/* from mdopen.c */
int open_mddev(char *dev, int report_errors/*unused*/)
{
int mdfd = open(dev, O_RDWR);
if (mdfd < 0)
fprintf(stderr, Name ": error opening %s: %s\n",
dev, strerror(errno));
else if (md_get_version(mdfd) <= 0) {
fprintf(stderr, Name ": %s does not appear to be an md device\n",
dev);
close(mdfd);
mdfd = -1;
}
return mdfd;
}
int Build(char *mddev, struct mddev_dev *devlist,
struct shape *s, struct context *c)
{
/* Build a linear or raid0 arrays without superblocks
* We cannot really do any checks, we just do it.
* For md_version < 0.90.0, we call REGISTER_DEV
* with the device numbers, and then
* START_MD giving the "geometry"
* geometry is 0xpp00cc
* where pp is personality: 1==linear, 2=raid0
* cc = chunk size factor: 0==4k, 1==8k etc.
*
* For md_version >= 0.90.0 we call
* SET_ARRAY_INFO, ADD_NEW_DISK, RUN_ARRAY
*
*/
int i;
int vers;
struct stat stb;
int subdevs = 0, missing_disks = 0;
struct mddev_dev *dv;
int bitmap_fd;
unsigned long long bitmapsize;
int mdfd;
char chosen_name[1024];
int uuid[4] = {0,0,0,0};
struct map_ent *map = NULL;
/* scan all devices, make sure they really are block devices */
for (dv = devlist; dv; dv=dv->next) {
subdevs++;
if (strcmp("missing", dv->devname) == 0) {
missing_disks++;
continue;
}
if (stat(dv->devname, &stb)) {
pr_err("Cannot find %s: %s\n",
dv->devname, strerror(errno));
return 1;
}
if ((stb.st_mode & S_IFMT) != S_IFBLK) {
pr_err("%s is not a block device.\n",
dv->devname);
return 1;
}
}
if (s->raiddisks != subdevs) {
pr_err("requested %d devices in array but listed %d\n",
s->raiddisks, subdevs);
return 1;
}
if (s->layout == UnSet)
switch(s->level) {
default: /* no layout */
s->layout = 0;
break;
case 10:
s->layout = 0x102; /* near=2, far=1 */
if (c->verbose > 0)
pr_err("layout defaults to n1\n");
break;
case 5:
case 6:
s->layout = map_name(r5layout, "default");
if (c->verbose > 0)
pr_err("layout defaults to %s\n", map_num(r5layout, s->layout));
break;
case LEVEL_FAULTY:
s->layout = map_name(faultylayout, "default");
if (c->verbose > 0)
pr_err("layout defaults to %s\n", map_num(faultylayout, s->layout));
break;
}
/* We need to create the device. It can have no name. */
map_lock(&map);
mdfd = create_mddev(mddev, NULL, c->autof, LOCAL,
chosen_name);
if (mdfd < 0) {
map_unlock(&map);
return 1;
}
mddev = chosen_name;
map_update(&map, fd2devnm(mdfd), "none", uuid, chosen_name);
map_unlock(&map);
vers = md_get_version(mdfd);
/* looks Ok, go for it */
if (vers >= 9000) {
mdu_array_info_t array;
array.level = s->level;
if (s->size == MAX_SIZE)
s->size = 0;
array.size = s->size;
array.nr_disks = s->raiddisks;
array.raid_disks = s->raiddisks;
array.md_minor = 0;
if (fstat(mdfd, &stb)==0)
array.md_minor = minor(stb.st_rdev);
array.not_persistent = 1;
array.state = 0; /* not clean, but no errors */
if (s->assume_clean)
array.state |= 1;
array.active_disks = s->raiddisks - missing_disks;
array.working_disks = s->raiddisks - missing_disks;
array.spare_disks = 0;
array.failed_disks = missing_disks;
if (s->chunk == 0 && (s->level==0 || s->level==LEVEL_LINEAR))
s->chunk = 64;
array.chunk_size = s->chunk*1024;
array.layout = s->layout;
if (ioctl(mdfd, SET_ARRAY_INFO, &array)) {
pr_err("SET_ARRAY_INFO failed for %s: %s\n",
mddev, strerror(errno));
goto abort;
}
} else if (s->bitmap_file) {
pr_err("bitmaps not supported with this kernel\n");
goto abort;
}
if (s->bitmap_file && strcmp(s->bitmap_file, "none") == 0)
s->bitmap_file = NULL;
if (s->bitmap_file && s->level <= 0) {
pr_err("bitmaps not meaningful with level %s\n",
map_num(pers, s->level)?:"given");
goto abort;
}
static int mdmon(char *devnm, int must_fork, int takeover)
{
int mdfd;
struct mdinfo *mdi, *di;
struct supertype *container;
sigset_t set;
struct sigaction act;
int pfd[2];
int status;
int ignore;
pid_t victim = -1;
int victim_sock = -1;
dprintf("starting mdmon for %s\n", devnm);
mdfd = open_dev(devnm);
if (mdfd < 0) {
pr_err("%s: %s\n", devnm, strerror(errno));
return 1;
}
if (md_get_version(mdfd) < 0) {
pr_err("%s: Not an md device\n", devnm);
return 1;
}
/* Fork, and have the child tell us when they are ready */
if (must_fork) {
if (pipe(pfd) != 0) {
pr_err("failed to create pipe\n");
return 1;
}
switch(fork()) {
case -1:
pr_err("failed to fork: %s\n", strerror(errno));
return 1;
case 0: /* child */
close(pfd[0]);
break;
default: /* parent */
close(pfd[1]);
if (read(pfd[0], &status, sizeof(status)) != sizeof(status)) {
wait(&status);
status = WEXITSTATUS(status);
}
close(pfd[0]);
return status;
}
} else
pfd[0] = pfd[1] = -1;
container = xcalloc(1, sizeof(*container));
strcpy(container->devnm, devnm);
container->arrays = NULL;
container->sock = -1;
mdi = sysfs_read(mdfd, container->devnm, GET_VERSION|GET_LEVEL|GET_DEVS);
if (!mdi) {
pr_err("failed to load sysfs info for %s\n", container->devnm);
exit(3);
}
if (mdi->array.level != UnSet) {
pr_err("%s is not a container - cannot monitor\n", devnm);
exit(3);
}
if (mdi->array.major_version != -1 ||
mdi->array.minor_version != -2) {
pr_err("%s does not use external metadata - cannot monitor\n",
devnm);
exit(3);
}
container->ss = version_to_superswitch(mdi->text_version);
if (container->ss == NULL) {
pr_err("%s uses unsupported metadata: %s\n",
devnm, mdi->text_version);
exit(3);
}
container->devs = NULL;
for (di = mdi->devs; di; di = di->next) {
struct mdinfo *cd = xmalloc(sizeof(*cd));
*cd = *di;
cd->next = container->devs;
container->devs = cd;
}
sysfs_free(mdi);
/* SIGUSR is sent between parent and child. So both block it
* and enable it only with pselect.
