diff options
Diffstat (limited to 'drivers/block/drbd/drbd_actlog.c')
| -rw-r--r-- | drivers/block/drbd/drbd_actlog.c | 1290 |
1 files changed, 609 insertions, 681 deletions
diff --git a/drivers/block/drbd/drbd_actlog.c b/drivers/block/drbd/drbd_actlog.c index 6608076dc39e..742b2908ff68 100644 --- a/drivers/block/drbd/drbd_actlog.c +++ b/drivers/block/drbd/drbd_actlog.c @@ -1,3 +1,4 @@ +// SPDX-License-Identifier: GPL-2.0-only /* drbd_actlog.c @@ -7,19 +8,6 @@ Copyright (C) 2003-2008, Philipp Reisner <philipp.reisner@linbit.com>. Copyright (C) 2003-2008, Lars Ellenberg <lars.ellenberg@linbit.com>. - drbd is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation; either version 2, or (at your option) - any later version. - - drbd is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with drbd; see the file COPYING. If not, write to - the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. */ @@ -27,9 +15,7 @@ #include <linux/crc32c.h> #include <linux/drbd.h> #include <linux/drbd_limits.h> -#include <linux/dynamic_debug.h> #include "drbd_int.h" -#include "drbd_wrappers.h" enum al_transaction_types { @@ -93,36 +79,30 @@ struct __packed al_transaction_on_disk { __be32 context[AL_CONTEXT_PER_TRANSACTION]; }; -struct update_odbm_work { - struct drbd_work w; - unsigned int enr; -}; - -struct update_al_work { - struct drbd_work w; - struct completion event; - int err; -}; - - -void *drbd_md_get_buffer(struct drbd_conf *mdev) +void *drbd_md_get_buffer(struct drbd_device *device, const char *intent) { int r; - wait_event(mdev->misc_wait, - (r = atomic_cmpxchg(&mdev->md_io_in_use, 0, 1)) == 0 || - mdev->state.disk <= D_FAILED); + wait_event(device->misc_wait, + (r = atomic_cmpxchg(&device->md_io.in_use, 0, 1)) == 0 || + device->state.disk <= D_FAILED); - return r ? NULL : page_address(mdev->md_io_page); + if (r) + return NULL; + + device->md_io.current_use = intent; + device->md_io.start_jif = jiffies; + device->md_io.submit_jif = device->md_io.start_jif - 1; + return page_address(device->md_io.page); } -void drbd_md_put_buffer(struct drbd_conf *mdev) +void drbd_md_put_buffer(struct drbd_device *device) { - if (atomic_dec_and_test(&mdev->md_io_in_use)) - wake_up(&mdev->misc_wait); + if (atomic_dec_and_test(&device->md_io.in_use)) + wake_up(&device->misc_wait); } -void wait_until_done_or_force_detached(struct drbd_conf *mdev, struct drbd_backing_dev *bdev, +void wait_until_done_or_force_detached(struct drbd_device *device, struct drbd_backing_dev *bdev, unsigned int *done) { long dt; @@ -134,98 +114,99 @@ void wait_until_done_or_force_detached(struct drbd_conf *mdev, struct drbd_backi if (dt == 0) dt = MAX_SCHEDULE_TIMEOUT; - dt = wait_event_timeout(mdev->misc_wait, - *done || test_bit(FORCE_DETACH, &mdev->flags), dt); + dt = wait_event_timeout(device->misc_wait, + *done || test_bit(FORCE_DETACH, &device->flags), dt); if (dt == 0) { - dev_err(DEV, "meta-data IO operation timed out\n"); - drbd_chk_io_error(mdev, 1, DRBD_FORCE_DETACH); + drbd_err(device, "meta-data IO operation timed out\n"); + drbd_chk_io_error(device, 1, DRBD_FORCE_DETACH); } } -static int _drbd_md_sync_page_io(struct drbd_conf *mdev, +static int _drbd_md_sync_page_io(struct drbd_device *device, struct drbd_backing_dev *bdev, - struct page *page, sector_t sector, - int rw, int size) + sector_t sector, enum req_op op) { struct bio *bio; + /* we do all our meta data IO in aligned 4k blocks. */ + const int size = 4096; int err; + blk_opf_t op_flags = 0; - mdev->md_io.done = 0; - mdev->md_io.error = -ENODEV; + device->md_io.done = 0; + device->md_io.error = -ENODEV; - if ((rw & WRITE) && !test_bit(MD_NO_FUA, &mdev->flags)) - rw |= REQ_FUA | REQ_FLUSH; - rw |= REQ_SYNC; + if ((op == REQ_OP_WRITE) && !test_bit(MD_NO_FUA, &device->flags)) + op_flags |= REQ_FUA | REQ_PREFLUSH; + op_flags |= REQ_SYNC; - bio = bio_alloc_drbd(GFP_NOIO); - bio->bi_bdev = bdev->md_bdev; - bio->bi_sector = sector; + bio = bio_alloc_bioset(bdev->md_bdev, 1, op | op_flags, GFP_NOIO, + &drbd_md_io_bio_set); + bio->bi_iter.bi_sector = sector; err = -EIO; - if (bio_add_page(bio, page, size, 0) != size) + if (bio_add_page(bio, device->md_io.page, size, 0) != size) goto out; - bio->bi_private = &mdev->md_io; - bio->bi_end_io = drbd_md_io_complete; - bio->bi_rw = rw; + bio->bi_private = device; + bio->bi_end_io = drbd_md_endio; - if (!(rw & WRITE) && mdev->state.disk == D_DISKLESS && mdev->ldev == NULL) + if (op != REQ_OP_WRITE && device->state.disk == D_DISKLESS && device->ldev == NULL) /* special case, drbd_md_read() during drbd_adm_attach(): no get_ldev */ ; - else if (!get_ldev_if_state(mdev, D_ATTACHING)) { - /* Corresponding put_ldev in drbd_md_io_complete() */ - dev_err(DEV, "ASSERT FAILED: get_ldev_if_state() == 1 in _drbd_md_sync_page_io()\n"); + else if (!get_ldev_if_state(device, D_ATTACHING)) { + /* Corresponding put_ldev in drbd_md_endio() */ + drbd_err(device, "ASSERT FAILED: get_ldev_if_state() == 1 in _drbd_md_sync_page_io()\n"); err = -ENODEV; goto out; } bio_get(bio); /* one bio_put() is in the completion handler */ - atomic_inc(&mdev->md_io_in_use); /* drbd_md_put_buffer() is in the completion handler */ - if (drbd_insert_fault(mdev, (rw & WRITE) ? DRBD_FAULT_MD_WR : DRBD_FAULT_MD_RD)) - bio_endio(bio, -EIO); + atomic_inc(&device->md_io.in_use); /* drbd_md_put_buffer() is in the completion handler */ + device->md_io.submit_jif = jiffies; + if (drbd_insert_fault(device, (op == REQ_OP_WRITE) ? DRBD_FAULT_MD_WR : DRBD_FAULT_MD_RD)) + bio_io_error(bio); else - submit_bio(rw, bio); - wait_until_done_or_force_detached(mdev, bdev, &mdev->md_io.done); - if (bio_flagged(bio, BIO_UPTODATE)) - err = mdev->md_io.error; + submit_bio(bio); + wait_until_done_or_force_detached(device, bdev, &device->md_io.done); + if (!bio->bi_status) + err = device->md_io.error; out: bio_put(bio); return err; } -int drbd_md_sync_page_io(struct drbd_conf *mdev, struct drbd_backing_dev *bdev, - sector_t sector, int rw) +int drbd_md_sync_page_io(struct drbd_device *device, struct drbd_backing_dev *bdev, + sector_t sector, enum req_op op) { int err; - struct page *iop = mdev->md_io_page; - - D_ASSERT(atomic_read(&mdev->md_io_in_use) == 1); + D_ASSERT(device, atomic_read(&device->md_io.in_use) == 1); BUG_ON(!bdev->md_bdev); - dev_dbg(DEV, "meta_data io: %s [%d]:%s(,%llus,%s) %pS\n", + dynamic_drbd_dbg(device, "meta_data io: %s [%d]:%s(,%llus,%s) %pS\n", current->comm, current->pid, __func__, - (unsigned long long)sector, (rw & WRITE) ? "WRITE" : "READ", + (unsigned long long)sector, (op == REQ_OP_WRITE) ? "WRITE" : "READ", (void*)_RET_IP_ ); if (sector < drbd_md_first_sector(bdev) || sector + 7 > drbd_md_last_sector(bdev)) - dev_alert(DEV, "%s [%d]:%s(,%llus,%s) out of range md access!\n", + drbd_alert(device, "%s [%d]:%s(,%llus,%s) out of range md access!