xref: /qemu/fsdev/virtfs-proxy-helper.c (revision e615c157)
1 /*
2  * Helper for QEMU Proxy FS Driver
3  * Copyright IBM, Corp. 2011
4  *
5  * Authors:
6  * M. Mohan Kumar <mohan@in.ibm.com>
7  *
8  * This work is licensed under the terms of the GNU GPL, version 2. See
9  * the COPYING file in the top-level directory.
10  */
11 
12 #include "qemu/osdep.h"
13 #include <sys/resource.h>
14 #include <getopt.h>
15 #include <syslog.h>
16 #include <sys/fsuid.h>
17 #include <sys/vfs.h>
18 #include <sys/ioctl.h>
19 #include <linux/fs.h>
20 #ifdef CONFIG_LINUX_MAGIC_H
21 #include <linux/magic.h>
22 #endif
23 #include <cap-ng.h>
24 #include "qemu-common.h"
25 #include "qemu/sockets.h"
26 #include "qemu/xattr.h"
27 #include "9p-iov-marshal.h"
28 #include "hw/9pfs/9p-proxy.h"
29 #include "fsdev/9p-iov-marshal.h"
30 
31 #define PROGNAME "virtfs-proxy-helper"
32 
33 #ifndef XFS_SUPER_MAGIC
34 #define XFS_SUPER_MAGIC  0x58465342
35 #endif
36 #ifndef EXT2_SUPER_MAGIC
37 #define EXT2_SUPER_MAGIC 0xEF53
38 #endif
39 #ifndef REISERFS_SUPER_MAGIC
40 #define REISERFS_SUPER_MAGIC 0x52654973
41 #endif
42 #ifndef BTRFS_SUPER_MAGIC
43 #define BTRFS_SUPER_MAGIC 0x9123683E
44 #endif
45 
46 static struct option helper_opts[] = {
47     {"fd", required_argument, NULL, 'f'},
48     {"path", required_argument, NULL, 'p'},
49     {"nodaemon", no_argument, NULL, 'n'},
50     {"socket", required_argument, NULL, 's'},
51     {"uid", required_argument, NULL, 'u'},
52     {"gid", required_argument, NULL, 'g'},
53     {},
54 };
55 
56 static bool is_daemon;
57 static bool get_version; /* IOC getversion IOCTL supported */
58 static char *prog_name;
59 
60 static void GCC_FMT_ATTR(2, 3) do_log(int loglevel, const char *format, ...)
61 {
62     va_list ap;
63 
64     va_start(ap, format);
65     if (is_daemon) {
66         vsyslog(LOG_CRIT, format, ap);
67     } else {
68         vfprintf(stderr, format, ap);
69     }
70     va_end(ap);
71 }
72 
73 static void do_perror(const char *string)
74 {
75     if (is_daemon) {
76         syslog(LOG_CRIT, "%s:%s", string, strerror(errno));
77     } else {
78         fprintf(stderr, "%s:%s\n", string, strerror(errno));
79     }
80 }
81 
82 static int init_capabilities(void)
83 {
84     /* helper needs following capabilities only */
85     int cap_list[] = {
86         CAP_CHOWN,
87         CAP_DAC_OVERRIDE,
88         CAP_FOWNER,
89         CAP_FSETID,
90         CAP_SETGID,
91         CAP_MKNOD,
92         CAP_SETUID,
93     };
94     int i;
95 
96     capng_clear(CAPNG_SELECT_BOTH);
97     for (i = 0; i < ARRAY_SIZE(cap_list); i++) {
98         if (capng_update(CAPNG_ADD, CAPNG_EFFECTIVE | CAPNG_PERMITTED,
99                          cap_list[i]) < 0) {
100             do_perror("capng_update");
101             return -1;
102         }
103     }
104     if (capng_apply(CAPNG_SELECT_BOTH) < 0) {
105         do_perror("capng_apply");
106         return -1;
107     }
108 
109     /* Prepare effective set for setugid.  */
110     for (i = 0; i < ARRAY_SIZE(cap_list); i++) {
111         if (cap_list[i] == CAP_DAC_OVERRIDE) {
112             continue;
113         }
114 
115         if (capng_update(CAPNG_DROP, CAPNG_EFFECTIVE,
116                          cap_list[i]) < 0) {
117             do_perror("capng_update");
118             return -1;
119         }
120     }
121     return 0;
122 }
123 
124 static int socket_read(int sockfd, void *buff, ssize_t size)
125 {
126     ssize_t retval, total = 0;
127 
128     while (size) {
129         retval = read(sockfd, buff, size);
130         if (retval == 0) {
131             return -EIO;
132         }
133         if (retval < 0) {
134             if (errno == EINTR) {
135                 continue;
136             }
137             return -errno;
138         }
139         size -= retval;
140         buff += retval;
141         total += retval;
142     }
143     return total;
144 }
145 
146 static int socket_write(int sockfd, void *buff, ssize_t size)
147 {
148     ssize_t retval, total = 0;
149 
150     while (size) {
151         retval = write(sockfd, buff, size);
152         if (retval < 0) {
153             if (errno == EINTR) {
154                 continue;
155             }
156             return -errno;
157         }
158         size -= retval;
159         buff += retval;
160         total += retval;
161     }
162     return total;
163 }
164 
165 static int read_request(int sockfd, struct iovec *iovec, ProxyHeader *header)
166 {
167     int retval;
168 
169     /*
170      * read the request header.
