3 * Handle communication with knfsd internal cache
5 * We open /proc/net/rpc/{auth.unix.ip,nfsd.export,nfsd.fh}/channel
6 * and listen for requests (using my_svc_run)
14 #include <sys/types.h>
15 #include <sys/select.h>
18 #include <netinet/in.h>
19 #include <arpa/inet.h>
33 #include "pseudoflavors.h"
36 #include "blkid/blkid.h"
52 * Support routines for text-based upcalls.
53 * Fields are separated by spaces.
54 * Fields are either mangled to quote space tab newline slosh with slosh
55 * or a hexified with a leading \x
56 * Record is terminated with newline.
59 int cache_export_ent(char *domain, struct exportent *exp, char *p);
65 void auth_unix_ip(FILE *f)
69 * Ignore if class != "nfsd"
70 * Otherwise find domainname and write back:
72 * "nfsd" IP-ADDR expiry domainname
80 if (readline(fileno(f), &lbuf, &lbuflen) != 1)
85 if (qword_get(&cp, class, 20) <= 0 ||
86 strcmp(class, "nfsd") != 0)
89 if (qword_get(&cp, ipaddr, 20) <= 0)
92 if (inet_aton(ipaddr, &addr)==0)
97 /* addr is a valid, interesting address, find the domain name... */
98 he = client_resolve(addr);
99 client = client_compose(he);
102 qword_print(f, "nfsd");
103 qword_print(f, ipaddr);
104 qword_printint(f, time(0)+30*60);
106 qword_print(f, *client?client:"DEFAULT");
109 if (client) free(client);
113 void auth_unix_gid(FILE *f)
118 * uid expiry count list of group ids
122 gid_t glist[100], *groups = glist;
127 if (readline(fileno(f), &lbuf, &lbuflen) != 1)
131 if (qword_get_int(&cp, &uid) != 0)
138 rv = getgrouplist(pw->pw_name, pw->pw_gid, groups, &ngroups);
139 if (rv == -1 && ngroups >= 100) {
140 groups = malloc(sizeof(gid_t)*ngroups);
144 rv = getgrouplist(pw->pw_name, pw->pw_gid,
148 qword_printint(f, uid);
149 qword_printint(f, time(0)+30*60);
151 qword_printint(f, ngroups);
152 for (i=0; i<ngroups; i++)
153 qword_printint(f, groups[i]);
161 int get_uuid(char *path, char *uuid, int uuidlen, char *u)
163 /* extract hex digits from uuidstr and compose a uuid
164 * of the given length (max 16), xoring bytes to make
165 * a smaller uuid. Then compare with uuid
171 static blkid_cache cache = NULL;
174 blkid_tag_iterate iter;
178 blkid_get_cache(&cache, NULL);
180 blkid_probe_all_new(cache);
182 if (stat(path, &stb) != 0)
184 devname = blkid_devno_to_devname(stb.st_dev);
187 dev = blkid_get_dev(cache, devname, BLKID_DEV_NORMAL);
191 iter = blkid_tag_iterate_begin(dev);
194 while (blkid_tag_next(iter, &type, &val) == 0)
195 if (strcmp(type, "UUID") == 0)
197 blkid_tag_iterate_end(iter);
204 memset(u, 0, uuidlen);
205 for ( ; *val ; val++) {
227 /* Iterate through /etc/mtab, finding mountpoints
228 * at or below a given path
230 static char *next_mnt(void **v, char *p)
236 f = setmntent("/etc/mtab", "r");
240 while ((me = getmntent(f)) != NULL &&
241 (strncmp(me->mnt_dir, p, l) != 0 ||
242 me->mnt_dir[l] != '/'))
252 void nfsd_fh(FILE *f)
255 * domain fsidtype fsid
256 * interpret fsid, find export point and options, and write:
257 * domain fsidtype fsid expiry path
263 unsigned int dev, major=0, minor=0;
264 unsigned int inode=0;
265 unsigned long long inode64;
266 unsigned int fsidnum=0;