\n", current->comm, current->pid, __func__, - (unsigned long long)sector, (rw & WRITE) ? "WRITE" : "READ"); + (unsigned long long)sector, + (op == REQ_OP_WRITE) ? "WRITE" : "READ"); - /* we do all our meta data IO in aligned 4k blocks. */ - err = _drbd_md_sync_page_io(mdev, bdev, iop, sector, rw, 4096); + err = _drbd_md_sync_page_io(device, bdev, sector, op); if (err) { - dev_err(DEV, "drbd_md_sync_page_io(,%llus,%s) failed with error %d\n", - (unsigned long long)sector, (rw & WRITE) ? "WRITE" : "READ", err); + drbd_err(device, "drbd_md_sync_page_io(,%llus,%s) failed with error %d\n", + (unsigned long long)sector, + (op == REQ_OP_WRITE) ? "WRITE" : "READ", err); } return err; } -static struct bm_extent *find_active_resync_extent(struct drbd_conf *mdev, unsigned int enr) +static struct bm_extent *find_active_resync_extent(struct drbd_device *device, unsigned int enr) { struct lc_element *tmp; - tmp = lc_find(mdev->resync, enr/AL_EXT_PER_BM_SECT); + tmp = lc_find(device->resync, enr/AL_EXT_PER_BM_SECT); if (unlikely(tmp != NULL)) { struct bm_extent *bm_ext = lc_entry(tmp, struct bm_extent, lce); if (test_bit(BME_NO_WRITES, &bm_ext->flags)) @@ -234,47 +215,47 @@ static struct bm_extent *find_active_resync_extent(struct drbd_conf *mdev, unsig return NULL; } -static struct lc_element *_al_get(struct drbd_conf *mdev, unsigned int enr, bool nonblock) +static struct lc_element *_al_get(struct drbd_device *device, unsigned int enr, bool nonblock) { struct lc_element *al_ext; struct bm_extent *bm_ext; int wake; - spin_lock_irq(&mdev->al_lock); - bm_ext = find_active_resync_extent(mdev, enr); + spin_lock_irq(&device->al_lock); + bm_ext = find_active_resync_extent(device, enr); if (bm_ext) { wake = !test_and_set_bit(BME_PRIORITY, &bm_ext->flags); - spin_unlock_irq(&mdev->al_lock); + spin_unlock_irq(&device->al_lock); if (wake) - wake_up(&mdev->al_wait); + wake_up(&device->al_wait); return NULL; } if (nonblock) - al_ext = lc_try_get(mdev->act_log, enr); + al_ext = lc_try_get(device->act_log, enr); else - al_ext = lc_get(mdev->act_log, enr); - spin_unlock_irq(&mdev->al_lock); + al_ext = lc_get(device->act_log, enr); + spin_unlock_irq(&device->al_lock); return al_ext; } -bool drbd_al_begin_io_fastpath(struct drbd_conf *mdev, struct drbd_interval *i) +bool drbd_al_begin_io_fastpath(struct drbd_device *device, struct drbd_interval *i) { /* for bios crossing activity log extent boundaries, * we may need to activate two extents in one go */ unsigned first = i->sector >> (AL_EXTENT_SHIFT-9); unsigned last = i->size == 0 ? first : (i->sector + (i->size >> 9) - 1) >> (AL_EXTENT_SHIFT-9); - D_ASSERT((unsigned)(last - first) <= 1); - D_ASSERT(atomic_read(&mdev->local_cnt) > 0); + D_ASSERT(device, first <= last); + D_ASSERT(device, atomic_read(&device->local_cnt) > 0); /* FIXME figure out a fast path for bios crossing AL extent boundaries */ if (first != last) return false; - return _al_get(mdev, first, true); + return _al_get(device, first, true); } -bool drbd_al_begin_io_prepare(struct drbd_conf *mdev, struct drbd_interval *i) +bool drbd_al_begin_io_prepare(struct drbd_device *device, struct drbd_interval *i) { /* for bios crossing activity log extent boundaries, * we may need to activate two extents in one go */ @@ -283,156 +264,19 @@ bool drbd_al_begin_io_prepare(struct drbd_conf *mdev, struct drbd_interval *i) unsigned enr; bool need_transaction = false; - D_ASSERT(first <= last); - D_ASSERT(atomic_read(&mdev->local_cnt) > 0); + D_ASSERT(device, first <= last); + D_ASSERT(device, atomic_read(&device->local_cnt) > 0); for (enr = first; enr <= last; enr++) { struct lc_element *al_ext; - wait_event(mdev->al_wait, - (al_ext = _al_get(mdev, enr, false)) != NULL); + wait_event(device->al_wait, + (al_ext = _al_get(device, enr, false)) != NULL); if (al_ext->lc_number != enr) need_transaction = true; } return need_transaction; } -static int al_write_transaction(struct drbd_conf *mdev, bool delegate); - -/* When called through generic_make_request(), we must delegate - * activity log I/O to the worker thread: a further request - * submitted via generic_make_request() within the same task - * would be queued on current->bio_list, and would only start - * after this function returns (see generic_make_request()). - * - * However, if we *are* the worker, we must not delegate to ourselves. - */ - -/* - * @delegate: delegate activity log I/O to the worker thread - */ -void drbd_al_begin_io_commit(struct drbd_conf *mdev, bool delegate) -{ - bool locked = false; - - BUG_ON(delegate && current == mdev->tconn->worker.task); - - /* Serialize multiple transactions. - * This uses test_and_set_bit, memory barrier is implicit. - */ - wait_event(mdev->al_wait, - mdev->act_log->pending_changes == 0 || - (locked = lc_try_lock_for_transaction(mdev->act_log))); - - if (locked) { - /* Double check: it may have been committed by someone else, - * while we have been waiting for the lock. */ - if (mdev->act_log->pending_changes) { - bool write_al_updates; - - rcu_read_lock(); - write_al_updates = rcu_dereference(mdev->ldev->disk_conf)->al_updates; - rcu_read_unlock(); - - if (write_al_updates) - al_write_transaction(mdev, delegate); - spin_lock_irq(&mdev->al_lock); - /* FIXME - if (err) - we need an "lc_cancel" here; - */ - lc_committed(mdev->act_log); - spin_unlock_irq(&mdev->al_lock); - } - lc_unlock(mdev->act_log); - wake_up(&mdev->al_wait); - } -} - -/* - * @delegate: delegate activity log I/O to the worker thread - */ -void drbd_al_begin_io(struct drbd_conf *mdev, struct drbd_interval *i, bool delegate) -{ - BUG_ON(delegate && current == mdev->tconn->worker.task); - - if (drbd_al_begin_io_prepare(mdev, i)) - drbd_al_begin_io_commit(mdev, delegate); -} - -int drbd_al_begin_io_nonblock(struct drbd_conf *mdev, struct drbd_interval *i) -{ - struct lru_cache *al = mdev->act_log; - /* for bios crossing activity log extent boundaries, - * we may need to activate two extents in one go */ - unsigned first = i->sector >> (AL_EXTENT_SHIFT-9); - unsigned last = i->size == 0 ? first : (i->sector + (i->size >> 9) - 1) >> (AL_EXTENT_SHIFT-9); - unsigned nr_al_extents; - unsigned available_update_slots; - unsigned enr; - - D_ASSERT(first <= last); - - nr_al_extents = 1 + last - first; /* worst case: all touched extends are cold. */ - available_update_slots = min(al->nr_elements - al->used, - al->max_pending_changes - al->pending_changes); - - /* We want all necessary updates for a given request within the same transaction - * We could first check how many updates are *actually* needed, - * and use that instead of the worst-case nr_al_extents */ - if (available_update_slots < nr_al_extents) - return -EWOULDBLOCK; - - /* Is resync active in this area? */ - for (enr = first; enr <= last; enr++) { - struct lc_element *tmp; - tmp = lc_find(mdev->resync, enr/AL_EXT_PER_BM_SECT); - if (unlikely(tmp != NULL)) { - struct bm_extent *bm_ext = lc_entry(tmp, struct bm_extent, lce); - if (test_bit(BME_NO_WRITES, &bm_ext->flags)) { - if (!test_and_set_bit(BME_PRIORITY, &bm_ext->flags)) - return -EBUSY; - return -EWOULDBLOCK; - } - } - } - - /* Checkout the refcounts. - * Given that we checked for available elements and update slots above, - * this has to be successful. */ - for (enr = first; enr <= last; enr++) { - struct lc_element *al_ext; - al_ext = lc_get_cumulative(mdev->act_log, enr); - if (!al_ext) - dev_info(DEV, "LOGIC BUG for enr=%u\n", enr); - } - return 0; -} - -void drbd_al_complete_io(struct drbd_conf *mdev, struct drbd_interval *i) -{ - /* for bios crossing activity log extent boundaries, - * we may need to activate two extents in one go */ - unsigned first = i->sector >> (AL_EXTENT_SHIFT-9); - unsigned last = i->size == 0 ? first : (i->sector + (i->size >> 9) - 1) >> (AL_EXTENT_SHIFT-9); - unsigned enr; - struct lc_element *extent; - unsigned long flags; - - D_ASSERT(first <= last); - spin_lock_irqsave(&mdev->al_lock, flags); - - for (enr = first; enr <= last; enr++) { - extent = lc_find(mdev->act_log, enr); - if (!extent) { - dev_err(DEV, "al_complete_io() called on inactive extent %u\n", enr); - continue; - } - lc_put(mdev->act_log, extent); - } - spin_unlock_irqrestore(&mdev->al_lock, flags); - wake_up(&mdev->al_wait); -} - #if (PAGE_SHIFT + 3) < (AL_EXTENT_SHIFT - BM_BLOCK_SHIFT) /* Currently BM_BLOCK_SHIFT, BM_EXT_SHIFT and AL_EXTENT_SHIFT * are still coupled, or assume too much about their relation. @@ -451,22 +295,13 @@ static unsigned int al_extent_to_bm_page(unsigned int al_enr) (AL_EXTENT_SHIFT - BM_BLOCK_SHIFT)); } -static unsigned int rs_extent_to_bm_page(unsigned int rs_enr) -{ - return rs_enr >> - /* bit to page */ - ((PAGE_SHIFT + 3) - - /* resync extent number to bit */ - (BM_EXT_SHIFT - BM_BLOCK_SHIFT)); -} - -static sector_t al_tr_number_to_on_disk_sector(struct drbd_conf *mdev) +static sector_t al_tr_number_to_on_disk_sector(struct drbd_device *device) { - const unsigned int stripes = mdev->ldev->md.al_stripes; - const unsigned int stripe_size_4kB = mdev->ldev->md.al_stripe_size_4k; + const unsigned int stripes = device->ldev->md.al_stripes; + const unsigned int stripe_size_4kB = device->ldev->md.al_stripe_size_4k; /* transaction number, modulo on-disk ring buffer wrap around */ - unsigned int t = mdev->al_tr_number % (mdev->ldev->md.al_size_4k); + unsigned int t = device->al_tr_number % (device->ldev->md.al_size_4k); /* ... to aligned 4k on disk block */ t = ((t % stripes) * stripe_size_4kB) + t/stripes; @@ -475,13 +310,11 @@ static sector_t al_tr_number_to_on_disk_sector(struct drbd_conf *mdev) t *= 8; /* ... plus offset to the on disk position */ - return mdev->ldev->md.md_offset + mdev->ldev->md.al_offset + t; + return device->ldev->md.md_offset + device->ldev->md.al_offset + t; } -static int -_al_write_transaction(struct drbd_conf *mdev) +static int __al_write_transaction(struct drbd_device *device, struct al_transaction_on_disk *buffer) { - struct al_transaction_on_disk *buffer; struct lc_element *e; sector_t sector; int i, mx; @@ -489,40 +322,20 @@ _al_write_transaction(struct drbd_conf *mdev) unsigned crc = 0; int err = 0; - if (!get_ldev(mdev)) { - dev_err(DEV, "disk is %s, cannot start al transaction\n", - drbd_disk_str(mdev->state.disk)); - return -EIO; - } - - /* The bitmap write may have failed, causing a state change. */ - if (mdev->state.disk < D_INCONSISTENT) { - dev_err(DEV, - "disk is %s, cannot write al transaction\n", - drbd_disk_str(mdev->state.disk)); - put_ldev(mdev); - return -EIO; - } - - buffer = drbd_md_get_buffer(mdev); /* protects md_io_buffer, al_tr_cycle, ... */ - if (!buffer) { - dev_err(DEV, "disk failed while waiting for md_io buffer\n"); - put_ldev(mdev); - return -ENODEV; - } - memset(buffer, 0, sizeof(*buffer)); buffer->magic = cpu_to_be32(DRBD_AL_MAGIC); - buffer->tr_number = cpu_to_be32(mdev->al_tr_number); + buffer->tr_number = cpu_to_be32(device->al_tr_number); i = 0; + drbd_bm_reset_al_hints(device); + /* Even though no one can start to change this list * once we set the LC_LOCKED -- from drbd_al_begin_io(), * lc_try_lock_for_transaction() --, someone may still * be in the process of changing it. */ - spin_lock_irq(&mdev->al_lock); - list_for_each_entry(e, &mdev->act_log->to_be_changed, list) { + spin_lock_irq(&device->al_lock); + list_for_each_entry(e, &device->act_log->to_be_changed, list) { if (i == AL_UPDATES_PER_TRANSACTION) { i++; break; @@ -530,11 +343,11 @@ _al_write_transaction(struct drbd_conf *mdev) buffer->update_slot_nr[i] = cpu_to_be16(e->lc_index); buffer->update_extent_nr[i] = cpu_to_be32(e->lc_new_number); if (e->lc_number != LC_FREE) - drbd_bm_mark_for_writeout(mdev, + drbd_bm_mark_for_writeout(device, al_extent_to_bm_page(e->lc_number)); i++; } - spin_unlock_irq(&mdev->al_lock); + spin_unlock_irq(&device->al_lock); BUG_ON(i > AL_UPDATES_PER_TRANSACTION); buffer->n_updates = cpu_to_be16(i); @@ -543,191 +356,332 @@ _al_write_transaction(struct drbd_conf *mdev) buffer->update_extent_nr[i] = cpu_to_be32(LC_FREE); } - buffer->context_size = cpu_to_be16(mdev->act_log->nr_elements); - buffer->context_start_slot_nr = cpu_to_be16(mdev->al_tr_cycle); + buffer->context_size = cpu_to_be16(device->act_log->nr_elements); + buffer->context_start_slot_nr = cpu_to_be16(device->al_tr_cycle); mx = min_t(int, AL_CONTEXT_PER_TRANSACTION, - mdev->act_log->nr_elements - mdev->al_tr_cycle); + device->act_log->nr_elements - device->al_tr_cycle); for (i = 0; i < mx; i++) { - unsigned idx = mdev->al_tr_cycle + i; - extent_nr = lc_element_by_index(mdev->act_log, idx)->lc_number; + unsigned idx = device->al_tr_cycle + i; + extent_nr = lc_element_by_index(device->act_log, idx)->lc_number; buffer->context[i] = cpu_to_be32(extent_nr); } for (; i < AL_CONTEXT_PER_TRANSACTION; i++) buffer->context[i] = cpu_to_be32(LC_FREE); - mdev->al_tr_cycle += AL_CONTEXT_PER_TRANSACTION; - if (mdev->al_tr_cycle >= mdev->act_log->nr_elements) - mdev->al_tr_cycle = 0; + device->al_tr_cycle += AL_CONTEXT_PER_TRANSACTION; + if (device->al_tr_cycle >= device->act_log->nr_elements) + device->al_tr_cycle = 0; - sector = al_tr_number_to_on_disk_sector(mdev); + sector = al_tr_number_to_on_disk_sector(device); crc = crc32c(0, buffer, 4096); buffer->crc32c = cpu_to_be32(crc); - if (drbd_bm_write_hinted(mdev)) + if (drbd_bm_write_hinted(device)) err = -EIO; else { bool write_al_updates; rcu_read_lock(); - write_al_updates = rcu_dereference(mdev->ldev->disk_conf)->al_updates; + write_al_updates = rcu_dereference(device->ldev->disk_conf)->al_updates; rcu_read_unlock(); if (write_al_updates) { - if (drbd_md_sync_page_io(mdev, mdev->ldev, sector, WRITE)) { + if (drbd_md_sync_page_io(device, device->ldev, sector, REQ_OP_WRITE)) { err = -EIO; - drbd_chk_io_error(mdev, 1, DRBD_META_IO_ERROR); + drbd_chk_io_error(device, 1, DRBD_META_IO_ERROR); } else { - mdev->al_tr_number++; - mdev->al_writ_cnt++; + device->al_tr_number++; + device->al_writ_cnt++; } } } - drbd_md_put_buffer(mdev); - put_ldev(mdev); + return err; +} + +static int al_write_transaction(struct drbd_device *device) +{ + struct al_transaction_on_disk *buffer; + int err; + + if (!get_ldev(device)) { + drbd_err(device, "disk is %s, cannot start al transaction\n", + drbd_disk_str(device->state.disk)); + return -EIO; + } + + /* The bitmap write may have failed, causing a state change. */ + if (device->state.disk < D_INCONSISTENT) { + drbd_err(device, + "disk is %s, cannot write al transaction\n", + drbd_disk_str(device->state.disk)); + put_ldev(device); + return -EIO; + } + + /* protects md_io_buffer, al_tr_cycle, ... */ + buffer = drbd_md_get_buffer(device, __func__); + if (!buffer) { + drbd_err(device, "disk failed while waiting for md_io buffer\n"); + put_ldev(device); + return -ENODEV; + } + + err = __al_write_transaction(device, buffer); + + drbd_md_put_buffer(device); + put_ldev(device); return err; } -static int w_al_write_transaction(struct drbd_work *w, int unused) +void drbd_al_begin_io_commit(struct drbd_device *device) { - struct update_al_work *aw = container_of(w, struct update_al_work, w); - struct drbd_conf *mdev = w->mdev; - int err; + bool locked = false; - err = _al_write_transaction(mdev); - aw->err = err; - complete(&aw->event); + /* Serialize multiple transactions. + * This uses test_and_set_bit, memory barrier is implicit. + */ + wait_event(device->al_wait, + device->act_log->pending_changes == 0 || + (locked = lc_try_lock_for_transaction(device->act_log))); + + if (locked) { + /* Double