/*
* mdadm - manage Linux "md" devices aka RAID arrays.
*
- * Copyright (C) 2001-2006 Neil Brown <neilb@suse.de>
+ * Copyright (C) 2001-2009 Neil Brown <neilb@suse.de>
*
*
* This program is free software; you can redistribute it and/or modify
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*
* Author: Neil Brown
- * Email: <neilb@cse.unsw.edu.au>
- * Paper: Neil Brown
- * School of Computer Science and Engineering
- * The University of New South Wales
- * Sydney, 2052
- * Australia
+ * Email: <neilb@suse.de>
*/
#include "mdadm.h"
#include "md_u.h"
#include "md_p.h"
+#include <ctype.h>
#define REGISTER_DEV _IO (MD_MAJOR, 1)
#define START_MD _IO (MD_MAJOR, 2)
*
*/
mdu_array_info_t array;
+#ifndef MDASSEMBLE
struct mdinfo *mdi;
+#endif
if (md_get_version(fd) < 9000) {
fprintf(stderr, Name ": 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
* metadata_version so that mdmon doesn't undo our change.
*/
}
return 0;
}
-
+#endif
if (ioctl(fd, GET_ARRAY_INFO, &array)) {
fprintf(stderr, Name ": %s does not appear to be active.\n",
devname);
#ifndef MDASSEMBLE
+static void remove_devices(int devnum, char *path)
+{
+ /*
+ * Remove names at 'path' - possibly with
+ * partition suffixes - which link to the 'standard'
+ * name for devnum. These were probably created
+ * by mdadm when the array was assembled.
+ */
+ char base[40];
+ char *path2;
+ char link[1024];
+ int n;
+ int part;
+ char *be;
+ char *pe;
+
+ if (!path)
+ return;
+
+ if (devnum >= 0)
+ sprintf(base, "/dev/md%d", devnum);
+ else
+ sprintf(base, "/dev/md_d%d", -1-devnum);
+ be = base + strlen(base);
+
+ path2 = malloc(strlen(path)+20);
+ strcpy(path2, path);
+ pe = path2 + strlen(path2);
+
+ for (part = 0; part < 16; part++) {
+ if (part) {
+ sprintf(be, "p%d", part);
+
+ if (isdigit(pe[-1]))
+ sprintf(pe, "p%d", part);
+ else
+ sprintf(pe, "%d", part);
+ }
+ n = readlink(path2, link, sizeof(link));
+ if (n && (int)strlen(base) == n &&
+ strncmp(link, base, n) == 0)
+ unlink(path2);
+ }
+ free(path2);
+}
+
+
int Manage_runstop(char *devname, int fd, int runstop, int quiet)
{
/* Run or stop the array. array must already be configured
struct map_ent *map = NULL;
struct stat stb;
struct mdinfo *mdi;
+ int devnum;
/* If this is an mdmon managed array, just write 'inactive'
* to the array state and let mdmon clear up.
*/
+ devnum = fd2devnum(fd);
mdi = sysfs_read(fd, -1, GET_LEVEL|GET_VERSION);
if (mdi &&
mdi->array.level > 0 &&
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);
+
+ /* 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;
+ }
}
- if (mdi)
- sysfs_free(mdi);
if (fd >= 0 && ioctl(fd, STOP_ARRAY, NULL)) {
if (quiet == 0) {
"process, mounted filesystem "
"or active volume group?\n");
}
+ if (mdi)
+ sysfs_free(mdi);
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...
+ */
+ 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 (quiet <= 0)
fprintf(stderr, Name ": stopped %s\n", devname);
- if (fd >= 0 && fstat(fd, &stb) == 0) {
- int devnum;
- if (major(stb.st_rdev) == MD_MAJOR)
- devnum = minor(stb.st_rdev);
- else
- devnum = -1-(minor(stb.st_rdev)>>6);
- map_delete(&map, devnum);
- map_write(map);
- map_free(map);
- }
+ map_lock(&map);
+ map_remove(&map, devnum);
+ map_unlock(&map);
}
return 0;
}
return 0;
}
-int Manage_reconfig(char *devname, int fd, int layout)
-{
- 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));
- return 1;
- }
- info.layout = layout;
- printf("layout set to %d\n", info.layout);
- if (ioctl(fd, SET_ARRAY_INFO, &info) != 0) {
- fprintf(stderr, Name ": Cannot set layout for %s: %s\n",
- devname, strerror(errno));
- return 1;
- }
- return 0;
-}
-
int Manage_subdevs(char *devname, int fd,
- mddev_dev_t devlist, int verbose)
+ mddev_dev_t devlist, int verbose, int test)
{
/* do something to each dev.
* devmode can be
* 'f' - set the device faulty SET_DISK_FAULTY
* device can be 'detached' in which case any device that
* is inaccessible will be marked faulty.
+ * For 'f' and 'r', the device can also be a kernel-internal
+ * name such as 'sdb'.
