From 0cf9d27e38447efe5e5edce155a66a782a5aac4a Mon Sep 17 00:00:00 2001 From: Andreas Gruenbacher Date: Tue, 7 Dec 2010 10:43:29 +0100 Subject: drbd: Get rid of unnecessary macros (2) The FAULT_ACTIVE macro just wraps the drbd_insert_fault macro for no apparent reason. Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_bitmap.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/block/drbd/drbd_bitmap.c') diff --git a/drivers/block/drbd/drbd_bitmap.c b/drivers/block/drbd/drbd_bitmap.c index 0645ca829a94..5dafbabe9617 100644 --- a/drivers/block/drbd/drbd_bitmap.c +++ b/drivers/block/drbd/drbd_bitmap.c @@ -502,7 +502,7 @@ int drbd_bm_resize(struct drbd_conf *mdev, sector_t capacity, int set_new_bits) D_ASSERT(b->bm_pages != NULL); npages = b->bm_pages; } else { - if (FAULT_ACTIVE(mdev, DRBD_FAULT_BM_ALLOC)) + if (drbd_insert_fault(mdev, DRBD_FAULT_BM_ALLOC)) npages = NULL; else npages = bm_realloc_pages(b, want); @@ -768,7 +768,7 @@ static void bm_page_io_async(struct drbd_conf *mdev, struct drbd_bitmap *b, int bio->bi_private = b; bio->bi_end_io = bm_async_io_complete; - if (FAULT_ACTIVE(mdev, (rw & WRITE) ? DRBD_FAULT_MD_WR : DRBD_FAULT_MD_RD)) { + if (drbd_insert_fault(mdev, (rw & WRITE) ? DRBD_FAULT_MD_WR : DRBD_FAULT_MD_RD)) { bio->bi_rw |= rw; bio_endio(bio, -EIO); } else { -- cgit v1.2.3 From 81e84650c200de0695372461964dd960365696db Mon Sep 17 00:00:00 2001 From: Andreas Gruenbacher Date: Thu, 9 Dec 2010 15:03:57 +0100 Subject: drbd: Use the standard bool, true, and false keywords Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_actlog.c | 8 +-- drivers/block/drbd/drbd_bitmap.c | 4 +- drivers/block/drbd/drbd_int.h | 13 +--- drivers/block/drbd/drbd_main.c | 30 ++++----- drivers/block/drbd/drbd_nl.c | 4 +- drivers/block/drbd/drbd_receiver.c | 128 ++++++++++++++++++------------------- drivers/block/drbd/drbd_req.c | 4 +- drivers/block/drbd/drbd_worker.c | 4 +- 8 files changed, 94 insertions(+), 101 deletions(-) (limited to 'drivers/block/drbd/drbd_bitmap.c') diff --git a/drivers/block/drbd/drbd_actlog.c b/drivers/block/drbd/drbd_actlog.c index 2e8a95ce79b1..e3f0f4d31d75 100644 --- a/drivers/block/drbd/drbd_actlog.c +++ b/drivers/block/drbd/drbd_actlog.c @@ -338,7 +338,7 @@ w_al_write_transaction(struct drbd_conf *mdev, struct drbd_work *w, int unused) + mdev->ldev->md.al_offset + mdev->al_tr_pos; if (!drbd_md_sync_page_io(mdev, mdev->ldev, sector, WRITE)) - drbd_chk_io_error(mdev, 1, TRUE); + drbd_chk_io_error(mdev, 1, true); if (++mdev->al_tr_pos > div_ceil(mdev->act_log->nr_elements, AL_EXTENTS_PT)) @@ -528,7 +528,7 @@ static void atodb_endio(struct bio *bio, int error) if (!error && !uptodate) error = -EIO; - drbd_chk_io_error(mdev, error, TRUE); + drbd_chk_io_error(mdev, error, true); if (error && wc->error == 0) wc->error = error; @@ -991,7 +991,7 @@ void __drbd_set_in_sync(struct drbd_conf *mdev, sector_t sector, int size, if (count && get_ldev(mdev)) { 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); + 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(), @@ -1441,7 +1441,7 @@ void drbd_rs_failed_io(struct drbd_conf *mdev, sector_t sector, int size) mdev->rs_failed += count; if (get_ldev(mdev)) { - drbd_try_clear_on_disk_bm(mdev, sector, count, FALSE); + drbd_try_clear_on_disk_bm(mdev, sector, count, false); put_ldev(mdev); } diff --git a/drivers/block/drbd/drbd_bitmap.c b/drivers/block/drbd/drbd_bitmap.c index 5dafbabe9617..9390e9526786 100644 --- a/drivers/block/drbd/drbd_bitmap.c +++ b/drivers/block/drbd/drbd_bitmap.c @@ -844,7 +844,7 @@ static int bm_rw(struct drbd_conf *mdev, int rw) __must_hold(local) if (test_bit(BM_MD_IO_ERROR, &b->bm_flags)) { dev_alert(DEV, "we had at least one MD IO ERROR during bitmap IO\n"); - drbd_chk_io_error(mdev, 1, TRUE); + drbd_chk_io_error(mdev, 1, true); err = -EIO; } @@ -916,7 +916,7 @@ int drbd_bm_write_sect(struct drbd_conf *mdev, unsigned long enr) __must_hold(lo dev_err(DEV, "IO ERROR writing bitmap sector %lu " "(meta-disk sector %llus)\n", enr, (unsigned long long)on_disk_sector); - drbd_chk_io_error(mdev, 1, TRUE); + drbd_chk_io_error(mdev, 1, true); for (i = 0; i < AL_EXT_PER_BM_SECT; i++) drbd_bm_ALe_set_all(mdev, enr*AL_EXT_PER_BM_SECT+i); } diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h index 749607a494e6..0855934e19d0 100644 --- a/drivers/block/drbd/drbd_int.h +++ b/drivers/block/drbd/drbd_int.h @@ -72,13 +72,6 @@ extern int fault_devs; extern char usermode_helper[]; -#ifndef TRUE -#define TRUE 1 -#endif -#ifndef FALSE -#define FALSE 0 -#endif - /* I don't remember why XCPU ... * This is used to wake the asender, * and to interrupt sending the sending task @@ -2002,17 +1995,17 @@ static inline int drbd_send_ping_ack(struct drbd_conf *mdev) static inline void drbd_thread_stop(struct drbd_thread *thi) { - _drbd_thread_stop(thi, FALSE, TRUE); + _drbd_thread_stop(thi, false, true); } static inline void drbd_thread_stop_nowait(struct drbd_thread *thi) { - _drbd_thread_stop(thi, FALSE, FALSE); + _drbd_thread_stop(thi, false, false); } static inline void drbd_thread_restart_nowait(struct drbd_thread *thi) { - _drbd_thread_stop(thi, TRUE, FALSE); + _drbd_thread_stop(thi, true, false); } /* counts how many answer packets packets we expect from our peer, diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index a101dceb2d49..7eb447d20cce 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -456,7 +456,7 @@ void tl_restart(struct drbd_conf *mdev, enum drbd_req_event what) } /** - * cl_wide_st_chg() - TRUE if the state change is a cluster wide one + * cl_wide_st_chg() - true if the state change is a cluster wide one * @mdev: DRBD device. * @os: old (current) state. * @ns: new (wanted) state. @@ -1623,7 +1623,7 @@ int drbd_thread_start(struct drbd_thread *thi) if (!try_module_get(THIS_MODULE)) { dev_err(DEV, "Failed to get module reference in drbd_thread_start\n"); spin_unlock_irqrestore(&thi->t_lock, flags); - return FALSE; + return false; } init_completion(&thi->stop); @@ -1640,7 +1640,7 @@ int drbd_thread_start(struct drbd_thread *thi) dev_err(DEV, "Couldn't start thread\n"); module_put(THIS_MODULE); - return FALSE; + return false; } spin_lock_irqsave(&thi->t_lock, flags); thi->task = nt; @@ -1660,7 +1660,7 @@ int drbd_thread_start(struct drbd_thread *thi) break; } - return TRUE; + return true; } @@ -1758,8 +1758,8 @@ int _drbd_send_cmd(struct drbd_conf *mdev, struct socket *sock, { int sent, ok; - ERR_IF(!h) return FALSE; - ERR_IF(!size) return FALSE; + ERR_IF(!h) return false; + ERR_IF(!size) return false; h->magic = BE_DRBD_MAGIC; h->command = cpu_to_be16(cmd); @@ -2196,14 +2196,14 @@ int _drbd_send_bitmap(struct drbd_conf *mdev) struct p_header80 *p; int ret; - ERR_IF(!mdev->bitmap) return FALSE; + ERR_IF(!mdev->bitmap) return false; /* maybe we should use some per thread scratch page, * and allocate that during initial device creation? */ p = (struct p_header80 *) __get_free_page(GFP_NOIO); if (!p) { dev_err(DEV, "failed to allocate one page buffer in %s\n", __func__); - return FALSE; + return false; } if (get_ldev(mdev)) { @@ -2256,7 +2256,7 @@ int drbd_send_b_ack(struct drbd_conf *mdev, u32 barrier_nr, u32 set_size) p.set_size = cpu_to_be32(set_size); if (mdev->state.conn < C_CONNECTED) - return FALSE; + return false; ok = drbd_send_cmd(mdev, USE_META_SOCKET, P_BARRIER_ACK, (struct p_header80 *)&p, sizeof(p)); return ok; @@ -2284,7 +2284,7 @@ static int _drbd_send_ack(struct drbd_conf *mdev, enum drbd_packets cmd, p.seq_num = cpu_to_be32(atomic_add_return(1, &mdev->packet_seq)); if (!mdev->meta.socket || mdev->state.conn < C_CONNECTED) - return FALSE; + return false; ok = drbd_send_cmd(mdev, USE_META_SOCKET, cmd, (struct p_header80 *)&p, sizeof(p)); return ok; @@ -2390,8 +2390,8 @@ int drbd_send_ov_request(struct drbd_conf *mdev, sector_t sector, int size) } /* called on sndtimeo - * returns FALSE if we should retry, - * TRUE if we think connection is dead + * returns false if we should retry, + * true if we think connection is dead */ static int we_should_drop_the_connection(struct drbd_conf *mdev, struct socket *sock) { @@ -2404,7 +2404,7 @@ static int we_should_drop_the_connection(struct drbd_conf *mdev, struct socket * || mdev->state.conn < C_CONNECTED; if (drop_it) - return TRUE; + return true; drop_it = !--mdev->ko_count; if (!drop_it) { @@ -3283,7 +3283,7 @@ struct drbd_conf *drbd_new_device(unsigned int minor) goto out_no_disk; mdev->vdisk = disk; - set_disk_ro(disk, TRUE); + set_disk_ro(disk, true); disk->queue = q; disk->major = DRBD_MAJOR; @@ -3560,7 +3560,7 @@ void drbd_md_sync(struct drbd_conf *mdev) if (!drbd_md_sync_page_io(mdev, mdev->ldev, sector, WRITE)) { /* this was a try anyways ... */ dev_err(DEV, "meta data update failed!\n"); - drbd_chk_io_error(mdev, 1, TRUE); + drbd_chk_io_error(mdev, 1, true); } /* Update mdev->ldev->md.la_size_sect, diff --git a/drivers/block/drbd/drbd_nl.c b/drivers/block/drbd/drbd_nl.c index 7a2faf6d48e2..9e94c348c42c 100644 --- a/drivers/block/drbd/drbd_nl.c +++ b/drivers/block/drbd/drbd_nl.c @@ -385,7 +385,7 @@ drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role, int force) wait_event(mdev->misc_wait, atomic_read(&mdev->ap_pending_cnt) == 0); if (new_role == R_SECONDARY) { - set_disk_ro(mdev->vdisk, TRUE); + set_disk_ro(mdev->vdisk, true); if (get_ldev(mdev)) { mdev->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; put_ldev(mdev); @@ -395,7 +395,7 @@ drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role, int force) mdev->net_conf->want_lose = 0; put_net_conf(mdev); } - set_disk_ro(mdev->vdisk, FALSE); + set_disk_ro(mdev->vdisk, false); if (get_ldev(mdev)) { if (((mdev->state.conn < C_CONNECTED || mdev->state.pdsk <= D_FAILED) diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c index 00dcb1172ca8..732aacb46a32 100644 --- a/drivers/block/drbd/drbd_receiver.c +++ b/drivers/block/drbd/drbd_receiver.c @@ -725,16 +725,16 @@ static int drbd_socket_okay(struct drbd_conf *mdev, struct socket **sock) char tb[4]; if (!*sock) - return FALSE; + return false; rr = drbd_recv_short(mdev, *sock, tb, 4, MSG_DONTWAIT | MSG_PEEK); if (rr > 0 || rr == -EAGAIN) { - return TRUE; + return true; } else { sock_release(*sock); *sock = NULL; - return FALSE; + return false; } } @@ -933,7 +933,7 @@ static int drbd_recv_header(struct drbd_conf *mdev, enum drbd_packets *cmd, unsi r = drbd_recv(mdev, h, sizeof(*h)); if (unlikely(r != sizeof(*h))) { dev_err(DEV, "short read expecting header on sock: r=%d\n", r); - return FALSE; + return false; } if (likely(h->h80.magic == BE_DRBD_MAGIC)) { @@ -947,11 +947,11 @@ static int drbd_recv_header(struct drbd_conf *mdev, enum drbd_packets *cmd, unsi be32_to_cpu(h->h80.magic), be16_to_cpu(h->h80.command), be16_to_cpu(h->h80.length)); - return FALSE; + return false; } mdev->last_received = jiffies; - return TRUE; + return true; } static void drbd_flush(struct drbd_conf *mdev) @@ -1160,7 +1160,7 @@ static int receive_Barrier(struct drbd_conf *mdev, enum drbd_packets cmd, unsign switch (mdev->write_ordering) { case WO_none: if (rv == FE_RECYCLED) - return TRUE; + return true; /* receiver context, in the writeout path of the other node. * avoid potential distributed deadlock */ @@ -1188,10 +1188,10 @@ static int receive_Barrier(struct drbd_conf *mdev, enum drbd_packets cmd, unsign D_ASSERT(atomic_read(&epoch->active) == 0); D_ASSERT(epoch->flags == 0); - return TRUE; + return true; default: dev_err(DEV, "Strangeness in mdev->write_ordering %d\n", mdev->write_ordering); - return FALSE; + return false; } epoch->flags = 0; @@ -1209,7 +1209,7 @@ static int receive_Barrier(struct drbd_conf *mdev, enum drbd_packets cmd, unsign } spin_unlock(&mdev->epoch_lock); - return TRUE; + return true; } /* used from receive_RSDataReply (recv_resync_read) @@ -1303,7 +1303,7 @@ static int drbd_drain_block(struct drbd_conf *mdev, int data_size) void *data; if (!data_size) - return TRUE; + return true; page = drbd_pp_alloc(mdev, 1, 1); @@ -1426,7 +1426,7 @@ static int recv_resync_read(struct drbd_conf *mdev, sector_t sector, int data_si atomic_add(data_size >> 9, &mdev->rs_sect_ev); if (drbd_submit_ee(mdev, e, WRITE, DRBD_FAULT_RS_WR) == 0) - return TRUE; + return true; /* drbd_submit_ee currently fails for one reason only: * not being able to allocate enough bios. @@ -1438,7 +1438,7 @@ static int recv_resync_read(struct drbd_conf *mdev, sector_t sector, int data_si drbd_free_ee(mdev, e); fail: put_ldev(mdev); - return FALSE; + return false; } static int receive_DataReply(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned int data_size) @@ -1455,7 +1455,7 @@ static int receive_DataReply(struct drbd_conf *mdev, enum drbd_packets cmd, unsi spin_unlock_irq(&mdev->req_lock); if (unlikely(!req)) { dev_err(DEV, "Got a corrupt block_id/sector pair(1).