Merge 1.8.0~pre4 packaging into master
[pkg-k5-afs_openafs.git] / src / vol / partition.c
blob272ada35d8d5b1ed76d0d3148d73753c623926d8
1 /*
2 * Copyright 2000, International Business Machines Corporation and others.
3 * All Rights Reserved.
5 * This software has been released under the terms of the IBM Public
6 * License. For details, see the LICENSE file in the top-level source
7 * directory or online at http://www.openafs.org/dl/license10.html
9 * Portions Copyright (c) 2003 Apple Computer, Inc.
10 * Portions Copyright (c) 2006 Sine Nomine Associates
15 System: VICE-TWO
16 Module: partition.c
17 Institution: The Information Technology Center, Carnegie-Mellon University
21 #include <afsconfig.h>
22 #include <afs/param.h>
24 #include <roken.h>
26 #include <ctype.h>
28 #ifdef AFS_NT40_ENV
29 #include <windows.h>
30 #include <winbase.h>
31 #include <winioctl.h>
32 #else
34 #if AFS_HAVE_STATVFS || AFS_HAVE_STATVFS64
35 #include <sys/statvfs.h>
36 #endif /* AFS_HAVE_STATVFS */
37 #if defined(AFS_DARWIN_ENV) || defined(AFS_XBSD_ENV)
38 #include <sys/mount.h>
39 #endif
41 #if !defined(AFS_SGI_ENV)
42 #ifdef AFS_OSF_ENV
43 #include <sys/mount.h>
44 #include <ufs/fs.h>
45 #else /* AFS_OSF_ENV */
46 #ifdef AFS_VFSINCL_ENV
47 #define VFS
48 #ifdef AFS_SUN5_ENV
49 #include <sys/fs/ufs_fs.h>
50 #else
51 #if defined(AFS_DARWIN_ENV) || defined(AFS_XBSD_ENV)
52 #include <ufs/ufs/dinode.h>
53 #include <ufs/ffs/fs.h>
54 #else
55 #include <ufs/fs.h>
56 #endif
57 #endif
58 #else /* AFS_VFSINCL_ENV */
59 #if !defined(AFS_AIX_ENV) && !defined(AFS_LINUX22_ENV) && !defined(AFS_DARWIN_ENV) && !defined(AFS_XBSD_ENV)
60 #include <sys/fs.h>
61 #endif
62 #endif /* AFS_VFSINCL_ENV */
63 #endif /* AFS_OSF_ENV */
64 #include <sys/file.h>
65 #ifdef AFS_AIX_ENV
66 #include <sys/vfs.h>
67 #include <sys/lockf.h>
68 #else
69 #ifdef AFS_HPUX_ENV
70 #include <sys/vfs.h>
71 #include <checklist.h>
72 #else
73 #if defined(AFS_SUN_ENV)
74 #include <sys/vfs.h>
75 #ifndef AFS_SUN5_ENV
76 #include <mntent.h>
77 #endif
78 #endif
79 #ifdef AFS_SUN5_ENV
80 #include <sys/mnttab.h>
81 #include <sys/mntent.h>
82 #else
83 #ifdef AFS_LINUX22_ENV
84 #include <mntent.h>
85 #include <sys/statfs.h>
86 #else
87 #include <fstab.h>
88 #endif
89 #endif
90 #endif
91 #endif
92 #endif /* AFS_SGI_ENV */
93 #endif /* AFS_NT40_ENV */
94 #if defined(AFS_SGI_ENV)
95 #include <sys/file.h>
96 #include <mntent.h>
97 #endif
99 #include <afs/opr.h>
100 #ifdef AFS_PTHREAD_ENV
101 # include <opr/lock.h>
102 #endif
103 #include <afs/afsint.h>
104 #include <rx/rx_queue.h>
105 #include "nfs.h"
106 #include <afs/errors.h>
107 #include "lock.h"
108 #include "lwp.h"
109 #include <afs/afssyscalls.h>
110 #include "ihandle.h"
111 #include "common.h"
112 #ifdef AFS_NAMEI_ENV
113 #ifdef AFS_NT40_ENV
114 #include "ntops.h"
115 #else
116 #include "namei_ops.h"
117 #endif /* AFS_NT40_ENV */
118 #endif /* AFS_NAMEI_ENV */
119 #include "vnode.h"
120 #include "volume.h"
121 #include "partition.h"
123 #if defined(AFS_HPUX_ENV)
124 #include <sys/privgrp.h>
125 #endif /* defined(AFS_HPUX_ENV) */
127 #ifdef AFS_AIX42_ENV
128 #include <jfs/filsys.h>
129 #endif
131 #ifdef AFS_NT40_ENV
132 extern int VValidVPTEntry(struct vptab *vptp);
133 #endif
135 int aixlow_water = 8; /* default 8% */
136 struct DiskPartition64 *DiskPartitionList;
138 #ifdef AFS_DEMAND_ATTACH_FS
139 /* file to lock to conceptually "lock" the vol headers on a partition */
140 #define AFS_PARTLOCK_FILE ".volheaders.lock"
141 #define AFS_VOLUMELOCK_FILE ".volume.lock"
143 static struct DiskPartition64 *DiskPartitionTable[VOLMAXPARTS+1];
145 static struct DiskPartition64 * VLookupPartition_r(char * path);
146 static void AddPartitionToTable_r(struct DiskPartition64 *);
147 #endif /* AFS_DEMAND_ATTACH_FS */
149 #ifdef AFS_SGI_XFS_IOPS_ENV
150 /* Verify that the on disk XFS inodes on the partition are large enough to
151 * hold the AFS attribute. Returns -1 if the attribute can't be set or is
152 * too small to fit in the inode. Returns 0 if the attribute does fit in
153 * the XFS inode.
