xref: /openbmc/qemu/hw/9pfs/9p-proxy.c (revision 095859e5)
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 /*
19  * NOTE: The 9p 'proxy' backend is deprecated (since QEMU 8.1) and will be
20  * removed in a future version of QEMU!
21  */
22 
23 #include "qemu/osdep.h"
24 #include <sys/socket.h>
25 #include <sys/un.h>
26 #include "9p.h"
27 #include "qapi/error.h"
28 #include "qemu/cutils.h"
29 #include "qemu/error-report.h"
30 #include "qemu/option.h"
31 #include "fsdev/qemu-fsdev.h"
32 #include "9p-proxy.h"
33 
34 typedef struct V9fsProxy {
35     int sockfd;
36     QemuMutex mutex;
37     struct iovec in_iovec;
38     struct iovec out_iovec;
39 } V9fsProxy;
40 
41 /*
42  * Return received file descriptor on success in *status.
43  * errno is also returned on *status (which will be < 0)
44  * return < 0 on transport error.
45  */
46 static int v9fs_receivefd(int sockfd, int *status)
47 {
48     struct iovec iov;
49     struct msghdr msg;
50     struct cmsghdr *cmsg;
51     int retval, data, fd;
52     union MsgControl msg_control;
53 
54     iov.iov_base = &data;
55     iov.iov_len = sizeof(data);
56 
57     memset(&msg, 0, sizeof(msg));
58     msg.msg_iov = &iov;
59     msg.msg_iovlen = 1;
60     msg.msg_control = &msg_control;
61     msg.msg_controllen = sizeof(msg_control);
62 
63     do {
64         retval = recvmsg(sockfd, &msg, 0);
65     } while (retval < 0 && errno == EINTR);
66     if (retval <= 0) {
67         return retval;
68     }
69     /*
70      * data is set to V9FS_FD_VALID, if ancillary data is sent.  If this
71      * request doesn't need ancillary data (fd) or an error occurred,
72      * data is set to negative errno value.
73      */
74     if (data != V9FS_FD_VALID) {
75         *status = data;
76         return 0;
77     }
78     /*
79      * File descriptor (fd) is sent in the ancillary data. Check if we
80      * indeed received it. One of the reasons to fail to receive it is if
81      * we exceeded the maximum number of file descriptors!
82      */
83     for (cmsg = CMSG_FIRSTHDR(&msg); cmsg; cmsg = CMSG_NXTHDR(&msg, cmsg)) {
84         if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)) ||
85             cmsg->cmsg_level != SOL_SOCKET ||
86             cmsg->cmsg_type != SCM_RIGHTS) {
87             continue;
88         }
89         fd = *((int *)CMSG_DATA(cmsg));
90         *status = fd;
91         return 0;
92     }
93     *status = -ENFILE;  /* Ancillary data sent but not received */
94     return 0;
95 }
96 
97 static ssize_t socket_read(int sockfd, void *buff, size_t size)
98 {
99     ssize_t retval, total = 0;
100 
101     while (size) {
102         retval = read(sockfd, buff, size);
103         if (retval == 0) {
104             return -EIO;
105         }
106         if (retval < 0) {
107             if (errno == EINTR) {
108                 continue;
109             }
110             return -errno;
111         }
112         size -= retval;
113         buff += retval;
114         total += retval;
115     }
116     return total;
117 }
118 
119 /* Converts proxy_statfs to VFS statfs structure */
120 static void prstatfs_to_statfs(struct statfs *stfs, ProxyStatFS *prstfs)
121 {
122     memset(stfs, 0, sizeof(*stfs));
123     stfs->f_type = prstfs->f_type;
124     stfs->f_bsize = prstfs->f_bsize;
125     stfs->f_blocks = prstfs->f_blocks;
126     stfs->f_bfree = prstfs->f_bfree;
127     stfs->f_bavail = prstfs->f_bavail;
128     stfs->f_files = prstfs->f_files;
129     stfs->f_ffree = prstfs->f_ffree;
130 #ifdef CONFIG_DARWIN
131     /* f_namelen and f_frsize do not exist on Darwin */
132     stfs->f_fsid.val[0] = prstfs->f_fsid[0] & 0xFFFFFFFFU;
133     stfs->f_fsid.val[1] = prstfs->f_fsid[1] >> 32 & 0xFFFFFFFFU;
134 #else
135     stfs->f_fsid.__val[0] = prstfs->f_fsid[0] & 0xFFFFFFFFU;
136     stfs->f_fsid.__val[1] = prstfs->f_fsid[1] >> 32 & 0xFFFFFFFFU;
137     stfs->f_namelen = prstfs->f_namelen;
138     stfs->f_frsize = prstfs->f_frsize;
139 #endif
140 }
141 
142 /* Converts proxy_stat structure to VFS stat structure */
143 static void prstat_to_stat(struct stat *stbuf, ProxyStat *prstat)
144 {
145    memset(stbuf, 0, sizeof(*stbuf));
146    stbuf->st_dev = prstat->st_dev;
147    stbuf->st_ino = prstat->st_ino;
148    stbuf->st_nlink = prstat->st_nlink;
149    stbuf->st_mode = prstat->st_mode;
150    stbuf->st_uid = prstat->st_uid;
151    stbuf->st_gid = prstat->st_gid;
152    stbuf->st_rdev = prstat->st_rdev;
153    stbuf->st_size = prstat->st_size;
154    stbuf->st_blksize = prstat->st_blksize;
155    stbuf->st_blocks = prstat->st_blocks;
156    stbuf->st_atime = prstat->st_atim_sec;
157    stbuf->st_mtime = prstat->st_mtim_sec;
158    stbuf->st_ctime = prstat->st_ctim_sec;
159 #ifdef CONFIG_DARWIN
160    stbuf->st_atimespec.tv_sec = prstat->st_atim_sec;
161    stbuf->st_mtimespec.tv_sec = prstat->st_mtim_sec;
162    stbuf->st_ctimespec.tv_sec = prstat->st_ctim_sec;
163    stbuf->st_atimespec.tv_nsec = prstat->st_atim_nsec;
164    stbuf->st_mtimespec.tv_nsec = prstat->st_mtim_nsec;
165    stbuf->st_ctimespec.tv_nsec = prstat->st_ctim_nsec;
166 #else
167    stbuf->st_atim.tv_sec = prstat->st_atim_sec;
168    stbuf->st_mtim.tv_sec = prstat->st_mtim_sec;
169    stbuf->st_ctim.tv_sec = prstat->st_ctim_sec;
170    stbuf->st_atim.tv_nsec = prstat->st_atim_nsec;
171    stbuf->st_mtim.tv_nsec = prstat->st_mtim_nsec;
172    stbuf->st_ctim.tv_nsec = prstat->st_ctim_nsec;
173 #endif
174 }
175 
176 /*
177  * Response contains two parts
178  * {header, data}
179  * header.type == T_ERROR, data -> -errno
180  * header.type == T_SUCCESS, data -> response
181  * size of errno/response is given by header.size
182  * returns < 0, on transport error. response is
183  * valid only if status >= 0.
