/*
- * mdctl - manage Linux "md" devices aka RAID arrays.
+ * mdadm - manage Linux "md" devices aka RAID arrays.
*
- * Copyright (C) 2001 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
* Australia
*/
-#include "mdctl.h"
-#include "md_p.h"
-#include "md_u.h"
+#include "mdadm.h"
+#include <ctype.h>
-int Assemble(char *mddev, int mdfd,
- int uuid[4], int uuidset,
- char *conffile, int scan,
- int subdevs, char **subdev,
+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,
+ mddev_dev_t devlist, char *backup_file,
int readonly, int runstop,
+ char *update, char *homehost,
int verbose, int force)
{
/*
- * The task of Assemble is to submit a
+ * The task of Assemble is to find a collection of
+ * devices that should (according to their superblocks)
+ * form an array, and to give this collection to the MD driver.
+ * In Linux-2.4 and later, this involves submitting a
* SET_ARRAY_INFO ioctl with no arg - to prepare
* the array - and then submit a number of
* ADD_NEW_DISK ioctls to add disks into
* Much of the work of Assemble is in finding and/or
* checking the disks to make sure they look right.
*
- * If mddev is not set, then scan must be and we
+ * If mddev is not set, then scan must be set and we
* read through the config file for dev+uuid mapping
* We recurse, setting mddev, for each device that
* - isn't running
- * - has a valid uuid (or any uuid if !uuidset
+ * - has a valid uuid (or any uuid if !uuidset)
*
* If mddev is set, we try to determine state of md.
* check version - must be at least 0.90.0
*
* If !uuidset and scan, look in conf-file for uuid
* If not found, give up
- * If !subdevs 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
* Check uuid (set if we don't have one) - discard if no match
- * Check superblock similarity if we have a superbloc - discard if different
- * Record events, devicenum, utime
+ * Check superblock similarity if we have a superblock - discard if different
+ * Record events, devicenum
* This should give us a list of devices for the array
- * We should collect the most recent event and utime numbers
+ * We should collect the most recent event number
*
* Count disks with recent enough event count
* While force && !enough disks
* START_ARRAY
*
*/
+ int clean = 0;
+ int must_close = 0;
int old_linux = 0;
- int vers;
- mdu_array_info_t array;
- mddev_dev_t devlist = NULL;
- mdp_super_t first_super, super;
+ int vers = 0; /* Keep gcc quite - it really is initialised */
struct {
char *devname;
- int major, minor;
- long long events;
- time_t utime;
- int uptodate;
- } devices[MD_SB_DISKS];
- int best[MD_SB_DISKS]; /* indexed by raid_disk */
- int devcnt = 0, okcnt;
- int i;
+ int uptodate; /* set once we decide that this device is as
+ * recent as everything else in the array.
+ */
+ struct mdinfo i;
+ } *devices;
+ int *best = NULL; /* indexed by raid_disk */
+ unsigned int bestcnt = 0;
+ int devcnt = 0;
+ unsigned int okcnt, sparecnt;
+ unsigned int req_cnt;
+ unsigned int i;
int most_recent = 0;
-
- if (!mddev && !scan) {
- fputs(Name ": internal error - Assemble called with no device or --scan\n", stderr);
- return 1;
- }
- if (!mddev) {
- mddev_uuid_t device_list;
- int found = 0;
- device_list = conf_get_uuids(conffile);
- if (!device_list) {
- fprintf(stderr, Name ": No devices found in config file\n");
- return 1;
- }
- for (; device_list; device_list=device_list->next) {
- if (!uuidset || same_uuid(device_list->uuid,uuid)) {