171      */
172     iovec->iov_len = 0;
173     retval = socket_read(sockfd, iovec->iov_base, PROXY_HDR_SZ);
174     if (retval < 0) {
175         return retval;
176     }
177     iovec->iov_len = PROXY_HDR_SZ;
178     retval = proxy_unmarshal(iovec, 0, "dd", &header->type, &header->size);
179     if (retval < 0) {
180         return retval;
181     }
182     /*
183      * We can't process message.size > PROXY_MAX_IO_SZ.
184      * Treat it as fatal error
185      */
186     if (header->size > PROXY_MAX_IO_SZ) {
187         return -ENOBUFS;
188     }
189     retval = socket_read(sockfd, iovec->iov_base + PROXY_HDR_SZ, header->size);
190     if (retval < 0) {
191         return retval;
192     }
193     iovec->iov_len += header->size;
194     return 0;
195 }
196 
197 static int send_fd(int sockfd, int fd)
198 {
199     struct msghdr msg;
200     struct iovec iov;
201     int retval, data;
202     struct cmsghdr *cmsg;
203     union MsgControl msg_control;
204 
205     iov.iov_base = &data;
206     iov.iov_len = sizeof(data);
207 
208     memset(&msg, 0, sizeof(msg));
209     msg.msg_iov = &iov;
210     msg.msg_iovlen = 1;
211     /* No ancillary data on error */
212     if (fd < 0) {
213         /* fd is really negative errno if the request failed  */
214         data = fd;
215     } else {
216         data = V9FS_FD_VALID;
217         msg.msg_control = &msg_control;
218         msg.msg_controllen = sizeof(msg_control);
219 
220         cmsg = &msg_control.cmsg;
221         cmsg->cmsg_len = CMSG_LEN(sizeof(fd));
222         cmsg->cmsg_level = SOL_SOCKET;
223         cmsg->cmsg_type = SCM_RIGHTS;
224         memcpy(CMSG_DATA(cmsg), &fd, sizeof(fd));
225     }
226 
227     do {
228         retval = sendmsg(sockfd, &msg, 0);
229     } while (retval < 0 && errno == EINTR);
230     if (fd >= 0) {
231         close(fd);
232     }
233     if (retval < 0) {
234         return retval;
235     }
236     return 0;
237 }
238 
239 static int send_status(int sockfd, struct iovec *iovec, int status)
240 {
241     ProxyHeader header;
242     int retval, msg_size;
243 
244     if (status < 0) {
245         header.type = T_ERROR;
246     } else {
247         header.type = T_SUCCESS;
248     }
249     header.size = sizeof(status);
250     /*
251      * marshal the return status. We don't check error.
252      * because we are sure we have enough space for the status
253      */
254     msg_size = proxy_marshal(iovec, 0, "ddd", header.type,
255                              header.size, status);
256     if (msg_size < 0) {
257         return msg_size;
258     }
259     retval = socket_write(sockfd, iovec->iov_base, msg_size);
260     if (retval < 0) {
261         return retval;
262     }
263     return 0;
264 }
265 
266 /*
267  * from man 7 capabilities, section
268  * Effect of User ID Changes on Capabilities:
269  * If the effective user ID is changed from nonzero to 0, then the permitted
270  * set is copied to the effective set.  If the effective user ID is changed
271  * from 0 to nonzero, then all capabilities are are cleared from the effective
272  * set.
273  *
274  * The setfsuid/setfsgid man pages warn that changing the effective user ID may
275  * expose the program to unwanted signals, but this is not true anymore: for an
276  * unprivileged (without CAP_KILL) program to send a signal, the real or
277  * effective user ID of the sending process must equal the real or saved user
278  * ID of the target process.  Even when dropping privileges, it is enough to
279  * keep the saved UID to a "privileged" value and virtfs-proxy-helper won't
280  * be exposed to signals.  So just use setresuid/setresgid.