check: it may have been committed by someone else, + * while we have been waiting for the lock. */ + if (device->act_log->pending_changes) { + bool write_al_updates; - return err != -EIO ? err : 0; + rcu_read_lock(); + write_al_updates = rcu_dereference(device->ldev->disk_conf)->al_updates; + rcu_read_unlock(); + + if (write_al_updates) + al_write_transaction(device); + spin_lock_irq(&device->al_lock); + /* FIXME + if (err) + we need an "lc_cancel" here; + */ + lc_committed(device->act_log); + spin_unlock_irq(&device->al_lock); + } + lc_unlock(device->act_log); + wake_up(&device->al_wait); + } } -/* Calls from worker context (see w_restart_disk_io()) need to write the - transaction directly. Others came through generic_make_request(), - those need to delegate it to the worker. */ -static int al_write_transaction(struct drbd_conf *mdev, bool delegate) +/* + * @delegate: delegate activity log I/O to the worker thread + */ +void drbd_al_begin_io(struct drbd_device *device, struct drbd_interval *i) { - if (delegate) { - struct update_al_work al_work; - init_completion(&al_work.event); - al_work.w.cb = w_al_write_transaction; - al_work.w.mdev = mdev; - drbd_queue_work_front(&mdev->tconn->sender_work, &al_work.w); - wait_for_completion(&al_work.event); - return al_work.err; - } else - return _al_write_transaction(mdev); + if (drbd_al_begin_io_prepare(device, i)) + drbd_al_begin_io_commit(device); } -static int _try_lc_del(struct drbd_conf *mdev, struct lc_element *al_ext) +int drbd_al_begin_io_nonblock(struct drbd_device *device, struct drbd_interval *i) +{ + struct lru_cache *al = device->act_log; + /* for bios crossing activity log extent boundaries, + * we may need to activate two extents in one go */ + unsigned first = i->sector >> (AL_EXTENT_SHIFT-9); + unsigned last = i->size == 0 ? first : (i->sector + (i->size >> 9) - 1) >> (AL_EXTENT_SHIFT-9); + unsigned nr_al_extents; + unsigned available_update_slots; + unsigned enr; + + D_ASSERT(device, first <= last); + + nr_al_extents = 1 + last - first; /* worst case: all touched extends are cold. */ + available_update_slots = min(al->nr_elements - al->used, + al->max_pending_changes - al->pending_changes); + + /* We want all necessary updates for a given request within the same transaction + * We could first check how many updates are *actually* needed, + * and use that instead of the worst-case nr_al_extents */ + if (available_update_slots < nr_al_extents) { + /* Too many activity log extents are currently "hot". + * + * If we have accumulated pending changes already, + * we made progress. + * + * If we cannot get even a single pending change through, + * stop the fast path until we made some progress, + * or requests to "cold" extents could be starved. */ + if (!al->pending_changes) + __set_bit(__LC_STARVING, &device->act_log->flags); + return -ENOBUFS; + } + + /* Is resync active in this area? */ + for (enr = first; enr <= last; enr++) { + struct lc_element *tmp; + tmp = lc_find(device->resync, enr/AL_EXT_PER_BM_SECT); + if (unlikely(tmp != NULL)) { + struct bm_extent *bm_ext = lc_entry(tmp, struct bm_extent, lce); + if (test_bit(BME_NO_WRITES, &bm_ext->flags)) { + if (!test_and_set_bit(BME_PRIORITY, &bm_ext->flags)) + return -EBUSY; + return -EWOULDBLOCK; + } + } + } + + /* Checkout the refcounts. + * Given that we checked for available elements and update slots above, + * this has to be successful. */ + for (enr = first; enr <= last; enr++) { + struct lc_element *al_ext; + al_ext = lc_get_cumulative(device->act_log, enr); + if (!al_ext) + drbd_info(device, "LOGIC BUG for enr=%u\n", enr); + } + return 0; +} + +void drbd_al_complete_io(struct drbd_device *device, struct drbd_interval *i) +{ + /* for bios crossing activity log extent boundaries, + * we may need to activate two extents in one go */ + unsigned first = i->sector >> (AL_EXTENT_SHIFT-9); + unsigned last = i->size == 0 ? first : (i->sector + (i->size >> 9) - 1) >> (AL_EXTENT_SHIFT-9); + unsigned enr; + struct lc_element *extent; + unsigned long flags; + + D_ASSERT(device, first <= last); + spin_lock_irqsave(&device->al_lock, flags); + + for (enr = first; enr <= last; enr++) { + extent = lc_find(device->act_log, enr); + if (!extent) { + drbd_err(device, "al_complete_io() called on inactive extent %u\n", enr); + continue; + } + lc_put(device->act_log, extent); + } + spin_unlock_irqrestore(&device->al_lock, flags); + wake_up(&device->al_wait); +} + +static int _try_lc_del(struct drbd_device *device, struct lc_element *al_ext) { int rv; - spin_lock_irq(&mdev->al_lock); + spin_lock_irq(&device->al_lock); rv = (al_ext->refcnt == 0); if (likely(rv)) - lc_del(mdev->act_log, al_ext); - spin_unlock_irq(&mdev->al_lock); + lc_del(device->act_log, al_ext); + spin_unlock_irq(&device->al_lock); return rv; } /** * drbd_al_shrink() - Removes all active extents form the activity log - * @mdev: DRBD device. + * @device: DRBD device. * * Removes all active extents form the activity log, waiting until * the reference count of each entry dropped to 0 first, of course. * - * You need to lock mdev->act_log with lc_try_lock() / lc_unlock() + * You need to lock device->act_log with lc_try_lock() / lc_unlock() */ -void drbd_al_shrink(struct drbd_conf *mdev) +void drbd_al_shrink(struct drbd_device *device) { struct lc_element *al_ext; int i; - D_ASSERT(test_bit(__LC_LOCKED, &mdev->act_log->flags)); + D_ASSERT(device, test_bit(__LC_LOCKED, &device->act_log->flags)); - for (i = 0; i < mdev->act_log->nr_elements; i++) { - al_ext = lc_element_by_index(mdev->act_log, i); + for (i = 0; i < device->act_log->nr_elements; i++) { + al_ext = lc_element_by_index(device->act_log, i); if (al_ext->lc_number == LC_FREE) continue; - wait_event(mdev->al_wait, _try_lc_del(mdev, al_ext)); + wait_event(device->al_wait, _try_lc_del(device, al_ext)); } - wake_up(&mdev->al_wait); + wake_up(&device->al_wait); } -static int w_update_odbm(struct drbd_work *w, int unused) +int drbd_al_initialize(struct drbd_device *device, void *buffer) { - struct update_odbm_work *udw = container_of(w, struct update_odbm_work, w); - struct drbd_conf *mdev = w->mdev; - struct sib_info sib = { .sib_reason = SIB_SYNC_PROGRESS, }; - - if (!get_ldev(mdev)) { - if (__ratelimit(&drbd_ratelimit_state)) - dev_warn(DEV, "Can not update on disk bitmap, local IO disabled.