On 8/30/21 8:45 PM, Jeff Layton wrote:
On Mon, 2021-08-30 at 20:33 +0800, xiubli@xxxxxxxxxx wrote:
From: Xiubo Li <xiubli@xxxxxxxxxx>
Signed-off-by: Xiubo Li <xiubli@xxxxxxxxxx>
---
fs/ceph/mds_client.c | 8 +++++---
1 file changed, 5 insertions(+), 3 deletions(-)
diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
index 7ddc36c14b92..aa0ab069db40 100644
--- a/fs/ceph/mds_client.c
+++ b/fs/ceph/mds_client.c
@@ -4434,7 +4434,7 @@ static void check_new_map(struct ceph_mds_client *mdsc,
struct ceph_mdsmap *newmap,
struct ceph_mdsmap *oldmap)
{
- int i, err;
+ int i, j, err;
int oldstate, newstate;
struct ceph_mds_session *s;
unsigned long targets[DIV_ROUND_UP(CEPH_MAX_MDS, sizeof(unsigned long))] = {0};
@@ -4443,8 +4443,10 @@ static void check_new_map(struct ceph_mds_client *mdsc,
newmap->m_epoch, oldmap->m_epoch);
if (newmap->m_info) {
- for (i = 0; i < newmap->m_info->num_export_targets; i++)
- set_bit(newmap->m_info->export_targets[i], targets);
+ for (i = 0; i < newmap->m_num_active_mds; i++) {
+ for (j = 0; j < newmap->m_info[i].num_export_targets; j++)
+ set_bit(newmap->m_info[i].export_targets[j], targets);
+ }
}
for (i = 0; i < oldmap->possible_max_rank && i < mdsc->max_sessions; i++) {
Looks sane. I'll plan to fold this into "ceph: reconnect to the export
targets on new mdsmaps".
Why my previous test worked well, it should be there has only one active
mds. And for the 'newmap->m_info->export_target' it will always equal to
'newmap->m_info[0].export_target', but I am not very sure why this could
cause crash.
Thanks,
