xref: /openbmc/qemu/hw/9pfs/9p-proxy.c (revision 89fc45d5)
1 /*
2  * 9p Proxy callback
3  *
4  * Copyright IBM, Corp. 2011
5  *
6  * Authors:
7  * M. Mohan Kumar <mohan@in.ibm.com>
8  *
9  * This work is licensed under the terms of the GNU GPL, version 2.  See
10  * the COPYING file in the top-level directory.
11  */
12 
13 /*
14  * Not so fast! You might want to read the 9p developer docs first:
15  * https://wiki.qemu.org/Documentation/9p
16  */
17 
18 #include "qemu/osdep.h"
19 #include <sys/socket.h>
20 #include <sys/un.h>
21 #include "qemu-common.h"
22 #include "9p.h"
23 #include "qapi/error.h"
24 #include "qemu/cutils.h"
25 #include "qemu/error-report.h"
26 #include "qemu/option.h"
27 #include "fsdev/qemu-fsdev.h"
28 #include "9p-proxy.h"
29 
30 typedef struct V9fsProxy {
31     int sockfd;
32     QemuMutex mutex;
33     struct iovec in_iovec;
34     struct iovec out_iovec;
35 } V9fsProxy;
36 
37 /*
38  * Return received file descriptor on success in *status.
39  * errno is also returned on *status (which will be < 0)
40  * return < 0 on transport error.
41  */
42 static int v9fs_receivefd(int sockfd, int *status)
43 {
44     struct iovec iov;
45     struct msghdr msg;
46     struct cmsghdr *cmsg;
47     int retval, data, fd;
48     union MsgControl msg_control;
49 
50     iov.iov_base = &data;
51     iov.iov_len = sizeof(data);
52 
53     memset(&msg, 0, sizeof(msg));
54     msg.msg_iov = &iov;
55     msg.msg_iovlen = 1;
56     msg.msg_control = &msg_control;
57     msg.msg_controllen = sizeof(msg_control);
58 
59     do {
60         retval = recvmsg(sockfd, &msg, 0);
61     } while (retval < 0 && errno == EINTR);
62     if (retval <= 0) {
63         return retval;
64     }
65     /*
66      * data is set to V9FS_FD_VALID, if ancillary data is sent.  If this
67      * request doesn't need ancillary data (fd) or an error occurred,
68      * data is set to negative errno value.
69      */
70     if (data != V9FS_FD_VALID) {
71         *status = data;
72         return 0;
73     }
74     /*
75      * File descriptor (fd) is sent in the ancillary data. Check if we
76      * indeed received it. One of the reasons to fail to receive it is if
77      * we exceeded the maximum number of file descriptors!
78      */
79     for (cmsg = CMSG_FIRSTHDR(&msg); cmsg; cmsg = CMSG_NXTHDR(&msg, cmsg)) {
80         if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)) ||
81             cmsg->cmsg_level != SOL_SOCKET ||
82             cmsg->cmsg_type != SCM_RIGHTS) {
83             continue;
84         }
85         fd = *((int *)CMSG_DATA(cmsg));
86         *status = fd;
87         return 0;
88     }
89     *status = -ENFILE;  /* Ancillary data sent but not received */
90     return 0;
91 }
92 
93 static ssize_t socket_read(int sockfd, void *buff, size_t size)
94 {
95     ssize_t retval, total = 0;
96 
97     while (size) {
98         retval = read(sockfd, buff, size);
99         if (retval == 0) {
100             return -EIO;
101         }
102         if (retval < 0) {
103             if (errno == EINTR) {
104                 continue;
105             }
106             return -errno;
107         }
108         size -= retval;
109         buff += retval;
110         total += retval;
111     }
112     return total;
113 }
114 
115 /* Converts proxy_statfs to VFS statfs structure */
116 static void prstatfs_to_statfs(struct statfs *stfs, ProxyStatFS *prstfs)
117 {
118     memset(stfs, 0, sizeof(*stfs));
119     stfs->f_type = prstfs->f_type;
120     stfs->f_bsize = prstfs->f_bsize;
121     stfs->f_blocks = prstfs->f_blocks;
122     stfs->f_bfree = prstfs->f_bfree;
123     stfs->f_bavail = prstfs->f_bavail;
124     stfs->f_files = prstfs->f_files;
125     stfs->f_ffree = prstfs->f_ffree;
126 #ifdef CONFIG_DARWIN
127     /* f_namelen and f_frsize do not exist on Darwin */
128     stfs->f_fsid.val[0] = prstfs->f_fsid[0] & 0xFFFFFFFFU;
129     stfs->f_fsid.val[1] = prstfs->f_fsid[1] >> 32 & 0xFFFFFFFFU;
130 #else
131     stfs->f_fsid.__val[0] = prstfs->f_fsid[0] & 0xFFFFFFFFU;
132     stfs->f_fsid.__val[1] = prstfs->f_fsid[1] >> 32 & 0xFFFFFFFFU;
133     stfs->f_namelen = prstfs->f_namelen;
134     stfs->f_frsize = prstfs->f_frsize;
135 #endif
136 }
137 
138 /* Converts proxy_stat structure to VFS stat structure */
139 static void prstat_to_stat(struct stat *stbuf, ProxyStat *prstat)
140 {
141    memset(stbuf, 0, sizeof(*stbuf));
142    stbuf->st_dev = prstat->st_dev;
143    stbuf->st_ino = prstat->st_ino;
144    stbuf->st_nlink = prstat->st_nlink;
145    stbuf->st_mode = prstat->st_mode;
146    stbuf->st_uid = prstat->st_uid;
147    stbuf->st_gid = prstat->st_gid;
148    stbuf->st_rdev = prstat->st_rdev;
149    stbuf->st_size = prstat->st_size;
150    stbuf->st_blksize = prstat->st_blksize;
151    stbuf->st_blocks = prstat->st_blocks;
152    stbuf->st_atime = prstat->st_atim_sec;
153    stbuf->st_mtime = prstat->st_mtim_sec;
154    stbuf->st_ctime = prstat->st_ctim_sec;
155 #ifdef CONFIG_DARWIN
156    stbuf->st_atimespec.tv_sec = prstat->st_atim_sec;
157    stbuf->st_mtimespec.tv_sec = prstat->st_mtim_sec;