- mdfd = open(device_list->devname, O_RDONLY, 0);
- if (mdfd < 0) {
- fprintf(stderr,
- Name ": error opening %s: %s\n",
- device_list->devname,
- strerror(errno));
- continue;
- }
- if (Assemble(device_list->devname, mdfd,
- device_list->uuid, 1,
- conffile, 1,
- subdevs, subdev,
- readonly, runstop, verbose, force)==0)
- found++;
- close(mdfd);
- }
- }
- if (found)
- return 0;
- fprintf(stderr,Name ": Did not successfully Assemble any devices\n");
- return 1;
- }
+ int chosen_drive;
+ int change = 0;
+ int inargv = 0;
+ int bitmap_done;
+ int start_partial_ok = (runstop >= 0) && (force || devlist==NULL || mdfd < 0);
+ unsigned int num_devs;
+ mddev_dev_t tmpdev;
+ struct mdinfo info;
+ char *avail;
+ int nextspare = 0;
- /*
- * Ok, we have an mddev, check it out
- */
- vers = md_get_version(mdfd);
- if (vers <= 0) {
- fprintf(stderr, Name ": %s appears not to be an md device.\n");
- 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, &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;
+ }
- /*
- * We have a valid mddev, check out uuid
- */
- if (!uuidset && scan) {
- /* device must be listed with uuid in conf file */
- mddev_uuid_t device_list;
- device_list = conf_get_uuids(conffile);
- while (device_list &&
- strcmp(device_list->devname, mddev) != 0)
- device_list = device_list->next;
-
- if (!device_list) {
- fprintf(stderr, Name ": --scan set and no uuid found for %s in config file.\n",
+ if (ioctl(mdfd, GET_ARRAY_INFO, &info.array)>=0) {
+ fprintf(stderr, Name ": device %s already active - cannot assemble it\n",
mddev);
return 1;
}
- /* the uuid is safe until next call to conf_get_uuids */
- uuid = device_list->uuid;
- uuidset = 1;
+ ioctl(mdfd, STOP_ARRAY, NULL); /* just incase it was started but has no content */
}
-
- /* Now to start looking at devices.
- * If no devices were given, but a uuid is available and
- * --scan was set, then we should scan all devices listed in the
- * config file
- *
+ /*
+ * If any subdevs are listed, then any that don't
+ * match ident are discarded. Remainder must all match and
+ * become the array.
+ * If no subdevs, then we scan all devices in the config file, but
+ * there must be something in the identity
*/
- if (subdevs==0 && scan && uuidset)
- devlist = conf_get_devs(conffile);
- if (subdevs == 0 && devlist == NULL) {
- fprintf(stderr, Name ": no devices given for %s\n", mddev);
+ if (!devlist &&
+ ident->uuid_set == 0 &&
+ ident->super_minor < 0 &&
+ ident->devices == NULL) {
+ fprintf(stderr, Name ": No identity information available for %s - cannot assemble.\n",
+ mddev ? mddev : "further assembly");
return 1;
}
- /* now for each device */
- first_super.md_magic = 0;
- for (i=0; i<MD_SB_DISKS; i++)
- best[i] = -1;
+ if (devlist == NULL)
+ devlist = conf_get_devs();
+ else if (mdfd >= 0)
+ inargv = 1;
+
+ try_again:
+
+ tmpdev = devlist; num_devs = 0;
+ while (tmpdev) {
+ if (tmpdev->used)
+ tmpdev->used = 2;
+ else
+ num_devs++;
+ tmpdev = tmpdev->next;
+ }
+ devices = malloc(num_devs * sizeof(*devices));
- if (verbose)
+ if (!st && ident->st) st = ident->st;
+
+ if (verbose>0)
fprintf(stderr, Name ": looking for devices for %s\n",
- mddev);
+ mddev ? mddev : "further assembly");
- while (subdevs || devlist) {
- char *devname;
- int this_uuid[4];
+ /* 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;
- int inargv;
- if (subdevs) {
- devname = *subdev++;
- subdevs--;
- inargv=1;
- } else {
- devname = devlist->devname;