\n"); - kfree(udw); - return 0; - } - - drbd_bm_write_page(mdev, rs_extent_to_bm_page(udw->enr)); - put_ldev(mdev); - - kfree(udw); + struct al_transaction_on_disk *al = buffer; + struct drbd_md *md = &device->ldev->md; + int al_size_4k = md->al_stripes * md->al_stripe_size_4k; + int i; - if (drbd_bm_total_weight(mdev) <= mdev->rs_failed) { - switch (mdev->state.conn) { - case C_SYNC_SOURCE: case C_SYNC_TARGET: - case C_PAUSED_SYNC_S: case C_PAUSED_SYNC_T: - drbd_resync_finished(mdev); - default: - /* nothing to do */ - break; - } + __al_write_transaction(device, al); + /* There may or may not have been a pending transaction. */ + spin_lock_irq(&device->al_lock); + lc_committed(device->act_log); + spin_unlock_irq(&device->al_lock); + + /* The rest of the transactions will have an empty "updates" list, and + * are written out only to provide the context, and to initialize the + * on-disk ring buffer. */ + for (i = 1; i < al_size_4k; i++) { + int err = __al_write_transaction(device, al); + if (err) + return err; } - drbd_bcast_event(mdev, &sib); - return 0; } +static const char *drbd_change_sync_fname[] = { + [RECORD_RS_FAILED] = "drbd_rs_failed_io", + [SET_IN_SYNC] = "drbd_set_in_sync", + [SET_OUT_OF_SYNC] = "drbd_set_out_of_sync" +}; /* ATTENTION. The AL's extents are 4MB each, while the extents in the * resync LRU-cache are 16MB each. * The caller of this function has to hold an get_ldev() reference. * + * Adjusts the caching members ->rs_left (success) or ->rs_failed (!success), + * potentially pulling in (and recounting the corresponding bits) + * this resync extent into the resync extent lru cache. + * + * Returns whether all bits have been cleared for this resync extent, + * precisely: (rs_left <= rs_failed) + * * TODO will be obsoleted once we have a caching lru of the on disk bitmap */ -static void drbd_try_clear_on_disk_bm(struct drbd_conf *mdev, sector_t sector, - int count, int success) +static bool update_rs_extent(struct drbd_device *device, + unsigned int enr, int count, + enum update_sync_bits_mode mode) { struct lc_element *e; - struct update_odbm_work *udw; - - unsigned int enr; - - D_ASSERT(atomic_read(&mdev->local_cnt)); - /* I simply assume that a sector/size pair never crosses - * a 16 MB extent border. (Currently this is true...) */ - enr = BM_SECT_TO_EXT(sector); - - e = lc_get(mdev->resync, enr); + D_ASSERT(device, atomic_read(&device->local_cnt)); + + /* When setting out-of-sync bits, + * we don't need it cached (lc_find). + * But if it is present in the cache, + * we should update the cached bit count. + * Otherwise, that extent should be in the resync extent lru cache + * already -- or we want to pull it in if necessary -- (lc_get), + * then update and check rs_left and rs_failed. */ + if (mode == SET_OUT_OF_SYNC) + e = lc_find(device->resync, enr); + else + e = lc_get(device->resync, enr); if (e) { struct bm_extent *ext = lc_entry(e, struct bm_extent, lce); if (ext->lce.lc_number == enr) { - if (success) + if (mode == SET_IN_SYNC) ext->rs_left -= count; + else if (mode == SET_OUT_OF_SYNC) + ext->rs_left += count; else ext->rs_failed += count; if (ext->rs_left < ext->rs_failed) { - dev_warn(DEV, "BAD! sector=%llus enr=%u rs_left=%d " + drbd_warn(device, "BAD! enr=%u rs_left=%d " "rs_failed=%d count=%d cstate=%s\n", - (unsigned long long)sector, ext->lce.lc_number, ext->rs_left, ext->rs_failed, count, - drbd_conn_str(mdev->state.conn)); + drbd_conn_str(device->state.conn)); /* We don't expect to be able to clear more bits * than have been set when we originally counted @@ -735,7 +689,7 @@ static void drbd_try_clear_on_disk_bm(struct drbd_conf *mdev, sector_t sector, * Whatever the reason (disconnect during resync, * delayed local completion of an application write), * try to fix it up by recounting here. */ - ext->rs_left = drbd_bm_e_weight(mdev, enr); + ext->rs_left = drbd_bm_e_weight(device, enr); } } else { /* Normally this element should be in the cache, @@ -744,229 +698,242 @@ static void drbd_try_clear_on_disk_bm(struct drbd_conf *mdev, sector_t sector, * But maybe an application write finished, and we set * something outside the resync lru_cache in sync. */ - int rs_left = drbd_bm_e_weight(mdev, enr); + int rs_left = drbd_bm_e_weight(device, enr); if (ext->flags != 0) { - dev_warn(DEV, "changing resync lce: %d[%u;%02lx]" + drbd_warn(device, "changing resync lce: %d[%u;%02lx]" " -> %d[%u;00]\n", ext->lce.lc_number, ext->rs_left, ext->flags, enr, rs_left); ext->flags = 0; } if (ext->rs_failed) { - dev_warn(DEV, "Kicking resync_lru element enr=%u " + drbd_warn(device, "Kicking resync_lru element enr=%u " "out with rs_failed=%d\n", ext->lce.lc_number, ext->rs_failed); } ext->rs_left = rs_left; - ext->rs_failed = success ? 0 : count; + ext->rs_failed = (mode == RECORD_RS_FAILED) ? count : 0; /* we don't keep a persistent log of the resync lru, * we can commit any change right away. */ - lc_committed(mdev->resync); + lc_committed(device->resync); } - lc_put(mdev->resync, &ext->lce); + if (mode != SET_OUT_OF_SYNC) + lc_put(device->resync, &ext->lce); /* no race, we are within the al_lock! */ - if (ext->rs_left == ext->rs_failed) { + if (ext->rs_left <= ext->rs_failed) { ext->rs_failed = 0; - - udw = kmalloc(sizeof(*udw), GFP_ATOMIC); - if (udw) { - udw->enr = ext->lce.lc_number; - udw->w.cb = w_update_odbm; - udw->w.mdev = mdev; - drbd_queue_work_front(&mdev->tconn->sender_work, &udw->w); - } else { - dev_warn(DEV, "Could not kmalloc an udw\n"); - } + return true; } - } else { - dev_err(DEV, "lc_get() failed! locked=%d/%d flags=%lu\n", - mdev->resync_locked, - mdev->resync->nr_elements, - mdev->resync->flags); + } else if (mode != SET_OUT_OF_SYNC) { + /* be quiet if lc_find() did not find it. */ + drbd_err(device, "lc_get() failed! locked=%d/%d flags=%lu\n", + device->resync_locked, + device->resync->nr_elements, + device->resync->flags); } + return false; } -void drbd_advance_rs_marks(struct drbd_conf *mdev, unsigned long still_to_go) +void drbd_advance_rs_marks(struct drbd_peer_device *peer_device, unsigned long still_to_go) { + struct drbd_device *device = peer_device->device; unsigned long now = jiffies; - unsigned long last = mdev->rs_mark_time[mdev->rs_last_mark]; - int next = (mdev->rs_last_mark + 1) % DRBD_SYNC_MARKS; + unsigned long last = device->rs_mark_time[device->rs_last_mark]; + int next = (device->rs_last_mark + 1) % DRBD_SYNC_MARKS; if (time_after_eq(now, last + DRBD_SYNC_MARK_STEP)) { - if (mdev->rs_mark_left[mdev->rs_last_mark] != still_to_go && - mdev->state.conn != C_PAUSED_SYNC_T && - mdev->state.conn != C_PAUSED_SYNC_S) { - mdev->rs_mark_time[next] = now; - mdev->rs_mark_left[next] = still_to_go; - mdev->rs_last_mark = next; + if (device->rs_mark_left[device->rs_last_mark] != still_to_go && + device->state.conn != C_PAUSED_SYNC_T && + device->state.conn != C_PAUSED_SYNC_S) { + device->rs_mark_time[next] = now; + device->rs_mark_left[next] = still_to_go; + device->rs_last_mark = next; } } } -/* 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. - * - * called by worker on C_SYNC_TARGET and receiver on SyncSource. - * - */ -void __drbd_set_in_sync(struct drbd_conf *mdev, sector_t sector, int size, - const char *file, const unsigned int line) +/* It is called lazy update, so don't do write-out too often. */ +static bool lazy_bitmap_update_due(struct drbd_device *device) { - /* Is called from worker and receiver context _only_ */ - unsigned long sbnr, ebnr, lbnr; - unsigned long count = 0; - sector_t esector, nr_sectors; - int wake_up = 0; - unsigned long flags; + return time_after(jiffies, device->rs_last_bcast + 2*HZ); +} - if (size <= 0 || !IS_ALIGNED(size, 512) || size > DRBD_MAX_BIO_SIZE) { - dev_err(DEV, "drbd_set_in_sync: sector=%llus size=%d nonsense!