\n"); - return FALSE; + return false; } /* hlist_del(&req->colision) is done in _req_may_be_done, to avoid @@ -1655,7 +1655,7 @@ static int receive_Data(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned e = read_in_block(mdev, p->block_id, sector, data_size); if (!e) { put_ldev(mdev); - return FALSE; + return false; } e->w.cb = e_end_block; @@ -1774,7 +1774,7 @@ static int receive_Data(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned put_ldev(mdev); wake_asender(mdev); finish_wait(&mdev->misc_wait, &wait); - return TRUE; + return true; } if (signal_pending(current)) { @@ -1830,7 +1830,7 @@ static int receive_Data(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned } if (drbd_submit_ee(mdev, e, rw, DRBD_FAULT_DT_WR) == 0) - return TRUE; + return true; /* drbd_submit_ee currently fails for one reason only: * not being able to allocate enough bios. @@ -1848,7 +1848,7 @@ out_interrupted: * receive a barrier... atomic_inc(&mdev->epoch_size); */ put_ldev(mdev); drbd_free_ee(mdev, e); - return FALSE; + return false; } /* We may throttle resync, if the lower device seems to be busy, @@ -1934,12 +1934,12 @@ static int receive_DataRequest(struct drbd_conf *mdev, enum drbd_packets cmd, un if (size <= 0 || (size & 0x1ff) != 0 || size > DRBD_MAX_BIO_SIZE) { dev_err(DEV, "%s:%d: sector: %llus, size: %u\n", __FILE__, __LINE__, (unsigned long long)sector, size); - return FALSE; + return false; } if (sector + (size>>9) > capacity) { dev_err(DEV, "%s:%d: sector: %llus, size: %u\n", __FILE__, __LINE__, (unsigned long long)sector, size); - return FALSE; + return false; } if (!get_ldev_if_state(mdev, D_UP_TO_DATE)) { @@ -1976,7 +1976,7 @@ static int receive_DataRequest(struct drbd_conf *mdev, enum drbd_packets cmd, un e = drbd_alloc_ee(mdev, p->block_id, sector, size, GFP_NOIO); if (!e) { put_ldev(mdev); - return FALSE; + return false; } switch (cmd) { @@ -2089,7 +2089,7 @@ submit: spin_unlock_irq(&mdev->req_lock); if (drbd_submit_ee(mdev, e, READ, fault_type) == 0) - return TRUE; + return true; /* drbd_submit_ee currently fails for one reason only: * not being able to allocate enough bios. @@ -2102,7 +2102,7 @@ submit: out_free_e: put_ldev(mdev); drbd_free_ee(mdev, e); - return FALSE; + return false; } static int drbd_asb_recover_0p(struct drbd_conf *mdev) __must_hold(local) @@ -2690,7 +2690,7 @@ static int receive_protocol(struct drbd_conf *mdev, enum drbd_packets cmd, unsig unsigned char *my_alg = mdev->net_conf->integrity_alg; if (drbd_recv(mdev, p_integrity_alg, data_size) != data_size) - return FALSE; + return false; p_integrity_alg[SHARED_SECRET_MAX-1] = 0; if (strcmp(p_integrity_alg, my_alg)) { @@ -2701,11 +2701,11 @@ static int receive_protocol(struct drbd_conf *mdev, enum drbd_packets cmd, unsig my_alg[0] ? my_alg : (unsigned char *)""); } - return TRUE; + return true; disconnect: drbd_force_state(mdev, NS(conn, C_DISCONNECTING)); - return FALSE; + return false; } /* helper function @@ -2737,7 +2737,7 @@ struct crypto_hash *drbd_crypto_alloc_digest_safe(const struct drbd_conf *mdev, static int receive_SyncParam(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned int packet_size) { - int ok = TRUE; + int ok = true; struct p_rs_param_95 *p = &mdev->data.rbuf.rs_param_95; unsigned int header_size, data_size, exp_max_sz; struct crypto_hash *verify_tfm = NULL; @@ -2755,7 +2755,7 @@ static int receive_SyncParam(struct drbd_conf *mdev, enum drbd_packets cmd, unsi if (packet_size > exp_max_sz) { dev_err(DEV, "SyncParam packet too long: received %u, expected <= %u bytes\n", packet_size, exp_max_sz); - return FALSE; + return false; } if (apv <= 88) { @@ -2775,7 +2775,7 @@ static int receive_SyncParam(struct drbd_conf *mdev, enum drbd_packets cmd, unsi memset(p->verify_alg, 0, 2 * SHARED_SECRET_MAX); if (drbd_recv(mdev, &p->head.payload, header_size) != header_size) - return FALSE; + return false; mdev->sync_conf.rate = be32_to_cpu(p->rate); @@ -2785,11 +2785,11 @@ static int receive_SyncParam(struct drbd_conf *mdev, enum drbd_packets cmd, unsi dev_err(DEV, "verify-alg too long, " "peer wants %u, accepting only %u byte\n", data_size, SHARED_SECRET_MAX); - return FALSE; + return false; } if (drbd_recv(mdev, p->verify_alg, data_size) != data_size) - return FALSE; + return false; /* we expect NUL terminated string */ /* but just in case someone tries to be evil */ @@ -2883,7 +2883,7 @@ disconnect: /* but free the verify_tfm again, if csums_tfm did not work out */ crypto_free_hash(verify_tfm); drbd_force_state(mdev, NS(conn, C_DISCONNECTING)); - return FALSE; + return false; } static void drbd_setup_order_type(struct drbd_conf *mdev, int peer) @@ -2920,7 +2920,7 @@ static int receive_sizes(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned if (p_size == 0 && mdev->state.disk == D_DISKLESS) { dev_err(DEV, "some backing storage is needed\n"); drbd_force_state(mdev, NS(conn, C_DISCONNECTING)); - return FALSE; + return false; } /* just store the peer's disk size for now. @@ -2957,7 +2957,7 @@ static int receive_sizes(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned drbd_force_state(mdev, NS(conn, C_DISCONNECTING)); mdev->ldev->dc.disk_size = my_usize; put_ldev(mdev); - return FALSE; + return false; } put_ldev(mdev); } @@ -2967,7 +2967,7 @@ static int receive_sizes(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned dd = drbd_determin_dev_size(mdev, ddsf); put_ldev(mdev); if (dd == dev_size_error) - return FALSE; + return false; drbd_md_sync(mdev); } else { /* I am diskless, need to accept the peer's size. */ @@ -3014,7 +3014,7 @@ static int receive_sizes(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned } } - return TRUE; + return true; } static int receive_uuids(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned int data_size) @@ -3038,7 +3038,7 @@ static int receive_uuids(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned dev_err(DEV, "Can only connect to data with current UUID=%016llX\n", (unsigned long long)mdev->ed_uuid); drbd_force_state(mdev, NS(conn, C_DISCONNECTING)); - return FALSE; + return false; } if (get_ldev(mdev)) { @@ -3073,7 +3073,7 @@ static int receive_uuids(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned if (mdev->state.conn >= C_CONNECTED && mdev->state.disk < D_INCONSISTENT) drbd_set_ed_uuid(mdev, p_uuid[UI_CURRENT]); - return TRUE; + return true; } /** @@ -3118,7 +3118,7 @@ static int receive_req_state(struct drbd_conf *mdev, enum drbd_packets cmd, unsi if (test_bit(DISCARD_CONCURRENT, &mdev->flags) && test_bit(CLUSTER_ST_CHANGE, &mdev->flags)) { drbd_send_sr_reply(mdev, SS_CONCURRENT_ST_CHG); - return TRUE; + return true; } mask = convert_state(mask); @@ -3129,7 +3129,7 @@ static int receive_req_state(struct drbd_conf *mdev, enum drbd_packets cmd, unsi drbd_send_sr_reply(mdev, rv); drbd_md_sync(mdev); - return TRUE; + return true; } static int receive_state(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned int data_size) @@ -3174,7 +3174,7 @@ static int receive_state(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned peer_state.conn == C_CONNECTED) { if (drbd_bm_total_weight(mdev) <= mdev->rs_failed) drbd_resync_finished(mdev); - return TRUE; + return true; } } @@ -3227,10 +3227,10 @@ static int receive_state(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned real_peer_disk = D_DISKLESS; } else { if (test_and_clear_bit(CONN_DRY_RUN, &mdev->flags)) - return FALSE; + return false; D_ASSERT(os.conn == C_WF_REPORT_PARAMS); drbd_force_state(mdev, NS(conn, C_DISCONNECTING)); - return FALSE; + return false; } } } @@ -3255,7 +3255,7 @@ static int receive_state(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned drbd_uuid_new_current(mdev); clear_bit(NEW_CUR_UUID, &mdev->flags); drbd_force_state(mdev, NS2(conn, C_PROTOCOL_ERROR, susp, 0)); - return FALSE; + return false; } rv = _drbd_set_state(mdev, ns, cs_flags, NULL); ns = mdev->state; @@ -3263,7 +3263,7 @@ static int receive_state(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned if (rv < SS_SUCCESS) { drbd_force_state(mdev, NS(conn, C_DISCONNECTING)); - return FALSE; + return false; } if (os.conn > C_WF_REPORT_PARAMS) { @@ -3281,7 +3281,7 @@ static int receive_state(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned drbd_md_sync(mdev); /* update connected indicator, la_size, ... */ - return TRUE; + return true; } static int receive_sync_uuid(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned int data_size) @@ -3308,7 +3308,7 @@ static int receive_sync_uuid(struct drbd_conf *mdev, enum drbd_packets cmd, unsi } else dev_err(DEV, "Ignoring SyncUUID packet!\n"); - return TRUE; + return true; } enum receive_bitmap_ret { OK, DONE, FAILED }; @@ -3462,7 +3462,7 @@ static int receive_bitmap(struct drbd_conf *mdev, enum drbd_packets cmd, unsigne struct bm_xfer_ctx c; void *buffer; enum receive_bitmap_ret ret; - int ok = FALSE; + int ok = false; struct p_header80 *h = &mdev->data.rbuf.header.h80; /* drbd_bm_lock(mdev, "receive bitmap"); By intention no bm_lock */ @@ -3535,7 +3535,7 @@ static int receive_bitmap(struct drbd_conf *mdev, enum drbd_packets cmd, unsigne drbd_conn_str(mdev->state.conn)); } - ok = TRUE; + ok = true; out: /* drbd_bm_unlock(mdev); by intention no lock */ if (ok && mdev->state.conn == C_WF_BITMAP_S) @@ -3569,7 +3569,7 @@ static int receive_UnplugRemote(struct drbd_conf *mdev, enum drbd_packets cmd, u * with the data requests being unplugged */ drbd_tcp_quickack(mdev->data.socket); - return TRUE; + return true; } static int receive_out_of_sync(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned int data_size) @@ -3578,7 +3578,7 @@ static int receive_out_of_sync(struct drbd_conf *mdev, enum drbd_packets cmd, un drbd_set_out_of_sync(mdev, be64_to_cpu(p->sector), be32_to_cpu(p->blksize)); - return TRUE; + return true; } typedef int (*drbd_cmd_handler_f)(struct drbd_conf *, enum drbd_packets cmd, unsigned int to_receive); @@ -4147,7 +4147,7 @@ static int got_RqSReply(struct drbd_conf *mdev, struct p_header80 *h) } wake_up(&mdev->state_wait); - return TRUE; + return true; } static int got_Ping(struct drbd_conf *mdev, struct p_header80 *h) @@ -4163,7 +4163,7 @@ static int got_PingAck(struct drbd_conf *mdev, struct p_header80 *h) if (!test_and_set_bit(GOT_PING_ACK, &mdev->flags)) wake_up(&mdev->misc_wait); - return TRUE; + return true; } static int got_IsInSync(struct drbd_conf *mdev, struct p_header80 *h) @@ -4186,7 +4186,7 @@ static int got_IsInSync(struct drbd_conf *mdev, struct p_header80 *h) dec_rs_pending(mdev); atomic_add(blksize >> 9, &mdev->rs_sect_in); - return TRUE; + return true; } /* when we receive the ACK for a write request, @@ -4230,14 +4230,14 @@ static int validate_req_change_req_state(struct drbd_conf *mdev, if (unlikely(!req)) { spin_unlock_irq(&mdev->req_lock); dev_err(DEV, "%s: got a corrupt block_id/sector pair\n", func); - return FALSE; + return false; } __req_mod(req, what, &m); spin_unlock_irq(&mdev->req_lock); if (m.bio) complete_master_bio(mdev, &m); - return TRUE; + return true; } static int got_BlockAck(struct drbd_conf *mdev, struct p_header80 *h) @@ -4252,7 +4252,7 @@ static int got_BlockAck(struct drbd_conf *mdev, struct p_header80 *h) if (is_syncer_block_id(p->block_id)) { drbd_set_in_sync(mdev, sector, blksize); dec_rs_pending(mdev); - return TRUE; + return true; } switch (be16_to_cpu(h->command)) { case P_RS_WRITE_ACK: @@ -4273,7 +4273,7 @@ static int got_BlockAck(struct drbd_conf *mdev, struct p_header80 *h) break; default: D_ASSERT(0); - return FALSE; + return false; } return validate_req_change_req_state(mdev, p->block_id, sector, @@ -4294,7 +4294,7 @@ static int got_NegAck(struct drbd_conf *mdev, struct p_header80 *h) int size = be32_to_cpu(p->blksize); dec_rs_pending(mdev); drbd_rs_failed_io(mdev, sector, size); - return TRUE; + return true; } return validate_req_change_req_state(mdev, p->block_id, sector, _ack_id_to_req, __func__ , neg_acked); @@ -4332,7 +4332,7 @@ static int got_NegRSDReply(struct drbd_conf *mdev, struct p_header80 *h) put_ldev(mdev); } - return TRUE; + return true; } static int got_BarrierAck(struct drbd_conf *mdev, struct p_header80 *h) @@ -4349,7 +4349,7 @@ static int got_BarrierAck(struct drbd_conf *mdev, struct p_header80 *h) drbd_queue_work_front(&mdev->data.work, w); } - return TRUE; + return true; } static int got_OVResult(struct drbd_conf *mdev, struct p_header80 *h) @@ -4370,7 +4370,7 @@ static int got_OVResult(struct drbd_conf *mdev, struct p_header80 *h) ov_oos_print(mdev); if (!get_ldev(mdev)) - return TRUE; + return true; drbd_rs_complete_io(mdev, sector); dec_rs_pending(mdev); @@ -4393,12 +4393,12 @@ static int got_OVResult(struct drbd_conf *mdev, struct p_header80 *h) } } put_ldev(mdev); - return TRUE; + return true; } static int got_skip(struct drbd_conf *mdev, struct p_header80 *h) { - return TRUE; + return true; } struct asender_cmd { diff --git a/drivers/block/drbd/drbd_req.c b/drivers/block/drbd/drbd_req.c index 53e7cc506456..528909090df7 100644 --- a/drivers/block/drbd/drbd_req.c +++ b/drivers/block/drbd/drbd_req.c @@ -445,7 +445,7 @@ int __req_mod(struct drbd_request *req, enum drbd_req_event what, req->rq_state |= RQ_LOCAL_COMPLETED; req->rq_state &= ~RQ_LOCAL_PENDING; - __drbd_chk_io_error(mdev, FALSE); + __drbd_chk_io_error(mdev, false); _req_may_be_done_not_susp(req, m); put_ldev(mdev); break; @@ -466,7 +466,7 @@ int __req_mod(struct drbd_request *req, enum drbd_req_event what, D_ASSERT(!