184  */
185 static int v9fs_receive_response(V9fsProxy *proxy, int type,
186                                  int *status, void *response)
187 {
188     int retval;
189     ProxyHeader header;
190     struct iovec *reply = &proxy->in_iovec;
191 
192     *status = 0;
193     reply->iov_len = 0;
194     retval = socket_read(proxy->sockfd, reply->iov_base, PROXY_HDR_SZ);
195     if (retval < 0) {
196         return retval;
197     }
198     reply->iov_len = PROXY_HDR_SZ;
199     retval = proxy_unmarshal(reply, 0, "dd", &header.type, &header.size);
200     assert(retval == 4 * 2);
201     /*
202      * if response size > PROXY_MAX_IO_SZ, read the response but ignore it and
203      * return -ENOBUFS
204      */
205     if (header.size > PROXY_MAX_IO_SZ) {
206         int count;
207         while (header.size > 0) {
208             count = MIN(PROXY_MAX_IO_SZ, header.size);
209             count = socket_read(proxy->sockfd, reply->iov_base, count);
210             if (count < 0) {
211                 return count;
212             }
213             header.size -= count;
214         }
215         *status = -ENOBUFS;
216         return 0;
217     }
218 
219     retval = socket_read(proxy->sockfd,
220                          reply->iov_base + PROXY_HDR_SZ, header.size);
221     if (retval < 0) {
222         return retval;
223     }
224     reply->iov_len += header.size;
225     /* there was an error during processing request */
226     if (header.type == T_ERROR) {
227         int ret;
228         ret = proxy_unmarshal(reply, PROXY_HDR_SZ, "d", status);
229         assert(ret == 4);
230         return 0;
231     }
232 
233     switch (type) {
234     case T_LSTAT: {
235         ProxyStat prstat;
236         retval = proxy_unmarshal(reply, PROXY_HDR_SZ,
237                                  "qqqdddqqqqqqqqqq", &prstat.st_dev,
238                                  &prstat.st_ino, &prstat.st_nlink,
239                                  &prstat.st_mode, &prstat.st_uid,
240                                  &prstat.st_gid, &prstat.st_rdev,
241                                  &prstat.st_size, &prstat.st_blksize,
242                                  &prstat.st_blocks,
243                                  &prstat.st_atim_sec, &prstat.st_atim_nsec,
244                                  &prstat.st_mtim_sec, &prstat.st_mtim_nsec,
245                                  &prstat.st_ctim_sec, &prstat.st_ctim_nsec);
246         assert(retval == 8 * 3 + 4 * 3 + 8 * 10);
247         prstat_to_stat(response, &prstat);
248         break;
249     }
250     case T_STATFS: {
251         ProxyStatFS prstfs;
252         retval = proxy_unmarshal(reply, PROXY_HDR_SZ,
253                                  "qqqqqqqqqqq", &prstfs.f_type,
254                                  &prstfs.f_bsize, &prstfs.f_blocks,
255                                  &prstfs.f_bfree, &prstfs.f_bavail,
256                                  &prstfs.f_files, &prstfs.f_ffree,
257                                  &prstfs.f_fsid[0], &prstfs.f_fsid[1],
258                                  &prstfs.f_namelen, &prstfs.f_frsize);
259         assert(retval == 8 * 11);
260         prstatfs_to_statfs(response, &prstfs);
261         break;
262     }
263     case T_READLINK: {
264         V9fsString target;
265         v9fs_string_init(&target);
266         retval = proxy_unmarshal(reply, PROXY_HDR_SZ, "s", &target);
267         strcpy(response, target.data);
268         v9fs_string_free(&target);
269         break;
270     }
271     case T_LGETXATTR:
272     case T_LLISTXATTR: {
273         V9fsString xattr;
274         v9fs_string_init(&xattr);
275         retval = proxy_unmarshal(reply, PROXY_HDR_SZ, "s", &xattr);
276         memcpy(response, xattr.data, xattr.size);
277         v9fs_string_free(&xattr);
278         break;
279     }
280     case T_GETVERSION:
281         retval = proxy_unmarshal(reply, PROXY_HDR_SZ, "q", response);
282         assert(retval == 8);
283         break;
284     default:
285         return -1;
286     }
287     if (retval < 0) {
288         *status  = retval;
289     }
290     return 0;
291 }
292 
293 /*
294  * return < 0 on transport error.
