copy(&fidp
->path
, &path
);
2065 } else if (perm
& P9_STAT_MODE_LINK
) {
2066 int32_t ofid
= atoi(extension
.data
);
2067 V9fsFidState
*ofidp
= get_fid(pdu
, ofid
);
2068 if (ofidp
== NULL
) {
2072 err
= v9fs_co_link(pdu
, ofidp
, fidp
, &name
);
2073 put_fid(pdu
, ofidp
);
2077 err
= v9fs_co_name_to_path(pdu
, &fidp
->path
, name
.data
, &path
);
2079 fidp
->fid_type
= P9_FID_NONE
;
2082 v9fs_path_copy(&fidp
->path
, &path
);
2083 err
= v9fs_co_lstat(pdu
, &fidp
->path
, &stbuf
);
2085 fidp
->fid_type
= P9_FID_NONE
;
2088 } else if (perm
& P9_STAT_MODE_DEVICE
) {
2090 uint32_t major
, minor
;
2093 if (sscanf(extension
.data
, "%c %u %u", &ctype
, &major
, &minor
) != 3) {
2110 nmode
|= perm
& 0777;
2111 err
= v9fs_co_mknod(pdu
, fidp
, &name
, fidp
->uid
, -1,
2112 makedev(major
, minor
), nmode
, &stbuf
);
2116 err
= v9fs_co_name_to_path(pdu
, &fidp
->path
, name
.data
, &path
);
2120 v9fs_path_copy(&fidp
->path
, &path
);
2121 } else if (perm
& P9_STAT_MODE_NAMED_PIPE
) {
2122 err
= v9fs_co_mknod(pdu
, fidp
, &name
, fidp
->uid
, -1,
2123 0, S_IFIFO
| (perm
& 0777), &stbuf
);
2127 err
= v9fs_co_name_to_path(pdu
, &fidp
->path
, name
.data
, &path
);
2131 v9fs_path_copy(&fidp
->path
, &path
);
2132 } else if (perm
& P9_STAT_MODE_SOCKET
) {
2133 err
= v9fs_co_mknod(pdu
, fidp
, &name
, fidp
->uid
, -1,
2134 0, S_IFSOCK
| (perm
& 0777), &stbuf
);
2138 err
= v9fs_co_name_to_path(pdu
, &fidp
->path
, name
.data
, &path
);
2142 v9fs_path_copy(&fidp
->path
, &path
);
2144 err
= v9fs_co_open2(pdu
, fidp
, &name
, -1,
2145 omode_to_uflags(mode
)|O_CREAT
, perm
, &stbuf
);
2149 fidp
->fid_type
= P9_FID_FILE
;
2150 fidp
->open_flags
= omode_to_uflags(mode
);
2151 if (fidp
->open_flags
& O_EXCL
) {
2153 * We let the host file system do O_EXCL check
2154 * We should not reclaim such fd
2156 fidp
->flags
|= FID_NON_RECLAIMABLE
;
2159 iounit
= get_iounit(pdu
, &fidp
->path
);
2160 stat_to_qid(&stbuf
, &qid
);
2161 err
= pdu_marshal(pdu
, offset
, "Qd", &qid
, iounit
);
2166 trace_v9fs_create_return(pdu
->tag
, pdu
->id
,
2167 qid
.type
, qid
.version
, qid
.path
, iounit
);
2171 complete_pdu(pdu
->s
, pdu
, err
);
2172 v9fs_string_free(&name
);
2173 v9fs_string_free(&extension
);
2174 v9fs_path_free(&path
);
2177 static void v9fs_symlink(void *opaque
)
2179 V9fsPDU
*pdu
= opaque
;
2182 V9fsFidState
*dfidp
;
2190 v9fs_string_init(&name
);
2191 v9fs_string_init(&symname
);
2192 err
= pdu_unmarshal(pdu
, offset
, "dssd", &dfid
, &name
, &symname
, &gid
);
2196 trace_v9fs_symlink(pdu
->tag
, pdu
->id
, dfid
, name
.data
, symname
.data
, gid
);
2198 dfidp
= get_fid(pdu
, dfid
);
2199 if (dfidp
== NULL
) {
2203 err
= v9fs_co_symlink(pdu
, dfidp
, &name
, symname
.data
, gid
, &stbuf
);
2207 stat_to_qid(&stbuf
, &qid
);
2208 err
= pdu_marshal(pdu
, offset
, "Q", &qid
);
2213 trace_v9fs_symlink_return(pdu
->tag
, pdu
->id
,
2214 qid
.type
, qid
.version
, qid
.path
);
2216 put_fid(pdu
, dfidp
);
2218 complete_pdu(pdu
->s
, pdu
, err
);
2219 v9fs_string_free(&name
);
2220 v9fs_string_free(&symname
);
2223 static void v9fs_flush(void *opaque
)
2228 V9fsPDU
*cancel_pdu
;
2229 V9fsPDU
*pdu
= opaque
;
2230 V9fsState
*s
= pdu
->s
;
2232 err
= pdu_unmarshal(pdu
, offset
, "w", &tag
);
2234 complete_pdu(s
, pdu
, err
);
2237 trace_v9fs_flush(pdu
->tag
, pdu
->id
, tag
);
2239 QLIST_FOREACH(cancel_pdu
, &s
->active_list
, next
) {
2240 if (cancel_pdu
->tag
== tag
) {
2245 cancel_pdu
->cancelled
= 1;
2247 * Wait for pdu to complete.
