X-Git-Url: http://git.ipfire.org/?a=blobdiff_plain;f=Manage.c;h=6ccd275919ca2fa7fd41b1cacdfa6854dd421393;hb=e2efe9e7bc73307f74a4c2e2197d6d4498dd46f0;hp=1b2b75afbd99b6717c16a61d08d554506e8e8992;hpb=11b391ece9fa284a151362537af093aa44883696;p=thirdparty%2Fmdadm.git diff --git a/Manage.c b/Manage.c index 1b2b75af..6ccd2759 100644 --- a/Manage.c +++ b/Manage.c @@ -1,7 +1,7 @@ /* * mdadm - manage Linux "md" devices aka RAID arrays. * - * Copyright (C) 2001-2009 Neil Brown + * Copyright (C) 2001-2013 Neil Brown * * * This program is free software; you can redistribute it and/or modify @@ -27,9 +27,9 @@ #include "md_p.h" #include -#define REGISTER_DEV _IO (MD_MAJOR, 1) -#define START_MD _IO (MD_MAJOR, 2) -#define STOP_MD _IO (MD_MAJOR, 3) +#define REGISTER_DEV _IO (MD_MAJOR, 1) +#define START_MD _IO (MD_MAJOR, 2) +#define STOP_MD _IO (MD_MAJOR, 3) int Manage_ro(char *devname, int fd, int readonly) { @@ -44,16 +44,17 @@ int Manage_ro(char *devname, int fd, int readonly) #ifndef MDASSEMBLE struct mdinfo *mdi; #endif + int rv = 0; if (md_get_version(fd) < 9000) { - fprintf(stderr, Name ": need md driver version 0.90.0 or later\n"); + pr_err("need md driver version 0.90.0 or later\n"); return 1; } #ifndef MDASSEMBLE - /* If this is an externally-manage array, we need to modify the + /* If this is an externally-managed array, we need to modify the * metadata_version so that mdmon doesn't undo our change. */ - mdi = sysfs_read(fd, -1, GET_LEVEL|GET_VERSION); + mdi = sysfs_read(fd, NULL, GET_LEVEL|GET_VERSION); if (mdi && mdi->array.major_version == -1 && is_subarray(mdi->text_version)) { @@ -70,12 +71,13 @@ int Manage_ro(char *devname, int fd, int readonly) rv = sysfs_set_str(mdi, NULL, "array_state", "readonly"); if (rv < 0) { - fprintf(stderr, Name ": failed to set readonly for %s: %s\n", + pr_err("failed to set readonly for %s: %s\n", devname, strerror(errno)); vers[9] = mdi->text_version[0]; sysfs_set_str(mdi, NULL, "metadata_version", vers); - return 1; + rv = 1; + goto out; } } else { char *cp; @@ -84,45 +86,53 @@ int Manage_ro(char *devname, int fd, int readonly) sysfs_set_str(mdi, NULL, "metadata_version", vers); cp = strchr(vers+10, '/'); - if (*cp) + if (cp) *cp = 0; ping_monitor(vers+10); if (mdi->array.level <= 0) sysfs_set_str(mdi, NULL, "array_state", "active"); } - return 0; + goto out; } #endif if (ioctl(fd, GET_ARRAY_INFO, &array)) { - fprintf(stderr, Name ": %s does not appear to be active.\n", + pr_err("%s does not appear to be active.\n", devname); - return 1; + rv = 1; + goto out; } - if (readonly>0) { + if (readonly > 0) { if (ioctl(fd, STOP_ARRAY_RO, NULL)) { - fprintf(stderr, Name ": failed to set readonly for %s: %s\n", + pr_err("failed to set readonly for %s: %s\n", devname, strerror(errno)); - return 1; + rv = 1; + goto out; } } else if (readonly < 0) { if (ioctl(fd, RESTART_ARRAY_RW, NULL)) { - fprintf(stderr, Name ": failed to set writable for %s: %s\n", + pr_err("failed to set writable for %s: %s\n", devname, strerror(errno)); - return 1; + rv = 1; + goto out; } } - return 0; +out: +#ifndef MDASSEMBLE + if (mdi) + sysfs_free(mdi); +#endif + return rv; } #ifndef MDASSEMBLE -static void remove_devices(int devnum, char *path) +static void remove_devices(char *devnm, char *path) { - /* + /* * Remove names at 'path' - possibly with * partition suffixes - which link to the 'standard' - * name for devnum. These were probably created + * name for devnm. These were probably created * by mdadm when the array was assembled. */ char base[40]; @@ -136,16 +146,13 @@ static void remove_devices(int devnum, char *path) if (!path) return; - if (devnum >= 0) - sprintf(base, "/dev/md%d", devnum); - else - sprintf(base, "/dev/md_d%d", -1-devnum); + sprintf(base, "/dev/%s", devnm); be = base + strlen(base); - path2 = malloc(strlen(path)+20); + path2 = xmalloc(strlen(path)+20); strcpy(path2, path); pe = path2 + strlen(path2); - + for (part = 0; part < 16; part++) { if (part) { sprintf(be, "p%d", part); @@ -156,260 +163,1125 @@ static void remove_devices(int devnum, char *path) sprintf(pe, "%d", part); } n = readlink(path2, link, sizeof(link)); - if (n && (int)strlen(base) == n && + if (n > 0 && (int)strlen(base) == n && strncmp(link, base, n) == 0) unlink(path2); } free(path2); } - -int Manage_runstop(char *devname, int fd, int runstop, int quiet) +int Manage_run(char *devname, int fd, struct context *c) { - /* Run or stop the array. array must already be configured - * required >= 0.90.0 - * Only print failure messages if quiet == 0; - * quiet > 0 means really be quiet - * quiet < 0 means we will try again if it fails. + /* Run the array. Array must already be configured + * Requires >= 0.90.0 */ - mdu_param_t param; /* unused */ - - if (runstop == -1 && md_get_version(fd) < 9000) { - if (ioctl(fd, STOP_MD, 0)) { - if (quiet == 0) fprintf(stderr, - Name ": stopping device %s " - "failed: %s\n", - devname, strerror(errno)); - return 1; - } + char nm[32], *nmp; + + if (md_get_version(fd) < 9000) { + pr_err("need md driver version 0.90.0 or later\n"); + return 1; + } + nmp = fd2devnm(fd); + if (!nmp) { + pr_err("Cannot find %s in sysfs!!\n", devname); + return 1; } + strcpy(nm, nmp); + return IncrementalScan(c, nm); +} + +int Manage_stop(char *devname, int fd, int verbose, int will_retry) +{ + /* Stop the array. Array must already be configured + * 'will_retry' means that error messages are not wanted. + */ + int rv = 0; + struct map_ent *map = NULL; + struct mdinfo *mdi; + char devnm[32]; + char container[32]; + int err; + int count; + char buf[32]; + unsigned long long rd1, rd2; + + if (will_retry && verbose == 0) + verbose = -1; if (md_get_version(fd) < 9000) { - fprintf(stderr, Name ": need md driver version 0.90.0 or later\n"); + if (ioctl(fd, STOP_MD, 0) == 0) + return 0; + pr_err("stopping device %s " + "failed: %s\n", + devname, strerror(errno)); return 1; } - /* - if (ioctl(fd, GET_ARRAY_INFO, &array)) { - fprintf(stderr, Name ": %s does not appear to be active.\n", - devname); + + /* If this is an mdmon managed array, just write 'inactive' + * to the array state and let mdmon clear up. + */ + strcpy(devnm, fd2devnm(fd)); + /* Get EXCL access first. If this fails, then attempting + * to stop is probably a bad idea. + */ + mdi = sysfs_read(fd, NULL, GET_LEVEL|GET_COMPONENT|GET_VERSION); + if (mdi && is_subarray(mdi->text_version)) { + char *sl; + strncpy(container, mdi->text_version+1, sizeof(container)); + container[sizeof(container)-1] = 0; + sl = strchr(container, '/'); + if (sl) + *sl = 0; + } else + container[0] = 0; + close(fd); + count = 5; + while (((fd = ((devnm[0] == '/') + ?open(devname, O_RDONLY|O_EXCL) + :open_dev_flags(devnm, O_RDONLY|O_EXCL))) < 0 + || strcmp(fd2devnm(fd), devnm) != 0) + && container[0] + && mdmon_running(container) + && count) { + if (fd >= 0) + close(fd); + flush_mdmon(container); + count--; + } + if (fd < 0 || strcmp(fd2devnm(fd), devnm) != 0) { + if (fd >= 0) + close(fd); + if (verbose >= 0) + pr_err("Cannot get exclusive access to %s:" + "Perhaps a running " + "process, mounted filesystem " + "or active volume group?