295  * *status is valid only if return >= 0
296  */
297 static int v9fs_receive_status(V9fsProxy *proxy,
298                                struct iovec *reply, int *status)
299 {
300     int retval;
301     ProxyHeader header;
302 
303     *status = 0;
304     reply->iov_len = 0;
305     retval = socket_read(proxy->sockfd, reply->iov_base, PROXY_HDR_SZ);
306     if (retval < 0) {
307         return retval;
308     }
309     reply->iov_len = PROXY_HDR_SZ;
310     retval = proxy_unmarshal(reply, 0, "dd", &header.type, &header.size);
311     assert(retval == 4 * 2);
312     retval = socket_read(proxy->sockfd,
313                          reply->iov_base + PROXY_HDR_SZ, header.size);
314     if (retval < 0) {
315         return retval;
316     }
317     reply->iov_len += header.size;
318     retval = proxy_unmarshal(reply, PROXY_HDR_SZ, "d", status);
319     assert(retval == 4);
320     return 0;
321 }
322 
323 /*
324  * Proxy->header and proxy->request written to socket by QEMU process.
325  * This request read by proxy helper process
326  * returns 0 on success and -errno on error
327  */
328 static int v9fs_request(V9fsProxy *proxy, int type, void *response, ...)
329 {
330     dev_t rdev;
331     va_list ap;
332     int size = 0;
333     int retval = 0;
334     uint64_t offset;
335     ProxyHeader header = { 0, 0};
336     struct timespec spec[2];
337     int flags, mode, uid, gid;
338     V9fsString *name, *value;
339     V9fsString *path, *oldpath;
340     struct iovec *iovec = NULL, *reply = NULL;
341 
342     qemu_mutex_lock(&proxy->mutex);
343 
344     if (proxy->sockfd == -1) {
345         retval = -EIO;
346         goto err_out;
347     }
348     iovec = &proxy->out_iovec;
349     reply = &proxy->in_iovec;
350     va_start(ap, response);
351     switch (type) {
352     case T_OPEN:
353         path = va_arg(ap, V9fsString *);
354         flags = va_arg(ap, int);
355         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sd", path, flags);
356         if (retval > 0) {
357             header.size = retval;
358             header.type = T_OPEN;
359         }
360         break;
361     case T_CREATE:
362         path = va_arg(ap, V9fsString *);
363         flags = va_arg(ap, int);
364         mode = va_arg(ap, int);
365         uid = va_arg(ap, int);
366         gid = va_arg(ap, int);
367         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sdddd", path,
368                                     flags, mode, uid, gid);
369         if (retval > 0) {
370             header.size = retval;
371             header.type = T_CREATE;
372         }
373         break;
374     case T_MKNOD:
375         path = va_arg(ap, V9fsString *);
376         mode = va_arg(ap, int);
377         rdev = va_arg(ap, long int);
378         uid = va_arg(ap, int);
379         gid = va_arg(ap, int);
380         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ddsdq",
381                                     uid, gid, path, mode, rdev);
382         if (retval > 0) {
383             header.size = retval;
384             header.type = T_MKNOD;
385         }
386         break;
387     case T_MKDIR:
388         path = va_arg(ap, V9fsString *);
389         mode = va_arg(ap, int);
390         uid = va_arg(ap, int);
391         gid = va_arg(ap, int);
392         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ddsd",
393                                     uid, gid, path, mode);
394         if (retval > 0) {
395             header.size = retval;
396             header.type = T_MKDIR;
397         }
398         break;
399     case T_SYMLINK:
400         oldpath = va_arg(ap, V9fsString *);
401         path = va_arg(ap, V9fsString *);
402         uid = va_arg(ap, int);
403         gid = va_arg(ap, int);
404         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ddss",
405                                     uid, gid, oldpath, path);
406         if (retval > 0) {
407             header.size = retval;
408             header.type = T_SYMLINK;
409         }
410         break;
411     case T_LINK:
412         oldpath = va_arg(ap, V9fsString *);
413         path = va_arg(ap, V9fsString *);
414         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ss",
415                                     oldpath, path);
416         if (retval > 0) {
417             header.size = retval;
418             header.type = T_LINK;
419         }
420         break;
421     case T_LSTAT:
422         path = va_arg(ap, V9fsString *);
423         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "s", path);
424         if (retval > 0) {
425             header.size = retval;
426             header.type = T_LSTAT;
427         }
428         break;
429     case T_READLINK:
430         path = va_arg(ap, V9fsString *);
431         size = va_arg(ap, int);
432         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sd", path, size);
433         if (retval > 0) {
434             header.size = retval;
435             header.type = T_READLINK;
436         }
437         break;
438     case T_STATFS:
439         path = va_arg(ap, V9fsString *);
440         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "s", path);
441         if (retval > 0) {
