/*
* mdadm - manage Linux "md" devices aka RAID arrays.
*
- * Copyright (C) 2001-2002 Neil Brown <neilb@cse.unsw.edu.au>
+ * Copyright (C) 2001-2006 Neil Brown <neilb@suse.de>
*
*
* This program is free software; you can redistribute it and/or modify
*/
#include "mdadm.h"
+#include <ctype.h>
+
+static int name_matches(char *found, char *required, char *homehost)
+{
+ /* See if the name found matches the required name, possibly
+ * prefixed with 'homehost'
+ */
+ char fnd[33];
+
+ strncpy(fnd, found, 32);
+ fnd[32] = 0;
+ if (strcmp(found, required)==0)
+ return 1;
+ if (homehost) {
+ int l = strlen(homehost);
+ if (l < 32 && fnd[l] == ':' &&
+ strcmp(fnd+l+1, required)==0)
+ return 1;
+ }
+ return 0;
+}
int Assemble(struct supertype *st, char *mddev, int mdfd,
- mddev_ident_t ident, char *conffile,
- mddev_dev_t devlist,
+ mddev_ident_t ident,
+ mddev_dev_t devlist, char *backup_file,
int readonly, int runstop,
- char *update,
+ char *update, char *homehost,
int verbose, int force)
{
/*
*
* If !uuidset and scan, look in conf-file for uuid
* If not found, give up
- * If !devlist and scan and uuidset, get list of devs from conf-file
+ * If !devlist and scan and uuidset, get list of devs from conf-file
*
* For each device:
* Check superblock - discard if bad
* START_ARRAY
*
*/
+ int must_close = 0;
int old_linux = 0;
- int vers;
+ int vers = 0; /* Keep gcc quite - it really is initialised */
void *first_super = NULL, *super = NULL;
struct {
char *devname;
int chosen_drive;
int change = 0;
int inargv = 0;
- int start_partial_ok = force || devlist==NULL;
+ int start_partial_ok = (runstop >= 0) && (force || devlist==NULL || mdfd < 0);
unsigned int num_devs;
mddev_dev_t tmpdev;
struct mdinfo info;
- struct mddev_ident_s ident2;
char *avail;
int nextspare = 0;
-
- vers = md_get_version(mdfd);
- if (vers <= 0) {
- fprintf(stderr, Name ": %s appears not to be an md device.\n", mddev);
- return 1;
- }
- if (vers < 9000) {
- fprintf(stderr, Name ": Assemble requires driver version 0.90.0 or later.\n"
- " Upgrade your kernel or try --build\n");
- return 1;
- }
+
if (get_linux_version() < 2004000)
old_linux = 1;
- if (ioctl(mdfd, GET_ARRAY_INFO, &info.array)>=0) {
- fprintf(stderr, Name ": device %s already active - cannot assemble it\n",
- mddev);
- return 1;
- }
- ioctl(mdfd, STOP_ARRAY, NULL); /* just incase it was started but has no content */
+ if (mdfd >= 0) {
+ vers = md_get_version(mdfd);
+ if (vers <= 0) {
+ fprintf(stderr, Name ": %s appears not to be an md device.\n", mddev);
+ return 1;
+ }
+ if (vers < 9000) {
+ fprintf(stderr, Name ": Assemble requires driver version 0.90.0 or later.\n"
+ " Upgrade your kernel or try --build\n");
+ return 1;
+ }
+ if (ioctl(mdfd, GET_ARRAY_INFO, &info.array)>=0) {
+ fprintf(stderr, Name ": device %s already active - cannot assemble it\n",
+ mddev);
+ return 1;
+ }
+ ioctl(mdfd, STOP_ARRAY, NULL); /* just incase it was started but has no content */
+ }
/*
* If any subdevs are listed, then any that don't
* match ident are discarded. Remainder must all match and
ident->super_minor < 0 &&
ident->devices == NULL) {
fprintf(stderr, Name ": No identity information available for %s - cannot assemble.\n",
- mddev);
+ mddev ? mddev : "further assembly");
return 1;
}
if (devlist == NULL)
- devlist = conf_get_devs(conffile);
- else inargv = 1;
+ devlist = conf_get_devs();
+ else if (mdfd >= 0)
+ inargv = 1;
+
+ try_again:
tmpdev = devlist; num_devs = 0;
while (tmpdev) {
- num_devs++;
+ if (tmpdev->used)
+ tmpdev->used = 2;
+ else
+ num_devs++;
tmpdev = tmpdev->next;
}
devices = malloc(num_devs * sizeof(*devices));
if (verbose>0)
fprintf(stderr, Name ": looking for devices for %s\n",
- mddev);
+ mddev ? mddev : "further assembly");
- while ( devlist) {
- char *devname;
+ /* first walk the list of devices to find a consistent set
+ * that match the criterea, if that is possible.