\n", + devname); return 1; } - */ - if (runstop>0) { - if (ioctl(fd, RUN_ARRAY, ¶m)) { - fprintf(stderr, Name ": failed to run array %s: %s\n", - devname, strerror(errno)); - return 1; - } - if (quiet <= 0) - fprintf(stderr, Name ": started %s\n", devname); - } else if (runstop < 0){ - struct map_ent *map = NULL; - struct stat stb; - struct mdinfo *mdi; - int devnum; + if (mdi && + mdi->array.level > 0 && + is_subarray(mdi->text_version)) { int err; - int count; - /* If this is an mdmon managed array, just write 'inactive' - * to the array state and let mdmon clear up. + /* This is mdmon managed. */ + close(fd); + + /* As we have an O_EXCL open, any use of the device + * which blocks STOP_ARRAY is probably a transient use, + * so it is reasonable to retry for a while - 5 seconds. */ - devnum = fd2devnum(fd); - /* Get EXCL access first. If this fails, then attempting - * to stop is probably a bad idea. + count = 25; + while (count && + (err = sysfs_set_str(mdi, NULL, + "array_state", + "inactive")) < 0 + && errno == EBUSY) { + usleep(200000); + count--; + } + if (err) { + if (verbose >= 0) + pr_err("failed to stop array %s: %s\n", + devname, strerror(errno)); + rv = 1; + goto out; + } + + /* Give monitor a chance to act */ + ping_monitor(mdi->text_version); + + fd = open_dev_excl(devnm); + if (fd < 0) { + if (verbose >= 0) + pr_err("failed to completely stop %s" + ": Device is busy\n", + devname); + rv = 1; + goto out; + } + } else if (mdi && + mdi->array.major_version == -1 && + mdi->array.minor_version == -2 && + !is_subarray(mdi->text_version)) { + struct mdstat_ent *mds, *m; + /* container, possibly mdmon-managed. + * Make sure mdmon isn't opening it, which + * would interfere with the 'stop' */ - close(fd); - fd = open(devname, O_RDONLY|O_EXCL); - if (fd < 0 || fd2devnum(fd) != devnum) { - if (fd >= 0) - close(fd); - fprintf(stderr, - Name ": Cannot get exclusive access to %s:" - "Perhaps a running " - "process, mounted filesystem " - "or active volume group?\n", - devname); - return 1; + ping_monitor(mdi->sys_name); + + /* now check that there are no existing arrays + * which are members of this array + */ + mds = mdstat_read(0, 0); + for (m = mds; m; m = m->next) + if (m->metadata_version && + strncmp(m->metadata_version, "external:", 9)==0 && + metadata_container_matches(m->metadata_version+9, + devnm)) { + if (verbose >= 0) + pr_err("Cannot stop container %s: " + "member %s still active\n", + devname, m->dev); + free_mdstat(mds); + rv = 1; + goto out; + } + } + + /* If the array is undergoing a reshape which changes the number + * of devices, then it would be nice to stop it at a point where + * it has completed a full number of stripes in both old and + * new layouts as this will allow the reshape to be reverted. + * So if 'sync_action' is "reshape" and 'raid_disks' shows two + * different numbers, then + * - freeze reshape + * - set sync_max to next multiple of both data_disks and + * chunk sizes (or next but one) + * - unfreeze reshape + * - wait on 'sync_completed' for that point to be reached. + */ + if (mdi && (mdi->array.level >= 4 && mdi->array.level <= 6) && + sysfs_attribute_available(mdi, NULL, "sync_action") && + sysfs_attribute_available(mdi, NULL, "reshape_direction") && + sysfs_get_str(mdi, NULL, "sync_action", buf, 20) > 0 && + strcmp(buf, "reshape\n") == 0 && + sysfs_get_two(mdi, NULL, "raid_disks", &rd1, &rd2) == 2 && + sysfs_set_str(mdi, NULL, "sync_action", "frozen") == 0) { + /* Array is frozen */ + unsigned long long position, curr; + unsigned long long chunk1, chunk2; + unsigned long long rddiv, chunkdiv; + unsigned long long sectors; + unsigned long long sync_max, old_sync_max; + unsigned long long completed; + int backwards = 0; + int delay; + int scfd; + + rd1 -= mdi->array.level == 6 ? 2 : 1; + rd2 -= mdi->array.level == 6 ? 2 : 1; + sysfs_get_str(mdi, NULL, "reshape_direction", buf, sizeof(buf)); + if (strncmp(buf, "back", 4) == 0) + backwards = 1; + sysfs_get_ll(mdi, NULL, "reshape_position", &position); + sysfs_get_two(mdi, NULL, "chunk_size", &chunk1, &chunk2); + chunk1 /= 512; + chunk2 /= 512; + rddiv = GCD(rd1, rd2); + chunkdiv = GCD(chunk1, chunk2); + sectors = (chunk1/chunkdiv) * chunk2 * (rd1/rddiv) * rd2; + + if (backwards) { + /* Need to subtract 'reshape_position' from + * array size to get equivalent of sync_max. + * Size calculation based on raid5_size in kernel. + */ + unsigned long long size = mdi->component_size; + size &= ~(chunk1-1); + size &= ~(chunk2-1); + /* rd1 must be smaller */ + position = (position / sectors - 1) * sectors; + sync_max = size - position/rd1; + } else { + position = (position / sectors + 2) * sectors; + sync_max = position/rd1; } - mdi = sysfs_read(fd, -1, GET_LEVEL|GET_VERSION); - if (mdi && - mdi->array.level > 0 && - is_subarray(mdi->text_version)) { - int err; - /* This is mdmon managed. */ - close(fd); + if (sysfs_get_ll(mdi, NULL, "sync_max", &old_sync_max) < 0) + old_sync_max = mdi->component_size; + /* Must not advance sync_max as that could confuse + * the reshape monitor */ + if (sync_max < old_sync_max) + sysfs_set_num(mdi, NULL, "sync_max", sync_max); + sysfs_set_str(mdi, NULL, "sync_action", "idle"); + + /* That should have set things going again. Now we + * wait a little while (3 second max) for sync_completed + * to reach the target. + * The reshape process can block for 500msec if + * the sync speed limit is hit, so we need to wait + * a lot longer than that. 1 second is usually + * enough. 3 is safe. + */ + delay = 3000; + scfd = sysfs_open(mdi->sys_name, NULL, "sync_completed"); + while (scfd >= 0 && delay > 0 && old_sync_max > 0) { + sysfs_get_ll(mdi, NULL, "reshape_position", &curr); + sysfs_fd_get_str(scfd, buf, sizeof(buf)); + if (strncmp(buf, "none", 4) == 0) { + /* Either reshape has aborted, or hasn't + * quite started yet. Wait a bit and + * check 'sync_action' to see. + */ + usleep(10000); + sysfs_get_str(mdi, NULL, "sync_action", buf, sizeof(buf)); + if (strncmp(buf, "reshape", 7) != 0) + break; + } - count = 25; - while (count && - (err = sysfs_set_str(mdi, NULL, - "array_state", - "inactive")) < 0 - && errno == EBUSY) { - usleep(200000); - count--; + if (sysfs_fd_get_ll(scfd, &completed) == 0 && + (completed > sync_max || + (completed == sync_max && curr != position))) { + while (completed > sync_max) { + sync_max += sectors / rd1; + if (backwards) + position -= sectors; + else + position += sectors; + } + if (sync_max < old_sync_max) + sysfs_set_num(mdi, NULL, "sync_max", sync_max); } - if (err && !quiet) { - fprintf(stderr, Name - ": failed to stop array %s: %s\n", - devname, strerror(errno)); - return 1; + + if (!backwards && curr >= position) + break; + if (backwards && curr <= position) + break; + sysfs_wait(scfd, &delay); + } + if (scfd >= 0) + close(scfd); + + } + + /* As we have an O_EXCL open, any use of the device + * which blocks STOP_ARRAY is probably a transient use, + * so it is reasonable to retry for a while - 5 seconds. + */ + count = 25; err = 0; + while (count && fd >= 0 + && (err = ioctl(fd, STOP_ARRAY, NULL)) < 0 + && errno == EBUSY) { + usleep(200000); + count --; + } + if (fd >= 0 && err) { + if (verbose >= 0) { + pr_err("failed to stop array %s: %s\n", + devname, strerror(errno)); + if (errno == EBUSY) + cont_err("Perhaps a running " + "process, mounted filesystem " + "or active volume group?