forked from ~ljy/RK356X_SDK_RELEASE
blame | history | patch | commit | commitdiff
hc
2024-02-20 ea08eeccae9297f7aabd2ef7f0c2517ac4549acc 
 kernel/drivers/block/drbd/drbd_main.c
....@@ -1,3 +1,4 @@
1
+// SPDX-License-Identifier: GPL-2.0-or-later
12 /*
23    drbd.c
34 
....@@ -10,19 +11,6 @@
1011    Thanks to Carter Burden, Bart Grantham and Gennadiy Nerubayev
1112    from Logicworks, Inc. for making SDP replication support possible.
1213 
13
-   drbd is free software; you can redistribute it and/or modify
14
-   it under the terms of the GNU General Public License as published by
15
-   the Free Software Foundation; either version 2, or (at your option)
16
-   any later version.
17
-
18
-   drbd is distributed in the hope that it will be useful,
19
-   but WITHOUT ANY WARRANTY; without even the implied warranty of
20
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
21
-   GNU General Public License for more details.
22
-
23
-   You should have received a copy of the GNU General Public License
24
-   along with drbd; see the file COPYING.  If not, write to
25
-   the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
2614 
2715  */
2816 
....@@ -144,9 +132,10 @@
144132 DEFINE_RATELIMIT_STATE(drbd_ratelimit_state, 5 * HZ, 5);
145133 
146134 static const struct block_device_operations drbd_ops = {
147
-	.owner =   THIS_MODULE,
148
-	.open =    drbd_open,
149
-	.release = drbd_release,
135
+	.owner		= THIS_MODULE,
136
+	.submit_bio	= drbd_submit_bio,
137
+	.open		= drbd_open,
138
+	.release	= drbd_release,
150139 };
151140 
152141 struct bio *bio_alloc_drbd(gfp_t gfp_mask)
....@@ -195,7 +184,7 @@
195184 		unsigned int set_size)
196185 {
197186 	struct drbd_request *r;
198
-	struct drbd_request *req = NULL;
187
+	struct drbd_request *req = NULL, *tmp = NULL;
199188 	int expect_epoch = 0;
200189 	int expect_size = 0;
201190 
....@@ -249,8 +238,11 @@
249238 	 * to catch requests being barrier-acked "unexpectedly".
250239 	 * It usually should find the same req again, or some READ preceding it. */
251240 	list_for_each_entry(req, &connection->transfer_log, tl_requests)
252
-		if (req->epoch == expect_epoch)
241
+		if (req->epoch == expect_epoch) {
242
+			tmp = req;
253243 			break;
244
+		}
245
+	req = list_prepare_entry(tmp, &connection->transfer_log, tl_requests);
254246 	list_for_each_entry_safe_from(req, r, &connection->transfer_log, tl_requests) {
255247 		if (req->epoch != expect_epoch)
256248 			break;
....@@ -441,7 +433,7 @@
441433 		thi->t_state = RESTARTING;
442434 		drbd_info(resource, "Restarting %s thread (from %s [%d])\n",
443435 				thi->name, current->comm, current->pid);
444
-		/* fall through */
436
+		fallthrough;
445437 	case RUNNING:
446438 	case RESTARTING:
447439 	default:
....@@ -479,7 +471,7 @@
479471 		smp_mb();
480472 		init_completion(&thi->stop);
481473 		if (thi->task != current)
482
-			force_sig(DRBD_SIGKILL, thi->task);
474
+			send_sig(DRBD_SIGKILL, thi->task, 1);
483475 	}
484476 
485477 	spin_unlock_irqrestore(&thi->t_lock, flags);
....@@ -672,7 +664,7 @@
672664 	/* DRBD protocol "pings" are latency critical.
