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 "qemu/osdep.h"
15 #include "hw/virtio/virtio.h"
16 #include "hw/i386/pc.h"
17 #include "qapi/error.h"
18 #include "qemu/error-report.h"
20 #include "qemu/sockets.h"
21 #include "virtio-9p.h"
22 #include "fsdev/qemu-fsdev.h"
26 #include "migration/migration.h"
30 static int open_fd_rc
;
44 ssize_t
pdu_marshal(V9fsPDU
*pdu
, size_t offset
, const char *fmt
, ...)
50 ret
= virtio_pdu_vmarshal(pdu
, offset
, fmt
, ap
);
56 ssize_t
pdu_unmarshal(V9fsPDU
*pdu
, size_t offset
, const char *fmt
, ...)
62 ret
= virtio_pdu_vunmarshal(pdu
, offset
, fmt
, ap
);
68 static void pdu_push_and_notify(V9fsPDU
*pdu
)
70 virtio_9p_push_and_notify(pdu
);
73 static int omode_to_uflags(int8_t mode
)
107 struct dotl_openflag_map
{
112 static int dotl_to_open_flags(int flags
)
116 * We have same bits for P9_DOTL_READONLY, P9_DOTL_WRONLY
117 * and P9_DOTL_NOACCESS
119 int oflags
= flags
& O_ACCMODE
;
121 struct dotl_openflag_map dotl_oflag_map
[] = {
122 { P9_DOTL_CREATE
, O_CREAT
},
123 { P9_DOTL_EXCL
, O_EXCL
},
124 { P9_DOTL_NOCTTY
, O_NOCTTY
},
125 { P9_DOTL_TRUNC
, O_TRUNC
},
126 { P9_DOTL_APPEND
, O_APPEND
},
127 { P9_DOTL_NONBLOCK
, O_NONBLOCK
} ,
128 { P9_DOTL_DSYNC
, O_DSYNC
},
129 { P9_DOTL_FASYNC
, FASYNC
},
130 { P9_DOTL_DIRECT
, O_DIRECT
},
131 { P9_DOTL_LARGEFILE
, O_LARGEFILE
},
132 { P9_DOTL_DIRECTORY
, O_DIRECTORY
},
133 { P9_DOTL_NOFOLLOW
, O_NOFOLLOW
},
134 { P9_DOTL_NOATIME
, O_NOATIME
},
135 { P9_DOTL_SYNC
, O_SYNC
},
138 for (i
= 0; i
< ARRAY_SIZE(dotl_oflag_map
); i
++) {
139 if (flags
& dotl_oflag_map
[i
].dotl_flag
) {
140 oflags
|= dotl_oflag_map
[i
].open_flag
;
147 void cred_init(FsCred
*credp
)
155 static int get_dotl_openflags(V9fsState
*s
, int oflags
)
159 * Filter the client open flags
161 flags
= dotl_to_open_flags(oflags
);
162 flags
&= ~(O_NOCTTY
| O_ASYNC
| O_CREAT
);
164 * Ignore direct disk access hint until the server supports it.
170 void v9fs_path_init(V9fsPath
*path
)
176 void v9fs_path_free(V9fsPath
*path
)
183 void v9fs_path_copy(V9fsPath
*lhs
, V9fsPath
*rhs
)
186 lhs
->data
= g_malloc(rhs
->size
);
187 memcpy(lhs
->data
, rhs
->data
, rhs
->size
);
188 lhs
->size
= rhs
->size
;
191 int v9fs_name_to_path(V9fsState
*s
, V9fsPath
*dirpath
,
192 const char *name
, V9fsPath
*path
)
195 err
= s
->ops
->name_to_path(&s
->ctx
, dirpath
, name
, path
);
203 * Return TRUE if s1 is an ancestor of s2.
205 * E.g. "a/b" is an ancestor of "a/b/c" but not of "a/bc/d".
206 * As a special case, We treat s1 as ancestor of s2 if they are same!
208 static int v9fs_path_is_ancestor(V9fsPath
*s1
, V9fsPath
*s2
)
210 if (!strncmp(s1
->data
, s2
->data
, s1
->size
- 1)) {
211 if (s2
->data
[s1
->size
- 1] == '\0' || s2
->data
[s1
->size
- 1] == '/') {
218 static size_t v9fs_string_size(V9fsString
*str
)
224 * returns 0 if fid got re-opened, 1 if not, < 0 on error */
225 static int v9fs_reopen_fid(V9fsPDU
*pdu
, V9fsFidState
*f
)
228 if (f
->fid_type
== P9_FID_FILE
) {
229 if (f
->fs
.fd
== -1) {
231 err
= v9fs_co_open(pdu
, f
, f
->open_flags
);
232 } while (err
== -EINTR
&& !pdu
->cancelled
);
234 } else if (f
->fid_type
== P9_FID_DIR
) {
235 if (f
->fs
.dir
== NULL
) {
237 err
= v9fs_co_opendir(pdu
, f
);
238 } while (err
== -EINTR
&& !pdu
->cancelled
);
244 static V9fsFidState
*get_fid(V9fsPDU
*pdu
, int32_t fid
)
248 V9fsState
*s
= pdu
->s
;
250 for (f
= s
->fid_list
; f
; f
= f
->next
) {
254 * Update the fid ref upfront so that
255 * we don't get reclaimed when we yield
260 * check whether we need to reopen the
261 * file. We might have closed the fd
262 * while trying to free up some file
265 err
= v9fs_reopen_fid(pdu
, f
);
271 * Mark the fid as referenced so that the LRU
272 * reclaim won't close the file descriptor
274 f
->flags
|= FID_REFERENCED
;
281 static V9fsFidState
*alloc_fid(V9fsState
*s
, int32_t fid
)
285 for (f
= s
->fid_list
; f
; f
= f
->next
) {
286 /* If fid is already there return NULL */
292 f
= g_malloc0(sizeof(V9fsFidState
));
294 f
->fid_type
= P9_FID_NONE
;
297 * Mark the fid as referenced so that the LRU
298 * reclaim won't close the file descriptor
300 f
->flags
|= FID_REFERENCED
;
301 f
->next
= s
->fid_list
;
307 static int v9fs_xattr_fid_clunk(V9fsPDU
*pdu
, V9fsFidState
*fidp
)
311 if (fidp
->fs
.xattr
.copied_len
== -1) {
312 /* getxattr/listxattr fid */
316 * if this is fid for setxattr. clunk should
317 * result in setxattr localcall
319 if (fidp
->fs
.xattr
.len
!= fidp
->fs
.xattr
.copied_len
) {
320 /* clunk after partial write */
324 if (fidp
->fs
.xattr
.len
) {
325 retval
= v9fs_co_lsetxattr(pdu
, &fidp
->path
, &fidp
->fs
.xattr
.name
,
326 fidp
->fs
.xattr
.value
,
328 fidp
->fs
.xattr
.flags
);
330 retval
= v9fs_co_lremovexattr(pdu
, &fidp
->path
, &fidp
->fs
.xattr
.name
);
333 v9fs_string_free(&fidp
->fs
.xattr
.name
);
335 g_free(fidp
->fs
.xattr
.value
);
339 static int free_fid(V9fsPDU
*pdu
, V9fsFidState
*fidp
)
343 if (fidp
->fid_type
== P9_FID_FILE
) {
344 /* If we reclaimed the fd no need to close */
345 if (fidp
->fs
.fd
!= -1) {
346 retval
= v9fs_co_close(pdu
, &fidp
->fs
);
348 } else if (fidp
->fid_type
== P9_FID_DIR
) {
349 if (fidp
->fs
.dir
!= NULL
) {
350 retval
= v9fs_co_closedir(pdu
, &fidp
->fs
);
352 } else if (fidp
->fid_type
== P9_FID_XATTR
) {
353 retval
= v9fs_xattr_fid_clunk(pdu
, fidp
);
355 v9fs_path_free(&fidp
->path
);
360 static int put_fid(V9fsPDU
*pdu
, V9fsFidState
*fidp
)
365 * Don't free the fid if it is in reclaim list
367 if (!fidp
->ref
&& fidp
->clunked
) {
368 if (fidp
->fid
== pdu
->s
->root_fid
) {
370 * if the clunked fid is root fid then we
371 * have unmounted the fs on the client side.
372 * delete the migration blocker. Ideally, this
373 * should be hooked to transport close notification
375 if (pdu
->s
->migration_blocker
) {
376 migrate_del_blocker(pdu
->s
->migration_blocker
);
377 error_free(pdu
->s
->migration_blocker
);
378 pdu
->s
->migration_blocker
= NULL
;
381 return free_fid(pdu
, fidp
);
386 static V9fsFidState
*clunk_fid(V9fsState
*s
, int32_t fid
)
388 V9fsFidState
**fidpp
, *fidp
;
390 for (fidpp
= &s
->fid_list
; *fidpp
; fidpp
= &(*fidpp
)->next
) {
391 if ((*fidpp
)->fid
== fid
) {
395 if (*fidpp
== NULL
) {
404 void v9fs_reclaim_fd(V9fsPDU
*pdu
)
406 int reclaim_count
= 0;
407 V9fsState
*s
= pdu
->s
;
408 V9fsFidState
*f
, *reclaim_list
= NULL
;
410 for (f
= s
->fid_list
; f
; f
= f
->next
) {
412 * Unlink fids cannot be reclaimed. Check
413 * for them and skip them. Also skip fids
414 * currently being operated on.
416 if (f
->ref
|| f
->flags
& FID_NON_RECLAIMABLE
) {
420 * if it is a recently referenced fid
421 * we leave the fid untouched and clear the
422 * reference bit. We come back to it later
423 * in the next iteration. (a simple LRU without
424 * moving list elements around)
426 if (f
->flags
& FID_REFERENCED
) {
427 f
->flags
&= ~FID_REFERENCED
;
431 * Add fids to reclaim list.
433 if (f
->fid_type
== P9_FID_FILE
) {
434 if (f
->fs
.fd
!= -1) {
436 * Up the reference count so that
437 * a clunk request won't free this fid
440 f
->rclm_lst
= reclaim_list
;
442 f
->fs_reclaim
.fd
= f
->fs
.fd
;
446 } else if (f
->fid_type
== P9_FID_DIR
) {
447 if (f
->fs
.dir
!= NULL
) {
449 * Up the reference count so that
450 * a clunk request won't free this fid
453 f
->rclm_lst
= reclaim_list
;
455 f
->fs_reclaim
.dir
= f
->fs
.dir
;
460 if (reclaim_count
>= open_fd_rc
) {
465 * Now close the fid in reclaim list. Free them if they
466 * are already clunked.
