1 /* SPDX-License-Identifier: GPL-2.0-or-later */
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>
72 #define ID_SYNCER (-1ULL)
80 	dev_printk(level, disk_to_dev((device)->vdisk), fmt, ## args)
82 	dev_printk(level, disk_to_dev((peer_device)->device->vdisk), fmt, ## args)
84 	printk(level "drbd %s: " fmt, (resource)->name, ## args)
86 	printk(level "drbd %s: " fmt, (connection)->resource->name, ## args)
124 	dynamic_dev_dbg(disk_to_dev(device->vdisk), fmt, ## args)
132  * expect  -  Make an assertion
198 	/* statistics; index: (h->command == P_BITMAP) */
216 	c->word_offset = c->bit_offset >> 6;  in bm_xfer_ctx_bit_to_word_offset()
218 	c->word_offset = c->bit_offset >> 5;  in bm_xfer_ctx_bit_to_word_offset()
219 	c->word_offset &= ~(1UL);  in bm_xfer_ctx_bit_to_word_offset()
251 	 *	--lge */  in get_t_state()
254 	return thi->t_state;  in get_t_state()
308 	/* Minimal set of time stamps to determine if we wait for activity log
309 	 * transactions, local disk or peer.  32 bit "jiffies" are good enough,
316 	/* local disk */
332 	 *      how long did we wait for activity log transactions
414  * non-atomic modification to ee->flags is ok.
422 	/* explicit zero-out requested, or
484 	MD_DIRTY,		/* current uuids and flags not yet on disk */
485 	USE_DEGR_WFC_T,		/* degr-wfc-timeout instead of wfc-timeout. */
498 	WAS_IO_ERROR,		/* Local disk failed, returned IO error */
499 	WAS_READ_ERROR,		/* Local disk READ failed (set additionally to the above) */
500 	FORCE_DETACH,		/* Force-detach from local disk, aborting any pending local IO */
505 	AL_SUSPENDED,		/* Activity logging is currently suspended. */
506 	AHEAD_TO_SYNC_SOURCE,   /* Ahead -> SyncSource queued */
511 	FLUSH_PENDING,		/* if set, device->flush_jif is when we submitted that flush
515 	GOING_DISKLESS,		/* Disk is being detached, because of io-error, or admin request. */
540 	 * and still allow all non-bulk operations */
548 	 * requires sending of "out-of-sync" information, though. */
581 	s32 al_offset;	/* signed relative sector offset to activity log */
584 	/* cached value of bdev->disk_conf->meta_dev_idx (see below) */
597 	struct disk_conf *disk_conf; /* RCU, for updates: resource->conf_update */
665 	struct mutex conf_update;	/* mutex for ready-copy-update of net_conf and disk_conf */
720 …pto_shash *integrity_tfm;  /* checksums we compute, updates protected by connection->data->mutex */
744 	 * protected by resource->req_lock */
769 		 * with req->epoch == current_epoch_nr.
780 	has_net_conf = rcu_dereference(connection->net_conf);  in has_net_conf()
793 	__update_timing_details(c->w_timing_details, &c->w_cb_nr, cb, __func__ , __LINE__ )
795 	__update_timing_details(c->r_timing_details, &c->r_cb_nr, cb, __func__ , __LINE__ )
801 	/* protected by ..->resource->req_lock */
836 	/* things that are stored as / read from meta data on disk */
842 	sector_t p_size;     /* partner's disk size */
854 	/* Used after attach while negotiating new disk state. */
867 	atomic_t ap_actlog_cnt;  /* Requests waiting for activity log */
897 	/* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
912 	/* size of out-of-sync range in sectors. */
929 	struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */
930 	struct list_head sync_ee;   /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
933 	struct list_head net_ee;    /* zero-copy network send in progress */
943 	struct lru_cache *act_log;	/* activity log */
954 …struct mutex *state_mutex; /* either own_state_mutex or first_peer_device(device)->connection->cst…
962 …struct fifo_buffer *rs_plan_s; /* correction values of resync planer (RCU, connection->conn_update…
969 	 * are deferred to this single-threaded work queue */
975 	struct list_head list; /* on device->pending_bitmap_io */;
993 #define VOLUME_UNSPECIFIED		(-1U)
1017 	return list_first_entry_or_null(&device->peer_devices, struct drbd_peer_device, peer_devices);  in first_peer_device()
1023 	return idr_find(&connection->peer_devices, volume_number);  in conn_peer_device()
1036 	list_for_each_entry(connection, &resource->connections, connections)
1039 	list_for_each_entry_rcu(connection, &resource->connections, connections)
1042 	list_for_each_entry_safe(connection, tmp, &resource->connections, connections)
1045 	list_for_each_entry(peer_device, &device->peer_devices, peer_devices)
1048 	list_for_each_entry_rcu(peer_device, &device->peer_devices, peer_devices)
1051 	list_for_each_entry_safe(peer_device, tmp, &device->peer_devices, peer_devices)
1055 	return device->minor;  in device_to_minor()
1157  *   |----------- md_size_sect ------------------|
1158  *   [ 4k superblock ][ activity log ][  Bitmap  ]
1161  *  ==> bitmap sectors = md_size_sect - bm_offset
1167  *            |----------- md_size_sect ------------------|
1168  * [data.....][  Bitmap  ][ activity log ][ 4k superblock ][padding*]
1170  *            | bm_offset = al_offset - Y |
1171  *  ==> bitmap sectors = Y = al_offset - bm_offset
1176  *  The activity log consists of 4k transaction blocks,
1177  *  which are written in a ring-buffer, or striped ring-buffer like fashion,
1189 /* One activity log extent represents 4M of storage */
1194  * variables at create-md time (or even re-configurable at runtime?).