- devlist = devlist->next;
- inargv=0;
+ struct supertype *tst = dup_super(st);
+
+ if (tmpdev->used > 1) continue;
+
+ if (ident->devices &&
+ !match_oneof(ident->devices, devname)) {
+ if ((inargv && verbose>=0) || verbose > 0)
+ fprintf(stderr, Name ": %s is not one of %s\n", devname, ident->devices);
+ continue;
}
- dfd = open(devname, O_RDONLY, 0);
+ dfd = dev_open(devname, O_RDONLY|O_EXCL);
if (dfd < 0) {
- if (inargv || verbose)
+ if ((inargv && verbose >= 0) || verbose > 0)
fprintf(stderr, Name ": cannot open device %s: %s\n",
devname, strerror(errno));
- continue;
- }
- if (fstat(dfd, &stb)< 0) {
- /* Impossible! */
- fprintf(stderr, Name ": fstat failed for %s: %s\n",
- devname, strerror(errno));
- close(dfd);
- continue;
- }
- if ((stb.st_mode & S_IFMT) != S_IFBLK) {
- fprintf(stderr, Name ": %d is not a block device.\n",
- devname);
- close(dfd);
- continue;
- }
- if (load_super(dfd, &super)) {
- if (inargv || verbose)
+ 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 on %s\n",
+ devname);
+ tmpdev->used = 2;
+ } else if (tst->ss->load_super(tst,dfd, NULL)) {
+ if ((inargv && verbose >= 0) || verbose > 0)
fprintf( stderr, Name ": no RAID superblock on %s\n",
devname);
- close(dfd);
- continue;
+ } else {
+ tst->ss->getinfo_super(tst, &info);
}
- close(dfd);
- uuid_from_super(this_uuid, &super);
- if (uuidset && !same_uuid(this_uuid, uuid)) {
- if (inargv || verbose)
- fprintf(stderr, Name ": %s has wrong uuid.\n",
- devname);
- continue;
+ if (dfd >= 0) close(dfd);
+
+ if (ident->uuid_set && (!update || strcmp(update, "uuid")!= 0) &&
+ (!tst || !tst->sb ||
+ 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);
+ goto loop;
}
- if (compare_super(&first_super, &super)) {
- if (inargv || verbose)
- fprintf(stderr, Name ": superblock on %s doesn't match\n",
+ if (ident->name[0] && (!update || strcmp(update, "name")!= 0) &&
+ (!tst || !tst->sb ||
+ name_matches(info.name, ident->name, homehost)==0)) {
+ if ((inargv && verbose >= 0) || verbose > 0)
+ fprintf(stderr, Name ": %s has wrong name.\n",
devname);
- continue;
+ goto loop;
+ }
+ if (ident->super_minor != UnSet &&
+ (!tst || !tst->sb ||
+ ident->super_minor != info.array.md_minor)) {
+ if ((inargv && verbose >= 0) || verbose > 0)
+ fprintf(stderr, Name ": %s has wrong super-minor.\n",
+ devname);
+ goto loop;
+ }
+ if (ident->level != UnSet &&
+ (!tst || !tst->sb ||
+ ident->level != info.array.level)) {
+ if ((inargv && verbose >= 0) || verbose > 0)
+ fprintf(stderr, Name ": %s has wrong raid level.\n",
+ devname);
+ goto loop;
+ }
+ if (ident->raid_disks != UnSet &&
+ (!tst || !tst->sb ||
+ ident->raid_disks!= info.array.raid_disks)) {
+ if ((inargv && verbose >= 0) || verbose > 0)
+ fprintf(stderr, Name ": %s requires wrong number of drives.\n",
+ devname);
+ goto loop;
}
- if (uuidset) {
- uuid_from_super(this_uuid, &first_super);
- if (!same_uuid(this_uuid, uuid)) {
- if (inargv || verbose)
- fprintf(stderr, Name ": %s has wrong uuid.\n",
- devname);
+ if (mdfd < 0) {
+ if (tst == NULL || tst->sb == NULL)
continue;
+ if (update == NULL &&
+ tst->ss->match_home(tst, 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... */
+ goto loop;
}
- } else {
- uuid_from_super(uuid, &first_super);
- uuidset = 1;
}
+ /* 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 probably cannot continue
+ * However if one of the arrays is for the homehost, and
+ * the other isn't that can disambiguate.