2249 qemu_co_queue_wait(&cancel_pdu
->complete
);
2250 cancel_pdu
->cancelled
= 0;
2251 free_pdu(pdu
->s
, cancel_pdu
);
2253 complete_pdu(s
, pdu
, 7);
2257 static void v9fs_link(void *opaque
)
2259 V9fsPDU
*pdu
= opaque
;
2260 V9fsState
*s
= pdu
->s
;
2261 int32_t dfid
, oldfid
;
2262 V9fsFidState
*dfidp
, *oldfidp
;
2267 v9fs_string_init(&name
);
2268 err
= pdu_unmarshal(pdu
, offset
, "dds", &dfid
, &oldfid
, &name
);
2272 trace_v9fs_link(pdu
->tag
, pdu
->id
, dfid
, oldfid
, name
.data
);
2274 dfidp
= get_fid(pdu
, dfid
);
2275 if (dfidp
== NULL
) {
2280 oldfidp
= get_fid(pdu
, oldfid
);
2281 if (oldfidp
== NULL
) {
2285 err
= v9fs_co_link(pdu
, oldfidp
, dfidp
, &name
);
2290 put_fid(pdu
, dfidp
);
2292 v9fs_string_free(&name
);
2293 complete_pdu(s
, pdu
, err
);
2296 /* Only works with path name based fid */
2297 static void v9fs_remove(void *opaque
)
2303 V9fsPDU
*pdu
= opaque
;
2305 err
= pdu_unmarshal(pdu
, offset
, "d", &fid
);
2309 trace_v9fs_remove(pdu
->tag
, pdu
->id
, fid
);
2311 fidp
= get_fid(pdu
, fid
);
2316 /* if fs driver is not path based, return EOPNOTSUPP */
2317 if (!(pdu
->s
->ctx
.export_flags
& V9FS_PATHNAME_FSCONTEXT
)) {
2322 * IF the file is unlinked, we cannot reopen
2323 * the file later. So don't reclaim fd
2325 err
= v9fs_mark_fids_unreclaim(pdu
, &fidp
->path
);
2329 err
= v9fs_co_remove(pdu
, &fidp
->path
);
2334 /* For TREMOVE we need to clunk the fid even on failed remove */
2335 clunk_fid(pdu
->s
, fidp
->fid
);
2338 complete_pdu(pdu
->s
, pdu
, err
);
2341 static void v9fs_unlinkat(void *opaque
)
2345 int32_t dfid
, flags
;
2348 V9fsFidState
*dfidp
;
2349 V9fsPDU
*pdu
= opaque
;
2351 v9fs_string_init(&name
);
2352 err
= pdu_unmarshal(pdu
, offset
, "dsd", &dfid
, &name
, &flags
);
2356 dfidp
= get_fid(pdu
, dfid
);
2357 if (dfidp
== NULL
) {
2362 * IF the file is unlinked, we cannot reopen
2363 * the file later. So don't reclaim fd
2365 v9fs_path_init(&path
);
2366 err
= v9fs_co_name_to_path(pdu
, &dfidp
->path
, name
.data
, &path
);
2370 err
= v9fs_mark_fids_unreclaim(pdu
, &path
);
2374 err
= v9fs_co_unlinkat(pdu
, &dfidp
->path
, &name
, flags
);
2379 put_fid(pdu
, dfidp
);
2380 v9fs_path_free(&path
);
2382 complete_pdu(pdu
->s
, pdu
, err
);
2383 v9fs_string_free(&name
);
2387 /* Only works with path name based fid */
2388 static int v9fs_complete_rename(V9fsPDU
*pdu
, V9fsFidState
*fidp
,
2389 int32_t newdirfid
, V9fsString
*name
)
2394 V9fsFidState
*tfidp
;
2395 V9fsState
*s
= pdu
->s
;
2396 V9fsFidState
*dirfidp
= NULL
;
2397 char *old_name
, *new_name
;
2399 v9fs_path_init(&new_path
);
2400 if (newdirfid
!= -1) {
2401 dirfidp
= get_fid(pdu
, newdirfid
);
2402 if (dirfidp
