Commit 9104d31a authored by Lars Ellenberg's avatar Lars Ellenberg Committed by Jens Axboe
Browse files

drbd: introduce WRITE_SAME support 



We will support WRITE_SAME, if
 * all peers support WRITE_SAME (both in kernel and DRBD version),
 * all peer devices support WRITE_SAME
 * logical_block_size is identical on all peers.

We may at some point introduce a fallback on the receiving side
for devices/kernels that do not support WRITE_SAME,
by open-coding a submit loop. But not yet.

Signed-off-by: default avatarPhilipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: default avatarLars Ellenberg <lars.ellenberg@linbit.com>
Signed-off-by: default avatarJens Axboe <axboe@fb.com>
parent 60bac040

drivers/block/drbd/drbd_actlog.c

+8 −1
Original line number Diff line number Diff line
@@ -840,6 +840,13 @@ static int update_sync_bits(struct drbd_device *device, 
	return count;

}



static bool plausible_request_size(int size)

{

	return size > 0

		&& size <= DRBD_MAX_BATCH_BIO_SIZE

		&& IS_ALIGNED(size, 512);

}



/* clear the bit corresponding to the piece of storage in question:

 * size byte of data starting from sector.  Only clear a bits of the affected

 * one ore more _aligned_ BM_BLOCK_SIZE blocks.
@@ -859,7 +866,7 @@ int __drbd_change_sync(struct drbd_device *device, sector_t sector, int size, 
	if ((mode == SET_OUT_OF_SYNC) && size == 0)

		return 0;



	if (size <= 0 || !IS_ALIGNED(size, 512) || size > DRBD_MAX_DISCARD_SIZE) {

	if (!plausible_request_size(size)) {

		drbd_err(device, "%s: sector=%llus size=%d nonsense!\n",

				drbd_change_sync_fname[mode],

				(unsigned long long)sector, size);

drivers/block/drbd/drbd_debugfs.c

+3 −8
Original line number Diff line number Diff line
@@ -237,14 +237,9 @@ static void seq_print_peer_request_flags(struct seq_file *m, struct drbd_peer_re 
	seq_print_rq_state_bit(m, f & EE_SEND_WRITE_ACK, &sep, "C");

	seq_print_rq_state_bit(m, f & EE_MAY_SET_IN_SYNC, &sep, "set-in-sync");



	if (f & EE_IS_TRIM) {

		seq_putc(m, sep);

		sep = '|';

		if (f & EE_IS_TRIM_USE_ZEROOUT)

			seq_puts(m, "zero-out");

		else

			seq_puts(m, "trim");

	}

	if (f & EE_IS_TRIM)

		__seq_print_rq_state_bit(m, f & EE_IS_TRIM_USE_ZEROOUT, &sep, "zero-out", "trim");

	seq_print_rq_state_bit(m, f & EE_WRITE_SAME, &sep, "write-same");

	seq_putc(m, '\n');

}

drivers/block/drbd/drbd_int.h

+9 −4
Original line number Diff line number Diff line
@@ -468,6 +468,9 @@ enum { 
	/* this is/was a write request */

	__EE_WRITE,



	/* this is/was a write same request */

	__EE_WRITE_SAME,



	/* this originates from application on peer

	 * (not some resync or verify or other DRBD internal request) */

	__EE_APPLICATION,
@@ -487,6 +490,7 @@ enum { 
#define EE_IN_INTERVAL_TREE	(1<<__EE_IN_INTERVAL_TREE)

#define EE_SUBMITTED		(1<<__EE_SUBMITTED)

#define EE_WRITE		(1<<__EE_WRITE)

#define EE_WRITE_SAME		(1<<__EE_WRITE_SAME)

#define EE_APPLICATION		(1<<__EE_APPLICATION)

#define EE_RS_THIN_REQ		(1<<__EE_RS_THIN_REQ)
@@ -1350,8 +1354,8 @@ struct bm_extent { 
/* For now, don't allow more than half of what we can "activate" in one

 * activity log transaction to be discarded in one go. We may need to rework

 * drbd_al_begin_io() to allow for even larger discard ranges */

#define DRBD_MAX_DISCARD_SIZE	(AL_UPDATES_PER_TRANSACTION/2*AL_EXTENT_SIZE)

#define DRBD_MAX_DISCARD_SECTORS (DRBD_MAX_DISCARD_SIZE >> 9)