*/
sigemptyset(&set);
sigaddset(&set, SIGUSR1);
sigaddset(&set, SIGTERM);
sigprocmask(SIG_BLOCK, &set, NULL);
act.sa_handler = wake_me;
act.sa_flags = 0;
sigaction(SIGUSR1, &act, NULL);
act.sa_handler = term;
sigaction(SIGTERM, &act, NULL);
act.sa_handler = SIG_IGN;
sigaction(SIGPIPE, &act, NULL);
victim = mdmon_pid(container->devnm);
if (victim >= 0)
victim_sock = connect_monitor(container->devnm);
ignore = chdir("/");
if (!takeover && victim > 0 && victim_sock >= 0) {
if (fping_monitor(victim_sock) == 0) {
pr_err("%s already managed\n", container->devnm);
exit(3);
}
close(victim_sock);
victim_sock = -1;
}
if (container->ss->load_container(container, mdfd, devnm)) {
pr_err("Cannot load metadata for %s\n", devnm);
exit(3);
}
close(mdfd);
/* Ok, this is close enough. We can say goodbye to our parent now.
*/
if (victim > 0)
remove_pidfile(devnm);
if (make_pidfile(devnm) < 0) {
exit(3);
}
container->sock = make_control_sock(devnm);
status = 0;
if (pfd[1] >= 0) {
if (write(pfd[1], &status, sizeof(status)) < 0)
pr_err("failed to notify our parent: %d\n",
getppid());
close(pfd[1]);
}
mlockall(MCL_CURRENT | MCL_FUTURE);
if (clone_monitor(container) < 0) {
pr_err("failed to start monitor process: %s\n",
strerror(errno));
exit(2);
}
if (victim > 0) {
try_kill_monitor(victim, container->devnm, victim_sock);
if (victim_sock >= 0)
close(victim_sock);
}
setsid();
close(0);
open("/dev/null", O_RDWR);
close(1);
ignore = dup(0);
#ifndef DEBUG
close(2);
ignore = dup(0);
#endif
/* This silliness is to stop the compiler complaining
* that we ignore 'ignore'
*/
if (ignore)
ignore++;
do_manager(container);
exit(0);
}
int Assemble(struct supertype *st, char *mddev, int mdfd,
mddev_ident_t ident, char *conffile,
mddev_dev_t devlist,
int readonly, int runstop,
char *update,
int verbose, int force)
{
/*
* 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
* the array. Finally RUN_ARRAY might
* be submitted to start the array.
*
* 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 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)
*
* If mddev is set, we try to determine state of md.
* check version - must be at least 0.90.0
* check kernel version. must be at least 2.4.
* If not, we can possibly fall back on START_ARRAY
* Try to GET_ARRAY_INFO.
* If possible, give up
* If not, try to STOP_ARRAY just to make sure
*
* 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
*
* 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 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 number
*
* Count disks with recent enough event count
* While force && !enough disks
* Choose newest rejected disks, update event count
* mark clean and rewrite superblock
* If recent kernel:
* SET_ARRAY_INFO
* foreach device with recent events : ADD_NEW_DISK
* if runstop == 1 || "enough" disks and runstop==0 -> RUN_ARRAY
* If old kernel:
* Check the device numbers in superblock are right
* update superblock if any changes
* START_ARRAY
*
*/
int old_linux = 0;
int vers;
void *first_super = NULL, *super = NULL;
struct {
char *devname;
unsigned int major, minor;
unsigned int oldmajor, oldminor;
long long events;
int uptodate;
int state;
int raid_disk;
int disk_nr;
} *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;
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 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 (!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);
return 1;
}
if (devlist == NULL)
devlist = conf_get_devs(conffile);
else inargv = 1;
tmpdev = devlist; num_devs = 0;
while (tmpdev) {
num_devs++;
tmpdev = tmpdev->next;
}
devices = malloc(num_devs * sizeof(*devices));
if (!st && ident->st) st = ident->st;
if (verbose>0)
fprintf(stderr, Name ": looking for devices for %s\n",
mddev);
while ( devlist) {
char *devname;
int dfd;
struct stat stb;
struct supertype *tst = st;
devname = devlist->devname;
devlist = devlist->next;
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;
}
if (super) {
free(super);
super = NULL;
}
dfd = dev_open(devname, O_RDONLY|O_EXCL);
if (dfd < 0) {
if ((inargv && verbose >= 0) || verbose > 0)
fprintf(stderr, Name ": cannot open device %s: %s\n",
devname, strerror(errno));
} else if (fstat(dfd, &stb)< 0) {
/* Impossible! */
fprintf(stderr, Name ": fstat failed for %s: %s\n",
devname, strerror(errno));
} else if ((stb.st_mode & S_IFMT) != S_IFBLK) {
fprintf(stderr, Name ": %s is not a block device.\n",
devname);
} else if (!tst && (tst = guess_super(dfd)) == NULL) {
if ((inargv && verbose >= 0) || verbose > 0)
fprintf(stderr, Name ": no recogniseable superblock\n");
} 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);
}
if (dfd >= 0) close(dfd);
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 ((inargv && verbose >= 0) || verbose > 0)
fprintf(stderr, Name ": %s has wrong name.\n",
devname);
continue;
}
if (ident->super_minor != UnSet &&
(!super || ident->super_minor != info.array.md_minor)) {
if ((inargv && verbose >= 0) || verbose > 0)
fprintf(stderr, Name ": %s has wrong super-minor.\n",
devname);
continue;
}
if (ident->level != UnSet &&
(!super|| ident->level != info.array.level)) {
if ((inargv && verbose >= 0) || verbose > 0)
fprintf(stderr, Name ": %s has wrong raid level.\n",
devname);
continue;
}
if (ident->raid_disks != UnSet &&
(!super || ident->raid_disks!= info.array.raid_disks)) {
if ((inargv && verbose >= 0) || verbose > 0)
fprintf(stderr, Name ": %s requires wrong number of drives.\n",
devname);
continue;
}
/* If we are this far, then we are commited to this device.