== NULL
) {
2406 BUG_ON(dirfidp
->fid_type
!= P9_FID_NONE
);
2407 v9fs_co_name_to_path(pdu
, &dirfidp
->path
, name
->data
, &new_path
);
2409 old_name
= fidp
->path
.data
;
2410 end
= strrchr(old_name
, '/');
2416 new_name
= g_malloc0(end
- old_name
+ name
->size
+ 1);
2417 strncat(new_name
, old_name
, end
- old_name
);
2418 strncat(new_name
+ (end
- old_name
), name
->data
, name
->size
);
2419 v9fs_co_name_to_path(pdu
, NULL
, new_name
, &new_path
);
2422 err
= v9fs_co_rename(pdu
, &fidp
->path
, &new_path
);
2427 * Fixup fid's pointing to the old name to
2428 * start pointing to the new name
2430 for (tfidp
= s
->fid_list
; tfidp
; tfidp
= tfidp
->next
) {
2431 if (v9fs_path_is_ancestor(&fidp
->path
, &tfidp
->path
)) {
2432 /* replace the name */
2433 v9fs_fix_path(&tfidp
->path
, &new_path
, strlen(fidp
->path
.data
));
2438 put_fid(pdu
, dirfidp
);
2440 v9fs_path_free(&new_path
);
2445 /* Only works with path name based fid */
2446 static void v9fs_rename(void *opaque
)
2454 V9fsPDU
*pdu
= opaque
;
2455 V9fsState
*s
= pdu
->s
;
2457 v9fs_string_init(&name
);
2458 err
= pdu_unmarshal(pdu
, offset
, "dds", &fid
, &newdirfid
, &name
);
2462 fidp
= get_fid(pdu
, fid
);
2467 BUG_ON(fidp
->fid_type
!= P9_FID_NONE
);
2468 /* if fs driver is not path based, return EOPNOTSUPP */
2469 if (!(pdu
->s
->ctx
.export_flags
& V9FS_PATHNAME_FSCONTEXT
)) {
2473 v9fs_path_write_lock(s
);
2474 err
= v9fs_complete_rename(pdu
, fidp
, newdirfid
, &name
);
2475 v9fs_path_unlock(s
);
2482 complete_pdu(s
, pdu
, err
);
2483 v9fs_string_free(&name
);
2486 static void v9fs_fix_fid_paths(V9fsPDU
*pdu
, V9fsPath
*olddir
,
2487 V9fsString
*old_name
, V9fsPath
*newdir
,
2488 V9fsString
*new_name
)
2490 V9fsFidState
*tfidp
;
2491 V9fsPath oldpath
, newpath
;
2492 V9fsState
*s
= pdu
->s
;
2495 v9fs_path_init(&oldpath
);
2496 v9fs_path_init(&newpath
);
2497 v9fs_co_name_to_path(pdu
, olddir
, old_name
->data
, &oldpath
);
2498 v9fs_co_name_to_path(pdu
, newdir
, new_name
->data
, &newpath
);
2501 * Fixup fid's pointing to the old name to
2502 * start pointing to the new name
2504 for (tfidp
= s
->fid_list
; tfidp
; tfidp
= tfidp
->next
) {
2505 if (v9fs_path_is_ancestor(&oldpath
, &tfidp
->path
)) {
2506 /* replace the name */
2507 v9fs_fix_path(&tfidp
->path
, &newpath
, strlen(oldpath
.data
));
2510 v9fs_path_free(&oldpath
);
2511 v9fs_path_free(&newpath
);
2514 static int v9fs_complete_renameat(V9fsPDU
*pdu
, int32_t olddirfid
,
2515 V9fsString
*old_name
, int32_t newdirfid
,
2516 V9fsString
*new_name
)
2519 V9fsState
*s
= pdu
->s
;
2520 V9fsFidState
*newdirfidp
= NULL
, *olddirfidp
= NULL
;
2522 olddirfidp
= get_fid(pdu
, olddirfid
