1 /* SPDX-License-Identifier: GPL-2.0-only */
7   Copyright (C) 2001-2008, LINBIT Information Technologies GmbH.
8   Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>.
9   Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>.
29 #include <linux/backing-dev.h>
63 #define ID_SYNCER (-1ULL)
125 	/* statistics; index: (h->command == P_BITMAP) */
143 	c->word_offset = c->bit_offset >> 6;  in bm_xfer_ctx_bit_to_word_offset()
145 	c->word_offset = c->bit_offset >> 5;  in bm_xfer_ctx_bit_to_word_offset()
146 	c->word_offset &= ~(1UL);  in bm_xfer_ctx_bit_to_word_offset()
178 	 *	--lge */  in get_t_state()
181 	return thi->t_state;  in get_t_state()
340 	((peer_req)->opf & REQ_OP_MASK)
346  * non-atomic modification to ee->flags is ok.
354 	/* explicit zero-out requested, or
417 	USE_DEGR_WFC_T,		/* degr-wfc-timeout instead of wfc-timeout. */
432 	FORCE_DETACH,		/* Force-detach from local disk, aborting any pending local IO */
438 	AHEAD_TO_SYNC_SOURCE,   /* Ahead -> SyncSource queued */
443 	FLUSH_PENDING,		/* if set, device->flush_jif is when we submitted that flush
447 	GOING_DISKLESS,		/* Disk is being detached, because of io-error, or admin request. */
472 	 * and still allow all non-bulk operations */
480 	 * requires sending of "out-of-sync" information, though. */
516 	/* cached value of bdev->disk_conf->meta_dev_idx (see below) */
531 	struct disk_conf *disk_conf; /* RCU, for updates: resource->conf_update */
597 	struct mutex conf_update;	/* mutex for ready-copy-update of net_conf and disk_conf */
652 …pto_shash *integrity_tfm;  /* checksums we compute, updates protected by connection->data->mutex */
676 	 * protected by resource->req_lock */
701 		 * with req->epoch == current_epoch_nr.
712 	has_net_conf = rcu_dereference(connection->net_conf);  in has_net_conf()
725 	__update_timing_details(c->w_timing_details, &c->w_cb_nr, cb, __func__ , __LINE__ )
727 	__update_timing_details(c->r_timing_details, &c->r_cb_nr, cb, __func__ , __LINE__ )
733 	/* protected by ..->resource->req_lock */
829 	/* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
844 	/* size of out-of-sync range in sectors. */
865 	struct list_head net_ee;    /* zero-copy network send in progress */
886 …struct mutex *state_mutex; /* either own_state_mutex or first_peer_device(device)->connection->cst…
894 …struct fifo_buffer *rs_plan_s; /* correction values of resync planer (RCU, connection->conn_update…
901 	 * are deferred to this single-threaded work queue */
907 	struct list_head list; /* on device->pending_bitmap_io */;
925 #define VOLUME_UNSPECIFIED		(-1U)
949 	return list_first_entry_or_null(&device->peer_devices, struct drbd_peer_device, peer_devices);  in first_peer_device()
955 	return idr_find(&connection->peer_devices, volume_number);  in conn_peer_device()
968 	list_for_each_entry(connection, &resource->connections, connections)
971 	list_for_each_entry_rcu(connection, &resource->connections, connections)
974 	list_for_each_entry_safe(connection, tmp, &resource->connections, connections)
977 	list_for_each_entry(peer_device, &device->peer_devices, peer_devices)
980 	list_for_each_entry_rcu(peer_device, &device->peer_devices, peer_devices)
983 	list_for_each_entry_safe(peer_device, tmp, &device->peer_devices, peer_devices)
987 	return device->minor;  in device_to_minor()
1094  *   |----------- md_size_sect ------------------|
1098  *  ==> bitmap sectors = md_size_sect - bm_offset
1104  *            |----------- md_size_sect ------------------|
1107  *            | bm_offset = al_offset - Y |
1108  *  ==> bitmap sectors = Y = al_offset - bm_offset
1114  *  which are written in a ring-buffer, or striped ring-buffer like fashion,
1116  *  but is about to become configurable.
