X-Git-Url: http://git.ipfire.org/?p=thirdparty%2Fmdadm.git;a=blobdiff_plain;f=Manage.c;h=65235bc1836f5e02c22c3e703cea2f39789b8023;hp=5219e6eb8760ec2b287f10f3d6567297bbe72838;hb=1124b3cf29dad544e8a5aa01e5f9e94e7be1765a;hpb=68c7d6d790f856b4e8301d0afa7fc6873a0d4bb8

diff --git a/Manage.c b/Manage.c
index 5219e6eb..65235bc1 100644
--- a/Manage.c
+++ b/Manage.c
@@ -30,6 +30,7 @@
 #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)
@@ -45,11 +46,57 @@ int Manage_ro(char *devname, int fd, int readonly)
 	 *
 	 */
 	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.
+	 */
+	mdi = sysfs_read(fd, -1, GET_LEVEL|GET_VERSION);
+	if (mdi &&
+	    mdi->array.major_version == -1 &&
+	    mdi->array.level > 0 &&
+	    is_subarray(mdi->text_version)) {
+		char vers[64];
+		strcpy(vers, "external:");
+		strcat(vers, mdi->text_version);
+		if (readonly > 0) {
+			int rv;
+			/* We set readonly ourselves. */
+			vers[9] = '-';
+			sysfs_set_str(mdi, NULL, "metadata_version", vers);
+
+			close(fd);
+			rv = sysfs_set_str(mdi, NULL, "array_state", "readonly");
+
+			if (rv < 0) {
+				fprintf(stderr, Name ": 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;
+			}
+		} else {
+			char *cp;
+			/* We cannot set read/write - must signal mdmon */
+			vers[9] = '/';
+			sysfs_set_str(mdi, NULL, "metadata_version", vers);
+
+			cp = strchr(vers+10, '/');
+			if (*cp)
+				*cp = 0;
+			ping_monitor(vers+10);
+		}
+		return 0;
+	}
+#endif
 	if (ioctl(fd, GET_ARRAY_INFO, &array)) {
 		fprintf(stderr, Name ": %s does not appear to be active.\n",
 			devname);
@@ -74,17 +121,70 @@ int Manage_ro(char *devname, int fd, int readonly)
 
 #ifndef MDASSEMBLE
 
+static void remove_devices(int devnum, char *path)
+{
+	/* Remove all 'standard' devices for 'devnum', including
+	 * partitions.  Also remove names at 'path' - possibly with
+	 * partition suffixes - which link to those names.
+	 */
+	char base[40];
+	char *path2;
+	char link[1024];
+	int n;
+	int part;
+	char *be;
+	char *pe;
+
+	if (devnum >= 0)
+		sprintf(base, "/dev/md%d", devnum);
+	else
+		sprintf(base, "/dev/md_d%d", -1-devnum);
+	be = base + strlen(base);
+	if (path) {
+		path2 = malloc(strlen(path)+20);
+		strcpy(path2, path);
+		pe = path2 + strlen(path2);
+	} else
+		path = NULL;
+	
+	for (part = 0; part < 16; part++) {
+		if (part) {
+			sprintf(be, "p%d", part);
+			if (path) {
+				if (isdigit(pe[-1]))
+					sprintf(pe, "p%d", part);
+				else
+					sprintf(pe, "%d", part);
+			}
+		}
+		/* FIXME test if really is md device ?? */
+		unlink(base);
+		if (path) {
+			n = readlink(path2, link, sizeof(link));
+			if (n && strlen(base) == n &&
+			    strncmp(link, base, n) == 0)
+				unlink(path2);
+		}
+	}
+}
+	
+
 int Manage_runstop(char *devname, int fd, int runstop, int quiet)
 {
 	/* 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.
 	 */
 	mdu_param_t param; /* unused */
 