\n", - (unsigned long long)sector, size); +static void maybe_schedule_on_disk_bitmap_update(struct drbd_device *device, bool rs_done) +{ + if (rs_done) { + struct drbd_connection *connection = first_peer_device(device)->connection; + if (connection->agreed_pro_version <= 95 || + is_sync_target_state(device->state.conn)) + set_bit(RS_DONE, &device->flags); + /* and also set RS_PROGRESS below */ + + /* Else: rather wait for explicit notification via receive_state, + * to avoid uuids-rotated-too-fast causing full resync + * in next handshake, in case the replication link breaks + * at the most unfortunate time... */ + } else if (!lazy_bitmap_update_due(device)) return; - } - - if (!get_ldev(mdev)) - return; /* no disk, no metadata, no bitmap to clear bits in */ - - nr_sectors = drbd_get_capacity(mdev->this_bdev); - esector = sector + (size >> 9) - 1; - - if (!expect(sector < nr_sectors)) - goto out; - if (!expect(esector < nr_sectors)) - esector = nr_sectors - 1; - lbnr = BM_SECT_TO_BIT(nr_sectors-1); - - /* we clear it (in sync). - * round up start sector, round down end sector. we make sure we only - * clear full, aligned, BM_BLOCK_SIZE (4K) blocks */ - if (unlikely(esector < BM_SECT_PER_BIT-1)) - goto out; - if (unlikely(esector == (nr_sectors-1))) - ebnr = lbnr; - else - ebnr = BM_SECT_TO_BIT(esector - (BM_SECT_PER_BIT-1)); - sbnr = BM_SECT_TO_BIT(sector + BM_SECT_PER_BIT-1); - - if (sbnr > ebnr) - goto out; + drbd_device_post_work(device, RS_PROGRESS); +} +static int update_sync_bits(struct drbd_device *device, + unsigned long sbnr, unsigned long ebnr, + enum update_sync_bits_mode mode) +{ /* - * ok, (capacity & 7) != 0 sometimes, but who cares... - * we count rs_{total,left} in bits, not sectors. + * We keep a count of set bits per resync-extent in the ->rs_left + * caching member, so we need to loop and work within the resync extent + * alignment. Typically this loop will execute exactly once. */ - count = drbd_bm_clear_bits(mdev, sbnr, ebnr); + unsigned long flags; + unsigned long count = 0; + unsigned int cleared = 0; + while (sbnr <= ebnr) { + /* set temporary boundary bit number to last bit number within + * the resync extent of the current start bit number, + * but cap at provided end bit number */ + unsigned long tbnr = min(ebnr, sbnr | BM_BLOCKS_PER_BM_EXT_MASK); + unsigned long c; + + if (mode == RECORD_RS_FAILED) + /* Only called from drbd_rs_failed_io(), bits + * supposedly still set. Recount, maybe some + * of the bits have been successfully cleared + * by application IO meanwhile. + */ + c = drbd_bm_count_bits(device, sbnr, tbnr); + else if (mode == SET_IN_SYNC) + c = drbd_bm_clear_bits(device, sbnr, tbnr); + else /* if (mode == SET_OUT_OF_SYNC) */ + c = drbd_bm_set_bits(device, sbnr, tbnr); + + if (c) { + spin_lock_irqsave(&device->al_lock, flags); + cleared += update_rs_extent(device, BM_BIT_TO_EXT(sbnr), c, mode); + spin_unlock_irqrestore(&device->al_lock, flags); + count += c; + } + sbnr = tbnr + 1; + } if (count) { - drbd_advance_rs_marks(mdev, drbd_bm_total_weight(mdev)); - spin_lock_irqsave(&mdev->al_lock, flags); - drbd_try_clear_on_disk_bm(mdev, sector, count, true); - spin_unlock_irqrestore(&mdev->al_lock, flags); - - /* just wake_up unconditional now, various lc_chaged(), - * lc_put() in drbd_try_clear_on_disk_bm(). */ - wake_up = 1; + if (mode == SET_IN_SYNC) { + unsigned long still_to_go = drbd_bm_total_weight(device); + bool rs_is_done = (still_to_go <= device->rs_failed); + drbd_advance_rs_marks(first_peer_device(device), still_to_go); + if (cleared || rs_is_done) + maybe_schedule_on_disk_bitmap_update(device, rs_is_done); + } else if (mode == RECORD_RS_FAILED) + device->rs_failed += count; + wake_up(&device->al_wait); } -out: - put_ldev(mdev); - if (wake_up) - wake_up(&mdev->al_wait); + return count; } -/* - * this is intended to set one request worth of data out of sync. - * affects at least 1 bit, - * and at most 1+DRBD_MAX_BIO_SIZE/BM_BLOCK_SIZE bits. +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 bits of the affected + * one or more _aligned_ BM_BLOCK_SIZE blocks. + * + * called by worker on C_SYNC_TARGET and receiver on SyncSource. * - * called by tl_clear and drbd_send_dblock (==drbd_make_request). - * so this can be _any_ process. */ -int __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector, int size, - const char *file, const unsigned int line) +int __drbd_change_sync(struct drbd_peer_device *peer_device, sector_t sector, int size, + enum update_sync_bits_mode mode) { - unsigned long sbnr, ebnr, flags; + /* Is called from worker and receiver context _only_ */ + struct drbd_device *device = peer_device->device; + unsigned long sbnr, ebnr, lbnr; + unsigned long count = 0; sector_t esector, nr_sectors; - unsigned int enr, count = 0; - struct lc_element *e; - /* this should be an empty REQ_FLUSH */ - if (size == 0) + /* This would be an empty REQ_PREFLUSH, be silent. */ + if ((mode == SET_OUT_OF_SYNC) && size == 0) return 0; - if (size < 0 || !IS_ALIGNED(size, 512) || size > DRBD_MAX_BIO_SIZE) { - dev_err(DEV, "sector: %llus, size: %d\n", - (unsigned long long)sector, 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); return 0; } - if (!get_ldev(mdev)) - return 0; /* no disk, no metadata, no bitmap to set bits in */ + if (!get_ldev(device)) + return 0; /* no disk, no metadata, no bitmap to manipulate bits in */ - nr_sectors = drbd_get_capacity(mdev->this_bdev); + nr_sectors = get_capacity(device->vdisk); esector = sector + (size >> 9) - 1; - if (!expect(sector < nr_sectors)) + if (!expect(device, sector < nr_sectors)) goto out; - if (!expect(esector < nr_sectors)) + if (!expect(device, esector < nr_sectors)) esector = nr_sectors - 1; - /* we set it out of sync, - * we do not need to round anything here */ - sbnr = BM_SECT_TO_BIT(sector); - ebnr = BM_SECT_TO_BIT(esector); - - /* ok, (capacity & 7) != 0 sometimes, but who cares... - * we count rs_{total,left} in bits, not sectors. */ - spin_lock_irqsave(&mdev->al_lock, flags); - count = drbd_bm_set_bits(mdev, sbnr, ebnr); + lbnr = BM_SECT_TO_BIT(nr_sectors-1); - enr = BM_SECT_TO_EXT(sector); - e = lc_find(mdev->resync, enr); - if (e) - lc_entry(e, struct bm_extent, lce)->rs_left += count; - spin_unlock_irqrestore(&mdev->al_lock, flags); + if (mode == SET_IN_SYNC) { + /* Round up start sector, round down end sector. We make sure + * we only clear full, aligned, BM_BLOCK_SIZE blocks. */ + if (unlikely(esector < BM_SECT_PER_BIT-1)) + goto out; + if (unlikely(esector == (nr_sectors-1))) + ebnr = lbnr; + else + ebnr = BM_SECT_TO_BIT(esector - (BM_SECT_PER_BIT-1)); + sbnr = BM_SECT_TO_BIT(sector + BM_SECT_PER_BIT-1); + } else { + /* We set it out of sync, or record resync failure. + * Should not round anything here. */ + sbnr = BM_SECT_TO_BIT(sector); + ebnr = BM_SECT_TO_BIT(esector); + } + count = update_sync_bits(device, sbnr, ebnr, mode); out: - put_ldev(mdev); - + put_ldev(device); return count; } static -struct bm_extent *_bme_get(struct drbd_conf *mdev, unsigned int enr) +struct bm_extent *_bme_get(struct drbd_device *device, unsigned int enr) { struct lc_element *e; struct bm_extent *bm_ext; int wakeup = 0; unsigned long rs_flags; - spin_lock_irq(&mdev->al_lock); - if (mdev->resync_locked > mdev->resync->nr_elements/2) { - spin_unlock_irq(&mdev->al_lock); + spin_lock_irq(&device->al_lock); + if (device->resync_locked > device->resync->nr_elements/2) { + spin_unlock_irq(&device->al_lock); return NULL; } - e = lc_get(mdev->resync, enr); + e = lc_get(device->resync, enr); bm_ext = e ? lc_entry(e, struct bm_extent, lce) : NULL; if (bm_ext) { if (bm_ext->lce.lc_number != enr) { - bm_ext->rs_left = drbd_bm_e_weight(mdev, enr); + bm_ext->rs_left = drbd_bm_e_weight(device, enr); bm_ext->rs_failed = 0; - lc_committed(mdev->resync); + lc_committed(device->resync); wakeup = 1; } if (bm_ext->lce.refcnt == 1) - mdev->resync_locked++; + device->resync_locked++; set_bit(BME_NO_WRITES, &bm_ext->flags); } - rs_flags = mdev->resync->flags; - spin_unlock_irq(&mdev->al_lock); + rs_flags = device->resync->flags; + spin_unlock_irq(&device->al_lock); if (wakeup) - wake_up(&mdev->al_wait); + wake_up(&device->al_wait); if (!bm_ext) { if (rs_flags & LC_STARVING) - dev_warn(DEV, "Have to wait for element" + drbd_warn(device, "Have to wait for element" " (resync LRU too small?)