int nfd, fd2;
int d, nd;
struct supertype *st = NULL;
-
+ char *subarray = NULL;
if (ioctl(fd, GET_ARRAY_INFO, &info.array) < 0) {
fprintf(stderr, Name ": cannot get array info for %s\n", devname);
return 1;
}
- st = super_by_fd(fd);
+ if (info.array.level != -1) {
+ fprintf(stderr, Name ": can only add devices to linear arrays\n");
+ return 1;
+ }
+
+ st = super_by_fd(fd, &subarray);
if (!st) {
fprintf(stderr, Name ": cannot handle arrays with superblock version %d\n", info.array.major_version);
return 1;
}
- if (info.array.level != -1) {
- fprintf(stderr, Name ": can only add devices to linear arrays\n");
- return 1;
+ if (subarray) {
+ fprintf(stderr, Name ": Cannot grow linear sub-arrays yet\n");
+ free(subarray);
+ free(st);
}
nfd = open(newdev, O_RDWR|O_EXCL|O_DIRECT);
if (nfd < 0) {
fprintf(stderr, Name ": cannot open %s\n", newdev);
+ free(st);
return 1;
}
fstat(nfd, &stb);
if ((stb.st_mode & S_IFMT) != S_IFBLK) {
fprintf(stderr, Name ": %s is not a block device!\n", newdev);
close(nfd);
+ free(st);
return 1;
}
/* now check out all the devices and make sure we can read the superblock */
mdu_disk_info_t disk;
char *dv;
+ st->ss->free_super(st);
+
disk.number = d;
if (ioctl(fd, GET_DISK_INFO, &disk) < 0) {
fprintf(stderr, Name ": cannot get device detail for device %d\n",
d);
+ close(nfd);
+ free(st);
return 1;
}
dv = map_dev(disk.major, disk.minor, 1);
if (!dv) {
fprintf(stderr, Name ": cannot find device file for device %d\n",
d);
+ close(nfd);
+ free(st);
return 1;
}
fd2 = dev_open(dv, O_RDWR);
if (!fd2) {
fprintf(stderr, Name ": cannot open device file %s\n", dv);
+ close(nfd);
+ free(st);
return 1;
}
- st->ss->free_super(st);
if (st->ss->load_super(st, fd2, NULL)) {
fprintf(stderr, Name ": cannot find super block on %s\n", dv);
+ close(nfd);
close(fd2);
+ free(st);
return 1;
}
close(fd2);
mdu_bitmap_file_t bmf;
mdu_array_info_t array;
struct supertype *st;
+ char *subarray = NULL;
int major = BITMAP_MAJOR_HI;
int vers = md_get_version(fd);
unsigned long long bitmapsize, array_size;
if (vers < 9003) {
major = BITMAP_MAJOR_HOSTENDIAN;
-#ifdef __BIG_ENDIAN
- fprintf(stderr, Name ": Warning - bitmaps created on this kernel are not portable\n"
- " between different architectured. Consider upgrading the Linux kernel.\n");
-#endif
+ fprintf(stderr, Name ": Warning - bitmaps created on this kernel"
+ " are not portable\n"
+ " between different architectures. Consider upgrading"
+ " the Linux kernel.\n");
}
if (ioctl(fd, GET_BITMAP_FILE, &bmf) != 0) {
devname);
return 1;
}
+
+ if (strcmp(file, "none") == 0) {
+ fprintf(stderr, Name ": no bitmap found on %s\n", devname);
+ return 1;
+ }
if (array.level <= 0) {
fprintf(stderr, Name ": Bitmaps not meaningful with level %s\n",
map_num(pers, array.level)?:"of this array");
bitmapsize = bitmapsize * array.raid_disks / ncopies;
}
- st = super_by_fd(fd);
+ st = super_by_fd(fd, &subarray);
if (!st) {
fprintf(stderr, Name ": Cannot understand version %d.%d\n",
array.major_version, array.minor_version);
return 1;
}
- if (strcmp(file, "none") == 0) {
- fprintf(stderr, Name ": no bitmap found on %s\n", devname);
+ if (subarray) {
+ fprintf(stderr, Name ": Cannot add bitmaps to sub-arrays yet\n");
+ free(subarray);
+ free(st);
return 1;
- } else if (strcmp(file, "internal") == 0) {
+ }
+ if (strcmp(file, "internal") == 0) {
int d;
+ if (st->ss->add_internal_bitmap == NULL) {
+ fprintf(stderr, Name ": Internal bitmaps not supported "
+ "with %s metadata\n", st->ss->name);
+ return 1;
+ }
for (d=0; d< st->max_devs; d++) {
mdu_disk_info_t disk;
char *dv;
}
array.state |= (1<<MD_SB_BITMAP_PRESENT);
if (ioctl(fd, SET_ARRAY_INFO, &array)!= 0) {
+ if (errno == EBUSY)
+ fprintf(stderr, Name
+ ": Cannot add bitmap while array is"
+ " resyncing or reshaping etc.\n");
fprintf(stderr, Name ": failed to set internal bitmap.\n");
return 1;
}
return 1;
}
if (ioctl(fd, SET_BITMAP_FILE, bitmap_fd) < 0) {
+ int err = errno;
+ if (errno == EBUSY)
+ fprintf(stderr, Name
+ ": Cannot add bitmap while array is"
+ " resyncing or reshaping etc.\n");
fprintf(stderr, Name ": Cannot set bitmap file for %s: %s\n",
- devname, strerror(errno));
+ devname, strerror(err));
return 1;
}
}
/*
* When reshaping an array we might need to backup some data.
* This is written to all spares with a 'super_block' describing it.
- * The superblock goes 1K form the end of the used space on the
+ * The superblock goes 4K from the end of the used space on the
* device.
* It if written after the backup is complete.
* It has the following structure.
__u8 pad[512-68-32];
} __attribute__((aligned(512))) bsb, bsb2;
-int bsb_csum(char *buf, int len)
+static __u32 bsb_csum(char *buf, int len)
{
int i;
int csum = 0;
int disks, int chunk, int level, int layout, int data,
int dests, int *destfd, unsigned long long *destoffsets);
-int freeze_array(struct mdinfo *sra)
+static int check_idle(struct supertype *st)
{
- /* Try to freeze resync on this array.
+ /* Check that all member arrays for this container, or the
+ * container of this array, are idle
+ */
+ int container_dev = (st->container_dev != NoMdDev
+ ? st->container_dev : st->devnum);
+ char container[40];
+ struct mdstat_ent *ent, *e;
+ int is_idle = 1;
+
+ fmt_devname(container, container_dev);
+ ent = mdstat_read(0, 0);
+ for (e = ent ; e; e = e->next) {
+ if (!is_container_member(e, container))
+ continue;
+ if (e->percent >= 0) {
+ is_idle = 0;
+ break;
+ }
+ }
+ free_mdstat(ent);
+ return is_idle;
+}
+
+static int freeze_container(struct supertype *st)
+{
+ int container_dev = (st->container_dev != NoMdDev
+ ? st->container_dev : st->devnum);
+ char container[40];
+
+ if (!check_idle(st))
+ return -1;
+
+ fmt_devname(container, container_dev);
+
+ if (block_monitor(container, 1)) {
+ fprintf(stderr, Name ": failed to freeze container\n");
+ return -2;
+ }
+
+ return 1;
+}
+
+static void unfreeze_container(struct supertype *st)
+{
+ int container_dev = (st->container_dev != NoMdDev
+ ? st->container_dev : st->devnum);
+ char container[40];
+
+ fmt_devname(container, container_dev);
+
+ unblock_monitor(container, 1);
+}
+
+static int freeze(struct supertype *st)
+{
+ /* Try to freeze resync/rebuild on this array/container.
* Return -1 if the array is busy,
+ * return -2 container cannot be frozen,
* return 0 if this kernel doesn't support 'frozen'
* return 1 if it worked.
