2 * Helper for QEMU Proxy FS Driver
3 * Copyright IBM, Corp. 2011
8 * This work is licensed under the terms of the GNU GPL, version 2. See
9 * the COPYING file in the top-level directory.
12 #include <sys/resource.h>
15 #include <sys/capability.h>
16 #include <sys/fsuid.h>
18 #include <sys/ioctl.h>
20 #ifdef CONFIG_LINUX_MAGIC_H
21 #include <linux/magic.h>
23 #include "qemu-common.h"
24 #include "qemu/sockets.h"
25 #include "qemu/xattr.h"
26 #include "virtio-9p-marshal.h"
27 #include "hw/9pfs/virtio-9p-proxy.h"
28 #include "fsdev/virtio-9p-marshal.h"
30 #define PROGNAME "virtfs-proxy-helper"
32 #ifndef XFS_SUPER_MAGIC
33 #define XFS_SUPER_MAGIC 0x58465342
35 #ifndef EXT2_SUPER_MAGIC
36 #define EXT2_SUPER_MAGIC 0xEF53
38 #ifndef REISERFS_SUPER_MAGIC
39 #define REISERFS_SUPER_MAGIC 0x52654973
41 #ifndef BTRFS_SUPER_MAGIC
42 #define BTRFS_SUPER_MAGIC 0x9123683E
45 static struct option helper_opts[] = {
46 {"fd", required_argument, NULL, 'f'},
47 {"path", required_argument, NULL, 'p'},
48 {"nodaemon", no_argument, NULL, 'n'},
49 {"socket", required_argument, NULL, 's'},
50 {"uid", required_argument, NULL, 'u'},
51 {"gid", required_argument, NULL, 'g'},
54 static bool is_daemon;
55 static bool get_version; /* IOC getversion IOCTL supported */
57 static void GCC_FMT_ATTR(2, 3) do_log(int loglevel, const char *format, ...)
63 vsyslog(LOG_CRIT, format, ap);
65 vfprintf(stderr, format, ap);
70 static void do_perror(const char *string)
73 syslog(LOG_CRIT, "%s:%s", string, strerror(errno));
75 fprintf(stderr, "%s:%s\n", string, strerror(errno));
79 static int do_cap_set(cap_value_t *cap_value, int size, int reset)
84 * Start with an empty set and set permitted and effective
88 do_perror("cap_init");
91 if (cap_set_flag(caps, CAP_PERMITTED, size, cap_value, CAP_SET) < 0) {
92 do_perror("cap_set_flag");
96 caps = cap_get_proc();
98 do_perror("cap_get_proc");
102 if (cap_set_flag(caps, CAP_EFFECTIVE, size, cap_value, CAP_SET) < 0) {
103 do_perror("cap_set_flag");
106 if (cap_set_proc(caps) < 0) {
107 do_perror("cap_set_proc");
118 static int init_capabilities(void)
120 /* helper needs following capbabilities only */
121 cap_value_t cap_list[] = {
130 return do_cap_set(cap_list, ARRAY_SIZE(cap_list), 1);
133 static int socket_read(int sockfd, void *buff, ssize_t size)
135 ssize_t retval, total = 0;
138 retval = read(sockfd, buff, size);
143 if (errno == EINTR) {
155 static int socket_write(int sockfd, void *buff, ssize_t size)
157 ssize_t retval, total = 0;
160 retval = write(sockfd, buff, size);
162 if (errno == EINTR) {
174 static int read_request(int sockfd, struct iovec *iovec, ProxyHeader *header)
179 * read the request header.
182 retval = socket_read(sockfd, iovec->iov_base, PROXY_HDR_SZ);
186 iovec->iov_len = PROXY_HDR_SZ;
187 retval = proxy_unmarshal(iovec, 0, "dd", &header->type, &header->size);
192 * We can't process message.size > PROXY_MAX_IO_SZ.
