else {
struct mdinfo *sra = sysfs_read(-1, st->devnum, GET_VERSION);
int err;
+ char buf[20];
if (!sra)
return -1;
+ /* Need to clear any 'read-auto' status */
+ if (sysfs_get_str(sra, NULL, "array_state", buf, 20) > 0 &&
+ strncmp(buf, "read-auto", 9) == 0)
+ sysfs_set_str(sra, NULL, "array_state", "clean");
+
err = sysfs_freeze_array(sra);
sysfs_free(sra);
return err;
static int reshape_array(char *container, int fd, char *devname,
struct supertype *st, struct mdinfo *info,
- int force, char *backup_file, int quiet, int forked,
+ int force, struct mddev_dev *devlist,
+ char *backup_file, int quiet, int forked,
int restart);
static int reshape_container(char *container, char *devname,
struct supertype *st,
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 force)
+ struct mddev_dev *devlist,
+ int assume_clean, int force)
{
/* Make some changes in the shape of an array.
* The kernel must support the change.
char container_buf[20];
int cfd = -1;
+ struct mddev_dev *dv;
+ int added_disks;
+
struct mdinfo info;
struct mdinfo *sra;
if (mdmon_running(container_dev))
st->update_tail = &st->updates;
- }
+ }
+ added_disks = 0;
+ for (dv = devlist; dv; dv = dv->next)
+ added_disks++;
if (raid_disks > array.raid_disks &&
- array.spare_disks < (raid_disks - array.raid_disks) &&
+ array.spare_disks +added_disks < (raid_disks - array.raid_disks) &&
!force) {
fprintf(stderr,
Name ": Need %d spare%s to avoid degraded array,"
" Use --force to over-ride this check.\n",
raid_disks - array.raid_disks,
raid_disks - array.raid_disks == 1 ? "" : "s",
- array.spare_disks);
+ array.spare_disks + added_disks);
return 1;
}
/* ========= set size =============== */
if (size >= 0 && (size == 0 || size != array.size)) {
- long long orig_size = array.size;
+ long long orig_size = get_component_size(fd)/2;
+ struct mdinfo *mdi;
+
+ if (orig_size == 0)
+ orig_size = array.size;
if (reshape_super(st, size, UnSet, UnSet, 0, 0, UnSet, NULL,
devname, !quiet)) {
goto release;
}
sync_metadata(st);
+
+ /* Update the size of each member device in case
+ * they have been resized. This will never reduce
+ * below the current used-size. The "size" attribute
+ * understand '0' to mean 'max'.
+ */
+ for (mdi = sra->devs; mdi; mdi = mdi->next)
+ sysfs_set_num(sra, mdi, "size", size);
+
array.size = size;
if (array.size != size) {
/* got truncated to 32bit, write to
rv = 1;
goto release;
}
+ if (assume_clean) {
+ /* This will fail on kernels newer than 2.6.40 unless
+ * a backport has been arranged.
+ */
+ if (sra == NULL ||
+ sysfs_set_str(sra, NULL, "resync_start", "none") < 0)
+ fprintf(stderr, Name ": --assume-clean not support with --grow on this kernel\n");
+ }
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 %lluK\n",
- devname, size);
+ if (!quiet) {
+ if (size == orig_size)
+ fprintf(stderr, Name ": component size of %s "
+ "unchanged at %lluK\n",
+ devname, size);
+ else
+ 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;
size = array.size;
}
+ /* See if there is anything else to do */
+ if ((level == UnSet || level == array.level) &&
+ (layout_str == NULL) &&
+ (chunksize == 0 || chunksize == array.chunk_size) &&
+ (raid_disks == 0 || raid_disks == array.raid_disks)) {
+ /* Nothing more to do */
+ if (!changed && !quiet)
+ fprintf(stderr, Name ": %s: no change requested\n",
+ devname);
+ goto release;
+ }
+
/* ========= check for Raid10/Raid1 -> Raid0 conversion ===============
* current implementation assumes that following conditions must be met:
* - RAID10:
ping_monitor(container);
}
+ memset(&info, 0, sizeof(info));
info.array = array;
sysfs_init(&info, fd, NoMdDev);
strcpy(info.text_version, sra->text_version);
}
}
- if (array.level == LEVEL_CONTAINER) {
+ if (array.level == LEVEL_FAULTY) {
+ if (level != UnSet && level != array.level) {
+ fprintf(stderr, Name ": cannot change level of Faulty device\n");
+ rv =1 ;
+ }
+ if (chunksize) {
+ fprintf(stderr, Name ": cannot set chunksize of Faulty device\n");
+ rv =1 ;
+ }
+ if (raid_disks && raid_disks != 1) {
+ fprintf(stderr, Name ": cannot set raid_disks of Faulty device\n");
+ rv =1 ;
+ }
+ if (layout_str) {
+ if (ioctl(fd, GET_ARRAY_INFO, &array) != 0) {
+ dprintf("Cannot get array information.\n");
+ goto release;
+ }
+ array.layout = info.new_layout;
+ if (ioctl(fd, SET_ARRAY_INFO, &array) != 0) {
+ fprintf(stderr, Name ": failed to set new layout\n");
+ rv = 1;
+ } else if (!quiet)
+ printf("layout for %s set to %d\n",
+ devname, array.layout);
+ }
+ } else if (array.level == LEVEL_CONTAINER) {
/* This change is to be applied to every array in the
* container. This is only needed when the metadata imposes
* restraints of the various arrays in the container.
