(Online Capacity Expansion for IMSM)
When mdmon calls grow_array() on disks increase detection,
grow_array() has to add new disks to active_array structure.
According to this memory for internal imsm data has to be reallocated
reallocate_imsm_dev() takes current memory contents of internal imsm data and copies it to new (larger) memory location. Manages memory relink then.
Signed-off-by: Adam Kwolek <adam.kwolek@xxxxxxxxx>
---
mdadm.h | 2 +
monitor.c | 50 -----------------------
super-intel.c | 126 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
util.c | 51 +++++++++++++++++++++++
4 files changed, 179 insertions(+), 50 deletions(-)
diff --git a/mdadm.h b/mdadm.h
index a846098..ecef25c 100644
--- a/mdadm.h
+++ b/mdadm.h
@@ -861,6 +861,8 @@ extern int conf_name_is_free(char *name); extern int devname_matches(char *name, char *match); extern struct mddev_ident_s *conf_match(struct mdinfo *info, struct supertype *st); extern void sleep2(unsigned int sec, unsigned int usec);
+extern int read_attr(char *buf, int len, int fd); extern int
+read_dev_state(int fd);
extern void free_line(char *line);
extern int match_oneof(char *devices, char *devname); diff --git a/monitor.c b/monitor.c index 4578718..3c99dff 100644
--- a/monitor.c
+++ b/monitor.c
@@ -45,27 +45,6 @@ static void add_fd(fd_set *fds, int *maxfd, int fd)
FD_SET(fd, fds);
}
-static int read_attr(char *buf, int len, int fd) -{
- int n;
-
- if (fd < 0) {
- buf[0] = 0;
- return 0;
- }
- lseek(fd, 0, 0);
- n = read(fd, buf, len - 1);
-
- if (n <= 0) {
- buf[0] = 0;
- return 0;
- }
- buf[n] = 0;
- if (buf[n-1] == '\n')
- buf[n-1] = 0;
- return n;
-}
-
static unsigned long long read_resync_start(int fd) {
char buf[30];
@@ -100,35 +79,6 @@ static enum sync_action read_action( int fd)
return (enum sync_action) sysfs_match_word(buf, sync_actions); }
-int read_dev_state(int fd)
-{
- char buf[60];
- int n = read_attr(buf, 60, fd);
- char *cp;
- int rv = 0;
-
- if (n <= 0)
- return 0;
-
- cp = buf;
- while (cp) {
- if (sysfs_attr_match(cp, "faulty"))
- rv |= DS_FAULTY;
- if (sysfs_attr_match(cp, "in_sync"))
- rv |= DS_INSYNC;
- if (sysfs_attr_match(cp, "write_mostly"))
- rv |= DS_WRITE_MOSTLY;
- if (sysfs_attr_match(cp, "spare"))
- rv |= DS_SPARE;
- if (sysfs_attr_match(cp, "blocked"))
- rv |= DS_BLOCKED;
- cp = strchr(cp, ',');
- if (cp)
- cp++;
- }
- return rv;
-}
-
static void signal_manager(void)
{
/* tgkill(getpid(), mon_tid, SIGUSR1); */ diff --git a/super-intel.c b/super-intel.c index a4ffb3f..115a5f0 100644
--- a/super-intel.c
+++ b/super-intel.c
@@ -5098,6 +5098,131 @@ static struct mdinfo *imsm_activate_spare(struct active_array *a,
return rv;
}
+int imsm_grow_array(struct active_array *a, int new_raid_disks) {
+ int ret_val = 0;
+ struct intel_super *super = a->container->sb;
+ int inst = a->info.container_member;
+ struct imsm_dev *dev = get_imsm_dev(super, inst);
+ struct imsm_map *map = get_imsm_map(dev, 0);
+ struct mdinfo *d;
+ struct mdinfo *di;
+ struct dl *dl;
+ int i;
+ int rv;
+ char dv[1024];
+ char nm[1024];
+ char *dname;
+
+ dprintf("imsm: grow array: inst=%d raid disks=%d level=%d\n",
+ inst, a->info.array.raid_disks, a->info.array.level);
+
+ /* grow support only */
+ if ((a->info.array.raid_disks >= new_raid_disks) &&
+ (new_raid_disks > 0))
+ return ret_val;
+
+ /* Look for all disks beyond current configuration
+ * To handle degradation after takeover
+ * look also on last disk in configuration.
+ */
+ for (i = a->info.array.raid_disks-1; i < new_raid_disks; i++) {
+ for (d = a->info.devs ; d ; d = d->next)
+ if (d->disk.raid_disk == i)
+ break;
+ dprintf("found %d: %p %x\n", i, d, d ? d->curr_state : 0);
+ if (d && (d->state_fd >= 0))
+ continue;
+
+ /* OK, this device can be added. Try to add.