*/
- char buf[20];
- if (sysfs_get_str(sra, NULL, "sync_action", buf, 20) <= 0)
- return 0;
- if (strcmp(buf, "idle\n") != 0 &&
- strcmp(buf, "frozen\n") != 0)
- return -1;
- if (sysfs_set_str(sra, NULL, "sync_action", "frozen") < 0)
- return 0;
- return 1;
+ if (st->ss->external)
+ return freeze_container(st);
+ else {
+ struct mdinfo *sra = sysfs_read(-1, st->devnum, GET_VERSION);
+ int err;
+
+ if (!sra)
+ return -1;
+ err = sysfs_freeze_array(sra);
+ sysfs_free(sra);
+ return err;
+ }
}
-void unfreeze_array(struct mdinfo *sra, int frozen)
+static void unfreeze(struct supertype *st, int frozen)
{
/* If 'frozen' is 1, unfreeze the array */
- if (frozen > 0)
- sysfs_set_str(sra, NULL, "sync_action", "idle");
+ if (frozen <= 0)
+ return;
+
+ if (st->ss->external)
+ return unfreeze_container(st);
+ else {
+ struct mdinfo *sra = sysfs_read(-1, st->devnum, GET_VERSION);
+
+ if (sra)
+ sysfs_set_str(sra, NULL, "sync_action", "idle");
+ else
+ fprintf(stderr, Name ": failed to unfreeze array\n");
+ sysfs_free(sra);
+ }
}
-void wait_reshape(struct mdinfo *sra)
+static void wait_reshape(struct mdinfo *sra)
{
int fd = sysfs_get_fd(sra, NULL, "sync_action");
char action[20];
- do {
+ if (fd < 0)
+ return;
+
+ while (sysfs_fd_get_str(fd, action, 20) > 0 &&
+ strncmp(action, "reshape", 7) == 0) {
fd_set rfds;
FD_ZERO(&rfds);
FD_SET(fd, &rfds);
select(fd+1, NULL, NULL, &rfds, NULL);
-
- if (sysfs_fd_get_str(fd, action, 20) < 0) {
- close(fd);
- return;
+ }
+ close(fd);
+}
+
+static int reshape_super(struct supertype *st, long long size, int level,
+ int layout, int chunksize, int raid_disks,
+ char *backup_file, char *dev, int verbose)
+{
+ /* nothing extra to check in the native case */
+ if (!st->ss->external)
+ return 0;
+ if (!st->ss->reshape_super ||
+ !st->ss->manage_reshape) {
+ fprintf(stderr, Name ": %s metadata does not support reshape\n",
+ st->ss->name);
+ return 1;
+ }
+
+ return st->ss->reshape_super(st, size, level, layout, chunksize,
+ raid_disks, backup_file, dev, verbose);
+}
+
+static void sync_metadata(struct supertype *st)
+{
+ if (st->ss->external) {
+ if (st->update_tail) {
+ flush_metadata_updates(st);
+ st->update_tail = &st->updates;
+ } else
+ st->ss->sync_metadata(st);
+ }
+}
+
+static int subarray_set_num(char *container, struct mdinfo *sra, char *name, int n)
+{
+ /* when dealing with external metadata subarrays we need to be
+ * prepared to handle EAGAIN. The kernel may need to wait for
+ * mdmon to mark the array active so the kernel can handle
+ * allocations/writeback when preparing the reshape action
+ * (md_allow_write()). We temporarily disable safe_mode_delay
+ * to close a race with the array_state going clean before the
+ * next write to raid_disks / stripe_cache_size
+ */
+ char safe[50];
+ int rc;
+
+ /* only 'raid_disks' and 'stripe_cache_size' trigger md_allow_write */
+ if (strcmp(name, "raid_disks") != 0 &&
+ strcmp(name, "stripe_cache_size") != 0)
+ return sysfs_set_num(sra, NULL, name, n);
+
+ rc = sysfs_get_str(sra, NULL, "safe_mode_delay", safe, sizeof(safe));
+ if (rc <= 0)
+ return -1;
+ sysfs_set_num(sra, NULL, "safe_mode_delay", 0);
+ rc = sysfs_set_num(sra, NULL, name, n);
+ if (rc < 0 && errno == EAGAIN) {
+ ping_monitor(container);
+ /* if we get EAGAIN here then the monitor is not active
+ * so stop trying
+ */
+ rc = sysfs_set_num(sra, NULL, name, n);
+ }
+ sysfs_set_str(sra, NULL, "safe_mode_delay", safe);
+ return rc;
+}
+
+int start_reshape(struct mdinfo *sra)
+{
+ int err;
+ err = sysfs_set_num(sra, NULL, "suspend_hi", 0);
+ err = err ?: sysfs_set_num(sra, NULL, "suspend_lo", 0);
+ err = err ?: sysfs_set_num(sra, NULL, "sync_min", 0);
+ err = err ?: sysfs_set_num(sra, NULL, "sync_max", 0);
+ err = err ?: sysfs_set_str(sra, NULL, "sync_action", "reshape");
+
+ return err;
+}
+
+void abort_reshape(struct mdinfo *sra)
+{
+ sysfs_set_str(sra, NULL, "sync_action", "idle");
+ sysfs_set_num(sra, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL);
+ sysfs_set_num(sra, NULL, "suspend_hi", 0);
+ sysfs_set_num(sra, NULL, "suspend_lo", 0);
+ sysfs_set_num(sra, NULL, "sync_min", 0);
+ sysfs_set_str(sra, NULL, "sync_max", "max");
+}
+
+static int reshape_container_raid_disks(struct supertype *st,
+ char *container, int raid_disks)
+{
+ /* for each subarray switch to a raid level that can
+ * support the reshape, and set raid disks
+ */
+ struct mdstat_ent *ent, *e;
+ int changed = 0, rv = 0, err = 0;
+
+ ent = mdstat_read(1, 0);
+ if (!ent) {
+ fprintf(stderr, Name ": unable to read /proc/mdstat\n");
+ return -1;
+ }
+
+ changed = 0;
+ for (e = ent; e; e = e->next) {
+ struct mdinfo *sub, *info;
+ unsigned int cache;
+ int level, takeover_delta = 0;
+ int parity_disks = 1;
+ unsigned int odata;
+ unsigned long blocks;
+ char *subarray;
+
+ if (!is_container_member(e, container))
+ continue;
+ subarray = strchr(e->metadata_version+10, '/')+1;
+
+ rv = -1;
+ level = map_name(pers, e->level);
+ if (level == 0) {
+ sub = sysfs_read(-1, e->devnum, GET_VERSION);
+ if (!sub)
+ break;
+ /* metadata records 'orig_level' */
+ rv = sysfs_set_num(sub, NULL, "level", 4);
+ if (rv < 0) {
+ err = errno;
+ break;
+ }
+ /* we want spares to be used for capacity
+ * expansion, not rebuild
+ */
+ takeover_delta = 1;
+
+ sysfs_free(sub);
+ level = 4;
+ }
+ rv = -1;
+ sub = NULL;
+ switch (level) {
+ default:
+ break;
+ case 6:
+ parity_disks++;
+ case 4:
+ case 5:
+ sub = sysfs_read(-1, e->devnum, GET_CHUNK|GET_CACHE|GET_DISKS);
+ if (!sub)
+ break;
+ cache = (sub->array.chunk_size / 4096) * 4;
+ odata = sub->array.raid_disks - parity_disks;
+ blocks = compute_backup_blocks(sub->array.chunk_size,
+ sub->array.chunk_size,
+ raid_disks - parity_disks,
+ odata);
+ if (cache < blocks / 8 / odata + 16)
+ /* Make it big enough to hold 'blocks' */
+ cache = blocks / 8 / odata + 16;
+ if (cache > sub->cache_size)
+ rv = subarray_set_num(container, sub,
+ "stripe_cache_size", cache);
+ if (rv) {
+ err = errno;
+ break;
+ }
+ rv = -1;
+ /* fall through */
+ case 1:
+ if (!sub)
+ sub = sysfs_read(-1, e->devnum, GET_VERSION);
+ if (!sub)
+ break;
+
+ rv = subarray_set_num(container, sub, "raid_disks",
+ raid_disks + takeover_delta);
+ if (rv)
+ err = errno;
+ else
+ changed++;
+ break;
+ }
+
+ /* add the devices that were chosen */
+ info = st->ss->container_content(st, subarray);
+ if (info) {
+ struct mdinfo *d;
+ for (d = info->devs; d; d = d->next) {
+ if (d->disk.state == 0 &&
+ d->disk.raid_disk >= 0) {
+ /* This is a spare that wants to
+ * be part of the array.
+ */
+ add_disk(-1, st, info, d);
+ }
+ }
}
- } while (strncmp(action, "reshape", 7) == 0);
+ sysfs_free(info);
+
+ if (!rv && level > 1)
+ start_reshape(sub);
+ sysfs_free(sub);
+ if (rv)
+ break;
+ }
+ free_mdstat(ent);
+ if (rv) {
+ fprintf(stderr, Name
+ ": failed to initiate container reshape%s%s\n",
+ err ? ": " : "", err ? strerror(err) : "");
+ return rv;
+ }
+
+ return changed;
}
-
-
+
+static void revert_container_raid_disks(struct supertype *st, int fd, char *container)
+{
+ /* we failed to prepare all subarrays in the container for
+ * reshape, so cancel the changes and restore the nominal raid
+ * level
+ */
+ struct mdstat_ent *ent, *e;
+
+ ent = mdstat_read(0, 0);
+ if (!ent) {
+ fprintf(stderr, Name
+ ": failed to read /proc/mdstat while aborting reshape\n");
+ return;
+ }
+
+ if (st->ss->load_container(st, fd, NULL)) {
+ fprintf(stderr, Name
+ ": failed read metadata while aborting reshape\n");
+ return ;
+ }
+
+
+ for (e = ent; e; e = e->next) {
+ int level_fixed = 0, disks_fixed = 0;
+ struct mdinfo *sub, *prev;
+ char *subarray;
+
+ if (!is_container_member(e, container))
+ continue;
+
+ subarray = to_subarray(e, container);
+ prev = st->ss->container_content(st, subarray);
+
+ /* changing level might change raid_disks so we do it
+ * first and then check if raid_disks still needs fixing
+ */
+ if (map_name(pers, e->level) != prev->array.level) {
+ sub = sysfs_read(-1, e->devnum, GET_VERSION);
+ if (sub &&
+ !sysfs_set_num(sub, NULL, "level", prev->array.level))
+ level_fixed = 1;
+ sysfs_free(sub);
+ } else
+ level_fixed = 1;
+
+ sub = sysfs_read(-1, e->devnum, GET_DISKS);
+ if (sub && sub->array.raid_disks != prev->array.raid_disks) {
+ if (!subarray_set_num(container, sub, "raid_disks",
+ prev->array.raid_disks))
+ disks_fixed = 1;
+ } else if (sub)
+ disks_fixed = 1;
+
+ if (sub)
+ abort_reshape(sub);