673665 	 * This is supposed to trigger tcp_push_pending_frames() */
674666 	if (!err && (cmd == P_PING || cmd == P_PING_ACK))
675
-		drbd_tcp_nodelay(sock->socket);
667
+		tcp_sock_set_nodelay(sock->socket->sk);
676668 
677669 	return err;
678670 }
....@@ -995,7 +987,10 @@
995987 
996988 	p->d_size = cpu_to_be64(d_size);
997989 	p->u_size = cpu_to_be64(u_size);
998
-	p->c_size = cpu_to_be64(trigger_reply ? 0 : drbd_get_capacity(device->this_bdev));
990
+	if (trigger_reply)
991
+		p->c_size = 0;
992
+	else
993
+		p->c_size = cpu_to_be64(get_capacity(device->vdisk));
999994 	p->max_bio_size = cpu_to_be32(max_bio_size);
1000995 	p->queue_order_type = cpu_to_be16(q_order_type);
1001996 	p->dds_flags = cpu_to_be16(flags);
....@@ -1378,7 +1373,7 @@
13781373 		      struct p_data *dp, int data_size)
13791374 {
13801375 	if (peer_device->connection->peer_integrity_tfm)
1381
-		data_size -= crypto_ahash_digestsize(peer_device->connection->peer_integrity_tfm);
1376
+		data_size -= crypto_shash_digestsize(peer_device->connection->peer_integrity_tfm);
13821377 	_drbd_send_ack(peer_device, cmd, dp->sector, cpu_to_be32(data_size),
13831378 		       dp->block_id);
13841379 }
....@@ -1564,7 +1559,7 @@
15641559 	 * put_page(); and would cause either a VM_BUG directly, or
15651560 	 * __page_cache_release a page that would actually still be referenced
15661561 	 * by someone, leading to some obscure delayed Oops somewhere else. */
1567
-	if (drbd_disable_sendpage || (page_count(page) < 1) || PageSlab(page))
1562
+	if (drbd_disable_sendpage || !sendpage_ok(page))
15681563 		return _drbd_no_send_page(peer_device, page, offset, size, msg_flags);
15691564 
15701565 	msg_flags |= MSG_NOSIGNAL;
....@@ -1669,12 +1664,16 @@
16691664 			(bio->bi_opf & REQ_PREFLUSH ? DP_FLUSH : 0) |
16701665 			(bio_op(bio) == REQ_OP_WRITE_SAME ? DP_WSAME : 0) |
16711666 			(bio_op(bio) == REQ_OP_DISCARD ? DP_DISCARD : 0) |
1672
-			(bio_op(bio) == REQ_OP_WRITE_ZEROES ? DP_DISCARD : 0);
1667
+			(bio_op(bio) == REQ_OP_WRITE_ZEROES ?
1668
+			  ((connection->agreed_features & DRBD_FF_WZEROES) ?
1669
+			   (DP_ZEROES |(!(bio->bi_opf & REQ_NOUNMAP) ? DP_DISCARD : 0))
1670
+			   : DP_DISCARD)
1671
+			: 0);
16731672 	else
16741673 		return bio->bi_opf & REQ_SYNC ? DP_RW_SYNC : 0;
16751674 }
16761675 
1677
-/* Used to send write or TRIM aka REQ_DISCARD requests
1676
+/* Used to send write or TRIM aka REQ_OP_DISCARD requests
16781677  * R_PRIMARY -> Peer	(P_DATA, P_TRIM)
16791678  */
16801679 int drbd_send_dblock(struct drbd_peer_device *peer_device, struct drbd_request *req)
....@@ -1691,7 +1690,7 @@
16911690 	sock = &peer_device->connection->data;
16921691 	p = drbd_prepare_command(peer_device, sock);
16931692 	digest_size = peer_device->connection->integrity_tfm ?