#define DRBD_MAX_BATCH_BIO_SIZE	 (AL_UPDATES_PER_TRANSACTION/2*AL_EXTENT_SIZE)

#define DRBD_MAX_BBIO_SECTORS    (DRBD_MAX_BATCH_BIO_SIZE >> 9)



extern int  drbd_bm_init(struct drbd_device *device);

extern int  drbd_bm_resize(struct drbd_device *device, sector_t sectors, int set_new_bits);
@@ -1488,7 +1492,8 @@ enum determine_dev_size { 
extern enum determine_dev_size

drbd_determine_dev_size(struct drbd_device *, enum dds_flags, struct resize_parms *) __must_hold(local);

extern void resync_after_online_grow(struct drbd_device *);

extern void drbd_reconsider_queue_parameters(struct drbd_device *device, struct drbd_backing_dev *bdev);

extern void drbd_reconsider_queue_parameters(struct drbd_device *device,

			struct drbd_backing_dev *bdev, struct o_qlim *o);

extern enum drbd_state_rv drbd_set_role(struct drbd_device *device,

					enum drbd_role new_role,

					int force);
@@ -1569,7 +1574,7 @@ extern int drbd_submit_peer_request(struct drbd_device *, 
extern int drbd_free_peer_reqs(struct drbd_device *, struct list_head *);

extern struct drbd_peer_request *drbd_alloc_peer_req(struct drbd_peer_device *, u64,

						     sector_t, unsigned int,

						     bool,

						     unsigned int,

						     gfp_t) __must_hold(local);

extern void __drbd_free_peer_req(struct drbd_device *, struct drbd_peer_request *,

				 int);

drivers/block/drbd/drbd_main.c

+72 −10
Original line number Diff line number Diff line
@@ -920,6 +920,31 @@ void drbd_gen_and_send_sync_uuid(struct drbd_peer_device *peer_device) 
	}

}



/* communicated if (agreed_features & DRBD_FF_WSAME) */

void assign_p_sizes_qlim(struct drbd_device *device, struct p_sizes *p, struct request_queue *q)

{

	if (q) {

		p->qlim->physical_block_size = cpu_to_be32(queue_physical_block_size(q));

		p->qlim->logical_block_size = cpu_to_be32(queue_logical_block_size(q));

		p->qlim->alignment_offset = cpu_to_be32(queue_alignment_offset(q));

		p->qlim->io_min = cpu_to_be32(queue_io_min(q));

		p->qlim->io_opt = cpu_to_be32(queue_io_opt(q));

		p->qlim->discard_enabled = blk_queue_discard(q);

		p->qlim->discard_zeroes_data = queue_discard_zeroes_data(q);

		p->qlim->write_same_capable = !!q->limits.max_write_same_sectors;

	} else {

		q = device->rq_queue;

		p->qlim->physical_block_size = cpu_to_be32(queue_physical_block_size(q));

		p->qlim->logical_block_size = cpu_to_be32(queue_logical_block_size(q));

		p->qlim->alignment_offset = 0;

		p->qlim->io_min = cpu_to_be32(queue_io_min(q));

		p->qlim->io_opt = cpu_to_be32(queue_io_opt(q));

		p->qlim->discard_enabled = 0;

		p->qlim->discard_zeroes_data = 0;

		p->qlim->write_same_capable = 0;

	}

}



int drbd_send_sizes(struct drbd_peer_device *peer_device, int trigger_reply, enum dds_flags flags)

{

	struct drbd_device *device = peer_device->device;
@@ -928,29 +953,37 @@ int drbd_send_sizes(struct drbd_peer_device *peer_device, int trigger_reply, enu 
	sector_t d_size, u_size;

	int q_order_type;

	unsigned int max_bio_size;

	unsigned int packet_size;



	sock = &peer_device->connection->data;

	p = drbd_prepare_command(peer_device, sock);

	if (!p)

		return -EIO;



	packet_size = sizeof(*p);

	if (peer_device->connection->agreed_features & DRBD_FF_WSAME)

		packet_size += sizeof(p->qlim[0]);



	memset(p, 0, packet_size);

	if (get_ldev_if_state(device, D_NEGOTIATING)) {

		D_ASSERT(device, device->ldev->backing_bdev);

		struct request_queue *q = bdev_get_queue(device->ldev->backing_bdev);

		d_size = drbd_get_max_capacity(device->ldev);

		rcu_read_lock();

		u_size = rcu_dereference(device->ldev->disk_conf)->disk_size;

		rcu_read_unlock();

		q_order_type = drbd_queue_order_type(device);

		max_bio_size = queue_max_hw_sectors(device->ldev->backing_bdev->bd_disk->queue) << 9;

		max_bio_size = queue_max_hw_sectors(q) << 9;

		max_bio_size = min(max_bio_size, DRBD_MAX_BIO_SIZE);

		assign_p_sizes_qlim(device, p, q);

		put_ldev(device);