442             header.size = retval;
443             header.type = T_STATFS;
444         }
445         break;
446     case T_CHMOD:
447         path = va_arg(ap, V9fsString *);
448         mode = va_arg(ap, int);
449         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sd", path, mode);
450         if (retval > 0) {
451             header.size = retval;
452             header.type = T_CHMOD;
453         }
454         break;
455     case T_CHOWN:
456         path = va_arg(ap, V9fsString *);
457         uid = va_arg(ap, int);
458         gid = va_arg(ap, int);
459         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sdd", path, uid, gid);
460         if (retval > 0) {
461             header.size = retval;
462             header.type = T_CHOWN;
463         }
464         break;
465     case T_TRUNCATE:
466         path = va_arg(ap, V9fsString *);
467         offset = va_arg(ap, uint64_t);
468         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sq", path, offset);
469         if (retval > 0) {
470             header.size = retval;
471             header.type = T_TRUNCATE;
472         }
473         break;
474     case T_UTIME:
475         path = va_arg(ap, V9fsString *);
476         spec[0].tv_sec = va_arg(ap, long);
477         spec[0].tv_nsec = va_arg(ap, long);
478         spec[1].tv_sec = va_arg(ap, long);
479         spec[1].tv_nsec = va_arg(ap, long);
480         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sqqqq", path,
481                                     spec[0].tv_sec, spec[1].tv_nsec,
482                                     spec[1].tv_sec, spec[1].tv_nsec);
483         if (retval > 0) {
484             header.size = retval;
485             header.type = T_UTIME;
486         }
487         break;
488     case T_RENAME:
489         oldpath = va_arg(ap, V9fsString *);
490         path = va_arg(ap, V9fsString *);
491         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ss", oldpath, path);
492         if (retval > 0) {
493             header.size = retval;
494             header.type = T_RENAME;
495         }
496         break;
497     case T_REMOVE:
498         path = va_arg(ap, V9fsString *);
499         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "s", path);
500         if (retval > 0) {
501             header.size = retval;
502             header.type = T_REMOVE;
503         }
504         break;
505     case T_LGETXATTR:
506         size = va_arg(ap, int);
507         path = va_arg(ap, V9fsString *);
508         name = va_arg(ap, V9fsString *);
509         retval = proxy_marshal(iovec, PROXY_HDR_SZ,
510                                     "dss", size, path, name);
511         if (retval > 0) {
512             header.size = retval;
513             header.type = T_LGETXATTR;
514         }
515         break;
516     case T_LLISTXATTR:
517         size = va_arg(ap, int);
518         path = va_arg(ap, V9fsString *);
519         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ds", size, path);
520         if (retval > 0) {
521             header.size = retval;
522             header.type = T_LLISTXATTR;
523         }
524         break;
525     case T_LSETXATTR:
526         path = va_arg(ap, V9fsString *);
527         name = va_arg(ap, V9fsString *);
528         value = va_arg(ap, V9fsString *);
529         size = va_arg(ap, int);
530         flags = va_arg(ap, int);
531         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sssdd",
532                                     path, name, value, size, flags);
533         if (retval > 0) {
534             header.size = retval;
535             header.type = T_LSETXATTR;
536         }
537         break;
538     case T_LREMOVEXATTR:
539         path = va_arg(ap, V9fsString *);
540         name = va_arg(ap, V9fsString *);
541         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ss", path, name);
542         if (retval > 0) {
543             header.size = retval;
544             header.type = T_LREMOVEXATTR;
545         }
546         break;
547     case T_GETVERSION:
548         path = va_arg(ap, V9fsString *);
549         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "s", path);
550         if (retval > 0) {
551             header.size = retval;
552             header.type = T_GETVERSION;
553         }
554         break;
555     default:
556         error_report("Invalid type %d", type);
557         retval = -EINVAL;
558         break;
559     }
560     va_end(ap);
561 
562     if (retval < 0) {
563         goto err_out;
564     }
565 
566     /* marshal the header details */
567     retval = proxy_marshal(iovec, 0, "dd", header.type, header.size);
568     assert(retval == 4 * 2);
569     header.size += PROXY_HDR_SZ;
570 
571     retval = qemu_write_full(proxy->sockfd, iovec->iov_base, header.size);
572     if (retval != header.size) {
573         goto close_error;
574     }
575 
576     switch (type) {
577     case T_OPEN:
578     case T_CREATE:
579         /*
580          * A file descriptor is returned as response for
581          * T_OPEN,T_CREATE on success
582          */
583         if (v9fs_receivefd(proxy->sockfd, &retval) < 0) {
584             goto close_error;
585         }