1694
-		      crypto_ahash_digestsize(peer_device->connection->integrity_tfm) : 0;
1693
+		      crypto_shash_digestsize(peer_device->connection->integrity_tfm) : 0;
16951694 
16961695 	if (!p)
16971696 		return -EIO;
....@@ -1713,10 +1712,11 @@
17131712 	}
17141713 	p->dp_flags = cpu_to_be32(dp_flags);
17151714 
1716
-	if (dp_flags & DP_DISCARD) {
1715
+	if (dp_flags & (DP_DISCARD|DP_ZEROES)) {
1716
+		enum drbd_packet cmd = (dp_flags & DP_ZEROES) ? P_ZEROES : P_TRIM;
17171717 		struct p_trim *t = (struct p_trim*)p;
17181718 		t->size = cpu_to_be32(req->i.size);
1719
-		err = __send_command(peer_device->connection, device->vnr, sock, P_TRIM, sizeof(*t), NULL, 0);
1719
+		err = __send_command(peer_device->connection, device->vnr, sock, cmd, sizeof(*t), NULL, 0);
17201720 		goto out;
17211721 	}
17221722 	if (dp_flags & DP_WSAME) {
....@@ -1797,7 +1797,7 @@
17971797 	p = drbd_prepare_command(peer_device, sock);
17981798 
17991799 	digest_size = peer_device->connection->integrity_tfm ?
1800
-		      crypto_ahash_digestsize(peer_device->connection->integrity_tfm) : 0;
1800
+		      crypto_shash_digestsize(peer_device->connection->integrity_tfm) : 0;
18011801 
18021802 	if (!p)
18031803 		return -EIO;
....@@ -1857,7 +1857,7 @@
18571857 
18581858 	/* THINK  if (signal_pending) return ... ? */
18591859 
1860
-	iov_iter_kvec(&msg.msg_iter, WRITE | ITER_KVEC, &iov, 1, size);
1860
+	iov_iter_kvec(&msg.msg_iter, WRITE, &iov, 1, size);
18611861 
18621862 	if (sock == connection->data.socket) {
18631863 		rcu_read_lock();
....@@ -2035,6 +2035,17 @@
20352035 	device->local_max_bio_size = DRBD_MAX_BIO_SIZE_SAFE;
20362036 }
20372037 
2038
+void drbd_set_my_capacity(struct drbd_device *device, sector_t size)
2039
+{
2040
+	char ppb[10];
2041
+
2042
+	set_capacity(device->vdisk, size);
2043
+	revalidate_disk_size(device->vdisk, false);
2044
+
2045
+	drbd_info(device, "size = %s (%llu KB)\n",
2046
+		ppsize(ppb, size>>1), (unsigned long long)size>>1);
2047
+}
2048
+
20382049 void drbd_device_cleanup(struct drbd_device *device)
20392050 {
20402051 	int i;
....@@ -2060,7 +2071,8 @@
20602071 	}
20612072 	D_ASSERT(device, first_peer_device(device)->connection->net_conf == NULL);
20622073 
2063
-	drbd_set_my_capacity(device, 0);
2074
+	set_capacity(device->vdisk, 0);
2075
+	revalidate_disk_size(device->vdisk, false);
20642076 	if (device->bitmap) {
20652077 		/* maybe never allocated. */
20662078 		drbd_bm_resize(device, 0, 1);
....@@ -2227,9 +2239,6 @@
22272239 	/* cleanup stuff that may have been allocated during
22282240 	 * device (re-)configuration or state changes */
22292241 
2230
-	if (device->this_bdev)
2231
-		bdput(device->this_bdev);
2232
-
22332242 	drbd_backing_dev_free(device, device->ldev);
22342243 	device->ldev = NULL;
22352244 
....@@ -2316,7 +2325,7 @@
23162325 		 * workqueues instead.
23172326 		 */
23182327 
2319
-		/* We are not just doing generic_make_request(),
2328
+		/* We are not just doing submit_bio_noacct(),
23202329 		 * as we want to keep the start_time information. */
23212330 		inc_ap_bio(device);
23222331 		__drbd_make_request(device, bio, start_jif);
....@@ -2404,62 +2413,6 @@
24042413 	idr_destroy(&drbd_devices);
24052414 
24062415 	pr_info("module cleanup done.\n");
2407
-}
2408
-
2409
-/**
2410
- * drbd_congested() - Callback for the flusher thread
2411
- * @congested_data:	User data
2412
- * @bdi_bits:		Bits the BDI flusher thread is currently interested in
2413
- *
2414
- * Returns 1<<WB_async_congested and/or 1<<WB_sync_congested if we are congested.