468 while (reclaim_list
) {
470 reclaim_list
= f
->rclm_lst
;
471 if (f
->fid_type
== P9_FID_FILE
) {
472 v9fs_co_close(pdu
, &f
->fs_reclaim
);
473 } else if (f
->fid_type
== P9_FID_DIR
) {
474 v9fs_co_closedir(pdu
, &f
->fs_reclaim
);
478 * Now drop the fid reference, free it
485 static int v9fs_mark_fids_unreclaim(V9fsPDU
*pdu
, V9fsPath
*path
)
488 V9fsState
*s
= pdu
->s
;
489 V9fsFidState
*fidp
, head_fid
;
491 head_fid
.next
= s
->fid_list
;
492 for (fidp
= s
->fid_list
; fidp
; fidp
= fidp
->next
) {
493 if (fidp
->path
.size
!= path
->size
) {
496 if (!memcmp(fidp
->path
.data
, path
->data
, path
->size
)) {
497 /* Mark the fid non reclaimable. */
498 fidp
->flags
|= FID_NON_RECLAIMABLE
;
500 /* reopen the file/dir if already closed */
501 err
= v9fs_reopen_fid(pdu
, fidp
);
506 * Go back to head of fid list because
507 * the list could have got updated when
508 * switched to the worker thread
518 static void virtfs_reset(V9fsPDU
*pdu
)
520 V9fsState
*s
= pdu
->s
;
521 V9fsFidState
*fidp
= NULL
;
524 while (s
->fid_list
) {
526 s
->fid_list
= fidp
->next
;
535 /* One or more unclunked fids found... */
536 error_report("9pfs:%s: One or more uncluncked fids "
537 "found during reset", __func__
);
541 #define P9_QID_TYPE_DIR 0x80
542 #define P9_QID_TYPE_SYMLINK 0x02
544 #define P9_STAT_MODE_DIR 0x80000000
545 #define P9_STAT_MODE_APPEND 0x40000000
546 #define P9_STAT_MODE_EXCL 0x20000000
547 #define P9_STAT_MODE_MOUNT 0x10000000
548 #define P9_STAT_MODE_AUTH 0x08000000
549 #define P9_STAT_MODE_TMP 0x04000000
550 #define P9_STAT_MODE_SYMLINK 0x02000000
551 #define P9_STAT_MODE_LINK 0x01000000
552 #define P9_STAT_MODE_DEVICE 0x00800000
553 #define P9_STAT_MODE_NAMED_PIPE 0x00200000
554 #define P9_STAT_MODE_SOCKET 0x00100000
555 #define P9_STAT_MODE_SETUID 0x00080000
556 #define P9_STAT_MODE_SETGID 0x00040000
557 #define P9_STAT_MODE_SETVTX 0x00010000
559 #define P9_STAT_MODE_TYPE_BITS (P9_STAT_MODE_DIR | \
560 P9_STAT_MODE_SYMLINK | \
561 P9_STAT_MODE_LINK | \
562 P9_STAT_MODE_DEVICE | \
563 P9_STAT_MODE_NAMED_PIPE | \
566 /* This is the algorithm from ufs in spfs */
567 static void stat_to_qid(const struct stat
*stbuf
, V9fsQID
*qidp
)
571 memset(&qidp
->path
, 0, sizeof(qidp
->path
));
572 size
= MIN(sizeof(stbuf
->st_ino
), sizeof(qidp
->path
));
573 memcpy(&qidp
->path
, &stbuf
->st_ino
, size
);
574 qidp
->version
= stbuf
->st_mtime
^ (stbuf
->st_size
<< 8);
576 if (S_ISDIR(stbuf
->st_mode
)) {
577 qidp
->type
|= P9_QID_TYPE_DIR
;
579 if (S_ISLNK(stbuf
->st_mode
)) {
580 qidp
->type
|= P9_QID_TYPE_SYMLINK
;
584 static int fid_to_qid(V9fsPDU
*pdu
, V9fsFidState
*fidp
, V9fsQID
*qidp
)
589 err
= v9fs_co_lstat(pdu
, &fidp
->path
, &stbuf
);
593 stat_to_qid(&stbuf
, qidp
);
597 V9fsPDU
*pdu_alloc(V9fsState
*s
)
601 if (!QLIST_EMPTY(&s
->free_list
)) {
602 pdu
= QLIST_FIRST(&s
->free_list
);
603 QLIST_REMOVE(pdu
, next
);
604 QLIST_INSERT_HEAD(&s
->active_list
, pdu
, next
);
609 void pdu_free(V9fsPDU
*pdu
)
612 V9fsState
*s
= pdu
->s
;
614 * Cancelled pdu are added back to the freelist
617 if (!pdu
->cancelled
) {
618 QLIST_REMOVE(pdu
, next
);
619 QLIST_INSERT_HEAD(&s
->free_list
, pdu
, next
);
625 * We don't do error checking for pdu_marshal/unmarshal here
626 * because we always expect to have enough space to encode
629 static void pdu_complete(V9fsPDU
*pdu
, ssize_t len
)
631 int8_t id
= pdu
->id
+ 1; /* Response */
632 V9fsState
*s
= pdu
->s
;
638 if (s
->proto_version
!= V9FS_PROTO_2000L
) {
641 str
.data
= strerror(err
);
642 str
.size
= strlen(str
.data
);
644 len
+= pdu_marshal(pdu
, len
, "s", &str
);
648 len
+= pdu_marshal(pdu
, len
, "d", err
);
650 if (s
->proto_version
== V9FS_PROTO_2000L
) {
653 trace_v9fs_rerror(pdu
->tag
, pdu
->id
, err
); /* Trace ERROR */
656 /* fill out the header */
657 pdu_marshal(pdu
, 0, "dbw", (int32_t)len
, id
, pdu
->tag
);
659 /* keep these in sync */
663 pdu_push_and_notify(pdu
);
665 /* Now wakeup anybody waiting in flush for this request */
666 qemu_co_queue_next(&pdu
->complete
);
671 static mode_t
v9mode_to_mode(uint32_t mode
, V9fsString
*extension
)
676 if (mode
& P9_STAT_MODE_DIR
) {
680 if (mode
& P9_STAT_MODE_SYMLINK
) {
683 if (mode
& P9_STAT_MODE_SOCKET
) {
686 if (mode
& P9_STAT_MODE_NAMED_PIPE
) {
689 if (mode
& P9_STAT_MODE_DEVICE
) {
690 if (extension
->size
&& extension
->data
[0] == 'c') {
701 if (mode
& P9_STAT_MODE_SETUID
) {
704 if (mode
& P9_STAT_MODE_SETGID
) {
707 if (mode
& P9_STAT_MODE_SETVTX
) {
714 static int donttouch_stat(V9fsStat
*stat
)
716 if (stat
->type
== -1 &&
718 stat
->qid
.type
== -1 &&
719 stat
->qid
.version
== -1 &&
720 stat
->qid
.path
== -1 &&
724 stat
->length
== -1 &&
731 stat
->n_muid
== -1) {
738 static void v9fs_stat_init(V9fsStat
*stat
)
740 v9fs_string_init(&stat
->name
);
741 v9fs_string_init(&stat
->uid
);
742 v9fs_string_init(&stat
->gid
);
743 v9fs_string_init(&stat
->muid
);
744 v9fs_string_init(&stat
->extension
);
747 static void v9fs_stat_free(V9fsStat
*stat
)
749 v9fs_string_free(&stat
->name
);
750 v9fs_string_free(&stat
->uid
);
751 v9fs_string_free(&stat
->gid
);
752 v9fs_string_free(&stat
->muid
);
753 v9fs_string_free(&stat
->extension
);
756 static uint32_t stat_to_v9mode(const struct stat
*stbuf
)
760 mode
= stbuf
->st_mode
& 0777;
761 if (S_ISDIR(stbuf
->st_mode
)) {
762 mode
|= P9_STAT_MODE_DIR
;
765 if (S_ISLNK(stbuf
->st_mode
)) {
766 mode
|= P9_STAT_MODE_SYMLINK
;
769 if (S_ISSOCK(stbuf
->st_mode
)) {
770 mode
|= P9_STAT_MODE_SOCKET
;
773 if (S_ISFIFO(stbuf
->st_mode
)) {
774 mode
|= P9_STAT_MODE_NAMED_PIPE
;
777 if (S_ISBLK(stbuf
->st_mode
) || S_ISCHR(stbuf
->st_mode
)) {
778 mode
|= P9_STAT_MODE_DEVICE
;
781 if (stbuf
->st_mode
& S_ISUID
) {
782 mode
|= P9_STAT_MODE_SETUID
;
785 if (stbuf
->st_mode
& S_ISGID
) {
786 mode
|= P9_STAT_MODE_SETGID
;
789 if (stbuf
->st_mode
& S_ISVTX
) {
790 mode
|= P9_STAT_MODE_SETVTX
;
796 static int stat_to_v9stat(V9fsPDU
*pdu
, V9fsPath
*name
,
797 const struct stat
*stbuf
,
803 memset(v9stat
, 0, sizeof(*v9stat
));
805 stat_to_qid(stbuf
, &v9stat
->qid
);
806 v9stat
->mode
= stat_to_v9mode(stbuf
);
807 v9stat
->atime
= stbuf
->st_atime
;
808 v9stat
->mtime
= stbuf
->st_mtime
;
809 v9stat
->length
= stbuf
->st_size
;
811 v9fs_string_null(&v9stat
->uid
);
812 v9fs_string_null(&v9stat
->gid
);
813 v9fs_string_null(&v9stat
->muid
);
815 v9stat
->n_uid
= stbuf
->st_uid
;
816 v9stat
->n_gid
= stbuf
->st_gid
;
819 v9fs_string_null(&v9stat
->extension
);
821 if (v9stat
->mode
& P9_STAT_MODE_SYMLINK
) {
822 err
= v9fs_co_readlink(pdu
, name
, &v9stat
->extension
);
826 } else if (v9stat
->mode
& P9_STAT_MODE_DEVICE
) {
827 v9fs_string_sprintf(&v9stat
->extension
, "%c %u %u",
828 S_ISCHR(stbuf
->st_mode
) ? 'c' : 'b',
829 major(stbuf
->st_rdev
), minor(stbuf
->st_rdev
));
830 } else if (S_ISDIR(stbuf
->st_mode
) || S_ISREG(stbuf
->st_mode
)) {
831 v9fs_string_sprintf(&v9stat
->extension
, "%s %lu",
832 "HARDLINKCOUNT", (unsigned long)stbuf
->st_nlink
);
835 str
= strrchr(name
->data
, '/');
842 v9fs_string_sprintf(&v9stat
->name
, "%s", str
);
845 v9fs_string_size(&v9stat
->name
) +
846 v9fs_string_size(&v9stat
->uid
) +
847 v9fs_string_size(&v9stat
->gid
) +
848 v9fs_string_size(&v9stat
->muid
) +
849 v9fs_string_size(&v9stat
->extension
);
853 #define P9_STATS_MODE 0x00000001ULL
854 #define P9_STATS_NLINK 0x00000002ULL
855 #define P9_STATS_UID 0x00000004ULL
856 #define P9_STATS_GID 0x00000008ULL
857 #define P9_STATS_RDEV 0x00000010ULL
858 #define P9_STATS_ATIME 0x00000020ULL
859 #define P9_STATS_MTIME 0x00000040ULL
860 #define P9_STATS_CTIME 0x00000080ULL
861 #define P9_STATS_INO 0x00000100ULL
862 #define P9_STATS_SIZE 0x00000200ULL
863 #define P9_STATS_BLOCKS 0x00000400ULL
865 #define P9_STATS_BTIME 0x00000800ULL
866 #define P9_STATS_GEN 0x00001000ULL
867 #define P9_STATS_DATA_VERSION 0x00002000ULL
869 #define P9_STATS_BASIC 0x000007ffULL /* Mask for fields up to BLOCKS */
870 #define P9_STATS_ALL 0x00003fffULL /* Mask for All fields above */
873 static void stat_to_v9stat_dotl(V9fsState
*s
, const struct stat
*stbuf
,
874 V9fsStatDotl
*v9lstat
)
876 memset(v9lstat
, 0, sizeof(*v9lstat
));
878 v9lstat
->st_mode
= stbuf
->st_mode