586         break;
587     case T_MKNOD:
588     case T_MKDIR:
589     case T_SYMLINK:
590     case T_LINK:
591     case T_CHMOD:
592     case T_CHOWN:
593     case T_RENAME:
594     case T_TRUNCATE:
595     case T_UTIME:
596     case T_REMOVE:
597     case T_LSETXATTR:
598     case T_LREMOVEXATTR:
599         if (v9fs_receive_status(proxy, reply, &retval) < 0) {
600             goto close_error;
601         }
602         break;
603     case T_LSTAT:
604     case T_READLINK:
605     case T_STATFS:
606     case T_GETVERSION:
607         if (v9fs_receive_response(proxy, type, &retval, response) < 0) {
608             goto close_error;
609         }
610         break;
611     case T_LGETXATTR:
612     case T_LLISTXATTR:
613         if (!size) {
614             if (v9fs_receive_status(proxy, reply, &retval) < 0) {
615                 goto close_error;
616             }
617         } else {
618             if (v9fs_receive_response(proxy, type, &retval, response) < 0) {
619                 goto close_error;
620             }
621         }
622         break;
623     }
624 
625 err_out:
626     qemu_mutex_unlock(&proxy->mutex);
627     return retval;
628 
629 close_error:
630     close(proxy->sockfd);
631     proxy->sockfd = -1;
632     qemu_mutex_unlock(&proxy->mutex);
633     return -EIO;
634 }
635 
636 static int proxy_lstat(FsContext *fs_ctx, V9fsPath *fs_path, struct stat *stbuf)
637 {
638     int retval;
639     retval = v9fs_request(fs_ctx->private, T_LSTAT, stbuf, fs_path);
640     if (retval < 0) {
641         errno = -retval;
642         return -1;
643     }
644     return retval;
645 }
646 
647 static ssize_t proxy_readlink(FsContext *fs_ctx, V9fsPath *fs_path,
648                               char *buf, size_t bufsz)
649 {
650     int retval;
651     retval = v9fs_request(fs_ctx->private, T_READLINK, buf, fs_path, bufsz);
652     if (retval < 0) {
653         errno = -retval;
654         return -1;
655     }
656     return strlen(buf);
657 }
658 
659 static int proxy_close(FsContext *ctx, V9fsFidOpenState *fs)
660 {
661     return close(fs->fd);
662 }
663 
664 static int proxy_closedir(FsContext *ctx, V9fsFidOpenState *fs)
665 {
666     return closedir(fs->dir.stream);
667 }
668 
669 static int proxy_open(FsContext *ctx, V9fsPath *fs_path,
670                       int flags, V9fsFidOpenState *fs)
671 {
672     fs->fd = v9fs_request(ctx->private, T_OPEN, NULL, fs_path, flags);
673     if (fs->fd < 0) {
674         errno = -fs->fd;
675         fs->fd = -1;
676     }
677     return fs->fd;
678 }
679 
680 static int proxy_opendir(FsContext *ctx,
681                          V9fsPath *fs_path, V9fsFidOpenState *fs)
682 {
683     int serrno, fd;
684 
685     fs->dir.stream = NULL;
686     fd = v9fs_request(ctx->private, T_OPEN, NULL, fs_path, O_DIRECTORY);
687     if (fd < 0) {
688         errno = -fd;
689         return -1;
690     }
691     fs->dir.stream = fdopendir(fd);
692     if (!fs->dir.stream) {
693         serrno = errno;
694         close(fd);
695         errno = serrno;
696         return -1;
697     }
698     return 0;
699 }
700 
701 static void proxy_rewinddir(FsContext *ctx, V9fsFidOpenState *fs)
702 {
703     rewinddir(fs->dir.stream);
704 }
705 
706 static off_t proxy_telldir(FsContext *ctx, V9fsFidOpenState *fs)
707 {
708     return telldir(fs->dir.stream);
709 }
710 
711 static struct dirent *proxy_readdir(FsContext *ctx, V9fsFidOpenState *fs)
712 {
713     struct dirent *entry;
714     entry = readdir(fs->dir.stream);
715 #ifdef CONFIG_DARWIN
716     if (!entry) {
717         return NULL;
718     }
719     int td;
720     td = telldir(fs->dir.stream);
721     /* If telldir fails, fail the entire readdir call */
722     if (td < 0) {
723         return NULL;
724     }
725     entry->d_seekoff = td;
726 #endif
727     return entry;
728 }
729 
730 static void proxy_seekdir(FsContext *ctx, V9fsFidOpenState *fs, off_t off)
731 {
732     seekdir(fs->dir.stream, off);
733 }
734 
735 static ssize_t proxy_preadv(FsContext *ctx, V9fsFidOpenState *fs,
736                             const struct iovec *iov,
737                             int iovcnt, off_t offset)
738 {
739     ssize_t ret;
740 #ifdef CONFIG_PREADV
741     ret = preadv(fs->fd, iov, iovcnt, offset);
742 #else
743     ret = lseek(fs->fd, offset, SEEK_SET);
744     if (ret >= 0) {
745         ret = readv(fs->fd, iov, iovcnt);
746     }
747 #endif
748     return ret;
749 }
750 
751 static ssize_t proxy_pwritev(FsContext *ctx, V9fsFidOpenState *fs,
752                              const struct iovec *iov,
753                              int iovcnt, off_t offset)
754 {
755     ssize_t ret;
756 
757 #ifdef CONFIG_PREADV
758     ret = pwritev(fs->fd, iov, iovcnt, offset);
759 #else
760     ret = lseek(fs->fd, offset, SEEK_SET);
761     if (ret >= 0) {
762         ret = writev(fs->fd, iov, iovcnt);
763     }
764 #endif
765 #ifdef CONFIG_SYNC_FILE_RANGE
766     if (ret > 0 && ctx->export_flags & V9FS_IMMEDIATE_WRITEOUT) {
767         /*
768          * Initiate a writeback. This is not a data integrity sync.
769          * We want to ensure that we don't leave dirty pages in the cache
770          * after write when writeout=immediate is specified.