+ sysfs_free(sub);
+
+ if (!disks_fixed || !level_fixed)
+ fprintf(stderr, Name
+ ": failed to restore %s to a %d-disk %s array\n",
+ e->dev, prev->array.raid_disks,
+ map_num(pers, prev->array.level));
+ free(prev);
+ }
+ st->ss->free_super(st);
+ free_mdstat(ent);
+}
+
+int remove_disks_on_raid10_to_raid0_takeover(struct supertype *st,
+ struct mdinfo *sra,
+ int layout)
+{
+ int nr_of_copies;
+ struct mdinfo *remaining;
+ int slot;
+
+ nr_of_copies = layout & 0xff;
+
+ remaining = sra->devs;
+ sra->devs = NULL;
+ /* for each 'copy', select one device and remove from the list. */
+ for (slot = 0; slot < sra->array.raid_disks; slot += nr_of_copies) {
+ struct mdinfo **diskp;
+ int found = 0;
+
+ /* Find a working device to keep */
+ for (diskp = &remaining; *diskp ; diskp = &(*diskp)->next) {
+ struct mdinfo *disk = *diskp;
+
+ if (disk->disk.raid_disk < slot)
+ continue;
+ if (disk->disk.raid_disk >= slot + nr_of_copies)
+ continue;
+ if (disk->disk.state & (1<<MD_DISK_REMOVED))
+ continue;
+ if (disk->disk.state & (1<<MD_DISK_FAULTY))
+ continue;
+ if (!(disk->disk.state & (1<<MD_DISK_SYNC)))
+ continue;
+
+ /* We have found a good disk to use! */
+ *diskp = disk->next;
+ disk->next = sra->devs;
+ sra->devs = disk;
+ found = 1;
+ break;
+ }
+ if (!found)
+ break;
+ }
+
+ if (slot < sra->array.raid_disks) {
+ /* didn't find all slots */
+ struct mdinfo **e;
+ e = &remaining;
+ while (*e)
+ e = &(*e)->next;
+ *e = sra->devs;
+ sra->devs = remaining;
+ return 1;
+ }
+
+ /* Remove all 'remaining' devices from the array */
+ while (remaining) {
+ struct mdinfo *sd = remaining;
+ remaining = sd->next;
+
+ sysfs_set_str(sra, sd, "state", "faulty");
+ sysfs_set_str(sra, sd, "slot", "none");
+ sysfs_set_str(sra, sd, "state", "remove");
+ sd->disk.state |= (1<<MD_DISK_REMOVED);
+ sd->disk.state &= ~(1<<MD_DISK_SYNC);
+ sd->next = sra->devs;
+ sra->devs = sd;
+ }
+ return 0;
+}
+
+void reshape_free_fdlist(int *fdlist,
+ unsigned long long *offsets,
+ int size)
+{
+ int i;
+
+ for (i = 0; i < size; i++)
+ if (fdlist[i] >= 0)
+ close(fdlist[i]);
+
+ free(fdlist);
+ free(offsets);
+}
+
+int reshape_prepare_fdlist(char *devname,
+ struct mdinfo *sra,
+ int raid_disks,
+ int nrdisks,
+ unsigned long blocks,
+ char *backup_file,
+ int *fdlist,
+ unsigned long long *offsets)
+{
+ int d = 0;
+ struct mdinfo *sd;
+
+ for (d = 0; d <= nrdisks; d++)
+ fdlist[d] = -1;
+ d = raid_disks;
+ for (sd = sra->devs; sd; sd = sd->next) {
+ if (sd->disk.state & (1<<MD_DISK_FAULTY))
+ continue;
+ if (sd->disk.state & (1<<MD_DISK_SYNC)) {
+ char *dn = map_dev(sd->disk.major,
+ sd->disk.minor, 1);
+ fdlist[sd->disk.raid_disk]
+ = dev_open(dn, O_RDONLY);
+ offsets[sd->disk.raid_disk] = sd->data_offset*512;
+ if (fdlist[sd->disk.raid_disk] < 0) {
+ fprintf(stderr,
+ Name ": %s: cannot open component %s\n",
+ devname, dn ? dn : "-unknown-");
+ d = -1;
+ goto release;
+ }
+ } else if (backup_file == NULL) {
+ /* spare */
+ char *dn = map_dev(sd->disk.major,
+ sd->disk.minor, 1);
+ fdlist[d] = dev_open(dn, O_RDWR);
+ offsets[d] = (sd->data_offset + sra->component_size - blocks - 8)*512;
+ if (fdlist[d] < 0) {
+ fprintf(stderr, Name ": %s: cannot open component %s\n",
+ devname, dn ? dn : "-unknown-");
+ d = -1;
+ goto release;
+ }
+ d++;
+ }
+ }
+release:
+ return d;
+}
+
+int reshape_open_backup_file(char *backup_file,
+ int fd,
+ char *devname,
+ long blocks,
+ int *fdlist,
+ unsigned long long *offsets)
+{
+ /* Return 1 on success, 0 on any form of failure */
+ /* need to check backup file is large enough */
+ char buf[512];
+ struct stat stb;
+ unsigned int dev;
+ int i;
+
+ *fdlist = open(backup_file, O_RDWR|O_CREAT|O_EXCL,
+ S_IRUSR | S_IWUSR);
+ *offsets = 8 * 512;
+ if (*fdlist < 0) {
+ fprintf(stderr, Name ": %s: cannot create backup file %s: %s\n",
+ devname, backup_file, strerror(errno));
+ return 0;
+ }
+ /* Guard against backup file being on array device.
+ * If array is partitioned or if LVM etc is in the
+ * way this will not notice, but it is better than
+ * nothing.
+ */
+ fstat(*fdlist, &stb);
+ dev = stb.st_dev;
+ fstat(fd, &stb);
+ if (stb.st_rdev == dev) {
+ fprintf(stderr, Name ": backup file must NOT be"
+ " on the array being reshaped.\n");
+ close(*fdlist);
+ return 0;
+ }
+
+ memset(buf, 0, 512);
+ for (i=0; i < blocks + 1 ; i++) {
+ if (write(*fdlist, buf, 512) != 512) {
+ fprintf(stderr, Name ": %s: cannot create"
+ " backup file %s: %s\n",
+ devname, backup_file, strerror(errno));
+ return 0;
+ }
+ }
+ if (fsync(*fdlist) != 0) {
+ fprintf(stderr, Name ": %s: cannot create backup file %s: %s\n",
+ devname, backup_file, strerror(errno));
+ return 0;
+ }
+
+ return 1;
+}
+
+unsigned long compute_backup_blocks(int nchunk, int ochunk,
+ unsigned int ndata, unsigned int odata)
+{
+ unsigned long a, b, blocks;
+ /* So how much do we need to backup.
+ * We need an amount of data which is both a whole number of
+ * old stripes and a whole number of new stripes.
+ * So LCM for (chunksize*datadisks).
+ */
+ a = (ochunk/512) * odata;
+ b = (nchunk/512) * ndata;
+ /* Find GCD */
+ while (a != b) {
+ if (a < b)
+ b -= a;
+ if (b < a)
+ a -= b;
+ }
+ /* LCM == product / GCD */
+ blocks = (ochunk/512) * (nchunk/512) * odata * ndata / a;
+
+ return blocks;
+}
+
+
int Grow_reshape(char *devname, int fd, int quiet, char *backup_file,
long long size,
- int level, char *layout_str, int chunksize, int raid_disks)
+ int level, char *layout_str, int chunksize, int raid_disks,
+ int force)
{
/* Make some changes in the shape of an array.
* The kernel must support the change.
char *c;
int rv = 0;
struct supertype *st;
+ char *subarray = NULL;
int nchunk, ochunk;
int nlayout, olayout;
int ndisks, odisks;
- int ndata, odata;
+ unsigned int ndata, odata;
int orig_level = UnSet;
char alt_layout[40];
int *fdlist;
unsigned long long *offsets;
- int d, i;
+ int d;
int nrdisks;
int err;
int frozen;
- unsigned long a,b, blocks, stripes;
- int cache;
+ unsigned long blocks, stripes;
+ unsigned long cache;
unsigned long long array_size;
int changed = 0;
+ char *container = NULL;
+ int cfd = -1;
int done;
struct mdinfo *sra;
devname);
return 1;
}
- sra = sysfs_read(fd, 0, GET_LEVEL);
- frozen = freeze_array(sra);
- if (frozen < 0) {
+
+ if (size >= 0 &&
+ (chunksize || level!= UnSet || layout_str || raid_disks)) {
+ fprintf(stderr, Name ": cannot change component size at the same time "
+ "as other changes.\n"
+ " Change size first, then check data is intact before "
+ "making other changes.\n");
+ return 1;
+ }
+
+ if (raid_disks && raid_disks < array.raid_disks && array.level > 1 &&
+ get_linux_version() < 2006032 &&
+ !check_env("MDADM_FORCE_FEWER")) {
+ fprintf(stderr, Name ": reducing the number of devices is not safe before Linux 2.6.32\n"
+ " Please use a newer kernel\n");
+ return 1;
+ }
+
+ st = super_by_fd(fd, &subarray);
+ if (!st) {
+ fprintf(stderr, Name ": Unable to determine metadata format for %s\n", devname);
+ return 1;
+ }
+ if (raid_disks > st->max_devs) {
+ fprintf(stderr, Name ": Cannot increase raid-disks on this array"
+ " beyond %d\n", st->max_devs);
+ return 1;
+ }
+
+ /* in the external case we need to check that the requested reshape is
+ * supported, and perform an initial check that the container holds the
+ * pre-requisite spare devices (mdmon owns final validation)
+ */
+ if (st->ss->external) {
+ int container_dev;
+ int rv;
+
+ if (subarray) {
+ container_dev = st->container_dev;
+ cfd = open_dev_excl(st->container_dev);
+ } else if (size >= 0 || layout_str != NULL || chunksize != 0 ||
+ level != UnSet) {
+ fprintf(stderr,
+ Name ": %s is a container, only 'raid-devices' can be changed\n",
+ devname);
+ return 1;
+ } else {
+ container_dev = st->devnum;
+ close(fd);
+ cfd = open_dev_excl(st->devnum);
+ fd = cfd;
+ }
+ if (cfd < 0) {
+ fprintf(stderr, Name ": Unable to open container for %s\n",
+ devname);
+ free(subarray);
+ return 1;
+ }
+
+ container = devnum2devname(st->devnum);
+ if (!container) {
+ fprintf(stderr, Name ": Could not determine container name\n");
+ free(subarray);
+ return 1;
+ }
+
+ if (subarray)
+ rv = st->ss->load_container(st, cfd, NULL);
+ else
+ rv = st->ss->load_super(st, cfd, NULL);
+ if (rv) {
+ fprintf(stderr, Name ": Cannot read superblock for %s\n",
+ devname);
+ free(subarray);
+ return 1;
+ }
+
+ if (mdmon_running(container_dev))
+ st->update_tail = &st->updates;
+ }
+