	} else {

		d_size = 0;

		u_size = 0;

		q_order_type = QUEUE_ORDERED_NONE;

		max_bio_size = DRBD_MAX_BIO_SIZE; /* ... multiple BIOs per peer_request */

		assign_p_sizes_qlim(device, p, NULL);

	}



	sock = &peer_device->connection->data;

	p = drbd_prepare_command(peer_device, sock);

	if (!p)

		return -EIO;



	if (peer_device->connection->agreed_pro_version <= 94)

		max_bio_size = min(max_bio_size, DRBD_MAX_SIZE_H80_PACKET);

	else if (peer_device->connection->agreed_pro_version < 100)
@@ -962,7 +995,8 @@ int drbd_send_sizes(struct drbd_peer_device *peer_device, int trigger_reply, enu 
	p->max_bio_size = cpu_to_be32(max_bio_size);

	p->queue_order_type = cpu_to_be16(q_order_type);

	p->dds_flags = cpu_to_be16(flags);

	return drbd_send_command(peer_device, sock, P_SIZES, sizeof(*p), NULL, 0);



	return drbd_send_command(peer_device, sock, P_SIZES, packet_size, NULL, 0);

}



/**
@@ -1577,6 +1611,9 @@ static int _drbd_send_bio(struct drbd_peer_device *peer_device, struct bio *bio) 
					 ? 0 : MSG_MORE);

		if (err)

			return err;

		/* REQ_OP_WRITE_SAME has only one segment */

		if (bio_op(bio) == REQ_OP_WRITE_SAME)

			break;

	}

	return 0;

}
@@ -1595,6 +1632,9 @@ static int _drbd_send_zc_bio(struct drbd_peer_device *peer_device, struct bio *b 
				      bio_iter_last(bvec, iter) ? 0 : MSG_MORE);

		if (err)

			return err;

		/* REQ_OP_WRITE_SAME has only one segment */

		if (bio_op(bio) == REQ_OP_WRITE_SAME)

			break;

	}

	return 0;

}
@@ -1626,6 +1666,7 @@ static u32 bio_flags_to_wire(struct drbd_connection *connection, 
		return  (bio->bi_rw & REQ_SYNC ? DP_RW_SYNC : 0) |

			(bio->bi_rw & REQ_FUA ? DP_FUA : 0) |

			(bio->bi_rw & REQ_PREFLUSH ? DP_FLUSH : 0) |

			(bio_op(bio) == REQ_OP_WRITE_SAME ? DP_WSAME : 0) |

			(bio_op(bio) == REQ_OP_DISCARD ? DP_DISCARD : 0);

	else

		return bio->bi_rw & REQ_SYNC ? DP_RW_SYNC : 0;
@@ -1639,6 +1680,8 @@ int drbd_send_dblock(struct drbd_peer_device *peer_device, struct drbd_request * 
	struct drbd_device *device = peer_device->device;

	struct drbd_socket *sock;

	struct p_data *p;

	struct p_wsame *wsame = NULL;

	void *digest_out;

	unsigned int dp_flags = 0;

	int digest_size;

	int err;
@@ -1674,12 +1717,29 @@ int drbd_send_dblock(struct drbd_peer_device *peer_device, struct drbd_request * 
		err = __send_command(peer_device->connection, device->vnr, sock, P_TRIM, sizeof(*t), NULL, 0);

		goto out;

	}

	if (dp_flags & DP_WSAME) {

		/* this will only work if DRBD_FF_WSAME is set AND the

		 * handshake agreed that all nodes and backend devices are

		 * WRITE_SAME capable and agree on logical_block_size */

		wsame = (struct p_wsame*)p;

		digest_out = wsame + 1;

		wsame->size = cpu_to_be32(req->i.size);

	} else

		digest_out = p + 1;



	/* our digest is still only over the payload.