;
879 v9lstat
->st_nlink
= stbuf
->st_nlink
;
880 v9lstat
->st_uid
= stbuf
->st_uid
;
881 v9lstat
->st_gid
= stbuf
->st_gid
;
882 v9lstat
->st_rdev
= stbuf
->st_rdev
;
883 v9lstat
->st_size
= stbuf
->st_size
;
884 v9lstat
->st_blksize
= stbuf
->st_blksize
;
885 v9lstat
->st_blocks
= stbuf
->st_blocks
;
886 v9lstat
->st_atime_sec
= stbuf
->st_atime
;
887 v9lstat
->st_atime_nsec
= stbuf
->st_atim
.tv_nsec
;
888 v9lstat
->st_mtime_sec
= stbuf
->st_mtime
;
889 v9lstat
->st_mtime_nsec
= stbuf
->st_mtim
.tv_nsec
;
890 v9lstat
->st_ctime_sec
= stbuf
->st_ctime
;
891 v9lstat
->st_ctime_nsec
= stbuf
->st_ctim
.tv_nsec
;
892 /* Currently we only support BASIC fields in stat */
893 v9lstat
->st_result_mask
= P9_STATS_BASIC
;
895 stat_to_qid(stbuf
, &v9lstat
->qid
);
898 static void print_sg(struct iovec
*sg
, int cnt
)
902 printf("sg[%d]: {", cnt
);
903 for (i
= 0; i
< cnt
; i
++) {
907 printf("(%p, %zd)", sg
[i
].iov_base
, sg
[i
].iov_len
);
912 /* Will call this only for path name based fid */
913 static void v9fs_fix_path(V9fsPath
*dst
, V9fsPath
*src
, int len
)
916 v9fs_path_init(&str
);
917 v9fs_path_copy(&str
, dst
);
918 v9fs_string_sprintf((V9fsString
*)dst
, "%s%s", src
->data
, str
.data
+len
);
919 v9fs_path_free(&str
);
920 /* +1 to include terminating NULL */
924 static inline bool is_ro_export(FsContext
*ctx
)
926 return ctx
->export_flags
& V9FS_RDONLY
;
929 static void v9fs_version(void *opaque
)
932 V9fsPDU
*pdu
= opaque
;
933 V9fsState
*s
= pdu
->s
;
937 v9fs_string_init(&version
);
938 err
= pdu_unmarshal(pdu
, offset
, "ds", &s
->msize
, &version
);
943 trace_v9fs_version(pdu
->tag
, pdu
->id
, s
->msize
, version
.data
);
947 if (!strcmp(version
.data
, "9P2000.u")) {
948 s
->proto_version
= V9FS_PROTO_2000U
;
949 } else if (!strcmp(version
.data
, "9P2000.L")) {
950 s
->proto_version
= V9FS_PROTO_2000L
;
952 v9fs_string_sprintf(&version
, "unknown");
955 err
= pdu_marshal(pdu
, offset
, "ds", s
->msize
, &version
);
961 trace_v9fs_version_return(pdu
->tag
, pdu
->id
, s
->msize
, version
.data
);
963 pdu_complete(pdu
, offset
);
964 v9fs_string_free(&version
);
967 static void v9fs_attach(void *opaque
)
969 V9fsPDU
*pdu
= opaque
;
970 V9fsState
*s
= pdu
->s
;
971 int32_t fid
, afid
, n_uname
;
972 V9fsString uname
, aname
;
978 v9fs_string_init(&uname
);
979 v9fs_string_init(&aname
);
980 err
= pdu_unmarshal(pdu
, offset
, "ddssd", &fid
,
981 &afid
, &uname
, &aname
, &n_uname
);
985 trace_v9fs_attach(pdu
->tag
, pdu
->id
, fid
, afid
, uname
.data
, aname
.data
);
987 fidp
= alloc_fid(s
, fid
);
993 err
= v9fs_co_name_to_path(pdu
, NULL
, "/", &fidp
->path
);
999 err
= fid_to_qid(pdu
, fidp
, &qid
);
1005 err
= pdu_marshal(pdu
, offset
, "Q", &qid
);
1011 trace_v9fs_attach_return(pdu
->tag
, pdu
->id
,
1012 qid
.type
, qid
.version
, qid
.path
);
1014 * disable migration if we haven't done already.
1015 * attach could get called multiple times for the same export.
1017 if (!s
->migration_blocker
) {
1019 error_setg(&s
->migration_blocker
,
1020 "Migration is disabled when VirtFS export path '%s' is mounted in the guest using mount_tag '%s'",
1021 s
->ctx
.fs_root
? s
->ctx
.fs_root
: "NULL", s
->tag
);
1022 migrate_add_blocker(s
->migration_blocker
);
1027 pdu_complete(pdu
, err
);
1028 v9fs_string_free(&uname
);
1029 v9fs_string_free(&aname
);
1032 static void v9fs_stat(void *opaque
)
1040 V9fsPDU
*pdu
= opaque
;
1042 err
= pdu_unmarshal(pdu
, offset
, "d", &fid
);
1046 trace_v9fs_stat(pdu
->tag
, pdu
->id
, fid
);
1048 fidp
= get_fid(pdu
, fid
);
1053 err
= v9fs_co_lstat(pdu
, &fidp
->path
, &stbuf
);
1057 err
= stat_to_v9stat(pdu
, &fidp
->path
, &stbuf
, &v9stat
);
1061 err
= pdu_marshal(pdu
, offset
, "wS", 0, &v9stat
);
1063 v9fs_stat_free(&v9stat
);
1066 trace_v9fs_stat_return(pdu
->tag
, pdu
->id
, v9stat
.mode
,
1067 v9stat
.atime
, v9stat
.mtime
, v9stat
.length
);
1069 v9fs_stat_free(&v9stat
);
1073 pdu_complete(pdu
, err
);
1076 static void v9fs_getattr(void *opaque
)
1083 uint64_t request_mask
;
1084 V9fsStatDotl v9stat_dotl
;
1085 V9fsPDU
*pdu
= opaque
;
1086 V9fsState
*s
= pdu
->s
;
1088 retval
= pdu_unmarshal(pdu
, offset
, "dq", &fid
, &request_mask
);
1092 trace_v9fs_getattr(pdu
->tag
, pdu
->id
, fid
, request_mask
);
1094 fidp
= get_fid(pdu
, fid
);
1100 * Currently we only support BASIC fields in stat, so there is no
1101 * need to look at request_mask.
1103 retval
= v9fs_co_lstat(pdu
, &fidp
->path
, &stbuf
);
1107 stat_to_v9stat_dotl(s
, &stbuf
, &v9stat_dotl
);
1109 /* fill st_gen if requested and supported by underlying fs */
1110 if (request_mask
& P9_STATS_GEN
) {
1111 retval
= v9fs_co_st_gen(pdu
, &fidp
->path
, stbuf
.st_mode
, &v9stat_dotl
);
1114 /* we have valid st_gen: update result mask */
1115 v9stat_dotl
.st_result_mask
|= P9_STATS_GEN
;
1118 /* request cancelled, e.g. by Tflush */
1121 /* failed to get st_gen: not fatal, ignore */
1125 retval
= pdu_marshal(pdu
, offset
, "A", &v9stat_dotl
);
1130 trace_v9fs_getattr_return(pdu
->tag
, pdu
->id
, v9stat_dotl
.st_result_mask
,
1131 v9stat_dotl
.st_mode
, v9stat_dotl
.st_uid
,
1132 v9stat_dotl
.st_gid
);
1136 pdu_complete(pdu
, retval
);
1139 /* Attribute flags */
1140 #define P9_ATTR_MODE (1 << 0)
1141 #define P9_ATTR_UID (1 << 1)
1142 #define P9_ATTR_GID (1 << 2)
1143 #define P9_ATTR_SIZE (1 << 3)
1144 #define P9_ATTR_ATIME (1 << 4)
1145 #define P9_ATTR_MTIME (1 << 5)
1146 #define P9_ATTR_CTIME (1 << 6)
1147 #define P9_ATTR_ATIME_SET (1 << 7)
1148 #define P9_ATTR_MTIME_SET (1 << 8)
1150 #define P9_ATTR_MASK 127
1152 static void v9fs_setattr(void *opaque
)
1159 V9fsPDU
*pdu
= opaque
;
1161 err
= pdu_unmarshal(pdu
, offset
, "dI", &fid
, &v9iattr
);
1166 fidp
= get_fid(pdu
, fid
);
1171 if (v9iattr
.valid
& P9_ATTR_MODE
) {
1172 err
= v9fs_co_chmod(pdu
, &fidp
->path
, v9iattr
.mode
);
1177 if (v9iattr
.valid
& (P9_ATTR_ATIME
| P9_ATTR_MTIME
)) {
1178 struct timespec times
[2];
1179 if (v9iattr
.valid
& P9_ATTR_ATIME
) {
1180 if (v9iattr
.valid
& P9_ATTR_ATIME_SET
) {
1181 times
[0].tv_sec
= v9iattr
.atime_sec
;
1182 times
[0].tv_nsec
= v9iattr
.atime_nsec
;
1184 times
[0].tv_nsec
= UTIME_NOW
;
1187 times
[0].tv_nsec
= UTIME_OMIT
;
1189 if (v9iattr
.valid
& P9_ATTR_MTIME
) {
1190 if (v9iattr
.valid
& P9_ATTR_MTIME_SET
) {
1191 times
[1].tv_sec
= v9iattr
.mtime_sec
;
1192 times
[1].tv_nsec
= v9iattr
.mtime_nsec
;
1194 times
[1].tv_nsec
= UTIME_NOW
;
1197 times
[1].tv_nsec
= UTIME_OMIT
;
1199 err
= v9fs_co_utimensat(pdu
, &fidp
->path
, times
);
1205 * If the only valid entry in iattr is ctime we can call
1206 * chown(-1,-1) to update the ctime of the file
1208 if ((v9iattr
.valid
& (P9_ATTR_UID
| P9_ATTR_GID
)) ||
1209 ((v9iattr
.valid
& P9_ATTR_CTIME
)
1210 && !((v9iattr
.valid
& P9_ATTR_MASK
) & ~P9_ATTR_CTIME
))) {
1211 if (!(v9iattr
.valid
& P9_ATTR_UID
)) {
1214 if (!(v9iattr
.valid
& P9_ATTR_GID
)) {
1217 err
= v9fs_co_chown(pdu
, &fidp
->path
, v9iattr
.uid
,
1223 if (v9iattr
.valid
& (P9_ATTR_SIZE
)) {
1224 err
= v9fs_co_truncate(pdu
, &fidp
->path
, v9iattr
.size
);
1233 pdu_complete(pdu
, err
);
1236 static int v9fs_walk_marshal(V9fsPDU
*pdu
, uint16_t nwnames
, V9fsQID
*qids
)
1242 err
= pdu_marshal(pdu
, offset
, "w", nwnames
);
1247 for (i
= 0; i
< nwnames
; i
++) {
1248 err
= pdu_marshal(pdu
, offset
, "Q", &qids
[i
]);
1257 static void v9fs_walk(void *opaque
)
1260 V9fsQID
*qids
= NULL
;
1262 V9fsPath dpath
, path
;
1266 int32_t fid
, newfid
;
1267 V9fsString
*wnames
= NULL
;
1269 V9fsFidState
*newfidp
= NULL
;
1270 V9fsPDU
*pdu
= opaque
;
1271 V9fsState
*s
= pdu
->s
;
1273 err
= pdu_unmarshal(pdu
, offset
, "ddw", &fid
, &newfid
, &nwnames
);
1275 pdu_complete(pdu
, err
);
1280 trace_v9fs_walk(pdu
->tag
, pdu
->id
, fid
, newfid
, nwnames
);
1282 if (nwnames
&& nwnames
<= P9_MAXWELEM
) {
1283 wnames
= g_malloc0(sizeof(wnames
[0]) * nwnames
);
1284 qids
= g_malloc0(sizeof(qids
[0]) * nwnames
);
1285 for (i
= 0; i
< nwnames
; i
++) {
1286 err
= pdu_unmarshal(pdu
, offset
, "s", &wnames
[i
]);
1292 } else if (nwnames
> P9_MAXWELEM
) {
1296 fidp
= get_fid(pdu
, fid
);
1301 v9fs_path_init(&dpath
);
1302 v9fs_path_init(&path
);
1304 * Both dpath and path initially poin to fidp.