+ * We flag the one we like with 'used'.
+ */
+ for (tmpdev = devlist;
+ tmpdev;
+ tmpdev = tmpdev->next) {
+ char *devname = tmpdev->devname;
int dfd;
struct stat stb;
struct supertype *tst = st;
- devname = devlist->devname;
- devlist = devlist->next;
+ if (tmpdev->used > 1) continue;
if (ident->devices &&
!match_oneof(ident->devices, devname)) {
if ((inargv && verbose >= 0) || verbose > 0)
fprintf(stderr, Name ": cannot open device %s: %s\n",
devname, strerror(errno));
+ tmpdev->used = 2;
} else if (fstat(dfd, &stb)< 0) {
/* Impossible! */
fprintf(stderr, Name ": fstat failed for %s: %s\n",
devname, strerror(errno));
+ tmpdev->used = 2;
} else if ((stb.st_mode & S_IFMT) != S_IFBLK) {
fprintf(stderr, Name ": %s is not a block device.\n",
devname);
+ tmpdev->used = 2;
} else if (!tst && (tst = guess_super(dfd)) == NULL) {
if ((inargv && verbose >= 0) || verbose > 0)
- fprintf(stderr, Name ": no recogniseable superblock\n");
+ fprintf(stderr, Name ": no recogniseable superblock on %s\n",
+ devname);
+ tmpdev->used = 2;
} else if (tst->ss->load_super(tst,dfd, &super, NULL)) {
if ((inargv && verbose >= 0) || verbose > 0)
fprintf( stderr, Name ": no RAID superblock on %s\n",
devname);
} else {
- tst->ss->getinfo_super(&info, &ident2, super);
+ tst->ss->getinfo_super(&info, super);
}
if (dfd >= 0) close(dfd);
- if (ident->uuid_set &&
+ if (ident->uuid_set && (!update || strcmp(update, "uuid")!= 0) &&
(!super || same_uuid(info.uuid, ident->uuid, tst->ss->swapuuid)==0)) {
if ((inargv && verbose >= 0) || verbose > 0)
fprintf(stderr, Name ": %s has wrong uuid.\n",
devname);
continue;
}
- if (ident->name[0] &&
- (!super || strncmp(ident2.name, ident->name, 32)!=0)) {
+ if (ident->name[0] && (!update || strcmp(update, "name")!= 0) &&
+ (!super || name_matches(info.name, ident->name, homehost)==0)) {
if ((inargv && verbose >= 0) || verbose > 0)
fprintf(stderr, Name ": %s has wrong name.\n",
devname);
devname);
continue;
}
-
- /* If we are this far, then we are commited to this device.
+ if (mdfd < 0) {
+ if (tst == NULL || super == NULL)
+ continue;
+ if (update == NULL &&
+ tst->ss->match_home(super, homehost)==0) {
+ if ((inargv && verbose >= 0) || verbose > 0)
+ fprintf(stderr, Name ": %s is not built for host %s.\n",
+ devname, homehost);
+ /* Auto-assemble, and this is not a usable host */
+ /* if update != NULL, we are updating the host
+ * name... */
+ continue;
+ }
+ }
+ /* If we are this far, then we are nearly commited to this device.