* If the super_block doesn't exist, or doesn't match others,
* then we cannot continue
*/
if (!super) {
fprintf(stderr, Name ": %s has no superblock - assembly aborted\n",
devname);
free(first_super);
return 1;
}
st = tst; /* commit to this format, if haven't already */
if (st->ss->compare_super(&first_super, super)) {
fprintf(stderr, Name ": superblock on %s doesn't match others - assembly aborted\n",
devname);
free(super);
free(first_super);
return 1;
}
/* looks like a good enough match to update the super block if needed */
if (update) {
/* prepare useful information in info structures */
struct stat stb2;
fstat(mdfd, &stb2);
info.array.md_minor = minor(stb2.st_rdev);
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);
ident->uuid_set = 1;
}
memcpy(info.uuid, ident->uuid, 16);
st->ss->update_super(&info, super, update, devname, verbose);
dfd = dev_open(devname, O_RDWR|O_EXCL);
if (dfd < 0)
fprintf(stderr, Name ": Cannot open %s for superblock update\n",
devname);
else if (st->ss->store_super(st, dfd, super))
fprintf(stderr, Name ": Could not re-write superblock on %s.\n",
devname);
if (dfd >= 0)
close(dfd);
}
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].oldmajor = info.disk.major;
devices[devcnt].oldminor = info.disk.minor;
devices[devcnt].events = info.events;
devices[devcnt].raid_disk = info.disk.raid_disk;
devices[devcnt].disk_nr = info.disk.number;
devices[devcnt].uptodate = 0;
devices[devcnt].state = info.disk.state;
if (most_recent < devcnt) {
if (devices[devcnt].events
> devices[most_recent].events)
most_recent = devcnt;
}
if (info.array.level == -4)
/* with multipath, the raid_disk from the superblock is meaningless */
i = devcnt;
else
i = devices[devcnt].raid_disk;
if (i+1 == 0) {
if (nextspare < info.array.raid_disks)
nextspare = info.array.raid_disks;
i = nextspare++;
}
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++;
}
if (super)
free(super);
super = NULL;
if (update && strcmp(update, "byteorder")==0)
st->minor_version = 90;
if (devcnt == 0) {
fprintf(stderr, Name ": no devices found for %s\n",
mddev);
free(first_super);
return 1;
}
st->ss->getinfo_super(&info, &ident2, first_super);
/* 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;
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;
/* note: we ignore error flags in multipath arrays
* as they don't make sense
*/
if (info.array.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;
if (i < info.array.raid_disks) {
okcnt++;
avail[i]=1;
} else
sparecnt++;
}
}
while (force && !enough(info.array.level, info.array.raid_disks,
info.array.layout,
avail, okcnt)) {
/* Choose the newest best drive which is
* not up-to-date, update the superblock
* and add it.
*/
int fd;
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].events > 0 &&
(chosen_drive < 0 ||
devices[j].events > devices[chosen_drive].events))
chosen_drive = j;
}
if (chosen_drive < 0)
break;
if (verbose >= 0)
fprintf(stderr, Name ": forcing event count in %s(%d) from %d upto %d\n",
devices[chosen_drive].devname, devices[chosen_drive].raid_disk,
(int)(devices[chosen_drive].events),
(int)(devices[most_recent].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].events = 0;
continue;
}
if (st->ss->load_super(st,fd, &super, NULL)) {
close(fd);
fprintf(stderr, Name ": RAID superblock disappeared from %s - not updating.\n",
devices[chosen_drive].devname);
devices[chosen_drive].events = 0;
continue;
}
info.events = devices[most_recent].events;
st->ss->update_super(&info, super, "force", devices[chosen_drive].devname, verbose);
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);
devices[chosen_drive].events = 0;
free(super);
continue;
}
close(fd);
devices[chosen_drive].events = devices[most_recent].events;
devices[chosen_drive].uptodate = 1;
avail[chosen_drive] = 1;
okcnt++;
free(super);
}
/* 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;
super = NULL;
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));
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);
return 1;
}
close(fd);
}
if (super == NULL) {
fprintf(stderr, Name ": No suitable drives found for %s\n", mddev);
return 1;
}
st->ss->getinfo_super(&info, &ident2, super);
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;
info.disk.number = devices[j].disk_nr;
info.disk.raid_disk = i;
info.disk.state = desired_state;
if (devices[j].uptodate &&
st->ss->update_super(&info, super, "assemble", NULL, verbose)) {
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 (!devices[j].uptodate &&
!(super.disks[i].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 && okcnt == info.array.raid_disks-1) {
/* FIXME check event count */
change += st->ss->update_super(&info, super, "force",
devices[chosen_drive].devname, verbose);
}
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",
devices[chosen_drive].devname);
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);
return 1;
}
close(fd);
}
/* 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) {
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));
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");
return 1;
}
}
/* 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 */) {
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",
devices[j].devname,
mddev,
strerror(errno));
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].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(info.array.level, info.array.raid_disks, info.array.layout, avail, okcnt) &&
(okcnt >= req_cnt || start_partial_ok)
))) {
if (ioctl(mdfd, RUN_ARRAY, NULL)==0) {