281  */
282 static int setugid(int uid, int gid, int *suid, int *sgid)
283 {
284     int retval;
285 
286     *suid = geteuid();
287     *sgid = getegid();
288 
289     if (setresgid(-1, gid, *sgid) == -1) {
290         retval = -errno;
291         goto err_out;
292     }
293 
294     if (setresuid(-1, uid, *suid) == -1) {
295         retval = -errno;
296         goto err_sgid;
297     }
298 
299     if (uid == 0 && gid == 0) {
300         /* Linux has already copied the permitted set to the effective set.  */
301         return 0;
302     }
303 
304     /*
305      * All capabilities have been cleared from the effective set.  However
306      * we still need DAC_OVERRIDE because we don't change supplementary
307      * group ids, and hence may be subject to DAC rules.  init_capabilities
308      * left the set of capabilities that we want in libcap-ng's state.
309      */
310     if (capng_apply(CAPNG_SELECT_CAPS) < 0) {
311         retval = -errno;
312         do_perror("capng_apply");
313         goto err_suid;
314     }
315     return 0;
316 
317 err_suid:
318     if (setresuid(-1, *suid, *suid) == -1) {
319         abort();
320     }
321 err_sgid:
322     if (setresgid(-1, *sgid, *sgid) == -1) {
323         abort();
324     }
325 err_out:
326     return retval;
327 }
328 
329 /*
330  * This is used to reset the ugid back with the saved values
331  * There is nothing much we can do checking error values here.
332  */
333 static void resetugid(int suid, int sgid)
334 {
335     if (setresgid(-1, sgid, sgid) == -1) {
336         abort();
337     }
338     if (setresuid(-1, suid, suid) == -1) {
339         abort();
340     }
341 }
342 
343 /*
344  * send response in two parts
345  * 1) ProxyHeader
346  * 2) Response or error status
347  * This function should be called with marshaled response
348  * send_response constructs header part and error part only.
349  * send response sends {ProxyHeader,Response} if the request was success
350  * otherwise sends {ProxyHeader,error status}
351  */
352 static int send_response(int sock, struct iovec *iovec, int size)
353 {
354     int retval;
355     ProxyHeader header;
356 
357     /*
358      * If response size exceeds available iovec->iov_len,
359      * we return ENOBUFS
360      */
361     if (size > PROXY_MAX_IO_SZ) {
362         size = -ENOBUFS;
363     }
364 
365     if (size < 0) {
366         /*
367          * In case of error we would not have got the error encoded
368          * already so encode the error here.
369          */
370         header.type = T_ERROR;
371         header.size = sizeof(size);
372         proxy_marshal(iovec, PROXY_HDR_SZ, "d", size);
373     } else {
374         header.type = T_SUCCESS;
375         header.size = size;
376     }
377     proxy_marshal(iovec, 0, "dd", header.type, header.size);
378     retval = socket_write(sock, iovec->iov_base, header.size + PROXY_HDR_SZ);
379     if (retval < 0) {
380         return retval;
381     }
382     return 0;
383 }
384 
385 /*
386  * gets generation number
387  * returns -errno on failure and sizeof(generation number) on success
388  */
389 static int do_getversion(struct iovec *iovec, struct iovec *out_iovec)
390 {
391     uint64_t version;
392     int retval = -ENOTTY;
393 #ifdef FS_IOC_GETVERSION
394     int fd;
395     V9fsString path;
396 #endif
397 
398 
399     /* no need to issue ioctl */
400     if (!get_version) {
401         version = 0;
402         retval = proxy_marshal(out_iovec, PROXY_HDR_SZ, "q", version);
403         return retval;
404     }
405 #ifdef FS_IOC_GETVERSION
406     retval = proxy_unmarshal(iovec, PROXY_HDR_SZ, "s", &path);
407     if (retval < 0) {
408         return retval;
409     }
410 
411     fd = open(path.data, O_RDONLY);
412     if (fd < 0) {
413         retval = -errno;
414         goto err_out;
415     }
416     if (ioctl(fd, FS_IOC_GETVERSION, &version) < 0) {
417         retval = -errno;
418     } else {
419         retval = proxy_marshal(out_iovec, PROXY_HDR_SZ, "q", version);
420     }
421     close(fd);
422 err_out:
423     v9fs_string_free(&path);
424 #endif
425     return retval;
426 }
427 
428 static int do_getxattr(int type, struct iovec *iovec, struct iovec *out_iovec)
429 {
430     int size = 0, offset, retval;
431     V9fsString path, name, xattr;
432 
433     v9fs_string_init(&xattr);
434     v9fs_string_init(&path);
435     retval = proxy_unmarshal(iovec, PROXY_HDR_SZ, "ds", &size, &path);
