close(aa->action_fd);
close(aa->info.state_fd);
close(aa->resync_start_fd);
+ close(aa->metadata_fd);
+ close(aa->sync_completed_fd);
}
static void free_aa(struct active_array *aa)
{
while (*update) {
struct metadata_update *this = *update;
+ void **space_list = this->space_list;
*update = this->next;
free(this->buf);
free(this->space);
+ while (space_list) {
+ void *space = space_list;
+ space_list = *space_list;
+ free(space);
+ }
free(this);
}
}
st->update_tail = NULL;
}
+/*
+ * Create and queue update structure about the removed disks.
+ * The update is prepared by super type handler and passed to the monitor
+ * thread.
+ */
+static void remove_disk_from_container(struct supertype *st, struct mdinfo *sd)
+{
+ struct metadata_update *update = NULL;
+ mdu_disk_info_t dk = {
+ .number = -1,
+ .major = sd->disk.major,
+ .minor = sd->disk.minor,
+ .raid_disk = -1,
+ .state = 0,
+ };
+ dprintf("%s: remove %d:%d from container\n",
+ __func__, sd->disk.major, sd->disk.minor);
+
+ st->update_tail = &update;
+ st->ss->remove_from_super(st, &dk);
+ /* FIXME this write_init_super shouldn't be here.
+ * We have it after add_to_super to write to new device,
+ * but with 'remove' we don't ant to write to that device!
+ */
+ st->ss->write_init_super(st);
+ queue_metadata_update(update);
+ st->update_tail = NULL;
+}
+
static void manage_container(struct mdstat_ent *mdstat,
struct supertype *container)
{
- /* The only thing of interest here is if a new device
- * has been added to the container. We add it to the
- * array ignoring any metadata on it.
+ /* Of interest here are:
+ * - if a new device has been added to the container, we
+ * add it to the array ignoring any metadata on it.
+ * - if a device has been removed from the container, we
+ * remove it from the device list and update the metadata.
* FIXME should we look for compatible metadata and take hints
* about spare assignment.... probably not.
*/
if (!found) {
cd = *cdp;
*cdp = (*cdp)->next;
+ remove_disk_from_container(container, cd);
free(cd);
} else
cdp = &(*cdp)->next;
* We do not need to look for device state changes here, that
* is dealt with by the monitor.
*
- * We just look for changes which suggest that a reshape is
- * being requested.
- * Unfortunately decreases in raid_disks don't show up in
- * mdstat until the reshape completes FIXME.
+ * If a reshape is being requested, monitor will have noticed
+ * that sync_action changed and will have set check_reshape.
+ * We just need to see if new devices have appeared. All metadata
+ * updates will already have been processed.
*
- * Actually, we also want to handle degraded arrays here by
+ * We also want to handle degraded arrays here by
* trying to find and assign a spare.
* We do that whenever the monitor tells us too.
*/
// FIXME
a->info.array.raid_disks = mdstat->raid_disks;
- a->info.array.chunk_size = mdstat->chunk_size;
// MORE
/* honor 'frozen' */
else
frozen = 1; /* can't read metadata_version assume the worst */
- if (a->check_degraded && !frozen) {
+ /* If sync_action is not 'idle' then don't try recovery now */
+ if (!frozen
+ && sysfs_get_str(&a->info, NULL, "sync_action", buf, sizeof(buf)) > 0
+ && strncmp(buf, "idle", 4) != 0)
+ frozen = 1;
+
+ if (mdstat->level) {
+ int level = map_name(pers, mdstat->level);
+ if (a->info.array.level != level && level >= 0) {
+ struct active_array *newa = duplicate_aa(a);
+ if (newa) {
+ newa->info.array.level = level;
+ replace_array(a->container, a, newa);
+ a = newa;
+ }
+ }
+ }
+
+ /* We don't check the array while any update is pending, as it
+ * might container a change (such as a spare assignment) which
+ * could affect our decisions.
+ */
+ if (a->check_degraded && !frozen &&
+ update_queue == NULL && update_queue_pending == NULL) {
struct metadata_update *updates = NULL;
struct mdinfo *newdev = NULL;
struct active_array *newa;
}
free_updates(&updates);
}
+
+ if (a->check_reshape) {
+ /* mdadm might have added some devices to the array.
+ * We want to disk_init_and_add any such device to a
+ * duplicate_aa and replace a with that.
+ * mdstat doesn't have enough info so we sysfs_read
+ * and look for new stuff.
+ */
+ struct mdinfo *info, *d, *d2, *newd;
+ unsigned long long array_size;
+ struct active_array *newa = NULL;
+ a->check_reshape = 0;
+ info = sysfs_read(-1, mdstat->devnum,
+ GET_DEVS|GET_OFFSET|GET_SIZE|GET_STATE);
+ if (!info)
+ goto out2;
+ for (d = info->devs; d; d = d->next) {
+ if (d->disk.raid_disk < 0)
+ continue;
+ for (d2 = a->info.devs; d2; d2 = d2->next)
+ if (d2->disk.raid_disk ==
+ d->disk.raid_disk)
+ break;
+ if (d2)
+ /* already have this one */
+ continue;
+ if (!newa) {
+ newa = duplicate_aa(a);
+ if (!newa)
+ break;
+ }
+ newd = malloc(sizeof(*newd));
+ if (!newd)
+ continue;
+ disk_init_and_add(newd, d, newa);
+ }
+ if (sysfs_get_ll(info, NULL, "array_size", &array_size) == 0
+ && a->info.custom_array_size > array_size*2) {
+ sysfs_set_num(info, NULL, "array_size",
+ a->info.custom_array_size/2);
+ }
+ out2:
+ sysfs_free(info);
+ if (newa)
+ replace_array(a->container, a, newa);
+ }
}
static int aa_ready(struct active_array *aa)
char *inst;
int i;
int failed = 0;
+ char buf[40];
/* check if array is ready to be monitored */
if (!mdstat->active)
dprintf("%s: inst: %d action: %d state: %d\n", __func__, atoi(inst),
new->action_fd, new->info.state_fd);
+ /* reshape_position is set by mdadm in sysfs
+ * read this information for new arrays only (empty victim)
+ */
+ if ((victim == NULL) &&
+ (sysfs_get_str(mdi, NULL, "sync_action", buf, 40) > 0) &&
+ (strncmp(buf, "reshape", 7) == 0)) {
+ if (sysfs_get_ll(mdi, NULL, "reshape_position",
+ &new->last_checkpoint) != 0)
+ new->last_checkpoint = 0;
+ else {
+ int data_disks = mdi->array.raid_disks;
+ if (mdi->array.level == 4 || mdi->array.level == 5)
+ data_disks--;
+ if (mdi->array.level == 6)
+ data_disks -= 2;
+
+ new->last_checkpoint /= data_disks;
+ }
+ dprintf("mdmon: New monitored array is under reshape.\n"
+ " Last checkpoint is: %llu\n",
+ new->last_checkpoint);
+ }
+
sysfs_free(mdi);
/* if everything checks out tell the metadata handler we want to
mu->buf = msg->buf;
msg->buf = NULL;
mu->space = NULL;
+ mu->space_list = NULL;
mu->next = NULL;
if (container->ss->prepare_update)
container->ss->prepare_update(container, mu);