Add support for --replace and --with

--replace can be used to replace a device without completely failing
it.  Once the replacement completes the device will be failed.
--with can indicate which of several spares to use.

Signed-off-by: NeilBrown <neilb@suse.de>
  

5 files changed, 193 insertions, 3 deletions

diff --git a/Manage.c b/Manage.c
index 296feeb7..832e84ce 100644
--- a/Manage.c
+++ b/Manage.c
@@ -954,6 +954,111 @@ int Manage_remove(struct supertype *tst, int fd, struct mddev_dev *dv,
 	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, -1, 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;
+	}
+	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, -1, 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<<MD_DISK_FAULTY)) {
+			pr_err("%s is faulty and cannot be a replacement\n",
+			       dv->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 %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;
+	}
+	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)
@@ -970,6 +1075,16 @@ int Manage_subdevs(char *devname, int fd,
 	 *  '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'
+	 *
 	 * For 'f' and 'r', the device can also be a kernel-internal
 	 * name such as 'sdb'.
 	 */
@@ -1209,6 +1324,38 @@ int Manage_subdevs(char *devname, int fd,
 				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)
diff --git a/ReadMe.c b/ReadMe.c
index 2fde2ac5..0aa8cbd7 100644
--- a/ReadMe.c
+++ b/ReadMe.c
@@ -146,6 +146,8 @@ struct option long_options[] = {
     {"remove",    0, 0, Remove},
     {"fail",      0, 0, Fail},
     {"set-faulty",0, 0, Fail},
+    {"replace",   0, 0, Replace},
+    {"with",      0, 0, With},
     {"run",       0, 0, 'R'},
     {"stop",      0, 0, 'S'},
     {"readonly",  0, 0, 'o'},
@@ -309,6 +311,7 @@ char OptionHelp[] =
 "  --remove      -r   : remove subsequent devices\n"
 "  --fail        -f   : mark subsequent devices as faulty\n"
 "  --set-faulty       : same as --fail\n"
+"  --replace          : mark a device for replacement\n"
 "  --run         -R   : start a partially built array\n"
 "  --stop        -S   : deactivate array, releasing all resources\n"
 "  --readonly    -o   : mark array as readonly\n"
@@ -462,6 +465,10 @@ char Help_manage[] =
 "  --remove      -r   : remove subsequent devices, which must not be active\n"
 "  --fail        -f   : mark subsequent devices a faulty\n"
 "  --set-faulty       : same as --fail\n"
+"  --replace          : mark device(s) to be replaced by spares.  Once\n"
+"                     : replacement completes, device will be marked faulty\n"
+"  --with             : Indicate which spare a previous '--replace' should\n"
+"                     : prefer to use\n"
 "  --run         -R   : start a partially built array\n"
 "  --stop        -S   : deactivate array, releasing all resources\n"
 "  --readonly    -o   : mark array as readonly\n"
diff --git a/mdadm.8.in b/mdadm.8.in
index d30963b6..23e31228 100644
--- a/mdadm.8.in
+++ b/mdadm.8.in
@@ -216,8 +216,9 @@ to detect and assemble arrays \(em possibly in an
 If a device is given before any options, or if the first option is
 .BR \-\-add ,
 .BR \-\-fail ,
-or
 .BR \-\-remove ,
+or
+.BR \-\-replace ,
 then the MANAGE mode is assumed.
 Anything other than these will cause the
 .B Misc
@@ -1283,7 +1284,7 @@ have already been marked as failed.
 
 .TP
 .BR \-f ", " \-\-fail
-mark listed devices as faulty.
+Mark listed devices as faulty.
 As well as the name of a device file, the word
 .B detached
 can be given.  This will cause any device that has been detached from
@@ -1295,6 +1296,25 @@ same as
 .BR \-\-fail .
 
 .TP
+.B \-\-replace
+Mark listed devices as requiring replacement.  As soon as a spare is
+available, it will be rebuilt and will replace the marked device.
+This is similar to marking a device as faulty, but the device remains
+in service during the recovery process to increase resilience against
+multiple failures.  When the replacement process finishes, the
+replaced device will be marked as faulty.
+
+.TP
+.B \-\-with
+This can follow a list of
+.B \-\-replace
+devices.  The devices listed after
+.B \-\-with
+will be preferentially used to replace the devices listed after
+.BR \-\-replace .
+These device must already be spare devices in the array.
+
+.TP
 .BR \-\-write\-mostly
 Subsequent devices that are added or re\-added will have the 'write-mostly'
 flag set.  This is only valid for RAID1 and means that the 'md' driver
diff --git a/mdadm.c b/mdadm.c
index 42544f1a..11016e7b 100644
--- a/mdadm.c
+++ b/mdadm.c
@@ -195,6 +195,8 @@ int main(int argc, char *argv[])
 		case Add:
 		case 'r':
 		case Remove:
+		case Replace:
+		case With:
 		case 'f':
 		case Fail:
 		case ReAdd: /* re-add */
@@ -928,6 +930,18 @@ int main(int argc, char *argv[])
 					   * remove the device */
 			devmode = 'f';
 			continue;
+		case O(MANAGE,Replace):
+			/* Mark these devices for replacement */
+			devmode = 'R';
+			continue;
+		case O(MANAGE,With):
+			/* These are the replacements to use */
+			if (devmode != 'R') {
+				pr_err("--with must follow --replace\n");
+				exit(2);
+			}
+			devmode = 'W';
+			continue;
 		case O(INCREMENTAL,'R'):
 		case O(MANAGE,'R'):
 		case O(ASSEMBLE,'R'):
diff --git a/mdadm.h b/mdadm.h
index ac051178..f1352e37 100644
--- a/mdadm.h
+++ b/mdadm.h
@@ -312,6 +312,8 @@ enum special_options {
 	Add,
 	Remove,
 	Fail,
+	Replace,
+	With,
 	MiscOpt,
 	WaitOpt,
 	ConfigFile,
@@ -432,7 +434,7 @@ struct mddev_dev {
 				 * Not set for names read from .config
 				 */
 	char writemostly;	/* 1 for 'set writemostly', 2 for 'clear writemostly' */
-	char used;		/* set when used */
+	int used;		/* set when used */
 	long long data_offset;
 	struct mddev_dev *next;
 };