771          */
772         sync_file_range(fs->fd, offset, ret,
773                         SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE);
774     }
775 #endif
776     return ret;
777 }
778 
779 static int proxy_chmod(FsContext *fs_ctx, V9fsPath *fs_path, FsCred *credp)
780 {
781     int retval;
782     retval = v9fs_request(fs_ctx->private, T_CHMOD, NULL, fs_path,
783                           credp->fc_mode);
784     if (retval < 0) {
785         errno = -retval;
786     }
787     return retval;
788 }
789 
790 static int proxy_mknod(FsContext *fs_ctx, V9fsPath *dir_path,
791                        const char *name, FsCred *credp)
792 {
793     int retval;
794     V9fsString fullname;
795 
796     v9fs_string_init(&fullname);
797     v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
798 
799     retval = v9fs_request(fs_ctx->private, T_MKNOD, NULL, &fullname,
800                           credp->fc_mode, credp->fc_rdev,
801                           credp->fc_uid, credp->fc_gid);
802     v9fs_string_free(&fullname);
803     if (retval < 0) {
804         errno = -retval;
805         retval = -1;
806     }
807     return retval;
808 }
809 
810 static int proxy_mkdir(FsContext *fs_ctx, V9fsPath *dir_path,
811                        const char *name, FsCred *credp)
812 {
813     int retval;
814     V9fsString fullname;
815 
816     v9fs_string_init(&fullname);
817     v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
818 
819     retval = v9fs_request(fs_ctx->private, T_MKDIR, NULL, &fullname,
820                           credp->fc_mode, credp->fc_uid, credp->fc_gid);
821     v9fs_string_free(&fullname);
822     if (retval < 0) {
823         errno = -retval;
824         retval = -1;
825     }
826     return retval;
827 }
828 
829 static int proxy_fstat(FsContext *fs_ctx, int fid_type,
830                        V9fsFidOpenState *fs, struct stat *stbuf)
831 {
832     int fd;
833 
834     if (fid_type == P9_FID_DIR) {
835         fd = dirfd(fs->dir.stream);
836     } else {
837         fd = fs->fd;
838     }
839     return fstat(fd, stbuf);
840 }
841 
842 static int proxy_open2(FsContext *fs_ctx, V9fsPath *dir_path, const char *name,
843                        int flags, FsCred *credp, V9fsFidOpenState *fs)
844 {
845     V9fsString fullname;
846 
847     v9fs_string_init(&fullname);
848     v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
849 
850     fs->fd = v9fs_request(fs_ctx->private, T_CREATE, NULL, &fullname, flags,
851                           credp->fc_mode, credp->fc_uid, credp->fc_gid);
852     v9fs_string_free(&fullname);
853     if (fs->fd < 0) {
854         errno = -fs->fd;
855         fs->fd = -1;
856     }
857     return fs->fd;
858 }
859 
860 static int proxy_symlink(FsContext *fs_ctx, const char *oldpath,
861                          V9fsPath *dir_path, const char *name, FsCred *credp)
862 {
863     int retval;
864     V9fsString fullname, target;
865 
866     v9fs_string_init(&fullname);
867     v9fs_string_init(&target);
868 
869     v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
870     v9fs_string_sprintf(&target, "%s", oldpath);
871 
872     retval = v9fs_request(fs_ctx->private, T_SYMLINK, NULL, &target, &fullname,
873                           credp->fc_uid, credp->fc_gid);
874     v9fs_string_free(&fullname);
875     v9fs_string_free(&target);
876     if (retval < 0) {
877         errno = -retval;
878         retval = -1;
879     }
880     return retval;
881 }
882 
883 static int proxy_link(FsContext *ctx, V9fsPath *oldpath,
884                       V9fsPath *dirpath, const char *name)
885 {
886     int retval;
887     V9fsString newpath;
888 
889     v9fs_string_init(&newpath);
890     v9fs_string_sprintf(&newpath, "%s/%s", dirpath->data, name);
891 
892     retval = v9fs_request(ctx->private, T_LINK, NULL, oldpath, &newpath);
893     v9fs_string_free(&newpath);
894     if (retval < 0) {
895         errno = -retval;
896         retval = -1;
897     }
898     return retval;
899 }
900 
901 static int proxy_truncate(FsContext *ctx, V9fsPath *fs_path, off_t size)
902 {
903     int retval;
904 
905     retval = v9fs_request(ctx->private, T_TRUNCATE, NULL, fs_path, size);
906     if (retval < 0) {
907         errno = -retval;
908         return -1;
909     }
910     return 0;
911 }
912 
913 static int proxy_rename(FsContext *ctx, const char *oldpath,
914                         const char *newpath)
915 {
916     int retval;
917     V9fsString oldname, newname;
918 
919     v9fs_string_init(&oldname);
920     v9fs_string_init(&newname);
921 
922     v9fs_string_sprintf(&oldname, "%s", oldpath);
923     v9fs_string_sprintf(&newname, "%s", newpath);
924     retval = v9fs_request(ctx->private, T_RENAME, NULL, &oldname, &newname);
925     v9fs_string_free(&oldname);
926     v9fs_string_free(&newname);
927     if (retval < 0) {
928         errno = -retval;
929     }
930     return retval;
931 }
932 
933 static int proxy_chown(FsContext *fs_ctx, V9fsPath *fs_path, FsCred *credp)
934 {
935     int retval;
936     retval = v9fs_request(fs_ctx->private, T_CHOWN, NULL, fs_path,
937                           credp->fc_uid, credp->fc_gid);
938     if (retval < 0) {
939         errno = -retval;
940     }
941     return retval;
942 }
943 
944 static int proxy_utimensat(FsContext *s, V9fsPath *fs_path,
945                            const struct timespec *buf)
946 {
947     int retval;