	 * TRIM does not carry any payload. */

	if (digest_size)

		drbd_csum_bio(peer_device->connection->integrity_tfm, req->master_bio, p + 1);

	err = __send_command(peer_device->connection, device->vnr, sock, P_DATA, sizeof(*p) + digest_size, NULL, req->i.size);

		drbd_csum_bio(peer_device->connection->integrity_tfm, req->master_bio, digest_out);

	if (wsame) {

		err =

		    __send_command(peer_device->connection, device->vnr, sock, P_WSAME,

				   sizeof(*wsame) + digest_size, NULL,

				   bio_iovec(req->master_bio).bv_len);

	} else

		err =

		    __send_command(peer_device->connection, device->vnr, sock, P_DATA,

				   sizeof(*p) + digest_size, NULL, req->i.size);

	if (!err) {

		/* For protocol A, we have to memcpy the payload into

		 * socket buffers, as we may complete right away
@@ -3660,6 +3720,8 @@ const char *cmdname(enum drbd_packet cmd) 
	 * one PRO_VERSION */

	static const char *cmdnames[] = {

		[P_DATA]	        = "Data",

		[P_WSAME]	        = "WriteSame",

		[P_TRIM]	        = "Trim",

		[P_DATA_REPLY]	        = "DataReply",

		[P_RS_DATA_REPLY]	= "RSDataReply",

		[P_BARRIER]	        = "Barrier",

drivers/block/drbd/drbd_nl.c

+79 −9
Original line number Diff line number Diff line
@@ -1174,6 +1174,17 @@ static void blk_queue_discard_granularity(struct request_queue *q, unsigned int 
{

	q->limits.discard_granularity = granularity;

}



static unsigned int drbd_max_discard_sectors(struct drbd_connection *connection)

{

	/* when we introduced REQ_WRITE_SAME support, we also bumped

	 * our maximum supported batch bio size used for discards. */

	if (connection->agreed_features & DRBD_FF_WSAME)

		return DRBD_MAX_BBIO_SECTORS;

	/* before, with DRBD <= 8.4.6, we only allowed up to one AL_EXTENT_SIZE. */

	return AL_EXTENT_SIZE >> 9;

}



static void decide_on_discard_support(struct drbd_device *device,

			struct request_queue *q,

			struct request_queue *b,
@@ -1190,7 +1201,7 @@ static void decide_on_discard_support(struct drbd_device *device, 
		can_do = false;

		drbd_info(device, "discard_zeroes_data=0 and discard_zeroes_if_aligned=no: disabling discards\n");

	}

	if (can_do && connection->cstate >= C_CONNECTED && !(connection->agreed_features & FF_TRIM)) {

	if (can_do && connection->cstate >= C_CONNECTED && !(connection->agreed_features & DRBD_FF_TRIM)) {

		can_do = false;

		drbd_info(connection, "peer DRBD too old, does not support TRIM: disabling discards\n");

	}
@@ -1202,7 +1213,7 @@ static void decide_on_discard_support(struct drbd_device *device, 
		 * you care, you need to use devices with similar

		 * topology on all peers. */

		blk_queue_discard_granularity(q, 512);

		q->limits.max_discard_sectors = DRBD_MAX_DISCARD_SECTORS;

		q->limits.max_discard_sectors = drbd_max_discard_sectors(connection);

		queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, q);

	} else {

		queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, q);
@@ -1223,8 +1234,67 @@ static void fixup_discard_if_not_supported(struct request_queue *q) 
	}

}



static void decide_on_write_same_support(struct drbd_device *device,

			struct request_queue *q,

			struct request_queue *b, struct o_qlim *o)

{

	struct drbd_peer_device *peer_device = first_peer_device(device);

	struct drbd_connection *connection = peer_device->connection;

	bool can_do = b ? b->limits.max_write_same_sectors : true;



	if (can_do && connection->cstate >= C_CONNECTED && !(connection->agreed_features & DRBD_FF_WSAME)) {

		can_do = false;

		drbd_info(peer_device, "peer does not support WRITE_SAME\n");

	}



	if (o) {

		/* logical block size; queue_logical_block_size(NULL) is 512 */

		unsigned int peer_lbs = be32_to_cpu(o->logical_block_size);

		unsigned int me_lbs_b = queue_logical_block_size(b);

		unsigned int me_lbs = queue_logical_block_size(q);



		if (me_lbs_b != me_lbs) {

			drbd_warn(device,

				"logical block size of local backend does not match (drbd:%u, backend:%u); was this a late attach?\n",

				me_lbs, me_lbs_b);

			/* rather disable write same than trigger some BUG_ON later in the scsi layer. */

			can_do = false;

		}

		if (me_lbs_b != peer_lbs) {

			drbd_warn(peer_device, "logical block sizes do not match (me:%u, peer:%u); this may cause problems.\n",

				me_lbs, peer_lbs);

			if (can_do) {

				drbd_dbg(peer_device, "logical block size mismatch: WRITE_SAME disabled.\n");

				can_do = false;

			}

			me_lbs = max(me_lbs, me_lbs_b);

			/* We cannot change the logical block size of an in-use queue.