158    stbuf->st_ctimespec.tv_sec = prstat->st_ctim_sec;
159    stbuf->st_atimespec.tv_nsec = prstat->st_atim_nsec;
160    stbuf->st_mtimespec.tv_nsec = prstat->st_mtim_nsec;
161    stbuf->st_ctimespec.tv_nsec = prstat->st_ctim_nsec;
162 #else
163    stbuf->st_atim.tv_sec = prstat->st_atim_sec;
164    stbuf->st_mtim.tv_sec = prstat->st_mtim_sec;
165    stbuf->st_ctim.tv_sec = prstat->st_ctim_sec;
166    stbuf->st_atim.tv_nsec = prstat->st_atim_nsec;
167    stbuf->st_mtim.tv_nsec = prstat->st_mtim_nsec;
168    stbuf->st_ctim.tv_nsec = prstat->st_ctim_nsec;
169 #endif
170 }
171 
172 /*
173  * Response contains two parts
174  * {header, data}
175  * header.type == T_ERROR, data -> -errno
176  * header.type == T_SUCCESS, data -> response
177  * size of errno/response is given by header.size
178  * returns < 0, on transport error. response is
179  * valid only if status >= 0.
180  */
181 static int v9fs_receive_response(V9fsProxy *proxy, int type,
182                                  int *status, void *response)
183 {
184     int retval;
185     ProxyHeader header;
186     struct iovec *reply = &proxy->in_iovec;
187 
188     *status = 0;
189     reply->iov_len = 0;
190     retval = socket_read(proxy->sockfd, reply->iov_base, PROXY_HDR_SZ);
191     if (retval < 0) {
192         return retval;
193     }
194     reply->iov_len = PROXY_HDR_SZ;
195     retval = proxy_unmarshal(reply, 0, "dd", &header.type, &header.size);
196     assert(retval == 4 * 2);
197     /*
198      * if response size > PROXY_MAX_IO_SZ, read the response but ignore it and
199      * return -ENOBUFS
200      */
201     if (header.size > PROXY_MAX_IO_SZ) {
202         int count;
203         while (header.size > 0) {
204             count = MIN(PROXY_MAX_IO_SZ, header.size);
205             count = socket_read(proxy->sockfd, reply->iov_base, count);
206             if (count < 0) {
207                 return count;
208             }
209             header.size -= count;
210         }
211         *status = -ENOBUFS;
212         return 0;
213     }
214 
215     retval = socket_read(proxy->sockfd,
216                          reply->iov_base + PROXY_HDR_SZ, header.size);
217     if (retval < 0) {
218         return retval;
219     }
220     reply->iov_len += header.size;
221     /* there was an error during processing request */
222     if (header.type == T_ERROR) {
223         int ret;
224         ret = proxy_unmarshal(reply, PROXY_HDR_SZ, "d", status);
225         assert(ret == 4);
226         return 0;
227     }
228 
229     switch (type) {
230     case T_LSTAT: {
231         ProxyStat prstat;
232         retval = proxy_unmarshal(reply, PROXY_HDR_SZ,
233                                  "qqqdddqqqqqqqqqq", &prstat.st_dev,
234                                  &prstat.st_ino, &prstat.st_nlink,
235                                  &prstat.st_mode, &prstat.st_uid,
236                                  &prstat.st_gid, &prstat.st_rdev,
237                                  &prstat.st_size, &prstat.st_blksize,
238                                  &prstat.st_blocks,
239                                  &prstat.st_atim_sec, &prstat.st_atim_nsec,
240                                  &prstat.st_mtim_sec, &prstat.st_mtim_nsec,
241                                  &prstat.st_ctim_sec, &prstat.st_ctim_nsec);
242         assert(retval == 8 * 3 + 4 * 3 + 8 * 10);
243         prstat_to_stat(response, &prstat);
244         break;
245     }
246     case T_STATFS: {
247         ProxyStatFS prstfs;
248         retval = proxy_unmarshal(reply, PROXY_HDR_SZ,
249                                  "qqqqqqqqqqq", &prstfs.f_type,
250                                  &prstfs.f_bsize, &prstfs.f_blocks,
251                                  &prstfs.f_bfree, &prstfs.f_bavail,
252                                  &prstfs.f_files, &prstfs.f_ffree,
253                                  &prstfs.f_fsid[0], &prstfs.f_fsid[1],
254                                  &prstfs.f_namelen, &prstfs.f_frsize);
255         assert(retval == 8 * 11);
256         prstatfs_to_statfs(response, &prstfs);
257         break;
258     }
259     case T_READLINK: {
260         V9fsString target;
261         v9fs_string_init(&target);
262         retval = proxy_unmarshal(reply, PROXY_HDR_SZ, "s", &target);
263         strcpy(response, target.data);
264         v9fs_string_free(&target);
265         break;
266     }
267     case T_LGETXATTR:
268     case T_LLISTXATTR: {
269         V9fsString xattr;
270         v9fs_string_init(&xattr);
271         retval = proxy_unmarshal(reply, PROXY_HDR_SZ, "s", &xattr);
272         memcpy(response, xattr.data, xattr.size);
273         v9fs_string_free(&xattr);
274         break;
275     }
276     case T_GETVERSION:
277         retval = proxy_unmarshal(reply, PROXY_HDR_SZ, "q", response);
278         assert(retval == 8);
279         break;
280     default:
281         return -1;
282     }
283     if (retval < 0) {
284         *status  = retval;
285     }
286     return 0;
287 }
288 
289 /*
290  * return < 0 on transport error.