}
sync_metadata(st);
rv = reshape_array(container, fd, devname, st, &info, force,
- backup_file, quiet, 0, 0);
+ devlist, backup_file, quiet, 0, 0);
frozen = 0;
}
release:
static int reshape_array(char *container, int fd, char *devname,
struct supertype *st, struct mdinfo *info,
- int force,
+ int force, struct mddev_dev *devlist,
char *backup_file, int quiet, int forked,
int restart)
{
struct mdu_array_info_s array;
char *c;
+ struct mddev_dev *dv;
+ int added_disks;
+
int *fdlist;
unsigned long long *offsets;
int d;
int done;
struct mdinfo *sra = NULL;
+ /* when reshaping a RAID0, the component_size might be zero.
+ * So try to fix that up.
+ */
+ if (ioctl(fd, GET_ARRAY_INFO, &array) != 0) {
+ dprintf("Cannot get array information.\n");
+ goto release;
+ }
+ if (array.level == 0 && info->component_size == 0) {
+ get_dev_size(fd, NULL, &array_size);
+ info->component_size = array_size / array.raid_disks;
+ }
+
if (info->reshape_active) {
int new_level = info->new_level;
info->new_level = UnSet;
+ info->array.raid_disks -= info->delta_disks;
msg = analyse_change(info, &reshape);
info->new_level = new_level;
+ info->array.raid_disks += info->delta_disks;
+ if (!restart)
+ /* Make sure the array isn't read-only */
+ ioctl(fd, RESTART_ARRAY_RW, 0);
} else
msg = analyse_change(info, &reshape);
if (msg) {
if (restart &&
(reshape.level != info->array.level ||
reshape.before.layout != info->array.layout ||
- reshape.before.data_disks + reshape.parity != info->array.raid_disks)) {
+ reshape.before.data_disks + reshape.parity
+ != info->array.raid_disks - info->delta_disks)) {
fprintf(stderr, Name ": reshape info is not in native format -"
" cannot continue.\n");
goto release;
}
- if (ioctl(fd, GET_ARRAY_INFO, &array) != 0) {
- dprintf("Cannot get array information.\n");
- goto release;
- }
if (restart) {
/* reshape already started. just skip to monitoring the reshape */
*/
sysfs_freeze_array(info);
/* Check we have enough spares to not be degraded */
+ added_disks = 0;
+ for (dv = devlist; dv ; dv=dv->next)
+ added_disks++;
spares_needed = max(reshape.before.data_disks,
reshape.after.data_disks)
+ reshape.parity - array.raid_disks;
if (!force &&
- info->new_level > 1 &&
- spares_needed > info->array.spare_disks) {
+ info->new_level > 1 && info->array.level > 1 &&
+ spares_needed > info->array.spare_disks + added_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",
- info->array.spare_disks);
+ info->array.spare_disks + added_disks);
goto release;
}
/* Check we have enough spares to not fail */
reshape.after.data_disks)
- array.raid_disks;
if ((info->new_level > 1 || info->new_level == 0) &&
- spares_needed > info->array.spare_disks) {
+ spares_needed > info->array.spare_disks +added_disks) {
fprintf(stderr,
Name ": Need %d spare%s to create working array,"
" and only have %d.\n",
spares_needed,
spares_needed == 1 ? "" : "s",
- info->array.spare_disks);
+ info->array.spare_disks + added_disks);
goto release;
}
if (info2) {
sysfs_init(info2, fd, st->devnum);
+ /* When increasing number of devices, we need to set
+ * new raid_disks before adding these, or they might
+ * be rejected.
+ */
+ if (reshape.backup_blocks &&
+ reshape.after.data_disks > reshape.before.data_disks)
+ subarray_set_num(container, info2, "raid_disks",
+ reshape.after.data_disks +
+ reshape.parity);
for (d = info2->devs; d; d = d->next) {
if (d->disk.state == 0 &&
d->disk.raid_disk >= 0) {
sysfs_free(info2);
}
}
+ /* We might have been given some devices to add to the
+ * array. Now that the array has been changed to the right
+ * level and frozen, we can safely add them.
+ */
+ if (devlist)
+ Manage_subdevs(devname, fd, devlist, !quiet,
+ 0,NULL);
if (reshape.backup_blocks == 0) {
/* No restriping needed, but we might need to impose
int quiet, int restart)
{
struct mdinfo *cc = NULL;
+ int rv = restart;
/* component_size is not meaningful for a container,
* so pass '-1' meaning 'no change'
* will take over the reshape.
