X-Git-Url: http://git.ipfire.org/?a=blobdiff_plain;f=Manage.c;h=21536f5ee4e92554a72b71bea5cfedd5d831d238;hb=783a4a93b9f01d1e9de98ea7d318a7885382cefa;hp=bdf2eaf3ea2b0da570e739ec4da47fea623ada2f;hpb=64a78416e36fe164fb5e33c0d3fd07416b00e7cc;p=thirdparty%2Fmdadm.git diff --git a/Manage.c b/Manage.c index bdf2eaf3..21536f5e 100644 --- a/Manage.c +++ b/Manage.c @@ -1,7 +1,7 @@ /* * mdadm - manage Linux "md" devices aka RAID arrays. * - * Copyright (C) 2001-2012 Neil Brown + * Copyright (C) 2001-2013 Neil Brown * * * This program is free software; you can redistribute it and/or modify @@ -27,10 +27,6 @@ #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) - int Manage_ro(char *devname, int fd, int readonly) { /* switch to readonly or rw @@ -40,17 +36,9 @@ int Manage_ro(char *devname, int fd, int readonly) * use RESTART_ARRAY_RW or STOP_ARRAY_RO * */ - mdu_array_info_t array; -#ifndef MDASSEMBLE struct mdinfo *mdi; -#endif int rv = 0; - if (md_get_version(fd) < 9000) { - pr_err("need md driver version 0.90.0 or later\n"); - return 1; - } -#ifndef MDASSEMBLE /* If this is an externally-managed array, we need to modify the * metadata_version so that mdmon doesn't undo our change. */ @@ -94,10 +82,9 @@ int Manage_ro(char *devname, int fd, int readonly) } goto out; } -#endif - if (ioctl(fd, GET_ARRAY_INFO, &array)) { - pr_err("%s does not appear to be active.\n", - devname); + + if (!md_array_active(fd)) { + pr_err("%s does not appear to be active.\n", devname); rv = 1; goto out; } @@ -118,15 +105,10 @@ int Manage_ro(char *devname, int fd, int readonly) } } out: -#ifndef MDASSEMBLE - if (mdi) - sysfs_free(mdi); -#endif + sysfs_free(mdi); return rv; } -#ifndef MDASSEMBLE - static void remove_devices(char *devnm, char *path) { /* @@ -170,186 +152,316 @@ static void remove_devices(char *devnm, char *path) free(path2); } -int Manage_runstop(char *devname, int fd, int runstop, - int verbose, int will_retry) +int Manage_run(char *devname, int fd, struct context *c) { - /* Run or stop the array. Array must already be configured - * 'Run' requires >= 0.90.0 - * 'will_retry' is only relevant for 'stop', and means - * that error messages are not wanted. + /* Run the array. Array must already be configured + * Requires >= 0.90.0 + */ + char nm[32], *nmp; + + 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. */ - mdu_param_t param; /* unused */ 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 (runstop == -1 && md_get_version(fd) < 9000) { - if (ioctl(fd, STOP_MD, 0) == 0) - return 0; - pr_err("stopping device %s " - "failed: %s\n", - devname, strerror(errno)); - return 1; + 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 = ((devname[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) { + /* Can't open, so something might be wrong. However it + * is a container, so we might be racing with mdmon, so + * retry for a bit. + */ + if (fd >= 0) + close(fd); + flush_mdmon(container); + count--; } - - if (md_get_version(fd) < 9000) { - pr_err("need md driver version 0.90.0 or later\n"); + 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)) { - if (verbose >= 0) - pr_err("failed to run array %s: %s\n", - devname, strerror(errno)); - return 1; - } - if (verbose >= 0) - pr_err("started %s\n", devname); - } else if (runstop < 0){ - struct map_ent *map = NULL; - struct stat stb; - struct mdinfo *mdi; - char devnm[32]; - char container[32]; + /* If this is an mdmon managed array, just write 'inactive' + * to the array state and let mdmon clear up. + */ + 