+ if (raid_disks > array.raid_disks &&
+ array.spare_disks < (raid_disks - array.raid_disks) &&
+ !force) {
+ fprintf(stderr,
+ Name ": Need %d spare%s to avoid degraded array,"
+ " and only have %d.\n"
+ " Use --force to over-ride this check.\n",
+ raid_disks - array.raid_disks,
+ raid_disks - array.raid_disks == 1 ? "" : "s",
+ array.spare_disks);
+ return 1;
+ }
+
+ sra = sysfs_read(fd, 0, GET_LEVEL | GET_DISKS | GET_DEVS | GET_STATE);
+ if (sra) {
+ if (st->ss->external && subarray == NULL) {
+ array.level = LEVEL_CONTAINER;
+ sra->array.level = LEVEL_CONTAINER;
+ }
+ } else {
+ fprintf(stderr, Name ": failed to read sysfs parameters for %s\n",
+ devname);
+ return 1;
+ }
+ frozen = freeze(st);
+ if (frozen < -1) {
+ /* freeze() already spewed the reason */
+ return 1;
+ } else if (frozen < 0) {
fprintf(stderr, Name ": %s is performing resync/recovery and cannot"
" be reshaped\n", devname);
return 1;
/* ========= set size =============== */
if (size >= 0 && (size == 0 || size != array.size)) {
+ long long orig_size = array.size;
+
+ if (reshape_super(st, size, UnSet, UnSet, 0, 0, NULL, devname, !quiet)) {
+ rv = 1;
+ goto release;
+ }
+ sync_metadata(st);
array.size = size;
if (array.size != size) {
/* got truncated to 32bit, write to
} else
rv = ioctl(fd, SET_ARRAY_INFO, &array);
if (rv != 0) {
+ int err = errno;
+
+ /* restore metadata */
+ if (reshape_super(st, orig_size, UnSet, UnSet, 0, 0,
+ NULL, devname, !quiet) == 0)
+ sync_metadata(st);
fprintf(stderr, Name ": Cannot set device size for %s: %s\n",
- devname, strerror(errno));
+ devname, strerror(err));
+ if (err == EBUSY &&
+ (array.state & (1<<MD_SB_BITMAP_PRESENT)))
+ fprintf(stderr, " Bitmap must be removed before size can be changed\n");
rv = 1;
goto release;
}
ioctl(fd, GET_ARRAY_INFO, &array);
+ size = get_component_size(fd)/2;
+ if (size == 0)
+ size = array.size;
if (!quiet)
- fprintf(stderr, Name ": component size of %s has been set to %dK\n",
- devname, array.size);
+ fprintf(stderr, Name ": component size of %s has been set to %lluK\n",
+ devname, size);
changed = 1;
+ } else if (array.level != LEVEL_CONTAINER) {
+ size = get_component_size(fd)/2;
+ if (size == 0)
+ size = array.size;
+ }
+
+ /* ========= check for Raid10 -> Raid0 conversion ===============
+ * current implemenation assumes that following conditions must be met:
+ * - far_copies == 1
+ * - near_copies == 2
+ */
+ if (level == 0 && array.level == 10 &&
+ array.layout == ((1 << 8) + 2) && !(array.raid_disks & 1)) {
+ int err;
+ err = remove_disks_on_raid10_to_raid0_takeover(st, sra, array.layout);
+ if (err) {
+ dprintf(Name": Array cannot be reshaped\n");
+ if (container)
+ free(container);
+ if (cfd > -1)
+ close(cfd);
+ return 1;
+ }
}
/* ======= set level =========== */
}
}
if (raid_disks)
- /* The find raid6->raid5 conversion
+ /* The final raid6->raid5 conversion
* will reduce the number of disks,
* so now we need to aim higher
*/
} else
layout_str = "parity-last";
} else {
+ /* Level change is a simple takeover. In the external
+ * case we don't check with the metadata handler until
+ * we establish what the final layout will be. If the
+ * level change is disallowed we will revert to
+ * orig_level without disturbing the metadata, otherwise
+ * we will send an update.
+ */
c = map_num(pers, level);
- if (c == NULL)
- return 1;/* not possible */
+ if (c == NULL) {
+ rv = 1;/* not possible */
+ goto release;
+ }
+ if (!force) {
+ /* Need to check there are enough spares */
+ int spares_needed = 0;
+ switch (array.level * 16 + level) {
+ case 0x05:
+ spares_needed = 1; break;
+ case 0x06:
+ spares_needed = 2; break;
+ case 0x15:
+ spares_needed = 1; break;
+ case 0x16:
+ spares_needed = 2; break;
+ case 0x56:
+ spares_needed = 1; break;
+ }
+ if (raid_disks > array.raid_disks)
+ spares_needed += raid_disks-array.raid_disks;
+ if (spares_needed > array.spare_disks) {
+ fprintf(stderr,
+ Name ": Need %d spare%s to avoid"
+ " degraded array, and only have %d.\n"
+ " Use --force to over-ride"
+ " this check.\n",
+ spares_needed,
+ spares_needed == 1 ? "" : "s",
+ array.spare_disks);
+ rv = 1;
+ goto release;
+ }
+ }
err = sysfs_set_str(sra, NULL, "level", c);
if (err) {
+ err = errno;
fprintf(stderr, Name ": %s: could not set level to %s\n",
devname, c);
+ if (err == EBUSY &&
+ (array.state & (1<<MD_SB_BITMAP_PRESENT)))
+ fprintf(stderr, " Bitmap must be removed before level can be changed\n");
rv = 1;
goto release;
}
/* ========= set shape (chunk_size / layout / ndisks) ============== */
/* Check if layout change is a no-op */
- if (layout_str) switch(array.level) {
+ switch (array.level) {
case 5:
- if (array.layout == map_name(r5layout, layout_str))
+ if (layout_str && array.layout == map_name(r5layout, layout_str))
layout_str = NULL;
break;
case 6:
rv = 1;
goto release;
}
- if (strcmp(layout_str, "normalise") == 0 ||
- strcmp(layout_str, "normalize") == 0) {
+ if (layout_str &&
+ (strcmp(layout_str, "normalise") == 0 ||
+ strcmp(layout_str, "normalize") == 0)) {
char *hyphen;
strcpy(alt_layout, map_num(r6layout, array.layout));
hyphen = strrchr(alt_layout, '-');
}
}
- if (array.layout == map_name(r6layout, layout_str))
+ if (layout_str && array.layout == map_name(r6layout, layout_str))
layout_str = NULL;
if (layout_str && strcmp(layout_str, "preserve") == 0)
layout_str = NULL;
if (layout_str == NULL
&& (chunksize == 0 || chunksize*1024 == array.chunk_size)
&& (raid_disks == 0 || raid_disks == array.raid_disks)) {
+ if (reshape_super(st, -1, level, UnSet, 0, 0, NULL, devname, !quiet)) {
+ rv = 1;
+ goto release;
+ }
+ sync_metadata(st);
rv = 0;
if (level != UnSet && level != array.level) {
/* Looks like this level change doesn't need
c = map_num(pers, level);
if (c) {
rv = sysfs_set_str(sra, NULL, "level", c);
- if (rv)
+ if (rv) {
+ int err = errno;
fprintf(stderr, Name ": %s: could not set level to %s\n",
devname, c);
+ if (err == EBUSY &&
+ (array.state & (1<<MD_SB_BITMAP_PRESENT)))
+ fprintf(stderr, " Bitmap must be removed before level can be changed\n");
+ rv = 1;
+ }
}
} else if (!changed && !quiet)
fprintf(stderr, Name ": %s: no change requested\n",
devname);
+
+ if (st->ss->external && !mdmon_running(st->container_dev) &&
+ level > 0) {
+ start_mdmon(st->container_dev);
+ ping_monitor(container);
+ }
goto release;
}
c = map_num(pers, array.level);
if (c == NULL) c = "-unknown-";
- switch(array.level) {
+ switch (array.level) {
default: /* raid0, linear, multipath cannot be reconfigured */
fprintf(stderr, Name ": %s array %s cannot be reshaped.\n",
c, devname);
+ /* TODO raid0 raiddisks can be reshaped via raid4 */
rv = 1;
break;
+ case LEVEL_CONTAINER: {
+ int count;
+
+ /* double check that we are not changing anything but raid_disks */
+ if (size >= 0 || layout_str != NULL || chunksize != 0 || level != UnSet) {
+ fprintf(stderr,
+ Name ": %s is a container, only 'raid-devices' can be changed\n",
+ devname);
+ rv = 1;
+ goto release;
+ }
+
+ if (reshape_super(st, -1, UnSet, UnSet, 0, raid_disks,
+ backup_file, devname, !quiet)) {
+ rv = 1;
+ goto release;
+ }
+
+ count = reshape_container_raid_disks(st, container, raid_disks);
+ if (count < 0) {
+ revert_container_raid_disks(st, fd, container);
+ rv = 1;
+ goto release;
+ } else if (count == 0) {
+ if (!quiet)
+ fprintf(stderr, Name
+ ": no active subarrays to reshape\n");
+ goto release;
+ }
+
+ sync_metadata(st);
+ if (!mdmon_running(st->devnum)) {
+ start_mdmon(st->devnum);
+ ping_monitor(container);
+ st->update_tail = &st->updates;
+ }
+
+ /* give mdmon a chance to allocate spares */
+ ping_manager(container);
+ /* manage_reshape takes care of releasing the array(s) */
+ st->ss->manage_reshape(st, backup_file);
+ frozen = 0;
+ goto release;
+ }
case LEVEL_FAULTY: /* only 'layout' change is permitted */
if (chunksize || raid_disks) {
array.layout = parse_layout_faulty(layout_str);
if (array.layout < 0) {
- int rv;
fprintf(stderr, Name ": %s: layout %s not understood for 'faulty' array\n",
devname, layout_str);
rv = 1;
break;
}
if (raid_disks > 0) {
+ if (reshape_super(st, -1, UnSet, UnSet, 0, raid_disks,
+ NULL, devname, !quiet)) {
+ rv = 1;
+ goto release;
+ }
+ sync_metadata(st);
array.raid_disks = raid_disks;
if (ioctl(fd, SET_ARRAY_INFO, &array) != 0) {
fprintf(stderr, Name ": Cannot set raid-devices for %s: %s\n",
* layout/chunksize/raid_disks can be changed
* though the kernel may not support it all.
