Skip to contentSkip to conversation

Incremental mode: remove safety verification

PATCH Superseded Kinga Tanska 2 participants 4 messages Dec 21, 2022 – Dec 22, 2022 MD RAID

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

v2 cover Incremental mode: remove safety verification

Changes in incremental mode. Removing verification if remove is safe, when this mode is triggered. Also moving commit description to obey kernel coding style.

Kinga Tanska (2):
incremental, manage: do not verify if remove is safe
manage: move comment with function description

 Incremental.c |  2 +-
 Manage.c      | 79 +++++++++++++++++++++++++++++++--------------------
 2 files changed, 49 insertions(+), 32 deletions(-)
Signature · 1 line
2.26.2
#1 lore raw

PATCH v2 1/2 incremental, manage: do not verify if remove is safe

Function is_remove_safe() was introduced to verify if removing member device won't cause failed state of the array. This verification should be used only with set-faulty command. Add special mode indicating that Incremental removal was executed. If this mode is used do not execute is_remove_safe() routine.

---
 Incremental.c | 2 +-
 Manage.c      | 7 ++++---
 2 files changed, 5 insertions(+), 4 deletions(-)
Incremental.c +1 -1
index 5a5f4c4c..bccfdeb9 100644
@@ -1744,7 +1744,7 @@ int IncrementalRemove(char *devname, char *id_path, int verbose)
 
 	memset(&devlist, 0, sizeof(devlist));
 	devlist.devname = devname;
-	devlist.disposition = 'f';
+	devlist.disposition = 'I';
 	/* for a container, we must fail each member array */
 	if (ent->metadata_version &&
 	    strncmp(ent->metadata_version, "external:", 9) == 0) {
Manage.c +4 -3
index b1d0e630..1e7f9b56 100644
@@ -1496,8 +1496,9 @@ int Manage_subdevs(char *devname, int fd,
 			/* Assume this is a kernel-internal name like 'sda1' */
 			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",
+			if (dv->disposition != 'r' && dv->disposition != 'f' &&
+			    dv->disposition != 'I') {
+				pr_err("%s only meaningful with -r, -f or -I, not -%c\n",
 					dv->devname, dv->disposition);
 				goto abort;
 			}
@@ -1649,7 +1650,7 @@ int Manage_subdevs(char *devname, int fd,
 					close(sysfd);
 				goto abort;
 			}
-
+		case 'I': /* incremental fail */
 			if ((sysfd >= 0 && write(sysfd, "faulty", 6) != 6) ||
 			    (sysfd < 0 && ioctl(fd, SET_DISK_FAULTY,
 						rdev))) {
Signature · 1 line
2.26.2
#2 lore raw

PATCH v2 2/2 manage: move comment with function description

In reply to Kinga Tanska

Move the function description from the function body to outside to obey kernel coding style.

---
 Manage.c | 72 ++++++++++++++++++++++++++++++++++----------------------
 1 file changed, 44 insertions(+), 28 deletions(-)
Manage.c +44 -28
index 1e7f9b56..9d85237c 100644
@@ -1329,38 +1329,54 @@ bool is_remove_safe(mdu_array_info_t *array, const int fd, char *devname, const
 	return is_enough;
 }
 
+/**
+ * Manage_subdevs() - Execute operation depending on devmode.
+ *
+ * @devname: name of the device.
+ * @fd: file descriptor.
+ * @devlist: list of sub-devices to manage.
+ * @verbose: verbose level.
+ * @test: test flag.
+ * @update: type of update.
+ * @force: force flag.
+ *
+ * This function executes operation defined by devmode
+ * for each dev from devlist.
+ * Devmode can be:
+ * 'a' - add the device
+ * '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
+ *       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.
+ * 'I' - remove device by using incremental fail
+ *       which is executed when device is removed surprisingly.
+ * '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.
+ * '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'.
+ *
+ * Return: 0 on success, otherwise 1 or 2.
+ */
 int Manage_subdevs(char *devname, int fd,
 		   struct mddev_dev *devlist, int verbose, int test,
 		   char *update, int force)
 {
-	/* Do something to each dev.
-	 * devmode can be
-	 *  'a' - add the device
-	 *  '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
-	 *	  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.
-	 *  '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'.
-	 */
 	mdu_array_info_t array;
 	unsigned long long array_size;
 	struct mddev_dev *dv;
Signature · 1 line
2.26.2
#3 lore raw
In reply to Kinga Tanska

Hi Kinga

For the series, it's ok for me.

Do you want to fix another regression problem? Now we can't set a member disk as faulty when
array is active.

The original patch is used to make sure the admin doesn't remove the disk and the raid can't work.
So it doesn't need to check the array state. How about

Manage.c +0 -0
index 9d85237ceab5..eee25fd8cda0 100644
@@ -1322,8 +1322,7 @@ bool is_remove_safe(mdu_array_info_t *array,

const int fd, char *devname, const

        sysfs_free(mdi);

        bool is_enough = enough(array->level, array->raid_disks,
-                               array->layout, (array->state & 1),
-                               avail);
+                               array->layout, 1, avail);
        free(avail);
        return is_enough;

Regards
Xiao

Quoted 16 lines · Kinga Tanska

Changes in incremental mode. Removing verification if remove is safe, when this mode is triggered. Also moving commit description to obey kernel coding style.

Kinga Tanska (2):
incremental, manage: do not verify if remove is safe
manage: move comment with function description

 Incremental.c |  2 +-
 Manage.c      | 79 +++++++++++++++++++++++++++++++--------------------
 2 files changed, 49 insertions(+), 32 deletions(-)

--
2.26.2