* If the super_block doesn't exist, or doesn't match others,
- * then we cannot continue
+ * then we probably cannot continue
+ * However if one of the arrays is for the homehost, and
+ * the other isn't that can disambiguate.
*/
if (!super) {
free(first_super);
return 1;
}
- st = tst; /* commit to this format, if haven't already */
- if (st->ss->compare_super(&first_super, super)) {
+
+ if (st == NULL)
+ st = tst;
+ if (st->ss != tst->ss ||
+ st->minor_version != tst->minor_version ||
+ st->ss->compare_super(&first_super, super) != 0) {
+ /* Some mismatch. If exactly one array matches this host,
+ * we can resolve on that one.
+ * Or, if we are auto assembling, we just ignore the second
+ * for now.
+ */
+ if (mdfd < 0)
+ continue;
+ if (homehost) {
+ int first = st->ss->match_home(first_super, homehost);
+ int last = tst->ss->match_home(super, homehost);
+ if (first+last == 1) {
+ /* We can do something */
+ if (first) {/* just ignore this one */
+ if ((inargv && verbose >= 0) || verbose > 0)
+ fprintf(stderr, Name ": %s misses out due to wrong homehost\n",
+ devname);
+ continue;
+ } else { /* reject all those sofar */
+ mddev_dev_t td;
+ if ((inargv && verbose >= 0) || verbose > 0)
+ fprintf(stderr, Name ": %s overrides previous devices due to good homehost\n",
+ devname);
+ for (td=devlist; td != tmpdev; td=td->next)
+ if (td->used == 1)
+ td->used = 0;
+ tmpdev->used = 1;
+ continue;
+ }
+ }
+ }
fprintf(stderr, Name ": superblock on %s doesn't match others - assembly aborted\n",
devname);
free(super);
return 1;
}
+ tmpdev->used = 1;
+ }
+
+ if (mdfd < 0) {
+ /* So... it is up to me to open the device.
+ * We create a name '/dev/md/XXX' based on the info in the
+ * superblock, and call open_mddev on that
+ */
+ mdu_array_info_t inf;
+ char *c;
+ if (!first_super) {
+ return 2;
+ }
+ st->ss->getinfo_super(&info, first_super);
+ c = strchr(info.name, ':');
+ if (c) c++; else c= info.name;
+ if (isdigit(*c) && ((ident->autof & 7)==4 || (ident->autof&7)==6))
+ /* /dev/md/d0 style for partitionable */
+ asprintf(&mddev, "/dev/md/d%s", c);
+ else
+ asprintf(&mddev, "/dev/md/%s", c);
+ mdfd = open_mddev(mddev, ident->autof);
+ if (mdfd < 0) {
+ free(first_super);
+ free(devices);
+ first_super = NULL;
+ goto try_again;
+ }
+ vers = md_get_version(mdfd);
+ if (ioctl(mdfd, GET_ARRAY_INFO, &inf)==0) {
+ for (tmpdev = devlist ;
+ tmpdev && tmpdev->used != 1;
+ tmpdev = tmpdev->next)
+ ;
+ fprintf(stderr, Name ": %s already active, cannot restart it!\n", mddev);
+ if (tmpdev)
+ fprintf(stderr, Name ": %s needed for %s...\n",
+ mddev, tmpdev->devname);
+ close(mdfd);
+ mdfd = -1;
+ free(first_super);
+ free(devices);
+ first_super = NULL;
+ goto try_again;
+ }
+ must_close = 1;
+ }
+
+ /* Ok, no bad inconsistancy, we can try updating etc */
+ for (tmpdev = devlist; tmpdev; tmpdev=tmpdev->next) if (tmpdev->used == 1) {
+ char *devname = tmpdev->devname;
+ struct stat stb;
/* looks like a good enough match to update the super block if needed */
+#ifndef MDASSEMBLE
if (update) {
+ int dfd;
/* prepare useful information in info structures */
struct stat stb2;
fstat(mdfd, &stb2);
- info.array.md_minor = minor(stb2.st_rdev);
-
- st->ss->update_super(&info, super, update, devname, verbose);
