xref: /openbmc/qemu/hw/9pfs/9p-proxy.c (revision 30b6852c)
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     stfs->f_fsid.__val[0] = prstfs->f_fsid[0] & 0xFFFFFFFFU;
127     stfs->f_fsid.__val[1] = prstfs->f_fsid[1] >> 32 & 0xFFFFFFFFU;
128     stfs->f_namelen = prstfs->f_namelen;
129     stfs->f_frsize = prstfs->f_frsize;
130 }
131 
132 /* Converts proxy_stat structure to VFS stat structure */
133 static void prstat_to_stat(struct stat *stbuf, ProxyStat *prstat)
134 {
135    memset(stbuf, 0, sizeof(*stbuf));
136    stbuf->st_dev = prstat->st_dev;
137    stbuf->st_ino = prstat->st_ino;
138    stbuf->st_nlink = prstat->st_nlink;
139    stbuf->st_mode = prstat->st_mode;
140    stbuf->st_uid = prstat->st_uid;
141    stbuf->st_gid = prstat->st_gid;
142    stbuf->st_rdev = prstat->st_rdev;
143    stbuf->st_size = prstat->st_size;
144    stbuf->st_blksize = prstat->st_blksize;
145    stbuf->st_blocks = prstat->st_blocks;
146    stbuf->st_atim.tv_sec = prstat->st_atim_sec;
147    stbuf->st_atim.tv_nsec = prstat->st_atim_nsec;
148    stbuf->st_mtime = prstat->st_mtim_sec;
149    stbuf->st_mtim.tv_nsec = prstat->st_mtim_nsec;
150    stbuf->st_ctime = prstat->st_ctim_sec;
151    stbuf->st_ctim.tv_nsec = prstat->st_ctim_nsec;
152 }
153 
154 /*
155  * Response contains two parts
156  * {header, data}
157  * header.type == T_ERROR, data -> -errno
158  * header.type == T_SUCCESS, data -> response
159  * size of errno/response is given by header.size
160  * returns < 0, on transport error. response is
161  * valid only if status >= 0.
162  */
163 static int v9fs_receive_response(V9fsProxy *proxy, int type,
164                                  int *status, void *response)
165 {
166     int retval;
167     ProxyHeader header;
168     struct iovec *reply = &proxy->in_iovec;
169 
170     *status = 0;
171     reply->iov_len = 0;
172     retval = socket_read(proxy->sockfd, reply->iov_base, PROXY_HDR_SZ);
173     if (retval < 0) {
174         return retval;
175     }
176     reply->iov_len = PROXY_HDR_SZ;
177     retval = proxy_unmarshal(reply, 0, "dd", &header.type, &header.size);
178     assert(retval == 4 * 2);
179     /*
180      * if response size > PROXY_MAX_IO_SZ, read the response but ignore it and
181      * return -ENOBUFS
182      */
183     if (header.size > PROXY_MAX_IO_SZ) {
184         int count;
185         while (header.size > 0) {
186             count = MIN(PROXY_MAX_IO_SZ, header.size);
187             count = socket_read(proxy->sockfd, reply->iov_base, count);
188             if (count < 0) {
189                 return count;
190             }
191             header.size -= count;
192         }
193         *status = -ENOBUFS;
194         return 0;
195     }
196 
197     retval = socket_read(proxy->sockfd,
198                          reply->iov_base + PROXY_HDR_SZ, header.size);
199     if (retval < 0) {
200         return retval;
201     }
202     reply->iov_len += header.size;
203     /* there was an error during processing request */
204     if (header.type == T_ERROR) {
205         int ret;
206         ret = proxy_unmarshal(reply, PROXY_HDR_SZ, "d", status);
207         assert(ret == 4);
208         return 0;
209     }
210 
211     switch (type) {
212     case T_LSTAT: {
213         ProxyStat prstat;
214         retval = proxy_unmarshal(reply, PROXY_HDR_SZ,
215                                  "qqqdddqqqqqqqqqq", &prstat.st_dev,
216                                  &prstat.st_ino, &prstat.st_nlink,
217                                  &prstat.st_mode, &prstat.st_uid,
218                                  &prstat.st_gid, &prstat.st_rdev,
219                                  &prstat.st_size, &prstat.st_blksize,
220                                  &prstat.st_blocks,
221                                  &prstat.st_atim_sec, &prstat.st_atim_nsec,
222                                  &prstat.st_mtim_sec, &prstat.st_mtim_nsec,
223                                  &prstat.st_ctim_sec, &prstat.st_ctim_nsec);
224         assert(retval == 8 * 3 + 4 * 3 + 8 * 10);
225         prstat_to_stat(response, &prstat);
226         break;
227     }
228     case T_STATFS: {
229         ProxyStatFS prstfs;
230         retval = proxy_unmarshal(reply, PROXY_HDR_SZ,
231                                  "qqqqqqqqqqq", &prstfs.f_type,
232                                  &prstfs.f_bsize, &prstfs.f_blocks,
233                                  &prstfs.f_bfree, &prstfs.f_bavail,
234                                  &prstfs.f_files, &prstfs.f_ffree,
235                                  &prstfs.f_fsid[0], &prstfs.f_fsid[1],
236                                  &prstfs.f_namelen, &prstfs.f_frsize);
237         assert(retval == 8 * 11);
238         prstatfs_to_statfs(response, &prstfs);
239         break;
240     }
241     case T_READLINK: {
242         V9fsString target;
243         v9fs_string_init(&target);
244         retval = proxy_unmarshal(reply, PROXY_HDR_SZ, "s", &target);
245         strcpy(response, target.data);
246         v9fs_string_free(&target);
247         break;
248     }
249     case T_LGETXATTR:
250     case T_LLISTXATTR: {
251         V9fsString xattr;
252         v9fs_string_init(&xattr);
253         retval = proxy_unmarshal(reply, PROXY_HDR_SZ, "s", &xattr);
254         memcpy(response, xattr.data, xattr.size);
255         v9fs_string_free(&xattr);
256         break;
257     }
258     case T_GETVERSION:
259         retval = proxy_unmarshal(reply, PROXY_HDR_SZ, "q", response);
260         assert(retval == 8);
261         break;
262     default:
263         return -1;
264     }
265     if (retval < 0) {
266         *status  = retval;
267     }
268     return 0;
269 }
270 
271 /*
272  * return < 0 on transport error.
