mddev_ident_t ident, char *conffile,
mddev_dev_t devlist,
int readonly, int runstop,
+ char *update,
int verbose, int force)
{
/*
* 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
+ * Check superblock similarity if we have a superblock - discard if different
* Record events, devicenum, utime
* This should give us a list of devices for the array
* We should collect the most recent event and utime numbers
mdp_super_t first_super, super;
struct {
char *devname;
- int major, minor;
- int oldmajor, oldminor;
+ unsigned int major, minor;
+ unsigned int oldmajor, oldminor;
long long events;
time_t utime;
int uptodate;
+ int state;
int raid_disk;
- } devices[MD_SB_DISKS];
- int best[MD_SB_DISKS]; /* indexed by raid_disk */
- int devcnt = 0, okcnt, sparecnt;
- int 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;
int chosen_drive;
int change = 0;
int inargv = 0;
int start_partial_ok = force || devlist==NULL;
+ unsigned int num_devs;
+ mddev_dev_t tmpdev;
vers = md_get_version(mdfd);
if (vers <= 0) {
devlist = conf_get_devs(conffile);
else inargv = 1;
+ tmpdev = devlist; num_devs = 0;
+ while (tmpdev) {
+ num_devs++;
+ tmpdev = tmpdev->next;
+ }
+ best = malloc(num_devs * sizeof(*best));
+ devices = malloc(num_devs * sizeof(*devices));
+
first_super.md_magic = 0;
- for (i=0; i<MD_SB_DISKS; i++)
+ for (i=0; i<num_devs; i++)
best[i] = -1;
if (verbose)
devlist = devlist->next;
if (ident->devices &&
- !match_oneof(ident->devices, devname))
+ !match_oneof(ident->devices, devname)) {
+ if (inargv || verbose)
+ fprintf(stderr, Name ": %s is not one of %s\n", devname, ident->devices);
continue;
+ }
- dfd = open(devname, O_RDONLY, 0);
+ dfd = open(devname, O_RDONLY|O_EXCL, 0);
if (dfd < 0) {
if (inargv || verbose)
fprintf(stderr, Name ": cannot open device %s: %s\n",
devname);
continue;
}
- if (ident->super_minor >= 0 &&
+ if (ident->super_minor != UnSet &&
(!havesuper || ident->super_minor != super.md_minor)) {
if (inargv || verbose)
fprintf(stderr, Name ": %s has wrong super-minor.\n",
devname);
continue;
}
- if (ident->level != -10 &&
- (!havesuper|| ident->level != super.level)) {
+ if (ident->level != UnSet &&
+ (!havesuper|| ident->level != (int)super.level)) {
if (inargv || verbose)
fprintf(stderr, Name ": %s has wrong raid level.\n",
devname);
continue;
}
- if (ident->raid_disks != -1 &&
+ if (ident->raid_disks != UnSet &&
(!havesuper || ident->raid_disks!= super.raid_disks)) {
if (inargv || verbose)
fprintf(stderr, Name ": %s requires wrong number of drives.\n",
return 1;
}
+
+ /* this is needed until we get a more relaxed super block format */
if (devcnt >= MD_SB_DISKS) {
fprintf(stderr, Name ": ouch - too many devices appear to be in this array. Ignoring %s\n",
devname);
continue;
}
+
+ /* looks like a good enough match to update the super block if needed */
+ if (update) {
+ if (strcmp(update, "sparc2.2")==0 ) {
+ /* 2.2 sparc put the events in the wrong place
+ * So we copy the tail of the superblock
+ * up 4 bytes before continuing
+ */
+ __u32 *sb32 = (__u32*)&super;
+ memcpy(sb32+MD_SB_GENERIC_CONSTANT_WORDS+7,
+ sb32+MD_SB_GENERIC_CONSTANT_WORDS+7+1,
+ (MD_SB_WORDS - (MD_SB_GENERIC_CONSTANT_WORDS+7+1))*4);
+ fprintf (stderr, Name ": adjusting superblock of %s for 2.2/sparc compatability.\n",
+ devname);
+ }
+ if (strcmp(update, "super-minor") ==0) {
+ struct stat stb2;
+ fstat(mdfd, &stb2);
+ super.md_minor = minor(stb2.st_rdev);
+ if (verbose)
+ fprintf(stderr, Name ": updating superblock of %s with minor number %d\n",
+ devname, super.md_minor);
+ }
+ if (strcmp(update, "summaries") == 0) {
+ /* set nr_disks, active_disks, working_disks,
+ * failed_disks, spare_disks based on disks[]
+ * array in superblock.