155 #include <afs/xfsattrs.h>
156 static int
157 VerifyXFSInodeSize(char *part, char *fstype)
159 afs_xfs_attr_t junk;
160 int length = SIZEOF_XFS_ATTR_T;
161 int fd = 0;
162 int code = -1;
163 struct fsxattr fsx;
165 if (strcmp("xfs", fstype))
166 return 0;
168 if (attr_set(part, AFS_XFS_ATTR, &junk, length, ATTR_ROOT) == 0) {
169 if (((fd = open(part, O_RDONLY, 0)) != -1)
170 && (fcntl(fd, F_FSGETXATTRA, &fsx) == 0)) {
172 if (fsx.fsx_nextents) {
173 Log("Partition %s: XFS inodes too small, exiting.\n", part);
174 Log("Run xfs_size_check utility and remake partitions.\n");
175 } else
176 code = 0;
179 if (fd > 0)
180 close(fd);
181 (void)attr_remove(part, AFS_XFS_ATTR, ATTR_ROOT);
183 return code;
185 #endif /* AFS_SGI_XFS_IOPS_ENV */
188 VInitPartitionPackage(void)
190 #ifdef AFS_DEMAND_ATTACH_FS
191 memset(&DiskPartitionTable, 0, sizeof(DiskPartitionTable));
192 #endif /* AFS_DEMAND_ATTACH_FS */
193 return 0;
196 static void
197 VInitPartition_r(char *path, char *devname, Device dev)
199 struct DiskPartition64 *dp, *op;
201 dp = malloc(sizeof(struct DiskPartition64));
202 /* Add it to the end, to preserve order when we print statistics */
203 for (op = DiskPartitionList; op; op = op->next) {
204 if (!op->next)
205 break;
207 if (op)
208 op->next = dp;
209 else
210 DiskPartitionList = dp;
211 dp->next = 0;
212 dp->name = strdup(path);
213 dp->index = volutil_GetPartitionID(path);
214 #if defined(AFS_NAMEI_ENV) && !defined(AFS_NT40_ENV)
215 /* Create a lockfile for the partition, of the form /vicepa/Lock/vicepa */
216 dp->devName = malloc(2 * strlen(path) + 6);
217 strcpy(dp->devName, path);
218 strcat(dp->devName, OS_DIRSEP);
219 strcat(dp->devName, "Lock");
220 mkdir(dp->devName, 0700);
221 strcat(dp->devName, path);
222 close(afs_open(dp->devName, O_RDWR | O_CREAT, 0600));
223 dp->device = dp->index;
224 #else
225 dp->devName = strdup(devname);
226 dp->device = dev;
227 #endif
228 dp->lock_fd = INVALID_FD;
229 dp->flags = 0;
230 dp->f_files = 1; /* just a default value */
231 #if defined(AFS_NAMEI_ENV) && !defined(AFS_NT40_ENV)
232 if (programType == fileServer)
233 (void)namei_ViceREADME(VPartitionPath(dp));
234 #endif
235 VSetPartitionDiskUsage_r(dp);
236 #ifdef AFS_DEMAND_ATTACH_FS
237 AddPartitionToTable_r(dp);
238 queue_Init(&dp->vol_list.head);
239 CV_INIT(&dp->vol_list.cv, "vol list", CV_DEFAULT, 0);
240 dp->vol_list.len = 0;
241 dp->vol_list.busy = 0;
243 char lockpath[MAXPATHLEN+1];
244 snprintf(lockpath, MAXPATHLEN, "%s/" AFS_PARTLOCK_FILE, dp->name);
245 lockpath[MAXPATHLEN] = '\0';
246 VLockFileInit(&dp->headerLockFile, lockpath);
248 snprintf(lockpath, MAXPATHLEN, "%s/" AFS_VOLUMELOCK_FILE, dp->name);
249 lockpath[MAXPATHLEN] = '\0';
250 VLockFileInit(&dp->volLockFile, lockpath);
252 VDiskLockInit(&dp->headerLock, &dp->headerLockFile, 1);
253 #endif /* AFS_DEMAND_ATTACH_FS */
256 static void
257 VInitPartition(char *path, char *devname, Device dev)
259 VOL_LOCK;
260 VInitPartition_r(path, devname, dev);
261 VOL_UNLOCK;
264 #ifndef AFS_NT40_ENV
265 /* VAttachPartitions() finds the vice partitions on this server. Calls
266 * VCheckPartition() to do some basic checks on the partition. If the partition
267 * is a valid vice partition, VCheckPartition will add it to the DiskPartition
268 * list.
269 * Returns the number of errors returned by VCheckPartition. An error in
270 * VCheckPartition means that partition is a valid vice partition but the
271 * fileserver should not start because of the error found on that partition.
273 * AFS_NAMEI_ENV
274 * No specific user space file system checks, since we don't know what
275 * is being used for vice partitions.
277 * Use partition name as devname.
279 static int
280 VCheckPartition(char *part, char *devname, int logging)
282 struct afs_stat_st status;
283 #if !defined(AFS_LINUX20_ENV) && !defined(AFS_NT40_ENV)
284 char AFSIDatPath[MAXPATHLEN];
285 #endif
287 /* Only keep track of "/vicepx" partitions since it can get hairy
288 * when NFS mounts are involved.. */
289 if (strncmp(part, VICE_PARTITION_PREFIX, VICE_PREFIX_SIZE)) {
290 return 0;
292 if (afs_stat(part, &status) < 0) {
293 Log("VInitVnodes: Couldn't find file system %s; ignored\n", part);
294 return 0;
296 if (logging) {
297 Log("This program is compiled without AFS_NAMEI_ENV, and "
298 "partition %s is mounted with the 'logging' option. "
299 "Using the inode fileserver backend with 'logging' UFS "
300 "partitions causes volume corruption, so please either "
301 "mount the partition without logging, or use the namei "
302 "fileserver backend. Aborting...\n", part);
303 return -1;
305 #ifndef AFS_AIX32_ENV
306 if (programType == fileServer) {
307 char salvpath[MAXPATHLEN];
308 strcpy(salvpath, part);
309 strcat(salvpath, "/FORCESALVAGE");
310 if (afs_stat(salvpath, &status) == 0) {
311 Log("VInitVnodes: Found %s; aborting\n", salvpath);
312 return -1;
315 #endif
317 #if !defined(AFS_LINUX20_ENV) && !defined(AFS_NT40_ENV)
318 strcpy(AFSIDatPath, part);
319 strcat(AFSIDatPath, "/AFSIDat");
320 #ifdef AFS_NAMEI_ENV
321 if (afs_stat(AFSIDatPath, &status) < 0) {
322 DIR *dirp;
323 struct dirent *dp;
325 dirp = opendir(part);
326 opr_Assert(dirp);
327 while ((dp = readdir(dirp))) {
328 if (dp->d_name[0] == 'V') {
329 Log("This program is compiled with AFS_NAMEI_ENV, but partition %s seems to contain volumes which don't use the namei-interface; aborting\n", part);
330 closedir(dirp);
331 return -1;
334 closedir(dirp);
336 #else /* AFS_NAMEI_ENV */
337 if (afs_stat(AFSIDatPath, &status) == 0) {
338 Log("This program is compiled without AFS_NAMEI_ENV, but partition %s seems to contain volumes which use the namei-interface; aborting\n", part);
339 return -1;
342 #ifdef AFS_SGI_XFS_IOPS_ENV
343 if (VerifyXFSInodeSize(part, status.st_fstype) < 0)
344 return -1;
345 #endif
346 #endif /* AFS_NAMEI_ENV */
347 #endif /* !AFS_LINUX20_ENV && !AFS_NT40_ENV */
349 #if defined(AFS_DUX40_ENV) && !defined(AFS_NAMEI_ENV)
350 if (status.st_ino != ROOTINO) {
351 Log("%s is not a mounted file system; ignored.\n", part);
352 return 0;
354 #endif
356 VInitPartition(part, devname, status.st_dev);
358 return 0;
361 /* VIsAlwaysAttach() checks whether a /vicepX directory should always be
362 * attached (return value 1), or only attached when it is a separately
363 * mounted partition (return value 0). For non-NAMEI environments, it
364 * always returns 0.