268 struct exportent *found = NULL;
269 char *found_path = NULL;
276 if (readline(fileno(f), &lbuf, &lbuflen) != 1)
281 dom = malloc(strlen(cp));
284 if (qword_get(&cp, dom, strlen(cp)) <= 0)
286 if (qword_get_int(&cp, &fsidtype) != 0)
288 if (fsidtype < 0 || fsidtype > 7)
289 goto out; /* unknown type */
290 if ((fsidlen = qword_get(&cp, fsid, 32)) <= 0)
293 case FSID_DEV: /* 4 bytes: 2 major, 2 minor, 4 inode */
296 memcpy(&dev, fsid, 4);
297 memcpy(&inode, fsid+4, 4);
298 major = ntohl(dev)>>16;
299 minor = ntohl(dev) & 0xFFFF;
302 case FSID_NUM: /* 4 bytes - fsid */
305 memcpy(&fsidnum, fsid, 4);
308 case FSID_MAJOR_MINOR: /* 12 bytes: 4 major, 4 minor, 4 inode
309 * This format is never actually used but was
310 * an historical accident
314 memcpy(&dev, fsid, 4); major = ntohl(dev);
315 memcpy(&dev, fsid+4, 4); minor = ntohl(dev);
316 memcpy(&inode, fsid+8, 4);
319 case FSID_ENCODE_DEV: /* 8 bytes: 4 byte packed device number, 4 inode */
320 /* This is *host* endian, not net-byte-order, because
321 * no-one outside this host has any business interpreting it
325 memcpy(&dev, fsid, 4);
326 memcpy(&inode, fsid+4, 4);
327 major = (dev & 0xfff00) >> 8;
328 minor = (dev & 0xff) | ((dev >> 12) & 0xfff00);
331 case FSID_UUID4_INUM: /* 4 byte inode number and 4 byte uuid */
334 memcpy(&inode, fsid, 4);
338 case FSID_UUID8: /* 8 byte uuid */
344 case FSID_UUID16: /* 16 byte uuid */
350 case FSID_UUID16_INUM: /* 8 byte inode number and 16 byte uuid */
353 memcpy(&inode64, fsid, 8);
362 /* Now determine export point for this fsid/domain */
363 for (i=0 ; i < MCL_MAXTYPES; i++) {
364 nfs_export *next_exp;
365 for (exp = exportlist[i]; exp; exp = next_exp) {
370 if (exp->m_export.e_flags & NFSEXP_CROSSMOUNT) {
371 static nfs_export *prev = NULL;
372 static void *mnt = NULL;
376 path = next_mnt(&mnt, exp->m_export.e_path);
378 next_exp = exp->m_next;
386 path = exp->m_export.e_path;
390 path = exp->m_export.e_path;
391 next_exp = exp->m_next;
394 if (!client_member(dom, exp->m_client->m_hostname))
396 if (exp->m_export.e_mountpoint &&
397 !is_mountpoint(exp->m_export.e_mountpoint[0]?
398 exp->m_export.e_mountpoint:
399 exp->m_export.e_path))
401 if (stat(path, &stb) != 0)
403 if (!S_ISDIR(stb.st_mode) && !S_ISREG(stb.st_mode)) {
408 case FSID_MAJOR_MINOR:
409 case FSID_ENCODE_DEV:
410 if (stb.st_ino != inode)
412 if (major != major(stb.st_dev) ||
413 minor != minor(stb.st_dev))
417 if (((exp->m_export.e_flags & NFSEXP_FSID) == 0 ||
418 exp->m_export.e_fsid != fsidnum))
421 case FSID_UUID4_INUM:
422 case FSID_UUID16_INUM:
423 if (stb.st_ino != inode)
428 if (!is_mountpoint(path))
432 if (exp->m_export.e_uuid)
433 get_uuid(NULL, exp->m_export.e_uuid,
435 else if (get_uuid(path, NULL,
439 if (memcmp(u, fhuuid, uuidlen) != 0)
446 /* It's a match !! */
448 found = &exp->m_export;
449 found_path = strdup(path);
450 if (found_path == NULL)
452 } else if (strcmp(found->e_path, exp->m_export.e_path)!= 0)
454 xlog(L_WARNING, "%s and %s have same filehandle for %s, using first",
455 found_path, path, dom);
460 found->e_mountpoint &&
461 !is_mountpoint(found->e_mountpoint[0]?