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");
}
return 0;
}
fprintf(stderr, Name ": failed to RUN_ARRAY %s: %s\n",
mddev, strerror(errno));
return 1;
}
if (runstop == -1) {
fprintf(stderr, Name ": %s assembled from %d drive%s, but not started.\n",
mddev, okcnt, okcnt==1?"":"s");
return 0;
}
if (verbose >= 0) {
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, avail, okcnt))
fprintf(stderr, " - not enough to start the array.\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");
}
}
return 1;
} else {
/* 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
*/
unsigned long dev;
dev = makedev(devices[chosen_drive].major,
devices[chosen_drive].minor);
if (ioctl(mdfd, START_ARRAY, dev)) {
fprintf(stderr, Name ": Cannot start array: %s\n",
strerror(errno));
}
}
return 0;
}
int Query(char *dev)
{
/* Give a brief description of the device,
* whether it is an md device and whether it has
* a superblock
*/
int fd = open(dev, O_RDONLY);
int vers;
int ioctlerr;
int superror;
struct mdinfo info;
mdu_array_info_t array;
struct supertype *st = NULL;
unsigned long long larray_size;
struct stat stb;
char *mddev;
mdu_disk_info_t disc;
char *activity;
if (fd < 0){
pr_err("cannot open %s: %s\n",
dev, strerror(errno));
return 1;
}
vers = md_get_version(fd);
if (ioctl(fd, GET_ARRAY_INFO, &array)<0)
ioctlerr = errno;
else ioctlerr = 0;
fstat(fd, &stb);
if (vers>=9000 && !ioctlerr) {
if (!get_dev_size(fd, NULL, &larray_size))
larray_size = 0;
}
if (vers < 0)
printf("%s: is not an md array\n", dev);
else if (vers < 9000)
printf("%s: is an md device, but kernel cannot provide details\n", dev);
else if (ioctlerr == ENODEV)
printf("%s: is an md device which is not active\n", dev);
else if (ioctlerr)
printf("%s: is an md device, but gives \"%s\" when queried\n",
dev, strerror(ioctlerr));
else {
printf("%s: %s %s %d devices, %d spare%s. Use mdadm --detail for more detail.\n",
dev,
human_size_brief(larray_size),
map_num(pers, array.level),
array.raid_disks,
array.spare_disks, array.spare_disks==1?"":"s");
}
st = guess_super(fd);
if (st)
superror = st->ss->load_super(st, fd, dev);
else
superror = -1;
close(fd);
if (superror == 0) {
/* array might be active... */
st->ss->getinfo_super(st, &info, NULL);
if (st->ss == &super0) {
mddev = get_md_name(info.array.md_minor);
disc.number = info.disk.number;
activity = "undetected";
if (mddev && (fd = open(mddev, O_RDONLY))>=0) {
if (md_get_version(fd) >= 9000 &&
ioctl(fd, GET_ARRAY_INFO, &array)>= 0) {
if (ioctl(fd, GET_DISK_INFO, &disc) >= 0 &&
makedev((unsigned)disc.major,(unsigned)disc.minor) == stb.st_rdev)
activity = "active";
else
activity = "mismatch";
}
close(fd);
}
} else {
activity = "unknown";
mddev = "array";
}
printf("%s: device %d in %d device %s %s %s. Use mdadm --examine for more detail.\n",
dev,
info.disk.number, info.array.raid_disks,
activity,
map_num(pers, info.array.level),
mddev);
if (st->ss == &super0)
put_md_name(mddev);
}
return 0;
}
int Manage_subdevs(char *devname, int fd,
struct mddev_dev *devlist, int verbose, int test,
char *update)
{
/* do something to each dev.
* devmode can be
* 'a' - add the device
* try HOT_ADD_DISK
* If that fails EINVAL, try ADD_NEW_DISK
* 'r' - remove the device HOT_REMOVE_DISK
* device can be 'faulty' or 'detached' in which case all
* matching devices are removed.
* 'f' - set the device faulty SET_DISK_FAULTY
* device can be 'detached' in which case any device that
* is inaccessible will be marked faulty.
* For 'f' and 'r', the device can also be a kernel-internal
* name such as 'sdb'.
*/
struct mddev_dev *add_devlist = NULL;
mdu_array_info_t array;
mdu_disk_info_t disc;
unsigned long long array_size;
struct mddev_dev *dv, *next = NULL;
struct stat stb;
int j, jnext = 0;
int tfd = -1;
struct supertype *st, *tst;
char *subarray = NULL;
int duuid[4];
int ouuid[4];
int lfd = -1;
int sysfd = -1;
int count = 0; /* number of actions taken */
if (ioctl(fd, GET_ARRAY_INFO, &array)) {
fprintf(stderr, Name ": cannot get array info for %s\n",
devname);
return 1;
}
/* array.size is only 32 bit and may be truncated.
* So read from sysfs if possible, and record number of sectors
*/
array_size = get_component_size(fd);
if (array_size <= 0)
array_size = array.size * 2;
tst = super_by_fd(fd, &subarray);
if (!tst) {
fprintf(stderr, Name ": unsupport array - version %d.%d\n",
array.major_version, array.minor_version);
return 1;
}
stb.st_rdev = 0;
for (dv = devlist, j=0 ; dv; dv = next, j = jnext) {
unsigned long long ldsize;
char dvname[20];
char *dnprintable = dv->devname;
char *add_dev = dv->devname;
int err;
int re_add_failed = 0;
next = dv->next;
jnext = 0;
if (strcmp(dv->devname, "failed")==0 ||
strcmp(dv->devname, "faulty")==0) {
int remaining_disks = array.nr_disks;
if (dv->disposition != 'r') {
fprintf(stderr, Name ": %s only meaningful "
"with -r, not -%c\n",
dv->devname, dv->disposition);
return 1;
}
for (; j < 1024 && remaining_disks > 0; j++) {
unsigned dev;
disc.number = j;
if (ioctl(fd, GET_DISK_INFO, &disc))
continue;
if (disc.major == 0 && disc.minor == 0)
continue;
remaining_disks --;
if ((disc.state & 1) == 0) /* faulty */
continue;
dev = makedev(disc.major, disc.minor);
if (stb.st_rdev == dev)
/* already did that one */
continue;
stb.st_rdev = dev;
next = dv;
/* same slot again next time - things might
* have reshuffled */
jnext = j;
sprintf(dvname,"%d:%d", disc.major, disc.minor);
dnprintable = dvname;
break;
}
if (next != dv)
continue;
} else if (strcmp(dv->devname, "detached") == 0) {
int remaining_disks = array.nr_disks;
if (dv->disposition != 'r' && dv->disposition != 'f') {
