will still barf -- I think this is a feature).
Some ChangeLog updates.
* version v0_5_16
* bug fixes:
- limit Lustre IOVs to PTL_MD_MAX_IOV (611336)
+ * protocol change to lustre_msg: move |version| and add |flags|
+ * added replay of create, unlink, link and rename operations during
+ MDS failover; recovery should be much more robust now
+ * remove failed OSCs from LOVs
2002-10-23 Phil Schwan <phil@clusterfs.com>
* version v0_5_15
__u64 transno;
__u32 status;
__u32 bufcount;
+ __u32 version;
__u32 buflens[0];
};
spin_unlock(&conn->c_lock);
request->rq_reqmsg->magic = PTLRPC_MSG_MAGIC;
+ request->rq_reqmsg->version = PTLRPC_MSG_VERSION;
request->rq_reqmsg->opc = HTON__u32(opcode);
request->rq_reqmsg->flags = 0;
goto out;
}
+ if (request->rq_reqmsg->version != PTLRPC_MSG_VERSION) {
+ CERROR("wrong lustre_msg version %d: ptl %d from "LPX64" xid "
+ LPD64"\n",
+ request->rq_reqmsg->version, svc->srv_req_portal,
+ event->initiator.nid, request->rq_xid);
+ goto out;
+ }
+
CDEBUG(D_NET, "got req "LPD64" (md: %p + %d)\n", request->rq_xid,
event->mem_desc.start, event->offset);