drbd: Protect accesses to the uuid set with a spinlock

There is at least the worker context, the receiver context, the context of
receiving netlink packts.

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

drivers/block/drbd/drbd_int.h

@@ -747,6 +747,7 @@ struct drbd_md {
 	u64 md_offset;		/* sector offset to 'super' block */
 
 	u64 la_size_sect;	/* last agreed size, unit sectors */
+	spinlock_t uuid_lock;
 	u64 uuid[UI_SIZE];
 	u64 device_uuid;
 	u32 flags;
@@ -1119,8 +1120,9 @@ extern int  drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev);
 extern void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
 extern void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
 extern void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
-extern void _drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
 extern void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local);
+extern void drbd_uuid_move_history(struct drbd_conf *mdev) __must_hold(local);
+extern void __drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
 extern void drbd_md_set_flag(struct drbd_conf *mdev, int flags) __must_hold(local);
 extern void drbd_md_clear_flag(struct drbd_conf *mdev, int flags)__must_hold(local);
 extern int drbd_md_test_flag(struct drbd_backing_dev *, int);

drivers/block/drbd/drbd_main.c

@@ -838,8 +838,10 @@ int _drbd_send_uuids(struct drbd_conf *mdev, u64 uuid_flags)
 		put_ldev(mdev);
 		return -EIO;
 	}
+	spin_lock_irq(&mdev->ldev->md.uuid_lock);
 	for (i = UI_CURRENT; i < UI_SIZE; i++)
 		p->uuid[i] = mdev->ldev ? cpu_to_be64(mdev->ldev->md.uuid[i]) : 0;
+	spin_unlock_irq(&mdev->ldev->md.uuid_lock);
 
 	mdev->comm_bm_set = drbd_bm_total_weight(mdev);
 	p->uuid[UI_SIZE] = cpu_to_be64(mdev->comm_bm_set);
@@ -3015,7 +3017,7 @@ void drbd_md_mark_dirty(struct drbd_conf *mdev)
 }
 #endif
 
-static void drbd_uuid_move_history(struct drbd_conf *mdev) __must_hold(local)
+void drbd_uuid_move_history(struct drbd_conf *mdev) __must_hold(local)
 {
 	int i;
 
@@ -3023,7 +3025,7 @@ static void drbd_uuid_move_history(struct drbd_conf *mdev) __must_hold(local)
 		mdev->ldev->md.uuid[i+1] = mdev->ldev->md.uuid[i];
 }
 
-void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local)
+void __drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local)
 {
 	if (idx == UI_CURRENT) {
 		if (mdev->state.role == R_PRIMARY)
@@ -3038,14 +3040,24 @@ void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local)
 	drbd_md_mark_dirty(mdev);
 }
 
+void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local)
+{
+	unsigned long flags;
+	spin_lock_irqsave(&mdev->ldev->md.uuid_lock, flags);
+	__drbd_uuid_set(mdev, idx, val);
+	spin_unlock_irqrestore(&mdev->ldev->md.uuid_lock, flags);
+}
 
 void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local)
 {
+	unsigned long flags;
+	spin_lock_irqsave(&mdev->ldev->md.uuid_lock, flags);
 	if (mdev->ldev->md.uuid[idx]) {
 		drbd_uuid_move_history(mdev);
 		mdev->ldev->md.uuid[UI_HISTORY_START] = mdev->ldev->md.uuid[idx];
 	}
-	_drbd_uuid_set(mdev, idx, val);
+	__drbd_uuid_set(mdev, idx, val);
+	spin_unlock_irqrestore(&mdev->ldev->md.uuid_lock, flags);
 }
 
 /**
@@ -3058,15 +3070,20 @@ void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local)
 void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local)
 {
 	u64 val;
-	unsigned long long bm_uuid = mdev->ldev->md.uuid[UI_BITMAP];
+	unsigned long long bm_uuid;
+
+	get_random_bytes(&val, sizeof(u64));
+
+	spin_lock_irq(&mdev->ldev->md.uuid_lock);
+	bm_uuid = mdev->ldev->md.uuid[UI_BITMAP];
 
 	if (bm_uuid)
 		dev_warn(DEV, "bm UUID was already set: %llX\n", bm_uuid);
 
 	mdev->ldev->md.uuid[UI_BITMAP] = mdev->ldev->md.uuid[UI_CURRENT];
+	__drbd_uuid_set(mdev, UI_CURRENT, val);
+	spin_unlock_irq(&mdev->ldev->md.uuid_lock);
 
-	get_random_bytes(&val, sizeof(u64));
-	_drbd_uuid_set(mdev, UI_CURRENT, val);
 	drbd_print_uuids(mdev, "new current UUID");
 	/* get it to stable storage _now_ */
 	drbd_md_sync(mdev);
@@ -3074,9 +3091,11 @@ void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local)
 
 void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local)
 {
+	unsigned long flags;
 	if (mdev->ldev->md.uuid[UI_BITMAP] == 0 && val == 0)
 		return;
 