273  * *status is valid only if return >= 0
274  */
275 static int v9fs_receive_status(V9fsProxy *proxy,
276                                struct iovec *reply, int *status)
277 {
278     int retval;
279     ProxyHeader header;
280 
281     *status = 0;
282     reply->iov_len = 0;
283     retval = socket_read(proxy->sockfd, reply->iov_base, PROXY_HDR_SZ);
284     if (retval < 0) {
285         return retval;
286     }
287     reply->iov_len = PROXY_HDR_SZ;
288     retval = proxy_unmarshal(reply, 0, "dd", &header.type, &header.size);
289     assert(retval == 4 * 2);
290     retval = socket_read(proxy->sockfd,
291                          reply->iov_base + PROXY_HDR_SZ, header.size);
292     if (retval < 0) {
293         return retval;
294     }
295     reply->iov_len += header.size;
296     retval = proxy_unmarshal(reply, PROXY_HDR_SZ, "d", status);
297     assert(retval == 4);
298     return 0;
299 }
300 
301 /*
302  * Proxy->header and proxy->request written to socket by QEMU process.
303  * This request read by proxy helper process
304  * returns 0 on success and -errno on error
305  */
306 static int v9fs_request(V9fsProxy *proxy, int type, void *response, ...)
307 {
308     dev_t rdev;
309     va_list ap;
310     int size = 0;
311     int retval = 0;
312     uint64_t offset;
313     ProxyHeader header = { 0, 0};
314     struct timespec spec[2];
315     int flags, mode, uid, gid;
316     V9fsString *name, *value;
317     V9fsString *path, *oldpath;
318     struct iovec *iovec = NULL, *reply = NULL;
319 
320     qemu_mutex_lock(&proxy->mutex);
321 
322     if (proxy->sockfd == -1) {
323         retval = -EIO;
324         goto err_out;
325     }
326     iovec = &proxy->out_iovec;
327     reply = &proxy->in_iovec;
328     va_start(ap, response);
329     switch (type) {
330     case T_OPEN:
331         path = va_arg(ap, V9fsString *);
332         flags = va_arg(ap, int);
333         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sd", path, flags);
334         if (retval > 0) {
335             header.size = retval;
336             header.type = T_OPEN;
337         }
338         break;
339     case T_CREATE:
340         path = va_arg(ap, V9fsString *);
341         flags = va_arg(ap, int);
342         mode = va_arg(ap, int);
343         uid = va_arg(ap, int);
344         gid = va_arg(ap, int);
345         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sdddd", path,
346                                     flags, mode, uid, gid);
347         if (retval > 0) {
348             header.size = retval;
349             header.type = T_CREATE;
350         }
351         break;
352     case T_MKNOD:
353         path = va_arg(ap, V9fsString *);
354         mode = va_arg(ap, int);
355         rdev = va_arg(ap, long int);
356         uid = va_arg(ap, int);
357         gid = va_arg(ap, int);
358         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ddsdq",
359                                     uid, gid, path, mode, rdev);
360         if (retval > 0) {
361             header.size = retval;
362             header.type = T_MKNOD;
363         }
364         break;
365     case T_MKDIR:
366         path = va_arg(ap, V9fsString *);
367         mode = va_arg(ap, int);
368         uid = va_arg(ap, int);
369         gid = va_arg(ap, int);
370         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ddsd",
371                                     uid, gid, path, mode);
372         if (retval > 0) {
373             header.size = retval;
374             header.type = T_MKDIR;
375         }
376         break;
377     case T_SYMLINK:
378         oldpath = va_arg(ap, V9fsString *);
379         path = va_arg(ap, V9fsString *);
380         uid = va_arg(ap, int);
381         gid = va_arg(ap, int);
382         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ddss",
383                                     uid, gid, oldpath, path);
384         if (retval > 0) {
385             header.size = retval;
386             header.type = T_SYMLINK;
387         }
388         break;
389     case T_LINK:
390         oldpath = va_arg(ap, V9fsString *);
391         path = va_arg(ap, V9fsString *);
392         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ss",
393                                     oldpath, path);
394         if (retval > 0) {
395             header.size = retval;
396             header.type = T_LINK;
397         }
398         break;
399     case T_LSTAT:
400         path = va_arg(ap, V9fsString *);
401         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "s", path);
402         if (retval > 0) {
403             header.size = retval;
404             header.type = T_LSTAT;
405         }
406         break;
407     case T_READLINK:
408         path = va_arg(ap, V9fsString *);
409         size = va_arg(ap, int);
410         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sd", path, size);
411         if (retval > 0) {
412             header.size = retval;
413             header.type = T_READLINK;
414         }
415         break;
416     case T_STATFS:
417         path = va_arg(ap, V9fsString *);
418         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "s", path);
419         if (retval > 0) {
420             header.size = retval;
421             header.type = T_STATFS;
422         }
423         break;
424     case T_CHMOD:
425         path = va_arg(ap, V9fsString *);
426         mode = va_arg(ap, int);
427         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sd", path, mode);
428         if (retval > 0) {
429             header.size = retval;
430             header.type = T_CHMOD;
431         }
432         break;
433     case T_CHOWN:
434         path = va_arg(ap, V9fsString *);
435         uid = va_arg(ap, int);
436         gid = va_arg(ap, int);
437         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sdd", path, uid, gid);
438         if (retval > 0) {
439             header.size = retval;
440             header.type = T_CHOWN;
441         }
442         break;
443     case T_TRUNCATE:
444         path = va_arg(ap, V9fsString *);
445         offset = va_arg(ap, uint64_t);
446         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sq", path, offset);
447         if (retval > 0) {
448             header.size = retval;
449             header.type = T_TRUNCATE;
450         }
451         break;
452     case T_UTIME:
453         path = va_arg(ap, V9fsString *);
454         spec[0].tv_sec = va_arg(ap, long);
455         spec[0].tv_nsec = va_arg(ap, long);
456         spec[1].tv_sec = va_arg(ap, long);
457         spec[1].tv_nsec = va_arg(ap, long);
458         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sqqqq", path,
459                                     spec[0].tv_sec, spec[1].tv_nsec,
460                                     spec[1].tv_sec, spec[1].tv_nsec);
461         if (retval > 0) {
462             header.size = retval;
463             header.type = T_UTIME;
464         }
465         break;
466     case T_RENAME:
467         oldpath = va_arg(ap, V9fsString *);
468         path = va_arg(ap, V9fsString *);
469         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ss", oldpath, path);
470         if (retval > 0) {
471             header.size = retval;
472             header.type = T_RENAME;
473         }
474         break;
475     case T_REMOVE:
476         path = va_arg(ap, V9fsString *);
477         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "s", path);
478         if (retval > 0) {
479             header.size = retval;
480             header.type = T_REMOVE;
481         }
482         break;
483     case T_LGETXATTR:
484         size = va_arg(ap, int);
485         path = va_arg(ap, V9fsString *);
486         name = va_arg(ap, V9fsString *);
487         retval = proxy_marshal(iovec, PROXY_HDR_SZ,
488                                     "dss", size, path, name);
489         if (retval > 0) {
490             header.size = retval;
491             header.type = T_LGETXATTR;
492         }
493         break;
494     case T_LLISTXATTR:
495         size = va_arg(ap, int);
496         path = va_arg(ap, V9fsString *);
497         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ds", size, path);
498         if (retval > 0) {
499             header.size = retval;
500             header.type = T_LLISTXATTR;
501         }
502         break;
503     case T_LSETXATTR:
504         path = va_arg(ap, V9fsString *);
505         name = va_arg(ap, V9fsString *);
506         value = va_arg(ap, V9fsString *);
507         size = va_arg(ap, int);
508         flags = va_arg(ap, int);
509         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "sssdd",
510                                     path, name, value, size, flags);
511         if (retval > 0) {
512             header.size = retval;
513             header.type = T_LSETXATTR;
514         }
515         break;
516     case T_LREMOVEXATTR:
517         path = va_arg(ap, V9fsString *);
518         name = va_arg(ap, V9fsString *);
519         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "ss", path, name);
520         if (retval > 0) {
521             header.size = retval;
522             header.type = T_LREMOVEXATTR;
523         }
524         break;
525     case T_GETVERSION:
526         path = va_arg(ap, V9fsString *);
527         retval = proxy_marshal(iovec, PROXY_HDR_SZ, "s", path);
528         if (retval > 0) {
529             header.size = retval;
530             header.type = T_GETVERSION;
531         }
532         break;
533     default:
534         error_report("Invalid type %d", type);
535         retval = -EINVAL;
536         break;
537     }
538     va_end(ap);
539 
540     if (retval < 0) {
541         goto err_out;
542     }
543 
544     /* marshal the header details */
545     retval = proxy_marshal(iovec, 0, "dd", header.type, header.size);
546     assert(retval == 4 * 2);
547     header.size += PROXY_HDR_SZ;
548 
549     retval = qemu_write_full(proxy->sockfd, iovec->iov_base, header.size);
550     if (retval != header.size) {
551         goto close_error;
552     }