193 * Treat it as fatal error
195 if (header->size > PROXY_MAX_IO_SZ) {
198 retval = socket_read(sockfd, iovec->iov_base + PROXY_HDR_SZ, header->size);
202 iovec->iov_len += header->size;
206 static int send_fd(int sockfd, int fd)
211 struct cmsghdr *cmsg;
212 union MsgControl msg_control;
214 iov.iov_base = &data;
215 iov.iov_len = sizeof(data);
217 memset(&msg, 0, sizeof(msg));
220 /* No ancillary data on error */
222 /* fd is really negative errno if the request failed */
225 data = V9FS_FD_VALID;
226 msg.msg_control = &msg_control;
227 msg.msg_controllen = sizeof(msg_control);
229 cmsg = &msg_control.cmsg;
230 cmsg->cmsg_len = CMSG_LEN(sizeof(fd));
231 cmsg->cmsg_level = SOL_SOCKET;
232 cmsg->cmsg_type = SCM_RIGHTS;
233 memcpy(CMSG_DATA(cmsg), &fd, sizeof(fd));
237 retval = sendmsg(sockfd, &msg, 0);
238 } while (retval < 0 && errno == EINTR);
248 static int send_status(int sockfd, struct iovec *iovec, int status)
251 int retval, msg_size;
254 header.type = T_ERROR;
256 header.type = T_SUCCESS;
258 header.size = sizeof(status);
260 * marshal the return status. We don't check error.
261 * because we are sure we have enough space for the status
263 msg_size = proxy_marshal(iovec, 0, "ddd", header.type,
264 header.size, status);
265 retval = socket_write(sockfd, iovec->iov_base, msg_size);
273 * from man 7 capabilities, section
274 * Effect of User ID Changes on Capabilities:
275 * If the effective user ID is changed from nonzero to 0, then the permitted
276 * set is copied to the effective set. If the effective user ID is changed
277 * from 0 to nonzero, then all capabilities are are cleared from the effective
280 * The setfsuid/setfsgid man pages warn that changing the effective user ID may
281 * expose the program to unwanted signals, but this is not true anymore: for an
282 * unprivileged (without CAP_KILL) program to send a signal, the real or
283 * effective user ID of the sending process must equal the real or saved user
284 * ID of the target process. Even when dropping privileges, it is enough to
285 * keep the saved UID to a "privileged" value and virtfs-proxy-helper won't
286 * be exposed to signals. So just use setresuid/setresgid.
288 static int setugid(int uid, int gid, int *suid, int *sgid)
293 * We still need DAC_OVERRIDE because we don't change
294 * supplementary group ids, and hence may be subjected DAC rules
296 cap_value_t cap_list[] = {
303 if (setresgid(-1, gid, *sgid) == -1) {
308 if (setresuid(-1, uid, *suid) == -1) {
313 if (uid != 0 || gid != 0) {
314 if (do_cap_set(cap_list, ARRAY_SIZE(cap_list), 0) < 0) {
322 if (setresuid(-1, *suid, *suid) == -1) {
326 if (setresgid(-1, *sgid, *sgid) == -1) {
334 * This is used to reset the ugid back with the saved values
335 * There is nothing much we can do checking error values here.
337 static void resetugid(int suid, int sgid)
339 if (setresgid(-1, sgid, sgid) == -1) {
342 if (setresuid(-1, suid, suid) == -1) {
348 * send response in two parts
350 * 2) Response or error status
351 * This function should be called with marshaled response
352 * send_response constructs header part and error part only.
353 * send response sends {ProxyHeader,Response} if the request was success
354 * otherwise sends {ProxyHeader,error status}
356 static int send_response(int sock, struct iovec *iovec, int size)
362 * If response size exceeds available iovec->iov_len,
365 if (size > PROXY_MAX_IO_SZ) {
371 * In case of error we would not have got the error encoded
372 * already so encode the error here.