+ */
- /* Ok, this one is at least worth considering */
- if (devcnt >= MD_SB_DISKS) {
- fprintf(stderr, Name ": ouch - too many devices appear to be in this array. Ignoring %s\n",
- devname);
- continue;
+ if (!tst || !tst->sb) {
+ fprintf(stderr, Name ": %s has no superblock - assembly aborted\n",
+ devname);
+ if (st)
+ st->ss->free_super(st);
+ return 1;
}
+
+ if (st == NULL)
+ st = dup_super(tst);
+ if (st->minor_version == -1)
+ st->minor_version = tst->minor_version;
+ if (st->ss != tst->ss ||
+ st->minor_version != tst->minor_version ||
+ st->ss->compare_super(st, tst) != 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)
+ goto loop;
+ if (homehost) {
+ int first = st->ss->match_home(st, homehost);
+ int last = tst->ss->match_home(tst, 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);
+ goto loop;
+ } 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;
+ goto loop;
+ }
+ }
+ }
+ fprintf(stderr, Name ": superblock on %s doesn't match others - assembly aborted\n",
+ devname);
+ tst->ss->free_super(tst);
+ st->ss->free_super(st);
+ return 1;
+ }
+
+ tmpdev->used = 1;
+
+ loop:
+ tst->ss->free_super(tst);
+ }
+
+ 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 (!st->sb) {
+ return 2;
+ }
+ st->ss->getinfo_super(st, &info);
+ 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) {
+ st->ss->free_super(st);
+ free(devices);
+ 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;
+ st->ss->free_super(st);
+ free(devices);
+ goto try_again;
+ }
+ must_close = 1;
+ }
+
+ /* Ok, no bad inconsistancy, we can try updating etc */
+ bitmap_done = 0;
+ 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;
+ struct supertype *tst;
+ fstat(mdfd, &stb2);
+
+ 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);
+
+ remove_partitions(dfd);
+
+ tst = dup_super(st);
+ tst->ss->load_super(tst, dfd, NULL);
+ tst->ss->getinfo_super(tst, &info);
+
+ memcpy(info.uuid, ident->uuid, 16);
+ strcpy(info.name, ident->name);
+ info.array.md_minor = minor(stb2.st_rdev);
+
+ tst->ss->update_super(tst, &info, 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 (tst->ss->store_super(tst, dfd))
+ fprintf(stderr, Name ": Could not re-write superblock on %s.\n",
+ devname);
+ if (dfd >= 0)
+ close(dfd);
+
+ if (strcmp(update, "uuid")==0 &&
+ ident->bitmap_fd >= 0 && !bitmap_done) {
+ if (bitmap_update_uuid(ident->bitmap_fd,
+ info.uuid,
+ tst->ss->swapuuid) != 0)
+ fprintf(stderr, Name ": Could not update uuid on external bitmap.\n");
+ else
+ bitmap_done = 1;
+ }
+ tst->ss->free_super(tst);
+ } else
+#endif
+ {
+ struct supertype *tst = dup_super(st);;
+ int dfd;
+ dfd = dev_open(devname, O_RDWR|O_EXCL);
+
+ remove_partitions(dfd);
+
+ tst->ss->load_super(tst, dfd, NULL);
+ tst->ss->getinfo_super(tst, &info);
+ tst->ss->free_super(tst);
+ 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[devcnt].devname = devname;
- devices[devcnt].major = MAJOR(stb.st_rdev);