-
+
+ if (strcmp(update, "uuid")==0 &&
+ !ident->uuid_set) {
+ int rfd;
+ if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 ||
+ read(rfd, ident->uuid, 16) != 16) {
+ *(__u32*)(ident->uuid) = random();
+ *(__u32*)(ident->uuid+1) = random();
+ *(__u32*)(ident->uuid+2) = random();
+ *(__u32*)(ident->uuid+3) = random();
+ }
+ if (rfd >= 0) close(rfd);
+ }
dfd = dev_open(devname, O_RDWR|O_EXCL);
- if (dfd < 0)
+
+ remove_partitions(dfd);
+
+ if (super) {
+ free(super);
+ super = NULL;
+ }
+
+ st->ss->load_super(st, dfd, &super, NULL);
+ st->ss->getinfo_super(&info, super);
+
+ memcpy(info.uuid, ident->uuid, 16);
+ strcpy(info.name, ident->name);
+ info.array.md_minor = minor(stb2.st_rdev);
+
+ st->ss->update_super(&info, super, update, devname, verbose,
+ ident->uuid_set, homehost);
+ if (strcmp(update, "uuid")==0 &&
+ !ident->uuid_set) {
+ ident->uuid_set = 1;
+ memcpy(ident->uuid, info.uuid, 16);
+ }
+ if (dfd < 0)
fprintf(stderr, Name ": Cannot open %s for superblock update\n",
devname);
else if (st->ss->store_super(st, dfd, super))
devname);
if (dfd >= 0)
close(dfd);
+
+ if (strcmp(update, "uuid")==0 &&
+ ident->bitmap_fd)
+ if (bitmap_update_uuid(ident->bitmap_fd, info.uuid) != 0)
+ fprintf(stderr, Name ": Could not update uuid on %s.\n",
+ devname);
+ } else
+#endif
+ {
+ int dfd;
+ dfd = dev_open(devname, O_RDWR|O_EXCL);
+
+ remove_partitions(dfd);
+
+ if (super) {
+ free(super);
+ super = NULL;
+ }
+
+ st->ss->load_super(st, dfd, &super, NULL);
+ st->ss->getinfo_super(&info, super);
+ close(dfd);
}
+ stat(devname, &stb);
+
if (verbose > 0)
fprintf(stderr, Name ": %s is identified as a member of %s, slot %d.\n",
devname, mddev, info.disk.raid_disk);
> devices[most_recent].events)
most_recent = devcnt;
}
- if (info.array.level == -4)
+ if (info.array.level == -4)
/* with multipath, the raid_disk from the superblock is meaningless */
i = devcnt;
else
if (nextspare < info.array.raid_disks)
nextspare = info.array.raid_disks;
i = nextspare++;
+ } else {
+ if (i >= info.array.raid_disks &&
+ i >= nextspare)
+ nextspare = i+1;
}
if (i < 10000) {
if (i >= bestcnt) {
best = newbest;
bestcnt = newbestcnt;
}
+ if (best[i] >=0 &&
+ devices[best[i]].events == devices[devcnt].events &&
+ devices[best[i]].minor != devices[devcnt].minor &&
+ st->ss->major == 0 &&
+ info.array.level != -4) {
+ /* two different devices with identical superblock.
+ * Could be a mis-detection caused by overlapping
+ * partitions. fail-safe.
+ */
+ fprintf(stderr, Name ": WARNING %s and %s appear"
+ " to have very similar superblocks.\n"
+ " If they are really different, "
+ "please --zero the superblock on one\n"
+ " If they are the same or overlap,"
+ " please remove one from %s.\n",
+ devices[best[i]].devname, devname,
+ inargv ? "the list" :
+ "the\n DEVICE list in mdadm.conf"
+ );
+ if (must_close) close(mdfd);
+ return 1;
+ }
if (best[i] == -1
|| devices[best[i]].events < devices[devcnt].events)
best[i] = devcnt;
fprintf(stderr, Name ": no devices found for %s\n",
mddev);
free(first_super);
+ if (must_close) close(mdfd);
return 1;
}
- st->ss->getinfo_super(&info, &ident2, first_super);
+ st->ss->getinfo_super(&info, first_super);
/* now we have some devices that might be suitable.