1305 * Needed to handle request with nwnames == 0
1307 v9fs_path_copy(&dpath
, &fidp
->path
);
1308 v9fs_path_copy(&path
, &fidp
->path
);
1309 for (name_idx
= 0; name_idx
< nwnames
; name_idx
++) {
1310 err
= v9fs_co_name_to_path(pdu
, &dpath
, wnames
[name_idx
].data
, &path
);
1314 err
= v9fs_co_lstat(pdu
, &path
, &stbuf
);
1318 stat_to_qid(&stbuf
, &qids
[name_idx
]);
1319 v9fs_path_copy(&dpath
, &path
);
1321 if (fid
== newfid
) {
1322 BUG_ON(fidp
->fid_type
!= P9_FID_NONE
);
1323 v9fs_path_copy(&fidp
->path
, &path
);
1325 newfidp
= alloc_fid(s
, newfid
);
1326 if (newfidp
== NULL
) {
1330 newfidp
->uid
= fidp
->uid
;
1331 v9fs_path_copy(&newfidp
->path
, &path
);
1333 err
= v9fs_walk_marshal(pdu
, nwnames
, qids
);
1334 trace_v9fs_walk_return(pdu
->tag
, pdu
->id
, nwnames
, qids
);
1338 put_fid(pdu
, newfidp
);
1340 v9fs_path_free(&dpath
);
1341 v9fs_path_free(&path
);
1343 pdu_complete(pdu
, err
);
1344 if (nwnames
&& nwnames
<= P9_MAXWELEM
) {
1345 for (name_idx
= 0; name_idx
< nwnames
; name_idx
++) {
1346 v9fs_string_free(&wnames
[name_idx
]);
1353 static int32_t get_iounit(V9fsPDU
*pdu
, V9fsPath
*path
)
1355 struct statfs stbuf
;
1357 V9fsState
*s
= pdu
->s
;
1360 * iounit should be multiples of f_bsize (host filesystem block size
1361 * and as well as less than (client msize - P9_IOHDRSZ))
1363 if (!v9fs_co_statfs(pdu
, path
, &stbuf
)) {
1364 iounit
= stbuf
.f_bsize
;
1365 iounit
*= (s
->msize
- P9_IOHDRSZ
)/stbuf
.f_bsize
;
1368 iounit
= s
->msize
- P9_IOHDRSZ
;
1373 static void v9fs_open(void *opaque
)
1384 V9fsPDU
*pdu
= opaque
;
1385 V9fsState
*s
= pdu
->s
;
1387 if (s
->proto_version
== V9FS_PROTO_2000L
) {
1388 err
= pdu_unmarshal(pdu
, offset
, "dd", &fid
, &mode
);
1391 err
= pdu_unmarshal(pdu
, offset
, "db", &fid
, &modebyte
);
1397 trace_v9fs_open(pdu
->tag
, pdu
->id
, fid
, mode
);
1399 fidp
= get_fid(pdu
, fid
);
1404 BUG_ON(fidp
->fid_type
!= P9_FID_NONE
);
1406 err
= v9fs_co_lstat(pdu
, &fidp
->path
, &stbuf
);
1410 stat_to_qid(&stbuf
, &qid
);
1411 if (S_ISDIR(stbuf
.st_mode
)) {
1412 err
= v9fs_co_opendir(pdu
, fidp
);
1416 fidp
->fid_type
= P9_FID_DIR
;
1417 err
= pdu_marshal(pdu
, offset
, "Qd", &qid
, 0);
1423 if (s
->proto_version
== V9FS_PROTO_2000L
) {
1424 flags
= get_dotl_openflags(s
, mode
);
1426 flags
= omode_to_uflags(mode
);
1428 if (is_ro_export(&s
->ctx
)) {
1429 if (mode
& O_WRONLY
|| mode
& O_RDWR
||
1430 mode
& O_APPEND
|| mode
& O_TRUNC
) {
1435 err
= v9fs_co_open(pdu
, fidp
, flags
);
1439 fidp
->fid_type
= P9_FID_FILE
;
1440 fidp
->open_flags
= flags
;
1441 if (flags
& O_EXCL
) {
1443 * We let the host file system do O_EXCL check
1444 * We should not reclaim such fd
1446 fidp
->flags
|= FID_NON_RECLAIMABLE
;
1448 iounit
= get_iounit(pdu
, &fidp
->path
);
1449 err
= pdu_marshal(pdu
, offset
, "Qd", &qid
, iounit
);
1455 trace_v9fs_open_return(pdu
->tag
, pdu
->id
,
1456 qid
.type
, qid
.version
, qid
.path
, iounit
);
1460 pdu_complete(pdu
, err
);
1463 static void v9fs_lcreate(void *opaque
)
1465 int32_t dfid
, flags
, mode
;
1474 V9fsPDU
*pdu
= opaque
;
1476 v9fs_string_init(&name
);
1477 err
= pdu_unmarshal(pdu
, offset
, "dsddd", &dfid
,
1478 &name
, &flags
, &mode
, &gid
);
1482 trace_v9fs_lcreate(pdu
->tag
, pdu
->id
, dfid
, flags
, mode
, gid
);
1484 fidp
= get_fid(pdu
, dfid
);
1490 flags
= get_dotl_openflags(pdu
->s
, flags
);
1491 err
= v9fs_co_open2(pdu
, fidp
, &name
, gid
,
1492 flags
| O_CREAT
, mode
, &stbuf
);
1496 fidp
->fid_type
= P9_FID_FILE
;
1497 fidp
->open_flags
= flags
;
1498 if (flags
& O_EXCL
) {
1500 * We let the host file system do O_EXCL check
1501 * We should not reclaim such fd
1503 fidp
->flags
|= FID_NON_RECLAIMABLE
;
1505 iounit
= get_iounit(pdu
, &fidp
->path
);
1506 stat_to_qid(&stbuf
, &qid
);
1507 err
= pdu_marshal(pdu
, offset
, "Qd", &qid
, iounit
);
1512 trace_v9fs_lcreate_return(pdu
->tag
, pdu
->id
,
1513 qid
.type
, qid
.version
, qid
.path
, iounit
);
1517 pdu_complete(pdu
, err
);
1518 v9fs_string_free(&name
);
1521 static void v9fs_fsync(void *opaque
)
1528 V9fsPDU
*pdu
= opaque
;
1530 err
= pdu_unmarshal(pdu
, offset
, "dd", &fid
, &datasync
);
1534 trace_v9fs_fsync(pdu
->tag
, pdu
->id
, fid
, datasync
);
1536 fidp
= get_fid(pdu
, fid
);
1541 err
= v9fs_co_fsync(pdu
, fidp
, datasync
);
1547 pdu_complete(pdu
, err
);
1550 static void v9fs_clunk(void *opaque
)
1556 V9fsPDU
*pdu
= opaque
;
1557 V9fsState
*s
= pdu
->s
;
1559 err
= pdu_unmarshal(pdu
, offset
, "d", &fid
);
1563 trace_v9fs_clunk(pdu
->tag
, pdu
->id
, fid
);
1565 fidp
= clunk_fid(s
, fid
);
1571 * Bump the ref so that put_fid will
1575 err
= put_fid(pdu
, fidp
);
1580 pdu_complete(pdu
, err
);
1583 static int v9fs_xattr_read(V9fsState
*s
, V9fsPDU
*pdu
, V9fsFidState
*fidp
,
1584 uint64_t off
, uint32_t max_count
)
1590 V9fsVirtioState
*v
= container_of(s
, V9fsVirtioState
, state
);
1591 VirtQueueElement
*elem
= v
->elems
[pdu
->idx
];
1593 xattr_len
= fidp
->fs
.xattr
.len
;
1594 read_count
= xattr_len
- off
;
1595 if (read_count
> max_count
) {
1596 read_count
= max_count
;
1597 } else if (read_count
< 0) {
1599 * read beyond XATTR value
1603 err
= pdu_marshal(pdu
, offset
, "d", read_count
);
1609 err
= v9fs_pack(elem
->in_sg
, elem
->in_num
, offset
,
1610 ((char *)fidp
->fs
.xattr
.value
) + off
,
1619 static int v9fs_do_readdir_with_stat(V9fsPDU
*pdu
,
1620 V9fsFidState
*fidp
, uint32_t max_count
)
1627 off_t saved_dir_pos
;
1628 struct dirent
*dent
, *result
;
1630 /* save the directory position */
1631 saved_dir_pos
= v9fs_co_telldir(pdu
, fidp
);
1632 if (saved_dir_pos
< 0) {
1633 return saved_dir_pos
;
1636 dent
= g_malloc(sizeof(struct dirent
));
1639 v9fs_path_init(&path
);
1640 err
= v9fs_co_readdir_r(pdu
, fidp
, dent
, &result
);
1641 if (err
|| !result
) {
1644 err
= v9fs_co_name_to_path(pdu
, &fidp
->path
, dent
->d_name
, &path
);
1648 err
= v9fs_co_lstat(pdu
, &path
, &stbuf
);
1652 err
= stat_to_v9stat(pdu
, &path
, &stbuf
, &v9stat
);
1656 /* 11 = 7 + 4 (7 = start offset, 4 = space for storing count) */
1657 len
= pdu_marshal(pdu
, 11 + count
, "S", &v9stat
);
1658 if ((len
!= (v9stat
.size
+ 2)) || ((count
+ len
) > max_count
)) {
1659 /* Ran out of buffer. Set dir back to old position and return */
1660 v9fs_co_seekdir(pdu
, fidp
, saved_dir_pos
);
1661 v9fs_stat_free(&v9stat
);
1662 v9fs_path_free(&path
);
1667 v9fs_stat_free(&v9stat
);
1668 v9fs_path_free(&path
);
1669 saved_dir_pos
= dent
->d_off
;
1673 v9fs_path_free(&path
);
1681 * Create a QEMUIOVector for a sub-region of PDU iovecs
1683 * @qiov: uninitialized QEMUIOVector
1684 * @skip: number of bytes to skip from beginning of PDU
1685 * @size: number of bytes to include
1686 * @is_write: true - write, false - read
1688 * The resulting QEMUIOVector has heap-allocated iovecs and must be cleaned up
1689 * with qemu_iovec_destroy().