*/
+ mddev_dev_t add_devlist = NULL;
mdu_array_info_t array;
mdu_disk_info_t disc;
unsigned long long array_size;
mddev_dev_t dv, next = NULL;
struct stat stb;
int j, jnext = 0;
- int tfd;
+ int tfd = -1;
struct supertype *st, *tst;
int duuid[4];
int ouuid[4];
int lfd = -1;
+ int sysfd = -1;
+ int count = 0; /* number of actions taken */
if (ioctl(fd, GET_ARRAY_INFO, &array)) {
fprintf(stderr, Name ": cannot get array info for %s\n",
return 1;
}
+ 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;
next = dv->next;
return 1;
}
for (; j < array.raid_disks + array.nr_disks ; j++) {
+ unsigned dev;
disc.number = j;
if (ioctl(fd, GET_DISK_INFO, &disc))
continue;
continue;
if ((disc.state & 1) == 0) /* faulty */
continue;
- stb.st_rdev = makedev(disc.major, disc.minor);
+ dev = makedev(disc.major, disc.minor);
+ if (stb.st_rdev == dev)
+ /* already did that one */
+ continue;
+ stb.st_rdev = dev;
next = dv;
- jnext = j+1;
+ /* same slot again next time - things might
+ * have reshuffled */
+ jnext = j;
sprintf(dvname,"%d:%d", disc.major, disc.minor);
dnprintable = dvname;
break;
}
for (; j < array.raid_disks + array.nr_disks; j++) {
int sfd;
+ unsigned dev;
disc.number = j;
if (ioctl(fd, GET_DISK_INFO, &disc))
continue;
continue;
if (errno != ENXIO)
continue;
- stb.st_rdev = makedev(disc.major, disc.minor);
+ dev = makedev(disc.major, disc.minor);
+ if (stb.st_rdev == dev)
+ /* already did that one */
+ continue;
+ stb.st_rdev = dev;
next = dv;
- jnext = j+1;
+ /* same slot again next time - things might
+ * have reshuffled */
+ jnext = j;
dnprintable = dvname;
break;
}
if (jnext == 0)
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;
+ }
+ if (add_devlist == NULL)
+ add_devlist = conf_get_devs();
+ if (add_devlist == NULL) {
+ fprintf(stderr, Name ": 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) {
+ /* 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 "
+ "with -r or -f, not -%c\n",
+ dv->devname, dv->disposition);
+ return 1;
+ }
+
+ sprintf(dname, "dev-%s", dv->devname);
+ sysfd = sysfs_open(fd2devnum(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);
+ found = 1;
+ }
+ close(sysfd);
+ sysfd = -1;
+ }
+ if (!found) {
+ sysfd = sysfs_open(fd2devnum(fd), dname, "state");
+ if (sysfd < 0) {
+ fprintf(stderr, Name ": %s does not appear "
+ "to be a component of %s\n",
+ dv->devname, devname);
+ return 1;
+ }
+ }
} else {
j = 0;
- if (stat(dv->devname, &stb)) {
- fprintf(stderr, Name ": cannot find %s: %s\n",
- dv->devname, strerror(errno));
- return 1;
+ 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
+ */
+ ;
+ 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;
+ }
+ close(tfd);
+ tfd = -1;
}
if ((stb.st_mode & S_IFMT) != S_IFBLK) {
fprintf(stderr, Name ": %s is not a "
return 1;
}
/* Make sure it isn't in use (in 2.6 or later) */
- tfd = open(dv->devname, O_RDONLY|O_EXCL|O_DIRECT);
+ 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;
}
- remove_partitions(tfd);
st = dup_super(tst);
if (array.not_persistent==0)
st->ss->load_super(st, tfd, NULL);
- if (!get_dev_size(tfd, dv->devname, &ldsize)) {
+ 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);
- return 1;
+ tfd = -1;
+ continue;
}
- close(tfd);
-
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",
- dv->devname);
+ add_dev);
continue;
}
fprintf(stderr, Name ": hot add failed for %s: %s\n",
- dv->devname, strerror(errno));
+ add_dev, strerror(errno));
return 1;
}
if (array.not_persistent == 0 || tst->ss->external) {
- /* Make sure device is large enough */
- if (tst->ss->avail_size(tst, ldsize/512) <
- array_size) {
- fprintf(stderr, Name ": %s not large enough to join array\n",
- dv->devname);
- return 1;
- }
/* 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->ss->external) {
+ if (tst->sb)
+ /* already loaded */;
+ else if (tst->ss->external) {
tst->ss->load_super(tst, fd, NULL);
} else for (j = 0; j < tst->max_devs; j++) {
char *dev;
}
/* FIXME this is a bad test to be using */
if (!tst->sb) {
+ close(tfd);
fprintf(stderr, Name ": cannot find valid superblock in this array - HELP\n");
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.