- devices[devcnt].minor = MINOR(stb.st_rdev);
- devices[devcnt].events = md_event(&super);
- devices[devcnt].utime = super.utime;
devices[devcnt].uptodate = 0;
+ devices[devcnt].i = info;
+ devices[devcnt].i.disk.major = major(stb.st_rdev);
+ devices[devcnt].i.disk.minor = minor(stb.st_rdev);
if (most_recent < devcnt) {
- if (devices[devcnt].events
- > devices[most_recent].events)
+ if (devices[devcnt].i.events
+ > devices[most_recent].i.events)
most_recent = devcnt;
}
- i = super.this_disk.raid_disk;
- if (best[i] == -1
- || devices[best[i]].events < devices[devcnt].events) {
- best[i] = devcnt;
+ if (info.array.level == -4)
+ /* with multipath, the raid_disk from the superblock is meaningless */
+ i = devcnt;
+ else
+ i = devices[devcnt].i.disk.raid_disk;
+ if (i+1 == 0) {
+ 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) {
+ unsigned int newbestcnt = i+10;
+ int *newbest = malloc(sizeof(int)*newbestcnt);
+ unsigned int c;
+ for (c=0; c < newbestcnt; c++)
+ if (c < bestcnt)
+ newbest[c] = best[c];
+ else
+ newbest[c] = -1;
+ if (best)free(best);
+ best = newbest;
+ bestcnt = newbestcnt;
+ }
+ if (best[i] >=0 &&
+ devices[best[i]].i.events
+ == devices[devcnt].i.events
+ && (devices[best[i]].i.disk.minor
+ != devices[devcnt].i.disk.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]].i.events
+ < devices[devcnt].i.events))
+ best[i] = devcnt;
}
devcnt++;
}
+ if (update && strcmp(update, "byteorder")==0)
+ st->minor_version = 90;
+
if (devcnt == 0) {
fprintf(stderr, Name ": no devices found for %s\n",
mddev);
+ st->ss->free_super(st);
+ if (must_close) close(mdfd);
return 1;
}
+
+ st->ss->getinfo_super(st, &info);
+ clean = info.array.state & 1;
+
/* now we have some devices that might be suitable.
* I wonder how many
*/
+ avail = malloc(info.array.raid_disks);
+ memset(avail, 0, info.array.raid_disks);
okcnt = 0;
- for (i=0; i< first_super.raid_disks;i++) {
+ sparecnt=0;
+ for (i=0; i< bestcnt ;i++) {
int j = best[i];
+ int event_margin = 1; /* always allow a difference of '1'
+ * like the kernel does
+ */
if (j < 0) continue;
- if (devices[j].events+1 >=
- devices[most_recent].events) {
+ /* note: we ignore error flags in multipath arrays
+ * as they don't make sense
+ */
+ if (info.array.level != -4)
+ if (!(devices[j].i.disk.state & (1<<MD_DISK_SYNC))) {
+ if (!(devices[j].i.disk.state
+ & (1<<MD_DISK_FAULTY)))
+ sparecnt++;
+ continue;
+ }
+ if (devices[j].i.events+event_margin >=
+ devices[most_recent].i.events) {
devices[j].uptodate = 1;
- okcnt++;
+ if (i < info.array.raid_disks) {
+ okcnt++;
+ avail[i]=1;
+ } else
+ sparecnt++;
}
}
- while (force && !enough(first_super.level, first_super.raid_disks, okcnt)) {
+ while (force && !enough(info.array.level, info.array.raid_disks,
+ info.array.layout, 1,
+ avail, okcnt)) {
/* Choose the newest best drive which is
* not up-to-date, update the superblock
* and add it.