(req->rq_state & RQ_NET_MASK)); - __drbd_chk_io_error(mdev, FALSE); + __drbd_chk_io_error(mdev, false); put_ldev(mdev); /* no point in retrying if there is no good remote data, diff --git a/drivers/block/drbd/drbd_worker.c b/drivers/block/drbd/drbd_worker.c index e201f6f82c0e..9fe3e890da0f 100644 --- a/drivers/block/drbd/drbd_worker.c +++ b/drivers/block/drbd/drbd_worker.c @@ -96,7 +96,7 @@ void drbd_endio_read_sec_final(struct drbd_epoch_entry *e) __releases(local) if (list_empty(&mdev->read_ee)) wake_up(&mdev->ee_wait); if (test_bit(__EE_WAS_ERROR, &e->flags)) - __drbd_chk_io_error(mdev, FALSE); + __drbd_chk_io_error(mdev, false); spin_unlock_irqrestore(&mdev->req_lock, flags); drbd_queue_work(&mdev->data.work, &e->w); @@ -139,7 +139,7 @@ static void drbd_endio_write_sec_final(struct drbd_epoch_entry *e) __releases(lo : list_empty(&mdev->active_ee); if (test_bit(__EE_WAS_ERROR, &e->flags)) - __drbd_chk_io_error(mdev, FALSE); + __drbd_chk_io_error(mdev, false); spin_unlock_irqrestore(&mdev->req_lock, flags); if (is_syncer_req) -- cgit v1.2.3 From 7777a8ba1fc980e5edfe492ebf5a1676497b8db2 Mon Sep 17 00:00:00 2001 From: Lars Ellenberg Date: Wed, 15 Dec 2010 23:21:39 +0100 Subject: drbd: bitmap: don't count unused bits (fix non-terminating resync) We trusted the on-disk bitmap to have unused bits cleared. In case that is not true for whatever reason, and we take a code path where the unused bits don't get cleared elsewhere (bm_clear_surplus is not called), we may miscount the bits, and get confused during resync, waiting for bits to get cleared that we don't even use: the resync process would not terminate. Fix this by masking out unused bits in __bm_count_bits. Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_bitmap.c | 25 ++++++++++++++++++++++--- 1 file changed, 22 insertions(+), 3 deletions(-) (limited to 'drivers/block/drbd/drbd_bitmap.c') diff --git a/drivers/block/drbd/drbd_bitmap.c b/drivers/block/drbd/drbd_bitmap.c index 9390e9526786..c5361487cf47 100644 --- a/drivers/block/drbd/drbd_bitmap.c +++ b/drivers/block/drbd/drbd_bitmap.c @@ -376,9 +376,16 @@ static unsigned long __bm_count_bits(struct drbd_bitmap *b, const int swap_endia unsigned long *p_addr, *bm, offset = 0; unsigned long bits = 0; unsigned long i, do_now; + unsigned long words; - while (offset < b->bm_words) { - i = do_now = min_t(size_t, b->bm_words-offset, LWPP); + /* due to 64bit alignment, the last long on a 32bit arch + * may be not used at all. The last used long will likely + * be only partially used, always. Don't count those bits, + * but mask them out. */ + words = (b->bm_bits + BITS_PER_LONG - 1) >> LN2_BPL; + + while (offset < words) { + i = do_now = min_t(size_t, words-offset, LWPP); p_addr = __bm_map_paddr(b, offset, KM_USER0); bm = p_addr + MLPP(offset); while (i--) { @@ -388,8 +395,20 @@ static unsigned long __bm_count_bits(struct drbd_bitmap *b, const int swap_endia #endif bits += hweight_long(*bm++); } - __bm_unmap(p_addr, KM_USER0); offset += do_now; + if (offset == words) { + /* last word may only be partially used, + * see also bm_clear_surplus. */ + i = (1UL << (b->bm_bits & (BITS_PER_LONG-1))) -1; + if (i) { + bits -= hweight_long(p_addr[do_now-1] & ~i); + p_addr[do_now-1] &= i; + } + /* 32bit arch, may have an unused padding long */ + if (words != b->bm_words) + p_addr[do_now] = 0; + } + __bm_unmap(p_addr, KM_USER0); cond_resched(); } -- cgit v1.2.3 From 95a0f10cddbf93ce89c175ac1c53dad2d20ad309 Mon Sep 17 00:00:00 2001 From: Lars Ellenberg Date: Wed, 15 Dec 2010 08:59:09 +0100 Subject: drbd: store in-core bitmap little endian, regardless of architecture Our on-disk bitmap is a little endian bitstream. Up to now, we have stored the in-core copy of that in native endian, applying byte order conversion when necessary. Instead, keep the bitmap pages little endian, as they are read from disk, and use the generic_*_le_bit family of functions. Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_bitmap.c | 166 ++++++++++++++++++++------------------- 1 file changed, 84 insertions(+), 82 deletions(-) (limited to 'drivers/block/drbd/drbd_bitmap.c') diff --git a/drivers/block/drbd/drbd_bitmap.c b/drivers/block/drbd/drbd_bitmap.c index c5361487cf47..8d959ed6c2cc 100644 --- a/drivers/block/drbd/drbd_bitmap.c +++ b/drivers/block/drbd/drbd_bitmap.c @@ -30,6 +30,7 @@ #include #include "drbd_int.h" + /* OPAQUE outside this file! * interface defined in drbd_int.h @@ -154,6 +155,14 @@ void drbd_bm_unlock(struct drbd_conf *mdev) mutex_unlock(&b->bm_change); } +static unsigned int bm_bit_to_page_idx(struct drbd_bitmap *b, u64 bitnr) +{ + /* page_nr = (bitnr/8) >> PAGE_SHIFT; */ + unsigned int page_nr = bitnr >> (PAGE_SHIFT + 3); + BUG_ON(page_nr >= b->bm_number_of_pages); + return page_nr; +} + /* word offset to long pointer */ static unsigned long *__bm_map_paddr(struct drbd_bitmap *b, unsigned long offset, const enum km_type km) { @@ -168,6 +177,17 @@ static unsigned long *__bm_map_paddr(struct drbd_bitmap *b, unsigned long offset return (unsigned long *) kmap_atomic(page, km); } +static unsigned long *__bm_map_pidx(struct drbd_bitmap *b, unsigned int idx, const enum km_type km) +{ + struct page *page = b->bm_pages[idx]; + return (unsigned long *) kmap_atomic(page, km); +} + +static unsigned long *bm_map_pidx(struct drbd_bitmap *b, unsigned int idx) +{ + return __bm_map_pidx(b, idx, KM_IRQ1); +} + static unsigned long * bm_map_paddr(struct drbd_bitmap *b, unsigned long offset) { return __bm_map_paddr(b, offset, KM_IRQ1); @@ -329,22 +349,42 @@ void drbd_bm_cleanup(struct drbd_conf *mdev) * this masks out the remaining bits. * Returns the number of bits cleared. */ +#define BITS_PER_PAGE (1UL << (PAGE_SHIFT + 3)) +#define BITS_PER_PAGE_MASK (BITS_PER_PAGE - 1) +#define BITS_PER_LONG_MASK (BITS_PER_LONG - 1) static int bm_clear_surplus(struct drbd_bitmap *b) { - const unsigned long mask = (1UL << (b->bm_bits & (BITS_PER_LONG-1))) - 1; - size_t w = b->bm_bits >> LN2_BPL; - int cleared = 0; + unsigned long mask; unsigned long *p_addr, *bm; + int tmp; + int cleared = 0; - p_addr = bm_map_paddr(b, w); - bm = p_addr + MLPP(w); - if (w < b->bm_words) { + /* number of bits modulo bits per page */ + tmp = (b->bm_bits & BITS_PER_PAGE_MASK); + /* mask the used bits of the word containing the last bit */ + mask = (1UL << (tmp & BITS_PER_LONG_MASK)) -1; + /* bitmap is always stored little endian, + * on disk and in core memory alike */ + mask = cpu_to_lel(mask); + + /* because of the "extra long to catch oob access" we allocate in + * drbd_bm_resize, bm_number_of_pages -1 is not necessarily the page + * containing the last _relevant_ bitmap word */ + p_addr = bm_map_pidx(b, bm_bit_to_page_idx(b, b->bm_bits - 1)); + bm = p_addr + (tmp/BITS_PER_LONG); + if (mask) { + /* If mask != 0, we are not exactly aligned, so bm now points + * to the long containing the last bit. + * If mask == 0, bm already points to the word immediately + * after the last (long word aligned) bit. */ cleared = hweight_long(*bm & ~mask); *bm &= mask; - w++; bm++; + bm++; } - if (w < b->bm_words) { + if (BITS_PER_LONG == 32 && ((bm - p_addr) & 1) == 1) { + /* on a 32bit arch, we may need to zero out + * a padding long to align with a 64bit remote */ cleared += hweight_long(*bm); *bm = 0; } @@ -354,24 +394,41 @@ static int bm_clear_surplus(struct drbd_bitmap *b) static void bm_set_surplus(struct drbd_bitmap *b) { - const unsigned long mask = (1UL << (b->bm_bits & (BITS_PER_LONG-1))) - 1; - size_t w = b->bm_bits >> LN2_BPL; + unsigned long mask; unsigned long *p_addr, *bm; - - p_addr = bm_map_paddr(b, w); - bm = p_addr + MLPP(w); - if (w < b->bm_words) { + int tmp; + + /* number of bits modulo bits per page */ + tmp = (b->bm_bits & BITS_PER_PAGE_MASK); + /* mask the used bits of the word containing the last bit */ + mask = (1UL << (tmp & BITS_PER_LONG_MASK)) -1; + /* bitmap is always stored little endian, + * on disk and in core memory alike */ + mask = cpu_to_lel(mask); + + /* because of the "extra long to catch oob access" we allocate in + * drbd_bm_resize, bm_number_of_pages -1 is not necessarily the page + * containing the last _relevant_ bitmap word */ + p_addr = bm_map_pidx(b, bm_bit_to_page_idx(b, b->bm_bits - 1)); + bm = p_addr + (tmp/BITS_PER_LONG); + if (mask) { + /* If mask != 0, we are not exactly aligned, so bm now points + * to the long containing the last bit. + * If mask == 0, bm already points to the word immediately + * after the last (long word aligned) bit. */ *bm |= ~mask; - bm++; w++; + bm++; } - if (w < b->bm_words) { - *bm = ~(0UL); + if (BITS_PER_LONG == 32 && ((bm - p_addr) & 1) == 1) { + /* on a 32bit arch, we may need to zero out + * a padding long to align with a 64bit remote */ + *bm = ~0UL; } bm_unmap(p_addr); } -static unsigned long __bm_count_bits(struct drbd_bitmap *b, const int swap_endian) +static unsigned long bm_count_bits(struct drbd_bitmap *b) { unsigned long *p_addr, *bm, offset = 0; unsigned long bits = 0; @@ -389,10 +446,6 @@ static unsigned long __bm_count_bits(struct drbd_bitmap *b, const int swap_endia p_addr = __bm_map_paddr(b, offset, KM_USER0); bm = p_addr + MLPP(offset); while (i--) { -#ifndef __LITTLE_ENDIAN - if (swap_endian) - *bm = lel_to_cpu(*bm); -#endif bits += hweight_long(*bm++); } offset += do_now; @@ -415,16 +468,6 @@ static unsigned long __bm_count_bits(struct drbd_bitmap *b, const int swap_endia return bits; } -static unsigned long bm_count_bits(struct drbd_bitmap *b) -{ - return __bm_count_bits(b, 0); -} - -static unsigned long bm_count_bits_swap_endian(struct drbd_bitmap *b) -{ - return __bm_count_bits(b, 1); -} - /* offset and len in long words.