374 header.type = T_ERROR;
375 header.size = sizeof(size);
376 proxy_marshal(iovec, PROXY_HDR_SZ, "d", size);
378 header.type = T_SUCCESS;
381 proxy_marshal(iovec, 0, "dd", header.type, header.size);
382 retval = socket_write(sock, iovec->iov_base, header.size + PROXY_HDR_SZ);
390 * gets generation number
391 * returns -errno on failure and sizeof(generation number) on success
393 static int do_getversion(struct iovec *iovec, struct iovec *out_iovec)
396 int retval = -ENOTTY;
397 #ifdef FS_IOC_GETVERSION
403 /* no need to issue ioctl */
406 retval = proxy_marshal(out_iovec, PROXY_HDR_SZ, "q", version);
409 #ifdef FS_IOC_GETVERSION
410 retval = proxy_unmarshal(iovec, PROXY_HDR_SZ, "s", &path);
415 fd = open(path.data, O_RDONLY);
420 if (ioctl(fd, FS_IOC_GETVERSION, &version) < 0) {
423 retval = proxy_marshal(out_iovec, PROXY_HDR_SZ, "q", version);
427 v9fs_string_free(&path);
432 static int do_getxattr(int type, struct iovec *iovec, struct iovec *out_iovec)
434 int size = 0, offset, retval;
435 V9fsString path, name, xattr;
437 v9fs_string_init(&xattr);
438 v9fs_string_init(&path);
439 retval = proxy_unmarshal(iovec, PROXY_HDR_SZ, "ds", &size, &path);
443 offset = PROXY_HDR_SZ + retval;
446 xattr.data = g_malloc(size);
451 v9fs_string_init(&name);
452 retval = proxy_unmarshal(iovec, offset, "s", &name);
454 retval = lgetxattr(path.data, name.data, xattr.data, size);
461 v9fs_string_free(&name);
464 retval = llistxattr(path.data, xattr.data, size);
477 proxy_marshal(out_iovec, PROXY_HDR_SZ, "d", retval);
478 retval = sizeof(retval);
480 retval = proxy_marshal(out_iovec, PROXY_HDR_SZ, "s", &xattr);
483 v9fs_string_free(&xattr);
484 v9fs_string_free(&path);
488 static void stat_to_prstat(ProxyStat *pr_stat, struct stat *stat)
490 memset(pr_stat, 0, sizeof(*pr_stat));
491 pr_stat->st_dev = stat->st_dev;
492 pr_stat->st_ino = stat->st_ino;
493 pr_stat->st_nlink = stat->st_nlink;
494 pr_stat->st_mode = stat->st_mode;
495 pr_stat->st_uid = stat->st_uid;
496 pr_stat->st_gid = stat->st_gid;
497 pr_stat->st_rdev = stat->st_rdev;
498 pr_stat->st_size = stat->st_size;
499 pr_stat->st_blksize = stat->st_blksize;
500 pr_stat->st_blocks = stat->st_blocks;
501 pr_stat->st_atim_sec = stat->st_atim.tv_sec;
502 pr_stat->st_atim_nsec = stat->st_atim.tv_nsec;
503 pr_stat->st_mtim_sec = stat->st_mtim.tv_sec;
504 pr_stat->st_mtim_nsec = stat->st_mtim.tv_nsec;
505 pr_stat->st_ctim_sec = stat->st_ctim.tv_sec;
506 pr_stat->st_ctim_nsec = stat->st_ctim.tv_nsec;
509 static void statfs_to_prstatfs(ProxyStatFS *pr_stfs, struct statfs *stfs)
511 memset(pr_stfs, 0, sizeof(*pr_stfs));
512 pr_stfs->f_type = stfs->f_type;
513 pr_stfs->f_bsize = stfs->f_bsize;
514 pr_stfs->f_blocks = stfs->f_blocks;
515 pr_stfs->f_bfree = stfs->f_bfree;
516 pr_stfs->f_bavail = stfs->f_bavail;
