drbd: New packet for Ahead/Behind mode: P_OUT_OF_SYNC

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

drivers/block/drbd/drbd_actlog.c

@@ -1007,22 +1007,22 @@ void __drbd_set_in_sync(struct drbd_conf *mdev, sector_t sector, int size,
  * called by tl_clear and drbd_send_dblock (==drbd_make_request).
  * so this can be _any_ process.
  */
-void __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector, int size,
+int __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector, int size,
 			    const char *file, const unsigned int line)
 {
 	unsigned long sbnr, ebnr, lbnr, flags;
 	sector_t esector, nr_sectors;
-	unsigned int enr, count;
+	unsigned int enr, count = 0;
 	struct lc_element *e;
 
 	if (size <= 0 || (size & 0x1ff) != 0 || size > DRBD_MAX_BIO_SIZE) {
 		dev_err(DEV, "sector: %llus, size: %d\n",
 			(unsigned long long)sector, size);
-		return;
+		return 0;
 	}
 
 	if (!get_ldev(mdev))
-		return; /* no disk, no metadata, no bitmap to set bits in */
+		return 0; /* no disk, no metadata, no bitmap to set bits in */
 
 	nr_sectors = drbd_get_capacity(mdev->this_bdev);
 	esector = sector + (size >> 9) - 1;
@@ -1052,6 +1052,8 @@ void __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector, int size,
 
 out:
 	put_ldev(mdev);
+
+	return count;
 }
 
 static

drivers/block/drbd/drbd_int.h

@@ -212,6 +212,7 @@ enum drbd_packets {
 	/* P_CKPT_FENCE_REQ      = 0x25, * currently reserved for protocol D */
 	/* P_CKPT_DISABLE_REQ    = 0x26, * currently reserved for protocol D */
 	P_DELAY_PROBE         = 0x27, /* is used on BOTH sockets */
+	P_OUT_OF_SYNC         = 0x28, /* Mark as out of sync (Outrunning), data socket */
 
 	P_MAX_CMD	      = 0x28,
 	P_MAY_IGNORE	      = 0x100, /* Flag to test if (cmd > P_MAY_IGNORE) ... */
@@ -269,6 +270,7 @@ static inline const char *cmdname(enum drbd_packets cmd)
 		[P_RS_IS_IN_SYNC]	= "CsumRSIsInSync",
 		[P_COMPRESSED_BITMAP]   = "CBitmap",
 		[P_DELAY_PROBE]         = "DelayProbe",
+		[P_OUT_OF_SYNC]		= "OutOfSync",
 		[P_MAX_CMD]	        = NULL,
 	};
 
@@ -550,6 +552,13 @@ struct p_discard {
 	u32	    pad;
 } __packed;
 
+struct p_block_desc {
+	struct p_header80 head;
+	u64 sector;
+	u32 blksize;
+	u32 pad;	/* to multiple of 8 Byte */
+} __packed;
+
 /* Valid values for the encoding field.
  * Bump proto version when changing this. */
 enum drbd_bitmap_code {
@@ -647,6 +656,7 @@ union p_polymorph {
         struct p_block_req       block_req;
 	struct p_delay_probe93   delay_probe93;
 	struct p_rs_uuid         rs_uuid;
+	struct p_block_desc      block_desc;
 } __packed;
 
 /**********************************************************************/
@@ -1221,6 +1231,7 @@ extern int drbd_send_ack_dp(struct drbd_conf *mdev, enum drbd_packets cmd,
 			struct p_data *dp, int data_size);
 extern int drbd_send_ack_ex(struct drbd_conf *mdev, enum drbd_packets cmd,
 			    sector_t sector, int blksize, u64 block_id);
+extern int drbd_send_oos(struct drbd_conf *mdev, struct drbd_request *req);
 extern int drbd_send_block(struct drbd_conf *mdev, enum drbd_packets cmd,
 			   struct drbd_epoch_entry *e);
 extern int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req);
@@ -1534,6 +1545,7 @@ extern int w_send_read_req(struct drbd_conf *, struct drbd_work *, int);
 extern int w_prev_work_done(struct drbd_conf *, struct drbd_work *, int);
 extern int w_e_reissue(struct drbd_conf *, struct drbd_work *, int);
 extern int w_restart_disk_io(struct drbd_conf *, struct drbd_work *, int);
+extern int w_send_oos(struct drbd_conf *, struct drbd_work *, int);
 
 extern void resync_timer_fn(unsigned long data);
 
@@ -1626,7 +1638,7 @@ extern void __drbd_set_in_sync(struct drbd_conf *mdev, sector_t sector,
 		int size, const char *file, const unsigned int line);
 #define drbd_set_in_sync(mdev, sector, size) \
 	__drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
-extern void __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector,
+extern int __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector,
 		int size, const char *file, const unsigned int line);
 #define drbd_set_out_of_sync(mdev, sector, size) \
 	__drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)

drivers/block/drbd/drbd_main.c

@@ -2634,6 +2634,16 @@ int drbd_send_block(struct drbd_conf *mdev, enum drbd_packets cmd,
 	return ok;
 }
 