366 * *awouldattach will be set to 1 if the given path at least looks like a vice
367 * partition (that is, if we return 0, the only thing preventing this partition
368 * from being attached is the existence of the AlwaysAttach file), or to 0
369 * otherwise. *awouldattach is set regardless of whether or not the partition
370 * should always be attached or not.
372 static int
373 VIsAlwaysAttach(char *part, int *awouldattach)
375 #ifdef AFS_NAMEI_ENV
376 struct afs_stat_st st;
377 char checkfile[256];
378 int ret;
379 #endif /* AFS_NAMEI_ENV */
381 if (awouldattach) {
382 *awouldattach = 0;
385 #ifdef AFS_NAMEI_ENV
386 if (strncmp(part, VICE_PARTITION_PREFIX, VICE_PREFIX_SIZE))
387 return 0;
389 if (awouldattach) {
390 *awouldattach = 1;
393 strncpy(checkfile, part, 100);
394 strcat(checkfile, OS_DIRSEP);
395 strcat(checkfile, VICE_ALWAYSATTACH_FILE);
397 ret = afs_stat(checkfile, &st);
398 return (ret < 0) ? 0 : 1;
399 #else /* AFS_NAMEI_ENV */
400 return 0;
401 #endif /* AFS_NAMEI_ENV */
404 /* VIsNeverAttach() checks whether a /vicepX directory should never be
405 * attached (return value 1), or follow the normal mounting logic. The
406 * Always Attach flag may override the NeverAttach flag.
408 static int
409 VIsNeverAttach(char *part)
411 struct afs_stat_st st;
412 char checkfile[256];
413 int ret;
415 if (strncmp(part, VICE_PARTITION_PREFIX, VICE_PREFIX_SIZE))
416 return 0;
418 strncpy(checkfile, part, 100);
419 strcat(checkfile, OS_DIRSEP);
420 strcat(checkfile, VICE_NEVERATTACH_FILE);
422 ret = afs_stat(checkfile, &st);
423 return (ret < 0) ? 0 : 1;
426 /* VAttachPartitions2() looks for and attaches /vicepX partitions
427 * where a special file (VICE_ALWAYSATTACH_FILE) exists. This is
428 * used to attach /vicepX directories which aren't on dedicated
429 * partitions, in the NAMEI fileserver.
431 static void
432 VAttachPartitions2(void)
434 #ifdef AFS_NAMEI_ENV
435 DIR *dirp;
436 struct dirent *de;
437 char pname[32];
438 int wouldattach;
440 dirp = opendir(OS_DIRSEP);
441 while ((de = readdir(dirp))) {
442 strcpy(pname, OS_DIRSEP);
443 strncat(pname, de->d_name, 20);
444 pname[sizeof(pname) - 1] = '\0';
446 /* Only keep track of "/vicepx" partitions since automounter
447 * may hose us */
448 if (VIsAlwaysAttach(pname, &wouldattach)) {
449 VCheckPartition(pname, "", 0);
450 } else {
451 struct afs_stat_st st;
452 if (wouldattach && VGetPartition(pname, 0) == NULL &&
453 afs_stat(pname, &st) == 0 && S_ISDIR(st.st_mode)) {
455 /* This is a /vicep* dir, and it has not been attached as a
456 * partition. This probably means that this is a /vicep* dir
457 * that is not a separate partition, so just give a notice so
458 * admins are not confused as to why their /vicep* dirs are not
459 * being attached.