464 /* Cannot export this yet
465 * should log a warning, but need to rate limit
466 xlog(L_WARNING, "%s not exported as %d not a mountpoint",
467 found->e_path, found->e_mountpoint);
469 /* FIXME we need to make sure we re-visit this later */
472 if (!found && dev_missing) {
473 /* The missing dev could be what we want, so just be
474 * quite rather than returning stale yet
480 if (cache_export_ent(dom, found, found_path) < 0)
484 qword_printint(f, fsidtype);
485 qword_printhex(f, fsid, fsidlen);
486 /* The fsid -> path lookup can be quite expensive as it
487 * potentially stats and reads lots of devices, and some of those
488 * might have spun-down. The Answer is not likely to
489 * change underneath us, and an 'exportfs -f' can always
490 * remove this from the kernel, so use a really log
491 * timeout. Maybe this should be configurable on the command
494 qword_printint(f, 0x7fffffff);
496 qword_print(f, found->e_path);
504 static void write_fsloc(FILE *f, struct exportent *ep, char *path)
506 struct servers *servers;
508 if (ep->e_fslocmethod == FSLOC_NONE)
511 servers = replicas_lookup(ep->e_fslocmethod, ep->e_fslocdata, path);
514 qword_print(f, "fsloc");
515 qword_printint(f, servers->h_num);
516 if (servers->h_num >= 0) {
518 for (i=0; i<servers->h_num; i++) {
519 qword_print(f, servers->h_mp[i]->h_host);
520 qword_print(f, servers->h_mp[i]->h_path);
523 qword_printint(f, servers->h_referral);
524 release_replicas(servers);
527 static void write_secinfo(FILE *f, struct exportent *ep)
531 for (p = ep->e_secinfo; p->flav; p++)
533 if (p == ep->e_secinfo) {
534 /* There was no sec= option */
537 qword_print(f, "secinfo");
538 qword_printint(f, p - ep->e_secinfo);
539 for (p = ep->e_secinfo; p->flav; p++) {
540 qword_printint(f, p->flav->fnum);
541 qword_printint(f, p->flags);
546 static int dump_to_cache(FILE *f, char *domain, char *path, struct exportent *exp)
548 qword_print(f, domain);
549 qword_print(f, path);
550 qword_printint(f, time(0)+30*60);
552 qword_printint(f, exp->e_flags);
553 qword_printint(f, exp->e_anonuid);
554 qword_printint(f, exp->e_anongid);
555 qword_printint(f, exp->e_fsid);
556 write_fsloc(f, exp, path);
557 write_secinfo(f, exp);
559 if (exp->e_uuid == NULL) {
561 if (get_uuid(path, NULL, 16, u)) {
562 qword_print(f, "uuid");
563 qword_printhex(f, u, 16);
565 } else if (exp->e_uuid) {
566 qword_print(f, "uuid");
567 qword_printhex(f, exp->e_uuid, 16);
574 void nfsd_export(FILE *f)
578 * determine export options and return:
579 * domain path expiry flags anonuid anongid fsid
585 nfs_export *exp, *found = NULL;
589 if (readline(fileno(f), &lbuf, &lbuflen) != 1)
593 dom = malloc(strlen(cp));
594 path = malloc(strlen(cp));
599 if (qword_get(&cp, dom, strlen(lbuf)) <= 0)
601 if (qword_get(&cp, path, strlen(lbuf)) <= 0)
606 /* now find flags for this export point in this domain */
607 for (i=0 ; i < MCL_MAXTYPES; i++) {
608 for (exp = exportlist[i]; exp; exp = exp->m_next) {
609 if (!client_member(dom, exp->m_client->m_hostname))
611 if (exp->m_export.e_flags & NFSEXP_CROSSMOUNT) {
612 /* if path is a mountpoint below e_path, then OK */
613 int l = strlen(exp->m_export.e_path);
614 if (strcmp(path, exp->m_export.e_path) == 0 ||
615 (strncmp(path, exp->m_export.e_path, l) == 0 &&
617 is_mountpoint(path)))
621 } else if (strcmp(path, exp->m_export.e_path) != 0)
628 /* If one is a CROSSMOUNT, then prefer the longest path */
629 if (((found->m_export.e_flags & NFSEXP_CROSSMOUNT) ||
630 (found->m_export.e_flags & NFSEXP_CROSSMOUNT)) &&
