drbd: Receiving of delay_probes

Delay_probes are new packets in the DRBD protocol, which allow
DRBD to know the current delay packets have on the data socket.
(relative to the meta data socket)

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>

1 files changed, 96 insertions, 0 deletions

diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c
index b27f4dd36f89..fee0d249adf7 100644
--- a/drivers/block/drbd/drbd_receiver.c
+++ b/drivers/block/drbd/drbd_receiver.c
@@ -3501,6 +3501,92 @@ static int receive_UnplugRemote(struct drbd_conf *mdev, struct p_header *h)
         return TRUE;
 }
 
+static void timeval_sub_us(struct timeval* tv, unsigned int us)
+{
+        tv->tv_sec -= us / 1000000;
+        us = us % 1000000;
+        if (tv->tv_usec > us) {
+                tv->tv_usec += 1000000;
+                tv->tv_sec--;
+        }
+        tv->tv_usec -= us;
+}
+
+static void got_delay_probe(struct drbd_conf *mdev, int from, struct p_delay_probe *p)
+{
+        struct delay_probe *dp;
+        struct list_head *le;
+        struct timeval now;
+        int seq_num;
+        int offset;
+        int data_delay;
+
+        seq_num = be32_to_cpu(p->seq_num);
+        offset  = be32_to_cpu(p->offset);
+
+        spin_lock(&mdev->peer_seq_lock);
+        if (!list_empty(&mdev->delay_probes)) {
+                if (from == USE_DATA_SOCKET)
+                        le = mdev->delay_probes.next;
+                else
+                        le = mdev->delay_probes.prev;
+
+                dp = list_entry(le, struct delay_probe, list);
+
+                if (dp->seq_num == seq_num) {
+                        list_del(le);
+                        spin_unlock(&mdev->peer_seq_lock);
+                        do_gettimeofday(&now);
+                        timeval_sub_us(&now, offset);
+                        data_delay =
+                                now.tv_usec - dp->time.tv_usec +
+                                (now.tv_sec - dp->time.tv_sec) * 1000000;
+
+                        if (data_delay > 0)
+                                mdev->data_delay = data_delay;
+
+                        kfree(dp);
+                        return;
+                }
+
+                if (dp->seq_num > seq_num) {
+                        spin_unlock(&mdev->peer_seq_lock);
+                        dev_warn(DEV, "Previous allocation failure of struct delay_probe?\n");
+                        return; /* Do not alloca a struct delay_probe.... */
+                }
+        }
+        spin_unlock(&mdev->peer_seq_lock);
+
+        dp = kmalloc(sizeof(struct delay_probe), GFP_NOIO);
+        if (!dp) {
+                dev_warn(DEV, "Failed to allocate a struct delay_probe, do not worry.\n");
+                return;
+        }
+
+        dp->seq_num = seq_num;
+        do_gettimeofday(&dp->time);
+        timeval_sub_us(&dp->time, offset);
+
+        spin_lock(&mdev->peer_seq_lock);
+        if (from == USE_DATA_SOCKET)
+                list_add(&dp->list, &mdev->delay_probes);
+        else
+                list_add_tail(&dp->list, &mdev->delay_probes);
+        spin_unlock(&mdev->peer_seq_lock);
+}
+
+static int receive_delay_probe(struct drbd_conf *mdev, struct p_header *h)
+{
+        struct p_delay_probe *p = (struct p_delay_probe *)h;
+
+        ERR_IF(h->length != (sizeof(*p)-sizeof(*h))) return FALSE;
+        if (drbd_recv(mdev, h->payload, h->length) != h->length)
+                return FALSE;
+
+        got_delay_probe(mdev, USE_DATA_SOCKET, p);
+        return TRUE;
+}
+
 typedef int (*drbd_cmd_handler_f)(struct drbd_conf *, struct p_header *);
 
 static drbd_cmd_handler_f drbd_default_handler[] = {
@@ -3524,6 +3610,7 @@ static drbd_cmd_handler_f drbd_default_handler[] = {
         [P_OV_REQUEST]      = receive_DataRequest,
         [P_OV_REPLY]        = receive_DataRequest,
         [P_CSUM_RS_REQUEST]    = receive_DataRequest,
+        [P_DELAY_PROBE]     = receive_delay_probe,
         /* anything missing from this table is in
          * the asender_tbl, see get_asender_cmd */
         [P_MAX_CMD]         = NULL,
@@ -4300,6 +4387,14 @@ static int got_OVResult(struct drbd_conf *mdev, struct p_header *h)
         return TRUE;
 }
 
+static int got_delay_probe_m(struct drbd_conf *mdev, struct p_header *h)
+{
+        struct p_delay_probe *p = (struct p_delay_probe *)h;
+
+        got_delay_probe(mdev, USE_META_SOCKET, p);
+        return TRUE;
+}
+
 struct asender_cmd {
         size_t pkt_size;
         int (*process)(struct drbd_conf *mdev, struct p_header *h);
@@ -4324,6 +4419,7 @@ static struct asender_cmd *get_asender_cmd(int cmd)
         [P_BARRIER_ACK]     = { sizeof(struct p_barrier_ack), got_BarrierAck },
         [P_STATE_CHG_REPLY] = { sizeof(struct p_req_state_reply), got_RqSReply },
         [P_RS_IS_IN_SYNC]   = { sizeof(struct p_block_ack), got_IsInSync },
+        [P_DELAY_PROBE]     = { sizeof(struct p_delay_probe), got_delay_probe_m },
         [P_MAX_CMD]         = { 0, NULL },
         };
         if (cmd > P_MAX_CMD || asender_tbl[cmd].process == NULL)