948     retval = v9fs_request(s->private, T_UTIME, NULL, fs_path,
949                           buf[0].tv_sec, buf[0].tv_nsec,
950                           buf[1].tv_sec, buf[1].tv_nsec);
951     if (retval < 0) {
952         errno = -retval;
953     }
954     return retval;
955 }
956 
957 static int proxy_remove(FsContext *ctx, const char *path)
958 {
959     int retval;
960     V9fsString name;
961     v9fs_string_init(&name);
962     v9fs_string_sprintf(&name, "%s", path);
963     retval = v9fs_request(ctx->private, T_REMOVE, NULL, &name);
964     v9fs_string_free(&name);
965     if (retval < 0) {
966         errno = -retval;
967     }
968     return retval;
969 }
970 
971 static int proxy_fsync(FsContext *ctx, int fid_type,
972                        V9fsFidOpenState *fs, int datasync)
973 {
974     int fd;
975 
976     if (fid_type == P9_FID_DIR) {
977         fd = dirfd(fs->dir.stream);
978     } else {
979         fd = fs->fd;
980     }
981 
982     if (datasync) {
983         return qemu_fdatasync(fd);
984     } else {
985         return fsync(fd);
986     }
987 }
988 
989 static int proxy_statfs(FsContext *s, V9fsPath *fs_path, struct statfs *stbuf)
990 {
991     int retval;
992     retval = v9fs_request(s->private, T_STATFS, stbuf, fs_path);
993     if (retval < 0) {
994         errno = -retval;
995         return -1;
996     }
997     return retval;
998 }
999 
1000 static ssize_t proxy_lgetxattr(FsContext *ctx, V9fsPath *fs_path,
1001                                const char *name, void *value, size_t size)
1002 {
1003     int retval;
1004     V9fsString xname;
1005 
1006     v9fs_string_init(&xname);
1007     v9fs_string_sprintf(&xname, "%s", name);
1008     retval = v9fs_request(ctx->private, T_LGETXATTR, value, size, fs_path,
1009                           &xname);
1010     v9fs_string_free(&xname);
1011     if (retval < 0) {
1012         errno = -retval;
1013     }
1014     return retval;
1015 }
1016 
1017 static ssize_t proxy_llistxattr(FsContext *ctx, V9fsPath *fs_path,
1018                                 void *value, size_t size)
1019 {
1020     int retval;
1021     retval = v9fs_request(ctx->private, T_LLISTXATTR, value, size, fs_path);
1022     if (retval < 0) {
1023         errno = -retval;
1024     }
1025     return retval;
1026 }
1027 
1028 static int proxy_lsetxattr(FsContext *ctx, V9fsPath *fs_path, const char *name,
1029                            void *value, size_t size, int flags)
1030 {
1031     int retval;
1032     V9fsString xname, xvalue;
1033 
1034     v9fs_string_init(&xname);
1035     v9fs_string_sprintf(&xname, "%s", name);
1036 
1037     v9fs_string_init(&xvalue);
1038     xvalue.size = size;
1039     xvalue.data = g_malloc(size);
1040     memcpy(xvalue.data, value, size);
1041 
1042     retval = v9fs_request(ctx->private, T_LSETXATTR, value, fs_path, &xname,
1043                           &xvalue, size, flags);
1044     v9fs_string_free(&xname);
1045     v9fs_string_free(&xvalue);
1046     if (retval < 0) {
1047         errno = -retval;
1048     }
1049     return retval;
1050 }
1051 
1052 static int proxy_lremovexattr(FsContext *ctx, V9fsPath *fs_path,
1053                               const char *name)
1054 {
1055     int retval;
1056     V9fsString xname;
1057 
1058     v9fs_string_init(&xname);
1059     v9fs_string_sprintf(&xname, "%s", name);
1060     retval = v9fs_request(ctx->private, T_LREMOVEXATTR, NULL, fs_path, &xname);
1061     v9fs_string_free(&xname);
1062     if (retval < 0) {
1063         errno = -retval;
1064     }
1065     return retval;
1066 }
1067 
1068 static int proxy_name_to_path(FsContext *ctx, V9fsPath *dir_path,
1069                               const char *name, V9fsPath *target)
1070 {
1071     if (dir_path) {
1072         v9fs_path_sprintf(target, "%s/%s", dir_path->data, name);
1073     } else {
1074         v9fs_path_sprintf(target, "%s", name);
1075     }
1076     return 0;
1077 }
1078 
1079 static int proxy_renameat(FsContext *ctx, V9fsPath *olddir,
1080                           const char *old_name, V9fsPath *newdir,
1081                           const char *new_name)
1082 {
1083     int ret;
1084     V9fsString old_full_name, new_full_name;
1085 
1086     v9fs_string_init(&old_full_name);
1087     v9fs_string_init(&new_full_name);
1088 
1089     v9fs_string_sprintf(&old_full_name, "%s/%s", olddir->data, old_name);
1090     v9fs_string_sprintf(&new_full_name, "%s/%s", newdir->data, new_name);
1091 
1092     ret = proxy_rename(ctx, old_full_name.data, new_full_name.data);
1093     v9fs_string_free(&old_full_name);
1094     v9fs_string_free(&new_full_name);
1095     return ret;
1096 }
1097 
1098 static int proxy_unlinkat(FsContext *ctx, V9fsPath *dir,
1099                           const char *name, int flags)
1100 {
1101     int ret;
1102     V9fsString fullname;
1103     v9fs_string_init(&fullname);
1104 
1105     v9fs_string_sprintf(&fullname, "%s/%s", dir->data, name);
1106     ret = proxy_remove(ctx, fullname.data);
1107     v9fs_string_free(&fullname);
1108 
1109     return ret;
1110 }
1111 
1112 static int proxy_ioc_getversion(FsContext *fs_ctx, V9fsPath *path,
1113                                 mode_t st_mode, uint64_t *st_gen)
1114 {
1115     int err;
1116 
1117     /* Do not try to open special files like device nodes, fifos etc
1118      * we can get fd for regular files and directories only
1119      */
1120     if (!S_ISREG(st_mode) && !S_ISDIR(st_mode)) {
1121         errno = ENOTTY;