631 strlen(found->m_export.e_path) !=
632 strlen(found->m_export.e_path)) {
634 if (strlen(exp->m_export.e_path) >
635 strlen(found->m_export.e_path)) {
641 } else if (found_type == i && found->m_warned == 0) {
642 xlog(L_WARNING, "%s exported to both %s and %s, "
643 "arbitrarily choosing options from first",
644 path, found->m_client->m_hostname, exp->m_client->m_hostname,
652 if (dump_to_cache(f, dom, path, &found->m_export) < 0) {
654 "Cannot export %s, possibly unsupported filesystem"
655 " or fsid= required", path);
656 dump_to_cache(f, dom, path, NULL);
659 dump_to_cache(f, dom, path, NULL);
663 if (path) free(path);
669 void (*cache_handle)(FILE *f);
672 { "auth.unix.ip", auth_unix_ip},
673 { "auth.unix.gid", auth_unix_gid},
674 { "nfsd.export", nfsd_export},
675 { "nfsd.fh", nfsd_fh},
679 extern int manage_gids;
680 void cache_open(void)
683 for (i=0; cachelist[i].cache_name; i++ ) {
685 if (!manage_gids && cachelist[i].cache_handle == auth_unix_gid)
687 sprintf(path, "/proc/net/rpc/%s/channel", cachelist[i].cache_name);
688 cachelist[i].f = fopen(path, "r+");
692 void cache_set_fds(fd_set *fdset)
695 for (i=0; cachelist[i].cache_name; i++) {
697 FD_SET(fileno(cachelist[i].f), fdset);
701 int cache_process_req(fd_set *readfds)
705 for (i=0; cachelist[i].cache_name; i++) {
706 if (cachelist[i].f != NULL &&
707 FD_ISSET(fileno(cachelist[i].f), readfds)) {
709 cachelist[i].cache_handle(cachelist[i].f);
710 FD_CLR(fileno(cachelist[i].f), readfds);
718 * Give IP->domain and domain+path->options to kernel
719 * % echo nfsd $IP $[now+30*60] $domain > /proc/net/rpc/auth.unix.ip/channel
720 * % echo $domain $path $[now+30*60] $options $anonuid $anongid $fsid > /proc/net/rpc/nfsd.export/channel
723 int cache_export_ent(char *domain, struct exportent *exp, char *path)
726 FILE *f = fopen("/proc/net/rpc/nfsd.export/channel", "w");
730 err = dump_to_cache(f, domain, exp->e_path, exp);
733 "Cannot export %s, possibly unsupported filesystem or"
734 " fsid= required", exp->e_path);
737 while (err == 0 && (exp->e_flags & NFSEXP_CROSSMOUNT) && path) {
738 /* really an 'if', but we can break out of
739 * a 'while' more easily */
740 /* Look along 'path' for other filesystems
741 * and export them with the same options
744 int l = strlen(exp->e_path);
747 if (strlen(path) <= l || path[l] != '/' ||
748 strncmp(exp->e_path, path, l) != 0)
750 if (stat(exp->e_path, &stb) != 0)
753 while(path[l] == '/') {
755 /* errors for submount should fail whole filesystem */
759 while (path[l] != '/' && path[l])
763 err2 = lstat(path, &stb);
767 if (stb.st_dev == dev)
771 dump_to_cache(f, domain, path, exp);
781 int cache_export(nfs_export *exp, char *path)
786 f = fopen("/proc/net/rpc/auth.unix.ip/channel", "w");
790 qword_print(f, "nfsd");
791 qword_print(f, inet_ntoa(exp->m_client->m_addrlist[0]));
792 qword_printint(f, time(0)+30*60);
793 qword_print(f, exp->m_client->m_hostname);
798 err = cache_export_ent(exp->m_client->m_hostname, &exp->m_export, path)
805 * echo $domain $path $length
806 * read filehandle <&0
807 * } <> /proc/fs/nfsd/filehandle
810 cache_get_filehandle(nfs_export *exp, int len, char *p)
812 FILE *f = fopen("/proc/fs/nfsd/filehandle", "r+");
816 static struct nfs_fh_len fh;
819 f = fopen("/proc/fs/nfs/filehandle", "r+");
823 qword_print(f, exp->m_client->m_hostname);
825 qword_printint(f, len);
826 failed = qword_eol(f);
829 failed = (fgets(buf, sizeof(buf), f) == NULL);
833 memset(fh.fh_handle, 0, sizeof(fh.fh_handle));
834 fh.fh_size = qword_get(&bp, (char *)fh.fh_handle, NFS3_FHSIZE);