*/ static void bm_memset(struct drbd_bitmap *b, size_t offset, int c, size_t len) { @@ -662,7 +705,7 @@ void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset, size_t number, offset += do_now; while (do_now--) { bits = hweight_long(*bm); - word = *bm | lel_to_cpu(*buffer++); + word = *bm | *buffer++; *bm++ = word; b->bm_set += hweight_long(word) - bits; } @@ -709,7 +752,7 @@ void drbd_bm_get_lel(struct drbd_conf *mdev, size_t offset, size_t number, bm = p_addr + MLPP(offset); offset += do_now; while (do_now--) - *buffer++ = cpu_to_lel(*bm++); + *buffer++ = *bm++; bm_unmap(p_addr); } } @@ -795,39 +838,6 @@ static void bm_page_io_async(struct drbd_conf *mdev, struct drbd_bitmap *b, int } } -# if defined(__LITTLE_ENDIAN) - /* nothing to do, on disk == in memory */ -# define bm_cpu_to_lel(x) ((void)0) -# else -static void bm_cpu_to_lel(struct drbd_bitmap *b) -{ - /* need to cpu_to_lel all the pages ... - * this may be optimized by using - * cpu_to_lel(-1) == -1 and cpu_to_lel(0) == 0; - * the following is still not optimal, but better than nothing */ - unsigned int i; - unsigned long *p_addr, *bm; - if (b->bm_set == 0) { - /* no page at all; avoid swap if all is 0 */ - i = b->bm_number_of_pages; - } else if (b->bm_set == b->bm_bits) { - /* only the last page */ - i = b->bm_number_of_pages - 1; - } else { - /* all pages */ - i = 0; - } - for (; i < b->bm_number_of_pages; i++) { - p_addr = kmap_atomic(b->bm_pages[i], KM_USER0); - for (bm = p_addr; bm < p_addr + PAGE_SIZE/sizeof(long); bm++) - *bm = cpu_to_lel(*bm); - kunmap_atomic(p_addr, KM_USER0); - } -} -# endif -/* lel_to_cpu == cpu_to_lel */ -# define bm_lel_to_cpu(x) bm_cpu_to_lel(x) - /* * bm_rw: read/write the whole bitmap from/to its on disk location. */ @@ -847,10 +857,6 @@ static int bm_rw(struct drbd_conf *mdev, int rw) __must_hold(local) bm_words = drbd_bm_words(mdev); num_pages = (bm_words*sizeof(long) + PAGE_SIZE-1) >> PAGE_SHIFT; - /* on disk bitmap is little endian */ - if (rw == WRITE) - bm_cpu_to_lel(b); - now = jiffies; atomic_set(&b->bm_async_io, num_pages); __clear_bit(BM_MD_IO_ERROR, &b->bm_flags); @@ -869,13 +875,9 @@ static int bm_rw(struct drbd_conf *mdev, int rw) __must_hold(local) now = jiffies; if (rw == WRITE) { - /* swap back endianness */ - bm_lel_to_cpu(b); - /* flush bitmap to stable storage */ drbd_md_flush(mdev); } else /* rw == READ */ { - /* just read, if necessary adjust endianness */ - b->bm_set = bm_count_bits_swap_endian(b); + b->bm_set = bm_count_bits(b); dev_info(DEV, "recounting of set bits took additional %lu jiffies\n", jiffies - now); } @@ -969,9 +971,9 @@ static unsigned long __bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo, p_addr = __bm_map_paddr(b, offset, km); if (find_zero_bit) - i = find_next_zero_bit(p_addr, PAGE_SIZE*8, bm_fo & BPP_MASK); + i = generic_find_next_zero_le_bit(p_addr, PAGE_SIZE*8, bm_fo & BPP_MASK); else - i = find_next_bit(p_addr, PAGE_SIZE*8, bm_fo & BPP_MASK); + i = generic_find_next_le_bit(p_addr, PAGE_SIZE*8, bm_fo & BPP_MASK); __bm_unmap(p_addr, km); if (i < PAGE_SIZE*8) { @@ -1064,9 +1066,9 @@ static int __bm_change_bits_to(struct drbd_conf *mdev, const unsigned long s, last_page_nr = page_nr; } if (val) - c += (0 == __test_and_set_bit(bitnr & BPP_MASK, p_addr)); + c += (0 == generic___test_and_set_le_bit(bitnr & BPP_MASK, p_addr)); else - c -= (0 != __test_and_clear_bit(bitnr & BPP_MASK, p_addr)); + c -= (0 != generic___test_and_clear_le_bit(bitnr & BPP_MASK, p_addr)); } if (p_addr) __bm_unmap(p_addr, km); @@ -1211,7 +1213,7 @@ int drbd_bm_test_bit(struct drbd_conf *mdev, const unsigned long bitnr) if (bitnr < b->bm_bits) { unsigned long offset = bitnr>>LN2_BPL; p_addr = bm_map_paddr(b, offset); - i = test_bit(bitnr & BPP_MASK, p_addr) ? 1 : 0; + i = generic_test_le_bit(bitnr & BPP_MASK, p_addr) ? 1 : 0; bm_unmap(p_addr); } else if (bitnr == b->bm_bits) { i = -1; @@ -1255,7 +1257,7 @@ int drbd_bm_count_bits(struct drbd_conf *mdev, const unsigned long s, const unsi ERR_IF (bitnr >= b->bm_bits) { dev_err(DEV, "bitnr=%lu bm_bits=%lu\n", bitnr, b->bm_bits); } else { - c += (0 != test_bit(bitnr - (page_nr << (PAGE_SHIFT+3)), p_addr)); + c += (0 != generic_test_le_bit(bitnr - (page_nr << (PAGE_SHIFT+3)), p_addr)); } } if (p_addr) -- cgit v1.2.3 From 19f843aa08e2d8f87a09b4c2edc43b00638423a8 Mon Sep 17 00:00:00 2001 From: Lars Ellenberg Date: Wed, 15 Dec 2010 08:59:11 +0100 Subject: drbd: bitmap keep track of changes vs on-disk bitmap When we set or clear bits in a bitmap page, also set a flag in the page->private pointer. This allows us to skip writes of unchanged pages. Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_actlog.c | 132 +++---------- drivers/block/drbd/drbd_bitmap.c | 409 +++++++++++++++++++++++++++++---------- drivers/block/drbd/drbd_int.h | 7 +- drivers/block/drbd/drbd_main.c | 27 ++- drivers/block/drbd/drbd_nl.c | 11 +- drivers/block/drbd/drbd_worker.c | 7 +- 6 files changed, 373 insertions(+), 220 deletions(-) (limited to 'drivers/block/drbd/drbd_bitmap.c') diff --git a/drivers/block/drbd/drbd_actlog.c b/drivers/block/drbd/drbd_actlog.c index e3f0f4d31d75..090fc2ce0df4 100644 --- a/drivers/block/drbd/drbd_actlog.c +++ b/drivers/block/drbd/drbd_actlog.c @@ -262,6 +262,33 @@ void drbd_al_complete_io(struct drbd_conf *mdev, sector_t sector) spin_unlock_irqrestore(&mdev->al_lock, flags); } +#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. + * Code below will not work if this is violated. + * Will be cleaned up with some followup patch. + */ +# error FIXME +#endif + +static unsigned int al_extent_to_bm_page(unsigned int al_enr) +{ + return al_enr >> + /* bit to page */ + ((PAGE_SHIFT + 3) - + /* al extent number to bit */ + (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) - + /* al extent number to bit */ + (BM_EXT_SHIFT - BM_BLOCK_SHIFT)); +} + int w_al_write_transaction(struct drbd_conf *mdev, struct drbd_work *w, int unused) { @@ -289,7 +316,7 @@ w_al_write_transaction(struct drbd_conf *mdev, struct drbd_work *w, int unused) * For now, we must not write the transaction, * if we cannot write out the bitmap of the evicted extent. */ if (mdev->state.conn < C_CONNECTED && evicted != LC_FREE) - drbd_bm_write_sect(mdev, evicted/AL_EXT_PER_BM_SECT); + drbd_bm_write_page(mdev, al_extent_to_bm_page(evicted)); /* The bitmap write may have failed, causing a state change. */ if (mdev->state.disk < D_INCONSISTENT) { @@ -635,105 +662,6 @@ out_bio_put: return err; } -/** - * drbd_al_to_on_disk_bm() - * Writes bitmap parts covered by active AL extents - * @mdev: DRBD device. - * - * Called when we detach (unconfigure) local storage, - * or when we go from R_PRIMARY to R_SECONDARY role. - */ -void drbd_al_to_on_disk_bm(struct drbd_conf *mdev) -{ - int i, nr_elements; - unsigned int enr; - struct bio **bios; - struct drbd_atodb_wait wc; - - ERR_IF (!get_ldev_if_state(mdev, D_ATTACHING)) - return; /* sorry, I don't have any act_log etc... */ - - wait_event(mdev->al_wait, lc_try_lock(mdev->act_log)); - - nr_elements = mdev->act_log->nr_elements; - - /* GFP_KERNEL, we are not in anyone's write-out path */ - bios = kzalloc(sizeof(struct bio *) * nr_elements, GFP_KERNEL); - if (!bios) - goto submit_one_by_one; - - atomic_set(&wc.count, 0); - init_completion(&wc.io_done); - wc.mdev = mdev; - wc.error = 0; - - for (i = 0; i < nr_elements; i++) { - enr = lc_element_by_index(mdev->act_log, i)->lc_number; - if (enr == LC_FREE) - continue; - /* next statement also does atomic_inc wc.count and local_cnt */ - if (atodb_prepare_unless_covered(mdev, bios, - enr/AL_EXT_PER_BM_SECT, - &wc)) - goto free_bios_submit_one_by_one; - } - - /* unnecessary optimization? */ - lc_unlock(mdev->act_log); - wake_up(&mdev->al_wait); - - /* all prepared, submit them */ - for (i = 0; i < nr_elements; i++) { - if (bios[i] == NULL) - break; - if (drbd_insert_fault(mdev, DRBD_FAULT_MD_WR)) { - bios[i]->bi_rw = WRITE; - bio_endio(bios[i], -EIO); - } else { - submit_bio(WRITE, bios[i]); - } - } - - /* always (try to) flush bitmap to stable storage */ - drbd_md_flush(mdev); - - /* In case we did not submit a single IO do not wait for - * them to complete. ( Because we would wait forever here. ) - * - * In case we had IOs and they are already complete, there - * is not point in waiting anyways. - * Therefore this if () ... */ - if (atomic_read(&wc.count)) - wait_for_completion(&wc.io_done); - - put_ldev(mdev); - - kfree(bios); - return; - - free_bios_submit_one_by_one: - /* free everything by calling the endio callback directly. */ - for (i = 0; i < nr_elements && bios[i]; i++) - bio_endio(bios[i], 0); - - kfree(bios); - - submit_one_by_one: - dev_warn(DEV, "Using the slow drbd_al_to_on_disk_bm()\n"); - - for (i = 0; i < mdev->act_log->nr_elements; i++) { - enr = lc_element_by_index(mdev->act_log, i)->lc_number; - if (enr == LC_FREE) - continue; - /* Really slow: if we have al-extents 16..19 active, - * sector 4 will be written four times! Synchronous! */ - drbd_bm_write_sect(mdev, enr/AL_EXT_PER_BM_SECT); - } - - lc_unlock(mdev->act_log); - wake_up(&mdev->al_wait); - put_ldev(mdev); -} - /** * drbd_al_apply_to_bm() - Sets the bitmap to diry(1) where covered ba active AL extents * @mdev: DRBD device. @@ -813,7 +741,7 @@ static int w_update_odbm(struct drbd_conf *mdev, struct drbd_work *w, int unused return 1; } - drbd_bm_write_sect(mdev, udw->enr); + drbd_bm_write_page(mdev, rs_extent_to_bm_page(udw->enr)); put_ldev(mdev); kfree(udw); @@ -893,7 +821,6 @@ static void drbd_try_clear_on_disk_bm(struct drbd_conf *mdev, sector_t sector, dev_warn(DEV, "Kicking resync_lru element enr=%u " "out with rs_failed=%d\n", ext->lce.lc_number, ext->rs_failed); - set_bit(WRITE_BM_AFTER_RESYNC, &mdev->flags); } ext->rs_left = rs_left; ext->rs_failed = success ? 0 : count; @@ -912,7 +839,6 @@ static void drbd_try_clear_on_disk_bm(struct drbd_conf *mdev, sector_t sector, drbd_queue_work_front(&mdev->data.work, &udw->w); } else { dev_warn(DEV, "Could not kmalloc an udw\n"); - set_bit(WRITE_BM_AFTER_RESYNC, &mdev->flags); } } } else { diff --git a/drivers/block/drbd/drbd_bitmap.c b/drivers/block/drbd/drbd_bitmap.c index 8d959ed6c2cc..72cd41a96ef9 100644 --- a/drivers/block/drbd/drbd_bitmap.c +++ b/drivers/block/drbd/drbd_bitmap.c @@ -70,8 +70,7 @@ struct drbd_bitmap { sector_t bm_dev_capacity; struct mutex bm_change; /* serializes resize operations */ - atomic_t bm_async_io; - wait_queue_head_t bm_io_wait; + wait_queue_head_t bm_io_wait; /* used to serialize IO of single pages */ unsigned long bm_flags; @@ -82,7 +81,7 @@ struct drbd_bitmap { /* definition of bits in bm_flags */ #define BM_LOCKED 0 -#define BM_MD_IO_ERROR 1 +// #define BM_MD_IO_ERROR 1 unused now. #define BM_P_VMALLOCED 2 static int __bm_change_bits_to(struct drbd_conf *mdev, const unsigned long s, @@ -155,26 +154,117 @@ void drbd_bm_unlock(struct drbd_conf *mdev) mutex_unlock(&b->bm_change); } -static unsigned int bm_bit_to_page_idx(struct drbd_bitmap *b, u64 bitnr) +/* we store some "meta" info about our pages in page->private */ +/* at a granularity of 4k storage per bitmap bit: + * one peta byte storage: 1<<50 byte, 1<<38 * 4k storage blocks + * 1<<38 bits, + * 1<<23 4k bitmap pages. + * Use 24 bits as page index, covers 2 peta byte storage + * at a granularity of 4k per bit. + * Used to report the failed page idx on io error from the endio handlers. + */ +#define BM_PAGE_IDX_MASK ((1UL<<24)-1) +/* this page is currently read in, or written back */ +#define BM_PAGE_IO_LOCK 31 +/* if there has been an IO error for this page */ +#define BM_PAGE_IO_ERROR 30 +/* this is to be able to intelligently skip disk IO, + * set if bits have been set since last IO. */ +#define BM_PAGE_NEED_WRITEOUT 29 +/* to mark for lazy writeout once syncer cleared all clearable bits, + * we if bits have been cleared since last IO. */ +#define BM_PAGE_LAZY_WRITEOUT 28 + +/* store_page_idx uses non-atomic assingment. It is only used directly after + * allocating the page. All other bm_set_page_* and bm_clear_page_* need to + * use atomic bit manipulation, as set_out_of_sync (and therefore bitmap + * changes) may happen from various contexts, and wait_on_bit/wake_up_bit + * requires it all to be atomic as well. */ +static void bm_store_page_idx(struct page *page, unsigned long idx) { - /* page_nr = (bitnr/8) >> PAGE_SHIFT; */ - unsigned int page_nr = bitnr >> (PAGE_SHIFT + 3); - BUG_ON(page_nr >= b->bm_number_of_pages); - return page_nr; + BUG_ON(0 != (idx & ~BM_PAGE_IDX_MASK)); + page_private(page) |= idx; } -/* word offset to long pointer */ -static unsigned long *__bm_map_paddr(struct drbd_bitmap *b, unsigned long offset, const enum km_type km) +static unsigned long bm_page_to_idx(struct page *page) { - struct page *page; - unsigned long page_nr; + return page_private(page) & BM_PAGE_IDX_MASK; +} + +/* As is very unlikely that the same page is under IO from more than one + * context, we can get away with a bit per page and one wait queue per bitmap. + */ +static void bm_page_lock_io(struct drbd_conf *mdev, int page_nr) +{ + struct drbd_bitmap *b = mdev->bitmap; + void *addr = &page_private(b->bm_pages[page_nr]); + wait_event(b->bm_io_wait, !test_and_set_bit(BM_PAGE_IO_LOCK, addr)); +} + +static void bm_page_unlock_io(struct drbd_conf *mdev, int page_nr) +{ + struct drbd_bitmap *b = mdev->bitmap; + void *addr = &page_private(b->bm_pages[page_nr]); + clear_bit(BM_PAGE_IO_LOCK, addr); + smp_mb__after_clear_bit(); + wake_up(&mdev->bitmap->bm_io_wait); +} + +/* set _before_ submit_io, so it may be reset due to being changed + * while this page is in flight... will get submitted later again */ +static void bm_set_page_unchanged(struct page *page) +{ + /* use cmpxchg? */ + clear_bit(BM_PAGE_NEED_WRITEOUT, &page_private(page)); + clear_bit(BM_PAGE_LAZY_WRITEOUT, &page_private(page)); +} + +static void bm_set_page_need_writeout(struct page *page) +{ + set_bit(BM_PAGE_NEED_WRITEOUT, &page_private(page)); +} + +static int bm_test_page_unchanged(struct page *page) +{ + volatile const unsigned long *addr = &page_private(page); + return (*addr & ((1UL<> PAGE_SHIFT; */ - page_nr = offset >> (PAGE_SHIFT - LN2_BPL + 3); + unsigned int page_nr = long_nr >> (PAGE_SHIFT - LN2_BPL + 3); BUG_ON(page_nr >= b->bm_number_of_pages); - page = b->bm_pages[page_nr]; + return page_nr; +} - return (unsigned long *) kmap_atomic(page, km); +static unsigned int bm_bit_to_page_idx(struct drbd_bitmap *b, u64 bitnr) +{ + /* page_nr = (bitnr/8) >> PAGE_SHIFT; */ + unsigned int page_nr = bitnr >> (PAGE_SHIFT + 3); + BUG_ON(page_nr >= b->bm_number_of_pages); + return page_nr; } static unsigned long *__bm_map_pidx(struct drbd_bitmap *b, unsigned int idx, const enum km_type km) @@ -188,11 +278,6 @@ static unsigned long *bm_map_pidx(struct drbd_bitmap *b, unsigned int idx) return __bm_map_pidx(b, idx, KM_IRQ1); } -static unsigned long * bm_map_paddr(struct drbd_bitmap *b, unsigned long offset) -{ - return __bm_map_paddr(b, offset, KM_IRQ1); -} - static void __bm_unmap(unsigned long *p_addr, const enum km_type km) { kunmap_atomic(p_addr, km); @@ -222,6 +307,7 @@ static void bm_unmap(unsigned long *p_addr) * to be able to report device specific. */ + static void bm_free_pages(struct page **pages, unsigned long number) { unsigned long i; @@ -289,6 +375,9 @@ static struct page **bm_realloc_pages(struct drbd_bitmap *b, unsigned long want) bm_vk_free(new_pages, vmalloced); return NULL; } + /* we want to know which page it is + * from the endio handlers */ + bm_store_page_idx(page, i); new_pages[i] = page; } } else { @@ -443,7 +532,7 @@ static unsigned long bm_count_bits(struct drbd_bitmap *b) while (offset < words) { i = do_now = min_t(size_t, words-offset, LWPP); - p_addr = __bm_map_paddr(b, offset, KM_USER0); + p_addr = __bm_map_pidx(b, bm_word_to_page_idx(b, offset), KM_USER0); bm = p_addr + MLPP(offset); while (i--) { bits += hweight_long(*bm++); @@ -472,6 +561,7 @@ static unsigned long bm_count_bits(struct drbd_bitmap *b) static void bm_memset(struct drbd_bitmap *b, size_t offset, int c, size_t len) { unsigned long *p_addr, *bm; + unsigned int idx; size_t do_now, end; #define BM_SECTORS_PER_BIT (BM_BLOCK_SIZE/512) @@ -485,7 +575,8 @@ static void bm_memset(struct drbd_bitmap *b, size_t offset, int c, size_t len) while (offset < end) { do_now = min_t(size_t, ALIGN(offset + 1, LWPP), end) - offset; - p_addr = bm_map_paddr(b, offset); + idx = bm_word_to_page_idx(b, offset); + p_addr = bm_map_pidx(b, idx); bm = p_addr + MLPP(offset); if (bm+do_now > p_addr + LWPP) { printk(KERN_ALERT "drbd: BUG BUG BUG! p_addr:%p bm:%p do_now:%d\n", @@ -494,6 +585,7 @@ static void bm_memset(struct drbd_bitmap *b, size_t offset, int c, size_t len) } memset(bm, c, do_now * sizeof(long)); bm_unmap(p_addr); + bm_set_page_need_writeout(b->bm_pages[idx]); offset += do_now; } } @@ -604,7 +696,7 @@ int drbd_bm_resize(struct drbd_conf *mdev, sector_t capacity, int set_new_bits) bm_free_pages(opages + want, have - want); } - p_addr = bm_map_paddr(b, words); + p_addr = bm_map_pidx(b, bm_word_to_page_idx(b, words)); bm = p_addr + MLPP(words); *bm = DRBD_MAGIC; bm_unmap(p_addr); @@ -616,7 +708,7 @@ int drbd_bm_resize(struct drbd_conf *mdev, sector_t capacity, int set_new_bits) bm_vk_free(opages, opages_vmalloced); if (!growing) b->bm_set = bm_count_bits(b); - dev_info(DEV, "resync bitmap: bits=%lu words=%lu\n", bits, words); + dev_info(DEV, "resync bitmap: bits=%lu words=%lu pages=%lu\n", bits, words, want); out: drbd_bm_unlock(mdev); @@ -686,6 +778,7 @@ void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset, size_t number, struct drbd_bitmap *b = mdev->bitmap; unsigned long *p_addr, *bm; unsigned long word, bits; + unsigned int idx; size_t end, do_now; end = offset + number; @@ -700,7 +793,8 @@ void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset, size_t number, spin_lock_irq(&b->bm_lock); while (offset < end) { do_now = min_t(size_t, ALIGN(offset+1, LWPP), end) - offset; - p_addr = bm_map_paddr(b, offset); + idx = bm_word_to_page_idx(b, offset); + p_addr = bm_map_pidx(b, idx); bm = p_addr + MLPP(offset); offset += do_now; while (do_now--) { @@ -710,6 +804,7 @@ void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset, size_t number, b->bm_set += hweight_long(word) - bits; } bm_unmap(p_addr); + bm_set_page_need_writeout(b->bm_pages[idx]); } /* with 32bit <-> 64bit cross-platform connect * this is only correct for current usage, @@ -748,7 +843,7 @@ void drbd_bm_get_lel(struct drbd_conf *mdev, size_t offset, size_t number, else { while (offset < end) { do_now = min_t(size_t, ALIGN(offset+1, LWPP), end) - offset; - p_addr = bm_map_paddr(b, offset); + p_addr = bm_map_pidx(b, bm_word_to_page_idx(b, offset)); bm = p_addr + MLPP(offset); offset += do_now; while (do_now--) @@ -786,9 +881,22 @@ void drbd_bm_clear_all(struct drbd_conf *mdev) spin_unlock_irq(&b->bm_lock); } +struct bm_aio_ctx { + struct drbd_conf *mdev; + atomic_t in_flight; + wait_queue_head_t io_wait; + unsigned flags; +#define BM_AIO_COPY_PAGES 1 + int error; +}; + +/* bv_page may be a copy, or may be the original */ static void bm_async_io_complete(struct bio *bio, int error) { - struct drbd_bitmap *b = bio->bi_private; + struct bm_aio_ctx *ctx = bio->bi_private; + struct drbd_conf *mdev = ctx->mdev; + struct drbd_bitmap *b = mdev->bitmap; + unsigned int idx = bm_page_to_idx(bio->bi_io_vec[0].bv_page); int uptodate = bio_flagged(bio, BIO_UPTODATE); @@ -799,35 +907,79 @@ static void bm_async_io_complete(struct bio *bio, int error) if (!error && !uptodate) error = -EIO; + if (!bm_test_page_unchanged(b->bm_pages[idx])) + dev_info(DEV, "bitmap page idx %u changed during IO!\n", idx); + if (error) { - /* doh. what now? - * for now, set all bits, and flag MD_IO_ERROR */ - __set_bit(BM_MD_IO_ERROR, &b->bm_flags); + /* ctx error will hold the completed-last non-zero error code, + * in case error codes differ. */ + ctx->error = error; + bm_set_page_io_err(b->bm_pages[idx]); + /* Not identical to on disk version of it. + * Is BM_PAGE_IO_ERROR enough? */ + if (__ratelimit(&drbd_ratelimit_state)) + dev_err(DEV, "IO ERROR %d on bitmap page idx %u\n", + error, idx); + } else { + bm_clear_page_io_err(b->bm_pages[idx]); + dynamic_dev_dbg(DEV, "bitmap page idx %u completed\n", idx); } - if (atomic_dec_and_test(&b->bm_async_io)) - wake_up(&b->bm_io_wait); + + bm_page_unlock_io(mdev, idx); + + /* FIXME give back to page pool */ + if (ctx->flags & BM_AIO_COPY_PAGES) + put_page(bio->bi_io_vec[0].bv_page); bio_put(bio); + + if (atomic_dec_and_test(&ctx->in_flight)) + wake_up(&ctx->io_wait); } -static void bm_page_io_async(struct drbd_conf *mdev, struct drbd_bitmap *b, int page_nr, int rw) __must_hold(local) +static void bm_page_io_async(struct bm_aio_ctx *ctx, int page_nr, int rw) __must_hold(local) { /* we are process context. we always get a bio */ struct bio *bio = bio_alloc(GFP_KERNEL, 1); + struct drbd_conf *mdev = ctx->mdev; + struct drbd_bitmap *b = mdev->bitmap; + struct page *page; unsigned int len; + sector_t on_disk_sector = mdev->ldev->md.md_offset + mdev->ldev->md.bm_offset; on_disk_sector += ((sector_t)page_nr) << (PAGE_SHIFT-9); /* this might happen with very small - * flexible external meta data device */ + * flexible external meta data device, + * or with PAGE_SIZE > 4k */ len = min_t(unsigned int, PAGE_SIZE, (drbd_md_last_sector(mdev->ldev) - on_disk_sector + 1)<<9); + /* serialize IO on this page */ + bm_page_lock_io(mdev, page_nr); + /* before memcpy and submit, + * so it can be redirtied any time */ + bm_set_page_unchanged(b->bm_pages[page_nr]); + + if (ctx->flags & BM_AIO_COPY_PAGES) { + /* FIXME alloc_page is good enough for now, but actually needs + * to use pre-allocated page pool */ + void *src, *dest; + page = alloc_page(__GFP_HIGHMEM|__GFP_WAIT); + dest = kmap_atomic(page, KM_USER0); + src = kmap_atomic(b->bm_pages[page_nr], KM_USER1); + memcpy(dest, src, PAGE_SIZE); + kunmap_atomic(src, KM_USER1); + kunmap_atomic(dest, KM_USER0); + bm_store_page_idx(page, page_nr); + } else + page = b->bm_pages[page_nr]; + bio->bi_bdev = mdev->ldev->md_bdev; bio->bi_sector = on_disk_sector; - bio_add_page(bio, b->bm_pages[page_nr], len, 0); - bio->bi_private = b; + bio_add_page(bio, page, len, 0); + bio->bi_private = ctx; bio->bi_end_io = bm_async_io_complete; if (drbd_insert_fault(mdev, (rw & WRITE) ? DRBD_FAULT_MD_WR : DRBD_FAULT_MD_RD)) { @@ -841,36 +993,72 @@ static void bm_page_io_async(struct drbd_conf *mdev, struct drbd_bitmap *b, int /* * bm_rw: read/write the whole bitmap from/to its on disk location. */ -static int bm_rw(struct drbd_conf *mdev, int rw) __must_hold(local) +static int bm_rw(struct drbd_conf *mdev, int rw, unsigned lazy_writeout_upper_idx) __must_hold(local) { + struct bm_aio_ctx ctx = + { .flags = lazy_writeout_upper_idx ? BM_AIO_COPY_PAGES : 0 }; struct drbd_bitmap *b = mdev->bitmap; - /* sector_t sector; */ - int bm_words, num_pages, i; + int last_page, i, count = 0; unsigned long now; char ppb[10]; int err = 0; - WARN_ON(!bm_is_locked(b)); - - /* no spinlock here, the drbd_bm_lock should be enough! */ + /* + * We are protected against bitmap disappearing/resizing by holding an + * ldev reference (caller must have called get_ldev()). + * For read/write, we are protected against changes to the bitmap by + * the bitmap lock (see drbd_bitmap_io). + * For lazy writeout, we don't care for ongoing changes to the bitmap, + * as we submit copies of pages anyways. + */ + if (!ctx.flags) + WARN_ON(!bm_is_locked(b)); - bm_words = drbd_bm_words(mdev); - num_pages = (bm_words*sizeof(long) + PAGE_SIZE-1) >> PAGE_SHIFT; + /* because of the "extra long to catch oob access" we allocate in + * drbd_bm_resize, bm_number_of_pages -1 is not necessarily the page + * containing the last _relevant_ bitmap word */ + last_page = bm_word_to_page_idx(b, b->bm_words - 1); now = jiffies; - atomic_set(&b->bm_async_io, num_pages); - __clear_bit(BM_MD_IO_ERROR, &b->bm_flags); + ctx.mdev = mdev; + atomic_set(&ctx.in_flight, 1); /* one extra ref */ + init_waitqueue_head(&ctx.io_wait); + ctx.error = 0; /* let the layers below us try to merge these bios... */ - for (i = 0; i < num_pages; i++) - bm_page_io_async(mdev, b, i, rw); + for (i = 0; i <= last_page; i++) { + /* ignore completely unchanged pages */ + if (lazy_writeout_upper_idx && i == lazy_writeout_upper_idx) + break; + if (rw & WRITE) { + if (bm_test_page_unchanged(b->bm_pages[i])) { + dynamic_dev_dbg(DEV, "skipped bm write for idx %u\n", i); + continue; + } + /* during lazy writeout, + * ignore those pages not marked for lazy writeout. */ + if (lazy_writeout_upper_idx && + !bm_test_page_lazy_writeout(b->bm_pages[i])) { + dynamic_dev_dbg(DEV, "skipped bm lazy write for idx %u\n", i); + continue; + } + } + atomic_inc(&ctx.in_flight); + bm_page_io_async(&ctx, i, rw); + ++count; + cond_resched(); + } - wait_event(b->bm_io_wait, atomic_read(&b->bm_async_io) == 0); + atomic_dec(&ctx.in_flight); /* drop the extra ref */ + wait_event(ctx.io_wait, atomic_read(&ctx.in_flight) == 0); + dev_info(DEV, "bitmap %s of %u pages took %lu jiffies\n", + rw == WRITE ? "WRITE" : "READ", + count, jiffies - now); - if (test_bit(BM_MD_IO_ERROR, &b->bm_flags)) { + if (ctx.error) { dev_alert(DEV, "we had at least one MD IO ERROR during bitmap IO\n"); drbd_chk_io_error(mdev, 1, true); - err = -EIO; + err = -EIO; /* ctx.error ? */ } now = jiffies; @@ -895,55 +1083,63 @@ static int bm_rw(struct drbd_conf *mdev, int rw) __must_hold(local) */ int drbd_bm_read(struct drbd_conf *mdev) __must_hold(local) { - return bm_rw(mdev, READ); + return bm_rw(mdev, READ, 0); } /** * drbd_bm_write() - Write the whole bitmap to its on disk location. * @mdev: DRBD device. + * + * Will only write pages that have changed since last IO. */ int drbd_bm_write(struct drbd_conf *mdev) __must_hold(local) { - return bm_rw(mdev, WRITE); + return bm_rw(mdev, WRITE, 0); } /** - * drbd_bm_write_sect: Writes a 512 (MD_SECTOR_SIZE) byte piece of the bitmap + * drbd_bm_lazy_write_out() - Write bitmap pages 0 to @upper_idx-1, if they have changed. * @mdev: DRBD device. - * @enr: Extent number in the resync lru (happens to be sector offset) + * @upper_idx: 0: write all changed pages; +ve: page index to stop scanning for changed pages + */ +int drbd_bm_write_lazy(struct drbd_conf *mdev, unsigned upper_idx) __must_hold(local) +{ + return