461 * It is possible that the dir _is_ a separate partition and we
462 * failed to attach it earlier, making this message a bit
463 * confusing. But that should be rare, and an error message
464 * about the failure will already be logged right before this,
465 * so it should be clear enough. */
467 Log("VAttachPartitions: not attaching %s; either it is not a "
468 "separate partition, or it failed to attach (create the "
469 "file %s/" VICE_ALWAYSATTACH_FILE " to force attachment)\n",
470 pname, pname);
474 closedir(dirp);
475 #endif /* AFS_NAMEI_ENV */
477 #endif /* AFS_NT40_ENV */
479 #ifdef AFS_SUN5_ENV
481 VAttachPartitions(void)
483 int errors = 0;
484 struct mnttab mnt;
485 FILE *mntfile;
487 if (!(mntfile = afs_fopen(MNTTAB, "r"))) {
488 Log("Can't open %s\n", MNTTAB);
489 perror(MNTTAB);
490 exit(-1);
492 while (!getmntent(mntfile, &mnt)) {
493 int logging = 0;
494 /* Ignore non ufs or non read/write partitions */
495 /* but allow zfs too if we're in the NAMEI environment */
496 if (
497 #ifdef AFS_NAMEI_ENV
498 (((strcmp(mnt.mnt_fstype, "ufs") &&
499 strcmp(mnt.mnt_fstype, "zfs"))))
500 #else
501 (strcmp(mnt.mnt_fstype, "ufs") != 0)
502 #endif
503 || (strncmp(mnt.mnt_mntopts, "ro,ignore", 9) == 0))
504 continue;
506 /* Skip this Partition? */
507 if (VIsNeverAttach(mnt.mnt_mountp))
508 continue;
510 /* If we're going to always attach this partition, do it later. */
511 if (VIsAlwaysAttach(mnt.mnt_mountp, NULL))
512 continue;
514 #ifndef AFS_NAMEI_ENV
515 if (hasmntopt(&mnt, "logging") != NULL) {
516 logging = 1;
518 #endif /* !AFS_NAMEI_ENV */
520 if (VCheckPartition(mnt.mnt_mountp, mnt.mnt_special, logging) < 0)
521 errors++;
524 (void)fclose(mntfile);
526 /* Process the always-attach partitions, if any. */
527 VAttachPartitions2();
529 return errors;
532 #endif /* AFS_SUN5_ENV */
533 #if defined(AFS_SGI_ENV) || (defined(AFS_SUN_ENV) && !defined(AFS_SUN5_ENV)) || defined(AFS_HPUX_ENV)
535 VAttachPartitions(void)
537 int errors = 0;
538 FILE *mfd;
539 struct mntent *mntent;
541 if ((mfd = setmntent(MOUNTED, "r")) == NULL) {
542 Log("Problems in getting mount entries(setmntent)\n");
543 exit(-1);
545 while (mntent = getmntent(mfd)) {
546 if (!hasmntopt(mntent, MNTOPT_RW))
547 continue;
549 /* Skip this Partition? */
550 if (VIsNeverAttach(mntent->mnt_dir))
551 continue;
553 /* If we're going to always attach this partition, do it later. */
554 if (VIsAlwaysAttach(mntent->mnt_dir, NULL))
555 continue;
557 if (VCheckPartition(mntent->mnt_dir, mntent->mnt_fsname, 0) < 0)
558 errors++;
561 endmntent(mfd);
563 /* Process the always-attach partitions, if any. */
564 VAttachPartitions2();
566 return errors;
568 #endif
569 #ifdef AFS_AIX_ENV
571 * (This function was grabbed from df.c)
574 getmount(struct vmount **vmountpp)
576 int size;
577 struct vmount *vm;
578 int nmounts;
580 /* set initial size of mntctl buffer to a MAGIC NUMBER */
581 size = BUFSIZ;
583 /* try the operation until ok or a fatal error */
584 while (1) {
585 if ((vm = malloc(size)) == NULL) {
586 /* failed getting memory for mount status buf */
587 perror("FATAL ERROR: get_stat malloc failed\n");
588 exit(-1);
592 * perform the QUERY mntctl - if it returns > 0, that is the
593 * number of vmount structures in the buffer. If it returns
594 * -1, an error occured. If it returned 0, then look in
595 * first word of buffer for needed size.
597 if ((nmounts = mntctl(MCTL_QUERY, size, (caddr_t) vm)) > 0) {
598 /* OK, got it, now return */
599 *vmountpp = vm;
600 return (nmounts);
602 } else if (nmounts == 0) {
603 /* the buffer wasn't big enough .... */
604 /* .... get required buffer size */
605 size = *(int *)vm;
606 free(vm);
608 } else {
609 /* some other kind of error occurred */
610 free(vm);
611 return (-1);
617 VAttachPartitions(void)
619 int errors = 0;
620 int nmounts;
621 struct vmount *vmountp;
623 if ((nmounts = getmount(&vmountp)) <= 0) {
624 Log("Problems in getting # of mount entries(getmount)\n");
625 exit(-1);
627 for (; nmounts;
628 nmounts--, vmountp =
629 (struct vmount *)((int)vmountp + vmountp->vmt_length)) {
630 char *part = vmt2dataptr(vmountp, VMT_STUB);
632 if (vmountp->vmt_flags & (MNT_READONLY | MNT_REMOVABLE | MNT_REMOTE))
633 continue; /* Ignore any "special" partitions */
635 #ifdef AFS_AIX42_ENV
636 #ifndef AFS_NAMEI_ENV
638 struct superblock fs;
639 /* The Log statements are non-sequiters in the SalvageLog and don't
640 * even appear in the VolserLog, so restrict them to the FileLog.
642 if (ReadSuper(&fs, vmt2dataptr(vmountp, VMT_OBJECT)) < 0) {
643 if (programType == fileServer)
644 Log("Can't read superblock for %s, ignoring it.\n", part);
645 continue;
647 if (IsBigFilesFileSystem(&fs)) {
648 if (programType == fileServer)
649 Log("%s is a big files filesystem, ignoring it.\n", part);
650 continue;
653 #endif
654 #endif
656 /* Skip this Partition? */
657 if (VIsNeverAttach(part))
658 continue;
660 /* If we're going to always attach this partition, do it later. */
661 if (VIsAlwaysAttach(part, NULL))
662 continue;
664 if (VCheckPartition(part, vmt2dataptr(vmountp, VMT_OBJECT), 0) < 0)
665 errors++;
668 /* Process the always-attach partitions, if any. */
669 VAttachPartitions2();
671 return errors;
673 #endif
674 #if defined(AFS_DUX40_ENV) || defined(AFS_DARWIN_ENV) || defined(AFS_XBSD_ENV)
676 VAttachPartitions(void)
678 int errors = 0;
679 struct fstab *fsent;
681 if (setfsent() < 0) {
682 Log("Error listing filesystems.\n");
683 exit(-1);
686 while ((fsent = getfsent())) {
687 if (strcmp(fsent->fs_type, "rw") != 0)
688 continue;
690 /* Skip this Partition? */
691 if (VIsNeverAttach(fsent->fs_file))
692 continue;
694 /* If we're going to always attach this partition, do it later. */
695 if (VIsAlwaysAttach(fsent->fs_file, NULL))
696 continue;
698 if (VCheckPartition(fsent->fs_file, fsent->fs_spec, 0) < 0)
699 errors++;
701 endfsent();
703 /* Process the always-attach partitions, if any. */
704 VAttachPartitions2();
706 return errors;
708 #endif
710 #ifdef AFS_NT40_ENV
711 /* VValidVPTEntry
713 * validate names in vptab.