disc.number = mdi.disk.number;
disc.raid_disk = mdi.disk.raid_disk;
disc.state = mdi.disk.state;
- if (dv->writemostly)
+ if (dv->writemostly == 1)
disc.state |= 1 << MD_DISK_WRITEMOSTLY;
- if (ioctl(fd, ADD_NEW_DISK, &disc) == 0) {
+ if (dv->writemostly == 2)
+ disc.state &= ~(1 << MD_DISK_WRITEMOSTLY);
+ remove_partitions(tfd);
+ close(tfd);
+ tfd = -1;
+ /* don't even try if disk is marked as faulty */
+ errno = 0;
+ if ((disc.state & 1) == 0 &&
+ ioctl(fd, ADD_NEW_DISK, &disc) == 0) {
if (verbose >= 0)
- fprintf(stderr, Name ": re-added %s\n", dv->devname);
+ 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;
+ }
/* fall back on normal-add */
}
}
+ 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;
+ }
} 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*/
+ if (tfd >= 0) {
+ remove_partitions(tfd);
+ close(tfd);
+ tfd = -1;
+ }
/* 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
disc.state = 0;
if (array.not_persistent==0 || tst->ss->external) {
int dfd;
- if (dv->writemostly)
+ if (dv->writemostly == 1)
disc.state |= 1 << MD_DISK_WRITEMOSTLY;
- dfd = open(dv->devname, O_RDWR | O_EXCL|O_DIRECT);
- tst->ss->add_to_super(tst, &disc, dfd,
- dv->devname);
+ 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;
+ }
/* write_init_super will close 'dfd' */
if (tst->ss->external)
/* mdmon will write the metadata */
disc.state |= (1<<MD_DISK_SYNC);
break;
}
+ free(used);
}
- if (dv->writemostly)
+ if (dv->writemostly == 1)
disc.state |= (1 << MD_DISK_WRITEMOSTLY);
if (tst->ss->external) {
/* add a disk to an external metadata container
tst->ss->getinfo_super(tst, &new_mdi);
new_mdi.disk.major = disc.major;
new_mdi.disk.minor = disc.minor;
- if (sysfs_add_disk(sra, &new_mdi) != 0) {
+ new_mdi.recovery_start = 0;
+ 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);
fprintf(stderr, Name ": 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) {
fprintf(stderr, Name
": Cannot get exclusive access "
" to container - odd\n");
+ if (sysfd >= 0)
+ close(sysfd);
return 1;
}
- if (!sysfs_unique_holder(dnum, stb.st_rdev)) {
+ /* 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,
}
}
/* FIXME check that it is a current member */
- 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 == major(stb.st_rdev) &&
- dv->disk.minor == minor(stb.st_rdev))
- break;
- if (dv)
- err = sysfs_set_str(sra, dv,
- "state", "remove");
- else
+ 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;
- if (sra)
- sysfs_free(sra);
+ else
+ err = 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 "
close(lfd);
return 1;
}
- close(lfd);
+ 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\n",
- dnprintable);
+ fprintf(stderr, Name ": hot removed %s from %s\n",
+ dnprintable, devname);
break;
case 'f': /* set faulty */
/* FIXME check current member */
- if (ioctl(fd, SET_DISK_FAULTY, (unsigned long) stb.st_rdev)) {
+ 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 (sysfd >= 0)
+ close(sysfd);
return 1;
}
+ if (sysfd >= 0)
+ close(sysfd);
+ sysfd = -1;
+ count++;
if (verbose >= 0)
fprintf(stderr, Name ": set %s faulty in %s\n",
dnprintable, devname);
break;
}
}
+ if (test && count == 0)
+ return 2;
return 0;
-
}
int autodetect(void)
}
return rv;
}
+
+int Update_subarray(char *dev, char *subarray, char *update, mddev_ident_t ident, int quiet)
+{
+ struct supertype supertype, *st = &supertype;
+ int fd, rv = 2;
+
+ memset(st, 0, sizeof(*st));
+ if (snprintf(st->subarray, sizeof(st->subarray), "%s", subarray) >=
+ (signed)sizeof(st->subarray)) {
+ if (!quiet)
+ fprintf(stderr,
+ Name ": Input overflow for subarray '%s' > %zu bytes\n",
+ subarray, sizeof(st->subarray) - 1);
+ return 2;
+ }
+
+ fd = open_subarray(dev, st, quiet);
+ 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);
+ goto free_super;
+ }
+
+ if (mdmon_running(st->devnum))
+ st->update_tail = &st->updates;
+
+ rv = st->ss->update_subarray(st, update, ident);
+
+ if (rv) {
+ if (!quiet)
+ fprintf(stderr, Name ": 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);
+
+ free_super:
+ st->ss->free_super(st);
+ close(fd);
+
+ return rv;
+}
#endif
projects / thirdparty / mdadm.git / blobdiff