From: lixiaokeng <lixiaokeng@xxxxxxxxxx>
When we add a new disk to a raid, it may return -EBUSY.
The main process of --add:
1. dev_open
2. store_super1(st, di->fd) in write_init_super1
3. fsync(di->fd) in write_init_super1
4. close(di->fd)
5. ioctl(ADD_NEW_DISK)
However, there will be some udev(change) event after step4. Then
"/usr/sbin/mdadm --incremental ..." will be run, and the new disk
will be add to md device. After that, ioctl will return -EBUSY.
Here we add map_lock before write_init_super in "mdadm --add"
to fix this race.
Signed-off-by: Li Xiao Keng <lixiaokeng@xxxxxxxxxx>
---
Manage.c | 22 ++++++++++++++--------
1 file changed, 14 insertions(+), 8 deletions(-)
diff --git a/Manage.c b/Manage.c
index fde6aba3..893a2ea8 100644
--- a/Manage.c
+++ b/Manage.c
@@ -720,6 +720,7 @@ int Manage_add(int fd, int tfd, struct mddev_dev *dv,
struct supertype *dev_st;
int j;
mdu_disk_info_t disc;
+ struct map_ent *map = NULL;
if (!get_dev_size(tfd, dv->devname, &ldsize)) {
if (dv->disposition == 'M')
@@ -917,6 +918,7 @@ int Manage_add(int fd, int tfd, struct mddev_dev *dv,
disc.raid_disk = 0;
}
+ map_lock(&map);
if (array->not_persistent==0) {
int dfd;
if (dv->disposition == 'j')
@@ -928,9 +930,9 @@ int Manage_add(int fd, int tfd, struct mddev_dev *dv,
dfd = dev_open(dv->devname, O_RDWR | O_EXCL|O_DIRECT);
if (tst->ss->add_to_super(tst, &disc, dfd,
dv->devname, INVALID_SECTORS))
- return -1;
+ goto unlock;
if (tst->ss->write_init_super(tst))
- return -1;
+ goto unlock;
} else if (dv->disposition == 'A') {
/* this had better be raid1.
* As we are "--re-add"ing we must find a spare slot
@@ -988,14 +990,14 @@ int Manage_add(int fd, int tfd, struct mddev_dev *dv,
pr_err("add failed for %s: could not get exclusive access to container\n",
dv->devname);
tst->ss->free_super(tst);
- return -1;
+ goto unlock;
}
/* Check if metadata handler is able to accept the drive */
if (!tst->ss->validate_geometry(tst, LEVEL_CONTAINER, 0, 1, NULL,
0, 0, dv->devname, NULL, 0, 1)) {
close(container_fd);
- return -1;
+ goto unlock;
}
Kill(dv->devname, NULL, 0, -1, 0);
@@ -1004,7 +1006,7 @@ int Manage_add(int fd, int tfd, struct mddev_dev *dv,
dv->devname, INVALID_SECTORS)) {
close(dfd);
close(container_fd);
- return -1;
+ goto unlock;
}
if (!mdmon_running(tst->container_devnm))
tst->ss->sync_metadata(tst);
@@ -1015,7 +1017,7 @@ int Manage_add(int fd, int tfd, struct mddev_dev *dv,
dv->devname);
close(container_fd);
tst->ss->free_super(tst);
- return -1;
+ goto unlock;
}
sra->array.level = LEVEL_CONTAINER;
/* Need to set data_offset and component_size */
@@ -1030,7 +1032,7 @@ int Manage_add(int fd, int tfd, struct mddev_dev *dv,
pr_err("add new device to external metadata failed for %s\n", dv->devname);
close(container_fd);
sysfs_free(sra);
- return -1;
+ goto unlock;
}
ping_monitor(devnm);
sysfs_free(sra);
@@ -1044,7 +1046,7 @@ int Manage_add(int fd, int tfd, struct mddev_dev *dv,
else
pr_err("add new device failed for %s as %d: %s\n",
dv->devname, j, strerror(errno));
- return -1;
+ goto unlock;
}
if (dv->disposition == 'j') {
pr_err("Journal added successfully, making %s read-write\n", devname);
@@ -1055,7 +1057,11 @@ int Manage_add(int fd, int tfd, struct mddev_dev *dv,
}
if (verbose >= 0)
pr_err("added %s\n", dv->devname);
+ map_unlock(&map);
return 1;
+unlock:
+ map_unlock(&map);
+ return -1;
}
int Manage_remove(struct supertype *tst, int fd, struct mddev_dev *dv,
--
2.33.0