 	if (runstop == -1 && md_get_version(fd) < 9000) {
 		if (ioctl(fd, STOP_MD, 0)) {
-			if (!quiet) fprintf(stderr, Name ": stopping device %s failed: %s\n",
-					    devname, strerror(errno));
+			if (quiet == 0) fprintf(stderr,
+						Name ": stopping device %s "
+						"failed: %s\n",
+						devname, strerror(errno));
 			return 1;
 		}
 	}
@@ -111,20 +211,77 @@ int Manage_runstop(char *devname, int fd, int runstop, int quiet)
 	} else if (runstop < 0){
 		struct map_ent *map = NULL;
 		struct stat stb;
-		if (ioctl(fd, STOP_ARRAY, NULL)) {
-			if (quiet==0)
-				fprintf(stderr, Name ": fail to stop array %s: %s\n",
+		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 &&
+		    is_subarray(mdi->text_version)) {
+			/* This is mdmon managed. */
+			close(fd);
+			if (sysfs_set_str(mdi, NULL,
+					  "array_state", "inactive") < 0) {
+				if (quiet == 0)
+					fprintf(stderr, Name
+						": failed to stop array %s: %s\n",
+						devname, strerror(errno));
+				return 1;
+			}
+
+			/* Give monitor a chance to act */
+			ping_monitor(mdi->text_version);
+
+			fd = open(devname, O_RDONLY);
+		} else if (mdi &&
+			   mdi->array.major_version == -1 &&
+			   mdi->array.minor_version == -2 &&
+			   !is_subarray(mdi->text_version)) {
+			/* container, possibly mdmon-managed.
+			 * Make sure mdmon isn't opening it, which
+			 * would interfere with the 'stop'
+			 */
+			ping_monitor(mdi->sys_name);
+		}
+
+		if (fd >= 0 && ioctl(fd, STOP_ARRAY, NULL)) {
+			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 (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 (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);
+		if (devnum != NoMdDev) {
 			map_delete(&map, devnum);
 			map_write(map);
 			map_free(map);
@@ -188,25 +345,42 @@ int Manage_subdevs(char *devname, int fd,
 	 */
 	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;
-	struct supertype *st;
-	void *dsuper = NULL;
-	void *osuper = NULL; /* original super */
+	struct supertype *st, *tst;
 	int duuid[4];
 	int ouuid[4];
+	int lfd = -1;
 
 	if (ioctl(fd, GET_ARRAY_INFO, &array)) {
 		fprintf(stderr, Name ": cannot get array info for %s\n",
 			devname);
 		return 1;
 	}
+
+	/* array.size is only 32 bit and may be truncated.
+	 * So read from sysfs if possible, and record number of sectors
+	 */
+
+	array_size = get_component_size(fd);
+	if (array_size <= 0)
+		array_size = array.size * 2;
+
+	tst = super_by_fd(fd);
+	if (!tst) {
+		fprintf(stderr, Name ": unsupport array - version %d.%d\n",
+			array.major_version, array.minor_version);
+		return 1;
+	}
+
 	for (dv = devlist, j=0 ; dv; dv = next, j = jnext) {
 		unsigned long long ldsize;
 		char dvname[20];
 		char *dnprintable = dv->devname;
+		int err;
 
 		next = dv->next;
 		jnext = 0;
@@ -291,32 +465,35 @@ int Manage_subdevs(char *devname, int fd,
 			return 1;
 		case 'a':
 			/* add the device */
-			st = super_by_version(array.major_version,
-					      array.minor_version);
-			if (!st) {
-				fprintf(stderr, Name ": unsupport array - version %d.%d\n",
-					array.major_version, array.minor_version);
+			if (tst->subarray[0]) {
+				fprintf(stderr, Name ": Cannot add disks to a"
+					" \'member\' array, perform this"
+					" operation on the parent container\n");
 				return 1;
 			}
-
 			/* Make sure it isn't in use (in 2.6 or later) */
-			tfd = open(dv->devname, O_RDONLY|O_EXCL);
+			tfd = open(dv->devname, O_RDONLY|O_EXCL|O_DIRECT);
 			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, &osuper, NULL);
-			/* will use osuper later */
+				st->ss->load_super(st, tfd, NULL);
+
 			if (!get_dev_size(tfd, dv->devname, &ldsize)) {
 				close(tfd);
 				return 1;
 			}
 			close(tfd);
 