+ * Also make sure extra slots aren't 'failed'
+ */
+ super.nr_disks = super.active_disks =
+ super.working_disks = super.failed_disks =
+ super.spare_disks = 0;
+ for (i=0; i < MD_SB_DISKS ; i++)
+ if (super.disks[i].major ||
+ super.disks[i].minor) {
+ int state = super.disks[i].state;
+ if (state & (1<<MD_DISK_REMOVED))
+ continue;
+ super.nr_disks++;
+ if (state & (1<<MD_DISK_ACTIVE))
+ super.active_disks++;
+ if (state & (1<<MD_DISK_FAULTY))
+ super.failed_disks++;
+ else
+ super.working_disks++;
+ if (state == 0)
+ super.spare_disks++;
+ } else if (i >= super.raid_disks && super.disks[i].number == 0)
+ super.disks[i].state = 0;
+ }
+ if (strcmp(update, "resync") == 0) {
+ /* make sure resync happens */
+ super.state &= ~(1<<MD_SB_CLEAN);
+ super.recovery_cp = 0;
+ }
+ super.sb_csum = calc_sb_csum(&super);
+ dfd = open(devname, O_RDWR|O_EXCL, 0);
+ if (dfd < 0)
+ fprintf(stderr, Name ": Cannot open %s for superblock update\n",
+ devname);
+ else if (store_super(dfd, &super))
+ fprintf(stderr, Name ": Could not re-write superblock on %s.\n",
+ devname);
+ if (dfd >= 0)
+ close(dfd);
+ }
+
if (verbose)
fprintf(stderr, Name ": %s is identified as a member of %s, slot %d.\n",
devname, mddev, super.this_disk.raid_disk);
devices[devcnt].devname = devname;
- devices[devcnt].major = MAJOR(stb.st_rdev);
- devices[devcnt].minor = MINOR(stb.st_rdev);
+ devices[devcnt].major = major(stb.st_rdev);
+ devices[devcnt].minor = minor(stb.st_rdev);
devices[devcnt].oldmajor = super.this_disk.major;
devices[devcnt].oldminor = super.this_disk.minor;
devices[devcnt].events = md_event(&super);
devices[devcnt].utime = super.utime;
devices[devcnt].raid_disk = super.this_disk.raid_disk;
devices[devcnt].uptodate = 0;
+ devices[devcnt].state = super.this_disk.state;
if (most_recent < devcnt) {
if (devices[devcnt].events
> devices[most_recent].events)
most_recent = devcnt;
}
- i = devices[devcnt].raid_disk;
- if (i>=0 && i < MD_SB_DISKS)
+ if ((int)super.level == -4)
+ /* with multipath, the raid_disk from the superblock is meaningless */
+ i = devcnt;
+ else
+ i = devices[devcnt].raid_disk;
+ 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] == -1
|| devices[best[i]].events < devices[devcnt].events)
best[i] = devcnt;
-
+ }
devcnt++;
}
*/
okcnt = 0;
sparecnt=0;
- for (i=0; i< MD_SB_DISKS;i++) {
+ for (i=0; i< bestcnt ;i++) {
int j = best[i];
int event_margin = !force;
if (j < 0) continue;
+ /* note: we ignore error flags in multipath arrays
+ * as they don't make sense
+ */
+ if ((int)first_super.level != -4)
+ if (!(devices[j].state & (1<<MD_DISK_SYNC))) {
+ if (!(devices[j].state & (1<<MD_DISK_FAULTY)))
+ sparecnt++;
+ continue;
+ }
if (devices[j].events+event_margin >=
devices[most_recent].events) {
devices[j].uptodate = 1;
*/
int fd;
chosen_drive = -1;
- for (i=0; i<first_super.raid_disks; i++) {
+ for (i=0; i<first_super.raid_disks && i < bestcnt; i++) {
int j = best[i];
if (j>=0 &&
!devices[j].uptodate &&
devices[chosen_drive].devname, devices[chosen_drive].raid_disk,
(int)(devices[chosen_drive].events),
(int)(devices[most_recent].events));
- fd = open(devices[chosen_drive].devname, O_RDWR);
+ fd = 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);
}
super.events_hi = (devices[most_recent].events>>32)&0xFFFFFFFF;
super.events_lo = (devices[most_recent].events)&0xFFFFFFFF;
+ if (super.level == 5 || super.level == 4) {
+ /* need to force clean */
+ super.state = (1<<MD_SB_CLEAN);
+ }
super.sb_csum = calc_sb_csum(&super);
/*DRYRUN*/ if (store_super(fd, &super)) {
close(fd);
* superblock.