291  * *status is valid only if return >= 0
292  */
293 static int v9fs_receive_status(V9fsProxy *proxy,
294                                struct iovec *reply, int *status)
295 {
296     int retval;
297     ProxyHeader header;
298 
299     *status = 0;
300     reply->iov_len = 0;
301     retval = socket_read(proxy->sockfd, reply->iov_base, PROXY_HDR_SZ);
302     if (retval < 0) {
303         return retval;
304     }
305     reply->iov_len = PROXY_HDR_SZ;
306     retval = proxy_unmarshal(reply, 0, "dd", &header.type, &header.size);
307     assert(retval == 4 * 2);
308     retval = socket_read(proxy->sockfd,
309                          reply->iov_base + PROXY_HDR_SZ, header.size);
310     if (retval < 0) {
311         return retval;
312     }
313     reply->iov_len += header.size;
314     retval = proxy_unmarshal(reply, PROXY_HDR_SZ, "d", status);
315     assert(retval == 4);
316     return 0;
317 }
318 
319 /*
320  * Proxy->header and proxy->request written to socket by QEMU process.
321  * This request read by proxy helper process
322  * returns 0 on success and -errno on error
323  */
324 static int v9fs_request(V9fsProxy *proxy, int type, void *response, ...)
325 {
326     dev_t rdev;
327     va_list ap;
328     int size = 0;
329     int retval = 0;
330     uint64_t offset;
331     ProxyHeader header = { 0, 0};
332     struct timespec spec[2];
333     int flags, mode, uid, gid;
334     V9fsString *name, *value;
335     V9fsString *path, *oldpath;
336     struct iovec *iovec = NULL, *reply = NULL;
337 
338     qemu_mutex_lock(&proxy->mutex);
339 
340     if (proxy->sockfd == -1) {
341         retval = -EIO;
342         goto err_out;
343     }
344     iovec = &proxy->out_iovec;
345     reply = &proxy->in_iovec;
346     va_start(ap, response);
347     switch (type) {
348     case T_OPEN:
349         path = va_arg(ap, V9fsString *);
350         flags = va_arg(ap, int);
351         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sd", path, flags);
352         if (retval > 0) {
353             header.size = retval;
354             header.type = T_OPEN;
355         }
356         break;
357     case T_CREATE:
358         path = va_arg(ap, V9fsString *);
359         flags = va_arg(ap, int);
360         mode = va_arg(ap, int);
361         uid = va_arg(ap, int);
362         gid = va_arg(ap, int);
363         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sdddd", path,
364                                     flags, mode, uid, gid);
365         if (retval > 0) {
366             header.size = retval;
367             header.type = T_CREATE;
368         }
369         break;
370     case T_MKNOD:
371         path = va_arg(ap, V9fsString *);
372         mode = va_arg(ap, int);
373         rdev = va_arg(ap, long int);
374         uid = va_arg(ap, int);
375         gid = va_arg(ap, int);
376         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ddsdq",
377                                     uid, gid, path, mode, rdev);
378         if (retval > 0) {
379             header.size = retval;
380             header.type = T_MKNOD;
381         }
382         break;
383     case T_MKDIR:
384         path = va_arg(ap, V9fsString *);
385         mode = va_arg(ap, int);
386         uid = va_arg(ap, int);
387         gid = va_arg(ap, int);
388         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ddsd",
389                                     uid, gid, path, mode);
390         if (retval > 0) {
391             header.size = retval;
392             header.type = T_MKDIR;
393         }
394         break;
395     case T_SYMLINK:
396         oldpath = va_arg(ap, V9fsString *);
397         path = va_arg(ap, V9fsString *);
398         uid = va_arg(ap, int);
399         gid = va_arg(ap, int);
400         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ddss",
401                                     uid, gid, oldpath, path);
402         if (retval > 0) {
403             header.size = retval;
404             header.type = T_SYMLINK;
405         }
406         break;
407     case T_LINK:
408         oldpath = va_arg(ap, V9fsString *);
409         path = va_arg(ap, V9fsString *);
410         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ss",
411                                     oldpath, path);
412         if (retval > 0) {
413             header.size = retval;
414             header.type = T_LINK;
415         }
416         break;
417     case T_LSTAT:
418         path = va_arg(ap, V9fsString *);
419         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "s", path);
420         if (retval > 0) {
421             header.size = retval;
422             header.type = T_LSTAT;
423         }
424         break;
425     case T_READLINK:
426         path = va_arg(ap, V9fsString *);
427         size = va_arg(ap, int);
428         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sd", path, size);
429         if (retval > 0) {
430             header.size = retval;
431             header.type = T_READLINK;
432         }
433         break;
434     case T_STATFS:
435         path = va_arg(ap, V9fsString *);
436         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "s", path);
437         if (retval > 0) {