1208 #define AL_CONTEXT_PER_TRANSACTION	919	// (4096 - 36 - 6*64)/4
1238  * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1260 #define BM_SECT_TO_BIT(x)   ((x)>>(BM_BLOCK_SHIFT-9))
1261 #define BM_BIT_TO_SECT(x)   ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1265 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1269 #define BM_SECT_TO_EXT(x)   ((x)>>(BM_EXT_SHIFT-9))
1270 #define BM_BIT_TO_EXT(x)    ((x) >> (BM_EXT_SHIFT - BM_BLOCK_SHIFT))
1273 #define BM_EXT_TO_SECT(x)   ((sector_t)(x) << (BM_EXT_SHIFT-9))
1277 #define BM_BITS_PER_EXT     (1UL << (BM_EXT_SHIFT - BM_BLOCK_SHIFT))
1279 #define BM_BLOCKS_PER_BM_EXT_MASK  (BM_BITS_PER_EXT - 1)
1283 #define AL_EXT_PER_BM_SECT  (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1285 /* the extent in "PER_EXTENT" below is an activity log extent
1290  * bit	 0	  bit 37   bit 38	     bit (512*8)-1
1292  * sect. 0	 `296	  `304			   ^(512*8*8)-1
1300 /* we have a certain meta data variant that has a fixed on-disk size of 128
1301  * MiB, of which 4k are our "superblock", and 32k are the fixed size activity
1306 	  ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9)))
1322  * Since we may live in a mixed-platform cluster,
1336  * activity log transaction to be discarded in one go. We may need to rework
1392 extern struct kmem_cache *drbd_al_ext_cache;	/* activity log extents */
1414 /* We also need a standard (emergency-reserve backed) page pool
1415  * for meta data IO (activity log, bitmap).
1465 	DS_ERROR_SHRINK = -3,
1466 	DS_ERROR_SPACE_MD = -2,
1467 	DS_ERROR = -1,
1510 	if (device->ov_last_oos_size) {  in ov_out_of_sync_print()
1512 		     (unsigned long long)device->ov_last_oos_start,  in ov_out_of_sync_print()
1513 		     (unsigned long)device->ov_last_oos_size);  in ov_out_of_sync_print()
1515 	device->ov_last_oos_size = 0;  in ov_out_of_sync_print()
1580 	if (!bio->bi_bdev) {  in drbd_submit_bio_noacct()
1581 		drbd_err(device, "drbd_submit_bio_noacct: bio->bi_bdev == NULL\n");  in drbd_submit_bio_noacct()
1582 		bio->bi_status = BLK_STS_IOERR;  in drbd_submit_bio_noacct()
1686 	struct page *page = peer_req->pages;  in drbd_peer_req_has_active_page()
1696 	struct drbd_resource *resource = device->resource;  in drbd_read_state()
1699 	rv.i = device->state.i;  in drbd_read_state()
1700 	rv.susp = resource->susp;  in drbd_read_state()
1701 	rv.susp_nod = resource->susp_nod;  in drbd_read_state()
1702 	rv.susp_fen = resource->susp_fen;  in drbd_read_state()
1722 	ep = rcu_dereference(device->ldev->disk_conf)->on_io_error;  in __drbd_chk_io_error_()
1729 			if (device->state.disk > D_INCONSISTENT)  in __drbd_chk_io_error_()
1730 				_drbd_set_state(_NS(device, disk, D_INCONSISTENT), CS_HARD, NULL);  in __drbd_chk_io_error_()
1739 		 * by the activity log.  in __drbd_chk_io_error_()
1742 		 * blocks, which triggers block re-allocation in lower layers.  in __drbd_chk_io_error_()
1747 		 * Force-detach is not really an IO error, but rather a  in __drbd_chk_io_error_()
1756 		set_bit(WAS_IO_ERROR, &device->flags);  in __drbd_chk_io_error_()
1758 			set_bit(WAS_READ_ERROR, &device->flags);  in __drbd_chk_io_error_()
1760 			set_bit(FORCE_DETACH, &device->flags);  in __drbd_chk_io_error_()
1761 		if (device->state.disk > D_FAILED) {  in __drbd_chk_io_error_()
1762 			_drbd_set_state(_NS(device, disk, D_FAILED), CS_HARD, NULL);  in __drbd_chk_io_error_()
1776  * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1784 		spin_lock_irqsave(&device->resource->req_lock, flags);  in drbd_chk_io_error_()
1786 		spin_unlock_irqrestore(&device->resource->req_lock, flags);  in drbd_chk_io_error_()
1792  * drbd_md_first_sector() - Returns the first sector number of the meta data area