+ */
+ dl = imsm_add_spare(super, i, a, 0);
+ if (!dl)
+ dl = imsm_add_spare(super, i, a, 1);
+ if (!dl)
+ continue;
+
+ if (dl->index < 0)
+ dl->index = i;
+ /* found a usable disk with enough space */
+ di = malloc(sizeof(*di));
+ if (!di)
+ continue;
+ memset(di, 0, sizeof(*di));
+
+ /* dl->index will be -1 in the case we are activating a
+ * pristine spare. imsm_process_update() will create a
+ * new index in this case. Once a disk is found to be
+ * failed in all member arrays it is kicked from the
+ * metadata
+ */
+ di->disk.number = dl->index;
+
+ dl->disk.status |= CONFIGURED_DISK;
+ dl->disk.status &= ~SPARE_DISK;
+
+ /* (ab)use di->devs to store a pointer to the device
+ * we chose
+ */
+ di->devs = (struct mdinfo *) dl;
+
+ di->disk.raid_disk = i;
+ di->disk.major = dl->major;
+ di->disk.minor = dl->minor;
+ di->disk.state = (1<<MD_DISK_SYNC) |
+ (1<<MD_DISK_ACTIVE);
+
+ /* set handle to state in sysfs
+ */
+ sprintf(dv, "/sys/dev/block/%d:%d", di->disk.major, di->disk.minor);
+ memset(nm, 0, sizeof(nm));
+ rv = readlink(dv, nm, sizeof(nm));
+ if (rv > 0) {
+ nm[rv] = '\0';
+ dname = strrchr(nm, '/');
+ if (dname) {
+ dname++;
+ sprintf(di->sys_name, "dev-%s", dname);
+ }
+ }
+
+ di->state_fd = sysfs_open(a->devnum,
+ di->sys_name,
+ "state");
+ di->recovery_fd = sysfs_open(a->devnum,
+ di->sys_name,
+ "recovery_start");
+
+ di->prev_state = read_dev_state(di->state_fd);
+ di->curr_state = di->prev_state;
+ di->next_state = 0;
+
+ di->recovery_start = 0;
+ di->data_offset = __le32_to_cpu(map->pba_of_lba0);
+ di->component_size = a->info.component_size;
+ di->container_member = inst;
+ super->random = random32();
+
+ di->next = a->info.devs;
+ a->info.devs = di;
+
+ ret_val++;
+ dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
+ i, di->data_offset);
+ }
+
+ if ((new_raid_disks - a->info.array.raid_disks) != ret_val) {
+ dprintf("imsm: ERROR: imsm_grow_array() configures %i raid disks\n",
+ ret_val);
+ }
+
+ /* prepare device to receive new disk(s)
+ */
+ reallocate_imsm_dev(super, inst, new_raid_disks);
+
+ return ret_val;
+}
+
static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u) {
struct imsm_dev *dev = get_imsm_dev(super, idx); @@ -5697,6 +5822,7 @@ struct superswitch super_imsm = {
.match_metadata_desc = match_metadata_desc_imsm,
.container_content = container_content_imsm,
.default_layout = imsm_level_to_layout,
+ .grow_array = imsm_grow_array,
.external = 1,
.name = "imsm",
diff --git a/util.c b/util.c
index 5779d52..7f5e20a 100644
--- a/util.c
+++ b/util.c
@@ -1656,3 +1656,54 @@ void sleep2(unsigned int sec, unsigned int usec)
tv.tv_usec = usec;
select(0, NULL, NULL, NULL, &tv);
}
+
+int read_attr(char *buf, int len, int fd) {
+ int n;
+
+ if (fd < 0) {
+ buf[0] = 0;
+ return 0;
+ }
+ lseek(fd, 0, 0);
+ n = read(fd, buf, len - 1);
+
+ if (n <= 0) {
+ buf[0] = 0;
+ return 0;
+ }
+ buf[n] = 0;
+ if (buf[n-1] == '\n')
+ buf[n-1] = 0;
+
+ return n;
+}
+
+int read_dev_state(int fd)
+{
+ char buf[60];
+ int n = read_attr(buf, 60, fd);
+ char *cp;
+ int rv = 0;
+
+ if (n <= 0)
+ return 0;
+
+ cp = buf;
+ while (cp) {
+ if (sysfs_attr_match(cp, "faulty"))
+ rv |= DS_FAULTY;
+ if (sysfs_attr_match(cp, "in_sync"))
+ rv |= DS_INSYNC;
+ if (sysfs_attr_match(cp, "write_mostly"))
+ rv |= DS_WRITE_MOSTLY;
+ if (sysfs_attr_match(cp, "spare"))
+ rv |= DS_SPARE;
+ if (sysfs_attr_match(cp, "blocked"))
+ rv |= DS_BLOCKED;
+ cp = strchr(cp, ',');
+ if (cp)
+ cp++;
+ }
+ return rv;
+}
��.n��������+%������w��{.n�����{����w��ܨ}���Ơz�j:+v�����w����ޙ��&�)ߡ�a����z�ޗ���ݢj��w�f