\n"); BUG_ON(rs_flags & LC_LOCKED); } @@ -974,61 +941,62 @@ struct bm_extent *_bme_get(struct drbd_conf *mdev, unsigned int enr) return bm_ext; } -static int _is_in_al(struct drbd_conf *mdev, unsigned int enr) +static int _is_in_al(struct drbd_device *device, unsigned int enr) { int rv; - spin_lock_irq(&mdev->al_lock); - rv = lc_is_used(mdev->act_log, enr); - spin_unlock_irq(&mdev->al_lock); + spin_lock_irq(&device->al_lock); + rv = lc_is_used(device->act_log, enr); + spin_unlock_irq(&device->al_lock); return rv; } /** * drbd_rs_begin_io() - Gets an extent in the resync LRU cache and sets it to BME_LOCKED - * @mdev: DRBD device. + * @device: DRBD device. * @sector: The sector number. * - * This functions sleeps on al_wait. Returns 0 on success, -EINTR if interrupted. + * This functions sleeps on al_wait. + * + * Returns: %0 on success, -EINTR if interrupted. */ -int drbd_rs_begin_io(struct drbd_conf *mdev, sector_t sector) +int drbd_rs_begin_io(struct drbd_device *device, sector_t sector) { unsigned int enr = BM_SECT_TO_EXT(sector); struct bm_extent *bm_ext; int i, sig; - int sa = 200; /* Step aside 200 times, then grab the extent and let app-IO wait. - 200 times -> 20 seconds. */ + bool sa; retry: - sig = wait_event_interruptible(mdev->al_wait, - (bm_ext = _bme_get(mdev, enr))); + sig = wait_event_interruptible(device->al_wait, + (bm_ext = _bme_get(device, enr))); if (sig) return -EINTR; if (test_bit(BME_LOCKED, &bm_ext->flags)) return 0; + /* step aside only while we are above c-min-rate; unless disabled. */ + sa = drbd_rs_c_min_rate_throttle(device); + for (i = 0; i < AL_EXT_PER_BM_SECT; i++) { - sig = wait_event_interruptible(mdev->al_wait, - !_is_in_al(mdev, enr * AL_EXT_PER_BM_SECT + i) || - test_bit(BME_PRIORITY, &bm_ext->flags)); + sig = wait_event_interruptible(device->al_wait, + !_is_in_al(device, enr * AL_EXT_PER_BM_SECT + i) || + (sa && test_bit(BME_PRIORITY, &bm_ext->flags))); - if (sig || (test_bit(BME_PRIORITY, &bm_ext->flags) && sa)) { - spin_lock_irq(&mdev->al_lock); - if (lc_put(mdev->resync, &bm_ext->lce) == 0) { + if (sig || (sa && test_bit(BME_PRIORITY, &bm_ext->flags))) { + spin_lock_irq(&device->al_lock); + if (lc_put(device->resync, &bm_ext->lce) == 0) { bm_ext->flags = 0; /* clears BME_NO_WRITES and eventually BME_PRIORITY */ - mdev->resync_locked--; - wake_up(&mdev->al_wait); + device->resync_locked--; + wake_up(&device->al_wait); } - spin_unlock_irq(&mdev->al_lock); + spin_unlock_irq(&device->al_lock); if (sig) return -EINTR; if (schedule_timeout_interruptible(HZ/10)) return -EINTR; - if (sa && --sa == 0) - dev_warn(DEV,"drbd_rs_begin_io() stepped aside for 20sec." - "Resync stalled?\n"); goto retry; } } @@ -1038,23 +1006,35 @@ retry: /** * drbd_try_rs_begin_io() - Gets an extent in the resync LRU cache, does not sleep - * @mdev: DRBD device. + * @peer_device: DRBD device. * @sector: The sector number. * * Gets an extent in the resync LRU cache, sets it to BME_NO_WRITES, then - * tries to set it to BME_LOCKED. Returns 0 upon success, and -EAGAIN + * tries to set it to BME_LOCKED. + * + * Returns: %0 upon success, and -EAGAIN * if there is still application IO going on in this area. */ -int drbd_try_rs_begin_io(struct drbd_conf *mdev, sector_t sector) +int drbd_try_rs_begin_io(struct drbd_peer_device *peer_device, sector_t sector) { + struct drbd_device *device = peer_device->device; unsigned int enr = BM_SECT_TO_EXT(sector); const unsigned int al_enr = enr*AL_EXT_PER_BM_SECT; struct lc_element *e; struct bm_extent *bm_ext; int i; + bool throttle = drbd_rs_should_slow_down(peer_device, sector, true); - spin_lock_irq(&mdev->al_lock); - if (mdev->resync_wenr != LC_FREE && mdev->resync_wenr != enr) { + /* If we need to throttle, a half-locked (only marked BME_NO_WRITES, + * not yet BME_LOCKED) extent needs to be kicked out explicitly if we + * need to throttle. There is at most one such half-locked extent, + * which is remembered in resync_wenr. */ + + if (throttle && device->resync_wenr != enr) + return -EAGAIN; + + spin_lock_irq(&device->al_lock); + if (device->resync_wenr != LC_FREE && device->resync_wenr != enr) { /* in case you have very heavy scattered io, it may * stall the syncer undefined if we give up the ref count * when we try again and requeue. @@ -1068,248 +1048,196 @@ int drbd_try_rs_begin_io(struct drbd_conf *mdev, sector_t sector) * the lc_put here... * we also have to wake_up */ - e = lc_find(mdev->resync, mdev->resync_wenr); + e = lc_find(device->resync, device->resync_wenr); bm_ext = e ? lc_entry(e, struct bm_extent, lce) : NULL; if (bm_ext) { - D_ASSERT(!test_bit(BME_LOCKED, &bm_ext->flags)); - D_ASSERT(test_bit(BME_NO_WRITES, &bm_ext->flags)); + D_ASSERT(device, !test_bit(BME_LOCKED, &bm_ext->flags)); + D_ASSERT(device, test_bit(BME_NO_WRITES, &bm_ext->flags)); clear_bit(BME_NO_WRITES, &bm_ext->flags); - mdev->resync_wenr = LC_FREE; - if (lc_put(mdev->resync, &bm_ext->lce) == 0) - mdev->resync_locked--; - wake_up(&mdev->al_wait); + device->resync_wenr = LC_FREE; + if (lc_put(device->resync, &bm_ext->lce) == 0) { + bm_ext->flags = 0; + device->resync_locked--; + } + wake_up(&device->al_wait); } else { - dev_alert(DEV, "LOGIC BUG\n"); + drbd_alert(device, "LOGIC BUG\n"); } } /* TRY. */ - e = lc_try_get(mdev->resync, enr); + e = lc_try_get(device->resync, enr); bm_ext = e ? lc_entry(e, struct bm_extent, lce) : NULL; if (bm_ext) { if (test_bit(BME_LOCKED, &bm_ext->flags)) goto proceed; if (!test_and_set_bit(BME_NO_WRITES, &bm_ext->flags)) { - mdev->resync_locked++; + device->resync_locked++; } else { /* we did set the BME_NO_WRITES, * but then could not set BME_LOCKED, * so we tried again. * drop the extra reference. */ bm_ext->lce.refcnt--; - D_ASSERT(bm_ext->lce.refcnt > 0); + D_ASSERT(device, bm_ext->lce.refcnt > 0); } goto check_al; } else { /* do we rather want to try later? */ - if (mdev->resync_locked > mdev->resync->nr_elements-3) + if (device->resync_locked > device->resync->nr_elements-3) goto try_again; /* Do or do not. There is no try. -- Yoda */ - e = lc_get(mdev->resync, enr); + e = lc_get(device->resync, enr); bm_ext = e ? lc_entry(e, struct bm_extent, lce) : NULL; if (!bm_ext) { - const unsigned long rs_flags = mdev->resync->flags; + const unsigned long rs_flags = device->resync->flags; if (rs_flags & LC_STARVING) - dev_warn(DEV, "Have to wait for element" + drbd_warn(device, "Have to wait for element" " (resync LRU too small?)