* I wonder how many
continue;
}
info.events = devices[most_recent].events;
- st->ss->update_super(&info, super, "force", devices[chosen_drive].devname, verbose);
+ st->ss->update_super(&info, super, "force-one",
+ devices[chosen_drive].devname, verbose,
+ 0, NULL);
if (st->ss->store_super(st, fd, super)) {
close(fd);
if ((fd=dev_open(devices[j].devname, O_RDONLY|O_EXCL))< 0) {
fprintf(stderr, Name ": Cannot open %s: %s\n",
devices[j].devname, strerror(errno));
+ if (must_close) close(mdfd);
return 1;
}
if (st->ss->load_super(st,fd, &super, NULL)) {
close(fd);
fprintf(stderr, Name ": RAID superblock has disappeared from %s\n",
devices[j].devname);
+ if (must_close) close(mdfd);
return 1;
}
close(fd);
}
if (super == NULL) {
fprintf(stderr, Name ": No suitable drives found for %s\n", mddev);
+ if (must_close) close(mdfd);
return 1;
}
- st->ss->getinfo_super(&info, &ident2, super);
+ st->ss->getinfo_super(&info, super);
for (i=0; i<bestcnt; i++) {
int j = best[i];
unsigned int desired_state;
info.disk.state = desired_state;
if (devices[j].uptodate &&
- st->ss->update_super(&info, super, "assemble", NULL, verbose)) {
+ st->ss->update_super(&info, super, "assemble", NULL, verbose, 0, NULL)) {
if (force) {
if (verbose >= 0)
fprintf(stderr, Name ": "
}
#endif
}
- if (force && okcnt == info.array.raid_disks-1) {
- /* FIXME check event count */
- change += st->ss->update_super(&info, super, "force",
- devices[chosen_drive].devname, verbose);
+ if (force && okcnt < info.array.raid_disks) {
+ change += st->ss->update_super(&info, super, "force-array",
+ devices[chosen_drive].devname, verbose,
+ 0, NULL);
}
if (change) {
int fd;
fd = dev_open(devices[chosen_drive].devname, O_RDWR|O_EXCL);
if (fd < 0) {
- fprintf(stderr, Name ": Could open %s for write - cannot Assemble array.\n",
+ fprintf(stderr, Name ": Could not open %s for write - cannot Assemble array.\n",
devices[chosen_drive].devname);
+ if (must_close) close(mdfd);
return 1;
}
if (st->ss->store_super(st, fd, super)) {
close(fd);
fprintf(stderr, Name ": Could not re-write superblock on %s\n",
devices[chosen_drive].devname);
+ if (must_close) close(mdfd);
return 1;
}
close(fd);
}
+ /* If we are in the middle of a reshape we may need to restore saved data
+ * that was moved aside due to the reshape overwriting live data
+ * The code of doing this lives in Grow.c
+ */
+#ifndef MDASSEMBLE
+ if (info.reshape_active) {
+ int err = 0;
+ int *fdlist = malloc(sizeof(int)* bestcnt);
+ for (i=0; i<bestcnt; i++) {
+ int j = best[i];
+ if (j >= 0) {
+ fdlist[i] = dev_open(devices[j].devname, O_RDWR|O_EXCL);
+ if (fdlist[i] < 0) {
+ fprintf(stderr, Name ": Could not open %s for write - cannot Assemble array.\n",
+ devices[j].devname);
+ err = 1;
+ break;
+ }
+ } else
+ fdlist[i] = -1;
+ }
+ if (!err)
+ err = Grow_restart(st, &info, fdlist, bestcnt, backup_file);
+ while (i>0) {
+ i--;
+ if (fdlist[i]>=0) close(fdlist[i]);
+ }
+ if (err) {
+ fprintf(stderr, Name ": Failed to restore critical section for reshape, sorry.\n");
+ if (must_close) close(mdfd);
+ return err;
+ }
+ }
+#endif
/* count number of in-sync devices according to the superblock.