553 
554     switch (type) {
555     case T_OPEN:
556     case T_CREATE:
557         /*
558          * A file descriptor is returned as response for
559          * T_OPEN,T_CREATE on success
560          */
561         if (v9fs_receivefd(proxy->sockfd, &retval) < 0) {
562             goto close_error;
563         }
564         break;
565     case T_MKNOD:
566     case T_MKDIR:
567     case T_SYMLINK:
568     case T_LINK:
569     case T_CHMOD:
570     case T_CHOWN:
571     case T_RENAME:
572     case T_TRUNCATE:
573     case T_UTIME:
574     case T_REMOVE:
575     case T_LSETXATTR:
576     case T_LREMOVEXATTR:
577         if (v9fs_receive_status(proxy, reply, &retval) < 0) {
578             goto close_error;
579         }
580         break;
581     case T_LSTAT:
582     case T_READLINK:
583     case T_STATFS:
584     case T_GETVERSION:
585         if (v9fs_receive_response(proxy, type, &retval, response) < 0) {
586             goto close_error;
587         }
588         break;
589     case T_LGETXATTR:
590     case T_LLISTXATTR:
591         if (!size) {
592             if (v9fs_receive_status(proxy, reply, &retval) < 0) {
593                 goto close_error;
594             }
595         } else {
596             if (v9fs_receive_response(proxy, type, &retval, response) < 0) {
597                 goto close_error;
598             }
599         }
600         break;
601     }
602 
603 err_out:
604     qemu_mutex_unlock(&proxy->mutex);
605     return retval;
606 
607 close_error:
608     close(proxy->sockfd);
609     proxy->sockfd = -1;
610     qemu_mutex_unlock(&proxy->mutex);
611     return -EIO;
612 }
613 
614 static int proxy_lstat(FsContext *fs_ctx, V9fsPath *fs_path, struct stat *stbuf)
615 {
616     int retval;
617     retval = v9fs_request(fs_ctx->private, T_LSTAT, stbuf, fs_path);
618     if (retval < 0) {
619         errno = -retval;
620         return -1;
621     }
622     return retval;
623 }
624 
625 static ssize_t proxy_readlink(FsContext *fs_ctx, V9fsPath *fs_path,
626                               char *buf, size_t bufsz)
627 {
628     int retval;
629     retval = v9fs_request(fs_ctx->private, T_READLINK, buf, fs_path, bufsz);
630     if (retval < 0) {
631         errno = -retval;
632         return -1;
633     }
634     return strlen(buf);
635 }
636 
637 static int proxy_close(FsContext *ctx, V9fsFidOpenState *fs)
638 {
639     return close(fs->fd);
640 }
641 
642 static int proxy_closedir(FsContext *ctx, V9fsFidOpenState *fs)
643 {
644     return closedir(fs->dir.stream);
645 }
646 
647 static int proxy_open(FsContext *ctx, V9fsPath *fs_path,
648                       int flags, V9fsFidOpenState *fs)
649 {
650     fs->fd = v9fs_request(ctx->private, T_OPEN, NULL, fs_path, flags);
651     if (fs->fd < 0) {
652         errno = -fs->fd;
653         fs->fd = -1;
654     }
655     return fs->fd;
656 }
657 
658 static int proxy_opendir(FsContext *ctx,
659                          V9fsPath *fs_path, V9fsFidOpenState *fs)
660 {
661     int serrno, fd;
662 
663     fs->dir.stream = NULL;
664     fd = v9fs_request(ctx->private, T_OPEN, NULL, fs_path, O_DIRECTORY);
665     if (fd < 0) {
666         errno = -fd;
667         return -1;
668     }
669     fs->dir.stream = fdopendir(fd);
670     if (!fs->dir.stream) {
671         serrno = errno;
672         close(fd);
673         errno = serrno;
674         return -1;
675     }
676     return 0;
677 }
678 
679 static void proxy_rewinddir(FsContext *ctx, V9fsFidOpenState *fs)
680 {
681     rewinddir(fs->dir.stream);
682 }
683 
684 static off_t proxy_telldir(FsContext *ctx, V9fsFidOpenState *fs)
685 {
686     return telldir(fs->dir.stream);
687 }
688 
689 static struct dirent *proxy_readdir(FsContext *ctx, V9fsFidOpenState *fs)
690 {
691     return readdir(fs->dir.stream);
692 }
693 
694 static void proxy_seekdir(FsContext *ctx, V9fsFidOpenState *fs, off_t off)
695 {
696     seekdir(fs->dir.stream, off);
697 }
698 
699 static ssize_t proxy_preadv(FsContext *ctx, V9fsFidOpenState *fs,
700                             const struct iovec *iov,
701                             int iovcnt, off_t offset)
702 {
703     ssize_t ret;
704 #ifdef CONFIG_PREADV
705     ret = preadv(fs->fd, iov, iovcnt, offset);
706 #else
707     ret = lseek(fs->fd, offset, SEEK_SET);
708     if (ret >= 0) {
709         ret = readv(fs->fd, iov, iovcnt);
710     }
711 #endif
712     return ret;
713 }
714 
715 static ssize_t proxy_pwritev(FsContext *ctx, V9fsFidOpenState *fs,
716                              const struct iovec *iov,
717                              int iovcnt, off_t offset)
718 {
719     ssize_t ret;
720 
721 #ifdef CONFIG_PREADV
722     ret = pwritev(fs->fd, iov, iovcnt, offset);
723 #else
724     ret = lseek(fs->fd, offset, SEEK_SET);
725     if (ret >= 0) {
726         ret = writev(fs->fd, iov, iovcnt);
727     }
728 #endif
729 #ifdef CONFIG_SYNC_FILE_RANGE
730     if (ret > 0 && ctx->export_flags & V9FS_IMMEDIATE_WRITEOUT) {
731         /*
732          * Initiate a writeback. This is not a data integrity sync.