\n"); BUG_ON(rs_flags & LC_LOCKED); goto try_again; } if (bm_ext->lce.lc_number != enr) { - bm_ext->rs_left = drbd_bm_e_weight(mdev, enr); + bm_ext->rs_left = drbd_bm_e_weight(device, enr); bm_ext->rs_failed = 0; - lc_committed(mdev->resync); - wake_up(&mdev->al_wait); - D_ASSERT(test_bit(BME_LOCKED, &bm_ext->flags) == 0); + lc_committed(device->resync); + wake_up(&device->al_wait); + D_ASSERT(device, test_bit(BME_LOCKED, &bm_ext->flags) == 0); } set_bit(BME_NO_WRITES, &bm_ext->flags); - D_ASSERT(bm_ext->lce.refcnt == 1); - mdev->resync_locked++; + D_ASSERT(device, bm_ext->lce.refcnt == 1); + device->resync_locked++; goto check_al; } check_al: for (i = 0; i < AL_EXT_PER_BM_SECT; i++) { - if (lc_is_used(mdev->act_log, al_enr+i)) + if (lc_is_used(device->act_log, al_enr+i)) goto try_again; } set_bit(BME_LOCKED, &bm_ext->flags); proceed: - mdev->resync_wenr = LC_FREE; - spin_unlock_irq(&mdev->al_lock); + device->resync_wenr = LC_FREE; + spin_unlock_irq(&device->al_lock); return 0; try_again: - if (bm_ext) - mdev->resync_wenr = enr; - spin_unlock_irq(&mdev->al_lock); + if (bm_ext) { + if (throttle) { + D_ASSERT(device, !test_bit(BME_LOCKED, &bm_ext->flags)); + D_ASSERT(device, test_bit(BME_NO_WRITES, &bm_ext->flags)); + clear_bit(BME_NO_WRITES, &bm_ext->flags); + device->resync_wenr = LC_FREE; + if (lc_put(device->resync, &bm_ext->lce) == 0) { + bm_ext->flags = 0; + device->resync_locked--; + } + wake_up(&device->al_wait); + } else + device->resync_wenr = enr; + } + spin_unlock_irq(&device->al_lock); return -EAGAIN; } -void drbd_rs_complete_io(struct drbd_conf *mdev, sector_t sector) +void drbd_rs_complete_io(struct drbd_device *device, sector_t sector) { unsigned int enr = BM_SECT_TO_EXT(sector); struct lc_element *e; struct bm_extent *bm_ext; unsigned long flags; - spin_lock_irqsave(&mdev->al_lock, flags); - e = lc_find(mdev->resync, enr); + spin_lock_irqsave(&device->al_lock, flags); + e = lc_find(device->resync, enr); bm_ext = e ? lc_entry(e, struct bm_extent, lce) : NULL; if (!bm_ext) { - spin_unlock_irqrestore(&mdev->al_lock, flags); - if (__ratelimit(&drbd_ratelimit_state)) - dev_err(DEV, "drbd_rs_complete_io() called, but extent not found\n"); + spin_unlock_irqrestore(&device->al_lock, flags); + if (drbd_ratelimit()) + drbd_err(device, "drbd_rs_complete_io() called, but extent not found\n"); return; } if (bm_ext->lce.refcnt == 0) { - spin_unlock_irqrestore(&mdev->al_lock, flags); - dev_err(DEV, "drbd_rs_complete_io(,%llu [=%u]) called, " + spin_unlock_irqrestore(&device->al_lock, flags); + drbd_err(device, "drbd_rs_complete_io(,%llu [=%u]) called, " "but refcnt is 0!?\n", (unsigned long long)sector, enr); return; } - if (lc_put(mdev->resync, &bm_ext->lce) == 0) { + if (lc_put(device->resync, &bm_ext->lce) == 0) { bm_ext->flags = 0; /* clear BME_LOCKED, BME_NO_WRITES and BME_PRIORITY */ - mdev->resync_locked--; - wake_up(&mdev->al_wait); + device->resync_locked--; + wake_up(&device->al_wait); } - spin_unlock_irqrestore(&mdev->al_lock, flags); + spin_unlock_irqrestore(&device->al_lock, flags); } /** * drbd_rs_cancel_all() - Removes all extents from the resync LRU (even BME_LOCKED) - * @mdev: DRBD device. + * @device: DRBD device. */ -void drbd_rs_cancel_all(struct drbd_conf *mdev) +void drbd_rs_cancel_all(struct drbd_device *device) { - spin_lock_irq(&mdev->al_lock); + spin_lock_irq(&device->al_lock); - if (get_ldev_if_state(mdev, D_FAILED)) { /* Makes sure ->resync is there. */ - lc_reset(mdev->resync); - put_ldev(mdev); + if (get_ldev_if_state(device, D_FAILED)) { /* Makes sure ->resync is there. */ + lc_reset(device->resync); + put_ldev(device); } - mdev->resync_locked = 0; - mdev->resync_wenr = LC_FREE; - spin_unlock_irq(&mdev->al_lock); - wake_up(&mdev->al_wait); + device->resync_locked = 0; + device->resync_wenr = LC_FREE; + spin_unlock_irq(&device->al_lock); + wake_up(&device->al_wait); } /** * drbd_rs_del_all() - Gracefully remove all extents from the resync LRU - * @mdev: DRBD device. + * @device: DRBD device. * - * Returns 0 upon success, -EAGAIN if at least one reference count was + * Returns: %0 upon success, -EAGAIN if at least one reference count was * not zero. */ -int drbd_rs_del_all(struct drbd_conf *mdev) +int drbd_rs_del_all(struct drbd_device *device) { struct lc_element *e; struct bm_extent *bm_ext; int i; - spin_lock_irq(&mdev->al_lock); + spin_lock_irq(&device->al_lock); - if (get_ldev_if_state(mdev, D_FAILED)) { + if (get_ldev_if_state(device, D_FAILED)) { /* ok, ->resync is there. */ - for (i = 0; i < mdev->resync->nr_elements; i++) { - e = lc_element_by_index(mdev->resync, i); + for (i = 0; i < device->resync->nr_elements; i++) { + e = lc_element_by_index(device->resync, i); bm_ext = lc_entry(e, struct bm_extent, lce); if (bm_ext->lce.lc_number == LC_FREE) continue; - if (bm_ext->lce.lc_number == mdev->resync_wenr) { - dev_info(DEV, "dropping %u in drbd_rs_del_all, apparently" + if (bm_ext->lce.lc_number == device->resync_wenr) { + drbd_info(device, "dropping %u in drbd_rs_del_all, apparently" " got 'synced' by application io\n", - mdev->resync_wenr); - D_ASSERT(!test_bit(BME_LOCKED, &bm_ext->flags)); - D_ASSERT(test_bit(BME_NO_WRITES, &bm_ext->flags)); + device->resync_wenr); + D_ASSERT(device, !test_bit(BME_LOCKED, &bm_ext->flags)); + D_ASSERT(device, test_bit(BME_NO_WRITES, &bm_ext->flags)); clear_bit(BME_NO_WRITES, &bm_ext->flags); - mdev->resync_wenr = LC_FREE; - lc_put(mdev->resync, &bm_ext->lce); + device->resync_wenr = LC_FREE; + lc_put(device->resync, &bm_ext->lce); } if (bm_ext->lce.refcnt != 0) { - dev_info(DEV, "Retrying drbd_rs_del_all() later. " + drbd_info(device, "Retrying drbd_rs_del_all() later. " "refcnt=%d\n", bm_ext->lce.refcnt); - put_ldev(mdev); - spin_unlock_irq(&mdev->al_lock); + put_ldev(device); + spin_unlock_irq(&device->al_lock); return -EAGAIN; } - D_ASSERT(!test_bit(BME_LOCKED, &bm_ext->flags)); - D_ASSERT(!test_bit(BME_NO_WRITES, &bm_ext->flags)); - lc_del(mdev->resync, &bm_ext->lce); + D_ASSERT(device, !test_bit(BME_LOCKED, &bm_ext->flags)); + D_ASSERT(device, !test_bit(BME_NO_WRITES, &bm_ext->flags)); + lc_del(device->resync, &bm_ext->lce); } - D_ASSERT(mdev->resync->used == 0); - put_ldev(mdev); + D_ASSERT(device, device->resync->used == 0); + put_ldev(device); } - spin_unlock_irq(&mdev->al_lock); - wake_up(&mdev->al_wait); + spin_unlock_irq(&device->al_lock); + wake_up(&device->al_wait); return 0; } - -/** - * drbd_rs_failed_io() - Record information on a failure to resync the specified blocks - * @mdev: DRBD device. - * @sector: The sector number. - * @size: Size of failed IO operation, in byte. - */ -void drbd_rs_failed_io(struct drbd_conf *mdev, sector_t sector, int size) -{ - /* Is called from worker and receiver context _only_ */ - unsigned long sbnr, ebnr, lbnr; - unsigned long count; - sector_t esector, nr_sectors; - int wake_up = 0; - - if (size <= 0 || !IS_ALIGNED(size, 512) || size > DRBD_MAX_BIO_SIZE) { - dev_err(DEV, "drbd_rs_failed_io: sector=%llus size=%d nonsense!\n", - (unsigned long long)sector, size); - return; - } - nr_sectors = drbd_get_capacity(mdev->this_bdev); - esector = sector + (size >> 9) - 1; - - if (!expect(sector < nr_sectors)) - return; - if (!expect(esector < nr_sectors)) - esector = nr_sectors - 1; - - lbnr = BM_SECT_TO_BIT(nr_sectors-1); - - /* - * round up start sector, round down end sector. we make sure we only - * handle full, aligned, BM_BLOCK_SIZE (4K) blocks */ - if (unlikely(esector < BM_SECT_PER_BIT-1)) - return; - if (unlikely(esector == (nr_sectors-1))) - ebnr = lbnr; - else - ebnr = BM_SECT_TO_BIT(esector - (BM_SECT_PER_BIT-1)); - sbnr = BM_SECT_TO_BIT(sector + BM_SECT_PER_BIT-1); - - if (sbnr > ebnr) - return; - - /* - * ok, (capacity & 7) != 0 sometimes, but who cares... - * we count rs_{total,left} in bits, not sectors. - */ - spin_lock_irq(&mdev->al_lock); - count = drbd_bm_count_bits(mdev, sbnr, ebnr); - if (count) { - mdev->rs_failed += count; - - if (get_ldev(mdev)) { - drbd_try_clear_on_disk_bm(mdev, sector, count, false); - put_ldev(mdev); - } - - /* just wake_up unconditional now, various lc_chaged(), - * lc_put() in drbd_try_clear_on_disk_bm(). */ - wake_up = 1; - } - spin_unlock_irq(&mdev->al_lock); - if (wake_up) - wake_up(&mdev->al_wait); -} |