715 * Return value:
716 * 1 valid entry
717 * 0 invalid entry
720 static int
721 VValidVPTEntry(struct vptab *vpe)
723 int len = strlen(vpe->vp_name);
724 int i;
726 if (len < VICE_PREFIX_SIZE + 1 || len > VICE_PREFIX_SIZE + 2)
727 return 0;
728 if (strncmp(vpe->vp_name, VICE_PARTITION_PREFIX, VICE_PREFIX_SIZE))
729 return 0;
731 for (i = VICE_PREFIX_SIZE; i < len; i++) {
732 if (vpe->vp_name[i] < 'a' || vpe->vp_name[i] > 'z') {
733 Log("Invalid partition name %s in registry, ignoring it.\n",
734 vpe->vp_name);
735 return 0;
738 if (len == VICE_PREFIX_SIZE + 2) {
739 i = (int)(vpe->vp_name[VICE_PREFIX_SIZE] - 'a') * 26 +
740 (int)(vpe->vp_name[VICE_PREFIX_SIZE + 1] - 'a');
741 if (i > 255) {
742 Log("Invalid partition name %s in registry, ignoring it.\n",
743 vpe->vp_name);
744 return 0;
748 len = strlen(vpe->vp_dev);
749 if (len != 2 || vpe->vp_dev[1] != ':' || vpe->vp_dev[0] < 'A'
750 || vpe->vp_dev[0] > 'Z') {
751 Log("Invalid device name %s in registry, ignoring it.\n",
752 vpe->vp_dev);
753 return 0;
756 return 1;
759 static int
760 VCheckPartition(char *partName)
762 char volRoot[4];
763 char volFsType[64];
764 DWORD dwDummy;
765 int err;
767 /* partName is presumed to be of the form "X:" */
768 (void)sprintf(volRoot, "%c:\\", *partName);
770 if (!GetVolumeInformation(volRoot, /* volume root directory */
771 NULL, /* volume name buffer */
772 0, /* volume name size */
773 NULL, /* volume serial number */
774 &dwDummy, /* max component length */
775 &dwDummy, /* file system flags */
776 volFsType, /* file system name */
777 sizeof(volFsType))) {
778 err = GetLastError();
779 Log("VCheckPartition: Failed to get partition information for %s, ignoring it.\n", partName);
780 return -1;
783 if (strcmp(volFsType, "NTFS")) {
784 Log("VCheckPartition: Partition %s is not an NTFS partition, ignoring it.\n", partName);
785 return -1;
788 return 0;
793 VAttachPartitions(void)
795 struct DiskPartition64 *partP, *prevP, *nextP;
796 struct vpt_iter iter;
797 struct vptab entry;
799 if (vpt_Start(&iter) < 0) {
800 Log("No partitions to attach.\n");
801 return 0;
804 while (0 == vpt_NextEntry(&iter, &entry)) {
805 if (!VValidVPTEntry(&entry)) {
806 continue;
809 /* This test for duplicates relies on the fact that the method
810 * of storing the partition names in the NT registry means the same
811 * partition name will never appear twice in the list.
813 for (partP = DiskPartitionList; partP; partP = partP->next) {
814 if (*partP->devName == *entry.vp_dev) {
815 Log("Same drive (%s) used for both partition %s and partition %s, ignoring both.\n", entry.vp_dev, partP->name, entry.vp_name);
816 partP->flags = PART_DUPLICATE;
817 break; /* Only one entry will ever be in this list. */
820 if (partP)
821 continue; /* found a duplicate */
823 if (VCheckPartition(entry.vp_dev) < 0)
824 continue;
825 /* This test allows for manually inserting the FORCESALVAGE flag
826 * and thereby invoking the salvager. scandisk obviously won't be
827 * doing this for us.
829 if (programType == fileServer) {
830 struct afs_stat_st status;
831 char salvpath[MAXPATHLEN];
832 strcpy(salvpath, entry.vp_dev);
833 strcat(salvpath, "\\FORCESALVAGE");
834 if (afs_stat(salvpath, &status) == 0) {
835 Log("VAttachPartitions: Found %s; aborting\n", salvpath);
836 exit(1);
839 VInitPartition(entry.vp_name, entry.vp_dev, *entry.vp_dev - 'A');
841 vpt_Finish(&iter);
843 /* Run through partition list and clear out the dupes. */
844 prevP = nextP = NULL;
845 for (partP = DiskPartitionList; partP; partP = nextP) {
846 nextP = partP->next;
847 if (partP->flags == PART_DUPLICATE) {
848 if (prevP)
849 prevP->next = partP->next;
850 else
851 DiskPartitionList = partP->next;
852 free(partP);
853 } else
854 prevP = partP;
857 return 0;
859 #endif
861 #ifdef AFS_LINUX22_ENV
863 VAttachPartitions(void)
865 int errors = 0;
866 FILE *mfd;
867 struct mntent *mntent;
869 if ((mfd = setmntent("/proc/mounts", "r")) == NULL) {
870 if ((mfd = setmntent("/etc/mtab", "r")) == NULL) {
871 Log("Problems in getting mount entries(setmntent)\n");
872 exit(-1);
875 while ((mntent = getmntent(mfd))) {
876 /* Skip this Partition? */
877 if (VIsNeverAttach(mntent->mnt_dir))
878 continue;
880 /* If we're going to always attach this partition, do it later. */
881 if (VIsAlwaysAttach(mntent->mnt_dir, NULL))
882 continue;
884 if (VCheckPartition(mntent->mnt_dir, mntent->mnt_fsname, 0) < 0)
885 errors++;
887 endmntent(mfd);
889 /* Process the always-attach partitions, if any. */
890 VAttachPartitions2();
892 return errors;
894 #endif /* AFS_LINUX22_ENV */
896 /* This routine is to be called whenever the actual name of the partition
897 * is required. The canonical name is still in part->name.