1122         return -1;
1123     }
1124     err = v9fs_request(fs_ctx->private, T_GETVERSION, st_gen, path);
1125     if (err < 0) {
1126         errno = -err;
1127         err = -1;
1128     }
1129     return err;
1130 }
1131 
1132 static int connect_namedsocket(const char *path, Error **errp)
1133 {
1134     int sockfd;
1135     struct sockaddr_un helper;
1136 
1137     if (strlen(path) >= sizeof(helper.sun_path)) {
1138         error_setg(errp, "socket name too long");
1139         return -1;
1140     }
1141     sockfd = socket(AF_UNIX, SOCK_STREAM, 0);
1142     if (sockfd < 0) {
1143         error_setg_errno(errp, errno, "failed to create client socket");
1144         return -1;
1145     }
1146     strcpy(helper.sun_path, path);
1147     helper.sun_family = AF_UNIX;
1148     if (connect(sockfd, (struct sockaddr *)&helper, sizeof(helper)) < 0) {
1149         error_setg_errno(errp, errno, "failed to connect to '%s'", path);
1150         close(sockfd);
1151         return -1;
1152     }
1153 
1154     /* remove the socket for security reasons */
1155     unlink(path);
1156     return sockfd;
1157 }
1158 
1159 static void error_append_socket_sockfd_hint(Error *const *errp)
1160 {
1161     error_append_hint(errp, "Either specify socket=/some/path where /some/path"
1162                       " points to a listening AF_UNIX socket or sock_fd=fd"
1163                       " where fd is a file descriptor to a connected AF_UNIX"
1164                       " socket\n");
1165 }
1166 
1167 static int proxy_parse_opts(QemuOpts *opts, FsDriverEntry *fs, Error **errp)
1168 {
1169     const char *socket = qemu_opt_get(opts, "socket");
1170     const char *sock_fd = qemu_opt_get(opts, "sock_fd");
1171 
1172     if (!socket && !sock_fd) {
1173         error_setg(errp, "both socket and sock_fd properties are missing");
1174         error_append_socket_sockfd_hint(errp);
1175         return -1;
1176     }
1177     if (socket && sock_fd) {
1178         error_setg(errp, "both socket and sock_fd properties are set");
1179         error_append_socket_sockfd_hint(errp);
1180         return -1;
1181     }
1182     if (socket) {
1183         fs->path = g_strdup(socket);
1184         fs->export_flags |= V9FS_PROXY_SOCK_NAME;
1185     } else {
1186         fs->path = g_strdup(sock_fd);
1187         fs->export_flags |= V9FS_PROXY_SOCK_FD;
1188     }
1189     return 0;
1190 }
1191 
1192 static int proxy_init(FsContext *ctx, Error **errp)
1193 {
1194     V9fsProxy *proxy = g_new(V9fsProxy, 1);
1195     int sock_id;
1196 
1197     if (ctx->export_flags & V9FS_PROXY_SOCK_NAME) {
1198         sock_id = connect_namedsocket(ctx->fs_root, errp);
1199     } else {
1200         sock_id = atoi(ctx->fs_root);
1201         if (sock_id < 0) {
1202             error_setg(errp, "socket descriptor not initialized");
1203         }
1204     }
1205     if (sock_id < 0) {
1206         g_free(proxy);
1207         return -1;
1208     }
1209     g_free(ctx->fs_root);
1210     ctx->fs_root = NULL;
1211 
1212     proxy->in_iovec.iov_base  = g_malloc(PROXY_MAX_IO_SZ + PROXY_HDR_SZ);
1213     proxy->in_iovec.iov_len   = PROXY_MAX_IO_SZ + PROXY_HDR_SZ;
1214     proxy->out_iovec.iov_base = g_malloc(PROXY_MAX_IO_SZ + PROXY_HDR_SZ);
1215     proxy->out_iovec.iov_len  = PROXY_MAX_IO_SZ + PROXY_HDR_SZ;
1216 
1217     ctx->private = proxy;
1218     proxy->sockfd = sock_id;
1219     qemu_mutex_init(&proxy->mutex);
1220 
1221     ctx->export_flags |= V9FS_PATHNAME_FSCONTEXT;
1222     ctx->exops.get_st_gen = proxy_ioc_getversion;
1223     return 0;
1224 }
1225 
1226 static void proxy_cleanup(FsContext *ctx)
1227 {
1228     V9fsProxy *proxy = ctx->private;
1229 
1230     if (!proxy) {
1231         return;
1232     }
1233 
1234     g_free(proxy->out_iovec.iov_base);
1235     g_free(proxy->in_iovec.iov_base);
1236     if (ctx->export_flags & V9FS_PROXY_SOCK_NAME) {
1237         close(proxy->sockfd);
1238     }
1239     g_free(proxy);
1240 }
1241 
1242 FileOperations proxy_ops = {
1243     .parse_opts   = proxy_parse_opts,
1244     .init         = proxy_init,
1245     .cleanup      = proxy_cleanup,
1246     .lstat        = proxy_lstat,
1247     .readlink     = proxy_readlink,
1248     .close        = proxy_close,
1249     .closedir     = proxy_closedir,
1250     .open         = proxy_open,
1251     .opendir      = proxy_opendir,
1252     .rewinddir    = proxy_rewinddir,
1253     .telldir      = proxy_telldir,
1254     .readdir      = proxy_readdir,
1255     .seekdir      = proxy_seekdir,
1256     .preadv       = proxy_preadv,
1257     .pwritev      = proxy_pwritev,
1258     .chmod        = proxy_chmod,
1259     .mknod        = proxy_mknod,
1260     .mkdir        = proxy_mkdir,
1261     .fstat        = proxy_fstat,
1262     .open2        = proxy_open2,
1263     .symlink      = proxy_symlink,
1264     .link         = proxy_link,
1265     .truncate     = proxy_truncate,
1266     .rename       = proxy_rename,
1267     .chown        = proxy_chown,
1268     .utimensat    = proxy_utimensat,
1269     .remove       = proxy_remove,
1270     .fsync        = proxy_fsync,
1271     .statfs       = proxy_statfs,
1272     .lgetxattr    = proxy_lgetxattr,
1273     .llistxattr   = proxy_llistxattr,
1274     .lsetxattr    = proxy_lsetxattr,
1275     .lremovexattr = proxy_lremovexattr,
1276     .name_to_path = proxy_name_to_path,
1277     .renameat     = proxy_renameat,
1278     .unlinkat     = proxy_unlinkat,
1279 };
1280