\n"); + } + rv = 1; + goto out; + } + /* prior to 2.6.28, KOBJ_CHANGE was not sent when an md array + * was stopped, so We'll do it here just to be sure. Drop any + * partitions as well... + */ + if (fd >= 0) + ioctl(fd, BLKRRPART, 0); + if (mdi) + sysfs_uevent(mdi, "change"); + + if (devnm[0] && use_udev()) { + struct map_ent *mp = map_by_devnm(&map, devnm); + remove_devices(devnm, mp ? mp->path : NULL); + } + + if (verbose >= 0) + pr_err("stopped %s\n", devname); + map_lock(&map); + map_remove(&map, devnm); + map_unlock(&map); +out: + if (mdi) + sysfs_free(mdi); + + return rv; +} + +static struct mddev_dev *add_one(struct mddev_dev *dv, char *name, char disp) +{ + struct mddev_dev *new; + new = xmalloc(sizeof(*new)); + memset(new, 0, sizeof(*new)); + new->devname = xstrdup(name); + new->disposition = disp; + new->next = dv->next; + dv->next = new; + return new; +} + +static void add_faulty(struct mddev_dev *dv, int fd, char disp) +{ + mdu_array_info_t array; + mdu_disk_info_t disk; + int remaining_disks; + int i; + + if (ioctl(fd, GET_ARRAY_INFO, &array) != 0) + return; + + remaining_disks = array.nr_disks; + for (i = 0; i < MAX_DISKS && remaining_disks > 0; i++) { + char buf[40]; + disk.number = i; + if (ioctl(fd, GET_DISK_INFO, &disk) != 0) + continue; + if (disk.major == 0 && disk.minor == 0) + continue; + remaining_disks--; + if ((disk.state & 1) == 0) /* not faulty */ + continue; + sprintf(buf, "%d:%d", disk.major, disk.minor); + dv = add_one(dv, buf, disp); + } +} + +static void add_detached(struct mddev_dev *dv, int fd, char disp) +{ + mdu_array_info_t array; + mdu_disk_info_t disk; + int remaining_disks; + int i; + + if (ioctl(fd, GET_ARRAY_INFO, &array) != 0) + return; + + remaining_disks = array.nr_disks; + for (i = 0; i < MAX_DISKS && remaining_disks > 0; i++) { + char buf[40]; + int sfd; + disk.number = i; + if (ioctl(fd, GET_DISK_INFO, &disk) != 0) + continue; + if (disk.major == 0 && disk.minor == 0) + continue; + remaining_disks--; + if (disp == 'f' && (disk.state & 1) != 0) /* already faulty */ + continue; + sprintf(buf, "%d:%d", disk.major, disk.minor); + sfd = dev_open(buf, O_RDONLY); + if (sfd >= 0) { + /* Not detached */ + close(sfd); + continue; + } + if (errno != ENXIO) + /* Probably not detached */ + continue; + dv = add_one(dv, buf, disp); + } +} + +static void add_set(struct mddev_dev *dv, int fd, char set_char) +{ + mdu_array_info_t array; + mdu_disk_info_t disk; + int remaining_disks; + int copies, set; + int i; + + if (ioctl(fd, GET_ARRAY_INFO, &array) != 0) + return; + if (array.level != 10) + return; + copies = ((array.layout & 0xff) * + ((array.layout >> 8) & 0xff)); + if (array.raid_disks % copies) + return; + + remaining_disks = array.nr_disks; + for (i = 0; i < MAX_DISKS && remaining_disks > 0; i++) { + char buf[40]; + disk.number = i; + if (ioctl(fd, GET_DISK_INFO, &disk) != 0) + continue; + if (disk.major == 0 && disk.minor == 0) + continue; + remaining_disks--; + set = disk.raid_disk % copies; + if (set_char != set + 'A') + continue; + sprintf(buf, "%d:%d", disk.major, disk.minor); + dv = add_one(dv, buf, dv->disposition); + } +} + +int attempt_re_add(int fd, int tfd, struct mddev_dev *dv, + struct supertype *dev_st, struct supertype *tst, + unsigned long rdev, + char *update, char *devname, int verbose, + mdu_array_info_t *array) +{ + struct mdinfo mdi; + int duuid[4]; + int ouuid[4]; + + dev_st->ss->getinfo_super(dev_st, &mdi, NULL); + dev_st->ss->uuid_from_super(dev_st, ouuid); + if (tst->sb) + tst->ss->uuid_from_super(tst, duuid); + else + /* Assume uuid matches: kernel will check */ + memcpy(duuid, ouuid, sizeof(ouuid)); + if ((mdi.disk.state & (1<major_version == 1 && + get_linux_version() <= 2006018) + goto skip_re_add; + disc.number = mdi.disk.number; + if (ioctl(fd, GET_DISK_INFO, &disc) != 0 + || disc.major != 0 || disc.minor != 0 + ) + goto skip_re_add; + disc.major = major(rdev); + disc.minor = minor(rdev); + disc.number = mdi.disk.number; + disc.raid_disk = mdi.disk.raid_disk; + disc.state = mdi.disk.state; + if (dv->writemostly == 1) + disc.state |= 1 << MD_DISK_WRITEMOSTLY; + if (dv->writemostly == 2) + disc.state &= ~(1 << MD_DISK_WRITEMOSTLY); + remove_partitions(tfd); + if (update || dv->writemostly > 0) { + int rv = -1; + tfd = dev_open(dv->devname, O_RDWR); + if (tfd < 0) { + pr_err("failed to open %s for" + " superblock update during re-add\n", dv->devname); + return -1; + } + + if (dv->writemostly == 1) + rv = dev_st->ss->update_super( + dev_st, NULL, "writemostly", + devname, verbose, 0, NULL); + if (dv->writemostly == 2) + rv = dev_st->ss->update_super( + dev_st, NULL, "readwrite", + devname, verbose, 0, NULL); + if (update) + rv = dev_st->ss->update_super( + dev_st, NULL, update, + devname, verbose, 0, NULL); + if (rv == 0) + rv = dev_st->ss->store_super(dev_st, tfd); + close(tfd); + if (rv != 0) { + pr_err("failed to update" + " superblock during re-add\n"); + return -1; } + } + /* don't even try if disk is marked as faulty */ + errno = 0; + if (ioctl(fd, ADD_NEW_DISK, &disc) == 0) { + if (verbose >= 0) + pr_err("re-added %s\n", dv->devname); + return 1; + } + if (errno == ENOMEM || errno == EROFS) { + pr_err("add new device failed for %s: %s\n", + dv->devname, strerror(errno)); + if (dv->disposition == 'M') + return 0; + return -1; + } + } +skip_re_add: + return 0; +} + +int Manage_add(int fd, int tfd, struct mddev_dev *dv, + struct supertype *tst, mdu_array_info_t *array, + int force, int verbose, char *devname, + char *update, unsigned long rdev, unsigned long long array_size) +{ + unsigned long long ldsize; + struct supertype *dev_st = NULL; + int j; + mdu_disk_info_t disc; - /* Give monitor a chance to act */ - ping_monitor(mdi->text_version); + if (!get_dev_size(tfd, dv->devname, &ldsize)) { + if (dv->disposition == 'M') + return 0; + else + return -1; + } - fd = open_dev_excl(devnum); - if (fd < 0) { - fprintf(stderr, Name - ": failed to completely stop %s" - ": Device is busy\n", - devname); - return 1; + if (tst->ss == &super0 && ldsize > 4ULL*1024*1024*1024*1024) { + /* More than 4TB is wasted on v0.90 */ + if (!force) { + pr_err("%s is larger than %s can " + "effectively use.\n" + " Add --force is you " + "really want to add this device.\n", + dv->devname, devname); + return -1; + } + pr_err("%s is larger than %s can " + "effectively use.\n" + " Adding anyway as --force " + "was given.\n", + dv->devname, devname); + } + if (!tst->ss->external && + array->major_version == 0 && + md_get_version(fd)%100 < 2) { + if (ioctl(fd, HOT_ADD_DISK, rdev)==0) { + if (verbose >= 0) + pr_err("hot added %s\n", + dv->devname); + return 1; + } + + pr_err("hot add failed for %s: %s\n", + dv->devname, strerror(errno)); + return -1; + } + + if (array->not_persistent == 0 || tst->ss->external) { + + /* need to find a sample superblock to copy, and + * a spare slot to use. + * For 'external' array (well, container based), + * We can just load the metadata for the array-> + */ + int array_failed; + if (tst->sb) + /* already loaded */; + else if (tst->ss->external) { + tst->ss->load_container(tst, fd, NULL); + } else for (j = 0; j < tst->max_devs; j++) { + char *dev; + int dfd; + disc.number = j; + if (ioctl(fd, GET_DISK_INFO, &disc)) + continue; + if (disc.major==0 && disc.minor==0) + continue; + if ((disc.state & 4)==0) /* sync */ + continue; + /* Looks like a good device to try */ + dev = map_dev(disc.major, disc.minor, 1); + if (!dev) + continue; + dfd = dev_open(dev, O_RDONLY); + if (dfd < 0) + continue; + if (tst->ss->load_super(tst, dfd, + NULL)) { + close(dfd); + continue; + } + close(dfd); + break; } - } else if (mdi && - mdi->array.major_version == -1 && - mdi->array.minor_version == -2 && - !is_subarray(mdi->text_version)) { - struct mdstat_ent *mds, *m; - /* container, possibly mdmon-managed. - * Make sure mdmon isn't opening it, which - * would interfere with the 'stop' + /* FIXME this is a bad test to be using */ + if (!tst->sb && (dv->disposition != 'a' + && dv->disposition != 'S')) { + /* we are re-adding a device to a + * completely dead array - have to depend + * on kernel to check */ - ping_monitor(mdi->sys_name); + } else if (!tst->sb) { + pr_err("cannot load array metadata from %s\n", devname); + return -1; + } - /* now check that there are no existing arrays - * which are members of this array - */ - mds = mdstat_read(0, 0); - for (m=mds; m; m=m->next) - if (m->metadata_version && - strncmp(m->metadata_version, "external:", 9)==0 && - is_subarray(m->metadata_version+9) && - devname2devnum(m->metadata_version+10) == devnum) { - if (!quiet) - fprintf(stderr, Name - ": Cannot stop container %s: " - "member %s still active\n", - devname, m->dev); - free_mdstat(mds); - if (mdi) - sysfs_free(mdi); - return 1; - } + /* Make sure device is large enough */ + if (tst->ss->avail_size(tst, ldsize/512, INVALID_SECTORS) < + array_size) { + if (dv->disposition == 'M') + return 0; + pr_err("%s not large enough to join array\n", + dv->devname); + return -1; } - /* As we have an O_EXCL open, any use of the device - * which blocks STOP_ARRAY is probably a transient use, - * so it is reasonable to retry for a while - 5 seconds. + /* Possibly this device was recently part of + * the array and was temporarily removed, and + * is now being re-added. If so, we can + * simply re-add it. */ - count = 25; err = 0; - while (count && fd >= 0 - && (err = ioctl(fd, STOP_ARRAY, NULL)) < 0 - && errno == EBUSY) { - usleep(200000); - count --; + + if (array->not_persistent==0) { + dev_st = dup_super(tst); + dev_st->ss->load_super(dev_st, tfd, NULL); } - if (fd >= 0 && err) { - if (quiet == 0) { - fprintf(stderr, Name - ": failed to stop array %s: %s\n", - devname, strerror(errno)); - if (errno == EBUSY) - fprintf(stderr, "Perhaps a running " - "process, mounted filesystem " - "or active volume group?\n"); + if (dev_st && dev_st->sb && dv->disposition != 'S') { + int rv = attempt_re_add(fd, tfd, dv, + dev_st, tst, + rdev, + update, devname, + verbose, + array); + dev_st->ss->free_super(dev_st); + if (rv) + return rv; + } + if (dv->disposition == 'M') { + if (verbose > 0) + pr_err("--re-add for %s to %s is not possible\n", + dv->devname, devname); + return 0; + } + if (dv->disposition == 'A') { + pr_err("--re-add for %s to %s is not possible\n", + dv->devname, devname); + return -1; + } + if (array->active_disks < array->raid_disks) { + char *avail = xcalloc(array->raid_disks, 1); + int d; + int found = 0; + + for (d = 0; d < MAX_DISKS && found < array->active_disks; d++) { + disc.number = d; + if (ioctl(fd, GET_DISK_INFO, &disc)) + continue; + if (disc.major == 0 && disc.minor == 0) + continue; + found++; + if (!(disc.state & (1<level, array->raid_disks, + array->layout, 1, avail); + free(avail); + } else + array_failed = 0; + if (array_failed) { + pr_err("%s has failed so using --add cannot work and might destroy\n", + devname); + pr_err("data on %s. You should stop the array and re-assemble it.\n", + dv->devname); + return -1; } - /* prior to 2.6.28, KOBJ_CHANGE was not sent when an md array - * was stopped, so We'll do it here just to be sure. Drop any - * partitions as well... + } else { + /* non-persistent. Must ensure that new drive + * is at least array->size big. */ - if (fd >= 0) - ioctl(fd, BLKRRPART, 0); - if (mdi) - sysfs_uevent(mdi, "change"); - - - if (devnum != NoMdDev && - (stat("/dev/.udev", &stb) != 0 || - check_env("MDADM_NO_UDEV"))) { - struct map_ent *mp = map_by_devnum(&map, devnum); - remove_devices(devnum, mp ? mp->path : NULL); + if (ldsize/512 < array_size) { + pr_err("%s not large enough to join array\n", + dv->devname); + return -1; } + } + /* committed to really trying this device now*/ + remove_partitions(tfd); + /* in 2.6.17 and earlier, version-1 superblocks won't + * use the number we write, but will choose a free number. + * we must choose the same free number, which requires + * starting at 'raid_disks' and counting up + */ + for (j = array->raid_disks; j < tst->max_devs; j++) { + disc.number = j; + if (ioctl(fd, GET_DISK_INFO, &disc)) + break; + if (disc.major==0 && disc.minor==0) + break; + if (disc.state & 8) /* removed */ + break; + } + disc.major = major(rdev); + disc.minor = minor(rdev); + disc.number =j; + disc.state = 0; + if (array->not_persistent==0) { + int dfd; + if (dv->writemostly == 1) + disc.state |= 1 << MD_DISK_WRITEMOSTLY; + dfd = dev_open(dv->devname, O_RDWR | O_EXCL|O_DIRECT); + if (tst->ss->add_to_super(tst, &disc, dfd, + dv->devname, INVALID_SECTORS)) + return -1; + if (tst->ss->write_init_super(tst)) + return -1; + } else if (dv->disposition == 'A') { + /* this had better be raid1. + * As we are "--re-add"ing we must find a spare slot + * to fill. + */ + char *used = xcalloc(array->raid_disks, 1); + for (j = 0; j < tst->max_devs; j++) { + mdu_disk_info_t disc2; + disc2.number = j; + if (ioctl(fd, GET_DISK_INFO, &disc2)) + continue; + if (disc2.major==0 && disc2.minor==0) + continue; + if (disc2.state & 8) /* removed */ + continue; + if (disc2.raid_disk < 0) + continue; + if (disc2.raid_disk > array->raid_disks) + continue; + used[disc2.raid_disk] = 1; + } + for (j = 0 ; j < array->raid_disks; j++) + if (!used[j]) { + disc.raid_disk = j; + disc.state |= (1<writemostly == 1) + disc.state |= (1 << MD_DISK_WRITEMOSTLY); + if (tst->ss->external) { + /* add a disk + * to an external metadata container */ + struct mdinfo new_mdi; + struct mdinfo *sra; + int container_fd; + char devnm[32]; + int dfd; + + strcpy(devnm, fd2devnm(fd)); + + container_fd = open_dev_excl(devnm); + if (container_fd < 0) { + pr_err("add failed for %s:" + " could not get exclusive access to container\n", + dv->devname); + tst->ss->free_super(tst); + return -1; + } - if (quiet <= 0) - fprintf(stderr, Name ": stopped %s\n", devname); - map_lock(&map); - map_remove(&map, devnum); - map_unlock(&map); + Kill(dv->devname, NULL, 0, -1, 0); + dfd = dev_open(dv->devname, O_RDWR | O_EXCL|O_DIRECT); + if (mdmon_running(tst->container_devnm)) + tst->update_tail = &tst->updates; + if (tst->ss->add_to_super(tst, &disc, dfd, + dv->devname, INVALID_SECTORS)) { + close(dfd); + close(container_fd); + return -1; + } + if (tst->update_tail) + flush_metadata_updates(tst); + else + tst->ss->sync_metadata(tst); + + sra = sysfs_read(container_fd, NULL, 0); + if (!sra) { + pr_err("add failed for %s: sysfs_read failed\n", + dv->devname); + close(container_fd); + tst->ss->free_super(tst); + return -1; + } + sra->array.level = LEVEL_CONTAINER; + /* Need to set data_offset and component_size */ + tst->ss->getinfo_super(tst, &new_mdi, NULL); + new_mdi.disk.major = disc.major; + new_mdi.disk.minor = disc.minor; + new_mdi.recovery_start = 0; + /* Make sure fds are closed as they are O_EXCL which + * would block add_disk */ + tst->ss->free_super(tst); + if (sysfs_add_disk(sra, &new_mdi, 0) != 0) { + pr_err("add new device to external metadata" + " failed for %s\n", dv->devname); + close(container_fd); + sysfs_free(sra); + return -1; + } + ping_monitor(devnm); + sysfs_free(sra); + close(container_fd); + } else { + tst->ss->free_super(tst); + if (ioctl(fd, ADD_NEW_DISK, &disc)) { + pr_err("add new device failed for %s as %d: %s\n", + dv->devname, j, strerror(errno)); + return -1; + } } - return 0; + if (verbose >= 0) + pr_err("added %s\n", dv->devname); + return 1; } -int Manage_resize(char *devname, int fd, long long size, int raid_disks) +int Manage_remove(struct supertype *tst, int fd, struct mddev_dev *dv, + int sysfd, unsigned long rdev, int verbose, char *devname) { - mdu_array_info_t info; - if (ioctl(fd, GET_ARRAY_INFO, &info) != 0) { - fprintf(stderr, Name ": Cannot get array information for %s: %s\n", - devname, strerror(errno)); + int lfd = -1; + int err; + + if (tst->ss->external) { + /* To remove a device from a container, we must + * check that it isn't in use in an array. + * This involves looking in the 'holders' + * directory - there must be just one entry, + * the container. + * To ensure that it doesn't get used as a + * hot spare while we are checking, we + * get an O_EXCL open on the container + */ + int ret; + char devnm[32]; + strcpy(devnm, fd2devnm(fd)); + lfd = open_dev_excl(devnm); + if (lfd < 0) { + pr_err("Cannot get exclusive access " + " to container - odd\n"); + return -1; + } + /* We may not be able to check on holders in + * sysfs, either because we don't have the dev num + * (rdev == 0) or because the device has been detached + * and the 'holders' directory no longer exists + * (ret == -1). In that case, assume it is OK to + * remove. + */ + if (rdev == 0) + ret = -1; + else + ret = sysfs_unique_holder(devnm, rdev); + if (ret == 0) { + pr_err("%s is not a member, cannot remove.