2415
- */
2416
-static int drbd_congested(void *congested_data, int bdi_bits)
2417
-{
2418
-	struct drbd_device *device = congested_data;
2419
-	struct request_queue *q;
2420
-	char reason = '-';
2421
-	int r = 0;
2422
-
2423
-	if (!may_inc_ap_bio(device)) {
2424
-		/* DRBD has frozen IO */
2425
-		r = bdi_bits;
2426
-		reason = 'd';
2427
-		goto out;
2428
-	}
2429
-
2430
-	if (test_bit(CALLBACK_PENDING, &first_peer_device(device)->connection->flags)) {
2431
-		r |= (1 << WB_async_congested);
2432
-		/* Without good local data, we would need to read from remote,
2433
-		 * and that would need the worker thread as well, which is
2434
-		 * currently blocked waiting for that usermode helper to
2435
-		 * finish.
2436
-		 */
2437
-		if (!get_ldev_if_state(device, D_UP_TO_DATE))
2438
-			r |= (1 << WB_sync_congested);
2439
-		else
2440
-			put_ldev(device);
2441
-		r &= bdi_bits;
2442
-		reason = 'c';
2443
-		goto out;
2444
-	}
2445
-
2446
-	if (get_ldev(device)) {
2447
-		q = bdev_get_queue(device->ldev->backing_bdev);
2448
-		r = bdi_congested(q->backing_dev_info, bdi_bits);
2449
-		put_ldev(device);
2450
-		if (r)
2451
-			reason = 'b';
2452
-	}
2453
-
2454
-	if (bdi_bits & (1 << WB_async_congested) &&
2455
-	    test_bit(NET_CONGESTED, &first_peer_device(device)->connection->flags)) {
2456
-		r |= (1 << WB_async_congested);
2457
-		reason = reason == 'b' ? 'a' : 'n';
2458
-	}
2459
-
2460
-out:
2461
-	device->congestion_reason = reason;
2462
-	return r;
24632416 }
24642417 
24652418 static void drbd_init_workqueue(struct drbd_work_queue* wq)
....@@ -2558,11 +2511,11 @@
25582511 {
25592512 	drbd_free_sock(connection);
25602513 
2561
-	crypto_free_ahash(connection->csums_tfm);
2562
-	crypto_free_ahash(connection->verify_tfm);
2514
+	crypto_free_shash(connection->csums_tfm);
2515
+	crypto_free_shash(connection->verify_tfm);
25632516 	crypto_free_shash(connection->cram_hmac_tfm);
2564
-	crypto_free_ahash(connection->integrity_tfm);
2565
-	crypto_free_ahash(connection->peer_integrity_tfm);
2517
+	crypto_free_shash(connection->integrity_tfm);
2518
+	crypto_free_shash(connection->peer_integrity_tfm);
25662519 	kfree(connection->int_dig_in);
25672520 	kfree(connection->int_dig_vv);
25682521 
....@@ -2767,7 +2720,7 @@
27672720 enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsigned int minor)
27682721 {
27692722 	struct drbd_resource *resource = adm_ctx->resource;
2770
-	struct drbd_connection *connection;
2723
+	struct drbd_connection *connection, *n;
27712724 	struct drbd_device *device;
27722725 	struct drbd_peer_device *peer_device, *tmp_peer_device;
27732726 	struct gendisk *disk;
....@@ -2793,11 +2746,10 @@
27932746 
27942747 	drbd_init_set_defaults(device);
27952748 
2796
-	q = blk_alloc_queue_node(GFP_KERNEL, NUMA_NO_NODE, &resource->req_lock);
2749
+	q = blk_alloc_queue(NUMA_NO_NODE);
27972750 	if (!q)
27982751 		goto out_no_q;