bm_rw(mdev, WRITE, upper_idx); +} + + +/** + * drbd_bm_write_page: Writes a PAGE_SIZE aligned piece of bitmap + * @mdev: DRBD device. + * @idx: bitmap page index * - * The BM_EXT_SIZE is on purpose exactly the amount of the bitmap covered - * by a single sector write. Therefore enr == sector offset from the - * start of the bitmap. + * We don't want to special case on logical_block_size of the underlaying + * device, so we submit PAGE_SIZE aligned pieces containing the requested enr. + * Note that on "most" systems, PAGE_SIZE is 4k. */ -int drbd_bm_write_sect(struct drbd_conf *mdev, unsigned long enr) __must_hold(local) +int drbd_bm_write_page(struct drbd_conf *mdev, unsigned int idx) __must_hold(local) { - sector_t on_disk_sector = enr + mdev->ldev->md.md_offset - + mdev->ldev->md.bm_offset; - int bm_words, num_words, offset; - int err = 0; + struct bm_aio_ctx ctx = { .flags = BM_AIO_COPY_PAGES, }; - mutex_lock(&mdev->md_io_mutex); - bm_words = drbd_bm_words(mdev); - offset = S2W(enr); /* word offset into bitmap */ - num_words = min(S2W(1), bm_words - offset); - if (num_words < S2W(1)) - memset(page_address(mdev->md_io_page), 0, MD_SECTOR_SIZE); - drbd_bm_get_lel(mdev, offset, num_words, - page_address(mdev->md_io_page)); - if (!drbd_md_sync_page_io(mdev, mdev->ldev, on_disk_sector, WRITE)) { - int i; - err = -EIO; - dev_err(DEV, "IO ERROR writing bitmap sector %lu " - "(meta-disk sector %llus)\n", - enr, (unsigned long long)on_disk_sector); - drbd_chk_io_error(mdev, 1, true); - for (i = 0; i < AL_EXT_PER_BM_SECT; i++) - drbd_bm_ALe_set_all(mdev, enr*AL_EXT_PER_BM_SECT+i); + if (bm_test_page_unchanged(mdev->bitmap->bm_pages[idx])) { + dev_info(DEV, "skipped bm page write for idx %u\n", idx); + return 0; } + + ctx.mdev = mdev; + atomic_set(&ctx.in_flight, 1); + init_waitqueue_head(&ctx.io_wait); + + bm_page_io_async(&ctx, idx, WRITE_SYNC); + wait_event(ctx.io_wait, atomic_read(&ctx.in_flight) == 0); + + if (ctx.error) + drbd_chk_io_error(mdev, 1, true); + /* that should force detach, so the in memory bitmap will be + * gone in a moment as well. */ + mdev->bm_writ_cnt++; - mutex_unlock(&mdev->md_io_mutex); - return err; + return ctx.error; } /* NOTE @@ -965,10 +1161,9 @@ static unsigned long __bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo, dev_err(DEV, "bm_fo=%lu bm_bits=%lu\n", bm_fo, b->bm_bits); } else { while (bm_fo < b->bm_bits) { - unsigned long offset; - bit_offset = bm_fo & ~BPP_MASK; /* bit offset of the page */ - offset = bit_offset >> LN2_BPL; /* word offset of the page */ - p_addr = __bm_map_paddr(b, offset, km); + /* bit offset of the first bit in the page */ + bit_offset = bm_fo & ~BPP_MASK; + p_addr = __bm_map_pidx(b, bm_bit_to_page_idx(b, bm_fo), km); if (find_zero_bit) i = generic_find_next_zero_le_bit(p_addr, PAGE_SIZE*8, bm_fo & BPP_MASK); @@ -1048,8 +1243,9 @@ static int __bm_change_bits_to(struct drbd_conf *mdev, const unsigned long s, struct drbd_bitmap *b = mdev->bitmap; unsigned long *p_addr = NULL; unsigned long bitnr; - unsigned long last_page_nr = -1UL; + unsigned int last_page_nr = -1U; int c = 0; + int changed_total = 0; if (e >= b->bm_bits) { dev_err(DEV, "ASSERT FAILED: bit_s=%lu bit_e=%lu bm_bits=%lu\n", @@ -1057,12 +1253,17 @@ static int __bm_change_bits_to(struct drbd_conf *mdev, const unsigned long s, e = b->bm_bits ? b->bm_bits -1 : 0; } for (bitnr = s; bitnr <= e; bitnr++) { - unsigned long offset = bitnr>>LN2_BPL; - unsigned long page_nr = offset >> (PAGE_SHIFT - LN2_BPL + 3); + unsigned int page_nr = bm_bit_to_page_idx(b, bitnr); if (page_nr != last_page_nr) { if (p_addr) __bm_unmap(p_addr, km); - p_addr = __bm_map_paddr(b, offset, km); + if (c < 0) + bm_set_page_lazy_writeout(b->bm_pages[last_page_nr]); + else if (c > 0) + bm_set_page_need_writeout(b->bm_pages[last_page_nr]); + changed_total += c; + c = 0; + p_addr = __bm_map_pidx(b, page_nr, km); last_page_nr = page_nr; } if (val) @@ -1072,8 +1273,13 @@ static int __bm_change_bits_to(struct drbd_conf *mdev, const unsigned long s, } if (p_addr) __bm_unmap(p_addr, km); - b->bm_set += c; - return c; + if (c < 0) + bm_set_page_lazy_writeout(b->bm_pages[last_page_nr]); + else if (c > 0) + bm_set_page_need_writeout(b->bm_pages[last_page_nr]); + changed_total += c; + b->bm_set += changed_total; + return changed_total; } /* returns number of bits actually changed. @@ -1211,8 +1417,7 @@ int drbd_bm_test_bit(struct drbd_conf *mdev, const unsigned long bitnr) if (bm_is_locked(b)) bm_print_lock_info(mdev); if (bitnr < b->bm_bits) { - unsigned long offset = bitnr>>LN2_BPL; - p_addr = bm_map_paddr(b, offset); + p_addr = bm_map_pidx(b, bm_bit_to_page_idx(b, bitnr)); i = generic_test_le_bit(bitnr & BPP_MASK, p_addr) ? 1 : 0; bm_unmap(p_addr); } else if (bitnr == b->bm_bits) { @@ -1231,10 +1436,10 @@ int drbd_bm_count_bits(struct drbd_conf *mdev, const unsigned long s, const unsi { unsigned long flags; struct drbd_bitmap *b = mdev->bitmap; - unsigned long *p_addr = NULL, page_nr = -1; + unsigned long *p_addr = NULL; unsigned long bitnr; + unsigned int page_nr = -1U; int c = 0; - size_t w; /* If this is called without a bitmap, that is a bug. But just to be * robust in case we screwed up elsewhere, in that case pretend there @@ -1247,12 +1452,12 @@ int drbd_bm_count_bits(struct drbd_conf *mdev, const unsigned long s, const unsi if (bm_is_locked(b)) bm_print_lock_info(mdev); for (bitnr = s; bitnr <= e; bitnr++) { - w = bitnr >> LN2_BPL; - if (page_nr != w >> (PAGE_SHIFT - LN2_BPL + 3)) { - page_nr = w >> (PAGE_SHIFT - LN2_BPL + 3); + unsigned int idx = bm_bit_to_page_idx(b, bitnr); + if (page_nr != idx) { + page_nr = idx; if (p_addr) bm_unmap(p_addr); - p_addr = bm_map_paddr(b, w); + p_addr = bm_map_pidx(b, idx); } ERR_IF (bitnr >= b->bm_bits) { dev_err(DEV, "bitnr=%lu bm_bits=%lu\n", bitnr, b->bm_bits); @@ -1300,7 +1505,7 @@ int drbd_bm_e_weight(struct drbd_conf *mdev, unsigned long enr) count = 0; if (s < b->bm_words) { int n = e-s; - p_addr = bm_map_paddr(b, s); + p_addr = bm_map_pidx(b, bm_word_to_page_idx(b, s)); bm = p_addr + MLPP(s); while (n--) count += hweight_long(*bm++); @@ -1335,7 +1540,7 @@ unsigned long drbd_bm_ALe_set_all(struct drbd_conf *mdev, unsigned long al_enr) count = 0; if (s < b->bm_words) { i = do_now = e-s; - p_addr = bm_map_paddr(b, s); + p_addr = bm_map_pidx(b, bm_word_to_page_idx(b, s)); bm = p_addr + MLPP(s); while (i--) { count += hweight_long(*bm); diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h index f6da48bb8c70..74cc50a21822 100644 --- a/drivers/block/drbd/drbd_int.h +++ b/drivers/block/drbd/drbd_int.h @@ -833,7 +833,7 @@ enum { CRASHED_PRIMARY, /* This node was a crashed primary. * Gets cleared when the state.conn * goes into C_CONNECTED state. */ - WRITE_BM_AFTER_RESYNC, /* A kmalloc() during resync failed */ + NO_BARRIER_SUPP, /* underlying block device doesn't implement barriers */ CONSIDER_RESYNC, MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */ @@ -1428,7 +1428,7 @@ extern void _drbd_bm_set_bits(struct drbd_conf *mdev, const unsigned long s, const unsigned long e); extern int drbd_bm_test_bit(struct drbd_conf *mdev, unsigned long bitnr); extern int drbd_bm_e_weight(struct drbd_conf *mdev, unsigned long enr); -extern int drbd_bm_write_sect(struct drbd_conf *mdev, unsigned long enr) __must_hold(local); +extern int drbd_bm_write_page(struct drbd_conf *mdev, unsigned int idx) __must_hold(local); extern int drbd_bm_read(struct drbd_conf *mdev) __must_hold(local); extern int drbd_bm_write(struct drbd_conf *mdev) __must_hold(local); extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf *mdev, @@ -1446,7 +1446,7 @@ extern int drbd_bm_rs_done(struct drbd_conf *mdev); /* for receive_bitmap */ extern void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset, size_t number, unsigned long *buffer); -/* for _drbd_send_bitmap and drbd_bm_write_sect */ +/* for _drbd_send_bitmap */ extern void drbd_bm_get_lel(struct drbd_conf *mdev, size_t offset, size_t number, unsigned long *buffer); @@ -1641,7 +1641,6 @@ extern int __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector, #define drbd_set_out_of_sync(mdev, sector, size) \ __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__) extern void drbd_al_apply_to_bm(struct drbd_conf *mdev); -extern void drbd_al_to_on_disk_bm(struct drbd_conf *mdev); extern void drbd_al_shrink(struct drbd_conf *mdev); diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index 67fffad213ec..57ed7181742d 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -1289,6 +1289,26 @@ static void abw_start_sync(struct drbd_conf *mdev, int rv) } } +int drbd_bitmap_io_from_worker(struct drbd_conf *mdev, int (*io_fn)(struct drbd_conf *), char *why) +{ + int rv; + + D_ASSERT(current == mdev->worker.task); + + /* open coded non-blocking drbd_suspend_io(mdev); */ + set_bit(SUSPEND_IO, &mdev->flags); + if (!is_susp(mdev->state)) + D_ASSERT(atomic_read(&mdev->ap_bio_cnt) == 0); + + drbd_bm_lock(mdev, why); + rv = io_fn(mdev); + drbd_bm_unlock(mdev); + + drbd_resume_io(mdev); + + return rv; +} + /** * after_state_ch() - Perform after state change actions that may sleep * @mdev: DRBD device. @@ -1404,7 +1424,12 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, /* D_DISKLESS Peer becomes secondary */ if (os.peer == R_PRIMARY && ns.peer == R_SECONDARY) - drbd_al_to_on_disk_bm(mdev); + drbd_bitmap_io_from_worker(mdev, &drbd_bm_write, "demote diskless peer"); + put_ldev(mdev); + } + + if (os.role == R_PRIMARY && ns.role == R_SECONDARY && get_ldev(mdev)) { + drbd_bitmap_io_from_worker(mdev, &drbd_bm_write, "demote"); put_ldev(mdev); } diff --git a/drivers/block/drbd/drbd_nl.c b/drivers/block/drbd/drbd_nl.c index 2f0724982143..77dc022eaf6b 100644 --- a/drivers/block/drbd/drbd_nl.c +++ b/drivers/block/drbd/drbd_nl.c @@ -407,10 +407,8 @@ drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role, int force) } } - if ((new_role == R_SECONDARY) && get_ldev(mdev)) { - drbd_al_to_on_disk_bm(mdev); - put_ldev(mdev); - } + /* writeout of activity log covered areas of the bitmap + * to stable storage done in after state change already */ if (mdev->state.conn >= C_WF_REPORT_PARAMS) { /* if this was forced, we should consider sync */ @@ -1174,7 +1172,10 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp if (cp_discovered) { drbd_al_apply_to_bm(mdev); - drbd_al_to_on_disk_bm(mdev); + if (drbd_bitmap_io(mdev, &drbd_bm_write, "crashed primary apply AL")) { + retcode = ERR_IO_MD_DISK; + goto force_diskless_dec; + } } if (_drbd_bm_total_weight(mdev) == drbd_bm_bits(mdev)) diff --git a/drivers/block/drbd/drbd_worker.c b/drivers/block/drbd/drbd_worker.c index 9fe3e890da0f..d17f2ed777ce 100644 --- a/drivers/block/drbd/drbd_worker.c +++ b/drivers/block/drbd/drbd_worker.c @@ -907,10 +907,8 @@ out: drbd_md_sync(mdev); - if (test_and_clear_bit(WRITE_BM_AFTER_RESYNC, &mdev->flags)) { - dev_info(DEV, "Writing the whole bitmap\n"); - drbd_queue_bitmap_io(mdev, &drbd_bm_write, NULL, "write from resync_finished"); - } + dev_info(DEV, "Writing changed bitmap pages\n"); + drbd_queue_bitmap_io(mdev, &drbd_bm_write, NULL, "write from resync_finished"); if (khelper_cmd) drbd_khelper(mdev, khelper_cmd); @@ -1127,7 +1125,6 @@ void drbd_ov_oos_found(struct drbd_conf *mdev, sector_t sector, int size) mdev->ov_last_oos_size = size>>9; } drbd_set_out_of_sync(mdev, sector, size); - set_bit(WRITE_BM_AFTER_RESYNC, &mdev->flags); } int w_e_end_ov_reply(struct drbd_conf *mdev, struct drbd_work *w, int cancel) -- cgit v1.2.3 From 4b0715f09655e76ca24c35a9e25e7c464c2f7346 Mon Sep 17 00:00:00 2001 From: Lars Ellenberg Date: Tue, 14 Dec 2010 15:13:04 +0100 Subject: drbd: allow petabyte storage on 64bit arch Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_bitmap.c | 170 ++++++++++++++++++++++++--------------- drivers/block/drbd/drbd_int.h | 41 ++++++---- drivers/block/drbd/drbd_nl.c | 8 +- drivers/block/drbd/drbd_proc.c | 6 +- drivers/block/drbd/drbd_worker.c | 2 +- 5 files changed, 142 insertions(+), 85 deletions(-) (limited to 'drivers/block/drbd/drbd_bitmap.c') diff --git a/drivers/block/drbd/drbd_bitmap.c b/drivers/block/drbd/drbd_bitmap.c index 72cd41a96ef9..0e31e573af72 100644 --- a/drivers/block/drbd/drbd_bitmap.c +++ b/drivers/block/drbd/drbd_bitmap.c @@ -37,10 +37,46 @@ * convention: * function name drbd_bm_... => used elsewhere, "public". * function name bm_... => internal to implementation, "private". + */ + - * Note that since find_first_bit returns