733          * We want to ensure that we don't leave dirty pages in the cache
734          * after write when writeout=immediate is sepcified.
735          */
736         sync_file_range(fs->fd, offset, ret,
737                         SYNC_FILE_RANGE_WAIT_BEFORE | SYNC_FILE_RANGE_WRITE);
738     }
739 #endif
740     return ret;
741 }
742 
743 static int proxy_chmod(FsContext *fs_ctx, V9fsPath *fs_path, FsCred *credp)
744 {
745     int retval;
746     retval = v9fs_request(fs_ctx->private, T_CHMOD, NULL, fs_path,
747                           credp->fc_mode);
748     if (retval < 0) {
749         errno = -retval;
750     }
751     return retval;
752 }
753 
754 static int proxy_mknod(FsContext *fs_ctx, V9fsPath *dir_path,
755                        const char *name, FsCred *credp)
756 {
757     int retval;
758     V9fsString fullname;
759 
760     v9fs_string_init(&fullname);
761     v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
762 
763     retval = v9fs_request(fs_ctx->private, T_MKNOD, NULL, &fullname,
764                           credp->fc_mode, credp->fc_rdev,
765                           credp->fc_uid, credp->fc_gid);
766     v9fs_string_free(&fullname);
767     if (retval < 0) {
768         errno = -retval;
769         retval = -1;
770     }
771     return retval;
772 }
773 
774 static int proxy_mkdir(FsContext *fs_ctx, V9fsPath *dir_path,
775                        const char *name, FsCred *credp)
776 {
777     int retval;
778     V9fsString fullname;
779 
780     v9fs_string_init(&fullname);
781     v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
782 
783     retval = v9fs_request(fs_ctx->private, T_MKDIR, NULL, &fullname,
784                           credp->fc_mode, credp->fc_uid, credp->fc_gid);
785     v9fs_string_free(&fullname);
786     if (retval < 0) {
787         errno = -retval;
788         retval = -1;
789     }
790     return retval;
791 }
792 
793 static int proxy_fstat(FsContext *fs_ctx, int fid_type,
794                        V9fsFidOpenState *fs, struct stat *stbuf)
795 {
796     int fd;
797 
798     if (fid_type == P9_FID_DIR) {
799         fd = dirfd(fs->dir.stream);
800     } else {
801         fd = fs->fd;
802     }
803     return fstat(fd, stbuf);
804 }
805 
806 static int proxy_open2(FsContext *fs_ctx, V9fsPath *dir_path, const char *name,
807                        int flags, FsCred *credp, V9fsFidOpenState *fs)
808 {
809     V9fsString fullname;
810 
811     v9fs_string_init(&fullname);
812     v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
813 
814     fs->fd = v9fs_request(fs_ctx->private, T_CREATE, NULL, &fullname, flags,
815                           credp->fc_mode, credp->fc_uid, credp->fc_gid);
816     v9fs_string_free(&fullname);
817     if (fs->fd < 0) {
818         errno = -fs->fd;
819         fs->fd = -1;
820     }
821     return fs->fd;
822 }
823 
824 static int proxy_symlink(FsContext *fs_ctx, const char *oldpath,
825                          V9fsPath *dir_path, const char *name, FsCred *credp)
826 {
827     int retval;
828     V9fsString fullname, target;
829 
830     v9fs_string_init(&fullname);
831     v9fs_string_init(&target);
832 
833     v9fs_string_sprintf(&fullname, "%s/%s", dir_path->data, name);
834     v9fs_string_sprintf(&target, "%s", oldpath);
835 
836     retval = v9fs_request(fs_ctx->private, T_SYMLINK, NULL, &target, &fullname,
837                           credp->fc_uid, credp->fc_gid);
838     v9fs_string_free(&fullname);
839     v9fs_string_free(&target);
840     if (retval < 0) {
841         errno = -retval;
842         retval = -1;
843     }
844     return retval;
845 }
846 
847 static int proxy_link(FsContext *ctx, V9fsPath *oldpath,
848                       V9fsPath *dirpath, const char *name)
849 {
850     int retval;
851     V9fsString newpath;
852 
853     v9fs_string_init(&newpath);
854     v9fs_string_sprintf(&newpath, "%s/%s", dirpath->data, name);
855 
856     retval = v9fs_request(ctx->private, T_LINK, NULL, oldpath, &newpath);
857     v9fs_string_free(&newpath);
858     if (retval < 0) {
859         errno = -retval;
860         retval = -1;