517 pr_stfs->f_files = stfs->f_files;
518 pr_stfs->f_ffree = stfs->f_ffree;
519 pr_stfs->f_fsid[0] = stfs->f_fsid.__val[0];
520 pr_stfs->f_fsid[1] = stfs->f_fsid.__val[1];
521 pr_stfs->f_namelen = stfs->f_namelen;
522 pr_stfs->f_frsize = stfs->f_frsize;
526 * Gets stat/statfs information and packs in out_iovec structure
527 * on success returns number of bytes packed in out_iovec struture
528 * otherwise returns -errno
530 static int do_stat(int type, struct iovec *iovec, struct iovec *out_iovec)
537 struct statfs stfs_buf;
539 v9fs_string_init(&path);
540 retval = proxy_unmarshal(iovec, PROXY_HDR_SZ, "s", &path);
547 retval = lstat(path.data, &st_buf);
551 stat_to_prstat(&pr_stat, &st_buf);
552 retval = proxy_marshal(out_iovec, PROXY_HDR_SZ,
553 "qqqdddqqqqqqqqqq", pr_stat.st_dev,
554 pr_stat.st_ino, pr_stat.st_nlink,
555 pr_stat.st_mode, pr_stat.st_uid,
556 pr_stat.st_gid, pr_stat.st_rdev,
557 pr_stat.st_size, pr_stat.st_blksize,
559 pr_stat.st_atim_sec, pr_stat.st_atim_nsec,
560 pr_stat.st_mtim_sec, pr_stat.st_mtim_nsec,
561 pr_stat.st_ctim_sec, pr_stat.st_ctim_nsec);
565 retval = statfs(path.data, &stfs_buf);
569 statfs_to_prstatfs(&pr_stfs, &stfs_buf);
570 retval = proxy_marshal(out_iovec, PROXY_HDR_SZ,
571 "qqqqqqqqqqq", pr_stfs.f_type,
572 pr_stfs.f_bsize, pr_stfs.f_blocks,
573 pr_stfs.f_bfree, pr_stfs.f_bavail,
574 pr_stfs.f_files, pr_stfs.f_ffree,
575 pr_stfs.f_fsid[0], pr_stfs.f_fsid[1],
576 pr_stfs.f_namelen, pr_stfs.f_frsize);
580 v9fs_string_free(&path);
584 static int do_readlink(struct iovec *iovec, struct iovec *out_iovec)
588 V9fsString target, path;
590 v9fs_string_init(&path);
591 retval = proxy_unmarshal(iovec, PROXY_HDR_SZ, "sd", &path, &size);
593 v9fs_string_free(&path);
596 buffer = g_malloc(size);
597 v9fs_string_init(&target);
598 retval = readlink(path.data, buffer, size);
600 buffer[retval] = '\0';
601 v9fs_string_sprintf(&target, "%s", buffer);
602 retval = proxy_marshal(out_iovec, PROXY_HDR_SZ, "s", &target);
607 v9fs_string_free(&target);
608 v9fs_string_free(&path);
613 * create other filesystem objects and send 0 on success
614 * return -errno on error
616 static int do_create_others(int type, struct iovec *iovec)
620 int offset = PROXY_HDR_SZ;
621 V9fsString oldpath, path;
622 int mode, uid, gid, cur_uid, cur_gid;
624 v9fs_string_init(&path);
625 v9fs_string_init(&oldpath);
627 retval = proxy_unmarshal(iovec, offset, "dd", &uid, &gid);
632 retval = setugid(uid, gid, &cur_uid, &cur_gid);
634 goto unmarshal_err_out;
638 retval = proxy_unmarshal(iovec, offset, "sdq", &path, &mode, &rdev);
642 retval = mknod(path.data, mode, rdev);
645 retval = proxy_unmarshal(iovec, offset, "sd", &path, &mode);
649 retval = mkdir(path.data, mode);
652 retval = proxy_unmarshal(iovec, offset, "ss", &oldpath, &path);