*/
chosen_drive = -1;
- for (i=0; chosen_drive < 0 && i<MD_SB_DISKS; i++) {
+ for (i=0; chosen_drive < 0 && i<bestcnt; i++) {
int j = best[i];
int fd;
if (j<0)
if (!devices[j].uptodate)
continue;
chosen_drive = j;
- if ((fd=open(devices[j].devname, O_RDONLY))< 0) {
+ if ((fd=open(devices[j].devname, O_RDONLY|O_EXCL))< 0) {
fprintf(stderr, Name ": Cannot open %s: %s\n",
devices[j].devname, strerror(errno));
return 1;
close(fd);
}
- for (i=0; i<MD_SB_DISKS; i++) {
+ for (i=0; i<bestcnt; i++) {
int j = best[i];
- int desired_state;
+ unsigned int desired_state;
if (i < super.raid_disks)
desired_state = (1<<MD_DISK_ACTIVE) | (1<<MD_DISK_SYNC);
i, mddev);
}
}
+ if (force && (super.level == 4 || super.level == 5) &&
+ okcnt == super.raid_disks-1) {
+ super.state = (1<< MD_SB_CLEAN);
+ change |= 2;
+ }
if ((force && (change & 2))
|| (old_linux && (change & 1))) {
int fd;
super.sb_csum = calc_sb_csum(&super);
- fd = open(devices[chosen_drive].devname, O_RDWR);
+ fd = open(devices[chosen_drive].devname, O_RDWR|O_EXCL);
if (fd < 0) {
fprintf(stderr, Name ": Could open %s for write - cannot Assemble array.\n",
devices[chosen_drive].devname);
change = 0;
}
+ /* 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 = 0;
+ for (i=0; i<MD_SB_DISKS; i++)
+ if ((first_super.disks[i].state & (1<<MD_DISK_SYNC)) &&
+ (first_super.disks[i].state & (1<<MD_DISK_ACTIVE)) &&
+ !(first_super.disks[i].state & (1<<MD_DISK_FAULTY)))
+ req_cnt ++;
+
+
/* Almost ready to actually *do* something */
if (!old_linux) {
if (ioctl(mdfd, SET_ARRAY_INFO, NULL) != 0) {
return 1;
}
/* First, add the raid disks, but add the chosen one last */
- for (i=0; i<=MD_SB_DISKS; i++) {
+ for (i=0; i<= bestcnt; i++) {
int j;
- if (i < MD_SB_DISKS) {
+ if (i < bestcnt) {
j = best[i];
if (j == chosen_drive)
continue;
} else
j = chosen_drive;
- if (j >= 0 && devices[j].uptodate) {
+ if (j >= 0 /* && devices[j].uptodate */) {
mdu_disk_info_t disk;
memset(&disk, 0, sizeof(disk));
disk.major = devices[j].major;
devices[j].devname,
mddev,
strerror(errno));
- if (i < first_super.raid_disks)
+ if (i < first_super.raid_disks || i == bestcnt)
okcnt--;
else
sparecnt--;
if (runstop == 1 ||
(runstop == 0 &&
- ( first_super.raid_disks == okcnt
- || (start_partial_ok && enough(first_super.level, first_super.raid_disks, okcnt)))
- )) {
+ ( enough(first_super.level, first_super.raid_disks, okcnt) &&
+ (okcnt >= req_cnt || start_partial_ok)
+ ))) {
if (ioctl(mdfd, RUN_ARRAY, NULL)==0) {
fprintf(stderr, Name ": %s has been started with %d drive%s",
mddev, okcnt, okcnt==1?"":"s");
+ if (okcnt < first_super.raid_disks)
+ fprintf(stderr, " (out of %d)", first_super.raid_disks);
if (sparecnt)
fprintf(stderr, " and %d spare%s", sparecnt, sparecnt==1?"":"s");
fprintf(stderr, ".\n");
mddev, okcnt, okcnt==1?"":"s");
return 0;
}
- fprintf(stderr, Name ": %s assembled from %d drive%s - not enough to start it (use --run to insist).\n",
- mddev, okcnt, okcnt==1?"":"s");
+ 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(first_super.level, first_super.raid_disks, okcnt))
+ fprintf(stderr, " - not enough to start the array.\n");
+ else {
+ if (req_cnt == first_super.raid_disks)
+ fprintf(stderr, " - need all %d to start it", req_cnt);
+ else
+ fprintf(stderr, " - need %d of %d to start", req_cnt, first_super.raid_disks);
+ fprintf(stderr, " (use --run to insist).\n");
+ }
return 1;
} else {
/* The "chosen_drive" is a good choice, and if necessary, the superblock has
* so we can just start the array
*/
unsigned long dev;
- dev = MKDEV(devices[chosen_drive].major,
+ dev = makedev(devices[chosen_drive].major,
devices[chosen_drive].minor);
if (ioctl(mdfd, START_ARRAY, dev)) {
fprintf(stderr, Name ": Cannot start array: %s\n",
projects / thirdparty / mdadm.git / blobdiff