436     if (retval < 0) {
437         return retval;
438     }
439     offset = PROXY_HDR_SZ + retval;
440 
441     if (size) {
442         xattr.data = g_malloc(size);
443         xattr.size = size;
444     }
445     switch (type) {
446     case T_LGETXATTR:
447         v9fs_string_init(&name);
448         retval = proxy_unmarshal(iovec, offset, "s", &name);
449         if (retval > 0) {
450             retval = lgetxattr(path.data, name.data, xattr.data, size);
451             if (retval < 0) {
452                 retval = -errno;
453             } else {
454                 xattr.size = retval;
455             }
456         }
457         v9fs_string_free(&name);
458         break;
459     case T_LLISTXATTR:
460         retval = llistxattr(path.data, xattr.data, size);
461         if (retval < 0) {
462             retval = -errno;
463         } else {
464             xattr.size = retval;
465         }
466         break;
467     }
468     if (retval < 0) {
469         goto err_out;
470     }
471 
472     if (!size) {
473         proxy_marshal(out_iovec, PROXY_HDR_SZ, "d", retval);
474         retval = sizeof(retval);
475     } else {
476         retval = proxy_marshal(out_iovec, PROXY_HDR_SZ, "s", &xattr);
477     }
478 err_out:
479     v9fs_string_free(&xattr);
480     v9fs_string_free(&path);
481     return retval;
482 }
483 
484 static void stat_to_prstat(ProxyStat *pr_stat, struct stat *stat)
485 {
486     memset(pr_stat, 0, sizeof(*pr_stat));
487     pr_stat->st_dev = stat->st_dev;
488     pr_stat->st_ino = stat->st_ino;
489     pr_stat->st_nlink = stat->st_nlink;
490     pr_stat->st_mode = stat->st_mode;
491     pr_stat->st_uid = stat->st_uid;
492     pr_stat->st_gid = stat->st_gid;
493     pr_stat->st_rdev = stat->st_rdev;
494     pr_stat->st_size = stat->st_size;
495     pr_stat->st_blksize = stat->st_blksize;
496     pr_stat->st_blocks = stat->st_blocks;
497     pr_stat->st_atim_sec = stat->st_atim.tv_sec;
498     pr_stat->st_atim_nsec = stat->st_atim.tv_nsec;
499     pr_stat->st_mtim_sec = stat->st_mtim.tv_sec;
500     pr_stat->st_mtim_nsec = stat->st_mtim.tv_nsec;
501     pr_stat->st_ctim_sec = stat->st_ctim.tv_sec;
502     pr_stat->st_ctim_nsec = stat->st_ctim.tv_nsec;
503 }
504 
505 static void statfs_to_prstatfs(ProxyStatFS *pr_stfs, struct statfs *stfs)
506 {
507     memset(pr_stfs, 0, sizeof(*pr_stfs));
508     pr_stfs->f_type = stfs->f_type;
509     pr_stfs->f_bsize = stfs->f_bsize;
510     pr_stfs->f_blocks = stfs->f_blocks;
511     pr_stfs->f_bfree = stfs->f_bfree;
512     pr_stfs->f_bavail = stfs->f_bavail;
513     pr_stfs->f_files = stfs->f_files;
514     pr_stfs->f_ffree = stfs->f_ffree;
515     pr_stfs->f_fsid[0] = stfs->f_fsid.__val[0];
516     pr_stfs->f_fsid[1] = stfs->f_fsid.__val[1];
517     pr_stfs->f_namelen = stfs->f_namelen;
518     pr_stfs->f_frsize = stfs->f_frsize;
519 }
520 
521 /*
522  * Gets stat/statfs information and packs in out_iovec structure
523  * on success returns number of bytes packed in out_iovec struture
524  * otherwise returns -errno
525  */
526 static int do_stat(int type, struct iovec *iovec, struct iovec *out_iovec)
527 {
528     int retval;
529     V9fsString path;
530     ProxyStat pr_stat;
531     ProxyStatFS pr_stfs;
532     struct stat st_buf;
533     struct statfs stfs_buf;
534 
535     v9fs_string_init(&path);
536     retval = proxy_unmarshal(iovec, PROXY_HDR_SZ, "s", &path);
537     if (retval < 0) {
538         return retval;
539     }
540 
541     switch (type) {
542     case T_LSTAT:
543         retval = lstat(path.data, &st_buf);
544         if (retval < 0) {
545             retval = -errno;
546         } else {
547             stat_to_prstat(&pr_stat, &st_buf);
548             retval = proxy_marshal(out_iovec, PROXY_HDR_SZ,
549                                    "qqqdddqqqqqqqqqq", pr_stat.st_dev,
550                                    pr_stat.st_ino, pr_stat.st_nlink,
551                                    pr_stat.st_mode, pr_stat.st_uid,
552                                    pr_stat.st_gid, pr_stat.st_rdev,
553                                    pr_stat.st_size, pr_stat.st_blksize,
554                                    pr_stat.st_blocks,
555                                    pr_stat.st_atim_sec, pr_stat.st_atim_nsec,
556                                    pr_stat.st_mtim_sec, pr_stat.st_mtim_nsec,
557                                    pr_stat.st_ctim_sec, pr_stat.st_ctim_nsec);