*/
- fprintf(stderr,"NotImplementedYet\n");
- /* FIXME */
- exit(2);
+ int fd;
+ struct supertype *tst;
+ long long current_events;
+ chosen_drive = -1;
+ for (i=0; i<info.array.raid_disks && i < bestcnt; i++) {
+ int j = best[i];
+ if (j>=0 &&
+ !devices[j].uptodate &&
+ devices[j].i.events > 0 &&
+ (chosen_drive < 0 ||
+ devices[j].i.events
+ > devices[chosen_drive].i.events))
+ chosen_drive = j;
+ }
+ if (chosen_drive < 0)
+ break;
+ current_events = devices[chosen_drive].i.events;
+ add_another:
+ if (verbose >= 0)
+ fprintf(stderr, Name ": forcing event count in %s(%d) from %d upto %d\n",
+ devices[chosen_drive].devname,
+ devices[chosen_drive].i.disk.raid_disk,
+ (int)(devices[chosen_drive].i.events),
+ (int)(devices[most_recent].i.events));
+ fd = dev_open(devices[chosen_drive].devname, O_RDWR|O_EXCL);
+ if (fd < 0) {
+ fprintf(stderr, Name ": Couldn't open %s for write - not updating\n",
+ devices[chosen_drive].devname);
+ devices[chosen_drive].i.events = 0;
+ continue;
+ }
+ tst = dup_super(st);
+ if (tst->ss->load_super(st,fd, NULL)) {
+ close(fd);
+ fprintf(stderr, Name ": RAID superblock disappeared from %s - not updating.\n",
+ devices[chosen_drive].devname);
+ devices[chosen_drive].i.events = 0;
+ continue;
+ }
+ info.events = devices[most_recent].i.events;
+ tst->ss->update_super(tst, &info, "force-one",
+ devices[chosen_drive].devname, verbose,
+ 0, NULL);
+
+ if (tst->ss->store_super(tst, fd)) {
+ close(fd);
+ fprintf(stderr, Name ": Could not re-write superblock on %s\n",
+ devices[chosen_drive].devname);
+ devices[chosen_drive].i.events = 0;
+ tst->ss->free_super(tst);
+ continue;
+ }
+ close(fd);
+ devices[chosen_drive].i.events = devices[most_recent].i.events;
+ devices[chosen_drive].uptodate = 1;
+ avail[chosen_drive] = 1;
+ okcnt++;
+ tst->ss->free_super(tst);
+
+ /* If there are any other drives of the same vintage,
+ * add them in as well. We can't lose and we might gain
+ */
+ for (i=0; i<info.array.raid_disks && i < bestcnt ; i++) {
+ int j = best[i];
+ if (j >= 0 &&
+ !devices[j].uptodate &&
+ devices[j].i.events > 0 &&
+ devices[j].i.events == current_events) {
+ chosen_drive = j;
+ goto add_another;
+ }
+ }
+ }
+
+ /* Now we want to look at the superblock which the kernel will base things on
+ * and compare the devices that we think are working with the devices that the
+ * superblock thinks are working.
+ * If there are differences and --force is given, then update this chosen
+ * superblock.
+ */
+ chosen_drive = -1;
+ st->ss->free_super(st);
+ for (i=0; chosen_drive < 0 && i<bestcnt; i++) {
+ int j = best[i];
+ int fd;
+
+ if (j<0)
+ continue;
+ if (!devices[j].uptodate)
+ continue;
+ chosen_drive = j;
+ 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, 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 (st->sb == NULL) {
+ fprintf(stderr, Name ": No suitable drives found for %s\n", mddev);
+ if (must_close) close(mdfd);
+ return 1;
+ }
+ st->ss->getinfo_super(st, &info);
+ for (i=0; i<bestcnt; i++) {
+ int j = best[i];
+ unsigned int desired_state;
+
+ if (i < info.array.raid_disks)
+ desired_state = (1<<MD_DISK_ACTIVE) | (1<<MD_DISK_SYNC);
+ else
+ desired_state = 0;
+
+ if (j<0)
+ continue;
+ if (!devices[j].uptodate)
+ continue;
+
+ devices[j].i.disk.state = desired_state;
+
+ if (st->ss->update_super(st, &devices[j].i, "assemble", NULL,
+ verbose, 0, NULL)) {
+ if (force) {
+ if (verbose >= 0)
+ fprintf(stderr, Name ": "