int, at the current granularity of - * the bitmap (4KB per byte), this implementation "only" supports up to - * 1<<(32+12) == 16 TB... +/* + * LIMITATIONS: + * We want to support >= peta byte of backend storage, while for now still using + * a granularity of one bit per 4KiB of storage. + * 1 << 50 bytes backend storage (1 PiB) + * 1 << (50 - 12) bits needed + * 38 --> we need u64 to index and count bits + * 1 << (38 - 3) bitmap bytes needed + * 35 --> we still need u64 to index and count bytes + * (that's 32 GiB of bitmap for 1 PiB storage) + * 1 << (35 - 2) 32bit longs needed + * 33 --> we'd even need u64 to index and count 32bit long words. + * 1 << (35 - 3) 64bit longs needed + * 32 --> we could get away with a 32bit unsigned int to index and count + * 64bit long words, but I rather stay with unsigned long for now. + * We probably should neither count nor point to bytes or long words + * directly, but either by bitnumber, or by page index and offset. + * 1 << (35 - 12) + * 22 --> we need that much 4KiB pages of bitmap. + * 1 << (22 + 3) --> on a 64bit arch, + * we need 32 MiB to store the array of page pointers. + * + * Because I'm lazy, and because the resulting patch was too large, too ugly + * and still incomplete, on 32bit we still "only" support 16 TiB (minus some), + * (1 << 32) bits * 4k storage. + * + + * bitmap storage and IO: + * Bitmap is stored little endian on disk, and is kept little endian in + * core memory. Currently we still hold the full bitmap in core as long + * as we are "attached" to a local disk, which at 32 GiB for 1PiB storage + * seems excessive. + * + * We plan to reduce the amount of in-core bitmap pages by pageing them in + * and out against their on-disk location as necessary, but need to make + * sure we don't cause too much meta data IO, and must not deadlock in + * tight memory situations. This needs some more work. */ /* @@ -56,13 +92,9 @@ struct drbd_bitmap { struct page **bm_pages; spinlock_t bm_lock; - /* WARNING unsigned long bm_*: - * 32bit number of bit offset is just enough for 512 MB bitmap. - * it will blow up if we make the bitmap bigger... - * not that it makes much sense to have a bitmap that large, - * rather change the granularity to 16k or 64k or something. - * (that implies other problems, however...) - */ + + /* see LIMITATIONS: above */ + unsigned long bm_set; /* nr of set bits; THINK maybe atomic_t? */ unsigned long bm_bits; size_t bm_words; @@ -517,43 +549,39 @@ static void bm_set_surplus(struct drbd_bitmap *b) bm_unmap(p_addr); } +/* you better not modify the bitmap while this is running, + * or its results will be stale */ static unsigned long bm_count_bits(struct drbd_bitmap *b) { - unsigned long *p_addr, *bm, offset = 0; + unsigned long *p_addr; unsigned long bits = 0; - unsigned long i, do_now; - unsigned long words; - - /* due to 64bit alignment, the last long on a 32bit arch - * may be not used at all. The last used long will likely - * be only partially used, always. Don't count those bits, - * but mask them out. */ - words = (b->bm_bits + BITS_PER_LONG - 1) >> LN2_BPL; - - while (offset < words) { - i = do_now = min_t(size_t, words-offset, LWPP); - p_addr = __bm_map_pidx(b, bm_word_to_page_idx(b, offset), KM_USER0); - bm = p_addr + MLPP(offset); - while (i--) { - bits += hweight_long(*bm++); - } - offset += do_now; - if (offset == words) { - /* last word may only be partially used, - * see also bm_clear_surplus. */ - i = (1UL << (b->bm_bits & (BITS_PER_LONG-1))) -1; - if (i) { - bits -= hweight_long(p_addr[do_now-1] & ~i); - p_addr[do_now-1] &= i; - } - /* 32bit arch, may have an unused padding long */ - if (words != b->bm_words) - p_addr[do_now] = 0; - } + unsigned long mask = (1UL << (b->bm_bits & BITS_PER_LONG_MASK)) -1; + int idx, last_page, i, last_word; + + /* because of the "extra long to catch oob access" we allocate in + * drbd_bm_resize, bm_number_of_pages -1 is not necessarily the page + * containing the last _relevant_ bitmap word */ + last_page = bm_bit_to_page_idx(b, b->bm_bits-1); + + /* all but last page */ + for (idx = 0; idx < last_page; idx++) { + p_addr = __bm_map_pidx(b, idx, KM_USER0); + for (i = 0; i < LWPP; i++) + bits += hweight_long(p_addr[i]); __bm_unmap(p_addr, KM_USER0); cond_resched(); } - + /* last (or only) page */ + last_word = ((b->bm_bits - 1) & BITS_PER_PAGE_MASK) >> LN2_BPL; + p_addr = __bm_map_pidx(b, idx, KM_USER0); + for (i = 0; i < last_word; i++) + bits += hweight_long(p_addr[i]); + p_addr[last_word] &= cpu_to_lel(mask); + bits += hweight_long(p_addr[last_word]); + /* 32bit arch, may have an unused padding long */ + if (BITS_PER_LONG == 32 && (last_word & 1) == 0) + p_addr[last_word+1] = 0; + __bm_unmap(p_addr, KM_USER0); return bits; } @@ -564,8 +592,6 @@ static void bm_memset(struct drbd_bitmap *b, size_t offset, int c, size_t len) unsigned int idx; size_t do_now, end; -#define BM_SECTORS_PER_BIT (BM_BLOCK_SIZE/512) - end = offset + len; if (end > b->bm_words) { @@ -645,8 +671,14 @@ int drbd_bm_resize(struct drbd_conf *mdev, sector_t capacity, int set_new_bits) words = ALIGN(bits, 64) >> LN2_BPL; if (get_ldev(mdev)) { - D_ASSERT((u64)bits <= (((u64)mdev->ldev->md.md_size_sect-MD_BM_OFFSET) << 12)); + u64 bits_on_disk = ((u64)mdev->ldev->md.md_size_sect-MD_BM_OFFSET) << 12; put_ldev(mdev); + if (bits > bits_on_disk) { + dev_info(DEV, "bits = %lu\n", bits); + dev_info(DEV, "bits_on_disk = %llu\n", bits_on_disk); + err = -ENOSPC; + goto out; + } } /* one extra long to catch off by one errors */ @@ -1113,9 +1145,12 @@ int drbd_bm_write_lazy(struct drbd_conf *mdev, unsigned upper_idx) __must_hold(l * @mdev: DRBD device. * @idx: bitmap page index * - * We don't want to special case on logical_block_size of the underlaying - * device, so we submit PAGE_SIZE aligned pieces containing the requested enr. + * We don't want to special case on logical_block_size of the backend device, + * so we submit PAGE_SIZE aligned pieces. * Note that on "most" systems, PAGE_SIZE is 4k. + * + * In case this becomes an issue on systems with larger PAGE_SIZE, + * we may want to change this again to write 4k aligned 4k pieces. */ int drbd_bm_write_page(struct drbd_conf *mdev, unsigned int idx) __must_hold(local) { @@ -1144,52 +1179,57 @@ int drbd_bm_write_page(struct drbd_conf *mdev, unsigned int idx) __must_hold(loc /* NOTE * find_first_bit returns int, we return unsigned long. - * should not make much difference anyways, but ... + * For this to work on 32bit arch with bitnumbers > (1<<32), + * we'd need to return u64, and get a whole lot of other places + * fixed where we still use unsigned long. * * this returns a bit number, NOT a sector! */ -#define BPP_MASK ((1UL << (PAGE_SHIFT+3)) - 1) static unsigned long __bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo, const int find_zero_bit, const enum km_type km) { struct drbd_bitmap *b = mdev->bitmap; - unsigned long i = -1UL; unsigned long *p_addr; - unsigned long bit_offset; /* bit offset of the mapped page. */ + unsigned long bit_offset; + unsigned i; + if (bm_fo > b->bm_bits) { dev_err(DEV, "bm_fo=%lu bm_bits=%lu\n", bm_fo, b->bm_bits); + bm_fo = DRBD_END_OF_BITMAP; } else { while (bm_fo < b->bm_bits) { /* bit offset of the first bit in the page */ - bit_offset = bm_fo & ~BPP_MASK; + bit_offset = bm_fo & ~BITS_PER_PAGE_MASK; p_addr = __bm_map_pidx(b, bm_bit_to_page_idx(b, bm_fo), km); if (find_zero_bit) - i = generic_find_next_zero_le_bit(p_addr, PAGE_SIZE*8, bm_fo & BPP_MASK); + i = generic_find_next_zero_le_bit(p_addr, + PAGE_SIZE*8, bm_fo & BITS_PER_PAGE_MASK); else - i = generic_find_next_le_bit(p_addr, PAGE_SIZE*8, bm_fo & BPP_MASK); + i = generic_find_next_le_bit(p_addr, + PAGE_SIZE*8, bm_fo & BITS_PER_PAGE_MASK); __bm_unmap(p_addr, km); if (i < PAGE_SIZE*8) { - i = bit_offset + i; - if (i >= b->bm_bits) + bm_fo = bit_offset + i; + if (bm_fo >= b->bm_bits) break; goto found; } bm_fo = bit_offset + PAGE_SIZE*8; } - i = -1UL; + bm_fo = DRBD_END_OF_BITMAP; } found: - return i; + return bm_fo; } static unsigned long bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo, const int find_zero_bit) { struct drbd_bitmap *b = mdev->bitmap; - unsigned long i = -1UL; + unsigned long i = DRBD_END_OF_BITMAP; ERR_IF(!b) return i; ERR_IF(!b->bm_pages) return i; @@ -1267,9 +1307,9 @@ static int __bm_change_bits_to(struct drbd_conf *mdev, const unsigned long s, last_page_nr = page_nr; } if (val) - c += (0 == generic___test_and_set_le_bit(bitnr & BPP_MASK, p_addr)); + c += (0 == generic___test_and_set_le_bit(bitnr & BITS_PER_PAGE_MASK, p_addr)); else - c -= (0 != generic___test_and_clear_le_bit(bitnr & BPP_MASK, p_addr)); + c -= (0 != generic___test_and_clear_le_bit(bitnr & BITS_PER_PAGE_MASK, p_addr)); } if (p_addr) __bm_unmap(p_addr, km); @@ -1418,7 +1458,7 @@ int drbd_bm_test_bit(struct drbd_conf *mdev, const unsigned long bitnr) bm_print_lock_info(mdev); if (bitnr < b->bm_bits) { p_addr = bm_map_pidx(b, bm_bit_to_page_idx(b, bitnr)); - i = generic_test_le_bit(bitnr & BPP_MASK, p_addr) ? 1 : 0; + i = generic_test_le_bit(bitnr & BITS_PER_PAGE_MASK, p_addr) ? 1 : 0; bm_unmap(p_addr); } else if (bitnr == b->bm_bits) { i = -1; @@ -1517,13 +1557,15 @@ int drbd_bm_e_weight(struct drbd_conf *mdev, unsigned long enr) return count; } -/* set all bits covered by the AL-extent al_enr */ +/* Set all bits covered by the AL-extent al_enr. + * Returns number of bits changed. */ unsigned long drbd_bm_ALe_set_all(struct drbd_conf *mdev, unsigned long al_enr) { struct drbd_bitmap *b = mdev->bitmap; unsigned long *p_addr, *bm; unsigned long weight; - int count, s, e, i, do_now; + unsigned long s, e; + int count, i, do_now; ERR_IF(!b) return 0; ERR_IF(!b->bm_pages) return 0; @@ -1552,7 +1594,7 @@ unsigned long drbd_bm_ALe_set_all(struct drbd_conf *mdev, unsigned long al_enr) if (e == b->bm_words) b->bm_set -= bm_clear_surplus(b); } else { - dev_err(DEV, "start offset (%d) too large in drbd_bm_ALe_set_all\n", s); + dev_err(DEV, "start offset (%lu) too large in drbd_bm_ALe_set_all\n", s); } weight = b->bm_set - weight; spin_unlock_irq(&b->bm_lock); diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h index 74cc50a21822..5a2d0ec72b34 100644 --- a/drivers/block/drbd/drbd_int.h +++ b/drivers/block/drbd/drbd_int.h @@ -1003,9 +1003,9 @@ struct drbd_conf { struct hlist_head *tl_hash; unsigned int tl_hash_s; - /* blocks to sync in this run [unit BM_BLOCK_SIZE] */ + /* blocks to resync in this run [unit BM_BLOCK_SIZE] */ unsigned long rs_total; - /* number of sync IOs that failed in this run */ + /* number of resync blocks that failed in this run */ unsigned long rs_failed; /* Syncer's start time [unit jiffies] */ unsigned long rs_start; @@ -1399,7 +1399,9 @@ struct bm_extent { * you should use 64bit OS for that much storage, anyways. */ #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff) #else -#define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0x1LU << 32) +/* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */ +#define DRBD_MAX_SECTORS_FLEX (1UL << 51) +/* corresponds to (1UL << 38) bits right now. */ #endif #endif @@ -1419,11 +1421,15 @@ extern int drbd_bm_resize(struct drbd_conf *mdev, sector_t sectors, int set_new extern void drbd_bm_cleanup(struct drbd_conf *mdev); extern void drbd_bm_set_all(struct drbd_conf *mdev); extern void drbd_bm_clear_all(struct drbd_conf *mdev); +/* set/clear/test only a few bits at a time */ extern int drbd_bm_set_bits( struct drbd_conf *mdev, unsigned long s, unsigned long e); extern int drbd_bm_clear_bits( struct drbd_conf *mdev, unsigned long s, unsigned long e); -/* bm_set_bits variant for use while holding drbd_bm_lock */ +extern int drbd_bm_count_bits( + struct drbd_conf *mdev, const unsigned long s, const unsigned long e); +/* bm_set_bits variant for use while holding drbd_bm_lock, + * may process the whole bitmap in one go */ extern void _drbd_bm_set_bits(struct drbd_conf *mdev, const unsigned long s, const unsigned long e); extern int drbd_bm_test_bit(struct drbd_conf *mdev, unsigned long bitnr); @@ -1436,6 +1442,8 @@ extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf *mdev, extern size_t drbd_bm_words(struct drbd_conf *mdev); extern unsigned long drbd_bm_bits(struct drbd_conf *mdev); extern sector_t drbd_bm_capacity(struct drbd_conf *mdev); + +#define DRBD_END_OF_BITMAP (~(unsigned long)0) extern unsigned long drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo); /* bm_find_next variants