438             header.size = retval;
439             header.type = T_STATFS;
440         }
441         break;
442     case T_CHMOD:
443         path = va_arg(ap, V9fsString *);
444         mode = va_arg(ap, int);
445         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sd", path, mode);
446         if (retval > 0) {
447             header.size = retval;
448             header.type = T_CHMOD;
449         }
450         break;
451     case T_CHOWN:
452         path = va_arg(ap, V9fsString *);
453         uid = va_arg(ap, int);
454         gid = va_arg(ap, int);
455         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sdd", path, uid, gid);
456         if (retval > 0) {
457             header.size = retval;
458             header.type = T_CHOWN;
459         }
460         break;
461     case T_TRUNCATE:
462         path = va_arg(ap, V9fsString *);
463         offset = va_arg(ap, uint64_t);
464         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sq", path, offset);
465         if (retval > 0) {
466             header.size = retval;
467             header.type = T_TRUNCATE;
468         }
469         break;
470     case T_UTIME:
471         path = va_arg(ap, V9fsString *);
472         spec[0].tv_sec = va_arg(ap, long);
473         spec[0].tv_nsec = va_arg(ap, long);
474         spec[1].tv_sec = va_arg(ap, long);
475         spec[1].tv_nsec = va_arg(ap, long);
476         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sqqqq", path,
477                                     spec[0].tv_sec, spec[1].tv_nsec,
478                                     spec[1].tv_sec, spec[1].tv_nsec);
479         if (retval > 0) {
480             header.size = retval;
481             header.type = T_UTIME;
482         }
483         break;
484     case T_RENAME:
485         oldpath = va_arg(ap, V9fsString *);
486         path = va_arg(ap, V9fsString *);
487         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ss", oldpath, path);
488         if (retval > 0) {
489             header.size = retval;
490             header.type = T_RENAME;
491         }
492         break;
493     case T_REMOVE:
494         path = va_arg(ap, V9fsString *);
495         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "s", path);
496         if (retval > 0) {
497             header.size = retval;
498             header.type = T_REMOVE;
499         }
500         break;
501     case T_LGETXATTR:
502         size = va_arg(ap, int);
503         path = va_arg(ap, V9fsString *);
504         name = va_arg(ap, V9fsString *);
505         retval = proxy_marshal(iovec, PROXY_HDR_SZ,
506                                     "dss", size, path, name);
507         if (retval > 0) {
508             header.size = retval;
509             header.type = T_LGETXATTR;
510         }
511         break;
512     case T_LLISTXATTR:
513         size = va_arg(ap, int);
514         path = va_arg(ap, V9fsString *);
515         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ds", size, path);
516         if (retval > 0) {
517             header.size = retval;
518             header.type = T_LLISTXATTR;
519         }
520         break;
521     case T_LSETXATTR:
522         path = va_arg(ap, V9fsString *);
523         name = va_arg(ap, V9fsString *);
524         value = va_arg(ap, V9fsString *);
525         size = va_arg(ap, int);
526         flags = va_arg(ap, int);
527         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sssdd",
528                                     path, name, value, size, flags);
529         if (retval > 0) {
530             header.size = retval;
531             header.type = T_LSETXATTR;
532         }
533         break;
534     case T_LREMOVEXATTR:
535         path = va_arg(ap, V9fsString *);
536         name = va_arg(ap, V9fsString *);
537         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ss", path, name);
538         if (retval > 0) {
539             header.size = retval;
540             header.type = T_LREMOVEXATTR;
541         }
542         break;
543     case T_GETVERSION:
544         path = va_arg(ap, V9fsString *);
545         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "s", path);
546         if (retval > 0) {
547             header.size = retval;
548             header.type = T_GETVERSION;
549         }
550         break;
551     default:
552         error_report("Invalid type %d", type);
553         retval = -EINVAL;
554         break;
555     }
556     va_end(ap);
557 
558     if (retval < 0) {
559         goto err_out;
560     }
561 
562     /* marshal the header details */
563     retval = proxy_marshal(iovec, 0, "dd", header.type, header.size);
564     assert(retval == 4 * 2);
565     header.size += PROXY_HDR_SZ;
566 
567     retval = qemu_write_full(proxy->sockfd, iovec->iov_base, header.size);
568     if (retval != header.size) {
569         goto close_error;
570     }
571 
572     switch (type) {
573     case T_OPEN:
574     case T_CREATE:
575         /*
576          * A file descriptor is returned as response for
577          * T_OPEN,T_CREATE on success
578          */
579         if (v9fs_receivefd(proxy->sockfd, &retval) < 0) {
580             goto close_error;
581         }