);
2523 if (olddirfidp
== NULL
) {
2527 if (newdirfid
!= -1) {
2528 newdirfidp
= get_fid(pdu
, newdirfid
);
2529 if (newdirfidp
== NULL
) {
2534 newdirfidp
= get_fid(pdu
, olddirfid
);
2537 err
= v9fs_co_renameat(pdu
, &olddirfidp
->path
, old_name
,
2538 &newdirfidp
->path
, new_name
);
2542 if (s
->ctx
.export_flags
& V9FS_PATHNAME_FSCONTEXT
) {
2543 /* Only for path based fid we need to do the below fixup */
2544 v9fs_fix_fid_paths(pdu
, &olddirfidp
->path
, old_name
,
2545 &newdirfidp
->path
, new_name
);
2549 put_fid(pdu
, olddirfidp
);
2552 put_fid(pdu
, newdirfidp
);
2557 static void v9fs_renameat(void *opaque
)
2561 V9fsPDU
*pdu
= opaque
;
2562 V9fsState
*s
= pdu
->s
;
2563 int32_t olddirfid
, newdirfid
;
2564 V9fsString old_name
, new_name
;
2566 v9fs_string_init(&old_name
);
2567 v9fs_string_init(&new_name
);
2568 err
= pdu_unmarshal(pdu
, offset
, "dsds", &olddirfid
,
2569 &old_name
, &newdirfid
, &new_name
);
2574 v9fs_path_write_lock(s
);
2575 err
= v9fs_complete_renameat(pdu
, olddirfid
,
2576 &old_name
, newdirfid
, &new_name
);
2577 v9fs_path_unlock(s
);
2583 complete_pdu(s
, pdu
, err
);
2584 v9fs_string_free(&old_name
);
2585 v9fs_string_free(&new_name
);
2588 static void v9fs_wstat(void *opaque
)
2597 V9fsPDU
*pdu
= opaque
;
2598 V9fsState
*s
= pdu
->s
;
2600 v9fs_stat_init(&v9stat
);
2601 err
= pdu_unmarshal(pdu
, offset
, "dwS", &fid
, &unused
, &v9stat
);
2605 trace_v9fs_wstat(pdu
->tag
, pdu
->id
, fid
,
2606 v9stat
.mode
, v9stat
.atime
, v9stat
.mtime
);
2608 fidp
= get_fid(pdu
, fid
);
2613 /* do we need to sync the file? */
2614 if (donttouch_stat(&v9stat
)) {
2615 err
= v9fs_co_fsync(pdu
, fidp
, 0);
2618 if (v9stat
.mode
!= -1) {
2620 err
= v9fs_co_lstat(pdu
, &fidp
->path
, &stbuf
);
2624 v9_mode
= stat_to_v9mode(&stbuf
);
2625 if ((v9stat
.mode
& P9_STAT_MODE_TYPE_BITS
) !=
2626 (v9_mode
& P9_STAT_MODE_TYPE_BITS
)) {
2627 /* Attempting to change the type */
2631 err
= v9fs_co_chmod(pdu
, &fidp
->path
,
2632 v9mode_to_mode(v9stat
.mode
,
2633 &v9stat
.extension
));
2638 if (v9stat
.mtime
!= -1 || v9stat
.atime
!= -1) {
2639 struct timespec times
[2];
2640 if (v9stat
.atime
!= -1) {
2641 times
[0].tv_sec
= v9stat
.atime
;
2642 times
[0].tv_nsec
= 0;
2644 times
[0].tv_nsec
= UTIME_OMIT
;
2646 if (v9stat
.mtime
!= -1) {
2647 times
[1].tv_sec
= v9stat
.mtime
;
2648 times
[1].tv_nsec
= 0;
2650 times
[1].tv_nsec
= UTIME_OMIT
;
2652 err
= v9fs_co_utimensat(pdu
, &fidp
->path
, times
);
2657 if (v9stat
.n_gid
!= -1 || v9stat
.n_uid
!= -1) {
2658 err
= v9fs_co_chown(pdu
, &fidp
->path
, v9stat
.n_uid
, v9stat
.n_gid
);
2663 if (v9stat
.name
.size
!= 0) {
2664 err
= v9fs_complete_rename(pdu
, fidp