			 * We can only hope that access happens to be properly aligned.

			 * If not, the peer will likely produce an IO error, and detach. */

			if (peer_lbs > me_lbs) {

				if (device->state.role != R_PRIMARY) {

					blk_queue_logical_block_size(q, peer_lbs);

					drbd_warn(peer_device, "logical block size set to %u\n", peer_lbs);

				} else {

					drbd_warn(peer_device,

						"current Primary must NOT adjust logical block size (%u -> %u); hope for the best.\n",

						me_lbs, peer_lbs);

				}

			}

		}

		if (can_do && !o->write_same_capable) {

			/* If we introduce an open-coded write-same loop on the receiving side,

			 * the peer would present itself as "capable". */

			drbd_dbg(peer_device, "WRITE_SAME disabled (peer device not capable)\n");

			can_do = false;

		}

	}



	blk_queue_max_write_same_sectors(q, can_do ? DRBD_MAX_BBIO_SECTORS : 0);

}



static void drbd_setup_queue_param(struct drbd_device *device, struct drbd_backing_dev *bdev,

				   unsigned int max_bio_size)

				   unsigned int max_bio_size, struct o_qlim *o)

{

	struct request_queue * const q = device->rq_queue;

	unsigned int max_hw_sectors = max_bio_size >> 9;
@@ -1244,15 +1314,15 @@ static void drbd_setup_queue_param(struct drbd_device *device, struct drbd_backi 
		rcu_read_unlock();



		blk_set_stacking_limits(&q->limits);

		blk_queue_max_write_same_sectors(q, 0);

	}



	blk_queue_logical_block_size(q, 512);

	blk_queue_max_hw_sectors(q, max_hw_sectors);

	/* This is the workaround for "bio would need to, but cannot, be split" */

	blk_queue_max_segments(q, max_segments ? max_segments : BLK_MAX_SEGMENTS);

	blk_queue_segment_boundary(q, PAGE_SIZE-1);

	decide_on_discard_support(device, q, b, discard_zeroes_if_aligned);

	decide_on_write_same_support(device, q, b, o);



	if (b) {

		blk_queue_stack_limits(q, b);
@@ -1266,7 +1336,7 @@ static void drbd_setup_queue_param(struct drbd_device *device, struct drbd_backi 
	fixup_discard_if_not_supported(q);

}



void drbd_reconsider_queue_parameters(struct drbd_device *device, struct drbd_backing_dev *bdev)

void drbd_reconsider_queue_parameters(struct drbd_device *device, struct drbd_backing_dev *bdev, struct o_qlim *o)

{

	unsigned int now, new, local, peer;
@@ -1309,7 +1379,7 @@ void drbd_reconsider_queue_parameters(struct drbd_device *device, struct drbd_ba 
	if (new != now)

		drbd_info(device, "max BIO size = %u\n", new);



	drbd_setup_queue_param(device, bdev, new);

	drbd_setup_queue_param(device, bdev, new, o);

}



/* Starts the worker thread */
@@ -1542,7 +1612,7 @@ int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info) 
		drbd_bump_write_ordering(device->resource, NULL, WO_BDEV_FLUSH);



	if (old_disk_conf->discard_zeroes_if_aligned != new_disk_conf->discard_zeroes_if_aligned)

		drbd_reconsider_queue_parameters(device, device->ldev);

		drbd_reconsider_queue_parameters(device, device->ldev, NULL);



	drbd_md_sync(device);
@@ -1922,7 +1992,7 @@ int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info) 
	device->read_cnt = 0;

	device->writ_cnt = 0;



	drbd_reconsider_queue_parameters(device, device->ldev);

	drbd_reconsider_queue_parameters(device, device->ldev, NULL);



	/* If I am currently not R_PRIMARY,

	 * but meta data primary indicator is set,

CRA Git | Maintained and supported by SUSTech CRA and CCSE