+int drbd_send_oos(struct drbd_conf *mdev, struct drbd_request *req)
+{
+	struct p_block_desc p;
+
+	p.sector  = cpu_to_be64(req->sector);
+	p.blksize = cpu_to_be32(req->size);
+
+	return drbd_send_cmd(mdev, USE_DATA_SOCKET, P_OUT_OF_SYNC, &p.head, sizeof(p));
+}
+
 /*
   drbd_send distinguishes two cases:
 

drivers/block/drbd/drbd_receiver.c

@@ -3562,6 +3562,15 @@ static int receive_UnplugRemote(struct drbd_conf *mdev, enum drbd_packets cmd, u
 	return TRUE;
 }
 
+static int receive_out_of_sync(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned int data_size)
+{
+	struct p_block_desc *p = &mdev->data.rbuf.block_desc;
+
+	drbd_set_out_of_sync(mdev, be64_to_cpu(p->sector), be32_to_cpu(p->blksize));
+
+	return TRUE;
+}
+
 typedef int (*drbd_cmd_handler_f)(struct drbd_conf *, enum drbd_packets cmd, unsigned int to_receive);
 
 struct data_cmd {
@@ -3592,6 +3601,7 @@ static struct data_cmd drbd_cmd_handler[] = {
 	[P_OV_REPLY]        = { 1, sizeof(struct p_block_req), receive_DataRequest },
 	[P_CSUM_RS_REQUEST] = { 1, sizeof(struct p_block_req), receive_DataRequest },
 	[P_DELAY_PROBE]     = { 0, sizeof(struct p_delay_probe93), receive_skip },
+	[P_OUT_OF_SYNC]     = { 0, sizeof(struct p_block_desc), receive_out_of_sync },
 	/* anything missing from this table is in
 	 * the asender_tbl, see get_asender_cmd */
 	[P_MAX_CMD]	    = { 0, 0, NULL },

drivers/block/drbd/drbd_req.c

@@ -142,7 +142,7 @@ static void _about_to_complete_local_write(struct drbd_conf *mdev,
 
 	/* before we can signal completion to the upper layers,
 	 * we may need to close the current epoch */
-	if (mdev->state.conn >= C_CONNECTED &&
+	if (mdev->state.conn >= C_CONNECTED && mdev->state.conn < C_AHEAD &&
 	    req->epoch == mdev->newest_tle->br_number)
 		queue_barrier(mdev);
 
@@ -545,6 +545,14 @@ int __req_mod(struct drbd_request *req, enum drbd_req_event what,
 
 		break;
 
+	case queue_for_send_oos:
+		req->rq_state |= RQ_NET_QUEUED;
+		req->w.cb =  w_send_oos;
+		drbd_queue_work(&mdev->data.work, &req->w);
+		break;
+
+	case oos_handed_to_network:
+		/* actually the same */
 	case send_canceled:
 		/* treat it the same */
 	case send_failed:
@@ -756,7 +764,7 @@ static int drbd_make_request_common(struct drbd_conf *mdev, struct bio *bio)
 	const sector_t sector = bio->bi_sector;
 	struct drbd_tl_epoch *b = NULL;
 	struct drbd_request *req;
-	int local, remote;
+	int local, remote, send_oos = 0;
 	int err = -EIO;
 	int ret = 0;
 
@@ -820,8 +828,11 @@ static int drbd_make_request_common(struct drbd_conf *mdev, struct bio *bio)
 	}
 
 	remote = remote && (mdev->state.pdsk == D_UP_TO_DATE ||
-			    (mdev->state.pdsk == D_INCONSISTENT &&
-			     mdev->state.conn >= C_CONNECTED));
+			    (mdev->state.pdsk >= D_INCONSISTENT &&
+			     mdev->state.conn >= C_CONNECTED &&
+			     mdev->state.conn < C_AHEAD));
+	send_oos = (rw == WRITE && mdev->state.conn == C_AHEAD &&
+		    mdev->state.pdsk >= D_INCONSISTENT);
 
 	if (!(local || remote) && !is_susp(mdev->state)) {
 		if (__ratelimit(&drbd_ratelimit_state))
@@ -835,7 +846,7 @@ static int drbd_make_request_common(struct drbd_conf *mdev, struct bio *bio)
 	 * but there is a race between testing the bit and pointer outside the
 	 * spinlock, and grabbing the spinlock.
 	 * if we lost that race, we retry.  */
-	if (rw == WRITE && remote &&
+	if (rw == WRITE && (remote || send_oos) &&
 	    mdev->unused_spare_tle == NULL &&
 	    test_bit(CREATE_BARRIER, &mdev->flags)) {
 allocate_barrier:
@@ -860,11 +871,15 @@ static int drbd_make_request_common(struct drbd_conf *mdev, struct bio *bio)
 		goto fail_free_complete;
 	}
 