899 char *
900 VPartitionPath(struct DiskPartition64 *part)
902 #ifdef AFS_NT40_ENV
903 return part->devName;
904 #else
905 return part->name;
906 #endif
909 /* get partition structure, abortp tells us if we should abort on failure */
910 struct DiskPartition64 *
911 VGetPartition_r(char *name, int abortp)
913 struct DiskPartition64 *dp;
914 #ifdef AFS_DEMAND_ATTACH_FS
915 dp = VLookupPartition_r(name);
916 #else /* AFS_DEMAND_ATTACH_FS */
917 for (dp = DiskPartitionList; dp; dp = dp->next) {
918 if (strcmp(dp->name, name) == 0)
919 break;
921 #endif /* AFS_DEMAND_ATTACH_FS */
922 if (abortp)
923 opr_Assert(dp != NULL);
924 return dp;
927 struct DiskPartition64 *
928 VGetPartition(char *name, int abortp)
930 struct DiskPartition64 *retVal;
931 VOL_LOCK;
932 retVal = VGetPartition_r(name, abortp);
933 VOL_UNLOCK;
934 return retVal;
937 #ifdef AFS_NT40_ENV
938 void
939 VSetPartitionDiskUsage_r(struct DiskPartition64 *dp)
941 ULARGE_INTEGER free_user, total, free_total;
942 int ufree, tot, tfree;
944 if (!GetDiskFreeSpaceEx
945 (VPartitionPath(dp), &free_user, &total, &free_total)) {
946 printf("Failed to get disk space info for %s, error = %d\n", dp->name,
947 GetLastError());
948 return;
951 /* Convert to 1K units. */
952 ufree = (int)Int64ShraMod32(free_user.QuadPart, 10);
953 tot = (int)Int64ShraMod32(total.QuadPart, 10);
954 tfree = (int)Int64ShraMod32(free_total.QuadPart, 10);
956 dp->minFree = tfree - ufree; /* only used in VPrintDiskStats_r */
957 dp->totalUsable = tot;
958 dp->free = tfree;
961 #else
962 void
963 VSetPartitionDiskUsage_r(struct DiskPartition64 *dp)
965 int bsize, code;
966 afs_int64 totalblks, free, used, availblks;
967 int reserved;
968 #ifdef afs_statvfs
969 struct afs_statvfs statbuf;
970 #else
971 struct afs_statfs statbuf;
972 #endif
974 if (dp->flags & PART_DONTUPDATE)
975 return;
976 /* Note: we don't bother syncing because it's only an estimate, update
977 * is syncing every 30 seconds anyway, we only have to keep the disk
978 * approximately 10% from full--you just can't get the stuff in from
979 * the net fast enough to worry */
980 #ifdef afs_statvfs
981 code = afs_statvfs(dp->name, &statbuf);
982 #else
983 code = afs_statfs(dp->name, &statbuf);
984 #endif
985 if (code < 0) {
986 Log("statfs of %s failed in VSetPartitionDiskUsage (errno = %d)\n",
987 dp->name, errno);
988 return;
990 if (statbuf.f_blocks == -1) { /* Undefined; skip stats.. */
991 Log("statfs of %s failed in VSetPartitionDiskUsage\n", dp->name);
992 return;
994 totalblks = statbuf.f_blocks;
995 free = statbuf.f_bfree;
996 reserved = free - statbuf.f_bavail;
997 #ifdef afs_statvfs
998 bsize = statbuf.f_frsize;
999 #else
1000 bsize = statbuf.f_bsize;
1001 #endif
1002 availblks = totalblks - reserved;
1003 dp->f_files = statbuf.f_files; /* max # of files in partition */
1005 /* Now free and totalblks are in fragment units, but we want them in
1006 * 1K units.
1008 if (bsize >= 1024) {
1009 free *= (bsize / 1024);
1010 totalblks *= (bsize / 1024);
1011 availblks *= (bsize / 1024);
1012 reserved *= (bsize / 1024);
1013 } else {
1014 free /= (1024 / bsize);
1015 totalblks /= (1024 / bsize);
1016 availblks /= (1024 / bsize);
1017 reserved /= (1024 / bsize);
1019 /* now compute remaining figures */
1020 used = totalblks - free;
1022 dp->minFree = reserved; /* only used in VPrintDiskStats_r */
1023 dp->totalUsable = availblks;
1024 dp->free = availblks - used; /* this is exactly f_bavail */
1026 #endif /* AFS_NT40_ENV */
1028 void
1029 VSetPartitionDiskUsage(struct DiskPartition64 *dp)
1031 VOL_LOCK;
1032 VSetPartitionDiskUsage_r(dp);
1033 VOL_UNLOCK;
1036 void
1037 VResetDiskUsage_r(void)
1039 struct DiskPartition64 *dp;
1040 for (dp = DiskPartitionList; dp; dp = dp->next) {
1041 VSetPartitionDiskUsage_r(dp);
1042 #ifndef AFS_PTHREAD_ENV
1043 IOMGR_Poll();
1044 #endif /* !AFS_PTHREAD_ENV */
1048 void
1049 VResetDiskUsage(void)
1051 VOL_LOCK;
1052 VResetDiskUsage_r();
1053 VOL_UNLOCK;
1056 void
1057 VAdjustDiskUsage_r(Error * ec, Volume * vp, afs_sfsize_t blocks,
1058 afs_sfsize_t checkBlocks)
1060 *ec = 0;
1061 /* why blocks instead of checkBlocks in the check below? Otherwise, any check
1062 * for less than BlocksSpare would skip the error-checking path, and we
1063 * could grow existing files forever, not just for another BlocksSpare
1064 * blocks. */
1065 if (blocks > 0) {
1066 #ifdef AFS_AIX32_ENV
1067 afs_int32 rem, minavail;
1069 if ((rem = vp->partition->free - checkBlocks) < (minavail =
1070 (vp->partition->
1071 totalUsable *
1072 aixlow_water) /
1073 100))
1074 #else
1075 if (vp->partition->free - checkBlocks < 0)
1076 #endif
1077 *ec = VDISKFULL;
1078 else if (V_maxquota(vp)
1079 && V_diskused(vp) + checkBlocks > V_maxquota(vp))
1080 *ec = VOVERQUOTA;
1082 vp->partition->free -= blocks;