-			if (array.major_version == 0 &&
+
+			if (!tst->ss->external &&
+			    array.major_version == 0 &&
 			    md_get_version(fd)%100 < 2) {
 				if (ioctl(fd, HOT_ADD_DISK,
 					  (unsigned long)stb.st_rdev)==0) {
@@ -331,20 +508,16 @@ int Manage_subdevs(char *devname, int fd,
 				return 1;
 			}
 
-			if (array.not_persistent == 0) {
-
-				/* Make sure device is large enough */
-				if (st->ss->avail_size(st, ldsize/512) <
-				    array.size) {
-					fprintf(stderr, Name ": %s not large enough to join array\n",
-						dv->devname);
-					return 1;
-				}
+			if (array.not_persistent == 0 || tst->ss->external) {
 
 				/* need to find a sample superblock to copy, and
-				 * a spare slot to use
+				 * a spare slot to use.
+				 * For 'external' array (well, container based),
+				 * We can just load the metadata for the array.
 				 */
-				for (j=0; j<st->max_devs; j++) {
+				if (tst->ss->external) {
+					tst->ss->load_super(tst, fd, NULL);
+				} else for (j = 0; j < tst->max_devs; j++) {
 					char *dev;
 					int dfd;
 					disc.number = j;
@@ -358,22 +531,33 @@ int Manage_subdevs(char *devname, int fd,
 					if (!dev) continue;
 					dfd = dev_open(dev, O_RDONLY);
 					if (dfd < 0) continue;
-					if (st->ss->load_super(st, dfd, &dsuper, NULL)) {
+					if (tst->ss->load_super(tst, dfd,
+								NULL)) {
 						close(dfd);
 						continue;
 					}
 					close(dfd);
 					break;
 				}
-				if (!dsuper) {
+				/* FIXME this is a bad test to be using */
+				if (!tst->sb) {
 					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) {
+					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.
 				 */
-				st->ss->uuid_from_super(st, duuid, dsuper);
+				tst->ss->uuid_from_super(tst, duuid);
 