*/
- st = super_by_fd(fd);
+ if (subarray) {
+ fprintf(stderr, Name ": Cannot reshape subarrays yet\n");
+ break;
+ }
/*
* There are three possibilities.
if (chunksize) {
nchunk = chunksize * 1024;
- if (array.size % chunksize) {
- fprintf(stderr, Name ": component size %dK is not"
+ if (size % chunksize) {
+ fprintf(stderr, Name ": component size %lluK is not"
" a multiple of chunksize %dK\n",
- array.size, chunksize);
+ size, chunksize);
break;
}
}
if (nlayout == UnSet) {
fprintf(stderr, Name ": layout %s not understood for raid5.\n",
layout_str);
- return 1;
+ rv = 1;
+ goto release;
}
break;
if (nlayout == UnSet) {
fprintf(stderr, Name ": layout %s not understood for raid6.\n",
layout_str);
- return 1;
+ rv = 1;
+ goto release;
}
break;
}
ndata--;
}
+ if (odata == ndata &&
+ get_linux_version() < 2006032) {
+ fprintf(stderr, Name ": in-place reshape is not safe before 2.6.32, sorry.\n");
+ break;
+ }
+
/* Check that we can hold all the data */
- size = ndata * array.size;
get_dev_size(fd, NULL, &array_size);
- if (size < (array_size/1024)) {
+ if (ndata * (unsigned long long)size < (array_size/1024)) {
fprintf(stderr, Name ": this change will reduce the size of the array.\n"
" use --grow --array-size first to truncate array.\n"
" e.g. mdadm --grow %s --array-size %llu\n",
- devname, size);
+ devname, ndata * size);
rv = 1;
break;
}
- /* So how much do we need to backup.
- * We need an amount of data which is both a whole number of
- * old stripes and a whole number of new stripes.
- * So LCM for (chunksize*datadisks).
- */
- a = ochunk/512 * odata;
- b = nchunk/512 * ndata;
- /* Find GCD */
- while (a != b) {
- if (a < b)
- b -= a;
- if (b < a)
- a -= b;
- }
- /* LCM == product / GCD */
- blocks = ochunk/512 * nchunk/512 * odata * ndata / a;
-
- if (ndata == odata)
- blocks *= 16;
- else
- fprintf(stderr, Name ": Need to backup %luK of critical "
- "section..\n", blocks/2);
+ blocks = compute_backup_blocks(nchunk, ochunk, ndata, odata);
sysfs_free(sra);
sra = sysfs_read(fd, 0,
GET_COMPONENT|GET_DEVS|GET_OFFSET|GET_STATE|
GET_CACHE);
+
if (!sra) {
fprintf(stderr, Name ": %s: Cannot get array details from sysfs\n",
devname);
break;
}
+ if (ndata == odata) {
+ /* Make 'blocks' bigger for better throughput, but
+ * not so big that we reject it below.
+ * Try for 16 megabytes
+ */
+ while (blocks * 32 < sra->component_size &&
+ blocks < 16*1024*2)
+ blocks *= 2;
+ } else
+ fprintf(stderr, Name ": Need to backup %luK of critical "
+ "section..\n", blocks/2);
+
if (blocks >= sra->component_size/2) {
fprintf(stderr, Name ": %s: Something wrong - reshape aborted\n",
devname);
rv = 1;
break;
}
- nrdisks = array.nr_disks + sra->array.spare_disks;
+ nrdisks = array.raid_disks + sra->array.spare_disks;
/* Now we need to open all these devices so we can read/write.
*/
fdlist = malloc((1+nrdisks) * sizeof(int));
rv = 1;
break;
}
- for (d=0; d <= nrdisks; d++)
- fdlist[d] = -1;
- d = array.raid_disks;
- for (sd = sra->devs; sd; sd=sd->next) {
- if (sd->disk.state & (1<<MD_DISK_FAULTY))
- continue;
- if (sd->disk.state & (1<<MD_DISK_SYNC)) {
- char *dn = map_dev(sd->disk.major,
- sd->disk.minor, 1);
- fdlist[sd->disk.raid_disk]
- = dev_open(dn, O_RDONLY);
- offsets[sd->disk.raid_disk] = sd->data_offset*512;
- if (fdlist[sd->disk.raid_disk] < 0) {
- fprintf(stderr, Name ": %s: cannot open component %s\n",
- devname, dn?dn:"-unknown-");
- rv = 1;
- goto release;
- }
- } else if (backup_file == NULL) {
- /* spare */
- char *dn = map_dev(sd->disk.major,
- sd->disk.minor, 1);
- fdlist[d] = dev_open(dn, O_RDWR);
- offsets[d] = (sra->component_size - blocks - 8)*512;
- if (fdlist[d]<0) {
- fprintf(stderr, Name ": %s: cannot open component %s\n",
- devname, dn?dn:"-unknown");
- rv = 1;
- goto release;
- }
- d++;
- }
+
+ d = reshape_prepare_fdlist(devname, sra, array.raid_disks,
+ nrdisks, blocks, backup_file,
+ fdlist, offsets);
+ if (d < 0) {
+ rv = 1;
+ goto release;
}
if (backup_file == NULL) {
+ if (st->ss->external && !st->ss->manage_reshape) {
+ fprintf(stderr, Name ": %s Grow operation not supported by %s metadata\n",
+ devname, st->ss->name);
+ rv = 1;
+ break;
+ }
if (ndata <= odata) {
fprintf(stderr, Name ": %s: Cannot grow - need backup-file\n",
devname);
break;
}
} else {
- /* need to check backup file is large enough */
- char buf[512];
- fdlist[d] = open(backup_file, O_RDWR|O_CREAT|O_EXCL,
- S_IRUSR | S_IWUSR);
- offsets[d] = 8 * 512;
- if (fdlist[d] < 0) {
- fprintf(stderr, Name ": %s: cannot create backup file %s: %s\n",
- devname, backup_file, strerror(errno));
- rv = 1;
- break;
- }
- memset(buf, 0, 512);
- for (i=0; i < blocks + 1 ; i++) {
- if (write(fdlist[d], buf, 512) != 512) {
- fprintf(stderr, Name ": %s: cannot create backup file %s: %s\n",
- devname, backup_file, strerror(errno));
- rv = 1;
- break;
- }
- }
- if (fsync(fdlist[d]) != 0) {
- fprintf(stderr, Name ": %s: cannot create backup file %s: %s\n",
- devname, backup_file, strerror(errno));
+ if (!reshape_open_backup_file(backup_file, fd, devname,
+ (signed)blocks,
+ fdlist+d, offsets+d)) {
rv = 1;
break;
}
d++;
}
+ /* check that the operation is supported by the metadata */
+ if (reshape_super(st, -1, level, nlayout, nchunk, ndisks,
+ backup_file, devname, !quiet)) {
+ rv = 1;
+ break;
+ }
+
+ /* ->reshape_super might have chosen some spares from the
+ * container that it wants to be part of the new array.
+ * We can collect them with ->container_content and give
+ * them to the kernel.
+ */
+ if (st->ss->reshape_super && st->ss->container_content) {
+ struct mdinfo *info =
+ st->ss->container_content(st, subarray);
+ struct mdinfo *d;
+
+ if (info)
+ for (d = info->devs; d; d = d->next) {
+ if (d->disk.state == 0 &&
+ d->disk.raid_disk >= 0) {
+ /* This is a spare that wants to
+ * be part of the array.
+ */
+ add_disk(fd, st, info, d);
+ }
+ }
+ sysfs_free(info);
+ }
+
/* lastly, check that the internal stripe cache is
* large enough, or it won't work.
*/
cache = (nchunk < ochunk) ? ochunk : nchunk;
cache = cache * 4 / 4096;
+ if (cache < blocks / 8 / odata + 16)
+ /* Make it big enough to hold 'blocks' */
+ cache = blocks / 8 / odata + 16;
if (sra->cache_size < cache)
sysfs_set_num(sra, NULL, "stripe_cache_size",
cache+1);
* If only changing raid_disks, use ioctl, else use
* sysfs.
*/
+ sync_metadata(st);
if (ochunk == nchunk && olayout == nlayout) {
array.raid_disks = ndisks;
if (ioctl(fd, SET_ARRAY_INFO, &array) != 0) {
+ int err = errno;
rv = 1;
fprintf(stderr, Name ": Cannot set device shape for %s: %s\n",
devname, strerror(errno));
if (ndisks < odisks &&
get_linux_version() < 2006030)
fprintf(stderr, Name ": linux 2.6.30 or later required\n");
+ if (err == EBUSY &&
+ (array.state & (1<<MD_SB_BITMAP_PRESENT)))
+ fprintf(stderr, " Bitmap must be removed before shape can be changed\n");
break;
}
/* set them all just in case some old 'new_*' value
* persists from some earlier problem
*/
+ int err = err; /* only used if rv==1, and always set if
+ * rv==1, so initialisation not needed,
+ * despite gcc warning
+ */
if (sysfs_set_num(sra, NULL, "chunk_size", nchunk) < 0)
- rv = 1;
- if (sysfs_set_num(sra, NULL, "layout", nlayout) < 0)
- rv = 1;
- if (sysfs_set_num(sra, NULL, "raid_disks", ndisks) < 0)
- rv = 1;
+ rv = 1, err = errno;
+ if (!rv && sysfs_set_num(sra, NULL, "layout", nlayout) < 0)
+ rv = 1, err = errno;
+ if (!rv && sysfs_set_num(sra, NULL, "raid_disks", ndisks) < 0)
+ rv = 1, err = errno;
if (rv) {
fprintf(stderr, Name ": Cannot set device shape for %s\n",
devname);
if (get_linux_version() < 2006030)
fprintf(stderr, Name ": linux 2.6.30 or later required\n");
+ if (err == EBUSY &&
+ (array.state & (1<<MD_SB_BITMAP_PRESENT)))
+ fprintf(stderr, " Bitmap must be removed before shape can be changed\n");
break;
}
}
break;
}
+ start_reshape(sra);
+ if (st->ss->external) {
+ /* metadata handler takes it from here */
+ ping_manager(container);
+ st->ss->manage_reshape(st, backup_file);
+ frozen = 0;
+ break;
+ }
+
/* set up the backup-super-block. This requires the
* uuid from the array.