fprintf(stderr, Name ": %s only meaningful "
"with -r of -f, not -%c\n",
dv->devname, dv->disposition);
return 1;
}
for (; j < 1024 && remaining_disks > 0; j++) {
int sfd;
unsigned dev;
disc.number = j;
if (ioctl(fd, GET_DISK_INFO, &disc))
continue;
if (disc.major == 0 && disc.minor == 0)
continue;
remaining_disks --;
sprintf(dvname,"%d:%d", disc.major, disc.minor);
sfd = dev_open(dvname, O_RDONLY);
if (sfd >= 0) {
close(sfd);
continue;
}
if (dv->disposition == 'f' &&
(disc.state & 1) == 1) /* already faulty */
continue;
if (errno != ENXIO)
continue;
dev = makedev(disc.major, disc.minor);
if (stb.st_rdev == dev)
/* already did that one */
continue;
stb.st_rdev = dev;
next = dv;
/* same slot again next time - things might
* have reshuffled */
jnext = j;
dnprintable = dvname;
break;
}
if (next != dv)
continue;
} else if (strcmp(dv->devname, "missing") == 0) {
if (dv->disposition != 'a' || dv->re_add == 0) {
fprintf(stderr, Name ": 'missing' only meaningful "
"with --re-add\n");
return 1;
}
if (add_devlist == NULL)
add_devlist = conf_get_devs();
if (add_devlist == NULL) {
fprintf(stderr, Name ": no devices to scan for missing members.");
continue;
}
add_dev = add_devlist->devname;
add_devlist = add_devlist->next;
if (add_devlist != NULL)
next = dv;
if (stat(add_dev, &stb) < 0)
continue;
} else if (strchr(dv->devname, '/') == NULL &&
strchr(dv->devname, ':') == NULL &&
strlen(dv->devname) < 50) {
/* Assume this is a kernel-internal name like 'sda1' */
int found = 0;
char dname[55];
if (dv->disposition != 'r' && dv->disposition != 'f') {
fprintf(stderr, Name ": %s only meaningful "
"with -r or -f, not -%c\n",
dv->devname, dv->disposition);
return 1;
}
sprintf(dname, "dev-%s", dv->devname);
sysfd = sysfs_open(fd2devnum(fd), dname, "block/dev");
if (sysfd >= 0) {
char dn[20];
int mj,mn;
if (sysfs_fd_get_str(sysfd, dn, 20) > 0 &&
sscanf(dn, "%d:%d", &mj,&mn) == 2) {
stb.st_rdev = makedev(mj,mn);
found = 1;
}
close(sysfd);
sysfd = -1;
}
if (!found) {
sysfd = sysfs_open(fd2devnum(fd), dname, "state");
if (sysfd < 0) {
fprintf(stderr, Name ": %s does not appear "
"to be a component of %s\n",
dv->devname, devname);
return 1;
}
}
} else {
j = 0;
tfd = dev_open(dv->devname, O_RDONLY);
if (tfd < 0 && dv->disposition == 'r' &&
lstat(dv->devname, &stb) == 0)
/* Be happy, the lstat worked, that is
* enough for --remove
*/
;
else {
if (tfd < 0 || fstat(tfd, &stb) != 0) {
fprintf(stderr, Name ": cannot find %s: %s\n",
dv->devname, strerror(errno));
if (tfd >= 0)
close(tfd);
return 1;
}
close(tfd);
tfd = -1;
}
if ((stb.st_mode & S_IFMT) != S_IFBLK) {
fprintf(stderr, Name ": %s is not a "
"block device.\n",
dv->devname);
return 1;
}
}
switch(dv->disposition){
default:
fprintf(stderr, Name ": internal error - devmode[%s]=%d\n",
dv->devname, dv->disposition);
return 1;
case 'a':
/* add the device */
if (subarray) {
fprintf(stderr, Name ": Cannot add disks to a"
" \'member\' array, perform this"
" operation on the parent container\n");
return 1;
}
/* Make sure it isn't in use (in 2.6 or later) */
tfd = dev_open(add_dev, O_RDONLY|O_EXCL|O_DIRECT);
if (tfd < 0 && add_dev != dv->devname)
continue;
if (tfd < 0) {
fprintf(stderr, Name ": Cannot open %s: %s\n",
dv->devname, strerror(errno));
return 1;
}
st = dup_super(tst);
if (array.not_persistent==0)
st->ss->load_super(st, tfd, NULL);
if (add_dev == dv->devname) {
if (!get_dev_size(tfd, dv->devname, &ldsize)) {
close(tfd);
return 1;
}
} else if (!get_dev_size(tfd, NULL, &ldsize)) {
close(tfd);
tfd = -1;
continue;
}
if (!tst->ss->external &&
array.major_version == 0 &&
md_get_version(fd)%100 < 2) {
close(tfd);
tfd = -1;
if (ioctl(fd, HOT_ADD_DISK,
(unsigned long)stb.st_rdev)==0) {
if (verbose >= 0)
fprintf(stderr, Name ": hot added %s\n",
add_dev);
continue;
}
fprintf(stderr, Name ": hot add failed for %s: %s\n",
add_dev, strerror(errno));
return 1;
}
if (array.not_persistent == 0 || tst->ss->external) {
/* need to find a sample superblock to copy, and
* a spare slot to use.
* For 'external' array (well, container based),
* We can just load the metadata for the array.
*/
if (tst->sb)
/* already loaded */;
else if (tst->ss->external) {
tst->ss->load_container(tst, fd, NULL);
} else for (j = 0; j < tst->max_devs; j++) {
char *dev;
int dfd;
disc.number = j;
if (ioctl(fd, GET_DISK_INFO, &disc))
continue;
if (disc.major==0 && disc.minor==0)
continue;
if ((disc.state & 4)==0) continue; /* sync */
/* Looks like a good device to try */
dev = map_dev(disc.major, disc.minor, 1);
if (!dev) continue;
dfd = dev_open(dev, O_RDONLY);
if (dfd < 0) continue;
if (tst->ss->load_super(tst, dfd,
NULL)) {
close(dfd);
continue;
}
close(dfd);
break;
}
/* FIXME this is a bad test to be using */
if (!tst->sb) {
close(tfd);
fprintf(stderr, Name ": cannot load array metadata from %s\n", devname);
return 1;
}
/* Make sure device is large enough */
if (tst->ss->avail_size(tst, ldsize/512) <
array_size) {
close(tfd);
tfd = -1;
if (add_dev != dv->devname)
continue;
fprintf(stderr, Name ": %s not large enough to join array\n",
dv->devname);
return 1;
}
/* Possibly this device was recently part of the array
* and was temporarily removed, and is now being re-added.
* If so, we can simply re-add it.
*/
tst->ss->uuid_from_super(tst, duuid);
if (st->sb) {
struct mdinfo mdi;
st->ss->getinfo_super(st, &mdi, NULL);
st->ss->uuid_from_super(st, ouuid);
if ((mdi.disk.state & (1<<MD_DISK_ACTIVE)) &&
!(mdi.disk.state & (1<<MD_DISK_FAULTY)) &&
memcmp(duuid, ouuid, sizeof(ouuid))==0) {
/* look like it is worth a try. Need to
* make sure kernel will accept it though.