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. - */ - 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_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; + /* This is mdmon managed. */ 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); + + /* As we had 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; + while (count && + (err = sysfs_set_str(mdi, NULL, + "array_state", + "inactive")) < 0 && + errno == EBUSY) { + usleep(200000); count--; } - if (fd < 0 || strcmp(fd2devnm(fd), devnm) != 0) { - if (fd >= 0) - close(fd); + 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("Cannot get exclusive access to %s:" - "Perhaps a running " - "process, mounted filesystem " - "or active volume group?\n", + pr_err("failed to completely stop %s: Device is busy\n", devname); - return 1; + rv = 1; + goto out; } - if (mdi && - mdi->array.level > 0 && - is_subarray(mdi->text_version)) { - int err; - /* This is mdmon managed. */ - close(fd); + } 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' + */ + ping_monitor(mdi->sys_name); - /* 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; - while (count && - (err = sysfs_set_str(mdi, NULL, - "array_state", - "inactive")) < 0 - && errno == EBUSY) { - usleep(200000); - count--; - } - if (err) { + /* 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("failed to stop array %s: %s\n", - devname, strerror(errno)); + pr_err("Cannot stop container %s: member %s still active\n", + devname, m->devnm); + free_mdstat(mds); rv = 1; goto out; } + } - /* Give monitor a chance to act */ - ping_monitor(mdi->text_version); + /* 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) { + 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; + + delay = 40; + while (rd1 > rd2 && delay > 0 && + sysfs_get_ll(mdi, NULL, "sync_max", &old_sync_max) == 0) { + /* must be in the critical section - wait a bit */ + delay -= 1; + usleep(100000); + } - 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' + if (sysfs_set_str(mdi, NULL, "sync_action", "frozen") != 0) + goto done; + /* Array is frozen */ + + 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; + if (sysfs_get_ll(mdi, NULL, "reshape_position", &position) != 0) { + /* reshape must have finished now */ + sysfs_set_str(mdi, NULL, "sync_action", "idle"); + goto done; + } + 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. */ - ping_monitor(mdi->sys_name); - - /* now check that there are no existing arrays - * which are members of this array + unsigned long long size = mdi->component_size; + size &= ~(chunk1-1); + size &= ~(chunk2-1); + /* rd1 must be smaller */ + /* Reshape may have progressed further backwards than + * recorded, so target even further back (hence "-1") */ - 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; - } + position = (position / sectors - 1) * sectors; + /* rd1 is always the conversion factor between 'sync' + * position and 'reshape' position. + * We read 1 "new" stripe worth of data from where-ever, + * and when write out that full stripe. + */ + sync_max = size - position/rd1; + } else { + /* Reshape will very likely be beyond position, and it may + * be too late to stop at '+1', so aim for '+2' + */ + position = (position / sectors + 2) * sectors; + sync_max = position/rd1; } - - /* 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. + 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. */ - 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) - fprintf(stderr, "Perhaps a running " - "process, mounted filesystem " - "or active volume group?