861     }
862     return retval;
863 }
864 
865 static int proxy_truncate(FsContext *ctx, V9fsPath *fs_path, off_t size)
866 {
867     int retval;
868 
869     retval = v9fs_request(ctx->private, T_TRUNCATE, NULL, fs_path, size);
870     if (retval < 0) {
871         errno = -retval;
872         return -1;
873     }
874     return 0;
875 }
876 
877 static int proxy_rename(FsContext *ctx, const char *oldpath,
878                         const char *newpath)
879 {
880     int retval;
881     V9fsString oldname, newname;
882 
883     v9fs_string_init(&oldname);
884     v9fs_string_init(&newname);
885 
886     v9fs_string_sprintf(&oldname, "%s", oldpath);
887     v9fs_string_sprintf(&newname, "%s", newpath);
888     retval = v9fs_request(ctx->private, T_RENAME, NULL, &oldname, &newname);
889     v9fs_string_free(&oldname);
890     v9fs_string_free(&newname);
891     if (retval < 0) {
892         errno = -retval;
893     }
894     return retval;
895 }
896 
897 static int proxy_chown(FsContext *fs_ctx, V9fsPath *fs_path, FsCred *credp)
898 {
899     int retval;
900     retval = v9fs_request(fs_ctx->private, T_CHOWN, NULL, fs_path,
901                           credp->fc_uid, credp->fc_gid);
902     if (retval < 0) {
903         errno = -retval;
904     }
905     return retval;
906 }
907 
908 static int proxy_utimensat(FsContext *s, V9fsPath *fs_path,
909                            const struct timespec *buf)
910 {
911     int retval;
912     retval = v9fs_request(s->private, T_UTIME, NULL, fs_path,
913                           buf[0].tv_sec, buf[0].tv_nsec,
914                           buf[1].tv_sec, buf[1].tv_nsec);
915     if (retval < 0) {
916         errno = -retval;
917     }
918     return retval;
919 }
920 
921 static int proxy_remove(FsContext *ctx, const char *path)
922 {
923     int retval;
924     V9fsString name;
925     v9fs_string_init(&name);
926     v9fs_string_sprintf(&name, "%s", path);
927     retval = v9fs_request(ctx->private, T_REMOVE, NULL, &name);
928     v9fs_string_free(&name);
929     if (retval < 0) {
930         errno = -retval;
931     }
932     return retval;
933 }
934 
935 static int proxy_fsync(FsContext *ctx, int fid_type,
936                        V9fsFidOpenState *fs, int datasync)
937 {
938     int fd;
939 
940     if (fid_type == P9_FID_DIR) {
941         fd = dirfd(fs->dir.stream);
942     } else {
943         fd = fs->fd;
944     }
945 
946     if (datasync) {
947         return qemu_fdatasync(fd);
948     } else {
949         return fsync(fd);
950     }
951 }
952 
953 static int proxy_statfs(FsContext *s, V9fsPath *fs_path, struct statfs *stbuf)
954 {
955     int retval;
956     retval = v9fs_request(s->private, T_STATFS, stbuf, fs_path);
957     if (retval < 0) {
958         errno = -retval;
959         return -1;
960     }
961     return retval;
962 }
963 
964 static ssize_t proxy_lgetxattr(FsContext *ctx, V9fsPath *fs_path,
965                                const char *name, void *value, size_t size)
966 {
967     int retval;
968     V9fsString xname;
969 
970     v9fs_string_init(&xname);
971     v9fs_string_sprintf(&xname, "%s", name);
972     retval = v9fs_request(ctx->private, T_LGETXATTR, value, size, fs_path,
973                           &xname);
974     v9fs_string_free(&xname);
975     if (retval < 0) {
976         errno = -retval;
977     }
978     return retval;
979 }
980 
981 static ssize_t proxy_llistxattr(FsContext *ctx, V9fsPath *fs_path,
982                                 void *value, size_t size)
983 {
984     int retval;
985     retval = v9fs_request(ctx->private, T_LLISTXATTR, value, size, fs_path);
986     if (retval < 0) {
987         errno = -retval;
988     }
989     return retval;
990 }
991 
992 static int proxy_lsetxattr(FsContext *ctx, V9fsPath *fs_path, const char *name,
993                            void *value, size_t size, int flags)
994 {
995     int retval;
996     V9fsString xname, xvalue;
997 
998     v9fs_string_init(&xname);
999     v9fs_string_sprintf(&xname, "%s", name);
1000 
1001     v9fs_string_init(&xvalue);
1002     xvalue.size = size;