1691 static void v9fs_init_qiov_from_pdu(QEMUIOVector
*qiov
, V9fsPDU
*pdu
,
1692 size_t skip
, size_t size
,
1699 virtio_init_iov_from_pdu(pdu
, &iov
, &niov
, is_write
);
1701 qemu_iovec_init_external(&elem
, iov
, niov
);
1702 qemu_iovec_init(qiov
, niov
);
1703 qemu_iovec_concat(qiov
, &elem
, skip
, size
);
1706 static void v9fs_read(void *opaque
)
1715 V9fsPDU
*pdu
= opaque
;
1716 V9fsState
*s
= pdu
->s
;
1718 err
= pdu_unmarshal(pdu
, offset
, "dqd", &fid
, &off
, &max_count
);
1722 trace_v9fs_read(pdu
->tag
, pdu
->id
, fid
, off
, max_count
);
1724 fidp
= get_fid(pdu
, fid
);
1729 if (fidp
->fid_type
== P9_FID_DIR
) {
1732 v9fs_co_rewinddir(pdu
, fidp
);
1734 count
= v9fs_do_readdir_with_stat(pdu
, fidp
, max_count
);
1739 err
= pdu_marshal(pdu
, offset
, "d", count
);
1743 err
+= offset
+ count
;
1744 } else if (fidp
->fid_type
== P9_FID_FILE
) {
1745 QEMUIOVector qiov_full
;
1749 v9fs_init_qiov_from_pdu(&qiov_full
, pdu
, offset
+ 4, max_count
, false);
1750 qemu_iovec_init(&qiov
, qiov_full
.niov
);
1752 qemu_iovec_reset(&qiov
);
1753 qemu_iovec_concat(&qiov
, &qiov_full
, count
, qiov_full
.size
- count
);
1755 print_sg(qiov
.iov
, qiov
.niov
);
1757 /* Loop in case of EINTR */
1759 len
= v9fs_co_preadv(pdu
, fidp
, qiov
.iov
, qiov
.niov
, off
);
1764 } while (len
== -EINTR
&& !pdu
->cancelled
);
1766 /* IO error return the error */
1770 } while (count
< max_count
&& len
> 0);
1771 err
= pdu_marshal(pdu
, offset
, "d", count
);
1775 err
+= offset
+ count
;
1776 qemu_iovec_destroy(&qiov
);
1777 qemu_iovec_destroy(&qiov_full
);
1778 } else if (fidp
->fid_type
== P9_FID_XATTR
) {
1779 err
= v9fs_xattr_read(s
, pdu
, fidp
, off
, max_count
);
1783 trace_v9fs_read_return(pdu
->tag
, pdu
->id
, count
, err
);
1787 pdu_complete(pdu
, err
);
1790 static size_t v9fs_readdir_data_size(V9fsString
*name
)
1793 * Size of each dirent on the wire: size of qid (13) + size of offset (8)
1794 * size of type (1) + size of name.size (2) + strlen(name.data)
1796 return 24 + v9fs_string_size(name
);
1799 static int v9fs_do_readdir(V9fsPDU
*pdu
,
1800 V9fsFidState
*fidp
, int32_t max_count
)
1807 off_t saved_dir_pos
;
1808 struct dirent
*dent
, *result
;
1810 /* save the directory position */
1811 saved_dir_pos
= v9fs_co_telldir(pdu
, fidp
);
1812 if (saved_dir_pos
< 0) {
1813 return saved_dir_pos
;
1816 dent
= g_malloc(sizeof(struct dirent
));
1819 err
= v9fs_co_readdir_r(pdu
, fidp
, dent
, &result
);
1820 if (err
|| !result
) {
1823 v9fs_string_init(&name
);
1824 v9fs_string_sprintf(&name
, "%s", dent
->d_name
);
1825 if ((count
+ v9fs_readdir_data_size(&name
)) > max_count
) {
1826 /* Ran out of buffer. Set dir back to old position and return */
1827 v9fs_co_seekdir(pdu
, fidp
, saved_dir_pos
);
1828 v9fs_string_free(&name
);
1833 * Fill up just the path field of qid because the client uses
1834 * only that. To fill the entire qid structure we will have
1835 * to stat each dirent found, which is expensive
1837 size
= MIN(sizeof(dent
->d_ino
), sizeof(qid
.path
));
1838 memcpy(&qid
.path
, &dent
->d_ino
, size
);
1839 /* Fill the other fields with dummy values */
1843 /* 11 = 7 + 4 (7 = start offset, 4 = space for storing count) */
1844 len
= pdu_marshal(pdu
, 11 + count
, "Qqbs",
1846 dent
->d_type
, &name
);
1848 v9fs_co_seekdir(pdu
, fidp
, saved_dir_pos
);
1849 v9fs_string_free(&name
);
1854 v9fs_string_free(&name
);
1855 saved_dir_pos
= dent
->d_off
;
1864 static void v9fs_readdir(void *opaque
)
1870 uint64_t initial_offset
;
1873 V9fsPDU
*pdu
= opaque
;
1875 retval
= pdu_unmarshal(pdu
, offset
, "dqd", &fid
,
1876 &initial_offset
, &max_count
);
1880 trace_v9fs_readdir(pdu
->tag
, pdu
->id
, fid
, initial_offset
, max_count
);
1882 fidp
= get_fid(pdu
, fid
);
1887 if (!fidp
->fs
.dir
) {
1891 if (initial_offset
== 0) {
1892 v9fs_co_rewinddir(pdu
, fidp
);
1894 v9fs_co_seekdir(pdu
, fidp
, initial_offset
);
1896 count
= v9fs_do_readdir(pdu
, fidp
, max_count
);
1901 retval
= pdu_marshal(pdu
, offset
, "d", count
);
1905 retval
+= count
+ offset
;
1906 trace_v9fs_readdir_return(pdu
->tag
, pdu
->id
, count
, retval
);
1910 pdu_complete(pdu
, retval
);
1913 static int v9fs_xattr_write(V9fsState
*s
, V9fsPDU
*pdu
, V9fsFidState
*fidp
,
1914 uint64_t off
, uint32_t count
,
1915 struct iovec
*sg
, int cnt
)
1924 xattr_len
= fidp
->fs
.xattr
.len
;
1925 write_count
= xattr_len
- off
;
1926 if (write_count
> count
) {
1927 write_count
= count
;
1928 } else if (write_count
< 0) {
1930 * write beyond XATTR value len specified in
1936 err
= pdu_marshal(pdu
, offset
, "d", write_count
);
1941 fidp
->fs
.xattr
.copied_len
+= write_count
;
1943 * Now copy the content from sg list
1945 for (i
= 0; i
< cnt
; i
++) {
1946 if (write_count
> sg
[i
].iov_len
) {
1947 to_copy
= sg
[i
].iov_len
;
1949 to_copy
= write_count
;
1951 memcpy((char *)fidp
->fs
.xattr
.value
+ off
, sg
[i
].iov_base
, to_copy
);
1952 /* updating vs->off since we are not using below */
1954 write_count
-= to_copy
;
1960 static void v9fs_write(void *opaque
)
1970 V9fsPDU
*pdu
= opaque
;
1971 V9fsState
*s
= pdu
->s
;
1972 QEMUIOVector qiov_full
;
1975 err
= pdu_unmarshal(pdu
, offset
, "dqd", &fid
, &off
, &count
);
1977 pdu_complete(pdu
, err
);
1981 v9fs_init_qiov_from_pdu(&qiov_full
, pdu
, offset
, count
, true);
1982 trace_v9fs_write(pdu
->tag
, pdu
->id
, fid
, off
, count
, qiov_full
.niov
);
1984 fidp
= get_fid(pdu
, fid
);
1989 if (fidp
->fid_type
== P9_FID_FILE
) {
1990 if (fidp
->fs
.fd
== -1) {
1994 } else if (fidp
->fid_type
== P9_FID_XATTR
) {
1996 * setxattr operation
1998 err
= v9fs_xattr_write(s
, pdu
, fidp
, off
, count
,
1999 qiov_full
.iov
, qiov_full
.niov
);
2005 qemu_iovec_init(&qiov
, qiov_full
.niov
);
2007 qemu_iovec_reset(&qiov
);
2008 qemu_iovec_concat(&qiov
, &qiov_full
, total
, qiov_full
.size
- total
);
2010 print_sg(qiov
.iov
, qiov
.niov
);
2012 /* Loop in case of EINTR */
2014 len
= v9fs_co_pwritev(pdu
, fidp
, qiov
.iov
, qiov
.niov
, off
);
2019 } while (len
== -EINTR
&& !pdu
->cancelled
);
2021 /* IO error return the error */
2025 } while (total
< count
&& len
> 0);
2028 err
= pdu_marshal(pdu
, offset
, "d", total
);
2033 trace_v9fs_write_return(pdu
->tag
, pdu
->id
, total
, err
);
2035 qemu_iovec_destroy(&qiov
);
2039 qemu_iovec_destroy(&qiov_full
);
2040 pdu_complete(pdu
, err
);
2043 static void v9fs_create(void *opaque
)
2055 V9fsString extension
;
2057 V9fsPDU
*pdu
= opaque
;
2059 v9fs_path_init(&path
);
2060 v9fs_string_init(&name
);
2061 v9fs_string_init(&extension
);
2062 err
= pdu_unmarshal(pdu
, offset
, "dsdbs", &fid
, &name
,
2063 &perm
, &mode
, &extension
);
2067 trace_v9fs_create(pdu
->tag
, pdu
->id
, fid
, name
.data
, perm
, mode
);
2069 fidp
= get_fid(pdu
, fid
);
2074 if (perm
& P9_STAT_MODE_DIR
) {
2075 err
= v9fs_co_mkdir(pdu
, fidp
, &name
, perm
& 0777,
2076 fidp
->uid
, -1, &stbuf
);
2080 err
= v9fs_co_name_to_path(pdu
, &fidp
->path
, name
.data
, &path
);
2084 v9fs_path_copy(&fidp
->path
, &path
);
2085 err
= v9fs_co_opendir(pdu
, fidp
);
2089 fidp
->fid_type
= P9_FID_DIR
;
2090 } else if (perm
& P9_STAT_MODE_SYMLINK
) {
2091 err
= v9fs_co_symlink(pdu
, fidp
, &name
,
2092 extension
.data
, -1 , &stbuf
);
2096 err
= v9fs_co_name_to_path(pdu
, &fidp
->path
, name
.data
, &path
);
2100 v9fs_path_copy(&fidp
->path
, &path
);
2101 } else if (perm
& P9_STAT_MODE_LINK
) {
2102 int32_t ofid
= atoi(extension
.data
);
2103 V9fsFidState
*ofidp
= get_fid(pdu
, ofid
);
2104 if (ofidp
== NULL
) {
2108 err
= v9fs_co_link(pdu
, ofidp
, fidp
, &name
);
2109 put_fid(pdu
, ofidp
);
2113 err
= v9fs_co_name_to_path(pdu
, &fidp
->path
, name
.data
, &path
);
2115 fidp
->fid_type
= P9_FID_NONE
;
2118 v9fs_path_copy(&fidp
->path
, &path
);
2119 err
= v9fs_co_lstat(pdu
, &fidp
->path
, &stbuf
);
2121 fidp
->fid_type
= P9_FID_NONE
;
2124 } else if (perm
& P9_STAT_MODE_DEVICE
) {
2126 uint32_t major
, minor
;
2129 if (sscanf(extension
.data
, "%c %u %u", &ctype
, &major
, &minor
) != 3) {
2146 nmode
|= perm
& 0777;
2147 err
= v9fs_co_mknod(pdu
, fidp
, &name
, fidp
->uid
, -1,
2148 makedev(major
, minor
), nmode
, &stbuf
);
2152 err
= v9fs_co_name_to_path(pdu
, &fidp
->path
, name
.data
, &path
);
2156 v9fs_path_copy(&fidp
->path
, &path
);
2157 } else if (perm
& P9_STAT_MODE_NAMED_PIPE
) {
2158 err
= v9fs_co_mknod(pdu
, fidp
, &name
, fidp
->uid
, -1,
2159 0, S_IFIFO
| (perm
& 0777), &stbuf
);
2163 err
= v9fs_co_name_to_path(pdu
, &fidp
->path
, name
.data
, &path
);
2167 v9fs_path_copy(&fidp
->path
, &path
);
2168 } else if (perm
& P9_STAT_MODE_SOCKET
) {
2169 err
= v9fs_co_mknod(pdu
, fidp
, &name
, fidp
->uid
, -1,
2170 0, S_IFSOCK
| (perm
& 0777), &stbuf
);
2174 err
= v9fs_co_name_to_path(pdu
, &fidp
->path
, name
.data
, &path
);
2178 v9fs_path_copy(&fidp
->path
, &path
);
2180 err
= v9fs_co_open2(pdu
, fidp
, &name
, -1,
2181 omode_to_uflags(mode
)|O_CREAT
, perm
, &stbuf
);
2185 fidp
->fid_type
= P9_FID_FILE
;
2186 fidp
->open_flags
= omode_to_uflags(mode
);
2187 if (fidp
->open_flags
& O_EXCL
) {
2189 * We let the host file system do O_EXCL check
2190 * We should not reclaim such fd
2192 fidp
->flags
|= FID_NON_RECLAIMABLE
;
2195 iounit
= get_iounit(pdu
, &fidp
->path
);
2196 stat_to_qid(&stbuf
, &qid
);
2197 err
= pdu_marshal(pdu
, offset
, "Qd", &qid
, iounit
);
2202 trace_v9fs_create_return(pdu
->tag
, pdu
->id
,
2203 qid
.type
, qid
.version
, qid
.path
, iounit
);
2207 pdu_complete(pdu
, err
);
2208 v9fs_string_free(&name
);
2209 v9fs_string_free(&extension
);
2210 v9fs_path_free(&path
);
2213 static void v9fs_symlink(void *opaque
)
2215 V9fsPDU
*pdu
= opaque
;
2218 V9fsFidState
*dfidp
;
2226 v9fs_string_init(&name
);
2227 v9fs_string_init(&symname
);
2228 err
= pdu_unmarshal(pdu
, offset
, "dssd", &dfid
, &name
, &symname
, &gid
);
2232 trace_v9fs_symlink(pdu
->tag
, pdu
->id
, dfid
, name
.data
, symname
.data
, gid
);
2234 dfidp
= get_fid(pdu
, dfid
);
2235 if (dfidp
== NULL
) {
2239 err
= v9fs_co_symlink(pdu
, dfidp
, &name
, symname
.data
, gid
, &stbuf
);
2243 stat_to_qid(&stbuf
, &qid
);
2244 err
= pdu_marshal(pdu
, offset
, "Q", &qid
);
2249 trace_v9fs_symlink_return(pdu
->tag
, pdu
->id
,
2250 qid
.type
, qid
.version
, qid
.path
);
2252 put_fid(pdu
, dfidp
);
2254 pdu_complete(pdu
, err
);
2255 v9fs_string_free(&name
);
2256 v9fs_string_free(&symname
);
2259 static void v9fs_flush(void *opaque
)
2264 V9fsPDU
*cancel_pdu
;
2265 V9fsPDU
*pdu
= opaque
;
2266 V9fsState
*s
= pdu
->s
;
2268 err
= pdu_unmarshal(pdu
, offset
, "w", &tag
);
2270 pdu_complete(pdu
, err
);
2273 trace_v9fs_flush(pdu
->tag
, pdu
->id
, tag
);
2275 QLIST_FOREACH(cancel_pdu
, &s
->active_list
, next
) {
2276 if (cancel_pdu
->tag
== tag
) {
2281 cancel_pdu
->cancelled
= 1;
2283 * Wait for pdu to complete.