\n"); + delay = 3000; + scfd = sysfs_open(mdi->sys_name, NULL, "sync_completed"); + while (scfd >= 0 && delay > 0 && old_sync_max > 0) { + unsigned long long max_completed; + 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; } - rv = 1; - goto out; + + if (sysfs_fd_get_two(scfd, &completed, + &max_completed) == 2 && + /* 'completed' sometimes reads as max-uulong */ + completed < max_completed && + (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 (!backwards && curr >= position) + break; + if (backwards && curr <= position) + break; + sysfs_wait(scfd, &delay); + } + if (scfd >= 0) + close(scfd); + + } +done: + + /* 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; + } + + if (get_linux_version() < 2006028) { /* 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... @@ -358,23 +470,21 @@ int Manage_runstop(char *devname, int fd, int runstop, ioctl(fd, BLKRRPART, 0); if (mdi) sysfs_uevent(mdi, "change"); + } - if (devnm[0] && - (stat("/dev/.udev", &stb) != 0 || - check_env("MDADM_NO_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); + 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: + sysfs_free(mdi); + return rv; } @@ -397,14 +507,14 @@ static void add_faulty(struct mddev_dev *dv, int fd, char disp) int remaining_disks; int i; - if (ioctl(fd, GET_ARRAY_INFO, &array) != 0) + if (md_get_array_info(fd, &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) + if (md_get_disk_info(fd, &disk) != 0) continue; if (disk.major == 0 && disk.minor == 0) continue; @@ -423,7 +533,7 @@ static void add_detached(struct mddev_dev *dv, int fd, char disp) int remaining_disks; int i; - if (ioctl(fd, GET_ARRAY_INFO, &array) != 0) + if (md_get_array_info(fd, &array) != 0) return; remaining_disks = array.nr_disks; @@ -431,7 +541,7 @@ static void add_detached(struct mddev_dev *dv, int fd, char disp) char buf[40]; int sfd; disk.number = i; - if (ioctl(fd, GET_DISK_INFO, &disk) != 0) + if (md_get_disk_info(fd, &disk) != 0) continue; if (disk.major == 0 && disk.minor == 0) continue; @@ -460,7 +570,7 @@ static void add_set(struct mddev_dev *dv, int fd, char set_char) int copies, set; int i; - if (ioctl(fd, GET_ARRAY_INFO, &array) != 0) + if (md_get_array_info(fd, &array) != 0) return; if (array.level != 10) return; @@ -473,7 +583,7 @@ static void add_set(struct mddev_dev *dv, int fd, char set_char) for (i = 0; i < MAX_DISKS && remaining_disks > 0; i++) { char buf[40]; disk.number = i; - if (ioctl(fd, GET_DISK_INFO, &disk) != 0) + if (md_get_disk_info(fd, &disk) != 0) continue; if (disk.major == 0 && disk.minor == 0) continue; @@ -519,37 +629,57 @@ int attempt_re_add(int fd, int tfd, struct mddev_dev *dv, 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 - ) + if (md_get_disk_info(fd, &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) + if (array->state & (1 << MD_SB_CLUSTERED)) { + /* extra flags are needed when adding to a cluster as + * there are two cases to distinguish + */ + if (dv->disposition == 'c') + disc.state |= (1 << MD_DISK_CANDIDATE); + else + disc.state |= (1 << MD_DISK_CLUSTER_ADD); + } + if (dv->writemostly == FlagSet) disc.state |= 1 << MD_DISK_WRITEMOSTLY; - if (dv->writemostly == 2) + if (dv->writemostly == FlagClear) disc.state &= ~(1 << MD_DISK_WRITEMOSTLY); + if (dv->failfast == FlagSet) + disc.state |= 1 << MD_DISK_FAILFAST; + if (dv->failfast == FlagClear) + disc.state &= ~(1 << MD_DISK_FAILFAST); remove_partitions(tfd); - if (update || dv->writemostly > 0) { + if (update || dv->writemostly != FlagDefault || + dv->failfast != FlagDefault) { 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); + pr_err("failed to open %s for superblock update during re-add\n", dv->devname); return -1; } - if (dv->writemostly == 1) + if (dv->writemostly == FlagSet) rv = dev_st->ss->update_super( dev_st, NULL, "writemostly", devname, verbose, 0, NULL); - if (dv->writemostly == 2) + if (dv->writemostly == FlagClear) rv = dev_st->ss->update_super( dev_st, NULL, "readwrite", devname, verbose, 0, NULL); + if (dv->failfast == FlagSet) + rv = dev_st->ss->update_super( + dev_st, NULL, "failfast", + devname, verbose, 0, NULL); + if (dv->failfast == FlagClear) + rv = dev_st->ss->update_super( + dev_st, NULL, "nofailfast", + devname, verbose, 0, NULL); if (update) rv = dev_st->ss->update_super( dev_st, NULL, update, @@ -558,8 +688,7 @@ int attempt_re_add(int fd, int tfd, struct mddev_dev *dv, rv = dev_st->ss->store_super(dev_st, tfd); close(tfd); if (rv != 0) { - pr_err("failed to update" - " superblock during re-add\n"); + pr_err("failed to update superblock during re-add\n"); return -1; } } @@ -585,10 +714,11 @@ skip_re_add: 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) + char *update, unsigned long rdev, unsigned long long array_size, + int raid_slot) { unsigned long long ldsize; - struct supertype *dev_st = NULL; + struct supertype *dev_st; int j; mdu_disk_info_t disc; @@ -599,27 +729,19 @@ int Manage_add(int fd, int tfd, struct mddev_dev *dv, return -1; } - if (tst->ss->validate_geometry( - tst, array->level, array->layout, - array->raid_disks, NULL, - ldsize >> 9, INVALID_SECTORS, NULL, NULL, 0) == 0) { + 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", + 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", + 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 (!tst->ss->external && array->major_version == 0) { if (ioctl(fd, HOT_ADD_DISK, rdev)==0) { if (verbose >= 0) pr_err("hot added %s\n", @@ -648,7 +770,7 @@ int Manage_add(int fd, int tfd, struct mddev_dev *dv, char *dev; int dfd; disc.number = j; - if (ioctl(fd, GET_DISK_INFO, &disc)) + if (md_get_disk_info(fd, &disc)) continue; if (disc.major==0 && disc.minor==0) continue; @@ -670,7 +792,8 @@ int Manage_add(int fd, int tfd, struct mddev_dev *dv, break; } /* FIXME this is a bad test to be using */ - if (!tst->sb && dv->disposition != 'a') { + 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 @@ -681,7 +804,9 @@ int Manage_add(int fd, int tfd, struct mddev_dev *dv, } /* Make sure device is large enough */ - if (tst->ss->avail_size(tst, ldsize/512, INVALID_SECTORS) < + if (dv->disposition != 'j' && /* skip size check for Journal */ + tst->sb && + tst->ss->avail_size(tst, ldsize/512, INVALID_SECTORS) < array_size) { if (dv->disposition == 'M') return 0; @@ -696,20 +821,19 @@ int Manage_add(int fd, int tfd, struct mddev_dev *dv, * simply re-add it. */ - if (array->not_persistent==0) { + if (array->not_persistent == 0) { dev_st = dup_super(tst); dev_st->ss->load_super(dev_st, tfd, NULL); - } - if (dev_st && dev_st->sb) { - 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 (dev_st->sb && dv->disposition != 'S') { + int rv; + + 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) @@ -727,9 +851,9 @@ int Manage_add(int fd, int tfd, struct mddev_dev *dv, int d; int found = 0; - for (d = 0; d < MAX_DISKS && found < array->active_disks; d++) { + for (d = 0; d < MAX_DISKS && found < array->nr_disks; d++) { disc.number = d; - if (ioctl(fd, GET_DISK_INFO, &disc)) + if (md_get_disk_info(fd, &disc)) continue; if (disc.major == 0 && disc.minor == 0) continue; @@ -740,6 +864,7 @@ int Manage_add(int fd, int tfd, struct mddev_dev *dv, } array_failed = !enough(array->level, array->raid_disks, array->layout, 1, avail); + free(avail); } else array_failed = 0; if (array_failed) { @@ -769,7 +894,7 @@ int Manage_add(int fd, int tfd, struct mddev_dev *dv, */ for (j = array->raid_disks; j < tst->max_devs; j++) { disc.number = j; - if (ioctl(fd, GET_DISK_INFO, &disc)) + if (md_get_disk_info(fd, &disc)) break; if (disc.major==0 && disc.minor==0) break; @@ -778,12 +903,41 @@ int Manage_add(int fd, int tfd, struct mddev_dev *dv, } disc.major = major(rdev); disc.minor = minor(rdev); - disc.number =j; + if (raid_slot < 0) + disc.number = j; + else + disc.number = raid_slot; disc.state = 0; + + /* only add journal to array that supports journaling */ + if (dv->disposition == 'j') { + struct mdinfo *mdp; + + mdp = sysfs_read(fd, NULL, GET_ARRAY_STATE); + if (!mdp) { + pr_err("%s unable to read array state.\n", devname); + return -1; + } + + if (mdp->array_state != ARRAY_READONLY) { + sysfs_free(mdp); + pr_err("%s is not readonly, cannot add journal.\n", devname); + return -1; + } + + sysfs_free(mdp); + + disc.raid_disk = 0; + } + if (array->not_persistent==0) { int dfd; - if (dv->writemostly == 1) + if (dv->disposition == 'j') + disc.state |= (1 << MD_DISK_JOURNAL) | (1 << MD_DISK_SYNC); + if (dv->writemostly == FlagSet) disc.state |= 1 << MD_DISK_WRITEMOSTLY; + if (dv->failfast == FlagSet) + disc.state |= 1 << MD_DISK_FAILFAST; dfd = dev_open(dv->devname, O_RDWR | O_EXCL|O_DIRECT); if (tst->ss->add_to_super(tst, &disc, dfd, dv->devname, INVALID_SECTORS)) @@ -799,7 +953,7 @@ int Manage_add(int fd, int tfd, struct mddev_dev *dv, for (j = 0; j < tst->max_devs; j++) { mdu_disk_info_t disc2; disc2.number = j; - if (ioctl(fd, GET_DISK_INFO, &disc2)) + if (md_get_disk_info(fd, &disc2)) continue; if (disc2.major==0 && disc2.minor==0) continue; @@ -819,8 +973,18 @@ int Manage_add(int fd, int tfd, struct mddev_dev *dv, } free(used); } - if (dv->writemostly == 1) + + if (array->state & (1 << MD_SB_CLUSTERED)) { + if (dv->disposition == 'c') + disc.state |= (1 << MD_DISK_CANDIDATE); + else + disc.state |= (1 << MD_DISK_CLUSTER_ADD); + } + + if (dv->writemostly == FlagSet) disc.state |= (1 << MD_DISK_WRITEMOSTLY); + if (dv->failfast == FlagSet) + disc.state |= (1 << MD_DISK_FAILFAST); if (tst->ss->external) { /* add a disk * to an external metadata container */ @@ -834,8 +998,7 @@ int Manage_add(int fd, int tfd, struct mddev_dev *dv, container_fd = open_dev_excl(devnm); if (container_fd < 0) { - pr_err("add failed for %s:" - " could not get exclusive access to container\n", + pr_err("add failed for %s: could not get exclusive access to container\n", dv->devname); tst->ss->free_super(tst); return -1; @@ -874,8 +1037,7 @@ int Manage_add(int fd, int tfd, struct mddev_dev *dv, * 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); + pr_err("add new device to external metadata failed for %s\n", dv->devname); close(container_fd); sysfs_free(sra); return -1; @@ -886,10 +1048,20 @@ int Manage_add(int fd, int tfd, struct mddev_dev *dv, } 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)); + if (dv->disposition == 'j') + pr_err("Failed to hot add %s as journal, " + "please try restart %s.