1003     xvalue.data = g_malloc(size);
1004     memcpy(xvalue.data, value, size);
1005 
1006     retval = v9fs_request(ctx->private, T_LSETXATTR, value, fs_path, &xname,
1007                           &xvalue, size, flags);
1008     v9fs_string_free(&xname);
1009     v9fs_string_free(&xvalue);
1010     if (retval < 0) {
1011         errno = -retval;
1012     }
1013     return retval;
1014 }
1015 
1016 static int proxy_lremovexattr(FsContext *ctx, V9fsPath *fs_path,
1017                               const char *name)
1018 {
1019     int retval;
1020     V9fsString xname;
1021 
1022     v9fs_string_init(&xname);
1023     v9fs_string_sprintf(&xname, "%s", name);
1024     retval = v9fs_request(ctx->private, T_LREMOVEXATTR, NULL, fs_path, &xname);
1025     v9fs_string_free(&xname);
1026     if (retval < 0) {
1027         errno = -retval;
1028     }
1029     return retval;
1030 }
1031 
1032 static int proxy_name_to_path(FsContext *ctx, V9fsPath *dir_path,
1033                               const char *name, V9fsPath *target)
1034 {
1035     if (dir_path) {
1036         v9fs_path_sprintf(target, "%s/%s", dir_path->data, name);
1037     } else {
1038         v9fs_path_sprintf(target, "%s", name);
1039     }
1040     return 0;
1041 }
1042 
1043 static int proxy_renameat(FsContext *ctx, V9fsPath *olddir,
1044                           const char *old_name, V9fsPath *newdir,
1045                           const char *new_name)
1046 {
1047     int ret;
1048     V9fsString old_full_name, new_full_name;
1049 
1050     v9fs_string_init(&old_full_name);
1051     v9fs_string_init(&new_full_name);
1052 
1053     v9fs_string_sprintf(&old_full_name, "%s/%s", olddir->data, old_name);
1054     v9fs_string_sprintf(&new_full_name, "%s/%s", newdir->data, new_name);
1055 
1056     ret = proxy_rename(ctx, old_full_name.data, new_full_name.data);
1057     v9fs_string_free(&old_full_name);
1058     v9fs_string_free(&new_full_name);
1059     return ret;
1060 }
1061 
1062 static int proxy_unlinkat(FsContext *ctx, V9fsPath *dir,
1063                           const char *name, int flags)
1064 {
1065     int ret;
1066     V9fsString fullname;
1067     v9fs_string_init(&fullname);
1068 
1069     v9fs_string_sprintf(&fullname, "%s/%s", dir->data, name);
1070     ret = proxy_remove(ctx, fullname.data);
1071     v9fs_string_free(&fullname);
1072 
1073     return ret;
1074 }
1075 
1076 static int proxy_ioc_getversion(FsContext *fs_ctx, V9fsPath *path,
1077                                 mode_t st_mode, uint64_t *st_gen)
1078 {
1079     int err;
1080 
1081     /* Do not try to open special files like device nodes, fifos etc
1082      * we can get fd for regular files and directories only
1083      */
1084     if (!S_ISREG(st_mode) && !S_ISDIR(st_mode)) {
1085         errno = ENOTTY;
1086         return -1;
1087     }
1088     err = v9fs_request(fs_ctx->private, T_GETVERSION, st_gen, path);
1089     if (err < 0) {
1090         errno = -err;
1091         err = -1;
1092     }
1093     return err;
1094 }
1095 
1096 static int connect_namedsocket(const char *path, Error **errp)
1097 {
1098     int sockfd;
1099     struct sockaddr_un helper;
1100 
1101     if (strlen(path) >= sizeof(helper.sun_path)) {
1102         error_setg(errp, "socket name too long");
1103         return -1;
1104     }
1105     sockfd = socket(AF_UNIX, SOCK_STREAM, 0);
1106     if (sockfd < 0) {
1107         error_setg_errno(errp, errno, "failed to create client socket");
1108         return -1;
1109     }
1110     strcpy(helper.sun_path, path);
1111     helper.sun_family = AF_UNIX;
1112     if (connect(sockfd, (struct sockaddr *)&helper, sizeof(helper)) < 0) {
1113         error_setg_errno(errp, errno, "failed to connect to '%s'", path);
1114         close(sockfd);
1115         return -1;
1116     }
1117 
1118     /* remove the socket for security reasons */
1119     unlink(path);
1120     return sockfd;
1121 }
1122 
1123 static void error_append_socket_sockfd_hint(Error *const *errp)
1124 {
1125     error_append_hint(errp, "Either specify socket=/some/path where /some/path"
1126                       " points to a listening AF_UNIX socket or sock_fd=fd"