582         break;
583     case T_MKNOD:
584     case T_MKDIR:
585     case T_SYMLINK:
586     case T_LINK:
587     case T_CHMOD:
588     case T_CHOWN:
589     case T_RENAME:
590     case T_TRUNCATE:
591     case T_UTIME:
592     case T_REMOVE:
593     case T_LSETXATTR:
594     case T_LREMOVEXATTR:
595         if (v9fs_receive_status(proxy, reply, &retval) < 0) {
596             goto close_error;
597         }
598         break;
599     case T_LSTAT:
600     case T_READLINK:
601     case T_STATFS:
602     case T_GETVERSION:
603         if (v9fs_receive_response(proxy, type, &retval, response) < 0) {
604             goto close_error;
605         }
606         break;
607     case T_LGETXATTR:
608     case T_LLISTXATTR:
609         if (!size) {
610             if (v9fs_receive_status(proxy, reply, &retval) < 0) {
611                 goto close_error;
612             }
613         } else {
614             if (v9fs_receive_response(proxy, type, &retval, response) < 0) {
615                 goto close_error;
616             }
617         }
618         break;
619     }
620 
621 err_out:
622     qemu_mutex_unlock(&proxy->mutex);
623     return retval;
624 
625 close_error:
626     close(proxy->sockfd);
627     proxy->sockfd = -1;
628     qemu_mutex_unlock(&proxy->mutex);
629     return -EIO;
630 }
631 
632 static int proxy_lstat(FsContext *fs_ctx, V9fsPath *fs_path, struct stat *stbuf)
633 {
634     int retval;
635     retval = v9fs_request(fs_ctx->private, T_LSTAT, stbuf, fs_path);
636     if (retval < 0) {
637         errno = -retval;
638         return -1;
639     }
640     return retval;
641 }
642 
643 static ssize_t proxy_readlink(FsContext *fs_ctx, V9fsPath *fs_path,
644                               char *buf, size_t bufsz)
645 {
646     int retval;
647     retval = v9fs_request(fs_ctx->private, T_READLINK, buf, fs_path, bufsz);
648     if (retval < 0) {
649         errno = -retval;
650         return -1;
651     }
652     return strlen(buf);
653 }
654 
655 static int proxy_close(FsContext *ctx, V9fsFidOpenState *fs)
656 {
657     return close(fs->fd);
658 }
659 
660 static int proxy_closedir(FsContext *ctx, V9fsFidOpenState *fs)
661 {
662     return closedir(fs->dir.stream);
663 }
664 
665 static int proxy_open(FsContext *ctx, V9fsPath *fs_path,
666                       int flags, V9fsFidOpenState *fs)
667 {
668     fs->fd = v9fs_request(ctx->private, T_OPEN, NULL, fs_path, flags);
669     if (fs->fd < 0) {
670         errno = -fs->fd;
671         fs->fd = -1;
672     }
673     return fs->fd;
674 }
675 
676 static int proxy_opendir(FsContext *ctx,
677                          V9fsPath *fs_path, V9fsFidOpenState *fs)
678 {
679     int serrno, fd;
680 
681     fs->dir.stream = NULL;
682     fd = v9fs_request(ctx->private, T_OPEN, NULL, fs_path, O_DIRECTORY);
683     if (fd < 0) {
684         errno = -fd;
685         return -1;
686     }
687     fs->dir.stream = fdopendir(fd);
688     if (!fs->dir.stream) {
689         serrno = errno;
690         close(fd);
691         errno = serrno;
692         return -1;
693     }
694     return 0;
695 }
696 
697 static void proxy_rewinddir(FsContext *ctx, V9fsFidOpenState *fs)
698 {
699     rewinddir(fs->dir.stream);
700 }
701 
702 static off_t proxy_telldir(FsContext *ctx, V9fsFidOpenState *fs)
703 {
704     return telldir(fs->dir.stream);
705 }
706 
707 static struct dirent *proxy_readdir(FsContext *ctx, V9fsFidOpenState *fs)
708 {
709     struct dirent *entry;
710     entry = readdir(fs->dir.stream);
711 #ifdef CONFIG_DARWIN
712     if (!entry) {
713         return NULL;
714     }
715     int td;
716     td = telldir(fs->dir.stream);
717     /* If telldir fails, fail the entire readdir call */
718     if (td < 0) {
719         return NULL;
720     }
721     entry->d_seekoff = td;
722 #endif
723     return entry;
724 }
725 
726 static void proxy_seekdir(FsContext *ctx, V9fsFidOpenState *fs, off_t off)
727 {
728     seekdir(fs->dir.stream, off);
729 }
730 
731 static ssize_t proxy_preadv(FsContext *ctx, V9fsFidOpenState *fs,
732                             const struct iovec *iov,
733                             int iovcnt, off_t offset)
734 {
735     ssize_t ret;
736 #ifdef CONFIG_PREADV
737     ret = preadv(fs->fd, iov, iovcnt, offset);
738 #else
739     ret = lseek(fs->fd, offset, SEEK_SET);
740     if (ret >= 0) {
741         ret = readv(fs->fd, iov, iovcnt);
742     }
743 #endif
744     return ret;
745 }
746 
747 static ssize_t proxy_pwritev(FsContext *ctx, V9fsFidOpenState *fs,
748                              const struct iovec *iov,
749                              int iovcnt, off_t offset)
750 {
751     ssize_t ret;
752 
753 #ifdef CONFIG_PREADV
754     ret = pwritev(fs->fd, iov, iovcnt, offset);
755 #else
756     ret = lseek(fs->fd, offset, SEEK_SET);
757     if (ret >= 0) {
758         ret = writev(fs->fd, iov, iovcnt);
759     }
760 #endif
761 #ifdef CONFIG_SYNC_FILE_RANGE
762     if (ret > 0 && ctx->export_flags & V9FS_IMMEDIATE_WRITEOUT) {
763         /*
764          * Initiate a writeback. This is not a data integrity sync.
765          * We want to ensure that we don't leave dirty pages in the cache
766          * after write when writeout=immediate is sepcified.