*/
/* re-add doesn't work for version-1 superblocks
* before 2.6.18 :-(
*/
if (array.major_version == 1 &&
get_linux_version() <= 2006018)
goto skip_re_add;
disc.number = mdi.disk.number;
if (ioctl(fd, GET_DISK_INFO, &disc) != 0
|| disc.major != 0 || disc.minor != 0
|| !enough_fd(fd))
goto skip_re_add;
disc.major = major(stb.st_rdev);
disc.minor = minor(stb.st_rdev);
disc.number = mdi.disk.number;
disc.raid_disk = mdi.disk.raid_disk;
disc.state = mdi.disk.state;
if (dv->writemostly == 1)
disc.state |= 1 << MD_DISK_WRITEMOSTLY;
if (dv->writemostly == 2)
disc.state &= ~(1 << MD_DISK_WRITEMOSTLY);
remove_partitions(tfd);
close(tfd);
tfd = -1;
if (update) {
int rv = -1;
tfd = dev_open(dv->devname, O_RDWR);
if (tfd >= 0)
rv = st->ss->update_super(
st, NULL, update,
devname, verbose, 0, NULL);
if (rv == 0)
rv = st->ss->store_super(st, tfd);
close(tfd);
tfd = -1;
if (rv != 0) {
fprintf(stderr, Name ": failed to update"
" superblock during re-add\n");
return 1;
}
}
/* don't even try if disk is marked as faulty */
errno = 0;
if (ioctl(fd, ADD_NEW_DISK, &disc) == 0) {
if (verbose >= 0)
fprintf(stderr, Name ": re-added %s\n", add_dev);
count++;
continue;
}
if (errno == ENOMEM || errno == EROFS) {
fprintf(stderr, Name ": add new device failed for %s: %s\n",
add_dev, strerror(errno));
if (add_dev != dv->devname)
continue;
return 1;
}
skip_re_add:
re_add_failed = 1;
}
st->ss->free_super(st);
}
if (add_dev != dv->devname) {
if (verbose > 0)
fprintf(stderr, Name
": --re-add for %s to %s is not possible\n",
add_dev, devname);
if (tfd >= 0) {
close(tfd);
tfd = -1;
}
continue;
}
if (dv->re_add) {
if (tfd >= 0)
close(tfd);
fprintf(stderr, Name
": --re-add for %s to %s is not possible\n",
dv->devname, devname);
return 1;
}
if (re_add_failed) {
fprintf(stderr, Name ": %s reports being an active member for %s, but a --re-add fails.\n",
dv->devname, devname);
fprintf(stderr, Name ": not performing --add as that would convert %s in to a spare.\n",
dv->devname);
fprintf(stderr, Name ": To make this a spare, use \"mdadm --zero-superblock %s\" first.\n",
dv->devname);
if (tfd >= 0)
close(tfd);
return 1;
}
} else {
/* non-persistent. Must ensure that new drive
* is at least array.size big.
*/
if (ldsize/512 < array_size) {
fprintf(stderr, Name ": %s not large enough to join array\n",
dv->devname);
if (tfd >= 0)
close(tfd);
return 1;
}
}
/* committed to really trying this device now*/
if (tfd >= 0) {
remove_partitions(tfd);
close(tfd);
tfd = -1;
}
/* in 2.6.17 and earlier, version-1 superblocks won't
* use the number we write, but will choose a free number.
* we must choose the same free number, which requires
* starting at 'raid_disks' and counting up
*/
for (j = array.raid_disks; j< tst->max_devs; j++) {
disc.number = j;
if (ioctl(fd, GET_DISK_INFO, &disc))
break;
if (disc.major==0 && disc.minor==0)
break;
if (disc.state & 8) /* removed */
break;
}
disc.major = major(stb.st_rdev);
disc.minor = minor(stb.st_rdev);
disc.number =j;
disc.state = 0;
if (array.not_persistent==0) {
int dfd;
if (dv->writemostly == 1)
disc.state |= 1 << MD_DISK_WRITEMOSTLY;
dfd = dev_open(dv->devname, O_RDWR | O_EXCL|O_DIRECT);
if (tst->ss->add_to_super(tst, &disc, dfd,
dv->devname)) {
close(dfd);
return 1;
}
if (tst->ss->write_init_super(tst)) {
close(dfd);
return 1;
}
} else if (dv->re_add) {
/* this had better be raid1.
* As we are "--re-add"ing we must find a spare slot
* to fill.
*/
char *used = malloc(array.raid_disks);
memset(used, 0, array.raid_disks);
for (j=0; j< tst->max_devs; j++) {
mdu_disk_info_t disc2;
disc2.number = j;
if (ioctl(fd, GET_DISK_INFO, &disc2))
continue;
if (disc2.major==0 && disc2.minor==0)
continue;
if (disc2.state & 8) /* removed */
continue;
if (disc2.raid_disk < 0)
continue;
if (disc2.raid_disk > array.raid_disks)
continue;
used[disc2.raid_disk] = 1;
}
for (j=0 ; j<array.raid_disks; j++)
if (!used[j]) {
disc.raid_disk = j;
disc.state |= (1<<MD_DISK_SYNC);
break;
}
free(used);
}
if (dv->writemostly == 1)
disc.state |= (1 << MD_DISK_WRITEMOSTLY);
if (tst->ss->external) {
/* add a disk
* to an external metadata container */
struct mdinfo new_mdi;
struct mdinfo *sra;
int container_fd;
int devnum = fd2devnum(fd);
int dfd;
container_fd = open_dev_excl(devnum);
if (container_fd < 0) {
fprintf(stderr, Name ": add failed for %s:"
" could not get exclusive access to container\n",
dv->devname);
tst->ss->free_super(tst);
return 1;
}
dfd = dev_open(dv->devname, O_RDWR | O_EXCL|O_DIRECT);
if (mdmon_running(tst->container_dev))
tst->update_tail = &tst->updates;
if (tst->ss->add_to_super(tst, &disc, dfd,
dv->devname)) {
close(dfd);
close(container_fd);
return 1;
}
if (tst->update_tail)
flush_metadata_updates(tst);
else
tst->ss->sync_metadata(tst);
sra = sysfs_read(container_fd, -1, 0);
if (!sra) {
fprintf(stderr, Name ": add failed for %s: sysfs_read failed\n",
dv->devname);
close(container_fd);
tst->ss->free_super(tst);
return 1;
}
sra->array.level = LEVEL_CONTAINER;
/* Need to set data_offset and component_size */
tst->ss->getinfo_super(tst, &new_mdi, NULL);
new_mdi.disk.major = disc.major;
new_mdi.disk.minor = disc.minor;
new_mdi.recovery_start = 0;
/* Make sure fds are closed as they are O_EXCL which
* would block add_disk */
tst->ss->free_super(tst);
if (sysfs_add_disk(sra, &new_mdi, 0) != 0) {
fprintf(stderr, Name ": add new device to external metadata"
" failed for %s\n", dv->devname);
close(container_fd);
sysfs_free(sra);
return 1;
}
ping_monitor_by_id(devnum);
sysfs_free(sra);
close(container_fd);
} else {
tst->ss->free_super(tst);
if (ioctl(fd, ADD_NEW_DISK, &disc)) {
fprintf(stderr, Name ": add new device failed for %s as %d: %s\n",
dv->devname, j, strerror(errno));
return 1;
}
}
if (verbose >= 0)
fprintf(stderr, Name ": added %s\n", dv->devname);
break;
case 'r':
/* hot remove */
if (subarray) {
fprintf(stderr, Name ": Cannot remove disks from a"
" \'member\' array, perform this"
" operation on the parent container\n");
if (sysfd >= 0)
close(sysfd);
return 1;
}
if (tst->ss->external) {
/* To remove a device from a container, we must
* check that it isn't in use in an array.