1127                       " where fd is a file descriptor to a connected AF_UNIX"
1128                       " socket\n");
1129 }
1130 
1131 static int proxy_parse_opts(QemuOpts *opts, FsDriverEntry *fs, Error **errp)
1132 {
1133     const char *socket = qemu_opt_get(opts, "socket");
1134     const char *sock_fd = qemu_opt_get(opts, "sock_fd");
1135 
1136     if (!socket && !sock_fd) {
1137         error_setg(errp, "both socket and sock_fd properties are missing");
1138         error_append_socket_sockfd_hint(errp);
1139         return -1;
1140     }
1141     if (socket && sock_fd) {
1142         error_setg(errp, "both socket and sock_fd properties are set");
1143         error_append_socket_sockfd_hint(errp);
1144         return -1;
1145     }
1146     if (socket) {
1147         fs->path = g_strdup(socket);
1148         fs->export_flags |= V9FS_PROXY_SOCK_NAME;
1149     } else {
1150         fs->path = g_strdup(sock_fd);
1151         fs->export_flags |= V9FS_PROXY_SOCK_FD;
1152     }
1153     return 0;
1154 }
1155 
1156 static int proxy_init(FsContext *ctx, Error **errp)
1157 {
1158     V9fsProxy *proxy = g_malloc(sizeof(V9fsProxy));
1159     int sock_id;
1160 
1161     if (ctx->export_flags & V9FS_PROXY_SOCK_NAME) {
1162         sock_id = connect_namedsocket(ctx->fs_root, errp);
1163     } else {
1164         sock_id = atoi(ctx->fs_root);
1165         if (sock_id < 0) {
1166             error_setg(errp, "socket descriptor not initialized");
1167         }
1168     }
1169     if (sock_id < 0) {
1170         g_free(proxy);
1171         return -1;
1172     }
1173     g_free(ctx->fs_root);
1174     ctx->fs_root = NULL;
1175 
1176     proxy->in_iovec.iov_base  = g_malloc(PROXY_MAX_IO_SZ + PROXY_HDR_SZ);
1177     proxy->in_iovec.iov_len   = PROXY_MAX_IO_SZ + PROXY_HDR_SZ;
1178     proxy->out_iovec.iov_base = g_malloc(PROXY_MAX_IO_SZ + PROXY_HDR_SZ);
1179     proxy->out_iovec.iov_len  = PROXY_MAX_IO_SZ + PROXY_HDR_SZ;
1180 
1181     ctx->private = proxy;
1182     proxy->sockfd = sock_id;
1183     qemu_mutex_init(&proxy->mutex);
1184 
1185     ctx->export_flags |= V9FS_PATHNAME_FSCONTEXT;
1186     ctx->exops.get_st_gen = proxy_ioc_getversion;
1187     return 0;
1188 }
1189 
1190 static void proxy_cleanup(FsContext *ctx)
1191 {
1192     V9fsProxy *proxy = ctx->private;
1193 
1194     if (!proxy) {
1195         return;
1196     }
1197 
1198     g_free(proxy->out_iovec.iov_base);
1199     g_free(proxy->in_iovec.iov_base);
1200     if (ctx->export_flags & V9FS_PROXY_SOCK_NAME) {
1201         close(proxy->sockfd);
1202     }
1203     g_free(proxy);
1204 }
1205 
1206 FileOperations proxy_ops = {
1207     .parse_opts   = proxy_parse_opts,
1208     .init         = proxy_init,
1209     .cleanup      = proxy_cleanup,
1210     .lstat        = proxy_lstat,
1211     .readlink     = proxy_readlink,
1212     .close        = proxy_close,
1213     .closedir     = proxy_closedir,
1214     .open         = proxy_open,
1215     .opendir      = proxy_opendir,
1216     .rewinddir    = proxy_rewinddir,
1217     .telldir      = proxy_telldir,
1218     .readdir      = proxy_readdir,
1219     .seekdir      = proxy_seekdir,
1220     .preadv       = proxy_preadv,
1221     .pwritev      = proxy_pwritev,
1222     .chmod        = proxy_chmod,
1223     .mknod        = proxy_mknod,
1224     .mkdir        = proxy_mkdir,
1225     .fstat        = proxy_fstat,
1226     .open2        = proxy_open2,
1227     .symlink      = proxy_symlink,
1228     .link         = proxy_link,
1229     .truncate     = proxy_truncate,
1230     .rename       = proxy_rename,
1231     .chown        = proxy_chown,
1232     .utimensat    = proxy_utimensat,
1233     .remove       = proxy_remove,
1234     .fsync        = proxy_fsync,
1235     .statfs       = proxy_statfs,
1236     .lgetxattr    = proxy_lgetxattr,
1237     .llistxattr   = proxy_llistxattr,
1238     .lsetxattr    = proxy_lsetxattr,
1239     .lremovexattr = proxy_lremovexattr,
1240     .name_to_path = proxy_name_to_path,
1241     .renameat     = proxy_renameat,
1242     .unlinkat     = proxy_unlinkat,
1243 };
1244