1130 /* We could make these currently hardcoded constants configurable
1131  * variables at create-md time (or even re-configurable at runtime?).
1145 #define AL_CONTEXT_PER_TRANSACTION	919	// (4096 - 36 - 6*64)/4
1197 #define BM_SECT_TO_BIT(x)   ((x)>>(BM_BLOCK_SHIFT-9))
1198 #define BM_BIT_TO_SECT(x)   ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1202 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1206 #define BM_SECT_TO_EXT(x)   ((x)>>(BM_EXT_SHIFT-9))
1207 #define BM_BIT_TO_EXT(x)    ((x) >> (BM_EXT_SHIFT - BM_BLOCK_SHIFT))
1210 #define BM_EXT_TO_SECT(x)   ((sector_t)(x) << (BM_EXT_SHIFT-9))
1214 #define BM_BITS_PER_EXT     (1UL << (BM_EXT_SHIFT - BM_BLOCK_SHIFT))
1216 #define BM_BLOCKS_PER_BM_EXT_MASK  (BM_BITS_PER_EXT - 1)
1220 #define AL_EXT_PER_BM_SECT  (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1227  * bit	 0	  bit 37   bit 38	     bit (512*8)-1
1229  * sect. 0	 `296	  `304			   ^(512*8*8)-1
1237 /* we have a certain meta data variant that has a fixed on-disk size of 128
1243 	  ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9)))
1259  * Since we may live in a mixed-platform cluster,
1355 /* We also need a standard (emergency-reserve backed) page pool
1406 	DS_ERROR_SHRINK = -3,
1407 	DS_ERROR_SPACE_MD = -2,
1408 	DS_ERROR = -1,
1452 	struct drbd_device *device = peer_device->device;  in ov_out_of_sync_print()
1454 	if (device->ov_last_oos_size) {  in ov_out_of_sync_print()
1456 		     (unsigned long long)device->ov_last_oos_start,  in ov_out_of_sync_print()
1457 		     (unsigned long)device->ov_last_oos_size);  in ov_out_of_sync_print()
1459 	device->ov_last_oos_size = 0;  in ov_out_of_sync_print()
1521 	if (!bio->bi_bdev) {  in drbd_submit_bio_noacct()
1522 		drbd_err(device, "drbd_submit_bio_noacct: bio->bi_bdev == NULL\n");  in drbd_submit_bio_noacct()
1523 		bio->bi_status = BLK_STS_IOERR;  in drbd_submit_bio_noacct()
1627 	struct page *page = peer_req->pages;  in drbd_peer_req_has_active_page()
1637 	struct drbd_resource *resource = device->resource;  in drbd_read_state()
1640 	rv.i = device->state.i;  in drbd_read_state()
1641 	rv.susp = resource->susp;  in drbd_read_state()
1642 	rv.susp_nod = resource->susp_nod;  in drbd_read_state()
1643 	rv.susp_fen = resource->susp_fen;  in drbd_read_state()
1663 	ep = rcu_dereference(device->ldev->disk_conf)->on_io_error;  in __drbd_chk_io_error_()
1670 			if (device->state.disk > D_INCONSISTENT)  in __drbd_chk_io_error_()
1683 		 * blocks, which triggers block re-allocation in lower layers.  in __drbd_chk_io_error_()
1688 		 * Force-detach is not really an IO error, but rather a  in __drbd_chk_io_error_()
1697 		set_bit(WAS_IO_ERROR, &device->flags);  in __drbd_chk_io_error_()
1699 			set_bit(WAS_READ_ERROR, &device->flags);  in __drbd_chk_io_error_()
1701 			set_bit(FORCE_DETACH, &device->flags);  in __drbd_chk_io_error_()
1702 		if (device->state.disk > D_FAILED) {  in __drbd_chk_io_error_()
1725 		spin_lock_irqsave(&device->resource->req_lock, flags);  in drbd_chk_io_error_()
1727 		spin_unlock_irqrestore(&device->resource->req_lock, flags);  in drbd_chk_io_error_()
1733  * drbd_md_first_sector() - Returns the first sector number of the meta data area
1741 	switch (bdev->md.meta_dev_idx) {  in drbd_md_first_sector()
1744 		return bdev->md.md_offset + bdev->md.bm_offset;  in drbd_md_first_sector()
1747 		return bdev->md.md_offset;  in drbd_md_first_sector()