558         }
559         break;
560     case T_STATFS:
561         retval = statfs(path.data, &stfs_buf);
562         if (retval < 0) {
563             retval = -errno;
564         } else {
565             statfs_to_prstatfs(&pr_stfs, &stfs_buf);
566             retval = proxy_marshal(out_iovec, PROXY_HDR_SZ,
567                                    "qqqqqqqqqqq", pr_stfs.f_type,
568                                    pr_stfs.f_bsize, pr_stfs.f_blocks,
569                                    pr_stfs.f_bfree, pr_stfs.f_bavail,
570                                    pr_stfs.f_files, pr_stfs.f_ffree,
571                                    pr_stfs.f_fsid[0], pr_stfs.f_fsid[1],
572                                    pr_stfs.f_namelen, pr_stfs.f_frsize);
573         }
574         break;
575     }
576     v9fs_string_free(&path);
577     return retval;
578 }
579 
580 static int do_readlink(struct iovec *iovec, struct iovec *out_iovec)
581 {
582     char *buffer;
583     int size, retval;
584     V9fsString target, path;
585 
586     v9fs_string_init(&path);
587     retval = proxy_unmarshal(iovec, PROXY_HDR_SZ, "sd", &path, &size);
588     if (retval < 0) {
589         v9fs_string_free(&path);
590         return retval;
591     }
592     buffer = g_malloc(size);
593     v9fs_string_init(&target);
594     retval = readlink(path.data, buffer, size - 1);
595     if (retval > 0) {
596         buffer[retval] = '\0';
597         v9fs_string_sprintf(&target, "%s", buffer);
598         retval = proxy_marshal(out_iovec, PROXY_HDR_SZ, "s", &target);
599     } else {
600         retval = -errno;
601     }
602     g_free(buffer);
603     v9fs_string_free(&target);
604     v9fs_string_free(&path);
605     return retval;
606 }
607 
608 /*
609  * create other filesystem objects and send 0 on success
610  * return -errno on error
611  */
612 static int do_create_others(int type, struct iovec *iovec)
613 {
614     dev_t rdev;
615     int retval = 0;
616     int offset = PROXY_HDR_SZ;
617     V9fsString oldpath, path;
618     int mode, uid, gid, cur_uid, cur_gid;
619 
620     v9fs_string_init(&path);
621     v9fs_string_init(&oldpath);
622 
623     retval = proxy_unmarshal(iovec, offset, "dd", &uid, &gid);
624     if (retval < 0) {
625         return retval;
626     }
627     offset += retval;
628     retval = setugid(uid, gid, &cur_uid, &cur_gid);
629     if (retval < 0) {
630         goto unmarshal_err_out;
631     }
632     switch (type) {
633     case T_MKNOD:
634         retval = proxy_unmarshal(iovec, offset, "sdq", &path, &mode, &rdev);
635         if (retval < 0) {
636             goto err_out;
637         }
638         retval = mknod(path.data, mode, rdev);
639         break;
640     case T_MKDIR:
641         retval = proxy_unmarshal(iovec, offset, "sd", &path, &mode);
642         if (retval < 0) {
643             goto err_out;
644         }
645         retval = mkdir(path.data, mode);
646         break;
647     case T_SYMLINK:
648         retval = proxy_unmarshal(iovec, offset, "ss", &oldpath, &path);
649         if (retval < 0) {
650             goto err_out;
651         }
652         retval = symlink(oldpath.data, path.data);
653         break;
654     }
655     if (retval < 0) {
656         retval = -errno;
657     }
658 
659 err_out:
660     resetugid(cur_uid, cur_gid);
661 unmarshal_err_out:
662     v9fs_string_free(&path);
663     v9fs_string_free(&oldpath);
664     return retval;
665 }
666 
667 /*
668  * create a file and send fd on success
669  * return -errno on error
670  */
671 static int do_create(struct iovec *iovec)
672 {
673     int ret;
674     V9fsString path;
675     int flags, mode, uid, gid, cur_uid, cur_gid;
676 
677     v9fs_string_init(&path);
678     ret = proxy_unmarshal(iovec, PROXY_HDR_SZ, "sdddd",
679                           &path, &flags, &mode, &uid, &gid);
680     if (ret < 0) {
681         goto unmarshal_err_out;
682     }
683     ret = setugid(uid, gid, &cur_uid, &cur_gid);
684     if (ret < 0) {
685         goto unmarshal_err_out;
686     }
687     ret = open(path.data, flags, mode);
688     if (ret < 0) {
689         ret = -errno;
690     }
691 
692     resetugid(cur_uid, cur_gid);
693 unmarshal_err_out:
694     v9fs_string_free(&path);
695     return ret;
696 }
697 
698 /*
699  * open a file and send fd on success
700  * return -errno on error
701  */
702 static int do_open(struct iovec *iovec)