*/
fprintf(stderr, Name ": %s: Cannot find a superblock\n",
devname);
rv = 1;
+ abort_reshape(sra);
break;
}
d - odisks, fdlist+odisks, offsets+odisks);
if (backup_file && done)
unlink(backup_file);
- if (level != UnSet && level != array.level) {
+ if (!done)
+ abort_reshape(sra);
+ else if (level != UnSet && level != array.level) {
/* We need to wait for the reshape to finish
* (which will have happened unless odata < ndata)
* and then set the level
fprintf(stderr, Name ": Cannot run child to monitor reshape: %s\n",
strerror(errno));
rv = 1;
+ abort_reshape(sra);
break;
default:
/* The child will take care of unfreezing the array */
if (c && sysfs_set_str(sra, NULL, "level", c) == 0)
fprintf(stderr, Name ": aborting level change\n");
}
- if (sra)
- unfreeze_array(sra, frozen);
+ unfreeze(st, frozen);
return rv;
}
*
*/
-int grow_backup(struct mdinfo *sra,
+/* FIXME return status is never checked */
+static int grow_backup(struct mdinfo *sra,
unsigned long long offset, /* per device */
unsigned long stripes, /* per device */
int *sources, unsigned long long *offsets,
int disks, int chunk, int level, int layout,
int dests, int *destfd, unsigned long long *destoffsets,
- int part,
+ int part, int *degraded,
char *buf)
{
/* Backup 'blocks' sectors at 'offset' on each device of the array,
int odata = disks;
int rv = 0;
int i;
+ unsigned long long ll;
+ int new_degraded;
//printf("offset %llu\n", offset);
if (level >= 4)
odata--;
if (level == 6)
odata--;
- sysfs_set_num(sra, NULL, "suspend_hi", (offset + stripes * chunk/512) * odata);
+ sysfs_set_num(sra, NULL, "suspend_hi", (offset + stripes * (chunk/512)) * odata);
+ /* Check that array hasn't become degraded, else we might backup the wrong data */
+ sysfs_get_ll(sra, NULL, "degraded", &ll);
+ new_degraded = (int)ll;
+ if (new_degraded != *degraded) {
+ /* check each device to ensure it is still working */
+ struct mdinfo *sd;
+ for (sd = sra->devs ; sd ; sd = sd->next) {
+ if (sd->disk.state & (1<<MD_DISK_FAULTY))
+ continue;
+ if (sd->disk.state & (1<<MD_DISK_SYNC)) {
+ char sbuf[20];
+ if (sysfs_get_str(sra, sd, "state", sbuf, 20) < 0 ||
+ strstr(sbuf, "faulty") ||
+ strstr(sbuf, "in_sync") == NULL) {
+ /* this device is dead */
+ sd->disk.state = (1<<MD_DISK_FAULTY);
+ if (sd->disk.raid_disk >= 0 &&
+ sources[sd->disk.raid_disk] >= 0) {
+ close(sources[sd->disk.raid_disk]);
+ sources[sd->disk.raid_disk] = -1;
+ }
+ }
+ }
+ }
+ *degraded = new_degraded;
+ }
if (part) {
bsb.arraystart2 = __cpu_to_le64(offset * odata);
- bsb.length2 = __cpu_to_le64(stripes * chunk/512 * odata);
+ bsb.length2 = __cpu_to_le64(stripes * (chunk/512) * odata);
} else {
bsb.arraystart = __cpu_to_le64(offset * odata);
- bsb.length = __cpu_to_le64(stripes * chunk/512 * odata);
+ bsb.length = __cpu_to_le64(stripes * (chunk/512) * odata);
}
if (part)
bsb.magic[15] = '2';
if (rv)
return rv;
+ bsb.mtime = __cpu_to_le64(time(0));
for (i = 0; i < dests; i++) {
bsb.devstart = __cpu_to_le64(destoffsets[i]/512);
bsb.sb_csum2 = bsb_csum((char*)&bsb,
((char*)&bsb.sb_csum2)-((char*)&bsb));
- lseek64(destfd[i], destoffsets[i] - 4096, 0);
- write(destfd[i], &bsb, 512);
+ rv = -1;
+ if ((unsigned long long)lseek64(destfd[i], destoffsets[i] - 4096, 0)
+ != destoffsets[i] - 4096)
+ break;
+ if (write(destfd[i], &bsb, 512) != 512)
+ break;
+ if (destoffsets[i] > 4096) {
+ if ((unsigned long long)lseek64(destfd[i], destoffsets[i]+stripes*chunk*odata, 0) !=
+ destoffsets[i]+stripes*chunk*odata)
+ break;
+ if (write(destfd[i], &bsb, 512) != 512)
+ break;
+ }
fsync(destfd[i]);
+ rv = 0;
}
- return 0;
+ return rv;
}
/* in 2.6.30, the value reported by sync_completed can be
* The various caller give appropriate values so that
* every works.
*/
-int wait_backup(struct mdinfo *sra,
+/* FIXME return value is often ignored */
+static int wait_backup(struct mdinfo *sra,
unsigned long long offset, /* per device */
unsigned long long blocks, /* per device */
unsigned long long blocks2, /* per device - hack */
int fd = sysfs_get_fd(sra, NULL, "sync_completed");
unsigned long long completed;
int i;
+ int rv;
if (fd < 0)
return -1;
sysfs_set_num(sra, NULL, "sync_max", offset + blocks + blocks2);
- if (offset == 0)
- sysfs_set_str(sra, NULL, "sync_action", "reshape");
- do {
+
+ if (sysfs_fd_get_ll(fd, &completed) < 0) {
+ close(fd);
+ return -1;
+ }
+ while (completed < offset + blocks) {
char action[20];
fd_set rfds;
FD_ZERO(&rfds);
action, 20) > 0 &&
strncmp(action, "reshape", 7) != 0)
break;
- } while (completed < offset + blocks);
+ }
close(fd);
if (part) {
bsb.arraystart = __cpu_to_le64(0);
bsb.length = __cpu_to_le64(0);
}
+ bsb.mtime = __cpu_to_le64(time(0));
+ rv = 0;
for (i = 0; i < dests; i++) {
bsb.devstart = __cpu_to_le64(destoffsets[i]/512);
bsb.sb_csum = bsb_csum((char*)&bsb, ((char*)&bsb.sb_csum)-((char*)&bsb));
if (memcmp(bsb.magic, "md_backup_data-2", 16) == 0)
bsb.sb_csum2 = bsb_csum((char*)&bsb,
((char*)&bsb.sb_csum2)-((char*)&bsb));
- lseek64(destfd[i], destoffsets[i]-4096, 0);
- write(destfd[i], &bsb, 512);
+ if ((unsigned long long)lseek64(destfd[i], destoffsets[i]-4096, 0) !=
+ destoffsets[i]-4096)
+ rv = -1;
+ if (rv == 0 &&
+ write(destfd[i], &bsb, 512) != 512)
+ rv = -1;
fsync(destfd[i]);
}
- return 0;
+ return rv;
}
static void fail(char *msg)
{
- write(2, msg, strlen(msg));
- write(2, "\n", 1);
- exit(1);
+ int rv;
+ rv = (write(2, msg, strlen(msg)) != (int)strlen(msg));
+ rv |= (write(2, "\n", 1) != 1);
+ exit(rv ? 1 : 2);
}
static char *abuf, *bbuf;
-static int abuflen;
+static unsigned long long abuflen;
static void validate(int afd, int bfd, unsigned long long offset)
{
/* check that the data in the backup against the array.