656 retval = symlink(oldpath.data, path.data);
664 resetugid(cur_uid, cur_gid);
666 v9fs_string_free(&path);
667 v9fs_string_free(&oldpath);
672 * create a file and send fd on success
673 * return -errno on error
675 static int do_create(struct iovec *iovec)
679 int flags, mode, uid, gid, cur_uid, cur_gid;
681 v9fs_string_init(&path);
682 ret = proxy_unmarshal(iovec, PROXY_HDR_SZ, "sdddd",
683 &path, &flags, &mode, &uid, &gid);
685 goto unmarshal_err_out;
687 ret = setugid(uid, gid, &cur_uid, &cur_gid);
689 goto unmarshal_err_out;
691 ret = open(path.data, flags, mode);
696 resetugid(cur_uid, cur_gid);
698 v9fs_string_free(&path);
703 * open a file and send fd on success
704 * return -errno on error
706 static int do_open(struct iovec *iovec)
711 v9fs_string_init(&path);
712 ret = proxy_unmarshal(iovec, PROXY_HDR_SZ, "sd", &path, &flags);
716 ret = open(path.data, flags);
721 v9fs_string_free(&path);
725 /* create unix domain socket and return the descriptor */
726 static int proxy_socket(const char *path, uid_t uid, gid_t gid)
729 struct sockaddr_un proxy, qemu;
732 /* requested socket already exists, refuse to start */
733 if (!access(path, F_OK)) {
734 do_log(LOG_CRIT, "socket already exists\n");
738 sock = socket(AF_UNIX, SOCK_STREAM, 0);
744 /* mask other part of mode bits */
747 proxy.sun_family = AF_UNIX;
748 strcpy(proxy.sun_path, path);
749 if (bind(sock, (struct sockaddr *)&proxy,
750 sizeof(struct sockaddr_un)) < 0) {
754 if (chown(proxy.sun_path, uid, gid) < 0) {
758 if (listen(sock, 1) < 0) {
763 client = accept(sock, (struct sockaddr *)&qemu, &size);
771 static void usage(char *prog)
773 fprintf(stderr, "usage: %s\n"
774 " -p|--path <path> 9p path to export\n"
775 " {-f|--fd <socket-descriptor>} socket file descriptor to be used\n"
776 " {-s|--socket <socketname> socket file used for communication\n"
777 " \t-u|--uid <uid> -g|--gid <gid>} - uid:gid combination to give "
778 " access to this socket\n"
779 " \tNote: -s & -f can not be used together\n"
780 " [-n|--nodaemon] Run as a normal program\n",
784 static int process_reply(int sock, int type,
785 struct iovec *out_iovec, int retval)
790 if (send_fd(sock, retval) < 0) {
806 if (send_status(sock, out_iovec, retval) < 0) {
816 if (send_response(sock, out_iovec, retval) < 0) {
827 static int process_requests(int sock)
835 V9fsString name, value;
836 struct timespec spec[2];
837 V9fsString oldpath, path;
838 struct iovec in_iovec, out_iovec;
840 in_iovec.iov_base = g_malloc(PROXY_MAX_IO_SZ + PROXY_HDR_SZ);
841 in_iovec.iov_len = PROXY_MAX_IO_SZ + PROXY_HDR_SZ;
842 out_iovec.iov_base = g_malloc(PROXY_MAX_IO_SZ + PROXY_HDR_SZ);
843 out_iovec.iov_len = PROXY_MAX_IO_SZ + PROXY_HDR_SZ;
847 * initialize the header type, so that we send
848 * response to proper request type.