703 {
704     int flags, ret;
705     V9fsString path;
706 
707     v9fs_string_init(&path);
708     ret = proxy_unmarshal(iovec, PROXY_HDR_SZ, "sd", &path, &flags);
709     if (ret < 0) {
710         goto err_out;
711     }
712     ret = open(path.data, flags);
713     if (ret < 0) {
714         ret = -errno;
715     }
716 err_out:
717     v9fs_string_free(&path);
718     return ret;
719 }
720 
721 /* create unix domain socket and return the descriptor */
722 static int proxy_socket(const char *path, uid_t uid, gid_t gid)
723 {
724     int sock, client;
725     struct sockaddr_un proxy, qemu;
726     socklen_t size;
727 
728     /* requested socket already exists, refuse to start */
729     if (!access(path, F_OK)) {
730         do_log(LOG_CRIT, "socket already exists\n");
731         return -1;
732     }
733 
734     if (strlen(path) >= sizeof(proxy.sun_path)) {
735         do_log(LOG_CRIT, "UNIX domain socket path exceeds %zu characters\n",
736                sizeof(proxy.sun_path));
737         return -1;
738     }
739 
740     sock = socket(AF_UNIX, SOCK_STREAM, 0);
741     if (sock < 0) {
742         do_perror("socket");
743         return -1;
744     }
745 
746     /* mask other part of mode bits */
747     umask(7);
748 
749     proxy.sun_family = AF_UNIX;
750     strcpy(proxy.sun_path, path);
751     if (bind(sock, (struct sockaddr *)&proxy,
752             sizeof(struct sockaddr_un)) < 0) {
753         do_perror("bind");
754         goto error;
755     }
756     if (chown(proxy.sun_path, uid, gid) < 0) {
757         do_perror("chown");
758         goto error;
759     }
760     if (listen(sock, 1) < 0) {
761         do_perror("listen");
762         goto error;
763     }
764 
765     size = sizeof(qemu);
766     client = accept(sock, (struct sockaddr *)&qemu, &size);
767     if (client < 0) {
768         do_perror("accept");
769         goto error;
770     }
771     close(sock);
772     return client;
773 
774 error:
775     close(sock);
776     return -1;
777 }
778 
779 static void usage(void)
780 {
781     fprintf(stderr, "usage: %s\n"
782             " -p|--path <path> 9p path to export\n"
783             " {-f|--fd <socket-descriptor>} socket file descriptor to be used\n"
784             " {-s|--socket <socketname> socket file used for communication\n"
785             " \t-u|--uid <uid> -g|--gid <gid>} - uid:gid combination to give "
786             " access to this socket\n"
787             " \tNote: -s & -f can not be used together\n"
788             " [-n|--nodaemon] Run as a normal program\n",
789             prog_name);
790 }
791 
792 static int process_reply(int sock, int type,
793                          struct iovec *out_iovec, int retval)
794 {
795     switch (type) {
796     case T_OPEN:
797     case T_CREATE:
798         if (send_fd(sock, retval) < 0) {
799             return -1;
800         }
801         break;
802     case T_MKNOD:
803     case T_MKDIR:
804     case T_SYMLINK:
805     case T_LINK:
806     case T_CHMOD:
807     case T_CHOWN:
808     case T_TRUNCATE:
809     case T_UTIME:
810     case T_RENAME:
811     case T_REMOVE:
812     case T_LSETXATTR:
813     case T_LREMOVEXATTR:
814         if (send_status(sock, out_iovec, retval) < 0) {
815             return -1;
816         }
817         break;
818     case T_LSTAT:
819     case T_STATFS:
820     case T_READLINK:
821     case T_LGETXATTR:
822     case T_LLISTXATTR:
823     case T_GETVERSION:
824         if (send_response(sock, out_iovec, retval) < 0) {
825             return -1;
826         }
827         break;
828     default:
829         return -1;
830         break;
831     }
832     return 0;
833 }
834 
835 static int process_requests(int sock)
836 {
837     int flags;
838     int size = 0;
839     int retval = 0;
840     uint64_t offset;
841     ProxyHeader header;
842     int mode, uid, gid;
843     V9fsString name, value;
844     struct timespec spec[2];
845     V9fsString oldpath, path;
846     struct iovec in_iovec, out_iovec;
847 
848     in_iovec.iov_base  = g_malloc(PROXY_MAX_IO_SZ + PROXY_HDR_SZ);
849     in_iovec.iov_len   = PROXY_MAX_IO_SZ + PROXY_HDR_SZ;
850     out_iovec.iov_base = g_malloc(PROXY_MAX_IO_SZ + PROXY_HDR_SZ);
851     out_iovec.iov_len  = PROXY_MAX_IO_SZ + PROXY_HDR_SZ;
852 
853     while (1) {
854         /*
855          * initialize the header type, so that we send
856          * response to proper request type.