, -1, &v9stat
.name
);
2669 if (v9stat
.length
!= -1) {
2670 err
= v9fs_co_truncate(pdu
, &fidp
->path
, v9stat
.length
);
2679 v9fs_stat_free(&v9stat
);
2680 complete_pdu(s
, pdu
, err
);
2683 static int v9fs_fill_statfs(V9fsState
*s
, V9fsPDU
*pdu
, struct statfs
*stbuf
)
2695 int32_t bsize_factor
;
2698 * compute bsize factor based on host file system block size
2701 bsize_factor
= (s
->msize
- P9_IOHDRSZ
)/stbuf
->f_bsize
;
2702 if (!bsize_factor
) {
2705 f_type
= stbuf
->f_type
;
2706 f_bsize
= stbuf
->f_bsize
;
2707 f_bsize
*= bsize_factor
;
2709 * f_bsize is adjusted(multiplied) by bsize factor, so we need to
2710 * adjust(divide) the number of blocks, free blocks and available
2711 * blocks by bsize factor
2713 f_blocks
= stbuf
->f_blocks
/bsize_factor
;
2714 f_bfree
= stbuf
->f_bfree
/bsize_factor
;
2715 f_bavail
= stbuf
->f_bavail
/bsize_factor
;
2716 f_files
= stbuf
->f_files
;
2717 f_ffree
= stbuf
->f_ffree
;
2718 fsid_val
= (unsigned int) stbuf
->f_fsid
.__val
[0] |
2719 (unsigned long long)stbuf
->f_fsid
.__val
[1] << 32;
2720 f_namelen
= stbuf
->f_namelen
;
2722 return pdu_marshal(pdu
, offset
, "ddqqqqqqd",
2723 f_type
, f_bsize
, f_blocks
, f_bfree
,
2724 f_bavail
, f_files
, f_ffree
,
2725 fsid_val
, f_namelen
);
2728 static void v9fs_statfs(void *opaque
)
2734 struct statfs stbuf
;
2735 V9fsPDU
*pdu
= opaque
;
2736 V9fsState
*s
= pdu
->s
;
2738 retval
= pdu_unmarshal(pdu
, offset
, "d", &fid
);
2742 fidp
= get_fid(pdu
, fid
);
2747 retval
= v9fs_co_statfs(pdu
, &fidp
->path
, &stbuf
);
2751 retval
= v9fs_fill_statfs(s
, pdu
, &stbuf
);
2759 complete_pdu(s
, pdu
, retval
);
2763 static void v9fs_mknod(void *opaque
)
2776 V9fsPDU
*pdu
= opaque
;
2777 V9fsState
*s
= pdu
->s
;
2779 v9fs_string_init(&name
);
2780 err
= pdu_unmarshal(pdu
, offset
, "dsdddd", &fid
, &name
, &mode
,
2781 &major
, &minor
, &gid
);
2785 trace_v9fs_mknod(pdu
->tag
, pdu
->id
, fid
, mode
, major
, minor
);
2787 fidp
= get_fid(pdu
, fid
);
2792 err
= v9fs_co_mknod(pdu
, fidp
, &name
, fidp
->uid
, gid
,
2793 makedev(major
, minor
), mode
, &stbuf
);
2797 stat_to_qid(&stbuf
, &qid
);
2798 err
= pdu_marshal(pdu
, offset
, "Q", &qid
);
2803 trace_v9fs_mknod_return(pdu
->tag
, pdu
->id
,
2804 qid
.type
, qid
.version
, qid
.path
);
2808 complete_pdu(s
, pdu
, err
);
2809 v9fs_string_free(&name
);
2813 * Implement posix byte range locking code
2814 * Server side handling of locking code is very simple, because 9p server in
2815 * QEMU can handle only one client. And most of the lock handling
2816 * (like conflict, merging) etc is done by the VFS layer itself, so no need to
2817 * do any thing in * qemu 9p server side lock code path.