\n", + dv->devname); + close(lfd); + return -1; + } + if (ret >= 2) { + pr_err("%s is still in use, cannot remove.\n", + dv->devname); + close(lfd); + return -1; + } + } + /* FIXME check that it is a current member */ + if (sysfd >= 0) { + /* device has been removed and we don't know + * the major:minor number + */ + int n = write(sysfd, "remove", 6); + if (n != 6) + err = -1; + else + err = 0; + } else { + err = ioctl(fd, HOT_REMOVE_DISK, rdev); + if (err && errno == ENODEV) { + /* Old kernels rejected this if no personality + * is registered */ + struct mdinfo *sra = sysfs_read(fd, NULL, GET_DEVS); + struct mdinfo *dv = NULL; + if (sra) + dv = sra->devs; + for ( ; dv ; dv=dv->next) + if (dv->disk.major == (int)major(rdev) && + dv->disk.minor == (int)minor(rdev)) + break; + if (dv) + err = sysfs_set_str(sra, dv, + "state", "remove"); + else + err = -1; + if (sra) + sysfs_free(sra); + } + } + if (err) { + pr_err("hot remove failed " + "for %s: %s\n", dv->devname, + strerror(errno)); + if (lfd >= 0) + close(lfd); + return -1; + } + if (tst->ss->external) { + /* + * Before dropping our exclusive open we make an + * attempt at preventing mdmon from seeing an + * 'add' event before reconciling this 'remove' + * event. + */ + char *devnm = fd2devnm(fd); + + if (!devnm) { + pr_err("unable to get container name\n"); + return -1; + } + + ping_manager(devnm); + } + if (lfd >= 0) + close(lfd); + if (verbose >= 0) + pr_err("hot removed %s from %s\n", + dv->devname, devname); + return 1; +} + +int Manage_replace(struct supertype *tst, int fd, struct mddev_dev *dv, + unsigned long rdev, int verbose, char *devname) +{ + struct mdinfo *mdi, *di; + if (tst->ss->external) { + pr_err("--replace only supported for native metadata (0.90 or 1.x)\n"); + return -1; + } + /* Need to find the device in sysfs and add 'want_replacement' to the + * status. + */ + mdi = sysfs_read(fd, NULL, GET_DEVS); + if (!mdi || !mdi->devs) { + pr_err("Cannot find status of %s to enable replacement - strange\n", + devname); + return -1; + } + for (di = mdi->devs; di; di = di->next) + if (di->disk.major == (int)major(rdev) && + di->disk.minor == (int)minor(rdev)) + break; + if (di) { + int rv; + if (di->disk.raid_disk < 0) { + pr_err("%s is not active and so cannot be replaced.\n", + dv->devname); + sysfs_free(mdi); + return -1; + } + rv = sysfs_set_str(mdi, di, + "state", "want_replacement"); + if (rv) { + sysfs_free(mdi); + pr_err("Failed to request replacement for %s\n", + dv->devname); + return -1; + } + if (verbose >= 0) + pr_err("Marked %s (device %d in %s) for replacement\n", + dv->devname, di->disk.raid_disk, devname); + /* If there is a matching 'with', we need to tell it which + * raid disk + */ + while (dv && dv->disposition != 'W') + dv = dv->next; + if (dv) { + dv->disposition = 'w'; + dv->used = di->disk.raid_disk; + } return 1; } - if (size >= 0) - info.size = size; - if (raid_disks > 0) - info.raid_disks = raid_disks; - if (ioctl(fd, SET_ARRAY_INFO, &info) != 0) { - fprintf(stderr, Name ": Cannot set device size/shape for %s: %s\n", - devname, strerror(errno)); + sysfs_free(mdi); + pr_err("%s not found in %s so cannot --replace it\n", + dv->devname, devname); + return -1; +} + +int Manage_with(struct supertype *tst, int fd, struct mddev_dev *dv, + unsigned long rdev, int verbose, char *devname) +{ + struct mdinfo *mdi, *di; + /* try to set 'slot' for 'rdev' in 'fd' to 'dv->used' */ + mdi = sysfs_read(fd, NULL, GET_DEVS|GET_STATE); + if (!mdi || !mdi->devs) { + pr_err("Cannot find status of %s to enable replacement - strange\n", + devname); + return -1; + } + for (di = mdi->devs; di; di = di->next) + if (di->disk.major == (int)major(rdev) && + di->disk.minor == (int)minor(rdev)) + break; + if (di) { + int rv; + if (di->disk.state & (1<devname); + sysfs_free(mdi); + return -1; + } + if (di->disk.raid_disk >= 0) { + pr_err("%s is active and cannot be a replacement\n", + dv->devname); + sysfs_free(mdi); + return -1; + } + rv = sysfs_set_num(mdi, di, + "slot", dv->used); + if (rv) { + sysfs_free(mdi); + pr_err("Failed to set %s as preferred replacement.\n", + dv->devname); + return -1; + } + if (verbose >= 0) + pr_err("Marked %s in %s as replacement for device %d\n", + dv->devname, devname, dv->used); return 1; } - return 0; + sysfs_free(mdi); + pr_err("%s not found in %s so cannot make it preferred replacement\n", + dv->devname, devname); + return -1; } int Manage_subdevs(char *devname, int fd, struct mddev_dev *devlist, int verbose, int test, char *update, int force) { - /* do something to each dev. + /* Do something to each dev. * devmode can be * 'a' - add the device * try HOT_ADD_DISK * If that fails EINVAL, try ADD_NEW_DISK - * 'r' - remove the device HOT_REMOVE_DISK + * 'S' - add the device as a spare - don't try re-add + * 'A' - re-add the device + * 'r' - remove the device: HOT_REMOVE_DISK * device can be 'faulty' or 'detached' in which case all * matching devices are removed. * 'f' - set the device faulty SET_DISK_FAULTY * device can be 'detached' in which case any device that * is inaccessible will be marked faulty. + * 'R' - mark this device as wanting replacement. + * 'W' - this device is added if necessary and activated as + * a replacement for a previous 'R' device. + * ----- + * 'w' - 'W' will be changed to 'w' when it is paired with + * a 'R' device. If a 'W' is found while walking the list + * it must be unpaired, and is an error. + * 'M' - this is created by a 'missing' target. It is a slight + * variant on 'A' + * 'F' - Another variant of 'A', where the device was faulty + * so must be removed from the array first. + * * For 'f' and 'r', the device can also be a kernel-internal * name such as 'sdb'. */ - struct mddev_dev *add_devlist = NULL; mdu_array_info_t array; - mdu_disk_info_t disc; unsigned long long array_size; - struct mddev_dev *dv, *next = NULL; + struct mddev_dev *dv; struct stat stb; - int j, jnext = 0; int tfd = -1; - struct supertype *st, *tst; + struct supertype *tst; char *subarray = NULL; - int duuid[4]; - int ouuid[4]; - int lfd = -1; int sysfd = -1; int count = 0; /* number of actions taken */ + struct mdinfo info; + int frozen = 0; + int busy = 0; if (ioctl(fd, GET_ARRAY_INFO, &array)) { - fprintf(stderr, Name ": cannot get array info for %s\n", + pr_err("Cannot get array info for %s\n", devname); - return 1; + goto abort; } + sysfs_init(&info, fd, NULL); - /* array.size is only 32 bit and may be truncated. + /* array.size is only 32 bits and may be truncated. * So read from sysfs if possible, and record number of sectors */ @@ -419,132 +1291,104 @@ int Manage_subdevs(char *devname, int fd, tst = super_by_fd(fd, &subarray); if (!tst) { - fprintf(stderr, Name ": unsupport array - version %d.%d\n", + pr_err("unsupport array - version %d.