857          */
858         header.type = 0;
859         retval = read_request(sock, &in_iovec, &header);
860         if (retval < 0) {
861             goto err_out;
862         }
863 
864         switch (header.type) {
865         case T_OPEN:
866             retval = do_open(&in_iovec);
867             break;
868         case T_CREATE:
869             retval = do_create(&in_iovec);
870             break;
871         case T_MKNOD:
872         case T_MKDIR:
873         case T_SYMLINK:
874             retval = do_create_others(header.type, &in_iovec);
875             break;
876         case T_LINK:
877             v9fs_string_init(&path);
878             v9fs_string_init(&oldpath);
879             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ,
880                                      "ss", &oldpath, &path);
881             if (retval > 0) {
882                 retval = link(oldpath.data, path.data);
883                 if (retval < 0) {
884                     retval = -errno;
885                 }
886             }
887             v9fs_string_free(&oldpath);
888             v9fs_string_free(&path);
889             break;
890         case T_LSTAT:
891         case T_STATFS:
892             retval = do_stat(header.type, &in_iovec, &out_iovec);
893             break;
894         case T_READLINK:
895             retval = do_readlink(&in_iovec, &out_iovec);
896             break;
897         case T_CHMOD:
898             v9fs_string_init(&path);
899             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ,
900                                      "sd", &path, &mode);
901             if (retval > 0) {
902                 retval = chmod(path.data, mode);
903                 if (retval < 0) {
904                     retval = -errno;
905                 }
906             }
907             v9fs_string_free(&path);
908             break;
909         case T_CHOWN:
910             v9fs_string_init(&path);
911             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "sdd", &path,
912                                      &uid, &gid);
913             if (retval > 0) {
914                 retval = lchown(path.data, uid, gid);
915                 if (retval < 0) {
916                     retval = -errno;
917                 }
918             }
919             v9fs_string_free(&path);
920             break;
921         case T_TRUNCATE:
922             v9fs_string_init(&path);
923             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "sq",
924                                      &path, &offset);
925             if (retval > 0) {
926                 retval = truncate(path.data, offset);
927                 if (retval < 0) {
928                     retval = -errno;
929                 }
930             }
931             v9fs_string_free(&path);
932             break;
933         case T_UTIME:
934             v9fs_string_init(&path);
935             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "sqqqq", &path,
936                                      &spec[0].tv_sec, &spec[0].tv_nsec,
937                                      &spec[1].tv_sec, &spec[1].tv_nsec);
938             if (retval > 0) {
939                 retval = utimensat(AT_FDCWD, path.data, spec,
940                                    AT_SYMLINK_NOFOLLOW);
941                 if (retval < 0) {
942                     retval = -errno;
943                 }
944             }
945             v9fs_string_free(&path);
946             break;
947         case T_RENAME:
948             v9fs_string_init(&path);
949             v9fs_string_init(&oldpath);
950             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ,
951                                      "ss", &oldpath, &path);
952             if (retval > 0) {
953                 retval = rename(oldpath.data, path.data);
954                 if (retval < 0) {
955                     retval = -errno;
956                 }
957             }
958             v9fs_string_free(&oldpath);
959             v9fs_string_free(&path);
960             break;
961         case T_REMOVE:
962             v9fs_string_init(&path);
963             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "s", &path);
964             if (retval > 0) {
965                 retval = remove(path.data);
966                 if (retval < 0) {
967                     retval = -errno;
968                 }
969             }
970             v9fs_string_free(&path);
971             break;
972         case T_LGETXATTR:
973         case T_LLISTXATTR:
974             retval = do_getxattr(header.type, &in_iovec, &out_iovec);
975             break;
976         case T_LSETXATTR:
977             v9fs_string_init(&path);
978             v9fs_string_init(&name);
979             v9fs_string_init(&value);
980             retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "sssdd", &path,
981                                      &name, &value, &size, &flags);
982             if (retval > 0) {
983                 retval = lsetxattr(path.data,
984                                    name.data, value.data, size, flags);
985                 if (retval < 0) {
986                     retval = -errno;