* This involves looking in the 'holders'
* directory - there must be just one entry,
* the container.
* To ensure that it doesn't get used as a
* hold spare while we are checking, we
* get an O_EXCL open on the container
*/
int dnum = fd2devnum(fd);
lfd = open_dev_excl(dnum);
if (lfd < 0) {
fprintf(stderr, Name
": Cannot get exclusive access "
" to container - odd\n");
if (sysfd >= 0)
close(sysfd);
return 1;
}
/* in the detached case it is not possible to
* check if we are the unique holder, so just
* rely on the 'detached' checks
*/
if (strcmp(dv->devname, "detached") == 0 ||
sysfd >= 0 ||
sysfs_unique_holder(dnum, stb.st_rdev))
/* pass */;
else {
fprintf(stderr, Name
": %s is %s, cannot remove.\n",
dnprintable,
errno == EEXIST ? "still in use":
"not a member");
close(lfd);
return 1;
}
}
/* FIXME check that it is a current member */
if (sysfd >= 0) {
/* device has been removed and we don't know
* the major:minor number
*/
int n = write(sysfd, "remove", 6);
if (n != 6)
err = -1;
else
err = 0;
close(sysfd);
sysfd = -1;
} else {
err = ioctl(fd, HOT_REMOVE_DISK, (unsigned long)stb.st_rdev);
if (err && errno == ENODEV) {
/* Old kernels rejected this if no personality
* registered */
struct mdinfo *sra = sysfs_read(fd, 0, GET_DEVS);
struct mdinfo *dv = NULL;
if (sra)
dv = sra->devs;
for ( ; dv ; dv=dv->next)
if (dv->disk.major == (int)major(stb.st_rdev) &&
dv->disk.minor == (int)minor(stb.st_rdev))
break;
if (dv)
err = sysfs_set_str(sra, dv,
"state", "remove");
else
err = -1;
if (sra)
sysfs_free(sra);
}
}
if (err) {
fprintf(stderr, Name ": hot remove failed "
"for %s: %s\n", dnprintable,
strerror(errno));
if (lfd >= 0)
close(lfd);
return 1;
}
if (tst->ss->external) {
/*
* Before dropping our exclusive open we make an
* attempt at preventing mdmon from seeing an
* 'add' event before reconciling this 'remove'
* event.
*/
char *name = devnum2devname(fd2devnum(fd));
if (!name) {
fprintf(stderr, Name ": unable to get container name\n");
return 1;
}
ping_manager(name);
free(name);
}
if (lfd >= 0)
close(lfd);
count++;
if (verbose >= 0)
fprintf(stderr, Name ": hot removed %s from %s\n",
dnprintable, devname);
break;
case 'f': /* set faulty */
/* FIXME check current member */
if ((sysfd >= 0 && write(sysfd, "faulty", 6) != 6) ||
(sysfd < 0 && ioctl(fd, SET_DISK_FAULTY,
(unsigned long) stb.st_rdev))) {
fprintf(stderr, Name ": set device faulty failed for %s: %s\n",
dnprintable, strerror(errno));
if (sysfd >= 0)
close(sysfd);
return 1;
}
if (sysfd >= 0)
close(sysfd);
sysfd = -1;
count++;
if (verbose >= 0)
fprintf(stderr, Name ": set %s faulty in %s\n",
dnprintable, devname);
break;
}
}
if (test && count == 0)
return 2;
return 0;
}
int Manage_ro(char *devname, int fd, int readonly)
{
/* switch to readonly or rw
*
* requires >= 0.90.0
* first check that array is runing
* use RESTART_ARRAY_RW or STOP_ARRAY_RO
*
*/
mdu_array_info_t array;
#ifndef MDASSEMBLE
struct mdinfo *mdi;
#endif
if (md_get_version(fd) < 9000) {
fprintf(stderr, Name ": need md driver version 0.90.0 or later\n");
return 1;
}
#ifndef MDASSEMBLE
/* If this is an externally-manage array, we need to modify the
* metadata_version so that mdmon doesn't undo our change.