2818 * So when a TLOCK request comes, always return success
2820 static void v9fs_lock(void *opaque
)
2827 int32_t fid
, err
= 0;
2828 V9fsPDU
*pdu
= opaque
;
2829 V9fsState
*s
= pdu
->s
;
2831 status
= P9_LOCK_ERROR
;
2832 v9fs_string_init(&flock
.client_id
);
2833 err
= pdu_unmarshal(pdu
, offset
, "dbdqqds", &fid
, &flock
.type
,
2834 &flock
.flags
, &flock
.start
, &flock
.length
,
2835 &flock
.proc_id
, &flock
.client_id
);
2839 trace_v9fs_lock(pdu
->tag
, pdu
->id
, fid
,
2840 flock
.type
, flock
.start
, flock
.length
);
2843 /* We support only block flag now (that too ignored currently) */
2844 if (flock
.flags
& ~P9_LOCK_FLAGS_BLOCK
) {
2848 fidp
= get_fid(pdu
, fid
);
2853 err
= v9fs_co_fstat(pdu
, fidp
, &stbuf
);
2857 status
= P9_LOCK_SUCCESS
;
2861 err
= pdu_marshal(pdu
, offset
, "b", status
);
2865 trace_v9fs_lock_return(pdu
->tag
, pdu
->id
, status
);
2866 complete_pdu(s
, pdu
, err
);
2867 v9fs_string_free(&flock
.client_id
);
2871 * When a TGETLOCK request comes, always return success because all lock
2872 * handling is done by client's VFS layer.
2874 static void v9fs_getlock(void *opaque
)
2880 int32_t fid
, err
= 0;
2881 V9fsPDU
*pdu
= opaque
;
2882 V9fsState
*s
= pdu
->s
;
2884 v9fs_string_init(&glock
.client_id
);
2885 err
= pdu_unmarshal(pdu
, offset
, "dbqqds", &fid
, &glock
.type
,
2886 &glock
.start
, &glock
.length
, &glock
.proc_id
,
2891 trace_v9fs_getlock(pdu
->tag
, pdu
->id
, fid
,
2892 glock
.type
, glock
.start
, glock
.length
);
2894 fidp
= get_fid(pdu
, fid
);
2899 err
= v9fs_co_fstat(pdu
, fidp
, &stbuf
);
2903 glock
.type
= P9_LOCK_TYPE_UNLCK
;
2904 err
= pdu_marshal(pdu
, offset
, "bqqds", glock
.type
,
2905 glock
.start
, glock
.length
, glock
.proc_id
,
2911 trace_v9fs_getlock_return(pdu
->tag
, pdu
->id
, glock
.type
, glock
.start
,
2912 glock
.length
, glock
.proc_id
);
2916 complete_pdu(s
, pdu
, err
);
2917 v9fs_string_free(&glock
.client_id
);
2920 static void v9fs_mkdir(void *opaque
)
2922 V9fsPDU
*pdu
= opaque
;
2933 v9fs_string_init(&name
);
2934 err
= pdu_unmarshal(pdu
, offset
, "dsdd", &fid
, &name
, &mode
, &gid
);
2938 trace_v9fs_mkdir(pdu
->tag
, pdu
->id
, fid
, name
.data
, mode
, gid
);
2940 fidp
= get_fid(pdu
, fid
);
2945 err
= v9fs_co_mkdir(pdu
, fidp
, &name
, mode
, fidp
->uid
, gid
, &stbuf
);
2949 stat_to_qid(&stbuf
, &qid
);
2950 err
= pdu_marshal(pdu
, offset
, "Q", &qid
);
2955 trace_v9fs_mkdir_return(pdu
->tag
, pdu
->id
,
2956 qid
.type
, qid
.version
, qid
.path
, err
);
2960 complete_pdu(pdu
->s
, pdu
, err
);
2961 v9fs_string_free(&name
);
2964 static void v9fs_xattrwalk(void *opaque
)
2970 int32_t fid
, newfid
;
2971 V9fsFidState
*file_fidp
;
2972 V9fsFidState
*xattr_fidp
= NULL
;
2973 V9fsPDU
*pdu
= opaque
;
2974 V9fsState
*s
= pdu
->s
;
2976 v9fs_string_init(&name
);
2977 err
= pdu_unmarshal(pdu
, offset
, "dds", &fid
, &newfid
, &name
);
2981 trace_v9fs_xattrwalk(pdu
->tag
, pdu
->id
, fid