27992752 	device->rq_queue = q;
2800
-	q->queuedata   = device;
28012753 
28022754 	disk = alloc_disk(1);
28032755 	if (!disk)
....@@ -2813,14 +2765,6 @@
28132765 	sprintf(disk->disk_name, "drbd%d", minor);
28142766 	disk->private_data = device;
28152767 
2816
-	device->this_bdev = bdget(MKDEV(DRBD_MAJOR, minor));
2817
-	/* we have no partitions. we contain only ourselves. */
2818
-	device->this_bdev->bd_contains = device->this_bdev;
2819
-
2820
-	q->backing_dev_info->congested_fn = drbd_congested;
2821
-	q->backing_dev_info->congested_data = device;
2822
-
2823
-	blk_queue_make_request(q, drbd_make_request);
28242768 	blk_queue_write_cache(q, true, true);
28252769 	/* Setting the max_hw_sectors to an odd value of 8kibyte here
28262770 	   This triggers a max_bio_size message upon first attach or connect */
....@@ -2875,7 +2819,7 @@
28752819 
28762820 	if (init_submitter(device)) {
28772821 		err = ERR_NOMEM;
2878
-		goto out_idr_remove_vol;
2822
+		goto out_idr_remove_from_resource;
28792823 	}
28802824 
28812825 	add_disk(disk);
....@@ -2892,10 +2836,8 @@
28922836 	drbd_debugfs_device_add(device);
28932837 	return NO_ERROR;
28942838 
2895
-out_idr_remove_vol:
2896
-	idr_remove(&connection->peer_devices, vnr);
28972839 out_idr_remove_from_resource:
2898
-	for_each_connection(connection, resource) {
2840
+	for_each_connection_safe(connection, n, resource) {
28992841 		peer_device = idr_remove(&connection->peer_devices, vnr);
29002842 		if (peer_device)
29012843 			kref_put(&connection->kref, drbd_destroy_connection);
....@@ -3002,8 +2944,7 @@
30022944 	spin_lock_init(&retry.lock);
30032945 	INIT_LIST_HEAD(&retry.writes);
30042946 
3005
-	if (drbd_debugfs_init())
3006
-		pr_notice("failed to initialize debugfs -- will not be available\n");
2947
+	drbd_debugfs_init();
30072948 
30082949 	pr_info("initialized. "
30092950 	       "Version: " REL_VERSION " (api:%d/proto:%d-%d)\n",
....@@ -3097,7 +3038,7 @@
30973038 
30983039 	memset(buffer, 0, sizeof(*buffer));
30993040 
3100
-	buffer->la_size_sect = cpu_to_be64(drbd_get_capacity(device->this_bdev));
3041
+	buffer->la_size_sect = cpu_to_be64(get_capacity(device->vdisk));
31013042 	for (i = UI_CURRENT; i < UI_SIZE; i++)
31023043 		buffer->uuid[i] = cpu_to_be64(device->ldev->md.uuid[i]);
31033044 	buffer->flags = cpu_to_be32(device->ldev->md.flags);
....@@ -3155,7 +3096,7 @@
31553096 
31563097 	/* Update device->ldev->md.la_size_sect,
31573098 	 * since we updated it on metadata. */
3158
-	device->ldev->md.la_size_sect = drbd_get_capacity(device->this_bdev);
3099
+	device->ldev->md.la_size_sect = get_capacity(device->vdisk);
31593100 
31603101 	drbd_md_put_buffer(device);
31613102 out:
....@@ -3407,22 +3348,11 @@
34073348  * the meta-data super block. This function sets MD_DIRTY, and starts a
34083349  * timer that ensures that within five seconds you have to call drbd_md_sync().