1800 	switch (bdev->md.meta_dev_idx) {  in drbd_md_first_sector()
1803 		return bdev->md.md_offset + bdev->md.bm_offset;  in drbd_md_first_sector()
1806 		return bdev->md.md_offset;  in drbd_md_first_sector()
1811  * drbd_md_last_sector() - Return the last sector number of the meta data area
1816 	switch (bdev->md.meta_dev_idx) {  in drbd_md_last_sector()
1819 		return bdev->md.md_offset + MD_4kB_SECT -1;  in drbd_md_last_sector()
1822 		return bdev->md.md_offset + bdev->md.md_size_sect -1;  in drbd_md_last_sector()
1829 	/* return bdev ? get_capacity(bdev->bd_disk) : 0; */  in drbd_get_capacity()
1830 	return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0;  in drbd_get_capacity()
1834  * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1845 	switch (bdev->md.meta_dev_idx) {  in drbd_get_max_capacity()
1848 		s = drbd_get_capacity(bdev->backing_bdev)  in drbd_get_max_capacity()
1855 				drbd_get_capacity(bdev->backing_bdev));  in drbd_get_max_capacity()
1858 			BM_EXT_TO_SECT(bdev->md.md_size_sect  in drbd_get_max_capacity()
1859 				     - bdev->md.bm_offset));  in drbd_get_max_capacity()
1863 				drbd_get_capacity(bdev->backing_bdev));  in drbd_get_max_capacity()
1869  * drbd_md_ss() - Return the sector number of our meta data super block
1874 	const int meta_dev_idx = bdev->md.meta_dev_idx;  in drbd_md_ss()
1883 		return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL) - 8;  in drbd_md_ss()
1886 	return MD_128MB_SECT * bdev->md.meta_dev_idx;  in drbd_md_ss()
1893 	spin_lock_irqsave(&q->q_lock, flags);  in drbd_queue_work()
1894 	list_add_tail(&w->list, &q->q);  in drbd_queue_work()
1895 	spin_unlock_irqrestore(&q->q_lock, flags);  in drbd_queue_work()
1896 	wake_up(&q->q_wait);  in drbd_queue_work()
1903 	spin_lock_irqsave(&q->q_lock, flags);  in drbd_queue_work_if_unqueued()
1904 	if (list_empty_careful(&w->list))  in drbd_queue_work_if_unqueued()
1905 		list_add_tail(&w->list, &q->q);  in drbd_queue_work_if_unqueued()
1906 	spin_unlock_irqrestore(&q->q_lock, flags);  in drbd_queue_work_if_unqueued()
1907 	wake_up(&q->q_wait);  in drbd_queue_work_if_unqueued()
1913 	if (!test_and_set_bit(work_bit, &device->flags)) {  in drbd_device_post_work()
1915 			first_peer_device(device)->connection;  in drbd_device_post_work()
1916 		struct drbd_work_queue *q = &connection->sender_work;  in drbd_device_post_work()
1917 		if (!test_and_set_bit(DEVICE_WORK_PENDING, &connection->flags))  in drbd_device_post_work()
1918 			wake_up(&q->q_wait);  in drbd_device_post_work()
1925  * so it can change its sk_rcvtimeo from idle- to ping-timeout,
1930 	struct task_struct *task = connection->ack_receiver.task;  in wake_ack_receiver()
1931 	if (task && get_t_state(&connection->ack_receiver) == RUNNING)  in wake_ack_receiver()
1937 	set_bit(SEND_PING, &connection->flags);  in request_ping()
1978  *    (drbd_make_request_common; recovery path on read io-error)
1994 	atomic_inc(&device->ap_pending_cnt);  in inc_ap_pending()
1998 	if (atomic_read(&device->which) < 0)				\
2001 			atomic_read(&device->which))
2006 	if (atomic_dec_and_test(&device->ap_pending_cnt))  in _dec_ap_pending()
2007 		wake_up(&device->misc_wait);  in _dec_ap_pending()
2011 /* counts how many resync-related answers we still expect from the peer
2019 	atomic_inc(&device->rs_pending_cnt);  in inc_rs_pending()
2025 	atomic_dec(&device->rs_pending_cnt);  in _dec_rs_pending()
2040 	atomic_inc(&device->unacked_cnt);  in inc_unacked()
2046 	atomic_dec(&device->unacked_cnt);  in _dec_unacked()
2053 	atomic_sub(n, &device->unacked_cnt);  in _sub_unacked()
2076  * get_ldev() - Increase the ref count on device->ldev. Returns 0 if there is no ldev
2080  * You have to call put_ldev() when finished working with device->ldev.