1083 V_diskused(vp) += blocks;
1086 void
1087 VAdjustDiskUsage(Error * ec, Volume * vp, afs_sfsize_t blocks,
1088 afs_sfsize_t checkBlocks)
1090 VOL_LOCK;
1091 VAdjustDiskUsage_r(ec, vp, blocks, checkBlocks);
1092 VOL_UNLOCK;
1096 VDiskUsage_r(Volume * vp, afs_sfsize_t blocks)
1098 if (blocks > 0) {
1099 #ifdef AFS_AIX32_ENV
1100 afs_int32 rem, minavail;
1102 if ((rem = vp->partition->free - blocks) < (minavail =
1103 (vp->partition->
1104 totalUsable *
1105 aixlow_water) / 100))
1106 #else
1107 if (vp->partition->free - blocks < 0)
1108 #endif
1109 return (VDISKFULL);
1111 vp->partition->free -= blocks;
1112 return 0;
1116 VDiskUsage(Volume * vp, afs_sfsize_t blocks)
1118 int retVal;
1119 VOL_LOCK;
1120 retVal = VDiskUsage_r(vp, blocks);
1121 VOL_UNLOCK;
1122 return retVal;
1125 void
1126 VPrintDiskStats_r(void)
1128 struct DiskPartition64 *dp;
1129 for (dp = DiskPartitionList; dp; dp = dp->next) {
1130 if (dp->free < 0) {
1131 Log("Partition %s: %lld "
1132 " available 1K blocks (minfree=%lld), "
1133 "overallocated by %lld blocks\n", dp->name,
1134 dp->totalUsable, dp->minFree, -dp->free);
1135 } else {
1136 Log("Partition %s: %lld"
1137 " available 1K blocks (minfree=%lld), "
1138 "%lld free blocks\n", dp->name,
1139 dp->totalUsable, dp->minFree, dp->free);
1144 void
1145 VPrintDiskStats(void)
1147 VOL_LOCK;
1148 VPrintDiskStats_r();
1149 VOL_UNLOCK;
1152 #ifdef AFS_NT40_ENV
1153 /* Need a separate lock file on NT, since NT only has mandatory file locks. */
1154 #define LOCKFILE "LOCKFILE"
1155 void
1156 VLockPartition_r(char *name)
1158 struct DiskPartition64 *dp = VGetPartition_r(name, 0);
1159 OVERLAPPED lap;
1161 if (!dp)
1162 return;
1163 if (dp->lock_fd == INVALID_FD) {
1164 char path[64];
1165 int rc;
1166 (void)sprintf(path, "%s\\%s", VPartitionPath(dp), LOCKFILE);
1167 dp->lock_fd =
1168 (FD_t)CreateFile(path, GENERIC_WRITE,
1169 FILE_SHARE_READ | FILE_SHARE_WRITE, NULL,
1170 CREATE_ALWAYS, FILE_ATTRIBUTE_HIDDEN, NULL);
1171 opr_Assert(dp->lock_fd != INVALID_FD);
1173 memset(&lap, 0, sizeof(lap));
1174 rc = LockFileEx((HANDLE) dp->lock_fd, LOCKFILE_EXCLUSIVE_LOCK, 0, 1,
1175 0, &lap);
1176 opr_Assert(rc);
1180 void
1181 VUnlockPartition_r(char *name)
1183 struct DiskPartition64 *dp = VGetPartition_r(name, 0);
1184 OVERLAPPED lap;
1186 if (!dp)
1187 return; /* no partition, will fail later */
1188 memset(&lap, 0, sizeof(lap));
1190 UnlockFileEx((HANDLE) dp->lock_fd, 0, 1, 0, &lap);
1191 CloseHandle((HANDLE) dp->lock_fd);
1192 dp->lock_fd = INVALID_FD;
1194 #else /* AFS_NT40_ENV */
1196 #if defined(AFS_HPUX_ENV)
1197 #define BITS_PER_CHAR (8)
1198 #define BITS(type) (sizeof(type) * BITS_PER_CHAR)
1200 #define LOCKRDONLY_OFFSET ((PRIV_LOCKRDONLY - 1) / BITS(int))
1201 #endif /* defined(AFS_HPUX_ENV) */
1203 void
1204 VLockPartition_r(char *name)
1206 struct DiskPartition64 *dp = VGetPartition_r(name, 0);
1207 char *partitionName;
1208 int retries, code;
1209 struct timeval pausing;
1210 #if defined(AFS_HPUX_ENV)
1211 int lockfRtn;
1212 struct privgrp_map privGrpList[PRIV_MAXGRPS];
1213 unsigned int *globalMask;
1214 int globalMaskIndex;
1215 #endif /* defined(AFS_HPUX_ENV) */
1216 #if defined(AFS_DARWIN_ENV)
1217 char lockfile[MAXPATHLEN];
1218 #endif /* defined(AFS_DARWIN_ENV) */
1219 #ifdef AFS_NAMEI_ENV
1220 #ifdef AFS_AIX42_ENV
1221 char LockFileName[MAXPATHLEN + 1];
1223 sprintf((char *)&LockFileName, "%s/AFSINODE_FSLock", name);
1224 partitionName = (char *)&LockFileName;
1225 #endif
1226 #endif
1228 if (!dp)
1229 return; /* no partition, will fail later */
1230 if (dp->lock_fd != INVALID_FD)
1231 return;
1233 #if defined(AFS_SUN5_ENV) || defined(AFS_AIX41_ENV)
1234 #if !defined(AFS_AIX42_ENV) || !defined(AFS_NAMEI_ENV)
1235 partitionName = dp->devName;
1236 #endif
1237 code = O_RDWR;
1238 #elif defined(AFS_DARWIN_ENV)
1239 strlcpy((partitionName = lockfile), dp->name, sizeof(lockfile));
1240 strlcat(lockfile, "/.lock.afs", sizeof(lockfile));
1241 code = O_RDONLY | O_CREAT;
1242 #else
1243 partitionName = dp->name;
1244 code = O_RDONLY;
1245 #endif
1247 for (retries = 25; retries; retries--) {
1248 if (code & O_CREAT)
1249 dp->lock_fd = afs_open(partitionName, code, 0644);
1250 else
1251 dp->lock_fd = afs_open(partitionName, code);
1253 if (dp->lock_fd != INVALID_FD)
1254 break;
1255 if (errno == ENOENT)
1256 code |= O_CREAT;
1257 pausing.tv_sec = 0;
1258 pausing.tv_usec = 500000;
1259 select(0, NULL, NULL, NULL, &pausing);
1261 opr_Assert(retries != 0);
1263 #if defined (AFS_HPUX_ENV)
1265 opr_Verify(getprivgrp(privGrpList) == 0);
1268 * In general, it will difficult and time-consuming ,if not impossible,
1269 * to try to find the privgroup to which this process belongs that has the
1270 * smallest membership, to minimise the security hole. So, we use the privgrp
1271 * to which everybody belongs.