767          */
768         sync_file_range(fs->fd, offset, ret,
769                         SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE);
770     }
771 #endif
772     return ret;
773 }
774 
775 static int proxy_chmod(FsContext *fs_ctx, V9fsPath *fs_path, FsCred *credp)
776 {
777     int retval;
778     retval = v9fs_request(fs_ctx->private, T_CHMOD, NULL, fs_path,
779                           credp->fc_mode);
780     if (retval < 0) {
781         errno = -retval;
782     }
783     return retval;
784 }
785 
786 static int proxy_mknod(FsContext *fs_ctx, V9fsPath *dir_path,
787                        const char *name, FsCred *credp)
788 {
789     int retval;
790     V9fsString fullname;
791 
792     v9fs_string_init(&fullname);
793     v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
794 
795     retval = v9fs_request(fs_ctx->private, T_MKNOD, NULL, &fullname,
796                           credp->fc_mode, credp->fc_rdev,
797                           credp->fc_uid, credp->fc_gid);
798     v9fs_string_free(&fullname);
799     if (retval < 0) {
800         errno = -retval;
801         retval = -1;
802     }
803     return retval;
804 }
805 
806 static int proxy_mkdir(FsContext *fs_ctx, V9fsPath *dir_path,
807                        const char *name, FsCred *credp)
808 {
809     int retval;
810     V9fsString fullname;
811 
812     v9fs_string_init(&fullname);
813     v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
814 
815     retval = v9fs_request(fs_ctx->private, T_MKDIR, NULL, &fullname,
816                           credp->fc_mode, credp->fc_uid, credp->fc_gid);
817     v9fs_string_free(&fullname);
818     if (retval < 0) {
819         errno = -retval;
820         retval = -1;
821     }
822     return retval;
823 }
824 
825 static int proxy_fstat(FsContext *fs_ctx, int fid_type,
826                        V9fsFidOpenState *fs, struct stat *stbuf)
827 {
828     int fd;
829 
830     if (fid_type == P9_FID_DIR) {
831         fd = dirfd(fs->dir.stream);
832     } else {
833         fd = fs->fd;
834     }
835     return fstat(fd, stbuf);
836 }
837 
838 static int proxy_open2(FsContext *fs_ctx, V9fsPath *dir_path, const char *name,
839                        int flags, FsCred *credp, V9fsFidOpenState *fs)
840 {
841     V9fsString fullname;
842 
843     v9fs_string_init(&fullname);
844     v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
845 
846     fs->fd = v9fs_request(fs_ctx->private, T_CREATE, NULL, &fullname, flags,
847                           credp->fc_mode, credp->fc_uid, credp->fc_gid);
848     v9fs_string_free(&fullname);
849     if (fs->fd < 0) {
850         errno = -fs->fd;
851         fs->fd = -1;
852     }
853     return fs->fd;
854 }
855 
856 static int proxy_symlink(FsContext *fs_ctx, const char *oldpath,
857                          V9fsPath *dir_path, const char *name, FsCred *credp)
858 {
859     int retval;
860     V9fsString fullname, target;
861 
862     v9fs_string_init(&fullname);
863     v9fs_string_init(&target);
864 
865     v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
866     v9fs_string_sprintf(&target, "%s", oldpath);
867 
868     retval = v9fs_request(fs_ctx->private, T_SYMLINK, NULL, &target, &fullname,
869                           credp->fc_uid, credp->fc_gid);
870     v9fs_string_free(&fullname);
871     v9fs_string_free(&target);
872     if (retval < 0) {
873         errno = -retval;
874         retval = -1;
875     }
876     return retval;
877 }
878 
879 static int proxy_link(FsContext *ctx, V9fsPath *oldpath,
880                       V9fsPath *dirpath, const char *name)
881 {
882     int retval;
883     V9fsString newpath;
884 
885     v9fs_string_init(&newpath);
886     v9fs_string_sprintf(&newpath, "%s/%s", dirpath->data, name);
887 
888     retval = v9fs_request(ctx->private, T_LINK, NULL, oldpath, &newpath);
889     v9fs_string_free(&newpath);
890     if (retval < 0) {
891         errno = -retval;
892         retval = -1;
893     }
894     return retval;
895 }
896 
897 static int proxy_truncate(FsContext *ctx, V9fsPath *fs_path, off_t size)
898 {
899     int retval;
900 
901     retval = v9fs_request(ctx->private, T_TRUNCATE, NULL, fs_path, size);
902     if (retval < 0) {
903         errno = -retval;
904         return -1;
905     }
906     return 0;
907 }
908 
909 static int proxy_rename(FsContext *ctx, const char *oldpath,
910                         const char *newpath)
911 {
912     int retval;
913     V9fsString oldname, newname;
914 
915     v9fs_string_init(&oldname);
916     v9fs_string_init(&newname);
917 
918     v9fs_string_sprintf(&oldname, "%s", oldpath);
919     v9fs_string_sprintf(&newname, "%s", newpath);
920     retval = v9fs_request(ctx->private, T_RENAME, NULL, &oldname, &newname);
921     v9fs_string_free(&oldname);
922     v9fs_string_free(&newname);
923     if (retval < 0) {
924         errno = -retval;
925     }
926     return retval;
927 }
928 
929 static int proxy_chown(FsContext *fs_ctx, V9fsPath *fs_path, FsCred *credp)
930 {
931     int retval;
932     retval = v9fs_request(fs_ctx->private, T_CHOWN, NULL, fs_path,
933                           credp->fc_uid, credp->fc_gid);
934     if (retval < 0) {
935         errno = -retval;
936     }
937     return retval;
938 }
939 
940 static int proxy_utimensat(FsContext *s, V9fsPath *fs_path,
941                            const struct timespec *buf)
942 {
943     int retval;