+ "clearing FAULTY flag for device %d in %s for %s\n",
+ j, mddev, devices[j].devname);
+ change = 1;
+ } else {
+ if (verbose >= -1)
+ fprintf(stderr, Name ": "
+ "device %d in %s has wrong state in superblock, but %s seems ok\n",
+ i, mddev, devices[j].devname);
+ }
+ }
+#if 0
+ if (!(super.disks[i].i.disk.state & (1 << MD_DISK_FAULTY))) {
+ fprintf(stderr, Name ": devices %d of %s is not marked FAULTY in superblock, but cannot be found\n",
+ i, mddev);
+ }
+#endif
+ }
+ if (force && !clean &&
+ !enough(info.array.level, info.array.raid_disks,
+ info.array.layout, clean,
+ avail, okcnt)) {
+ change += st->ss->update_super(st, &info, "force-array",
+ devices[chosen_drive].devname, verbose,
+ 0, NULL);
+ clean = 1;
+ }
+
+ if (change) {
+ int fd;
+ fd = dev_open(devices[chosen_drive].devname, O_RDWR|O_EXCL);
+ if (fd < 0) {
+ 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)) {
+ 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
+ */
+ req_cnt = info.array.working_disks;
+
/* Almost ready to actually *do* something */
if (!old_linux) {
- if (ioctl(mdfd, SET_ARRAY_INFO, NULL) != 0) {
+ int rv;
+ if ((vers % 100) >= 1) { /* can use different versions */
+ mdu_array_info_t inf;
+ memset(&inf, 0, sizeof(inf));
+ inf.major_version = st->ss->major;
+ inf.minor_version = st->minor_version;
+ rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
+ } 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;
}
- /* First, add the raid disks */
- for (i=0; i<first_super.raid_disks; i++) {
- int j = best[i];
- if (devices[j].uptodate) {
- mdu_disk_info_t disk;
- memset(&disk, 0, sizeof(disk));
- disk.major = devices[j].major;
- disk.minor = devices[j].minor;
- if (ioctl(mdfd, ADD_NEW_DISK, &disk)!=0) {
- fprintf(stderr, Name ": failed to add %s to %s: %s\n",
+ 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 (i < bestcnt) {
+ j = best[i];
+ if (j == chosen_drive)
+ continue;
+ } else
+ j = chosen_drive;
+
+ if (j >= 0 /* && devices[j].uptodate */) {
+ if (ioctl(mdfd, ADD_NEW_DISK,
+ &devices[j].i.disk)!=0) {
+ fprintf(stderr, Name ": failed to add "
+ "%s to %s: %s\n",
devices[j].devname,
mddev,
strerror(errno));
- okcnt--;
- }
- } else if (verbose)
- fprintf(stderr, Name ": no uptodate device for slot %d of %s\n",
+ if (i < info.array.raid_disks
+ || i == bestcnt)
+ okcnt--;
+ else
+ sparecnt--;
+ } else if (verbose > 0)
+ fprintf(stderr, Name ": added %s "
+ "to %s as %d\n",
+ devices[j].devname, mddev,
+ devices[j].i.disk.raid_disk);
+ } else if (verbose > 0 && i < info.array.raid_disks)
+ fprintf(stderr, Name ": no uptodate device for "
+ "slot %d of %s\n",
i, mddev);
}
+
if (runstop == 1 ||
- (runstop == 0 &&
- enough(first_super.level, first_super.raid_disks, okcnt))) {
+ (runstop <= 0 &&
+ ( enough(info.array.level, info.array.raid_disks,
+ info.array.layout, clean, avail, okcnt) &&
+ (okcnt >= req_cnt || start_partial_ok)
+ ))) {
if (ioctl(mdfd, RUN_ARRAY, NULL)==0) {
- fprintf(stderr, Name ": %s has been started with %d drives\n",
- mddev, okcnt);
+ 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)
+ 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 long size;
+ if (get_dev_size(mdfd, NULL, &size) &&
+ 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 (!enough(info.array.level, info.array.raid_disks,
+ info.array.layout, 1, avail, okcnt))
+ fprintf(stderr, Name ": Not enough devices to "
+ "start the array.\n");
+ else if (!enough(info.array.level,
+ info.array.raid_disks,