2285 qemu_co_queue_wait(&cancel_pdu
->complete
);
2286 cancel_pdu
->cancelled
= 0;
2287 pdu_free(cancel_pdu
);
2289 pdu_complete(pdu
, 7);
2292 static void v9fs_link(void *opaque
)
2294 V9fsPDU
*pdu
= opaque
;
2295 int32_t dfid
, oldfid
;
2296 V9fsFidState
*dfidp
, *oldfidp
;
2301 v9fs_string_init(&name
);
2302 err
= pdu_unmarshal(pdu
, offset
, "dds", &dfid
, &oldfid
, &name
);
2306 trace_v9fs_link(pdu
->tag
, pdu
->id
, dfid
, oldfid
, name
.data
);
2308 dfidp
= get_fid(pdu
, dfid
);
2309 if (dfidp
== NULL
) {
2314 oldfidp
= get_fid(pdu
, oldfid
);
2315 if (oldfidp
== NULL
) {
2319 err
= v9fs_co_link(pdu
, oldfidp
, dfidp
, &name
);
2324 put_fid(pdu
, dfidp
);
2326 v9fs_string_free(&name
);
2327 pdu_complete(pdu
, err
);
2330 /* Only works with path name based fid */
2331 static void v9fs_remove(void *opaque
)
2337 V9fsPDU
*pdu
= opaque
;
2339 err
= pdu_unmarshal(pdu
, offset
, "d", &fid
);
2343 trace_v9fs_remove(pdu
->tag
, pdu
->id
, fid
);
2345 fidp
= get_fid(pdu
, fid
);
2350 /* if fs driver is not path based, return EOPNOTSUPP */
2351 if (!(pdu
->s
->ctx
.export_flags
& V9FS_PATHNAME_FSCONTEXT
)) {
2356 * IF the file is unlinked, we cannot reopen
2357 * the file later. So don't reclaim fd
2359 err
= v9fs_mark_fids_unreclaim(pdu
, &fidp
->path
);
2363 err
= v9fs_co_remove(pdu
, &fidp
->path
);
2368 /* For TREMOVE we need to clunk the fid even on failed remove */
2369 clunk_fid(pdu
->s
, fidp
->fid
);
2372 pdu_complete(pdu
, err
);
2375 static void v9fs_unlinkat(void *opaque
)
2379 int32_t dfid
, flags
;
2382 V9fsFidState
*dfidp
;
2383 V9fsPDU
*pdu
= opaque
;
2385 v9fs_string_init(&name
);
2386 err
= pdu_unmarshal(pdu
, offset
, "dsd", &dfid
, &name
, &flags
);
2390 dfidp
= get_fid(pdu
, dfid
);
2391 if (dfidp
== NULL
) {
2396 * IF the file is unlinked, we cannot reopen
2397 * the file later. So don't reclaim fd
2399 v9fs_path_init(&path
);
2400 err
= v9fs_co_name_to_path(pdu
, &dfidp
->path
, name
.data
, &path
);
2404 err
= v9fs_mark_fids_unreclaim(pdu
, &path
);
2408 err
= v9fs_co_unlinkat(pdu
, &dfidp
->path
, &name
, flags
);
2413 put_fid(pdu
, dfidp
);
2414 v9fs_path_free(&path
);
2416 pdu_complete(pdu
, err
);
2417 v9fs_string_free(&name
);
2421 /* Only works with path name based fid */
2422 static int v9fs_complete_rename(V9fsPDU
*pdu
, V9fsFidState
*fidp
,
2423 int32_t newdirfid
, V9fsString
*name
)
2428 V9fsFidState
*tfidp
;
2429 V9fsState
*s
= pdu
->s
;
2430 V9fsFidState
*dirfidp
= NULL
;
2431 char *old_name
, *new_name
;
2433 v9fs_path_init(&new_path
);
2434 if (newdirfid
!= -1) {
2435 dirfidp
= get_fid(pdu
, newdirfid
);
2436 if (dirfidp
== NULL
) {
2440 BUG_ON(dirfidp
->fid_type
!= P9_FID_NONE
);
2441 v9fs_co_name_to_path(pdu
, &dirfidp
->path
, name
->data
, &new_path
);
2443 old_name
= fidp
->path
.data
;
2444 end
= strrchr(old_name
, '/');
2450 new_name
= g_malloc0(end
- old_name
+ name
->size
+ 1);
2451 strncat(new_name
, old_name
, end
- old_name
);
2452 strncat(new_name
+ (end
- old_name
), name
->data
, name
->size
);
2453 v9fs_co_name_to_path(pdu
, NULL
, new_name
, &new_path
);
2456 err
= v9fs_co_rename(pdu
, &fidp
->path
, &new_path
);
2461 * Fixup fid's pointing to the old name to
2462 * start pointing to the new name
2464 for (tfidp
= s
->fid_list
; tfidp
; tfidp
= tfidp
->next
) {
2465 if (v9fs_path_is_ancestor(&fidp
->path
, &tfidp
->path
)) {
2466 /* replace the name */
2467 v9fs_fix_path(&tfidp
->path
, &new_path
, strlen(fidp
->path
.data
));
2472 put_fid(pdu
, dirfidp
);
2474 v9fs_path_free(&new_path
);
2479 /* Only works with path name based fid */
2480 static void v9fs_rename(void *opaque
)
2488 V9fsPDU
*pdu
= opaque
;
2489 V9fsState
*s
= pdu
->s
;
2491 v9fs_string_init(&name
);
2492 err
= pdu_unmarshal(pdu
, offset
, "dds", &fid
, &newdirfid
, &name
);
2496 fidp
= get_fid(pdu
, fid
);
2501 BUG_ON(fidp
->fid_type
!= P9_FID_NONE
);
2502 /* if fs driver is not path based, return EOPNOTSUPP */
2503 if (!(pdu
->s
->ctx
.export_flags
& V9FS_PATHNAME_FSCONTEXT
)) {
2507 v9fs_path_write_lock(s
);
2508 err
= v9fs_complete_rename(pdu
, fidp
, newdirfid
, &name
);
2509 v9fs_path_unlock(s
);
2516 pdu_complete(pdu
, err
);
2517 v9fs_string_free(&name
);
2520 static void v9fs_fix_fid_paths(V9fsPDU
*pdu
, V9fsPath
*olddir
,
2521 V9fsString
*old_name
, V9fsPath
*newdir
,
2522 V9fsString
*new_name
)
2524 V9fsFidState
*tfidp
;
2525 V9fsPath oldpath
, newpath
;
2526 V9fsState
*s
= pdu
->s
;
2529 v9fs_path_init(&oldpath
);
2530 v9fs_path_init(&newpath
);
2531 v9fs_co_name_to_path(pdu
, olddir
, old_name
->data
, &oldpath
);
2532 v9fs_co_name_to_path(pdu
, newdir
, new_name
->data
, &newpath
);
2535 * Fixup fid's pointing to the old name to
2536 * start pointing to the new name
2538 for (tfidp
= s
->fid_list
; tfidp
; tfidp
= tfidp
->next
) {
2539 if (v9fs_path_is_ancestor(&oldpath
, &tfidp
->path
)) {
2540 /* replace the name */
2541 v9fs_fix_path(&tfidp
->path
, &newpath
, strlen(oldpath
.data
));
2544 v9fs_path_free(&oldpath
);
2545 v9fs_path_free(&newpath
);
2548 static int v9fs_complete_renameat(V9fsPDU
*pdu
, int32_t olddirfid
,
2549 V9fsString
*old_name
, int32_t newdirfid
,
2550 V9fsString
*new_name
)
2553 V9fsState
*s
= pdu
->s
;
2554 V9fsFidState
*newdirfidp
= NULL
, *olddirfidp
= NULL
;
2556 olddirfidp
= get_fid(pdu
, olddirfid
);
2557 if (olddirfidp
== NULL
) {
2561 if (newdirfid
!= -1) {
2562 newdirfidp
= get_fid(pdu
, newdirfid
);
2563 if (newdirfidp
== NULL
) {
2568 newdirfidp
= get_fid(pdu
, olddirfid
);
2571 err
= v9fs_co_renameat(pdu
, &olddirfidp
->path
, old_name
,
2572 &newdirfidp
->path
, new_name
);
2576 if (s
->ctx
.export_flags
& V9FS_PATHNAME_FSCONTEXT
) {
2577 /* Only for path based fid we need to do the below fixup */
2578 v9fs_fix_fid_paths(pdu
, &olddirfidp
->path
, old_name
,
2579 &newdirfidp
->path
, new_name
);
2583 put_fid(pdu