*/
mdi = sysfs_read(fd, -1, GET_LEVEL|GET_VERSION);
if (mdi &&
mdi->array.major_version == -1 &&
is_subarray(mdi->text_version)) {
char vers[64];
strcpy(vers, "external:");
strcat(vers, mdi->text_version);
if (readonly > 0) {
int rv;
/* We set readonly ourselves. */
vers[9] = '-';
sysfs_set_str(mdi, NULL, "metadata_version", vers);
close(fd);
rv = sysfs_set_str(mdi, NULL, "array_state", "readonly");
if (rv < 0) {
fprintf(stderr, Name ": failed to set readonly for %s: %s\n",
devname, strerror(errno));
vers[9] = mdi->text_version[0];
sysfs_set_str(mdi, NULL, "metadata_version", vers);
return 1;
}
} else {
char *cp;
/* We cannot set read/write - must signal mdmon */
vers[9] = '/';
sysfs_set_str(mdi, NULL, "metadata_version", vers);
cp = strchr(vers+10, '/');
if (*cp)
*cp = 0;
ping_monitor(vers+10);
if (mdi->array.level <= 0)
sysfs_set_str(mdi, NULL, "array_state", "active");
}
return 0;
}
#endif
if (ioctl(fd, GET_ARRAY_INFO, &array)) {
fprintf(stderr, Name ": %s does not appear to be active.\n",
devname);
return 1;
}
if (readonly>0) {
if (ioctl(fd, STOP_ARRAY_RO, NULL)) {
fprintf(stderr, Name ": failed to set readonly for %s: %s\n",
devname, strerror(errno));
return 1;
}
} else if (readonly < 0) {
if (ioctl(fd, RESTART_ARRAY_RW, NULL)) {
fprintf(stderr, Name ": failed to set writable for %s: %s\n",
devname, strerror(errno));
return 1;
}
}
return 0;
}
int Manage_runstop(char *devname, int fd, int runstop, int quiet)
{
/* Run or stop the array. array must already be configured
* required >= 0.90.0
* Only print failure messages if quiet == 0;
* quiet > 0 means really be quiet
* quiet < 0 means we will try again if it fails.
*/
mdu_param_t param; /* unused */
if (runstop == -1 && md_get_version(fd) < 9000) {
if (ioctl(fd, STOP_MD, 0)) {
if (quiet == 0) fprintf(stderr,
Name ": stopping device %s "
"failed: %s\n",
devname, strerror(errno));
return 1;
}
}
if (md_get_version(fd) < 9000) {
fprintf(stderr, Name ": need md driver version 0.90.0 or later\n");
return 1;
}
/*
if (ioctl(fd, GET_ARRAY_INFO, &array)) {
fprintf(stderr, Name ": %s does not appear to be active.\n",
devname);
return 1;
}
*/
if (runstop>0) {
if (ioctl(fd, RUN_ARRAY, ¶m)) {
fprintf(stderr, Name ": failed to run array %s: %s\n",
devname, strerror(errno));
return 1;
}
if (quiet <= 0)
fprintf(stderr, Name ": started %s\n", devname);
} else if (runstop < 0){
struct map_ent *map = NULL;
struct stat stb;
struct mdinfo *mdi;
int devnum;
int err;
int count;
/* If this is an mdmon managed array, just write 'inactive'
* to the array state and let mdmon clear up.
*/
devnum = fd2devnum(fd);
/* Get EXCL access first. If this fails, then attempting
* to stop is probably a bad idea.
*/
close(fd);
fd = open(devname, O_RDONLY|O_EXCL);
if (fd < 0 || fd2devnum(fd) != devnum) {
if (fd >= 0)
close(fd);
fprintf(stderr,
Name ": Cannot get exclusive access to %s:"
"Perhaps a running "
"process, mounted filesystem "
"or active volume group?\n",
devname);
return 1;
}
mdi = sysfs_read(fd, -1, GET_LEVEL|GET_VERSION);
if (mdi &&
mdi->array.level > 0 &&
is_subarray(mdi->text_version)) {
int err;
/* This is mdmon managed. */
close(fd);
count = 25;
while (count &&
(err = sysfs_set_str(mdi, NULL,
"array_state",
"inactive")) < 0
&& errno == EBUSY) {
usleep(200000);
count--;
}
if (err && !quiet) {
fprintf(stderr, Name
": failed to stop array %s: %s\n",
devname, strerror(errno));
return 1;
}
/* Give monitor a chance to act */
ping_monitor(mdi->text_version);
fd = open_dev_excl(devnum);
if (fd < 0) {
fprintf(stderr, Name
": failed to completely stop %s"
": Device is busy\n",
devname);
return 1;
}
} else if (mdi &&
mdi->array.major_version == -1 &&
mdi->array.minor_version == -2 &&
!is_subarray(mdi->text_version)) {
struct mdstat_ent *mds, *m;
/* container, possibly mdmon-managed.
* Make sure mdmon isn't opening it, which
* would interfere with the 'stop'
*/
ping_monitor(mdi->sys_name);
/* now check that there are no existing arrays
* which are members of this array
*/
mds = mdstat_read(0, 0);
for (m=mds; m; m=m->next)
if (m->metadata_version &&
strncmp(m->metadata_version, "external:", 9)==0 &&
is_subarray(m->metadata_version+9) &&
devname2devnum(m->metadata_version+10) == devnum) {
if (!quiet)
fprintf(stderr, Name
": Cannot stop container %s: "
"member %s still active\n",
devname, m->dev);
free_mdstat(mds);
if (mdi)
sysfs_free(mdi);
return 1;
}
}
/* As we have an O_EXCL open, any use of the device
* which blocks STOP_ARRAY is probably a transient use,
* so it is reasonable to retry for a while - 5 seconds.
*/
count = 25; err = 0;
while (count && fd >= 0
&& (err = ioctl(fd, STOP_ARRAY, NULL)) < 0
&& errno == EBUSY) {
usleep(200000);
count --;
}
if (fd >= 0 && err) {
if (quiet == 0) {
fprintf(stderr, Name
": failed to stop array %s: %s\n",
devname, strerror(errno));
if (errno == EBUSY)
fprintf(stderr, "Perhaps a running "
"process, mounted filesystem "
"or active volume group?\n");
}
if (mdi)
sysfs_free(mdi);
return 1;
}
/* prior to 2.6.28, KOBJ_CHANGE was not sent when an md array
* was stopped, so We'll do it here just to be sure. Drop any
* partitions as well...
*/
if (fd >= 0)
ioctl(fd, BLKRRPART, 0);
if (mdi)
sysfs_uevent(mdi, "change");
if (devnum != NoMdDev &&
(stat("/dev/.udev", &stb) != 0 ||
check_env("MDADM_NO_UDEV"))) {
struct map_ent *mp = map_by_devnum(&map, devnum);
remove_devices(devnum, mp ? mp->path : NULL);
}
if (quiet <= 0)
fprintf(stderr, Name ": stopped %s\n", devname);
map_lock(&map);
map_remove(&map, devnum);
map_unlock(&map);
}
return 0;
}