, newfid
, name
.data
);
2983 file_fidp
= get_fid(pdu
, fid
);
2984 if (file_fidp
== NULL
) {
2988 xattr_fidp
= alloc_fid(s
, newfid
);
2989 if (xattr_fidp
== NULL
) {
2993 v9fs_path_copy(&xattr_fidp
->path
, &file_fidp
->path
);
2994 if (name
.data
== NULL
) {
2996 * listxattr request. Get the size first
2998 size
= v9fs_co_llistxattr(pdu
, &xattr_fidp
->path
, NULL
, 0);
3001 clunk_fid(s
, xattr_fidp
->fid
);
3005 * Read the xattr value
3007 xattr_fidp
->fs
.xattr
.len
= size
;
3008 xattr_fidp
->fid_type
= P9_FID_XATTR
;
3009 xattr_fidp
->fs
.xattr
.copied_len
= -1;
3011 xattr_fidp
->fs
.xattr
.value
= g_malloc(size
);
3012 err
= v9fs_co_llistxattr(pdu
, &xattr_fidp
->path
,
3013 xattr_fidp
->fs
.xattr
.value
,
3014 xattr_fidp
->fs
.xattr
.len
);
3016 clunk_fid(s
, xattr_fidp
->fid
);
3020 err
= pdu_marshal(pdu
, offset
, "q", size
);
3027 * specific xattr fid. We check for xattr
3028 * presence also collect the xattr size
3030 size
= v9fs_co_lgetxattr(pdu
, &xattr_fidp
->path
,
3034 clunk_fid(s
, xattr_fidp
->fid
);
3038 * Read the xattr value
3040 xattr_fidp
->fs
.xattr
.len
= size
;
3041 xattr_fidp
->fid_type
= P9_FID_XATTR
;
3042 xattr_fidp
->fs
.xattr
.copied_len
= -1;
3044 xattr_fidp
->fs
.xattr
.value
= g_malloc(size
);
3045 err
= v9fs_co_lgetxattr(pdu
, &xattr_fidp
->path
,
3046 &name
, xattr_fidp
->fs
.xattr
.value
,
3047 xattr_fidp
->fs
.xattr
.len
);
3049 clunk_fid(s
, xattr_fidp
->fid
);
3053 err
= pdu_marshal(pdu
, offset
, "q", size
);
3059 trace_v9fs_xattrwalk_return(pdu
->tag
, pdu
->id
, size
);
3061 put_fid(pdu
, file_fidp
);
3063 put_fid(pdu
, xattr_fidp
);
3066 complete_pdu(s
, pdu
, err
);
3067 v9fs_string_free(&name
);
3070 static void v9fs_xattrcreate(void *opaque
)
3078 V9fsFidState
*file_fidp
;
3079 V9fsFidState
*xattr_fidp
;
3080 V9fsPDU
*pdu
= opaque
;
3081 V9fsState
*s
= pdu
->s
;
3083 v9fs_string_init(&name
);
3084 err
= pdu_unmarshal(pdu
, offset
, "dsqd", &fid
, &name
, &size
, &flags
);
3088 trace_v9fs_xattrcreate(pdu
->tag
, pdu
->id
, fid
, name
.data
, size
, flags
);
3090 file_fidp
= get_fid(pdu
, fid
);
3091 if (file_fidp
== NULL
) {
3095 /* Make the file fid point to xattr */
3096 xattr_fidp
= file_fidp
;
3097 xattr_fidp
->fid_type
= P9_FID_XATTR
;
3098 xattr_fidp
->fs
.xattr
.copied_len
= 0;
3099 xattr_fidp
->fs
.xattr
.len
= size
;
3100 xattr_fidp
->fs
.xattr
.flags
= flags
;
3101 v9fs_string_init(&xattr_fidp
->fs
.xattr
.name
);
3102 v9fs_string_copy(&xattr_fidp
->fs
.xattr
.name
, &name
);
3104 xattr_fidp
->fs
.xattr
.value
= g_malloc(size
);
3106 xattr_fidp
->fs
.xattr
.value
= NULL
;
3109 put_fid(pdu
, file_fidp
);
3111 complete_pdu(s
, pdu
, err
);
3112 v9fs_string_free(&name
);
3115 static void v9fs_readlink(void *opaque
)
3117 V9fsPDU
*pdu
= opaque
;
3124 err
= pdu_unmarshal(pdu
, offset
, "d", &fid
);
3128 trace_v9fs_readlink(pdu
->tag
, pdu
->id
, fid
);
3129 fidp
= get_fid(pdu
, fid
);
3135 v9fs_string_init(&target