1752  * drbd_md_last_sector() - Return the last sector number of the meta data area
1757 	switch (bdev->md.meta_dev_idx) {  in drbd_md_last_sector()
1760 		return bdev->md.md_offset + MD_4kB_SECT -1;  in drbd_md_last_sector()
1763 		return bdev->md.md_offset + bdev->md.md_size_sect -1;  in drbd_md_last_sector()
1774  * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1785 	switch (bdev->md.meta_dev_idx) {  in drbd_get_max_capacity()
1788 		s = drbd_get_capacity(bdev->backing_bdev)  in drbd_get_max_capacity()
1795 				drbd_get_capacity(bdev->backing_bdev));  in drbd_get_max_capacity()
1798 			BM_EXT_TO_SECT(bdev->md.md_size_sect  in drbd_get_max_capacity()
1799 				     - bdev->md.bm_offset));  in drbd_get_max_capacity()
1803 				drbd_get_capacity(bdev->backing_bdev));  in drbd_get_max_capacity()
1809  * drbd_md_ss() - Return the sector number of our meta data super block
1814 	const int meta_dev_idx = bdev->md.meta_dev_idx;  in drbd_md_ss()
1823 		return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL) - 8;  in drbd_md_ss()
1826 	return MD_128MB_SECT * bdev->md.meta_dev_idx;  in drbd_md_ss()
1833 	spin_lock_irqsave(&q->q_lock, flags);  in drbd_queue_work()
1834 	list_add_tail(&w->list, &q->q);  in drbd_queue_work()
1835 	spin_unlock_irqrestore(&q->q_lock, flags);  in drbd_queue_work()
1836 	wake_up(&q->q_wait);  in drbd_queue_work()
1843 	spin_lock_irqsave(&q->q_lock, flags);  in drbd_queue_work_if_unqueued()
1844 	if (list_empty_careful(&w->list))  in drbd_queue_work_if_unqueued()
1845 		list_add_tail(&w->list, &q->q);  in drbd_queue_work_if_unqueued()
1846 	spin_unlock_irqrestore(&q->q_lock, flags);  in drbd_queue_work_if_unqueued()
1847 	wake_up(&q->q_wait);  in drbd_queue_work_if_unqueued()
1853 	if (!test_and_set_bit(work_bit, &device->flags)) {  in drbd_device_post_work()
1855 			first_peer_device(device)->connection;  in drbd_device_post_work()
1856 		struct drbd_work_queue *q = &connection->sender_work;  in drbd_device_post_work()
1857 		if (!test_and_set_bit(DEVICE_WORK_PENDING, &connection->flags))  in drbd_device_post_work()
1858 			wake_up(&q->q_wait);  in drbd_device_post_work()
1865  * so it can change its sk_rcvtimeo from idle- to ping-timeout,
1870 	struct task_struct *task = connection->ack_receiver.task;  in wake_ack_receiver()
1871 	if (task && get_t_state(&connection->ack_receiver) == RUNNING)  in wake_ack_receiver()
1877 	set_bit(SEND_PING, &connection->flags);  in request_ping()
1918  *    (drbd_make_request_common; recovery path on read io-error)
1934 	atomic_inc(&device->ap_pending_cnt);  in inc_ap_pending()
1940 	int ap_pending_cnt = atomic_dec_return(&device->ap_pending_cnt);  in __dec_ap_pending()
1943 		wake_up(&device->misc_wait);  in __dec_ap_pending()
1947 /* counts how many resync-related answers we still expect from the peer
1955 	atomic_inc(&peer_device->device->rs_pending_cnt);  in inc_rs_pending()
1962 	return atomic_dec_return(&peer_device->device->rs_pending_cnt);  in __dec_rs_pending()
1976 	atomic_inc(&device->unacked_cnt);  in inc_unacked()
1982 	return atomic_dec_return(&device->unacked_cnt);  in __dec_unacked()
1988 	return atomic_sub_return(n, &device->unacked_cnt);  in __sub_unacked()
2010  * get_ldev() - Increase the ref count on device->ldev. Returns 0 if there is no ldev
2014  * You have to call put_ldev() when finished working with device->ldev.