, olddirfidp
);
2586 put_fid(pdu
, newdirfidp
);
2591 static void v9fs_renameat(void *opaque
)
2595 V9fsPDU
*pdu
= opaque
;
2596 V9fsState
*s
= pdu
->s
;
2597 int32_t olddirfid
, newdirfid
;
2598 V9fsString old_name
, new_name
;
2600 v9fs_string_init(&old_name
);
2601 v9fs_string_init(&new_name
);
2602 err
= pdu_unmarshal(pdu
, offset
, "dsds", &olddirfid
,
2603 &old_name
, &newdirfid
, &new_name
);
2608 v9fs_path_write_lock(s
);
2609 err
= v9fs_complete_renameat(pdu
, olddirfid
,
2610 &old_name
, newdirfid
, &new_name
);
2611 v9fs_path_unlock(s
);
2617 pdu_complete(pdu
, err
);
2618 v9fs_string_free(&old_name
);
2619 v9fs_string_free(&new_name
);
2622 static void v9fs_wstat(void *opaque
)
2631 V9fsPDU
*pdu
= opaque
;
2633 v9fs_stat_init(&v9stat
);
2634 err
= pdu_unmarshal(pdu
, offset
, "dwS", &fid
, &unused
, &v9stat
);
2638 trace_v9fs_wstat(pdu
->tag
, pdu
->id
, fid
,
2639 v9stat
.mode
, v9stat
.atime
, v9stat
.mtime
);
2641 fidp
= get_fid(pdu
, fid
);
2646 /* do we need to sync the file? */
2647 if (donttouch_stat(&v9stat
)) {
2648 err
= v9fs_co_fsync(pdu
, fidp
, 0);
2651 if (v9stat
.mode
!= -1) {
2653 err
= v9fs_co_lstat(pdu
, &fidp
->path
, &stbuf
);
2657 v9_mode
= stat_to_v9mode(&stbuf
);
2658 if ((v9stat
.mode
& P9_STAT_MODE_TYPE_BITS
) !=
2659 (v9_mode
& P9_STAT_MODE_TYPE_BITS
)) {
2660 /* Attempting to change the type */
2664 err
= v9fs_co_chmod(pdu
, &fidp
->path
,
2665 v9mode_to_mode(v9stat
.mode
,
2666 &v9stat
.extension
));
2671 if (v9stat
.mtime
!= -1 || v9stat
.atime
!= -1) {
2672 struct timespec times
[2];
2673 if (v9stat
.atime
!= -1) {
2674 times
[0].tv_sec
= v9stat
.atime
;
2675 times
[0].tv_nsec
= 0;
2677 times
[0].tv_nsec
= UTIME_OMIT
;
2679 if (v9stat
.mtime
!= -1) {
2680 times
[1].tv_sec
= v9stat
.mtime
;
2681 times
[1].tv_nsec
= 0;
2683 times
[1].tv_nsec
= UTIME_OMIT
;
2685 err
= v9fs_co_utimensat(pdu
, &fidp
->path
, times
);
2690 if (v9stat
.n_gid
!= -1 || v9stat
.n_uid
!= -1) {
2691 err
= v9fs_co_chown(pdu
, &fidp
->path
, v9stat
.n_uid
, v9stat
.n_gid
);
2696 if (v9stat
.name
.size
!= 0) {
2697 err
= v9fs_complete_rename(pdu
, fidp
, -1, &v9stat
.name
);
2702 if (v9stat
.length
!= -1) {
2703 err
= v9fs_co_truncate(pdu
, &fidp
->path
, v9stat
.length
);
2712 v9fs_stat_free(&v9stat
);
2713 pdu_complete(pdu
, err
);
2716 static int v9fs_fill_statfs(V9fsState
*s
, V9fsPDU
*pdu
, struct statfs
*stbuf
)
2728 int32_t bsize_factor
;
2731 * compute bsize factor based on host file system block size
2734 bsize_factor
= (s
->msize
- P9_IOHDRSZ
)/stbuf
->f_bsize
;
2735 if (!bsize_factor
) {
2738 f_type
= stbuf
->f_type
;
2739 f_bsize
= stbuf
->f_bsize
;
2740 f_bsize
*= bsize_factor
;
2742 * f_bsize is adjusted(multiplied) by bsize factor, so we need to
2743 * adjust(divide) the number of blocks, free blocks and available
2744 * blocks by bsize factor
2746 f_blocks
= stbuf
->f_blocks
/bsize_factor
;
2747 f_bfree
= stbuf
->f_bfree
/bsize_factor
;
2748 f_bavail
= stbuf
->f_bavail
/bsize_factor
;
2749 f_files
= stbuf
->f_files
;
2750 f_ffree
= stbuf
->f_ffree
;
2751 fsid_val
= (unsigned int) stbuf
->f_fsid
.__val
[0] |
2752 (unsigned long long)stbuf
->f_fsid
.__val
[1] << 32;
2753 f_namelen
= stbuf
->f_namelen
;
2755 return pdu_marshal(pdu
, offset
, "ddqqqqqqd",
2756 f_type
, f_bsize
, f_blocks
, f_bfree
,
2757 f_bavail
, f_files
, f_ffree
,
2758 fsid_val
, f_namelen
);
2761 static void v9fs_statfs(void *opaque
)
2767 struct statfs stbuf
;
2768 V9fsPDU
*pdu
= opaque
;
2769 V9fsState
*s
= pdu
->s
;
2771 retval
= pdu_unmarshal(pdu
, offset
, "d", &fid
);
2775 fidp
= get_fid(pdu
, fid
);
2780 retval
= v9fs_co_statfs(pdu
, &fidp
->path
, &stbuf
);
2784 retval
= v9fs_fill_statfs(s
, pdu
, &stbuf
);
2792 pdu_complete(pdu
, retval
);
2795 static void v9fs_mknod(void *opaque
)
2808 V9fsPDU
*pdu
= opaque
;
2810 v9fs_string_init(&name
);
2811 err
= pdu_unmarshal(pdu
, offset
, "dsdddd", &fid
, &name
, &mode
,
2812 &major
, &minor
, &gid
);
2816 trace_v9fs_mknod(pdu
->tag
, pdu
->id
, fid
, mode
, major
, minor
);
2818 fidp
= get_fid(pdu
, fid
);
2823 err
= v9fs_co_mknod(pdu
, fidp
, &name
, fidp
->uid
, gid
,
2824 makedev(major
, minor
), mode
, &stbuf
);
2828 stat_to_qid(&stbuf
, &qid
);
2829 err
= pdu_marshal(pdu
, offset
, "Q", &qid
);
2834 trace_v9fs_mknod_return(pdu
->tag
, pdu
->id
,
2835 qid
.type
, qid
.version
, qid
.path
);
2839 pdu_complete(pdu
, err
);
2840 v9fs_string_free(&name
);
2844 * Implement posix byte range locking code
2845 * Server side handling of locking code is very simple, because 9p server in
2846 * QEMU can handle only one client. And most of the lock handling
2847 * (like conflict, merging) etc is done by the VFS layer itself, so no need to
2848 * do any thing in * qemu 9p server side lock code path.
2849 * So when a TLOCK request comes, always return success
2851 static void v9fs_lock(void *opaque
)
2858 int32_t fid
, err
= 0;
2859 V9fsPDU
*pdu
= opaque
;
2861 status
= P9_LOCK_ERROR
;
2862 v9fs_string_init(&flock
.client_id
);
2863 err
= pdu_unmarshal(pdu
, offset
, "dbdqqds", &fid
, &flock
.type
,
2864 &flock
.flags
, &flock
.start
, &flock
.length
,
2865 &flock
.proc_id
, &flock
.client_id
);
2869 trace_v9fs_lock(pdu
->tag
, pdu
->id
, fid
,
2870 flock
.type
, flock
.start
, flock
.length
);
2873 /* We support only block flag now (that too ignored currently) */
2874 if (flock
.flags
& ~P9_LOCK_FLAGS_BLOCK
) {
2878 fidp
= get_fid(pdu
, fid
);
2883 err
= v9fs_co_fstat(pdu
, fidp
, &stbuf
);
2887 status
= P9_LOCK_SUCCESS
;
2891 err
= pdu_marshal(pdu
, offset
, "b", status
);
2895 trace_v9fs_lock_return(pdu
->tag
, pdu
->id
, status
);
2896 pdu_complete(pdu
, err
);
2897 v9fs_string_free(&flock
.client_id
);
2901 * When a TGETLOCK request comes, always return success because all lock
2902 * handling is done by client's VFS layer.