2089 	enum drbd_disk_state disk_state = device->state.disk;  in put_ldev()
2094 	int i = atomic_dec_return(&device->local_cnt);  in put_ldev()
2107 			if (!test_and_set_bit(GOING_DISKLESS, &device->flags))  in put_ldev()
2109 		wake_up(&device->misc_wait);  in put_ldev()
2119 	if (device->state.disk == D_DISKLESS)  in _get_ldev_if_state()
2122 	atomic_inc(&device->local_cnt);  in _get_ldev_if_state()
2123 	io_allowed = (device->state.disk >= mins);  in _get_ldev_if_state()
2132 /* this throttles on-the-fly application requests
2134  * maybe re-implement using semaphores? */
2141 	nc = rcu_dereference(first_peer_device(device)->connection->net_conf);  in drbd_get_max_buffers()
2142 	mxb = nc ? nc->max_buffers : 1000000;  /* arbitrary limit on open requests */  in drbd_get_max_buffers()
2150 	union drbd_dev_state s = device->state;  in drbd_state_is_stable()
2184 		if (first_peer_device(device)->connection->agreed_pro_version < 96)  in drbd_state_is_stable()
2196 	switch ((enum drbd_disk_state)s.disk) {  in drbd_state_is_stable()
2203 		/* disk state is stable as well. */  in drbd_state_is_stable()
2220 	struct drbd_resource *resource = device->resource;  in drbd_suspended()
2222 	return resource->susp || resource->susp_fen || resource->susp_nod;  in drbd_suspended()
2231 	if (atomic_read(&device->suspend_cnt))  in may_inc_ap_bio()
2238 	/* no new io accepted when attaching or detaching the disk */  in may_inc_ap_bio()
2244 	if (atomic_read(&device->ap_bio_cnt) > mxb)  in may_inc_ap_bio()
2246 	if (test_bit(BITMAP_IO, &device->flags))  in may_inc_ap_bio()
2255 	spin_lock_irq(&device->resource->req_lock);  in inc_ap_bio_cond()
2258 		atomic_inc(&device->ap_bio_cnt);  in inc_ap_bio_cond()
2259 	spin_unlock_irq(&device->resource->req_lock);  in inc_ap_bio_cond()
2274 	wait_event(device->misc_wait, inc_ap_bio_cond(device));  in inc_ap_bio()
2280 	int ap_bio = atomic_dec_return(&device->ap_bio_cnt);  in dec_ap_bio()
2284 	if (ap_bio == 0 && test_bit(BITMAP_IO, &device->flags)) {  in dec_ap_bio()
2285 		if (!test_and_set_bit(BITMAP_IO_QUEUED, &device->flags))  in dec_ap_bio()
2286 			drbd_queue_work(&first_peer_device(device)->  in dec_ap_bio()
2287 				connection->sender_work,  in dec_ap_bio()
2288 				&device->bm_io_work.w);  in dec_ap_bio()
2295 		wake_up(&device->misc_wait);  in dec_ap_bio()
2300 	return first_peer_device(device)->connection->agreed_pro_version >= 97 &&  in verify_can_do_stop_sector()
2301 		first_peer_device(device)->connection->agreed_pro_version != 100;  in verify_can_do_stop_sector()
2306 	int changed = device->ed_uuid != val;  in drbd_set_ed_uuid()
2307 	device->ed_uuid = val;  in drbd_set_ed_uuid()
2323 	return list_first_entry_or_null(&resource->connections,  in first_connection()