\n", dv->devname, devname); + else + pr_err("add new device failed for %s as %d: %s\n", + dv->devname, j, strerror(errno)); return -1; } + if (dv->disposition == 'j') { + pr_err("Journal added successfully, making %s read-write\n", devname); + if (Manage_ro(devname, fd, -1)) + pr_err("Failed to make %s read-write\n", devname); + } + } if (verbose >= 0) pr_err("added %s\n", dv->devname); @@ -897,7 +1069,7 @@ int Manage_add(int fd, int tfd, struct mddev_dev *dv, } int Manage_remove(struct supertype *tst, int fd, struct mddev_dev *dv, - int sysfd, unsigned long rdev, int verbose, char *devname) + int sysfd, unsigned long rdev, int force, int verbose, char *devname) { int lfd = -1; int err; @@ -917,8 +1089,7 @@ int Manage_remove(struct supertype *tst, int fd, struct mddev_dev *dv, strcpy(devnm, fd2devnm(fd)); lfd = open_dev_excl(devnm); if (lfd < 0) { - pr_err("Cannot get exclusive access " - " to container - odd\n"); + pr_err("Cannot get exclusive access to container - odd\n"); return -1; } /* We may not be able to check on holders in @@ -930,19 +1101,34 @@ int Manage_remove(struct supertype *tst, int fd, struct mddev_dev *dv, */ 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; + else { + /* + * The drive has already been set to 'faulty', however + * monitor might not have had time to process it and the + * drive might still have an entry in the 'holders' + * directory. Try a few times to avoid a false error + */ + int count = 20; + + do { + ret = sysfs_unique_holder(devnm, rdev); + if (ret < 2) + break; + usleep(100 * 1000); /* 100ms */ + } while (--count > 0); + + 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 */ @@ -950,13 +1136,9 @@ int Manage_remove(struct supertype *tst, int fd, struct mddev_dev *dv, /* 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; + err = sys_hot_remove_disk(sysfd, force); } else { - err = ioctl(fd, HOT_REMOVE_DISK, rdev); + err = hot_remove_disk(fd, rdev, force); if (err && errno == ENODEV) { /* Old kernels rejected this if no personality * is registered */ @@ -973,13 +1155,11 @@ int Manage_remove(struct supertype *tst, int fd, struct mddev_dev *dv, "state", "remove"); else err = -1; - if (sra) - sysfs_free(sra); + sysfs_free(sra); } } if (err) { - pr_err("hot remove failed " - "for %s: %s\n", dv->devname, + pr_err("hot remove failed for %s: %s\n", dv->devname, strerror(errno)); if (lfd >= 0) close(lfd); @@ -1099,7 +1279,7 @@ int Manage_with(struct supertype *tst, int fd, struct mddev_dev *dv, "slot", dv->used); if (rv) { sysfs_free(mdi); - pr_err("Failed to %s as preferred replacement.\n", + pr_err("Failed to set %s as preferred replacement.\n", dv->devname); return -1; } @@ -1123,6 +1303,8 @@ int Manage_subdevs(char *devname, int fd, * 'a' - add the device * try HOT_ADD_DISK * If that fails EINVAL, try ADD_NEW_DISK + * 'S' - add the device as a spare - don't try re-add + * 'j' - add the device as a journal device * 'A' - re-add the device * 'r' - remove the device: HOT_REMOVE_DISK * device can be 'faulty' or 'detached' in which case all @@ -1141,6 +1323,7 @@ int Manage_subdevs(char *devname, int fd, * variant on 'A' * 'F' - Another variant of 'A', where the device was faulty * so must be removed from the array first. + * 'c' - confirm the device as found (for clustered environments) * * For 'f' and 'r', the device can also be a kernel-internal * name such as 'sdb'. @@ -1148,23 +1331,26 @@ int Manage_subdevs(char *devname, int fd, mdu_array_info_t array; unsigned long long array_size; struct mddev_dev *dv; - struct stat stb; int tfd = -1; struct supertype *tst; char *subarray = NULL; int sysfd = -1; int count = 0; /* number of actions taken */ struct mdinfo info; + struct mdinfo devinfo; int frozen = 0; int busy = 0; + int raid_slot = -1; - if (ioctl(fd, GET_ARRAY_INFO, &array)) { - pr_err("Cannot get array info for %s\n", - devname); + if (sysfs_init(&info, fd, NULL)) { + pr_err("sysfs not availabile for %s\n", devname); goto abort; } - sysfs_init(&info, fd, NULL); + if (md_get_array_info(fd, &array)) { + pr_err("Cannot get array info for %s\n", devname); + goto abort; + } /* array.size is only 32 bits and may be truncated. * So read from sysfs if possible, and record number of sectors */ @@ -1180,16 +1366,26 @@ int Manage_subdevs(char *devname, int fd, goto abort; } - stb.st_rdev = 0; for (dv = devlist; dv; dv = dv->next) { + dev_t rdev = 0; /* device to add/remove etc */ int rv; + int mj,mn; + + raid_slot = -1; + if (dv->disposition == 'c') { + rv = parse_cluster_confirm_arg(dv->devname, + &dv->devname, + &raid_slot); + if (rv) { + pr_err("Could not get the devname of cluster\n"); + goto abort; + } + } 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", + if (dv->disposition != 'A' && dv->disposition != 'r') { + pr_err("%s only meaningful with -r or --re-add, not -%c\n", dv->devname, dv->disposition); goto abort; } @@ -1199,8 +1395,7 @@ int Manage_subdevs(char *devname, int fd, } if (strcmp(dv->devname, "detached") == 0) { if (dv->disposition != 'r' && dv->disposition != 'f') { - pr_err("%s only meaningful " - "with -r of -f, not -%c\n", + pr_err("%s only meaningful with -r of -f, not -%c\n", dv->devname, dv->disposition); goto abort; } @@ -1209,16 +1404,20 @@ int Manage_subdevs(char *devname, int fd, } if (strcmp(dv->devname, "missing") == 0) { - struct mddev_dev *add_devlist = NULL; + struct mddev_dev *add_devlist; struct mddev_dev **dp; + if (dv->disposition == 'c') { + rv = ioctl(fd, CLUSTERED_DISK_NACK, NULL); + break; + } + if (dv->disposition != 'A') { - pr_err("'missing' only meaningful " - "with --re-add\n"); + pr_err("'missing' only meaningful with --re-add\n"); goto abort; } add_devlist = conf_get_devs(); if (add_devlist == NULL) { - pr_err("no devices to scan for missing members."); + pr_err("no devices to scan for missing members.\n"); continue; } for (dp = &add_devlist; *dp; dp = & (*dp)->next) @@ -1265,8 +1464,7 @@ int Manage_subdevs(char *devname, int fd, int found = 0; char dname[55]; if (dv->disposition != 'r' && dv->disposition != 'f') { - pr_err("%s only meaningful " - "with -r or -f, not -%c\n", + pr_err("%s only meaningful with -r or -f, not -%c\n", dv->devname, dv->disposition); goto abort; } @@ -1275,10 +1473,9 @@ int Manage_subdevs(char *devname, int fd, sysfd = sysfs_open(fd2devnm(fd), dname, "block/dev"); if (sysfd >= 0) { char dn[20]; - int mj,mn; if (sysfs_fd_get_str(sysfd, dn, 20) > 0 && sscanf(dn, "%d:%d", &mj,&mn) == 2) { - stb.st_rdev = makedev(mj,mn); + rdev = makedev(mj,mn); found = 1; } close(sysfd); @@ -1287,29 +1484,29 @@ int Manage_subdevs(char *devname, int fd, if (!found) { sysfd = sysfs_open(fd2devnm(fd), dname, "state"); if (sysfd < 0) { - pr_err("%s does not appear " - "to be a component of %s\n", + pr_err("%s does not appear to be a component of %s\n", dv->devname, devname); goto abort; } } + } else if ((dv->disposition == 'r' || + dv->disposition == 'f') && + get_maj_min(dv->devname, &mj, &mn)) { + /* for 'fail' and 'remove', the device might + * not exist. + */ + rdev = makedev(mj, mn); } else { tfd = dev_open(dv->devname, O_RDONLY); - if (tfd >= 0) - fstat(tfd, &stb); - else { + if (tfd >= 0) { + fstat_is_blkdev(tfd, dv->devname, &rdev); + close(tfd); + } else { 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 (!stat_is_blkdev(dv->devname, &rdev)) { 