987                 }
988             }
989             v9fs_string_free(&path);
990             v9fs_string_free(&name);
991             v9fs_string_free(&value);
992             break;
993         case T_LREMOVEXATTR:
994             v9fs_string_init(&path);
995             v9fs_string_init(&name);
996             retval = proxy_unmarshal(&in_iovec,
997                                      PROXY_HDR_SZ, "ss", &path, &name);
998             if (retval > 0) {
999                 retval = lremovexattr(path.data, name.data);
1000                 if (retval < 0) {
1001                     retval = -errno;
1002                 }
1003             }
1004             v9fs_string_free(&path);
1005             v9fs_string_free(&name);
1006             break;
1007         case T_GETVERSION:
1008             retval = do_getversion(&in_iovec, &out_iovec);
1009             break;
1010         default:
1011             goto err_out;
1012             break;
1013         }
1014 
1015         if (process_reply(sock, header.type, &out_iovec, retval) < 0) {
1016             goto err_out;
1017         }
1018     }
1019 err_out:
1020     g_free(in_iovec.iov_base);
1021     g_free(out_iovec.iov_base);
1022     return -1;
1023 }
1024 
1025 int main(int argc, char **argv)
1026 {
1027     int sock;
1028     uid_t own_u;
1029     gid_t own_g;
1030     char *rpath = NULL;
1031     char *sock_name = NULL;
1032     struct stat stbuf;
1033     int c, option_index;
1034 #ifdef FS_IOC_GETVERSION
1035     int retval;
1036     struct statfs st_fs;
1037 #endif
1038 
1039     prog_name = g_path_get_basename(argv[0]);
1040 
1041     is_daemon = true;
1042     sock = -1;
1043     own_u = own_g = -1;
1044     while (1) {
1045         option_index = 0;
1046         c = getopt_long(argc, argv, "p:nh?f:s:u:g:", helper_opts,
1047                         &option_index);
1048         if (c == -1) {
1049             break;
1050         }
1051         switch (c) {
1052         case 'p':
1053             rpath = g_strdup(optarg);
1054             break;
1055         case 'n':
1056             is_daemon = false;
1057             break;
1058         case 'f':
1059             sock = atoi(optarg);
1060             break;
1061         case 's':
1062             sock_name = g_strdup(optarg);
1063             break;
1064         case 'u':
1065             own_u = atoi(optarg);
1066             break;
1067         case 'g':
1068             own_g = atoi(optarg);
1069             break;
1070         case '?':
1071         case 'h':
1072         default:
1073             usage();
1074             exit(EXIT_FAILURE);
1075         }
1076     }
1077 
1078     /* Parameter validation */
1079     if ((sock_name == NULL && sock == -1) || rpath == NULL) {
1080         fprintf(stderr, "socket, socket descriptor or path not specified\n");
1081         usage();
1082         return -1;
1083     }
1084 
1085     if (sock_name && sock != -1) {
1086         fprintf(stderr, "both named socket and socket descriptor specified\n");
1087         usage();
1088         exit(EXIT_FAILURE);
1089     }
1090 
1091     if (sock_name && (own_u == -1 || own_g == -1)) {
1092         fprintf(stderr, "owner uid:gid not specified, ");
1093         fprintf(stderr,
1094                 "owner uid:gid specifies who can access the socket file\n");
1095         usage();
1096         exit(EXIT_FAILURE);
1097     }
1098 
1099     if (lstat(rpath, &stbuf) < 0) {
1100         fprintf(stderr, "invalid path \"%s\" specified, %s\n",
1101                 rpath, strerror(errno));
1102         exit(EXIT_FAILURE);
1103     }
1104 
1105     if (!S_ISDIR(stbuf.st_mode)) {
1106         fprintf(stderr, "specified path \"%s\" is not directory\n", rpath);
1107         exit(EXIT_FAILURE);
1108     }
1109 
1110     if (is_daemon) {
1111         if (daemon(0, 0) < 0) {
1112             fprintf(stderr, "daemon call failed\n");
1113             exit(EXIT_FAILURE);
1114         }
1115         openlog(PROGNAME, LOG_PID, LOG_DAEMON);
1116     }
1117 
1118     do_log(LOG_INFO, "Started\n");
1119     if (sock_name) {
1120         sock = proxy_socket(sock_name, own_u, own_g);
1121         if (sock < 0) {
1122             goto error;
1123         }
1124     }
1125 
1126     if (chroot(rpath) < 0) {
1127         do_perror("chroot");
1128         goto error;
1129     }
1130     if (chdir("/") < 0) {
1131         do_perror("chdir");
1132         goto error;
1133     }
1134 
1135     get_version = false;
1136 #ifdef FS_IOC_GETVERSION
1137     /* check whether underlying FS support IOC_GETVERSION */
1138     retval = statfs("/", &st_fs);
1139     if (!retval) {
1140         switch (st_fs.f_type) {
1141         case EXT2_SUPER_MAGIC:
1142         case BTRFS_SUPER_MAGIC:
1143         case REISERFS_SUPER_MAGIC:
1144         case XFS_SUPER_MAGIC:
1145             get_version = true;
1146             break;
1147         }
1148     }
1149 #endif
1150 
1151     umask(0);
1152     if (init_capabilities() < 0) {
1153         goto error;
1154     }
1155 
1156     process_requests(sock);
1157 error:
1158     g_free(rpath);
1159     g_free(sock_name);
1160     do_log(LOG_INFO, "Done\n");
1161     closelog();
1162     return 0;
1163 }
1164