);
3136 err
= v9fs_co_readlink(pdu
, &fidp
->path
, &target
);
3140 err
= pdu_marshal(pdu
, offset
, "s", &target
);
3142 v9fs_string_free(&target
);
3146 trace_v9fs_readlink_return(pdu
->tag
, pdu
->id
, target
.data
);
3147 v9fs_string_free(&target
);
3151 complete_pdu(pdu
->s
, pdu
, err
);
3154 static CoroutineEntry
*pdu_co_handlers
[] = {
3155 [P9_TREADDIR
] = v9fs_readdir
,
3156 [P9_TSTATFS
] = v9fs_statfs
,
3157 [P9_TGETATTR
] = v9fs_getattr
,
3158 [P9_TSETATTR
] = v9fs_setattr
,
3159 [P9_TXATTRWALK
] = v9fs_xattrwalk
,
3160 [P9_TXATTRCREATE
] = v9fs_xattrcreate
,
3161 [P9_TMKNOD
] = v9fs_mknod
,
3162 [P9_TRENAME
] = v9fs_rename
,
3163 [P9_TLOCK
] = v9fs_lock
,
3164 [P9_TGETLOCK
] = v9fs_getlock
,
3165 [P9_TRENAMEAT
] = v9fs_renameat
,
3166 [P9_TREADLINK
] = v9fs_readlink
,
3167 [P9_TUNLINKAT
] = v9fs_unlinkat
,
3168 [P9_TMKDIR
] = v9fs_mkdir
,
3169 [P9_TVERSION
] = v9fs_version
,
3170 [P9_TLOPEN
] = v9fs_open
,
3171 [P9_TATTACH
] = v9fs_attach
,
3172 [P9_TSTAT
] = v9fs_stat
,
3173 [P9_TWALK
] = v9fs_walk
,
3174 [P9_TCLUNK
] = v9fs_clunk
,
3175 [P9_TFSYNC
] = v9fs_fsync
,
3176 [P9_TOPEN
] = v9fs_open
,
3177 [P9_TREAD
] = v9fs_read
,
3179 [P9_TAUTH
] = v9fs_auth
,
3181 [P9_TFLUSH
] = v9fs_flush
,
3182 [P9_TLINK
] = v9fs_link
,
3183 [P9_TSYMLINK
] = v9fs_symlink
,
3184 [P9_TCREATE
] = v9fs_create
,
3185 [P9_TLCREATE
] = v9fs_lcreate
,
3186 [P9_TWRITE
] = v9fs_write
,
3187 [P9_TWSTAT
] = v9fs_wstat
,
3188 [P9_TREMOVE
] = v9fs_remove
,
3191 static void v9fs_op_not_supp(void *opaque
)
3193 V9fsPDU
*pdu
= opaque
;
3194 complete_pdu(pdu
->s
, pdu
, -EOPNOTSUPP
);
3197 static void v9fs_fs_ro(void *opaque
)
3199 V9fsPDU
*pdu
= opaque
;
3200 complete_pdu(pdu
->s
, pdu
, -EROFS
);
3203 static inline bool is_read_only_op(V9fsPDU
*pdu
)
3230 static void submit_pdu(V9fsState
*s
, V9fsPDU
*pdu
)
3233 CoroutineEntry
*handler
;
3235 if (pdu
->id
>= ARRAY_SIZE(pdu_co_handlers
) ||
3236 (pdu_co_handlers
[pdu
->id
] == NULL
)) {
3237 handler
= v9fs_op_not_supp
;
3239 handler
= pdu_co_handlers
[pdu
->id
];
3242 if (is_ro_export(&s
->ctx
) && !is_read_only_op(pdu
)) {
3243 handler
= v9fs_fs_ro
;
3245 co
= qemu_coroutine_create(handler
);
3246 qemu_coroutine_enter(co
, pdu
);
3249 void handle_9p_output(VirtIODevice
*vdev
, VirtQueue
*vq
)
3251 V9fsState
*s
= (V9fsState
*)vdev
;
3255 while ((pdu
= alloc_pdu(s
)) &&
3256 (len
= virtqueue_pop(vq
, &pdu
->elem
)) != 0) {
3259 BUG_ON(pdu
->elem
.out_num
== 0 || pdu
->elem
.in_num
== 0);
3260 BUG_ON(pdu
->elem
.out_sg
[0].iov_len
< 7);
3262 ptr
= pdu
->elem
.out_sg
[0].iov_base
;
3264 memcpy(&pdu
->size
, ptr
, 4);
3266 memcpy(&pdu
->tag
, ptr
+ 5, 2);
3267 qemu_co_queue_init(&pdu
->complete
);
3273 void virtio_9p_set_fd_limit(void)
3276 if (getrlimit(RLIMIT_NOFILE
, &rlim
) < 0) {
3277 fprintf(stderr
, "Failed to get the resource limit\n");
3280 open_fd_hw
= rlim
.rlim_cur
- MIN(400, rlim
.rlim_cur
/3);
3281 open_fd_rc
= rlim
.rlim_cur
/2;