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') @@ -1333,20 +1530,32 @@ int Manage_subdevs(char *devname, int fd, dv->devname, dv->disposition); goto abort; case 'a': + case 'S': /* --add-spare */ + case 'j': /* --add-journal */ case 'A': case 'M': /* --re-add missing */ case 'F': /* --re-add faulty */ + case 'c': /* --cluster-confirm */ /* add the device */ if (subarray) { - pr_err("Cannot add disks to a" - " \'member\' array, perform this" - " operation on the parent container\n"); + pr_err("Cannot add disks to a \'member\' array, perform this operation on the parent container\n"); goto abort; } + + /* Let's first try to write re-add to sysfs */ + if (rdev != 0 && + (dv->disposition == 'A' || dv->disposition == 'F')) { + sysfs_init_dev(&devinfo, rdev); + if (sysfs_set_str(&info, &devinfo, "state", "re-add") == 0) { + pr_err("re-add %s to %s succeed\n", + dv->devname, info.sys_name); + break; + } + } + if (dv->disposition == 'F') /* Need to remove first */ - ioctl(fd, HOT_REMOVE_DISK, - (unsigned long)stb.st_rdev); + hot_remove_disk(fd, rdev, force); /* Make sure it isn't in use (in 2.6 or later) */ tfd = dev_open(dv->devname, O_RDONLY|O_EXCL); if (tfd >= 0) { @@ -1356,7 +1565,7 @@ int Manage_subdevs(char *devname, int fd, */ close(tfd); tfd = dev_open(dv->devname, O_RDONLY); - } + } if (tfd < 0) { if (dv->disposition == 'M') continue; @@ -1372,7 +1581,7 @@ int Manage_subdevs(char *devname, int fd, } rv = Manage_add(fd, tfd, dv, tst, &array, force, verbose, devname, update, - stb.st_rdev, array_size); + rdev, array_size, raid_slot); close(tfd); tfd = -1; if (rv < 0) @@ -1384,13 +1593,11 @@ int Manage_subdevs(char *devname, int fd, case 'r': /* hot remove */ if (subarray) { - pr_err("Cannot remove disks from a" - " \'member\' array, perform this" - " operation on the parent container\n"); + pr_err("Cannot remove disks from a \'member\' array, perform this operation on the parent container\n"); rv = -1; } else rv = Manage_remove(tst, fd, dv, sysfd, - stb.st_rdev, verbose, + rdev, verbose, force, devname); if (sysfd >= 0) close(sysfd); @@ -1405,7 +1612,7 @@ int Manage_subdevs(char *devname, int fd, /* FIXME check current member */ if ((sysfd >= 0 && write(sysfd, "faulty", 6) != 6) || (sysfd < 0 && ioctl(fd, SET_DISK_FAULTY, - (unsigned long) stb.st_rdev))) { + rdev))) { if (errno == EBUSY) busy = 1; pr_err("set device faulty failed for %s: %s\n", @@ -1424,9 +1631,7 @@ int Manage_subdevs(char *devname, int fd, 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"); + pr_err("Cannot replace disks in a \'member\' array, perform this operation on the parent container\n"); rv = -1; } else { if (!frozen) { @@ -1436,7 +1641,7 @@ int Manage_subdevs(char *devname, int fd, frozen = -1; } rv = Manage_replace(tst, fd, dv, - stb.st_rdev, verbose, + rdev, verbose, devname); } if (rv < 0) @@ -1450,7 +1655,7 @@ int Manage_subdevs(char *devname, int fd, goto abort; case 'w': /* --with device which was matched */ rv = Manage_with(tst, fd, dv, - stb.st_rdev, verbose, devname); + rdev, verbose, devname); if (rv < 0) goto abort; break; @@ -1544,7 +1749,8 @@ int move_spare(char *from_devname, char *to_devname, dev_t devid) devlist.next = NULL; devlist.used = 0; - devlist.writemostly = 0; + devlist.writemostly = FlagDefault; + devlist.failfast = FlagDefault; devlist.devname = devname; sprintf(devname, "%d:%d", major(devid), minor(devid)); @@ -1565,4 +1771,3 @@ int move_spare(char *from_devname, char *to_devname, dev_t devid) close(fd2); return 0; } -#endif