%d\n", array.major_version, array.minor_version); - return 1; + goto abort; } stb.st_rdev = 0; - for (dv = devlist, j=0 ; dv; dv = next, j = jnext) { - unsigned long long ldsize; - char dvname[20]; - char *dnprintable = dv->devname; - char *add_dev = dv->devname; - int err; - int re_add_failed = 0; - - next = dv->next; - jnext = 0; - - if (strcmp(dv->devname, "failed")==0 || - strcmp(dv->devname, "faulty")==0) { - int remaining_disks = array.nr_disks; - if (dv->disposition != 'r') { - fprintf(stderr, Name ": %s only meaningful " - "with -r, not -%c\n", + for (dv = devlist; dv; dv = dv->next) { + int rv; + + if (strcmp(dv->devname, "failed") == 0 || + strcmp(dv->devname, "faulty") == 0) { + if (dv->disposition != 'A' + && dv->disposition != 'r') { + pr_err("%s only meaningful " + "with -r or --re-add, not -%c\n", dv->devname, dv->disposition); - return 1; + goto abort; } - for (; j < 1024 && remaining_disks > 0; j++) { - unsigned dev; - disc.number = j; - if (ioctl(fd, GET_DISK_INFO, &disc)) - continue; - if (disc.major == 0 && disc.minor == 0) - continue; - remaining_disks --; - if ((disc.state & 1) == 0) /* faulty */ - continue; - dev = makedev(disc.major, disc.minor); - if (stb.st_rdev == dev) - /* already did that one */ - continue; - stb.st_rdev = dev; - next = dv; - /* same slot again next time - things might - * have reshuffled */ - jnext = j; - sprintf(dvname,"%d:%d", disc.major, disc.minor); - dnprintable = dvname; - break; - } - if (next != dv) - continue; - } else if (strcmp(dv->devname, "detached") == 0) { - int remaining_disks = array.nr_disks; + add_faulty(dv, fd, (dv->disposition == 'A' + ? 'F' : 'r')); + continue; + } + if (strcmp(dv->devname, "detached") == 0) { if (dv->disposition != 'r' && dv->disposition != 'f') { - fprintf(stderr, Name ": %s only meaningful " + pr_err("%s only meaningful " "with -r of -f, not -%c\n", dv->devname, dv->disposition); - return 1; + goto abort; } - for (; j < 1024 && remaining_disks > 0; j++) { - int sfd; - unsigned dev; - disc.number = j; - if (ioctl(fd, GET_DISK_INFO, &disc)) - continue; - if (disc.major == 0 && disc.minor == 0) - continue; - remaining_disks --; - sprintf(dvname,"%d:%d", disc.major, disc.minor); - sfd = dev_open(dvname, O_RDONLY); - if (sfd >= 0) { - close(sfd); - continue; - } - if (dv->disposition == 'f' && - (disc.state & 1) == 1) /* already faulty */ - continue; - if (errno != ENXIO) - continue; - dev = makedev(disc.major, disc.minor); - if (stb.st_rdev == dev) - /* already did that one */ - continue; - stb.st_rdev = dev; - next = dv; - /* same slot again next time - things might - * have reshuffled */ - jnext = j; - dnprintable = dvname; - break; - } - if (next != dv) - continue; - } else if (strcmp(dv->devname, "missing") == 0) { - if (dv->disposition != 'a' || dv->re_add == 0) { - fprintf(stderr, Name ": 'missing' only meaningful " - "with --re-add\n"); - return 1; + add_detached(dv, fd, dv->disposition); + continue; + } + + if (strcmp(dv->devname, "missing") == 0) { + struct mddev_dev *add_devlist = NULL; + struct mddev_dev **dp; + if (dv->disposition != 'A') { + pr_err("'missing' only meaningful " + "with --re-add\n"); + goto abort; } - if (add_devlist == NULL) - add_devlist = conf_get_devs(); + add_devlist = conf_get_devs(); if (add_devlist == NULL) { - fprintf(stderr, Name ": no devices to scan for missing members."); + pr_err("no devices to scan for missing members."); continue; } - add_dev = add_devlist->devname; - add_devlist = add_devlist->next; - if (add_devlist != NULL) - next = dv; - if (stat(add_dev, &stb) < 0) - continue; - } else if (strchr(dv->devname, '/') == NULL && - strchr(dv->devname, ':') == NULL && - strlen(dv->devname) < 50) { + for (dp = &add_devlist; *dp; dp = & (*dp)->next) + /* 'M' (for 'missing') is like 'A' without errors */ + (*dp)->disposition = 'M'; + *dp = dv->next; + dv->next = add_devlist; + continue; + } + + if (strncmp(dv->devname, "set-", 4) == 0 && + strlen(dv->devname) == 5) { + int copies; + + if (dv->disposition != 'r' && + dv->disposition != 'f') { + pr_err("'%s' only meaningful with -r or -f\n", + dv->devname); + goto abort; + } + if (array.level != 10) { + pr_err("'%s' only meaningful with RAID10 arrays\n", + dv->devname); + goto abort; + } + copies = ((array.layout & 0xff) * + ((array.layout >> 8) & 0xff)); + if (array.raid_disks % copies != 0 || + dv->devname[4] < 'A' || + dv->devname[4] >= 'A' + copies || + copies > 26) { + pr_err("'%s' not meaningful with this array\n", + dv->devname); + goto abort; + } + add_set(dv, fd, dv->devname[4]); + continue; + } + + if (strchr(dv->devname, '/') == NULL && + strchr(dv->devname, ':') == NULL && + strlen(dv->devname) < 50) { /* Assume this is a kernel-internal name like 'sda1' */ int found = 0; char dname[55]; if (dv->disposition != 'r' && dv->disposition != 'f') { - fprintf(stderr, Name ": %s only meaningful " + pr_err("%s only meaningful " "with -r or -f, not -%c\n", dv->devname, dv->disposition); - return 1; + goto abort; } sprintf(dname, "dev-%s", dv->devname); - sysfd = sysfs_open(fd2devnum(fd), dname, "block/dev"); + sysfd = sysfs_open(fd2devnm(fd), dname, "block/dev"); if (sysfd >= 0) { char dn[20]; int mj,mn; @@ -557,545 +1401,121 @@ int Manage_subdevs(char *devname, int fd, sysfd = -1; } if (!found) { - sysfd = sysfs_open(fd2devnum(fd), dname, "state"); + sysfd = sysfs_open(fd2devnm(fd), dname, "state"); if (sysfd < 0) { - fprintf(stderr, Name ": %s does not appear " + pr_err("%s does not appear " "to be a component of %s\n", dv->devname, devname); - return 1; + goto abort; } } } else { - j = 0; - tfd = dev_open(dv->devname, O_RDONLY); - if (tfd < 0 && dv->disposition == 'r' && - lstat(dv->devname, &stb) == 0) - /* Be happy, the lstat worked, that is - * enough for --remove - */ - ; + if (tfd >= 0) + fstat(tfd, &stb); else { - if (tfd < 0 || fstat(tfd, &stb) != 0) { - fprintf(stderr, Name ": cannot find %s: %s\n", - dv->devname, strerror(errno)); - if (tfd >= 0) - close(tfd); - return 1; + int open_err = errno; + if (stat(dv->devname, &stb) != 0) { + pr_err("Cannot find %s: %s\n", + dv->devname, strerror(errno)); + goto abort; + } + if ((stb.st_mode & S_IFMT) != S_IFBLK) { + if (dv->disposition == 'M') + /* non-fatal. Also improbable */ + continue; + pr_err("%s is not a block device.\n", + dv->devname); + goto abort; + } + if (dv->disposition == 'r') + /* Be happy, the stat worked, that is + * enough for --remove + */ + ; + else { + if (dv->disposition == 'M') + /* non-fatal */ + continue; + pr_err("Cannot open %s: %s\n", + dv->devname, strerror(open_err)); + goto abort; } - close(tfd); - tfd = -1; - } - if ((stb.st_mode & S_IFMT) != S_IFBLK) { - fprintf(stderr, Name ": %s is not a " - "block device.\n", - dv->devname); - return 1; } } switch(dv->disposition){ default: - fprintf(stderr, Name ": internal error - devmode[%s]=%d\n", + pr_err("internal error - devmode[%s]=%d\n", dv->devname, dv->disposition); - return 1; + goto abort; case 'a': + case 'S': /* --add-spare */ + case 'A': + case 'M': /* --re-add missing */ + case 'F': /* --re-add faulty */ /* add the device */ if (subarray) { - fprintf(stderr, Name ": Cannot add disks to a" + pr_err("Cannot add disks to a" " \'member\' array, perform this" " operation on the parent container\n"); - return 1; + goto abort; } + if (dv->disposition == 'F') + /* Need to remove first */ + ioctl(fd, HOT_REMOVE_DISK, + (unsigned long)stb.st_rdev); /* Make sure it isn't in use (in 2.6 or later) */ - tfd = dev_open(add_dev, O_RDONLY|O_EXCL|O_DIRECT); - if (tfd < 0 && add_dev != dv->devname) - continue; - if (tfd < 0) { - fprintf(stderr, Name ": Cannot open %s: %s\n", - dv->devname, strerror(errno)); - return 1; - } - - st = dup_super(tst); - - if (array.not_persistent==0) - st->ss->load_super(st, tfd, NULL); - - if (add_dev == dv->devname) { - if (!get_dev_size(tfd, dv->devname, &ldsize)) { - close(tfd); - return 1; - } - } else if (!get_dev_size(tfd, NULL, &ldsize)) { - close(tfd); - tfd = -1; - continue; - } - - if (tst->ss->validate_geometry( - tst, array.level, array.layout, - array.raid_disks, NULL, - ldsize >> 9, NULL, NULL, 0) == 0) { - if (!force) { - fprintf(stderr, Name - ": %s is larger than %s can " - "effectively use.