+ info.array.layout, clean,
+ avail, okcnt))
+ fprintf(stderr, Name ": Not enough devices to "
+ "start the array while not clean "
+ "- consider --force.\n");
+
+ if (must_close) close(mdfd);
return 1;
}
if (runstop == -1) {
- fprintf(stderr, Name ": %s assembled from %d drives, but not started.\n",
- mddev, okcnt);
+ 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;
}
- fprintf(stderr, Name ": %s assembled from %d drives - not enough to start it.\n",
- mddev, okcnt);
+ if (verbose >= -1) {
+ fprintf(stderr, Name ": %s assembled from %d drive%s", mddev, okcnt, okcnt==1?"":"s");
+ if (sparecnt)
+ fprintf(stderr, " and %d spare%s", sparecnt, sparecnt==1?"":"s");
+ if (!enough(info.array.level, info.array.raid_disks,
+ info.array.layout, 1, avail, okcnt))
+ fprintf(stderr, " - not enough to start the array.\n");
+ else if (!enough(info.array.level,
+ info.array.raid_disks,
+ info.array.layout, clean,
+ avail, okcnt))
+ fprintf(stderr, " - not enough to start the "
+ "array while not clean - consider "
+ "--force.\n");
+ else {
+ if (req_cnt == info.array.raid_disks)
+ fprintf(stderr, " - need all %d to start it", req_cnt);
+ else
+ fprintf(stderr, " - need %d of %d to start", req_cnt, info.array.raid_disks);
+ fprintf(stderr, " (use --run to insist).\n");
+ }
+ }
+ if (must_close) close(mdfd);
return 1;
} else {
- /* It maybe just a case of calling START_ARRAY, but it may not..
- * We need to pick a working device, read it's super block, and make
- * sure all the device numbers and the minor number are right.
- * Then we might need to re-write the super block.
- * THEN we call START_ARRAY
- * If the md_minor is wrong, wejust give up for now. The alternate is to
- * re-write ALL super blocks.
+ /* The "chosen_drive" is a good choice, and if necessary, the superblock has
+ * been updated to point to the current locations of devices.
+ * so we can just start the array
*/
- int chosen_drive = -1;
- int change = 0;
- int dev;
- for (i=0; i<first_super.nr_disks; i++) {
- if (!devices[i].uptodate)
- continue;
- if (chosen_drive == -1) {
- int fd;
- chosen_drive = i;
- if (open(devices[i].devname, O_RDONLY)>= 0) {
- fprintf(stderr, Name ": Cannot open %s: %s\n",
- devices[i].devname, strerror(errno));
- return 1;
- }
- if (load_super(fd, &super)) {
- close(fd);
- fprintf(stderr, Name ": RAID superblock has disappeared from %s\n",
- devices[i].devname);
- return 1;
- }
- close(fd);
- }
- if (devices[i].major != super.disks[i].major ||
- devices[i].minor != super.disks[i].minor) {
- change = 1;
- super.disks[i].major = devices[i].major;
- super.disks[i].minor = devices[i].minor;
- }
- }
- if (change) {
- int fd;
- super.sb_csum = calc_sb_csum(super);
- fd = open(devices[chosen_drive].devname, O_RDWR);
- if (fd < 0) {
- fprintf(stderr, Name ": Could open %s for write - cannot Assemble array.\n",
- devices[chosen_drive].devname);
- return 1;
- }
- if (store_super(fd, &super)) {
- close(fd);
- fprintf(stderr, Name ": Could not re-write superblock on %s\n",
- devices[chosen_drive].devname);
- return 1;
- }
- close(fd);
- }
- dev = MKDEV(devices[chosen_drive].major,
- devices[chosen_drive].minor);
+ unsigned long dev;
+ dev = makedev(devices[chosen_drive].i.disk.major,
+ devices[chosen_drive].i.disk.minor);
if (ioctl(mdfd, START_ARRAY, dev)) {
fprintf(stderr, Name ": Cannot start array: %s\n",
strerror(errno));
}
-
+
}
+ if (must_close) close(mdfd);
+ return 0;
}
projects / thirdparty / mdadm.git / blobdiff