2904 static void v9fs_getlock(void *opaque
)
2910 int32_t fid
, err
= 0;
2911 V9fsPDU
*pdu
= opaque
;
2913 v9fs_string_init(&glock
.client_id
);
2914 err
= pdu_unmarshal(pdu
, offset
, "dbqqds", &fid
, &glock
.type
,
2915 &glock
.start
, &glock
.length
, &glock
.proc_id
,
2920 trace_v9fs_getlock(pdu
->tag
, pdu
->id
, fid
,
2921 glock
.type
, glock
.start
, glock
.length
);
2923 fidp
= get_fid(pdu
, fid
);
2928 err
= v9fs_co_fstat(pdu
, fidp
, &stbuf
);
2932 glock
.type
= P9_LOCK_TYPE_UNLCK
;
2933 err
= pdu_marshal(pdu
, offset
, "bqqds", glock
.type
,
2934 glock
.start
, glock
.length
, glock
.proc_id
,
2940 trace_v9fs_getlock_return(pdu
->tag
, pdu
->id
, glock
.type
, glock
.start
,
2941 glock
.length
, glock
.proc_id
);
2945 pdu_complete(pdu
, err
);
2946 v9fs_string_free(&glock
.client_id
);
2949 static void v9fs_mkdir(void *opaque
)
2951 V9fsPDU
*pdu
= opaque
;
2962 v9fs_string_init(&name
);
2963 err
= pdu_unmarshal(pdu
, offset
, "dsdd", &fid
, &name
, &mode
, &gid
);
2967 trace_v9fs_mkdir(pdu
->tag
, pdu
->id
, fid
, name
.data
, mode
, gid
);
2969 fidp
= get_fid(pdu
, fid
);
2974 err
= v9fs_co_mkdir(pdu
, fidp
, &name
, mode
, fidp
->uid
, gid
, &stbuf
);
2978 stat_to_qid(&stbuf
, &qid
);
2979 err
= pdu_marshal(pdu
, offset
, "Q", &qid
);
2984 trace_v9fs_mkdir_return(pdu
->tag
, pdu
->id
,
2985 qid
.type
, qid
.version
, qid
.path
, err
);
2989 pdu_complete(pdu
, err
);
2990 v9fs_string_free(&name
);
2993 static void v9fs_xattrwalk(void *opaque
)
2999 int32_t fid
, newfid
;
3000 V9fsFidState
*file_fidp
;
3001 V9fsFidState
*xattr_fidp
= NULL
;
3002 V9fsPDU
*pdu
= opaque
;
3003 V9fsState
*s
= pdu
->s
;
3005 v9fs_string_init(&name
);
3006 err
= pdu_unmarshal(pdu
, offset
, "dds", &fid
, &newfid
, &name
);
3010 trace_v9fs_xattrwalk(pdu
->tag
, pdu
->id
, fid
, newfid
, name
.data
);
3012 file_fidp
= get_fid(pdu
, fid
);
3013 if (file_fidp
== NULL
) {
3017 xattr_fidp
= alloc_fid(s
, newfid
);
3018 if (xattr_fidp
== NULL
) {
3022 v9fs_path_copy(&xattr_fidp
->path
, &file_fidp
->path
);
3023 if (name
.data
== NULL
) {
3025 * listxattr request. Get the size first
3027 size
= v9fs_co_llistxattr(pdu
, &xattr_fidp
->path
, NULL
, 0);
3030 clunk_fid(s
, xattr_fidp
->fid
);
3034 * Read the xattr value
3036 xattr_fidp
->fs
.xattr
.len
= size
;
3037 xattr_fidp
->fid_type
= P9_FID_XATTR
;
3038 xattr_fidp
->fs
.xattr
.copied_len
= -1;
3040 xattr_fidp
->fs
.xattr
.value
= g_malloc(size
);
3041 err
= v9fs_co_llistxattr(pdu
, &xattr_fidp
->path
,
3042 xattr_fidp
->fs
.xattr
.value
,
3043 xattr_fidp
->fs
.xattr
.len
);
3045 clunk_fid(s
, xattr_fidp
->fid
);
3049 err
= pdu_marshal(pdu
, offset
, "q", size
);
3056 * specific xattr fid. We check for xattr
3057 * presence also collect the xattr size
3059 size
= v9fs_co_lgetxattr(pdu
, &xattr_fidp
->path
,
3063 clunk_fid(s
, xattr_fidp
->fid
);
3067 * Read the xattr value
3069 xattr_fidp
->fs
.xattr
.len
= size
;
3070 xattr_fidp
->fid_type
= P9_FID_XATTR
;
3071 xattr_fidp
->fs
.xattr
.copied_len
= -1;
3073 xattr_fidp
->fs
.xattr
.value
= g_malloc(size
);
3074 err
= v9fs_co_lgetxattr(pdu
, &xattr_fidp
->path
,
3075 &name
, xattr_fidp
->fs
.xattr
.value
,
3076 xattr_fidp
->fs
.xattr
.len
);
3078 clunk_fid(s
, xattr_fidp
->fid
);
3082 err
= pdu_marshal(pdu
, offset
, "q", size
);
3088 trace_v9fs_xattrwalk_return(pdu
->tag
, pdu
->id
, size
);
3090 put_fid(pdu
, file_fidp
);
3092 put_fid(pdu
, xattr_fidp
);
3095 pdu_complete(pdu
, err
);
3096 v9fs_string_free(&name
);
3099 static void v9fs_xattrcreate(void *opaque
)
3107 V9fsFidState
*file_fidp
;
3108 V9fsFidState
*xattr_fidp
;
3109 V9fsPDU
*pdu
= opaque
;
3111 v9fs_string_init(&name
);
3112 err
= pdu_unmarshal(pdu
, offset
, "dsqd", &fid
, &name
, &size
, &flags
);
3116 trace_v9fs_xattrcreate(pdu
->tag
, pdu
->id
, fid
, name
.data
, size
, flags
);
3118 file_fidp
= get_fid(pdu
, fid
);
3119 if (file_fidp
== NULL
) {
3123 /* Make the file fid point to xattr */
3124 xattr_fidp
= file_fidp
;
3125 xattr_fidp
->fid_type
= P9_FID_XATTR
;
3126 xattr_fidp
->fs
.xattr
.copied_len
= 0;
3127 xattr_fidp
->fs
.xattr
.len
= size
;
3128 xattr_fidp
->fs
.xattr
.flags
= flags
;
3129 v9fs_string_init(&xattr_fidp
->fs
.xattr
.name
);
3130 v9fs_string_copy(&xattr_fidp
->fs
.xattr
.name
, &name
);
3131 xattr_fidp
->fs
.xattr
.value
= g_malloc(size
);
3133 put_fid(pdu
, file_fidp
);
3135 pdu_complete(pdu
, err
);
3136 v9fs_string_free(&name
);
3139 static void v9fs_readlink(void *opaque
)
3141 V9fsPDU
*pdu
= opaque
;
3148 err
= pdu_unmarshal(pdu
, offset
, "d", &fid
);
3152 trace_v9fs_readlink(pdu
->tag
, pdu
->id
, fid
);
3153 fidp
= get_fid(pdu
, fid
);
3159 v9fs_string_init(&target
);
3160 err
= v9fs_co_readlink(pdu
, &fidp
->path
, &target
);
3164 err
= pdu_marshal(pdu
, offset
, "s", &target
);
3166 v9fs_string_free(&target
);
3170 trace_v9fs_readlink_return(pdu
->tag
, pdu
->id
, target
.data
);
3171 v9fs_string_free(&target
);
3175 pdu_complete(pdu
, err
);
3178 static CoroutineEntry
*pdu_co_handlers
[] = {
3179 [P9_TREADDIR
] = v9fs_readdir
,
3180 [P9_TSTATFS
] = v9fs_statfs
,
3181 [P9_TGETATTR
] = v9fs_getattr
,
3182 [P9_TSETATTR
] = v9fs_setattr
,
3183 [P9_TXATTRWALK
] = v9fs_xattrwalk
,
3184 [P9_TXATTRCREATE
] = v9fs_xattrcreate
,
3185 [P9_TMKNOD
] = v9fs_mknod
,
3186 [P9_TRENAME
] = v9fs_rename
,
3187 [P9_TLOCK
] = v9fs_lock
,
3188 [P9_TGETLOCK
] = v9fs_getlock
,
3189 [P9_TRENAMEAT
] = v9fs_renameat
,
3190 [P9_TREADLINK
] = v9fs_readlink
,
3191 [P9_TUNLINKAT
] = v9fs_unlinkat
,
3192 [P9_TMKDIR
] = v9fs_mkdir
,
3193 [P9_TVERSION
] = v9fs_version
,
3194 [P9_TLOPEN
] = v9fs_open
,
3195 [P9_TATTACH
] = v9fs_attach
,
3196 [P9_TSTAT
] = v9fs_stat
,
3197 [P9_TWALK
] = v9fs_walk
,
3198 [P9_TCLUNK
] = v9fs_clunk
,
3199 [P9_TFSYNC
] = v9fs_fsync
,
3200 [P9_TOPEN
] = v9fs_open
,
3201 [P9_TREAD
] = v9fs_read
,
3203 [P9_TAUTH
] = v9fs_auth
,
3205 [P9_TFLUSH
] = v9fs_flush
,
3206 [P9_TLINK
] = v9fs_link
,
3207 [P9_TSYMLINK
] = v9fs_symlink
,
3208 [P9_TCREATE
] = v9fs_create
,
3209 [P9_TLCREATE
] = v9fs_lcreate
,
3210 [P9_TWRITE
] = v9fs_write
,
3211 [P9_TWSTAT
] = v9fs_wstat
,
3212 [P9_TREMOVE
] = v9fs_remove
,
3215 static void v9fs_op_not_supp(void *opaque
)
3217 V9fsPDU
*pdu
= opaque
;
3218 pdu_complete(pdu
, -EOPNOTSUPP
);
3221 static void v9fs_fs_ro(void *opaque
)
3223 V9fsPDU
*pdu
= opaque
;
3224 pdu_complete(pdu
, -EROFS
);
3227 static inline bool is_read_only_op(V9fsPDU
*pdu
)
3254 void pdu_submit(V9fsPDU
*pdu
)
3257 CoroutineEntry
*handler
;
3258 V9fsState
*s
= pdu
->s
;
3260 if (pdu
->id
>= ARRAY_SIZE(pdu_co_handlers
) ||
3261 (pdu_co_handlers
[pdu
->id
] == NULL
)) {
3262 handler
= v9fs_op_not_supp
;
3264 handler
= pdu_co_handlers
[pdu
->id
];
3267 if (is_ro_export(&s
->ctx
) && !is_read_only_op(pdu
)) {
3268 handler
= v9fs_fs_ro
;
3270 co
= qemu_coroutine_create(handler
);
3271 qemu_coroutine_enter(co
, pdu
);
3274 /* Returns 0 on success, 1 on failure. */
3275 int v9fs_device_realize_common(V9fsState
*s
, Error
**errp
)
3277 V9fsVirtioState
*v
= container_of(s
, V9fsVirtioState
, state
);
3284 /* initialize pdu allocator */
3285 QLIST_INIT(&s
->free_list
);
3286 QLIST_INIT(&s
->active_list
);
3287 for (i
= 0; i
< (MAX_REQ
- 1); i
++) {
3288 QLIST_INSERT_HEAD(&s
->free_list
, &v
->pdus
[i
], next
);
3293 v9fs_path_init(&path
);
3295 fse
= get_fsdev_fsentry(s
->fsconf
.fsdev_id
);
3298 /* We don't have a fsdev identified by fsdev_id */
3299 error_setg(errp
, "9pfs device couldn't find fsdev with the "
3301 s
->fsconf
.fsdev_id
? s
->fsconf
.fsdev_id
: "NULL");
3305 if (!s
->fsconf
.tag
) {
3306 /* we haven't specified a mount_tag */
3307 error_setg(errp
, "fsdev with id %s needs mount_tag arguments",
3308 s
->fsconf
.fsdev_id
);
3312 s
->ctx
.export_flags
= fse
->export_flags
;
3313 s
->ctx
.fs_root
= g_strdup(fse
->path
);
3314 s
->ctx
.exops
.get_st_gen
= NULL
;
3315 len
= strlen(s
->fsconf
.tag
);
3316 if (len
> MAX_TAG_LEN
- 1) {
3317 error_setg(errp
, "mount tag '%s' (%d bytes) is longer than "
3318 "maximum (%d bytes)", s
->fsconf
.tag
, len
, MAX_TAG_LEN
- 1);
3322 s
->tag
= g_strdup(s
->fsconf
.tag
);
3328 qemu_co_rwlock_init(&s
->rename_lock
);
3330 if (s
->ops
->init(&s
->ctx
) < 0) {
3331 error_setg(errp
, "9pfs Failed to initialize fs-driver with id:%s"
3332 " and export path:%s", s
->fsconf
.fsdev_id
, s
->ctx
.fs_root
);
3337 * Check details of export path, We need to use fs driver
3338 * call back to do that. Since we are in the init path, we don't
3339 * use co-routines here.
3341 if (s
->ops
->name_to_path(&s
->ctx
, NULL
, "/", &path
) < 0) {
3343 "error in converting name to path %s", strerror(errno
));
3346 if (s
->ops
->lstat(&s
->ctx
, &path
, &stat
)) {
3347 error_setg(errp
, "share path %s does not exist", fse
->path
);
3349 } else if (!S_ISDIR(stat
.st_mode
)) {
3350 error_setg(errp
, "share path %s is not a directory", fse
->path
);
3353 v9fs_path_free(&path
);
3358 g_free(s
->ctx
.fs_root
);
3360 v9fs_path_free(&path
);
3365 void v9fs_device_unrealize_common(V9fsState
*s
, Error
**errp
)
3367 g_free(s
->ctx
.fs_root
);
3371 static void __attribute__((__constructor__
)) v9fs_set_fd_limit(void)
3374 if (getrlimit(RLIMIT_NOFILE
, &rlim
) < 0) {
3375 error_report("Failed to get the resource limit");
3378 open_fd_hw
= rlim
.rlim_cur
- MIN(400, rlim
.rlim_cur
/3);
3379 open_fd_rc
= rlim
.rlim_cur
/2;