-	if (remote) {
+	if (remote || send_oos) {
 		remote = (mdev->state.pdsk == D_UP_TO_DATE ||
-			    (mdev->state.pdsk == D_INCONSISTENT &&
-			     mdev->state.conn >= C_CONNECTED));
-		if (!remote)
+			    (mdev->state.pdsk >= D_INCONSISTENT &&
+			     mdev->state.conn >= C_CONNECTED &&
+			     mdev->state.conn < C_AHEAD));
+		send_oos = (rw == WRITE && mdev->state.conn == C_AHEAD &&
+			    mdev->state.pdsk >= D_INCONSISTENT);
+
+		if (!(remote || send_oos))
 			dev_warn(DEV, "lost connection while grabbing the req_lock!\n");
 		if (!(local || remote)) {
 			dev_err(DEV, "IO ERROR: neither local nor remote disk\n");
@@ -877,7 +892,7 @@ static int drbd_make_request_common(struct drbd_conf *mdev, struct bio *bio)
 		mdev->unused_spare_tle = b;
 		b = NULL;
 	}
-	if (rw == WRITE && remote &&
+	if (rw == WRITE && (remote || send_oos) &&
 	    mdev->unused_spare_tle == NULL &&
 	    test_bit(CREATE_BARRIER, &mdev->flags)) {
 		/* someone closed the current epoch
@@ -900,7 +915,7 @@ static int drbd_make_request_common(struct drbd_conf *mdev, struct bio *bio)
 	 * barrier packet.  To get the write ordering right, we only have to
 	 * make sure that, if this is a write request and it triggered a
 	 * barrier packet, this request is queued within the same spinlock. */
-	if (remote && mdev->unused_spare_tle &&
+	if ((remote || send_oos) && mdev->unused_spare_tle &&
 	    test_and_clear_bit(CREATE_BARRIER, &mdev->flags)) {
 		_tl_add_barrier(mdev, mdev->unused_spare_tle);
 		mdev->unused_spare_tle = NULL;
@@ -948,8 +963,11 @@ static int drbd_make_request_common(struct drbd_conf *mdev, struct bio *bio)
 				? queue_for_net_write
 				: queue_for_net_read);
 	}
+	if (send_oos && drbd_set_out_of_sync(mdev, sector, size))
+		_req_mod(req, queue_for_send_oos);
 
-	if (remote && mdev->net_conf->on_congestion != OC_BLOCK) {
+	if (remote &&
+	    mdev->net_conf->on_congestion != OC_BLOCK && mdev->agreed_pro_version >= 96) {
 		int congested = 0;
 
 		if (mdev->net_conf->cong_fill &&
@@ -964,6 +982,8 @@ static int drbd_make_request_common(struct drbd_conf *mdev, struct bio *bio)
 		}
 
 		if (congested) {
+			queue_barrier(mdev);
+
 			if (mdev->net_conf->on_congestion == OC_PULL_AHEAD)
 				_drbd_set_state(_NS(mdev, conn, C_AHEAD), 0, NULL);
 			else  /*mdev->net_conf->on_congestion == OC_DISCONNECT */

drivers/block/drbd/drbd_req.h

@@ -82,14 +82,16 @@ enum drbd_req_event {
 	to_be_submitted,
 
 	/* XXX yes, now I am inconsistent...
-	 * these two are not "events" but "actions"
+	 * these are not "events" but "actions"
 	 * oh, well... */
 	queue_for_net_write,
 	queue_for_net_read,
+	queue_for_send_oos,
 
 	send_canceled,
 	send_failed,
 	handed_over_to_network,
+	oos_handed_to_network,
 	connection_lost_while_pending,
 	read_retry_remote_canceled,
 	recv_acked_by_peer,

drivers/block/drbd/drbd_worker.c

@@ -1237,6 +1237,22 @@ int w_send_write_hint(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
 	return drbd_send_short_cmd(mdev, P_UNPLUG_REMOTE);
 }
 
+int w_send_oos(struct drbd_conf *mdev, struct drbd_work *w, int cancel)
+{
+	struct drbd_request *req = container_of(w, struct drbd_request, w);
+	int ok;
+
+	if (unlikely(cancel)) {
+		req_mod(req, send_canceled);
+		return 1;
+	}
+
+	ok = drbd_send_oos(mdev, req);
+	req_mod(req, oos_handed_to_network);
+
+	return ok;
+}
+
 /**
  * w_send_dblock() - Worker callback to send a P_DATA packet in order to mirror a write request
  * @mdev:	DRBD device.

include/linux/drbd.h

@@ -56,7 +56,7 @@ extern const char *drbd_buildtag(void);
 #define REL_VERSION "8.3.9"
 #define API_VERSION 88
 #define PRO_VERSION_MIN 86
-#define PRO_VERSION_MAX 95
+#define PRO_VERSION_MAX 96
 
 
 enum drbd_io_error_p {