free(abuf);
free(bbuf);
abuflen = len;
- posix_memalign((void**)&abuf, 4096, abuflen);
- posix_memalign((void**)&bbuf, 4096, abuflen);
+ if (posix_memalign((void**)&abuf, 4096, abuflen) ||
+ posix_memalign((void**)&bbuf, 4096, abuflen)) {
+ abuflen = 0;
+ /* just stop validating on mem-alloc failure */
+ return;
+ }
}
lseek64(bfd, offset, 0);
- if (read(bfd, bbuf, len) != len) {
- printf("len %llu\n", len);
+ if ((unsigned long long)read(bfd, bbuf, len) != len) {
+ //printf("len %llu\n", len);
fail("read first backup failed");
}
lseek64(afd, __le64_to_cpu(bsb2.arraystart)*512, 0);
- if (read(afd, abuf, len) != len)
+ if ((unsigned long long)read(afd, abuf, len) != len)
fail("read first from array failed");
if (memcmp(bbuf, abuf, len) != 0) {
+ #if 0
int i;
printf("offset=%llu len=%llu\n",
- __le64_to_cpu(bsb2.arraystart)*512, len);
+ (unsigned long long)__le64_to_cpu(bsb2.arraystart)*512, len);
for (i=0; i<len; i++)
if (bbuf[i] != abuf[i]) {
printf("first diff byte %d\n", i);
break;
}
+ #endif
fail("data1 compare failed");
}
}
}
lseek64(bfd, offset+__le64_to_cpu(bsb2.devstart2)*512, 0);
- if (read(bfd, bbuf, len) != len)
+ if ((unsigned long long)read(bfd, bbuf, len) != len)
fail("read second backup failed");
lseek64(afd, __le64_to_cpu(bsb2.arraystart2)*512, 0);
- if (read(afd, abuf, len) != len)
+ if ((unsigned long long)read(afd, abuf, len) != len)
fail("read second from array failed");
if (memcmp(bbuf, abuf, len) != 0)
fail("data2 compare failed");
int dests, int *destfd, unsigned long long *destoffsets)
{
char *buf;
+ int degraded = 0;
- posix_memalign((void**)&buf, 4096, disks * chunk);
- sysfs_set_num(sra, NULL, "suspend_hi", 0);
- sysfs_set_num(sra, NULL, "suspend_lo", 0);
+ if (posix_memalign((void**)&buf, 4096, disks * chunk))
+ /* Don't start the 'reshape' */
+ return 0;
grow_backup(sra, 0, stripes,
fds, offsets, disks, chunk, level, layout,
dests, destfd, destoffsets,
- 0, buf);
+ 0, °raded, buf);
validate(afd, destfd[0], destoffsets[0]);
- wait_backup(sra, 0, stripes * chunk / 512, stripes * chunk / 512,
+ wait_backup(sra, 0, stripes * (chunk / 512), stripes * (chunk / 512),
dests, destfd, destoffsets,
0);
- sysfs_set_num(sra, NULL, "suspend_lo", (stripes * chunk/512) * data);
+ sysfs_set_num(sra, NULL, "suspend_lo", (stripes * (chunk/512)) * data);
free(buf);
/* FIXME this should probably be numeric */
sysfs_set_str(sra, NULL, "sync_max", "max");
char *buf;
unsigned long long start;
int rv;
+ int degraded = 0;
- posix_memalign((void**)&buf, 4096, disks * chunk);
- start = sra->component_size - stripes * chunk/512;
+ if (posix_memalign((void**)&buf, 4096, disks * chunk))
+ return 0;
+ start = sra->component_size - stripes * (chunk/512);
sysfs_set_num(sra, NULL, "sync_max", start);
- sysfs_set_str(sra, NULL, "sync_action", "reshape");
- sysfs_set_num(sra, NULL, "suspend_lo", 0);
- sysfs_set_num(sra, NULL, "suspend_hi", 0);
- rv = wait_backup(sra, 0, start - stripes * chunk/512, stripes * chunk/512,
+ rv = wait_backup(sra, 0, start - stripes * (chunk/512), stripes * (chunk/512),
dests, destfd, destoffsets, 0);
if (rv < 0)
return 0;
fds, offsets,
disks, chunk, level, layout,
dests, destfd, destoffsets,
- 0, buf);
+ 0, °raded, buf);
validate(afd, destfd[0], destoffsets[0]);
- wait_backup(sra, start, stripes*chunk/512, 0,
+ wait_backup(sra, start, stripes*(chunk/512), 0,
dests, destfd, destoffsets, 0);
- sysfs_set_num(sra, NULL, "suspend_lo", (stripes * chunk/512) * data);
+ sysfs_set_num(sra, NULL, "suspend_lo", (stripes * (chunk/512)) * data);
free(buf);
/* FIXME this should probably be numeric */
sysfs_set_str(sra, NULL, "sync_max", "max");
int part;
char *buf;
unsigned long long speed;
+ int degraded = 0;
- posix_memalign((void**)&buf, 4096, disks * chunk);
-
- sysfs_set_num(sra, NULL, "suspend_lo", 0);
- sysfs_set_num(sra, NULL, "suspend_hi", 0);
+ if (posix_memalign((void**)&buf, 4096, disks * chunk))
+ return 0;
sysfs_get_ll(sra, NULL, "sync_speed_min", &speed);
sysfs_set_num(sra, NULL, "sync_speed_min", 200000);
fds, offsets,
disks, chunk, level, layout,
dests, destfd, destoffsets,
- 0, buf);
- grow_backup(sra, (start + stripes) * chunk/512, stripes,
+ 0, °raded, buf);
+ grow_backup(sra, (start + stripes) * (chunk/512), stripes,
fds, offsets,
disks, chunk, level, layout,
dests, destfd, destoffsets,
- 1, buf);
+ 1, °raded, buf);
validate(afd, destfd[0], destoffsets[0]);
part = 0;
start += stripes * 2; /* where to read next */
size = sra->component_size / (chunk/512);
while (start < size) {
- if (wait_backup(sra, (start-stripes*2)*chunk/512,
- stripes*chunk/512, 0,
+ if (wait_backup(sra, (start-stripes*2)*(chunk/512),
+ stripes*(chunk/512), 0,
dests, destfd, destoffsets,
part) < 0)
return 0;
- sysfs_set_num(sra, NULL, "suspend_lo", start*chunk/512 * data);
+ sysfs_set_num(sra, NULL, "suspend_lo", start*(chunk/512) * data);
if (start + stripes > size)
tailstripes = (size - start);
- grow_backup(sra, start*chunk/512, tailstripes,
+ grow_backup(sra, start*(chunk/512), tailstripes,
fds, offsets,
disks, chunk, level, layout,
dests, destfd, destoffsets,
- part, buf);
+ part, °raded, buf);
start += stripes;
part = 1 - part;
validate(afd, destfd[0], destoffsets[0]);
}
- if (wait_backup(sra, (start-stripes*2) * chunk/512, stripes * chunk/512, 0,
+ if (wait_backup(sra, (start-stripes*2) * (chunk/512), stripes * (chunk/512), 0,
dests, destfd, destoffsets,
part) < 0)
return 0;
- sysfs_set_num(sra, NULL, "suspend_lo", ((start-stripes)*chunk/512) * data);
- wait_backup(sra, (start-stripes) * chunk/512, tailstripes * chunk/512, 0,
+ sysfs_set_num(sra, NULL, "suspend_lo", ((start-stripes)*(chunk/512)) * data);
+ wait_backup(sra, (start-stripes) * (chunk/512), tailstripes * (chunk/512), 0,
dests, destfd, destoffsets,
1-part);
- sysfs_set_num(sra, NULL, "suspend_lo", (size*chunk/512) * data);
+ sysfs_set_num(sra, NULL, "suspend_lo", (size*(chunk/512)) * data);
sysfs_set_num(sra, NULL, "sync_speed_min", speed);
free(buf);
return 1;
* write that data into the array and update the super blocks with
* the new reshape_progress
*/
-int Grow_restart(struct supertype *st, struct mdinfo *info, int *fdlist, int cnt, char *backup_file)
+int Grow_restart(struct supertype *st, struct mdinfo *info, int *fdlist, int cnt,
+ char *backup_file, int verbose)
{
int i, j;
int old_disks;
unsigned long long *offsets;
- unsigned long long nstripe, ostripe, last_block;
+ unsigned long long nstripe, ostripe;
int ndata, odata;
if (info->new_level != info->array.level)
old_disks = cnt;
for (i=old_disks-(backup_file?1:0); i<cnt; i++) {
struct mdinfo dinfo;
- char buf[4096];
int fd;
int bsbsize;
+ char *devname, namebuf[20];
/* This was a spare and may have some saved data on it.
* Load the superblock, find and load the
backup_file, strerror(errno));
continue;
}
+ devname = backup_file;
} else {
fd = fdlist[i];
if (fd < 0)
if (st->ss->load_super(st, fd, NULL))
continue;
- st->ss->getinfo_super(st, &dinfo);
+ st->ss->getinfo_super(st, &dinfo, NULL);
st->ss->free_super(st);
if (lseek64(fd,
(dinfo.data_offset + dinfo.component_size - 8) <<9,
- 0) < 0)
+ 0) < 0) {
+ fprintf(stderr, Name ": Cannot seek on device %d\n", i);
continue; /* Cannot seek */
+ }
+ sprintf(namebuf, "device-%d", i);
+ devname = namebuf;
}
- if (read(fd, &bsb, sizeof(bsb)) != sizeof(bsb))
+ if (read(fd, &bsb, sizeof(bsb)) != sizeof(bsb)) {
+ if (verbose)
+ fprintf(stderr, Name ": Cannot read from %s\n", devname);
continue; /* Cannot read */
+ }
if (memcmp(bsb.magic, "md_backup_data-1", 16) != 0 &&
- memcmp(bsb.magic, "md_backup_data-2", 16) != 0)
+ memcmp(bsb.magic, "md_backup_data-2", 16) != 0) {
+ if (verbose)
+ fprintf(stderr, Name ": No backup metadata on %s\n", devname);
continue;
- if (bsb.sb_csum != bsb_csum((char*)&bsb, ((char*)&bsb.sb_csum)-((char*)&bsb)))
+ }
+ if (bsb.sb_csum != bsb_csum((char*)&bsb, ((char*)&bsb.sb_csum)-((char*)&bsb))) {
+ if (verbose)
+ fprintf(stderr, Name ": Bad backup-metadata checksum on %s\n", devname);
continue; /* bad checksum */
+ }
if (memcmp(bsb.magic, "md_backup_data-2", 16) == 0 &&
- bsb.sb_csum2 != bsb_csum((char*)&bsb, ((char*)&bsb.sb_csum2)-((char*)&bsb)))
- if (memcmp(bsb.set_uuid,info->uuid, 16) != 0)
+ bsb.sb_csum2 != bsb_csum((char*)&bsb, ((char*)&bsb.sb_csum2)-((char*)&bsb))) {
+ if (verbose)
+ fprintf(stderr, Name ": Bad backup-metadata checksum2 on %s\n", devname);
+ continue; /* Bad second checksum */
+ }
+ if (memcmp(bsb.set_uuid,info->uuid, 16) != 0) {
+ if (verbose)
+ fprintf(stderr, Name ": Wrong uuid on backup-metadata on %s\n", devname);
continue; /* Wrong uuid */
+ }
- if (info->array.utime > __le64_to_cpu(bsb.mtime) + 3600 ||
- info->array.utime < __le64_to_cpu(bsb.mtime))
- continue; /* time stamp is too bad */
+ /* array utime and backup-mtime should be updated at much the same time, but it seems that
+ * sometimes they aren't... So allow considerable flexability in matching, and allow
+ * this test to be overridden by an environment variable.