* We must have this number to start the array without -s or -R
*/
inf.major_version = st->ss->major;
inf.minor_version = st->minor_version;
rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
- } else
+ } else
rv = ioctl(mdfd, SET_ARRAY_INFO, NULL);
if (rv) {
fprintf(stderr, Name ": SET_ARRAY_INFO failed for %s: %s\n",
mddev, strerror(errno));
+ if (must_close) close(mdfd);
return 1;
}
if (ident->bitmap_fd >= 0) {
if (ioctl(mdfd, SET_BITMAP_FILE, ident->bitmap_fd) != 0) {
fprintf(stderr, Name ": SET_BITMAP_FILE failed.\n");
+ if (must_close) close(mdfd);
+ return 1;
+ }
+ } else if (ident->bitmap_file) {
+ /* From config file */
+ int bmfd = open(ident->bitmap_file, O_RDWR);
+ if (bmfd < 0) {
+ fprintf(stderr, Name ": Could not open bitmap file %s\n",
+ ident->bitmap_file);
+ if (must_close) close(mdfd);
+ return 1;
+ }
+ if (ioctl(mdfd, SET_BITMAP_FILE, bmfd) != 0) {
+ fprintf(stderr, Name ": Failed to set bitmapfile for %s\n", mddev);
+ close(bmfd);
+ if (must_close) close(mdfd);
return 1;
}
+ close(bmfd);
}
-
+
/* First, add the raid disks, but add the chosen one last */
for (i=0; i<= bestcnt; i++) {
int j;
}
if (runstop == 1 ||
- (runstop == 0 &&
+ (runstop <= 0 &&
( enough(info.array.level, info.array.raid_disks, info.array.layout, avail, okcnt) &&
(okcnt >= req_cnt || start_partial_ok)
))) {
if (verbose >= 0) {
fprintf(stderr, Name ": %s has been started with %d drive%s",
mddev, okcnt, okcnt==1?"":"s");
- if (okcnt < info.array.raid_disks)
+ if (okcnt < info.array.raid_disks)
fprintf(stderr, " (out of %d)", info.array.raid_disks);
if (sparecnt)
fprintf(stderr, " and %d spare%s", sparecnt, sparecnt==1?"":"s");
fprintf(stderr, ".\n");
}
+ if (must_close) {
+ int usecs = 1;
+ close(mdfd);
+ /* There is a nasty race with 'mdadm --monitor'.
+ * If it opens this device before we close it,
+ * it gets an incomplete open on which IO
+ * doesn't work and the capacity if wrong.
+ * If we reopen (to check for layered devices)
+ * before --monitor closes, we loose.
+ *
+ * So: wait upto 1 second for there to be
+ * a non-zero capacity.
+ */
+ while (usecs < 1000) {
+ mdfd = open(mddev, O_RDONLY);
+ if (mdfd >= 0) {
+ unsigned long size;
+ if (ioctl(mdfd, BLKGETSIZE, &size) == 0 &&
+ size > 0)
+ break;
+ close(mdfd);
+ }
+ usleep(usecs);
+ usecs <<= 1;
+ }
+ }
return 0;
}
fprintf(stderr, Name ": failed to RUN_ARRAY %s: %s\n",
mddev, strerror(errno));
+ if (must_close) close(mdfd);
return 1;
}
if (runstop == -1) {
- fprintf(stderr, Name ": %s assembled from %d drive%s, but not started.\n",
+ fprintf(stderr, Name ": %s assembled from %d drive%s",
mddev, okcnt, okcnt==1?"":"s");
+ if (okcnt != info.array.raid_disks)
+ fprintf(stderr, " (out of %d)", info.array.raid_disks);
+ fprintf(stderr, ", but not started.\n");
+ if (must_close) close(mdfd);
return 0;
}
if (verbose >= 0) {
fprintf(stderr, " (use --run to insist).\n");
}
}
+ if (must_close) close(mdfd);
return 1;
} else {
/* The "chosen_drive" is a good choice, and if necessary, the superblock has
}
}
+ if (must_close) close(mdfd);
return 0;
}