1273 /* first, we have to find the global mask */
1274 for (globalMaskIndex = 0; globalMaskIndex < PRIV_MAXGRPS;
1275 globalMaskIndex++) {
1276 if (privGrpList[globalMaskIndex].priv_groupno == PRIV_GLOBAL) {
1277 globalMask =
1278 &(privGrpList[globalMaskIndex].priv_mask[LOCKRDONLY_OFFSET]);
1279 break;
1283 if (((*globalMask) & privmask(PRIV_LOCKRDONLY)) == 0) {
1284 /* allow everybody to set a lock on a read-only file descriptor */
1285 (*globalMask) |= privmask(PRIV_LOCKRDONLY);
1286 opr_Verify(setprivgrp(PRIV_GLOBAL,
1287 privGrpList[globalMaskIndex].priv_mask) == 0);
1289 lockfRtn = lockf(dp->lock_fd, F_LOCK, 0);
1291 /* remove the privilege granted to everybody to lock a read-only fd */
1292 (*globalMask) &= ~(privmask(PRIV_LOCKRDONLY));
1293 opr_Verify(setprivgrp(PRIV_GLOBAL,
1294 privGrpList[globalMaskIndex].priv_mask) == 0);
1295 } else {
1296 /* in this case, we should be able to do this with impunity, anyway */
1297 lockfRtn = lockf(dp->lock_fd, F_LOCK, 0);
1300 opr_Assert(lockfRtn != -1);
1301 #else
1302 #if defined(AFS_AIX_ENV) || defined(AFS_SUN5_ENV)
1303 opr_Verify(lockf(dp->lock_fd, F_LOCK, 0) != -1);
1304 #else
1305 opr_Verify(flock(dp->lock_fd, LOCK_EX) == 0);
1306 #endif /* defined(AFS_AIX_ENV) || defined(AFS_SUN5_ENV) */
1307 #endif
1310 void
1311 VUnlockPartition_r(char *name)
1313 struct DiskPartition64 *dp = VGetPartition_r(name, 0);
1314 if (!dp)
1315 return; /* no partition, will fail later */
1316 close(dp->lock_fd);
1317 dp->lock_fd = INVALID_FD;
1320 #endif /* AFS_NT40_ENV */
1322 void
1323 VLockPartition(char *name)
1325 VOL_LOCK;
1326 VLockPartition_r(name);
1327 VOL_UNLOCK;
1330 void
1331 VUnlockPartition(char *name)
1333 VOL_LOCK;
1334 VUnlockPartition_r(name);
1335 VOL_UNLOCK;
1338 #ifdef AFS_DEMAND_ATTACH_FS
1340 /* new-style partition locks; these are only to have some mutual exclusion
1341 * between the VGC scanner and volume utilies creating/altering vol headers
1345 * lock a partition's vol headers.
1347 * @param[in] dp the partition to lock
1348 * @param[in] locktype READ_LOCK or WRITE_LOCK
1350 * @return operation status
1351 * @retval 0 success
1354 VPartHeaderLock(struct DiskPartition64 *dp, int locktype)
1356 int code;
1358 /* block on acquiring the lock */
1359 int nonblock = 0;
1361 code = VGetDiskLock(&dp->headerLock, locktype, nonblock);
1362 if (code) {
1363 Log("VPartHeaderLock: error %d locking partititon %s\n", code,
1364 VPartitionPath(dp));
1366 return code;
1370 * unlock a partition's vol headers.
1372 * @param[in] dp the partition to unlock
1373 * @param[in] locktype READ_LOCK or WRITE_LOCK
1375 void
1376 VPartHeaderUnlock(struct DiskPartition64 *dp, int locktype)
1378 VReleaseDiskLock(&dp->headerLock, locktype);
1381 /* XXX not sure this will work on AFS_NT40_ENV
1382 * needs to be tested!
1386 * lookup a disk partition object by its index number.
1388 * @param[in] id partition index number
1389 * @param[in] abortp see abortp usage note below
1391 * @return disk partition object
1392 * @retval NULL no such disk partition
1394 * @note when abortp is non-zero, lookups which would return
1395 * NULL will result in an assertion failure
1397 * @pre VOL_LOCK must be held
1399 * @internal volume package internal use only
1402 struct DiskPartition64 *
1403 VGetPartitionById_r(afs_int32 id, int abortp)
1405 struct DiskPartition64 *dp = NULL;
1407 if ((id >= 0) && (id <= VOLMAXPARTS)) {
1408 dp = DiskPartitionTable[id];
1411 if (abortp) {
1412 opr_Assert(dp != NULL);
1414 return dp;
1418 * lookup a disk partition object by its index number.
1420 * @param[in] id partition index number
1421 * @param[in] abortp see abortp usage note below
1423 * @return disk partition object
1424 * @retval NULL no such disk partition
1426 * @note when abortp is non-zero, lookups which would return
1427 * NULL will result in an assertion failure
1430 struct DiskPartition64 *
1431 VGetPartitionById(afs_int32 id, int abortp)
1433 struct DiskPartition64 * dp;
1435 VOL_LOCK;
1436 dp = VGetPartitionById_r(id, abortp);
1437 VOL_UNLOCK;
1439 return dp;
1442 static struct DiskPartition64 *
1443 VLookupPartition_r(char * path)
1445 afs_int32 id = volutil_GetPartitionID(path);
1447 if (id < 0 || id > VOLMAXPARTS)
1448 return NULL;
1450 return DiskPartitionTable[id];
1453 static void
1454 AddPartitionToTable_r(struct DiskPartition64 *dp)
1456 opr_Assert(dp->index >= 0 && dp->index <= VOLMAXPARTS);
1457 DiskPartitionTable[dp->index] = dp;
1460 #if 0
1461 static void
1462 DeletePartitionFromTable_r(struct DiskPartition64 *dp)
1464 opr_Assert(dp->index >= 0 && dp->index <= VOLMAXPARTS);
1465 DiskPartitionTable[dp->index] = NULL;
1467 #endif
1468 #endif /* AFS_DEMAND_ATTACH_FS */