+	spin_lock_irqsave(&mdev->ldev->md.uuid_lock, flags);
 	if (val == 0) {
 		drbd_uuid_move_history(mdev);
 		mdev->ldev->md.uuid[UI_HISTORY_START] = mdev->ldev->md.uuid[UI_BITMAP];
@@ -3088,6 +3107,8 @@ void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local)
 
 		mdev->ldev->md.uuid[UI_BITMAP] = val & ~((u64)1);
 	}
+	spin_unlock_irqrestore(&mdev->ldev->md.uuid_lock, flags);
+
 	drbd_md_mark_dirty(mdev);
 }
 

drivers/block/drbd/drbd_nl.c

@@ -1320,6 +1320,8 @@ int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info)
 		retcode = ERR_NOMEM;
 		goto fail;
 	}
+	spin_lock_init(&nbc->md.uuid_lock);
+
 	new_disk_conf = kzalloc(sizeof(struct disk_conf), GFP_KERNEL);
 	if (!new_disk_conf) {
 		retcode = ERR_NOMEM;
@@ -2679,8 +2681,16 @@ int nla_put_status_info(struct sk_buff *skb, struct drbd_conf *mdev,
 		goto nla_put_failure;
 
 	if (got_ldev) {
+		int err;
+
+		spin_lock_irq(&mdev->ldev->md.uuid_lock);
+		err = nla_put(skb, T_uuids, sizeof(si->uuids), mdev->ldev->md.uuid);
+		spin_unlock_irq(&mdev->ldev->md.uuid_lock);
+
+		if (err)
+			goto nla_put_failure;
+
 		if (nla_put_u32(skb, T_disk_flags, mdev->ldev->md.flags) ||
-		    nla_put(skb, T_uuids, sizeof(si->uuids), mdev->ldev->md.uuid) ||
 		    nla_put_u64(skb, T_bits_total, drbd_bm_bits(mdev)) ||
 		    nla_put_u64(skb, T_bits_oos, drbd_bm_total_weight(mdev)))
 			goto nla_put_failure;

drivers/block/drbd/drbd_receiver.c

@@ -2796,7 +2796,9 @@ static int drbd_uuid_compare(struct drbd_conf *mdev, int *rule_nr) __must_hold(l
 			if ((mdev->ldev->md.uuid[UI_BITMAP] & ~((u64)1)) == (mdev->p_uuid[UI_HISTORY_START] & ~((u64)1)) &&
 			    (mdev->ldev->md.uuid[UI_HISTORY_START] & ~((u64)1)) == (mdev->p_uuid[UI_HISTORY_START + 1] & ~((u64)1))) {
 				dev_info(DEV, "was SyncSource, missed the resync finished event, corrected myself:\n");
-				drbd_uuid_set_bm(mdev, 0UL);
+				drbd_uuid_move_history(mdev);
+				mdev->ldev->md.uuid[UI_HISTORY_START] = mdev->ldev->md.uuid[UI_BITMAP];
+				mdev->ldev->md.uuid[UI_BITMAP] = 0;
 
 				drbd_uuid_dump(mdev, "self", mdev->ldev->md.uuid,
 					       mdev->state.disk >= D_NEGOTIATING ? drbd_bm_total_weight(mdev) : 0, 0);
@@ -2904,8 +2906,8 @@ static int drbd_uuid_compare(struct drbd_conf *mdev, int *rule_nr) __must_hold(l
 			if (mdev->tconn->agreed_pro_version < 91)
 				return -1091;
 
-			_drbd_uuid_set(mdev, UI_BITMAP, mdev->ldev->md.uuid[UI_HISTORY_START]);
-			_drbd_uuid_set(mdev, UI_HISTORY_START, mdev->ldev->md.uuid[UI_HISTORY_START + 1]);
+			__drbd_uuid_set(mdev, UI_BITMAP, mdev->ldev->md.uuid[UI_HISTORY_START]);
+			__drbd_uuid_set(mdev, UI_HISTORY_START, mdev->ldev->md.uuid[UI_HISTORY_START + 1]);
 
 			dev_info(DEV, "Last syncUUID did not get through, corrected:\n");
 			drbd_uuid_dump(mdev, "self", mdev->ldev->md.uuid,
@@ -2959,11 +2961,14 @@ static enum drbd_conns drbd_sync_handshake(struct drbd_conf *mdev, enum drbd_rol
 		mydisk = mdev->new_state_tmp.disk;
 
 	dev_info(DEV, "drbd_sync_handshake:\n");
+
+	spin_lock_irq(&mdev->ldev->md.uuid_lock);
 	drbd_uuid_dump(mdev, "self", mdev->ldev->md.uuid, mdev->comm_bm_set, 0);
 	drbd_uuid_dump(mdev, "peer", mdev->p_uuid,
 		       mdev->p_uuid[UI_SIZE], mdev->p_uuid[UI_FLAGS]);
 
 	hg = drbd_uuid_compare(mdev, &rule_nr);
+	spin_unlock_irq(&mdev->ldev->md.uuid_lock);
 
 	dev_info(DEV, "uuid_compare()=%d by rule %d\n", hg, rule_nr);