944     retval = v9fs_request(s->private, T_UTIME, NULL, fs_path,
945                           buf[0].tv_sec, buf[0].tv_nsec,
946                           buf[1].tv_sec, buf[1].tv_nsec);
947     if (retval < 0) {
948         errno = -retval;
949     }
950     return retval;
951 }
952 
953 static int proxy_remove(FsContext *ctx, const char *path)
954 {
955     int retval;
956     V9fsString name;
957     v9fs_string_init(&name);
958     v9fs_string_sprintf(&name, "%s", path);
959     retval = v9fs_request(ctx->private, T_REMOVE, NULL, &name);
960     v9fs_string_free(&name);
961     if (retval < 0) {
962         errno = -retval;
963     }
964     return retval;
965 }
966 
967 static int proxy_fsync(FsContext *ctx, int fid_type,
968                        V9fsFidOpenState *fs, int datasync)
969 {
970     int fd;
971 
972     if (fid_type == P9_FID_DIR) {
973         fd = dirfd(fs->dir.stream);
974     } else {
975         fd = fs->fd;
976     }
977 
978     if (datasync) {
979         return qemu_fdatasync(fd);
980     } else {
981         return fsync(fd);
982     }
983 }
984 
985 static int proxy_statfs(FsContext *s, V9fsPath *fs_path, struct statfs *stbuf)
986 {
987     int retval;
988     retval = v9fs_request(s->private, T_STATFS, stbuf, fs_path);
989     if (retval < 0) {
990         errno = -retval;
991         return -1;
992     }
993     return retval;
994 }
995 
996 static ssize_t proxy_lgetxattr(FsContext *ctx, V9fsPath *fs_path,
997                                const char *name, void *value, size_t size)
998 {
999     int retval;
1000     V9fsString xname;
1001 
1002     v9fs_string_init(&xname);
1003     v9fs_string_sprintf(&xname, "%s", name);
1004     retval = v9fs_request(ctx->private, T_LGETXATTR, value, size, fs_path,
1005                           &xname);
1006     v9fs_string_free(&xname);
1007     if (retval < 0) {
1008         errno = -retval;
1009     }
1010     return retval;
1011 }
1012 
1013 static ssize_t proxy_llistxattr(FsContext *ctx, V9fsPath *fs_path,
1014                                 void *value, size_t size)
1015 {
1016     int retval;
1017     retval = v9fs_request(ctx->private, T_LLISTXATTR, value, size, fs_path);
1018     if (retval < 0) {
1019         errno = -retval;
1020     }
1021     return retval;
1022 }
1023 
1024 static int proxy_lsetxattr(FsContext *ctx, V9fsPath *fs_path, const char *name,
1025                            void *value, size_t size, int flags)
1026 {
1027     int retval;
1028     V9fsString xname, xvalue;
1029 
1030     v9fs_string_init(&xname);
1031     v9fs_string_sprintf(&xname, "%s", name);
1032 
1033     v9fs_string_init(&xvalue);
1034     xvalue.size = size;
1035     xvalue.data = g_malloc(size);
1036     memcpy(xvalue.data, value, size);
1037 
1038     retval = v9fs_request(ctx->private, T_LSETXATTR, value, fs_path, &xname,
1039                           &xvalue, size, flags);
1040     v9fs_string_free(&xname);
1041     v9fs_string_free(&xvalue);
1042     if (retval < 0) {
1043         errno = -retval;
1044     }
1045     return retval;
1046 }
1047 
1048 static int proxy_lremovexattr(FsContext *ctx, V9fsPath *fs_path,
1049                               const char *name)
1050 {
1051     int retval;
1052     V9fsString xname;
1053 
1054     v9fs_string_init(&xname);
1055     v9fs_string_sprintf(&xname, "%s", name);
1056     retval = v9fs_request(ctx->private, T_LREMOVEXATTR, NULL, fs_path, &xname);
1057     v9fs_string_free(&xname);
1058     if (retval < 0) {
1059         errno = -retval;
1060     }
1061     return retval;
1062 }
1063 
1064 static int proxy_name_to_path(FsContext *ctx, V9fsPath *dir_path,
1065                               const char *name, V9fsPath *target)
1066 {
1067     if (dir_path) {
1068         v9fs_path_sprintf(target, "%s/%s", dir_path->data, name);
1069     } else {
1070         v9fs_path_sprintf(target, "%s", name);
1071     }
1072     return 0;
1073 }
1074 
1075 static int proxy_renameat(FsContext *ctx, V9fsPath *olddir,
1076                           const char *old_name, V9fsPath *newdir,
1077                           const char *new_name)
1078 {
1079     int ret;
1080     V9fsString old_full_name, new_full_name;
1081 
1082     v9fs_string_init(&old_full_name);
1083     v9fs_string_init(&new_full_name);
1084 
1085     v9fs_string_sprintf(&old_full_name, "%s/%s", olddir->data, old_name);
1086     v9fs_string_sprintf(&new_full_name, "%s/%s", newdir->data, new_name);
1087 
1088     ret = proxy_rename(ctx, old_full_name.data, new_full_name.data);
1089     v9fs_string_free(&old_full_name);
1090     v9fs_string_free(&new_full_name);
1091     return ret;
1092 }
1093 
1094 static int proxy_unlinkat(FsContext *ctx, V9fsPath *dir,
1095                           const char *name, int flags)
1096 {
1097     int ret;
1098     V9fsString fullname;
1099     v9fs_string_init(&fullname);
1100 
1101     v9fs_string_sprintf(&fullname, "%s/%s", dir->data, name);
1102     ret = proxy_remove(ctx, fullname.data);
1103     v9fs_string_free(&fullname);
1104 
1105     return ret;
1106 }
1107 
1108 static int proxy_ioc_getversion(FsContext *fs_ctx, V9fsPath *path,
1109                                 mode_t st_mode, uint64_t *st_gen)
1110 {
1111     int err;
1112 
1113     /* Do not try to open special files like device nodes, fifos etc
1114      * we can get fd for regular files and directories only
1115      */
1116     if (!S_ISREG(st_mode) && !S_ISDIR(st_mode)) {
1117         errno = ENOTTY;