851 retval = read_request(sock, &in_iovec, &header);
856 switch (header.type) {
858 retval = do_open(&in_iovec);
861 retval = do_create(&in_iovec);
866 retval = do_create_others(header.type, &in_iovec);
869 v9fs_string_init(&path);
870 v9fs_string_init(&oldpath);
871 retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ,
872 "ss", &oldpath, &path);
874 retval = link(oldpath.data, path.data);
879 v9fs_string_free(&oldpath);
880 v9fs_string_free(&path);
884 retval = do_stat(header.type, &in_iovec, &out_iovec);
887 retval = do_readlink(&in_iovec, &out_iovec);
890 v9fs_string_init(&path);
891 retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ,
894 retval = chmod(path.data, mode);
899 v9fs_string_free(&path);
902 v9fs_string_init(&path);
903 retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "sdd", &path,
906 retval = lchown(path.data, uid, gid);
911 v9fs_string_free(&path);
914 v9fs_string_init(&path);
915 retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "sq",
918 retval = truncate(path.data, offset);
923 v9fs_string_free(&path);
926 v9fs_string_init(&path);
927 retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "sqqqq", &path,
928 &spec[0].tv_sec, &spec[0].tv_nsec,
929 &spec[1].tv_sec, &spec[1].tv_nsec);
931 retval = qemu_utimens(path.data, spec);
936 v9fs_string_free(&path);
939 v9fs_string_init(&path);
940 v9fs_string_init(&oldpath);
941 retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ,
942 "ss", &oldpath, &path);
944 retval = rename(oldpath.data, path.data);
949 v9fs_string_free(&oldpath);
950 v9fs_string_free(&path);
953 v9fs_string_init(&path);
954 retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "s", &path);
956 retval = remove(path.data);
961 v9fs_string_free(&path);
965 retval = do_getxattr(header.type, &in_iovec, &out_iovec);
968 v9fs_string_init(&path);
969 v9fs_string_init(&name);
970 v9fs_string_init(&value);
971 retval = proxy_unmarshal(&in_iovec, PROXY_HDR_SZ, "sssdd", &path,
972 &name, &value, &size, &flags);
974 retval = lsetxattr(path.data,
975 name.data, value.data, size, flags);
980 v9fs_string_free(&path);
981 v9fs_string_free(&name);
982 v9fs_string_free(&value);
985 v9fs_string_init(&path);
986 v9fs_string_init(&name);
987 retval = proxy_unmarshal(&in_iovec,
988 PROXY_HDR_SZ, "ss", &path, &name);
990 retval = lremovexattr(path.data, name.data);
995 v9fs_string_free(&path);
996 v9fs_string_free(&name);
999 retval = do_getversion(&in_iovec, &out_iovec);
1006 if (process_reply(sock, header.type, &out_iovec, retval) < 0) {
1011 g_free(in_iovec.iov_base);
1012 g_free(out_iovec.iov_base);
1016 int main(int argc, char **argv)
1022 char *sock_name = NULL;
1024 int c, option_index;
1025 #ifdef FS_IOC_GETVERSION
1027 struct statfs st_fs;
1035 c = getopt_long(argc, argv, "p:nh?f:s:u:g:", helper_opts,
1042 rpath = g_strdup(optarg);
1048 sock = atoi(optarg);
1051 sock_name = g_strdup(optarg);
1054 own_u = atoi(optarg);
1057 own_g = atoi(optarg);
1067 /* Parameter validation */
1068 if ((sock_name == NULL && sock == -1) || rpath == NULL) {
1069 fprintf(stderr, "socket, socket descriptor or path not specified\n");
1074 if (sock_name && sock != -1) {
1075 fprintf(stderr, "both named socket and socket descriptor specified\n");
1080 if (sock_name && (own_u == -1 || own_g == -1)) {
1081 fprintf(stderr, "owner uid:gid not specified, ");
1083 "owner uid:gid specifies who can access the socket file\n");
1088 if (lstat(rpath, &stbuf) < 0) {
1089 fprintf(stderr, "invalid path \"%s\" specified, %s\n",
1090 rpath, strerror(errno));
1094 if (!S_ISDIR(stbuf.st_mode)) {
1095 fprintf(stderr, "specified path \"%s\" is not directory\n", rpath);
1100 if (daemon(0, 0) < 0) {
1101 fprintf(stderr, "daemon call failed\n");
1104 openlog(PROGNAME, LOG_PID, LOG_DAEMON);
1107 do_log(LOG_INFO, "Started\n");
1109 sock = proxy_socket(sock_name, own_u, own_g);
1115 get_version = false;
1116 #ifdef FS_IOC_GETVERSION
1117 /* check whether underlying FS support IOC_GETVERSION */
1118 retval = statfs(rpath, &st_fs);
1120 switch (st_fs.f_type) {
1121 case EXT2_SUPER_MAGIC:
1122 case BTRFS_SUPER_MAGIC:
1123 case REISERFS_SUPER_MAGIC:
1124 case XFS_SUPER_MAGIC:
1131 if (chdir("/") < 0) {
1135 if (chroot(rpath) < 0) {
1136 do_perror("chroot");
1141 if (init_capabilities() < 0) {
1145 process_requests(sock);
1147 do_log(LOG_INFO, "Done\n");