+ */
+ if (info->array.utime > (int)__le64_to_cpu(bsb.mtime) + 2*60*60 ||
+ info->array.utime < (int)__le64_to_cpu(bsb.mtime) - 10*60) {
+ if (check_env("MDADM_GROW_ALLOW_OLD")) {
+ fprintf(stderr, Name ": accepting backup with timestamp %lu "
+ "for array with timestamp %lu\n",
+ (unsigned long)__le64_to_cpu(bsb.mtime),
+ (unsigned long)info->array.utime);
+ } else {
+ if (verbose)
+ fprintf(stderr, Name ": too-old timestamp on "
+ "backup-metadata on %s\n", devname);
+ continue; /* time stamp is too bad */
+ }
+ }
if (bsb.magic[15] == '1') {
if (info->delta_disks >= 0) {
/* reshape_progress is increasing */
if (__le64_to_cpu(bsb.arraystart) + __le64_to_cpu(bsb.length) <
- info->reshape_progress)
+ info->reshape_progress) {
+ nonew:
+ if (verbose)
+ fprintf(stderr, Name ": backup-metadata found on %s but is not needed\n", devname);
continue; /* No new data here */
+ }
} else {
/* reshape_progress is decreasing */
if (__le64_to_cpu(bsb.arraystart) >=
info->reshape_progress)
- continue; /* No new data here */
+ goto nonew; /* No new data here */
}
} else {
if (info->delta_disks >= 0) {
info->reshape_progress &&
__le64_to_cpu(bsb.arraystart2) + __le64_to_cpu(bsb.length2) <
info->reshape_progress)
- continue; /* No new data here */
+ goto nonew; /* No new data here */
} else {
/* reshape_progress is decreasing */
if (__le64_to_cpu(bsb.arraystart) >=
info->reshape_progress &&
__le64_to_cpu(bsb.arraystart2) >=
info->reshape_progress)
- continue; /* No new data here */
+ goto nonew; /* No new data here */
}
}
- if (lseek64(fd, __le64_to_cpu(bsb.devstart)*512, 0)< 0)
+ if (lseek64(fd, __le64_to_cpu(bsb.devstart)*512, 0)< 0) {
+ second_fail:
+ if (verbose)
+ fprintf(stderr, Name ": Failed to verify secondary backup-metadata block on %s\n",
+ devname);
continue; /* Cannot seek */
+ }
/* There should be a duplicate backup superblock 4k before here */
if (lseek64(fd, -4096, 1) < 0 ||
- read(fd, buf, 4096) != 4096)
- continue; /* Cannot find leading superblock */
+ read(fd, &bsb2, sizeof(bsb2)) != sizeof(bsb2))
+ goto second_fail; /* Cannot find leading superblock */
if (bsb.magic[15] == '1')
bsbsize = offsetof(struct mdp_backup_super, pad1);
else
bsbsize = offsetof(struct mdp_backup_super, pad);
- if (memcmp(buf, &bsb, bsbsize) != 0)
- continue; /* Cannot find leading superblock */
+ if (memcmp(&bsb2, &bsb, bsbsize) != 0)
+ goto second_fail; /* Cannot find leading superblock */
/* Now need the data offsets for all devices. */
offsets = malloc(sizeof(*offsets)*info->array.raid_disks);
if (st->ss->load_super(st, fdlist[j], NULL))
/* FIXME should be this be an error */
continue;
- st->ss->getinfo_super(st, &dinfo);
+ st->ss->getinfo_super(st, &dinfo, NULL);
st->ss->free_super(st);
- offsets[j] = dinfo.data_offset;
+ offsets[j] = dinfo.data_offset * 512;
}
printf(Name ": restoring critical section\n");
info->new_level,
info->new_layout,
fd, __le64_to_cpu(bsb.devstart)*512,
- __le64_to_cpu(bsb.arraystart),
+ __le64_to_cpu(bsb.arraystart)*512,
__le64_to_cpu(bsb.length)*512)) {
/* didn't succeed, so giveup */
+ if (verbose)
+ fprintf(stderr, Name ": Error restoring backup from %s\n",
+ devname);
return 1;
}
info->new_layout,
fd, __le64_to_cpu(bsb.devstart)*512 +
__le64_to_cpu(bsb.devstart2)*512,
- __le64_to_cpu(bsb.arraystart2),
+ __le64_to_cpu(bsb.arraystart2)*512,
__le64_to_cpu(bsb.length2)*512)) {
/* didn't succeed, so giveup */
+ if (verbose)
+ fprintf(stderr, Name ": Error restoring second backup from %s\n",
+ devname);
return 1;
}
if (fdlist[j] < 0) continue;
if (st->ss->load_super(st, fdlist[j], NULL))
continue;
- st->ss->getinfo_super(st, &dinfo);
+ st->ss->getinfo_super(st, &dinfo, NULL);
dinfo.reshape_progress = info->reshape_progress;
st->ss->update_super(st, &dinfo,
"_reshape_progress",
/* Didn't find any backup data, try to see if any
* was needed.
*/
- if (info->delta_disks == 0)
- /* Alway need backup data when size doesn't change */
- return 1;
- nstripe = ostripe = 0;
- last_block = 0;
- while (nstripe >= ostripe) {
- nstripe += info->new_chunk / 512;
- last_block = nstripe * ndata;
- ostripe = last_block / odata / (info->array.chunk_size/512) *
- (info->array.chunk_size/512);
+ if (info->delta_disks < 0) {
+ /* When shrinking, the critical section is at the end.
+ * So see if we are before the critical section.
+ */
+ unsigned long long first_block;
+ nstripe = ostripe = 0;
+ first_block = 0;
+ while (ostripe >= nstripe) {
+ ostripe += info->array.chunk_size / 512;
+ first_block = ostripe * odata;
+ nstripe = first_block / ndata / (info->new_chunk/512) *
+ (info->new_chunk/512);
+ }
+
+ if (info->reshape_progress >= first_block)
+ return 0;
}
+ if (info->delta_disks > 0) {
+ /* See if we are beyond the critical section. */
+ unsigned long long last_block;
+ nstripe = ostripe = 0;
+ last_block = 0;
+ while (nstripe >= ostripe) {
+ nstripe += info->new_chunk / 512;
+ last_block = nstripe * ndata;
+ ostripe = last_block / odata / (info->array.chunk_size/512) *
+ (info->array.chunk_size/512);
+ }
- if (info->reshape_progress >= last_block)
- return 0;
+ if (info->reshape_progress >= last_block)
+ return 0;
+ }
/* needed to recover critical section! */
+ if (verbose)
+ fprintf(stderr, Name ": Failed to find backup of critical section\n");
return 1;
}
int d;
struct mdinfo *sra, *sd;
int rv;
+ unsigned long cache;
int done = 0;
err = sysfs_set_str(info, NULL, "array_state", "readonly");
/* make sure reshape doesn't progress until we are ready */
sysfs_set_str(info, NULL, "sync_max", "0");
sysfs_set_str(info, NULL, "array_state", "active"); /* FIXME or clean */
-
+
+ sra = sysfs_read(-1, devname2devnum(info->sys_name),
+ GET_COMPONENT|GET_DEVS|GET_OFFSET|GET_STATE|
+ GET_CACHE);
+ if (!sra)
+ return 1;
+
/* ndisks is not growing, so raid_disks is old and +delta is new */
odisks = info->array.raid_disks;
ndisks = odisks + info->delta_disks;
ochunk = info->array.chunk_size;
nchunk = info->new_chunk;
-
- a = ochunk/512 * odata;
- b = nchunk/512 * ndata;
+ a = (ochunk/512) * odata;
+ b = (nchunk/512) * ndata;
/* Find GCD */
while (a != b) {
if (a < b)
a -= b;
}
/* LCM == product / GCD */
- blocks = ochunk/512 * nchunk/512 * odata * ndata / a;
+ blocks = (ochunk/512) * (nchunk/512) * odata * ndata / a;
if (ndata == odata)
- blocks *= 16;
+ while (blocks * 32 < sra->component_size &&
+ blocks < 16*1024*2)
+ blocks *= 2;
stripes = blocks / (info->array.chunk_size/512) / odata;
+ /* check that the internal stripe cache is
+ * large enough, or it won't work.
+ */
+ cache = (nchunk < ochunk) ? ochunk : nchunk;
+ cache = cache * 4 / 4096;
+ if (cache < blocks / 8 / odisks + 16)
+ /* Make it big enough to hold 'blocks' */
+ cache = blocks / 8 / odisks + 16;
+ if (sra->cache_size < cache)
+ sysfs_set_num(sra, NULL, "stripe_cache_size",
+ cache+1);
memset(&bsb, 0, 512);
memcpy(bsb.magic, "md_backup_data-1", 16);
bsb.devstart2 = blocks;
backup_fd = open(backup_file, O_RDWR|O_CREAT, S_IRUSR | S_IWUSR);
+ if (backup_fd < 0) {
+ fprintf(stderr, Name ": Cannot open backup file %s\n",
+ backup_file ?: "- no backup-file given");
+ return 1;
+ }
backup_list[0] = backup_fd;
backup_offsets[0] = 8 * 512;
fds = malloc(odisks * sizeof(fds[0]));
for (d=0; d<odisks; d++)
fds[d] = -1;
- sra = sysfs_read(-1, devname2devnum(info->sys_name),
- GET_COMPONENT|GET_DEVS|GET_OFFSET|GET_STATE|
- GET_CACHE);
-
for (sd = sra->devs; sd; sd = sd->next) {
if (sd->disk.state & (1<<MD_DISK_FAULTY))
continue;