1118         return -1;
1119     }
1120     err = v9fs_request(fs_ctx->private, T_GETVERSION, st_gen, path);
1121     if (err < 0) {
1122         errno = -err;
1123         err = -1;
1124     }
1125     return err;
1126 }
1127 
1128 static int connect_namedsocket(const char *path, Error **errp)
1129 {
1130     int sockfd;
1131     struct sockaddr_un helper;
1132 
1133     if (strlen(path) >= sizeof(helper.sun_path)) {
1134         error_setg(errp, "socket name too long");
1135         return -1;
1136     }
1137     sockfd = socket(AF_UNIX, SOCK_STREAM, 0);
1138     if (sockfd < 0) {
1139         error_setg_errno(errp, errno, "failed to create client socket");
1140         return -1;
1141     }
1142     strcpy(helper.sun_path, path);
1143     helper.sun_family = AF_UNIX;
1144     if (connect(sockfd, (struct sockaddr *)&helper, sizeof(helper)) < 0) {
1145         error_setg_errno(errp, errno, "failed to connect to '%s'", path);
1146         close(sockfd);
1147         return -1;
1148     }
1149 
1150     /* remove the socket for security reasons */
1151     unlink(path);
1152     return sockfd;
1153 }
1154 
1155 static void error_append_socket_sockfd_hint(Error *const *errp)
1156 {
1157     error_append_hint(errp, "Either specify socket=/some/path where /some/path"
1158                       " points to a listening AF_UNIX socket or sock_fd=fd"
1159                       " where fd is a file descriptor to a connected AF_UNIX"
1160                       " socket\n");
1161 }
1162 
1163 static int proxy_parse_opts(QemuOpts *opts, FsDriverEntry *fs, Error **errp)
1164 {
1165     const char *socket = qemu_opt_get(opts, "socket");
1166     const char *sock_fd = qemu_opt_get(opts, "sock_fd");
1167 
1168     if (!socket && !sock_fd) {
1169         error_setg(errp, "both socket and sock_fd properties are missing");
1170         error_append_socket_sockfd_hint(errp);
1171         return -1;
1172     }
1173     if (socket && sock_fd) {
1174         error_setg(errp, "both socket and sock_fd properties are set");
1175         error_append_socket_sockfd_hint(errp);
1176         return -1;
1177     }
1178     if (socket) {
1179         fs->path = g_strdup(socket);
1180         fs->export_flags |= V9FS_PROXY_SOCK_NAME;
1181     } else {
1182         fs->path = g_strdup(sock_fd);
1183         fs->export_flags |= V9FS_PROXY_SOCK_FD;
1184     }
1185     return 0;
1186 }
1187 
1188 static int proxy_init(FsContext *ctx, Error **errp)
1189 {
1190     V9fsProxy *proxy = g_new(V9fsProxy, 1);
1191     int sock_id;
1192 
1193     if (ctx->export_flags & V9FS_PROXY_SOCK_NAME) {
1194         sock_id = connect_namedsocket(ctx->fs_root, errp);
1195     } else {
1196         sock_id = atoi(ctx->fs_root);
1197         if (sock_id < 0) {
1198             error_setg(errp, "socket descriptor not initialized");
1199         }
1200     }
1201     if (sock_id < 0) {
1202         g_free(proxy);
1203         return -1;
1204     }
1205     g_free(ctx->fs_root);
1206     ctx->fs_root = NULL;
1207 
1208     proxy->in_iovec.iov_base  = g_malloc(PROXY_MAX_IO_SZ + PROXY_HDR_SZ);
1209     proxy->in_iovec.iov_len   = PROXY_MAX_IO_SZ + PROXY_HDR_SZ;
1210     proxy->out_iovec.iov_base = g_malloc(PROXY_MAX_IO_SZ + PROXY_HDR_SZ);
1211     proxy->out_iovec.iov_len  = PROXY_MAX_IO_SZ + PROXY_HDR_SZ;
1212 
1213     ctx->private = proxy;
1214     proxy->sockfd = sock_id;
1215     qemu_mutex_init(&proxy->mutex);
1216 
1217     ctx->export_flags |= V9FS_PATHNAME_FSCONTEXT;
1218     ctx->exops.get_st_gen = proxy_ioc_getversion;
1219     return 0;
1220 }
1221 
1222 static void proxy_cleanup(FsContext *ctx)
1223 {
1224     V9fsProxy *proxy = ctx->private;
1225 
1226     if (!proxy) {
1227         return;
1228     }
1229 
1230     g_free(proxy->out_iovec.iov_base);
1231     g_free(proxy->in_iovec.iov_base);
1232     if (ctx->export_flags & V9FS_PROXY_SOCK_NAME) {
1233         close(proxy->sockfd);
1234     }
1235     g_free(proxy);
1236 }
1237 
1238 FileOperations proxy_ops = {
1239     .parse_opts   = proxy_parse_opts,
1240     .init         = proxy_init,
1241     .cleanup      = proxy_cleanup,
1242     .lstat        = proxy_lstat,
1243     .readlink     = proxy_readlink,
1244     .close        = proxy_close,
1245     .closedir     = proxy_closedir,
1246     .open         = proxy_open,
1247     .opendir      = proxy_opendir,
1248     .rewinddir    = proxy_rewinddir,
1249     .telldir      = proxy_telldir,
1250     .readdir      = proxy_readdir,
1251     .seekdir      = proxy_seekdir,
1252     .preadv       = proxy_preadv,
1253     .pwritev      = proxy_pwritev,
1254     .chmod        = proxy_chmod,
1255     .mknod        = proxy_mknod,
1256     .mkdir        = proxy_mkdir,
1257     .fstat        = proxy_fstat,
1258     .open2        = proxy_open2,
1259     .symlink      = proxy_symlink,
1260     .link         = proxy_link,
1261     .truncate     = proxy_truncate,
1262     .rename       = proxy_rename,
1263     .chown        = proxy_chown,
1264     .utimensat    = proxy_utimensat,
1265     .remove       = proxy_remove,
1266     .fsync        = proxy_fsync,
1267     .statfs       = proxy_statfs,
1268     .lgetxattr    = proxy_lgetxattr,
1269     .llistxattr   = proxy_llistxattr,
1270     .lsetxattr    = proxy_lsetxattr,
1271     .lremovexattr = proxy_lremovexattr,
1272     .name_to_path = proxy_name_to_path,
1273     .renameat     = proxy_renameat,
1274     .unlinkat     = proxy_unlinkat,
1275 };
1276