2023 	enum drbd_disk_state disk_state = device->state.disk;  in put_ldev()
2028 	int i = atomic_dec_return(&device->local_cnt);  in put_ldev()
2041 			if (!test_and_set_bit(GOING_DISKLESS, &device->flags))  in put_ldev()
2043 		wake_up(&device->misc_wait);  in put_ldev()
2053 	if (device->state.disk == D_DISKLESS)  in _get_ldev_if_state()
2056 	atomic_inc(&device->local_cnt);  in _get_ldev_if_state()
2057 	io_allowed = (device->state.disk >= mins);  in _get_ldev_if_state()
2066 /* this throttles on-the-fly application requests
2068  * maybe re-implement using semaphores? */
2075 	nc = rcu_dereference(first_peer_device(device)->connection->net_conf);  in drbd_get_max_buffers()
2076 	mxb = nc ? nc->max_buffers : 1000000;  /* arbitrary limit on open requests */  in drbd_get_max_buffers()
2084 	union drbd_dev_state s = device->state;  in drbd_state_is_stable()
2118 		if (first_peer_device(device)->connection->agreed_pro_version < 96)  in drbd_state_is_stable()
2154 	struct drbd_resource *resource = device->resource;  in drbd_suspended()
2156 	return resource->susp || resource->susp_fen || resource->susp_nod;  in drbd_suspended()
2165 	if (atomic_read(&device->suspend_cnt))  in may_inc_ap_bio()
2178 	if (atomic_read(&device->ap_bio_cnt) > mxb)  in may_inc_ap_bio()
2180 	if (test_bit(BITMAP_IO, &device->flags))  in may_inc_ap_bio()
2189 	spin_lock_irq(&device->resource->req_lock);  in inc_ap_bio_cond()
2192 		atomic_inc(&device->ap_bio_cnt);  in inc_ap_bio_cond()
2193 	spin_unlock_irq(&device->resource->req_lock);  in inc_ap_bio_cond()
2208 	wait_event(device->misc_wait, inc_ap_bio_cond(device));  in inc_ap_bio()
2214 	int ap_bio = atomic_dec_return(&device->ap_bio_cnt);  in dec_ap_bio()
2218 	if (ap_bio == 0 && test_bit(BITMAP_IO, &device->flags)) {  in dec_ap_bio()
2219 		if (!test_and_set_bit(BITMAP_IO_QUEUED, &device->flags))  in dec_ap_bio()
2220 			drbd_queue_work(&first_peer_device(device)->  in dec_ap_bio()
2221 				connection->sender_work,  in dec_ap_bio()
2222 				&device->bm_io_work.w);  in dec_ap_bio()
2229 		wake_up(&device->misc_wait);  in dec_ap_bio()
2234 	return first_peer_device(device)->connection->agreed_pro_version >= 97 &&  in verify_can_do_stop_sector()
2235 		first_peer_device(device)->connection->agreed_pro_version != 100;  in verify_can_do_stop_sector()
2240 	int changed = device->ed_uuid != val;  in drbd_set_ed_uuid()
2241 	device->ed_uuid = val;  in drbd_set_ed_uuid()
2257 	return list_first_entry_or_null(&resource->connections,  in first_connection()