*/
struct mdinfo *content;
- int rv;
int fd;
struct mdstat_ent *mdstat;
char *adev;
sysfs_init(content, fd, mdstat->devnum);
rv = reshape_array(container, fd, adev, st,
- content, force,
+ content, force, NULL,
backup_file, quiet, 1, restart);
close(fd);
restart = 0;
if (rv)
break;
}
- unfreeze(st);
+ if (!rv)
+ unfreeze(st);
sysfs_free(cc);
exit(0);
}
if (fd < 0)
goto check_progress;
- if (sysfs_fd_get_ll(fd, &completed) < 0) {
- close(fd);
+ if (sysfs_fd_get_ll(fd, &completed) < 0)
goto check_progress;
- }
+
while (completed < max_progress && completed < wait_point) {
/* Check that sync_action is still 'reshape' to avoid
* waiting forever on a dead array
FD_ZERO(&rfds);
FD_SET(fd, &rfds);
select(fd+1, NULL, NULL, &rfds, NULL);
- if (sysfs_fd_get_ll(fd, &completed) < 0) {
- close(fd);
+ if (sysfs_fd_get_ll(fd, &completed) < 0)
goto check_progress;
- }
}
/* Some kernels reset 'sync_completed' to zero,
* we need to have real point we are in md
/* if we couldn't read a number from sync_completed, then
* either the reshape did complete, or it aborted.
* We can tell which by checking for 'none' in reshape_position.
+ * If it did abort, then it might immediately restart if it
+ * it was just a device failure that leaves us degraded but
+ * functioning.
*/
strcpy(buf, "hi");
if (sysfs_get_str(info, NULL, "reshape_position", buf, sizeof(buf)) < 0
- || strncmp(buf, "none", 4) != 0)
- return -2; /* abort */
- else {
+ || strncmp(buf, "none", 4) != 0) {
+ /* The abort might only be temporary. Wait up to 10
+ * seconds for fd to contain a valid number again.
+ */
+ struct timeval tv;
+ int rv = -2;
+ tv.tv_sec = 10;
+ tv.tv_usec = 0;
+ while (fd >= 0 && rv < 0) {
+ fd_set rfds;
+ FD_ZERO(&rfds);
+ FD_SET(fd, &rfds);
+ if (select(fd+1, NULL, NULL, &rfds, &tv) != 1)
+ break;
+ if (sysfs_fd_get_ll(fd, &completed) >= 0)
+ /* all good again */
+ rv = 1;
+ }
+ if (fd >= 0)
+ close(fd);
+ return rv; /* abort */
+ } else {
/* Maybe racing with array shutdown - check state */
+ if (fd >= 0)
+ close(fd);
if (sysfs_get_str(info, NULL, "array_state", buf, sizeof(buf)) < 0
|| strncmp(buf, "inactive", 8) == 0
|| strncmp(buf, "clear",5) == 0)
continue;
ok = st->ss->load_super(st, devfd, NULL);
close(devfd);
- if (ok >= 0)
+ if (ok == 0)
break;
}
if (!sd) {
/* Clear any backup region that is before 'here' */
if (increasing) {
- if (reshape_completed >= (__le64_to_cpu(bsb.arraystart) +
+ if (__le64_to_cpu(bsb.length) > 0 &&
+ reshape_completed >= (__le64_to_cpu(bsb.arraystart) +
__le64_to_cpu(bsb.length)))
forget_backup(dests, destfd,
destoffsets, 0);
- if (reshape_completed >= (__le64_to_cpu(bsb.arraystart2) +
+ if (__le64_to_cpu(bsb.length2) > 0 &&
+ reshape_completed >= (__le64_to_cpu(bsb.arraystart2) +
__le64_to_cpu(bsb.length2)))
forget_backup(dests, destfd,
destoffsets, 1);
} else {
- if (reshape_completed <= (__le64_to_cpu(bsb.arraystart)))
+ if (__le64_to_cpu(bsb.length) > 0 &&
+ reshape_completed <= (__le64_to_cpu(bsb.arraystart)))
forget_backup(dests, destfd,
destoffsets, 0);
- if (reshape_completed <= (__le64_to_cpu(bsb.arraystart2)))
+ if (__le64_to_cpu(bsb.length2) > 0 &&
+ reshape_completed <= (__le64_to_cpu(bsb.arraystart2)))
forget_backup(dests, destfd,
destoffsets, 1);
}
done = 1;
break;
}
+ if (rv == 0 && increasing && !st->ss->external) {
+ /* No longer need to monitor this reshape */
+ done = 1;
+ break;
+ }
while (rv) {
unsigned long long offset;
unsigned long long nstripe, ostripe;
int ndata, odata;
- if (info->new_level != info->array.level)
- return 1; /* Cannot handle level changes (they are instantaneous) */
-
odata = info->array.raid_disks - info->delta_disks - 1;
if (info->array.level == 6) odata--; /* number of data disks */
ndata = info->array.raid_disks - 1;
if (!mdmon_running(st->container_dev))
start_mdmon(st->container_dev);
- ping_monitor(devnum2devname(st->container_dev));
+ ping_monitor_by_id(st->container_dev);
if (info->reshape_active == 2) {
}
}
return reshape_array(container, mdfd, "array", st, info, 1,
- backup_file, 0, 0, 1);
+ NULL, backup_file, 0, 0, 1);
}
projects / thirdparty / mdadm.git / blobdiff