 				/* re-add doesn't work for version-1 superblocks
 				 * before 2.6.18 :-(
@@ -381,27 +565,34 @@ int Manage_subdevs(char *devname, int fd,
 				if (array.major_version == 1 &&
 				    get_linux_version() <= 2006018)
 					;
-				else if (osuper) {
-					st->ss->uuid_from_super(st, ouuid, osuper);
+				else if (st->sb) {
+					st->ss->uuid_from_super(st, ouuid);
 					if (memcmp(duuid, ouuid, sizeof(ouuid))==0) {
 						/* looks close enough for now.  Kernel
 						 * will worry about whether a bitmap
 						 * based reconstruction is possible.
 						 */
 						struct mdinfo mdi;
-						st->ss->getinfo_super(st, &mdi, osuper);
+						st->ss->getinfo_super(st, &mdi);
 						disc.major = major(stb.st_rdev);
 						disc.minor = minor(stb.st_rdev);
 						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 (dv->writemostly == 2)
+							disc.state &= ~(1 << MD_DISK_WRITEMOSTLY);
 						if (ioctl(fd, ADD_NEW_DISK, &disc) == 0) {
 							if (verbose >= 0)
 								fprintf(stderr, Name ": re-added %s\n", dv->devname);
 							continue;
 						}
+						if (errno == ENOMEM || errno == EROFS) {
+							fprintf(stderr, Name ": add new device failed for %s: %s\n",
+								dv->devname, strerror(errno));
+							return 1;
+						}
 						/* fall back on normal-add */
 					}
 				}
@@ -409,7 +600,7 @@ int Manage_subdevs(char *devname, int fd,
 				/* non-persistent. Must ensure that new drive
 				 * is at least array.size big.
 				 */
-				if (ldsize/512 < array.size) {
+				if (ldsize/512 < array_size) {
 					fprintf(stderr, Name ": %s not large enough to join array\n",
 						dv->devname);
 					return 1;
@@ -420,7 +611,7 @@ int Manage_subdevs(char *devname, int fd,
 			 * we must choose the same free number, which requires
 			 * starting at 'raid_disks' and counting up
 			 */
-			for (j = array.raid_disks; j< st->max_devs; j++) {
+			for (j = array.raid_disks; j< tst->max_devs; j++) {
 				disc.number = j;
 				if (ioctl(fd, GET_DISK_INFO, &disc))
 					break;
@@ -433,11 +624,21 @@ int Manage_subdevs(char *devname, int fd,
 			disc.minor = minor(stb.st_rdev);
 			disc.number =j;
 			disc.state = 0;
-			if (array.not_persistent==0) {
-				if (dv->writemostly)
+			if (array.not_persistent==0 || tst->ss->external) {
+				int dfd;
+				if (dv->writemostly == 1)
 					disc.state |= 1 << MD_DISK_WRITEMOSTLY;
-				st->ss->add_to_super(st, dsuper, &disc);
-				if (st->ss->write_init_super(st, dsuper, &disc, dv->devname))
+				dfd = 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 */
+					close(dfd);
+				else if (tst->ss->write_init_super(tst))
 					return 1;
 			} else if (dv->re_add) {
 				/*  this had better be raid1.
@@ -446,7 +647,7 @@ int Manage_subdevs(char *devname, int fd,
 				 */
 				char *used = malloc(array.raid_disks);
 				memset(used, 0, array.raid_disks);
-				for (j=0; j< st->max_devs; j++) {
+				for (j=0; j< tst->max_devs; j++) {
 					mdu_disk_info_t disc2;
 					disc2.number = j;
 					if (ioctl(fd, GET_DISK_INFO, &disc2))
@@ -468,9 +669,54 @@ int Manage_subdevs(char *devname, int fd,
 						break;
 					}
 			}
-			if (dv->writemostly)
+			if (dv->writemostly == 1)
 				disc.state |= (1 << MD_DISK_WRITEMOSTLY);
-			if (ioctl(fd,ADD_NEW_DISK, &disc)) {
+			if (tst->ss->external) {
+				/* add a disk to an external metadata container
+				 * only if mdmon is around to see it
+				 */
+				struct mdinfo new_mdi;
+				struct mdinfo *sra;
+				int container_fd;
+				int devnum = fd2devnum(fd);
+
+				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);
+					return 1;
+				}
+
+				if (!mdmon_running(devnum)) {
+					fprintf(stderr, Name ": add failed for %s: mdmon not running\n",
+						dv->devname);
+					close(container_fd);
+					return 1;
+				}
+
+				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);
+					return 1;
+				}
+				sra->array.level = LEVEL_CONTAINER;
+				/* Need to set data_offset and component_size */
+				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) != 0) {
+					fprintf(stderr, Name ": add new device to external metadata"
+						" failed for %s\n", dv->devname);
+					close(container_fd);
+					return 1;
+				}
+				ping_monitor(devnum2devname(devnum));
+				sysfs_free(sra);
+				close(container_fd);
+			} else 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;
@@ -481,13 +727,94 @@ int Manage_subdevs(char *devname, int fd,
 
 		case 'r':
 			/* hot remove */
+			if (tst->subarray[0]) {
+				fprintf(stderr, Name ": Cannot remove disks from a"
+					" \'member\' array, perform this"
+					" operation on the parent container\n");
+				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");
+					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 ||
+				    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 (ioctl(fd, HOT_REMOVE_DISK, (unsigned long)stb.st_rdev)) {
+			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
+					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);
+			}
+			close(lfd);
 			if (verbose >= 0)
 				fprintf(stderr, Name ": hot removed %s\n",
 					dnprintable);