\n" - " Add --force is you " - "really wan to add this device.\n", - add_dev, devname); - close(tfd); - return 1; - } - fprintf(stderr, Name - ": %s is larger than %s can " - "effectively use.\n" - " Adding anyway as --force " - "was given.\n", - add_dev, devname); - } - if (!tst->ss->external && - array.major_version == 0 && - md_get_version(fd)%100 < 2) { - close(tfd); - tfd = -1; - if (ioctl(fd, HOT_ADD_DISK, - (unsigned long)stb.st_rdev)==0) { - if (verbose >= 0) - fprintf(stderr, Name ": hot added %s\n", - add_dev); - continue; - } - - fprintf(stderr, Name ": hot add failed for %s: %s\n", - add_dev, strerror(errno)); - return 1; - } - - if (array.not_persistent == 0 || tst->ss->external) { - - /* need to find a sample superblock to copy, and - * a spare slot to use. - * For 'external' array (well, container based), - * We can just load the metadata for the array. - */ - if (tst->sb) - /* already loaded */; - else if (tst->ss->external) { - tst->ss->load_container(tst, fd, NULL); - } else for (j = 0; j < tst->max_devs; j++) { - char *dev; - int dfd; - disc.number = j; - if (ioctl(fd, GET_DISK_INFO, &disc)) - continue; - if (disc.major==0 && disc.minor==0) - continue; - if ((disc.state & 4)==0) continue; /* sync */ - /* Looks like a good device to try */ - dev = map_dev(disc.major, disc.minor, 1); - if (!dev) continue; - dfd = dev_open(dev, O_RDONLY); - if (dfd < 0) continue; - if (tst->ss->load_super(tst, dfd, - NULL)) { - close(dfd); - continue; - } - close(dfd); - break; - } - /* FIXME this is a bad test to be using */ - if (!tst->sb) { - close(tfd); - fprintf(stderr, Name ": cannot load array metadata from %s\n", devname); - return 1; - } - - /* Make sure device is large enough */ - if (tst->ss->avail_size(tst, ldsize/512) < - array_size) { - close(tfd); - tfd = -1; - if (add_dev != dv->devname) - continue; - fprintf(stderr, Name ": %s not large enough to join array\n", - dv->devname); - return 1; - } - - /* Possibly this device was recently part of the array - * and was temporarily removed, and is now being re-added. - * If so, we can simply re-add it. - */ - tst->ss->uuid_from_super(tst, duuid); - - if (st->sb) { - struct mdinfo mdi; - st->ss->getinfo_super(st, &mdi, NULL); - st->ss->uuid_from_super(st, ouuid); - if ((mdi.disk.state & (1<writemostly == 1) - disc.state |= 1 << MD_DISK_WRITEMOSTLY; - if (dv->writemostly == 2) - disc.state &= ~(1 << MD_DISK_WRITEMOSTLY); - remove_partitions(tfd); - close(tfd); - tfd = -1; - if (update) { - int rv = -1; - tfd = dev_open(dv->devname, O_RDWR); - - if (tfd >= 0) - rv = st->ss->update_super( - st, NULL, update, - devname, verbose, 0, NULL); - if (rv == 0) - rv = st->ss->store_super(st, tfd); - close(tfd); - tfd = -1; - if (rv != 0) { - fprintf(stderr, Name ": failed to update" - " superblock during re-add\n"); - return 1; - } - } - /* don't even try if disk is marked as faulty */ - errno = 0; - if (ioctl(fd, ADD_NEW_DISK, &disc) == 0) { - if (verbose >= 0) - fprintf(stderr, Name ": re-added %s\n", add_dev); - count++; - continue; - } - if (errno == ENOMEM || errno == EROFS) { - fprintf(stderr, Name ": add new device failed for %s: %s\n", - add_dev, strerror(errno)); - if (add_dev != dv->devname) - continue; - return 1; - } - skip_re_add: - re_add_failed = 1; - } - st->ss->free_super(st); - } - if (add_dev != dv->devname) { - if (verbose > 0) - fprintf(stderr, Name - ": --re-add for %s to %s is not possible\n", - add_dev, devname); - if (tfd >= 0) { - close(tfd); - tfd = -1; - } - continue; - } - if (dv->re_add) { - if (tfd >= 0) - close(tfd); - fprintf(stderr, Name - ": --re-add for %s to %s is not possible\n", - dv->devname, devname); - return 1; - } - if (re_add_failed) { - fprintf(stderr, Name ": %s reports being an active member for %s, but a --re-add fails.\n", - dv->devname, devname); - fprintf(stderr, Name ": not performing --add as that would convert %s in to a spare.\n", - dv->devname); - fprintf(stderr, Name ": To make this a spare, use \"mdadm --zero-superblock %s\" first.\n", - dv->devname); - if (tfd >= 0) - close(tfd); - return 1; - } - } else { - /* non-persistent. Must ensure that new drive - * is at least array.size big. - */ - if (ldsize/512 < array_size) { - fprintf(stderr, Name ": %s not large enough to join array\n", - dv->devname); - if (tfd >= 0) - close(tfd); - return 1; - } - } - /* committed to really trying this device now*/ + tfd = dev_open(dv->devname, O_RDONLY|O_EXCL); if (tfd >= 0) { - remove_partitions(tfd); + /* We know no-one else is using it. We'll + * need non-exclusive access to add it, so + * do that now. + */ close(tfd); - tfd = -1; + tfd = dev_open(dv->devname, O_RDONLY); } - /* in 2.6.17 and earlier, version-1 superblocks won't - * use the number we write, but will choose a free number. - * we must choose the same free number, which requires - * starting at 'raid_disks' and counting up - */ - for (j = array.raid_disks; j< tst->max_devs; j++) { - disc.number = j; - if (ioctl(fd, GET_DISK_INFO, &disc)) - break; - if (disc.major==0 && disc.minor==0) - break; - if (disc.state & 8) /* removed */ - break; - } - disc.major = major(stb.st_rdev); - disc.minor = minor(stb.st_rdev); - disc.number =j; - disc.state = 0; - if (array.not_persistent==0) { - int dfd; - if (dv->writemostly == 1) - disc.state |= 1 << MD_DISK_WRITEMOSTLY; - dfd = dev_open(dv->devname, O_RDWR | O_EXCL|O_DIRECT); - if (tst->ss->add_to_super(tst, &disc, dfd, - dv->devname)) { - close(dfd); - return 1; - } - if (tst->ss->write_init_super(tst)) { - close(dfd); - return 1; - } - } else if (dv->re_add) { - /* this had better be raid1. - * As we are "--re-add"ing we must find a spare slot - * to fill. - */ - char *used = malloc(array.raid_disks); - memset(used, 0, array.raid_disks); - for (j=0; j< tst->max_devs; j++) { - mdu_disk_info_t disc2; - disc2.number = j; - if (ioctl(fd, GET_DISK_INFO, &disc2)) - continue; - if (disc2.major==0 && disc2.minor==0) - continue; - if (disc2.state & 8) /* removed */ - continue; - if (disc2.raid_disk < 0) - continue; - if (disc2.raid_disk > array.raid_disks) - continue; - used[disc2.raid_disk] = 1; - } - for (j=0 ; jdisposition == 'M') + continue; + pr_err("Cannot open %s: %s\n", + dv->devname, strerror(errno)); + goto abort; } - if (dv->writemostly == 1) - disc.state |= (1 << MD_DISK_WRITEMOSTLY); - if (tst->ss->external) { - /* add a disk - * to an external metadata container */ - struct mdinfo new_mdi; - struct mdinfo *sra; - int container_fd; - int devnum = fd2devnum(fd); - int dfd; - - container_fd = open_dev_excl(devnum); - if (container_fd < 0) { - fprintf(stderr, Name ": add failed for %s:" - " could not get exclusive access to container\n", - dv->devname); - tst->ss->free_super(tst); - return 1; - } - - dfd = dev_open(dv->devname, O_RDWR | O_EXCL|O_DIRECT); - if (mdmon_running(tst->container_dev)) - tst->update_tail = &tst->updates; - if (tst->ss->add_to_super(tst, &disc, dfd, - dv->devname)) { - close(dfd); - close(container_fd); - return 1; - } - if (tst->update_tail) - flush_metadata_updates(tst); + if (!frozen) { + if (sysfs_freeze_array(&info) == 1) + frozen = 1; else - tst->ss->sync_metadata(tst); - - sra = sysfs_read(container_fd, -1, 0); - if (!sra) { - fprintf(stderr, Name ": add failed for %s: sysfs_read failed\n", - dv->devname); - close(container_fd); - tst->ss->free_super(tst); - return 1; - } - sra->array.level = LEVEL_CONTAINER; - /* Need to set data_offset and component_size */ - tst->ss->getinfo_super(tst, &new_mdi, NULL); - new_mdi.disk.major = disc.major; - new_mdi.disk.minor = disc.minor; - new_mdi.recovery_start = 0; - /* Make sure fds are closed as they are O_EXCL which - * would block add_disk */ - tst->ss->free_super(tst); - if (sysfs_add_disk(sra, &new_mdi, 0) != 0) { - fprintf(stderr, Name ": add new device to external metadata" - " failed for %s\n", dv->devname); - close(container_fd); - sysfs_free(sra); - return 1; - } - ping_monitor_by_id(devnum); - sysfs_free(sra); - close(container_fd); - } else { - tst->ss->free_super(tst); - if (ioctl(fd, ADD_NEW_DISK, &disc)) { - fprintf(stderr, Name ": add new device failed for %s as %d: %s\n", - dv->devname, j, strerror(errno)); - return 1; - } + frozen = -1; } - if (verbose >= 0) - fprintf(stderr, Name ": added %s\n", dv->devname); + rv = Manage_add(fd, tfd, dv, tst, &array, + force, verbose, devname, update, + stb.st_rdev, array_size); + close(tfd); + tfd = -1; + if (rv < 0) + goto abort; + if (rv > 0) + count++; break; case 'r': /* hot remove */ if (subarray) { - fprintf(stderr, Name ": Cannot remove disks from a" + pr_err("Cannot remove disks from a" " \'member\' array, perform this" " operation on the parent container\n"); - if (sysfd >= 0) - close(sysfd); - return 1; - } - if (tst->ss->external) { - /* To remove a device from a container, we must - * check that it isn't in use in an array. - * This involves looking in the 'holders' - * directory - there must be just one entry, - * the container. - * To ensure that it doesn't get used as a - * hold spare while we are checking, we - * get an O_EXCL open on the container - */ - int dnum = fd2devnum(fd); - lfd = open_dev_excl(dnum); - if (lfd < 0) { - fprintf(stderr, Name - ": Cannot get exclusive access " - " to container - odd\n"); - if (sysfd >= 0) - close(sysfd); - return 1; - } - /* in the detached case it is not possible to - * check if we are the unique holder, so just - * rely on the 'detached' checks - */ - if (strcmp(dv->devname, "detached") == 0 || - sysfd >= 0 || - sysfs_unique_holder(dnum, stb.st_rdev)) - /* pass */; - else { - fprintf(stderr, Name - ": %s is %s, cannot remove.\n", - dnprintable, - errno == EEXIST ? "still in use": - "not a member"); - close(lfd); - return 1; - } - } - /* FIXME check that it is a current member */ - if (sysfd >= 0) { - /* device has been removed and we don't know - * the major:minor number - */ - int n = write(sysfd, "remove", 6); - if (n != 6) - err = -1; - else - err = 0; + rv = -1; + } else + rv = Manage_remove(tst, fd, dv, sysfd, + stb.st_rdev, verbose, + devname); + if (sysfd >= 0) close(sysfd); - sysfd = -1; - } else { - err = ioctl(fd, HOT_REMOVE_DISK, (unsigned long)stb.st_rdev); - if (err && errno == ENODEV) { - /* Old kernels rejected this if no personality - * registered */ - struct mdinfo *sra = sysfs_read(fd, 0, GET_DEVS); - struct mdinfo *dv = NULL; - if (sra) - dv = sra->devs; - for ( ; dv ; dv=dv->next) - if (dv->disk.major == (int)major(stb.st_rdev) && - dv->disk.minor == (int)minor(stb.st_rdev)) - break; - if (dv) - err = sysfs_set_str(sra, dv, - "state", "remove"); - else - err = -1; - if (sra) - sysfs_free(sra); - } - } - if (err) { - fprintf(stderr, Name ": hot remove failed " - "for %s: %s\n", dnprintable, - strerror(errno)); - if (lfd >= 0) - close(lfd); - return 1; - } - if (tst->ss->external) { - /* - * Before dropping our exclusive open we make an - * attempt at preventing mdmon from seeing an - * 'add' event before reconciling this 'remove' - * event. - */ - char *name = devnum2devname(fd2devnum(fd)); - - if (!name) { - fprintf(stderr, Name ": unable to get container name\n"); - return 1; - } - - ping_manager(name); - free(name); - } - if (lfd >= 0) - close(lfd); - count++; - if (verbose >= 0) - fprintf(stderr, Name ": hot removed %s from %s\n", - dnprintable, devname); + sysfd = -1; + if (rv < 0) + goto abort; + if (rv > 0) + count++; break; case 'f': /* set faulty */ @@ -1103,25 +1523,66 @@ int Manage_subdevs(char *devname, int fd, if ((sysfd >= 0 && write(sysfd, "faulty", 6) != 6) || (sysfd < 0 && ioctl(fd, SET_DISK_FAULTY, (unsigned long) stb.st_rdev))) { - fprintf(stderr, Name ": set device faulty failed for %s: %s\n", - dnprintable, strerror(errno)); + if (errno == EBUSY) + busy = 1; + pr_err("set device faulty failed for %s: %s\n", + dv->devname, strerror(errno)); if (sysfd >= 0) close(sysfd); - return 1; + goto abort; } if (sysfd >= 0) close(sysfd); sysfd = -1; count++; if (verbose >= 0) - fprintf(stderr, Name ": set %s faulty in %s\n", - dnprintable, devname); + pr_err("set %s faulty in %s\n", + dv->devname, devname); + break; + case 'R': /* Mark as replaceable */ + if (subarray) { + pr_err("Cannot replace disks in a" + " \'member\' array, perform this" + " operation on the parent container\n"); + rv = -1; + } else { + if (!frozen) { + if (sysfs_freeze_array(&info) == 1) + frozen = 1; + else + frozen = -1; + } + rv = Manage_replace(tst, fd, dv, + stb.st_rdev, verbose, + devname); + } + if (rv < 0) + goto abort; + if (rv > 0) + count++; + break; + case 'W': /* --with device that doesn't match */ + pr_err("No matching --replace device for --with %s\n", + dv->devname); + goto abort; + case 'w': /* --with device which was matched */ + rv = Manage_with(tst, fd, dv, + stb.st_rdev, verbose, devname); + if (rv < 0) + goto abort; break; } } + if (frozen > 0) + sysfs_set_str(&info, NULL, "sync_action","idle"); if (test && count == 0) return 2; return 0; + +abort: + if (frozen > 0) + sysfs_set_str(&info, NULL, "sync_action","idle"); + return !test && busy ? 2 : 1; } int autodetect(void) @@ -1137,43 +1598,41 @@ int autodetect(void) return rv; } -int Update_subarray(char *dev, char *subarray, char *update, struct mddev_ident *ident, int quiet) +int Update_subarray(char *dev, char *subarray, char *update, struct mddev_ident *ident, int verbose) { struct supertype supertype, *st = &supertype; int fd, rv = 2; memset(st, 0, sizeof(*st)); - fd = open_subarray(dev, subarray, st, quiet); + fd = open_subarray(dev, subarray, st, verbose < 0); if (fd < 0) return 2; if (!st->ss->update_subarray) { - if (!quiet) - fprintf(stderr, - Name ": Operation not supported for %s metadata\n", - st->ss->name); + if (verbose >= 0) + pr_err("Operation not supported for %s metadata\n", + st->ss->name); goto free_super; } - if (mdmon_running(st->devnum)) + if (mdmon_running(st->devnm)) st->update_tail = &st->updates; rv = st->ss->update_subarray(st, subarray, update, ident); if (rv) { - if (!quiet) - fprintf(stderr, Name ": Failed to update %s of subarray-%s in %s\n", + if (verbose >= 0) + pr_err("Failed to update %s of subarray-%s in %s\n", update, subarray, dev); } else if (st->update_tail) flush_metadata_updates(st); else st->ss->sync_metadata(st); - if (rv == 0 && strcmp(update, "name") == 0 && !quiet) - fprintf(stderr, - Name ": Updated subarray-%s name from %s, UUIDs may have changed\n", - subarray, dev); + if (rv == 0 && strcmp(update, "name") == 0 && verbose >= 0) + pr_err("Updated subarray-%s name from %s, UUIDs may have changed\n", + subarray, dev); free_super: st->ss->free_super(st); @@ -1182,9 +1641,8 @@ int Update_subarray(char *dev, char *subarray, char *update, struct mddev_ident return rv; } -/* Move spare from one array to another - * If adding to destination array fails - * add back to original array +/* Move spare from one array to another If adding to destination array fails + * add back to original array. * Returns 1 on success, 0 on failure */ int move_spare(char *from_devname, char *to_devname, dev_t devid) { @@ -1203,7 +1661,6 @@ int move_spare(char *from_devname, char *to_devname, dev_t devid) devlist.next = NULL; devlist.used = 0; - devlist.re_add = 0; devlist.writemostly = 0; devlist.devname = devname; sprintf(devname, "%d:%d", major(devid), minor(devid));