34093350  */
3410
-#ifdef DEBUG
3411
-void drbd_md_mark_dirty_(struct drbd_device *device, unsigned int line, const char *func)
3412
-{
3413
-	if (!test_and_set_bit(MD_DIRTY, &device->flags)) {
3414
-		mod_timer(&device->md_sync_timer, jiffies + HZ);
3415
-		device->last_md_mark_dirty.line = line;
3416
-		device->last_md_mark_dirty.func = func;
3417
-	}
3418
-}
3419
-#else
34203351 void drbd_md_mark_dirty(struct drbd_device *device)
34213352 {
34223353 	if (!test_and_set_bit(MD_DIRTY, &device->flags))
34233354 		mod_timer(&device->md_sync_timer, jiffies + 5*HZ);
34243355 }
3425
-#endif
34263356 
34273357 void drbd_uuid_move_history(struct drbd_device *device) __must_hold(local)
34283358 {
....@@ -3699,9 +3629,8 @@
36993629 	 * when we want to support more than
37003630 	 * one PRO_VERSION */
37013631 	static const char *cmdnames[] = {
3632
+
37023633 		[P_DATA]	        = "Data",
3703
-		[P_WSAME]	        = "WriteSame",
3704
-		[P_TRIM]	        = "Trim",
37053634 		[P_DATA_REPLY]	        = "DataReply",
37063635 		[P_RS_DATA_REPLY]	= "RSDataReply",
37073636 		[P_BARRIER]	        = "Barrier",
....@@ -3712,7 +3641,6 @@
37123641 		[P_DATA_REQUEST]	= "DataRequest",
37133642 		[P_RS_DATA_REQUEST]     = "RSDataRequest",
37143643 		[P_SYNC_PARAM]	        = "SyncParam",
3715
-		[P_SYNC_PARAM89]	= "SyncParam89",
37163644 		[P_PROTOCOL]            = "ReportProtocol",
37173645 		[P_UUIDS]	        = "ReportUUIDs",
37183646 		[P_SIZES]	        = "ReportSizes",
....@@ -3720,6 +3648,7 @@
37203648 		[P_SYNC_UUID]           = "ReportSyncUUID",
37213649 		[P_AUTH_CHALLENGE]      = "AuthChallenge",
37223650 		[P_AUTH_RESPONSE]	= "AuthResponse",
3651
+		[P_STATE_CHG_REQ]       = "StateChgRequest",
37233652 		[P_PING]		= "Ping",
37243653 		[P_PING_ACK]	        = "PingAck",
37253654 		[P_RECV_ACK]	        = "RecvAck",
....@@ -3730,24 +3659,26 @@
37303659 		[P_NEG_DREPLY]	        = "NegDReply",
37313660 		[P_NEG_RS_DREPLY]	= "NegRSDReply",
37323661 		[P_BARRIER_ACK]	        = "BarrierAck",
3733
-		[P_STATE_CHG_REQ]       = "StateChgRequest",
37343662 		[P_STATE_CHG_REPLY]     = "StateChgReply",
37353663 		[P_OV_REQUEST]          = "OVRequest",
37363664 		[P_OV_REPLY]            = "OVReply",
37373665 		[P_OV_RESULT]           = "OVResult",
37383666 		[P_CSUM_RS_REQUEST]     = "CsumRSRequest",
37393667 		[P_RS_IS_IN_SYNC]	= "CsumRSIsInSync",
3668
+		[P_SYNC_PARAM89]	= "SyncParam89",
37403669 		[P_COMPRESSED_BITMAP]   = "CBitmap",
37413670 		[P_DELAY_PROBE]         = "DelayProbe",
37423671 		[P_OUT_OF_SYNC]		= "OutOfSync",
3743
-		[P_RETRY_WRITE]		= "RetryWrite",
37443672 		[P_RS_CANCEL]		= "RSCancel",
37453673 		[P_CONN_ST_CHG_REQ]	= "conn_st_chg_req",
37463674 		[P_CONN_ST_CHG_REPLY]	= "conn_st_chg_reply",
37473675 		[P_RETRY_WRITE]		= "retry_write",
37483676 		[P_PROTOCOL_UPDATE]	= "protocol_update",
3677
+		[P_TRIM]	        = "Trim",
37493678 		[P_RS_THIN_REQ]         = "rs_thin_req",
37503679 		[P_RS_DEALLOCATED]      = "rs_deallocated",
3680
+		[P_WSAME]	        = "WriteSame",
3681
+		[P_ZEROES]		= "Zeroes",
37513682 
37523683 		/* enum drbd_packet, but not commands - obsoleted flags:
37533684 		 *	P_MAY_IGNORE