X-Git-Url: http://git.ipfire.org/?a=blobdiff_plain;f=managemon.c;h=15b6e692efcbcf78a54afc9708db5d528ba742b6;hb=4dd968cc546aadc168efa0c6dd96d789a7b0a08f;hp=c9b054ff4a0605953a28792a463e959a067c24f1;hpb=a54d52625a7b06dde6e00a91398c517ecf205fdd;p=thirdparty%2Fmdadm.git diff --git a/managemon.c b/managemon.c index c9b054ff..15b6e692 100644 --- a/managemon.c +++ b/managemon.c @@ -1,8 +1,8 @@ /* * mdmon - monitor external metadata arrays * - * Copyright (C) 2007-2008 Neil Brown - * Copyright (C) 2007-2008 Intel Corporation + * Copyright (C) 2007-2009 Neil Brown + * Copyright (C) 2007-2009 Intel Corporation * * This program is free software; you can redistribute it and/or modify it * under the terms and conditions of the GNU General Public License, @@ -112,12 +112,16 @@ static void close_aa(struct active_array *aa) { struct mdinfo *d; - for (d = aa->info.devs; d; d = d->next) + for (d = aa->info.devs; d; d = d->next) { + close(d->recovery_fd); close(d->state_fd); + } 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) @@ -209,16 +213,28 @@ struct metadata_update *update_queue = NULL; struct metadata_update *update_queue_handled = NULL; struct metadata_update *update_queue_pending = NULL; -void check_update_queue(struct supertype *container) +static void free_updates(struct metadata_update **update) { - while (update_queue_handled) { - struct metadata_update *this = update_queue_handled; - update_queue_handled = this->next; + while (*update) { + struct metadata_update *this = *update; + void **space_list = this->space_list; + + *update = this->next; free(this->buf); - if (this->space) - free(this->space); + free(this->space); + while (space_list) { + void *space = space_list; + space_list = *space_list; + free(space); + } free(this); } +} + +void check_update_queue(struct supertype *container) +{ + free_updates(&update_queue_handled); + if (update_queue == NULL && update_queue_pending) { update_queue = update_queue_pending; @@ -241,7 +257,9 @@ static void add_disk_to_container(struct supertype *st, struct mdinfo *sd) { int dfd; char nm[20]; + struct supertype *st2; struct metadata_update *update = NULL; + struct mdinfo info; mdu_disk_info_t dk = { .number = -1, .major = sd->disk.major, @@ -253,11 +271,33 @@ static void add_disk_to_container(struct supertype *st, struct mdinfo *sd) dprintf("%s: add %d:%d to container\n", __func__, sd->disk.major, sd->disk.minor); + sd->next = st->devs; + st->devs = sd; + sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); dfd = dev_open(nm, O_RDWR); if (dfd < 0) return; + /* Check the metadata and see if it is already part of this + * array + */ + st2 = dup_super(st); + if (st2->ss->load_super(st2, dfd, NULL) == 0) { + st2->ss->getinfo_super(st, &info, NULL); + if (st->ss->compare_super(st, st2) == 0 && + info.disk.raid_disk >= 0) { + /* Looks like a good member of array. + * Just accept it. + * mdadm will incorporate any parts into + * active arrays. + */ + st2->ss->free_super(st2); + return; + } + } + st2->ss->free_super(st2); + st->update_tail = &update; st->ss->add_to_super(st, &dk, dfd, NULL); st->ss->write_init_super(st); @@ -265,12 +305,43 @@ static void add_disk_to_container(struct supertype *st, struct mdinfo *sd) 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. */ @@ -302,6 +373,7 @@ static void manage_container(struct mdstat_ent *mdstat, if (!found) { cd = *cdp; *cdp = (*cdp)->next; + remove_disk_from_container(container, cd); free(cd); } else cdp = &(*cdp)->next; @@ -313,14 +385,39 @@ static void manage_container(struct mdstat_ent *mdstat, if (di->disk.major == cd->disk.major && di->disk.minor == cd->disk.minor) break; - if (!cd) - add_disk_to_container(container, di); + if (!cd) { + struct mdinfo *newd = malloc(sizeof(*newd)); + + if (!newd) { + container->devcnt = -1; + continue; + } + *newd = *di; + add_disk_to_container(container, newd); + } } sysfs_free(mdi); container->devcnt = mdstat->devcnt; } } +static int disk_init_and_add(struct mdinfo *disk, struct mdinfo *clone, + struct active_array *aa) +{ + if (!disk || !clone) + return -1; + + *disk = *clone; + disk->recovery_fd = sysfs_open(aa->devnum, disk->sys_name, "recovery_start"); + disk->state_fd = sysfs_open(aa->devnum, disk->sys_name, "state"); + disk->prev_state = read_dev_state(disk->state_fd); + disk->curr_state = disk->prev_state; + disk->next = aa->info.devs; + aa->info.devs = disk; + + return 0; +} + static void manage_member(struct mdstat_ent *mdstat, struct active_array *a) { @@ -328,24 +425,51 @@ static void manage_member(struct mdstat_ent *mdstat, * 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. */ + char buf[64]; + int frozen; + // FIXME a->info.array.raid_disks = mdstat->raid_disks; - a->info.array.chunk_size = mdstat->chunk_size; // MORE - if (a->check_degraded) { + /* honor 'frozen' */ + if (sysfs_get_str(&a->info, NULL, "metadata_version", buf, sizeof(buf)) > 0) + frozen = buf[9] == '-'; + else + frozen = 1; /* can't read metadata_version assume the worst */ + + /* 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; + } + } + } + + if (a->check_degraded && !frozen) { struct metadata_update *updates = NULL; - struct mdinfo *newdev; + struct mdinfo *newdev = NULL; struct active_array *newa; + struct mdinfo *d; a->check_degraded = 0; @@ -353,34 +477,85 @@ static void manage_member(struct mdstat_ent *mdstat, * to check. */ newdev = a->container->ss->activate_spare(a, &updates); - if (newdev) { - struct mdinfo *d; - /* Cool, we can add a device or several. */ - newa = duplicate_aa(a); - /* suspend recovery - maybe not needed */ - - /* Add device to array and set offset/size/slot. - * and open files for each newdev */ - for (d = newdev; d ; d = d->next) { - struct mdinfo *newd; - if (sysfs_add_disk(&newa->info, d) < 0) - continue; - newd = malloc(sizeof(*newd)); - *newd = *d; - newd->next = newa->info.devs; - newa->info.devs = newd; - - newd->state_fd = sysfs_open(a->devnum, - newd->sys_name, - "state"); - newd->prev_state - = read_dev_state(newd->state_fd); - newd->curr_state = newd->prev_state; + if (!newdev) + return; + + newa = duplicate_aa(a); + if (!newa) + goto out; + /* Cool, we can add a device or several. */ + + /* Add device to array and set offset/size/slot. + * and open files for each newdev */ + for (d = newdev; d ; d = d->next) { + struct mdinfo *newd; + + newd = malloc(sizeof(*newd)); + if (!newd) + continue; + if (sysfs_add_disk(&newa->info, d, 0) < 0) { + free(newd); + continue; } - queue_metadata_update(updates); - replace_array(a->container, a, newa); - sysfs_set_str(&a->info, NULL, "sync_action", "recover"); + disk_init_and_add(newd, d, newa); + } + queue_metadata_update(updates); + updates = NULL; + replace_array(a->container, a, newa); + sysfs_set_str(&a->info, NULL, "sync_action", "recover"); + out: + while (newdev) { + d = newdev->next; + free(newdev); + newdev = d; + } + 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); } } @@ -420,6 +595,7 @@ static void manage_new(struct mdstat_ent *mdstat, char *inst; int i; int failed = 0; + char buf[40]; /* check if array is ready to be monitored */ if (!mdstat->active) @@ -448,7 +624,7 @@ static void manage_new(struct mdstat_ent *mdstat, new->container = container; - inst = &mdstat->metadata_version[10+strlen(container->devname)+1]; + inst = to_subarray(mdstat, container->devname); new->info.array = mdi->array; new->info.component_size = mdi->component_size; @@ -460,37 +636,50 @@ static void manage_new(struct mdstat_ent *mdstat, if (i == di->disk.raid_disk) break; - if (di) { - memcpy(newd, di, sizeof(*newd)); - - newd->state_fd = sysfs_open(new->devnum, - newd->sys_name, - "state"); + if (disk_init_and_add(newd, di, new) != 0) { + if (newd) + free(newd); - newd->prev_state = read_dev_state(newd->state_fd); - newd->curr_state = newd->prev_state; - } else if (failed + 1 > new->info.array.failed_disks) { - /* we cannot properly monitor without all working disks */ - new->container = NULL; - break; - } else { failed++; - free(newd); - continue; + if (failed > new->info.array.failed_disks) { + /* we cannot properly monitor without all working disks */ + new->container = NULL; + break; + } } - sprintf(newd->sys_name, "rd%d", i); - newd->next = new->info.devs; - new->info.devs = newd; } new->action_fd = sysfs_open(new->devnum, NULL, "sync_action"); new->info.state_fd = sysfs_open(new->devnum, NULL, "array_state"); new->resync_start_fd = sysfs_open(new->devnum, NULL, "resync_start"); new->metadata_fd = sysfs_open(new->devnum, NULL, "metadata_version"); - get_resync_start(new); + new->sync_completed_fd = sysfs_open(new->devnum, NULL, "sync_completed"); 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 @@ -574,6 +763,7 @@ static void handle_message(struct supertype *container, struct metadata_update * 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); @@ -603,7 +793,13 @@ void read_sock(struct supertype *container) /* read and validate the message */ if (receive_message(fd, &msg, tmo) == 0) { handle_message(container, &msg); - if (ack(fd, tmo) < 0) + if (msg.len == 0) { + /* ping reply with version */ + msg.buf = Version; + msg.len = strlen(Version) + 1; + if (send_message(fd, &msg, tmo) < 0) + terminate = 1; + } else if (ack(fd, tmo) < 0) terminate = 1; } else terminate = 1; @@ -619,14 +815,10 @@ void do_manager(struct supertype *container) { struct mdstat_ent *mdstat; sigset_t set; - int proc_fd; sigprocmask(SIG_UNBLOCK, NULL, &set); sigdelset(&set, SIGUSR1); - sigdelset(&set, SIGHUP); - sigdelset(&set, SIGALRM); sigdelset(&set, SIGTERM); - proc_fd = open("/proc/mounts", O_RDONLY); do { @@ -644,15 +836,6 @@ void do_manager(struct supertype *container) read_sock(container); - if (container->sock < 0 || socket_hup_requested) { - close(container->sock); - container->sock = make_control_sock(container->devname); - make_pidfile(container->devname, 0); - socket_hup_requested = 0; - } - if (container->sock < 0) - alarm(30); - free_mdstat(mdstat); } remove_old(); @@ -664,12 +847,9 @@ void do_manager(struct supertype *container) if (sigterm) wakeup_monitor(); - if (update_queue == NULL) { - if (container->sock < 0) - mdstat_wait_fd(proc_fd, &set); - else - mdstat_wait_fd(container->sock, &set); - } else + if (update_queue == NULL) + mdstat_wait_fd(container->sock, &set); + else /* If an update is happening, just wait for signal */ pselect(0, NULL, NULL, NULL, NULL, &set); } while(1);