for use while you hold drbd_bm_lock() */ extern unsigned long _drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo); @@ -1452,8 +1460,6 @@ extern void drbd_bm_get_lel(struct drbd_conf *mdev, size_t offset, extern void drbd_bm_lock(struct drbd_conf *mdev, char *why); extern void drbd_bm_unlock(struct drbd_conf *mdev); - -extern int drbd_bm_count_bits(struct drbd_conf *mdev, const unsigned long s, const unsigned long e); /* drbd_main.c */ extern struct kmem_cache *drbd_request_cache; @@ -2158,10 +2164,8 @@ extern int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins) static inline void drbd_get_syncer_progress(struct drbd_conf *mdev, unsigned long *bits_left, unsigned int *per_mil_done) { - /* - * this is to break it at compile time when we change that - * (we may feel 4TB maximum storage per drbd is not enough) - */ + /* this is to break it at compile time when we change that, in case we + * want to support more than (1<<32) bits on a 32bit arch. */ typecheck(unsigned long, mdev->rs_total); /* note: both rs_total and rs_left are in bits, i.e. in @@ -2186,10 +2190,19 @@ static inline void drbd_get_syncer_progress(struct drbd_conf *mdev, *bits_left, mdev->rs_total, mdev->rs_failed); *per_mil_done = 0; } else { - /* make sure the calculation happens in long context */ - unsigned long tmp = 1000UL - - (*bits_left >> 10)*1000UL - / ((mdev->rs_total >> 10) + 1UL); + /* Make sure the division happens in long context. + * We allow up to one petabyte storage right now, + * at a granularity of 4k per bit that is 2**38 bits. + * After shift right and multiplication by 1000, + * this should still fit easily into a 32bit long, + * so we don't need a 64bit division on 32bit arch. + * Note: currently we don't support such large bitmaps on 32bit + * arch anyways, but no harm done to be prepared for it here. + */ + unsigned int shift = mdev->rs_total >= (1ULL << 32) ? 16 : 10; + unsigned long left = *bits_left >> shift; + unsigned long total = 1UL + (mdev->rs_total >> shift); + unsigned long tmp = 1000UL - left * 1000UL/total; *per_mil_done = tmp; } } diff --git a/drivers/block/drbd/drbd_nl.c b/drivers/block/drbd/drbd_nl.c index 77dc022eaf6b..a46bc0287e21 100644 --- a/drivers/block/drbd/drbd_nl.c +++ b/drivers/block/drbd/drbd_nl.c @@ -527,17 +527,19 @@ static void drbd_md_set_sector_offsets(struct drbd_conf *mdev, } } +/* input size is expected to be in KB */ char *ppsize(char *buf, unsigned long long size) { - /* Needs 9 bytes at max. */ + /* Needs 9 bytes at max including trailing NUL: + * -1ULL ==> "16384 EB" */ static char units[] = { 'K', 'M', 'G', 'T', 'P', 'E' }; int base = 0; - while (size >= 10000) { + while (size >= 10000 && base < sizeof(units)-1) { /* shift + round */ size = (size >> 10) + !!(size & (1<<9)); base++; } - sprintf(buf, "%lu %cB", (long)size, units[base]); + sprintf(buf, "%u %cB", (unsigned)size, units[base]); return buf; } diff --git a/drivers/block/drbd/drbd_proc.c b/drivers/block/drbd/drbd_proc.c index efba62cd2e58..2959cdfb77f5 100644 --- a/drivers/block/drbd/drbd_proc.c +++ b/drivers/block/drbd/drbd_proc.c @@ -91,9 +91,9 @@ static void drbd_syncer_progress(struct drbd_conf *mdev, struct seq_file *seq) seq_printf(seq, "sync'ed:"); seq_printf(seq, "%3u.%u%% ", res / 10, res % 10); - /* if more than 1 GB display in MB */ - if (mdev->rs_total > 0x100000L) - seq_printf(seq, "(%lu/%lu)M\n\t", + /* if more than a few GB, display in MB */ + if (mdev->rs_total > (4UL << (30 - BM_BLOCK_SHIFT))) + seq_printf(seq, "(%lu/%lu)M", (unsigned long) Bit2KB(rs_left >> 10), (unsigned long) Bit2KB(mdev->rs_total >> 10)); else diff --git a/drivers/block/drbd/drbd_worker.c b/drivers/block/drbd/drbd_worker.c index d17f2ed777ce..be46084c254e 100644 --- a/drivers/block/drbd/drbd_worker.c +++ b/drivers/block/drbd/drbd_worker.c @@ -577,7 +577,7 @@ next_sector: size = BM_BLOCK_SIZE; bit = drbd_bm_find_next(mdev, mdev->bm_resync_fo); - if (bit == -1UL) { + if (bit == DRBD_END_OF_BITMAP) { mdev->bm_resync_fo = drbd_bm_bits(mdev); mdev->resync_work.cb = w_resync_inactive; put_ldev(mdev); -- cgit v1.2.3 From 6850c4421481139dc2cf982358e79c833a50d73c Mon Sep 17 00:00:00 2001 From: Lars Ellenberg Date: Thu, 16 Dec 2010 00:32:38 +0100 Subject: drbd: get rid of unused debug code Long time ago, we had paranoia code in the bitmap that allocated one extra word, assigned a magic value, and checked on every occasion that the magic value was still unchanged. That debug code is unused, the extra long word complicates code a bit. Get rid of it. Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_bitmap.c | 39 +++++++++------------------------------ 1 file changed, 9 insertions(+), 30 deletions(-) (limited to 'drivers/block/drbd/drbd_bitmap.c') diff --git a/drivers/block/drbd/drbd_bitmap.c b/drivers/block/drbd/drbd_bitmap.c index 0e31e573af72..09d208b858e7 100644 --- a/drivers/block/drbd/drbd_bitmap.c +++ b/drivers/block/drbd/drbd_bitmap.c @@ -488,10 +488,7 @@ static int bm_clear_surplus(struct drbd_bitmap *b) * on disk and in core memory alike */ mask = cpu_to_lel(mask); - /* because of the "extra long to catch oob access" we allocate in - * drbd_bm_resize, bm_number_of_pages -1 is not necessarily the page - * containing the last _relevant_ bitmap word */ - p_addr = bm_map_pidx(b, bm_bit_to_page_idx(b, b->bm_bits - 1)); + p_addr = bm_map_pidx(b, b->bm_number_of_pages - 1); bm = p_addr + (tmp/BITS_PER_LONG); if (mask) { /* If mask != 0, we are not exactly aligned, so bm now points @@ -527,10 +524,7 @@ static void bm_set_surplus(struct drbd_bitmap *b) * on disk and in core memory alike */ mask = cpu_to_lel(mask); - /* because of the "extra long to catch oob access" we allocate in - * drbd_bm_resize, bm_number_of_pages -1 is not necessarily the page - * containing the last _relevant_ bitmap word */ - p_addr = bm_map_pidx(b, bm_bit_to_page_idx(b, b->bm_bits - 1)); + p_addr = bm_map_pidx(b, b->bm_number_of_pages - 1); bm = p_addr + (tmp/BITS_PER_LONG); if (mask) { /* If mask != 0, we are not exactly aligned, so bm now points @@ -556,15 +550,10 @@ static unsigned long bm_count_bits(struct drbd_bitmap *b) unsigned long *p_addr; unsigned long bits = 0; unsigned long mask = (1UL << (b->bm_bits & BITS_PER_LONG_MASK)) -1; - int idx, last_page, i, last_word; - - /* because of the "extra long to catch oob access" we allocate in - * drbd_bm_resize, bm_number_of_pages -1 is not necessarily the page - * containing the last _relevant_ bitmap word */ - last_page = bm_bit_to_page_idx(b, b->bm_bits-1); + int idx, i, last_word; /* all but last page */ - for (idx = 0; idx < last_page; idx++) { + for (idx = 0; idx < b->bm_number_of_pages - 1; idx++) { p_addr = __bm_map_pidx(b, idx, KM_USER0); for (i = 0; i < LWPP; i++) bits += hweight_long(p_addr[i]); @@ -627,7 +616,7 @@ static void bm_memset(struct drbd_bitmap *b, size_t offset, int c, size_t len) int drbd_bm_resize(struct drbd_conf *mdev, sector_t capacity, int set_new_bits) { struct drbd_bitmap *b = mdev->bitmap; - unsigned long bits, words, owords, obits, *p_addr, *bm; + unsigned long bits, words, owords, obits; unsigned long want, have, onpages; /* number of pages */ struct page **npages, **opages = NULL; int err = 0, growing; @@ -681,8 +670,7 @@ int drbd_bm_resize(struct drbd_conf *mdev, sector_t capacity, int set_new_bits) } } - /* one extra long to catch off by one errors */ - want = ALIGN((words+1)*sizeof(long), PAGE_SIZE) >> PAGE_SHIFT; + want = ALIGN(words*sizeof(long), PAGE_SIZE) >> PAGE_SHIFT; have = b->bm_number_of_pages; if (want == have) { D_ASSERT(b->bm_pages != NULL); @@ -728,11 +716,6 @@ int drbd_bm_resize(struct drbd_conf *mdev, sector_t capacity, int set_new_bits) bm_free_pages(opages + want, have - want); } - p_addr = bm_map_pidx(b, bm_word_to_page_idx(b, words)); - bm = p_addr + MLPP(words); - *bm = DRBD_MAGIC; - bm_unmap(p_addr); - (void)bm_clear_surplus(b); spin_unlock_irq(&b->bm_lock); @@ -845,7 +828,6 @@ void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset, size_t number, */ if (end == b->bm_words) b->bm_set -= bm_clear_surplus(b); - spin_unlock_irq(&b->bm_lock); } @@ -1030,7 +1012,7 @@ static int bm_rw(struct drbd_conf *mdev, int rw, unsigned lazy_writeout_upper_id struct bm_aio_ctx ctx = { .flags = lazy_writeout_upper_idx ? BM_AIO_COPY_PAGES : 0 }; struct drbd_bitmap *b = mdev->bitmap; - int last_page, i, count = 0; + int num_pages, i, count = 0; unsigned long now; char ppb[10]; int err = 0; @@ -1046,10 +1028,7 @@ static int bm_rw(struct drbd_conf *mdev, int rw, unsigned lazy_writeout_upper_id if (!ctx.flags) WARN_ON(!bm_is_locked(b)); - /* because of the "extra long to catch oob access" we allocate in - * drbd_bm_resize, bm_number_of_pages -1 is not necessarily the page - * containing the last _relevant_ bitmap word */ - last_page = bm_word_to_page_idx(b, b->bm_words - 1); + num_pages = b->bm_number_of_pages; now = jiffies; ctx.mdev = mdev; @@ -1058,7 +1037,7 @@ static int bm_rw(struct drbd_conf *mdev, int rw, unsigned lazy_writeout_upper_id ctx.error = 0; /* let the layers below us try to merge these bios... */ - for (i = 0; i <= last_page; i++) { + for (i = 0; i < num_pages; i++) { /* ignore completely unchanged pages */ if (lazy_writeout_upper_idx && i == lazy_writeout_upper_idx) break; -- cgit v1.2.3 From 84e7c0f7d123d64d97e1f789ad2f23a72fe8981f Mon Sep 17 00:00:00 2001 From: Lars Ellenberg Date: Thu, 16 Dec 2010 00:37:57 +0100 Subject: drbd: Removed a reference to debug macros removed long time ago Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_bitmap.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) (limited to 'drivers/block/drbd/drbd_bitmap.c') diff --git a/drivers/block/drbd/drbd_bitmap.c b/drivers/block/drbd/drbd_bitmap.c index 09d208b858e7..423dede45844 100644 --- a/drivers/block/drbd/drbd_bitmap.c +++ b/drivers/block/drbd/drbd_bitmap.c @@ -596,9 +596,8 @@ static void bm_memset(struct drbd_bitmap *b, size_t offset, int c, size_t len) if (bm+do_now > p_addr + LWPP) { printk(KERN_ALERT "drbd: BUG BUG BUG! p_addr:%p bm:%p do_now:%d\n", p_addr, bm, (int)do_now); - break; /* breaks to after catch_oob_access_end() only! */ - } - memset(bm, c, do_now * sizeof(long)); + } else + memset(bm, c, do_now * sizeof(long)); bm_unmap(p_addr); bm_set_page_need_writeout(b->bm_pages[idx]); offset += do_now; -- cgit v1.2.3 From 7648cdfe52daf0ca4fa9489879dea9e089b0dfe1 Mon Sep 17 00:00:00 2001 From: Lars Ellenberg Date: Fri, 17 Dec 2010 23:58:41 +0100 Subject: drbd: be less noisy with some log messages We expect changes to a bitmap page in drbd_bm_write_page, that's why we submit a copy page. If a page changes during global writeout, that would be unexpected, and reason to warn, though. Also, often page writeout can be skipped (on activity log transactions during normal operation, for example), no need to log that everytime. Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_bitmap.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'drivers/block/drbd/drbd_bitmap.c') diff --git a/drivers/block/drbd/drbd_bitmap.c b/drivers/block/drbd/drbd_bitmap.c index 423dede45844..314a3632303b 100644 --- a/drivers/block/drbd/drbd_bitmap.c +++ b/drivers/block/drbd/drbd_bitmap.c @@ -920,8 +920,9 @@ static void bm_async_io_complete(struct bio *bio, int error) if (!error && !uptodate) error = -EIO; - if (!bm_test_page_unchanged(b->bm_pages[idx])) - dev_info(DEV, "bitmap page idx %u changed during IO!\n", idx); + if ((ctx->flags & BM_AIO_COPY_PAGES) == 0 && + !bm_test_page_unchanged(b->bm_pages[idx])) + dev_warn(DEV, "bitmap page idx %u changed during IO!\n", idx); if (error) { /* ctx error will hold the completed-last non-zero error code, @@ -1135,7 +1136,7 @@ int drbd_bm_write_page(struct drbd_conf *mdev, unsigned int idx) __must_hold(loc struct bm_aio_ctx ctx = { .flags = BM_AIO_COPY_PAGES, }; if (bm_test_page_unchanged(mdev->bitmap->bm_pages[idx])) { - dev_info(DEV, "skipped bm page write for idx %u\n", idx); + dynamic_dev_dbg(DEV, "skipped bm page write for idx %u\n", idx); return 0; } -- cgit v1.2.3 From 725a97e43ee945cc813fffd9e628e50d703b973b Mon Sep 17 00:00:00 2001 From: Lars Ellenberg Date: Sun, 19 Dec 2010 11:29:55 +0100 Subject: drbd: fix potential access of on-stack wait_queue_head_t after return I run into something declaring itself as "spinlock deadlock", BUG: spinlock lockup on CPU#1, kjournald/27816, ffff88000ad6bca0 Pid: 27816, comm: kjournald Tainted: G W 2.6.34.6 #2 Call Trace: [] do_raw_spin_lock+0x11e/0x14d [] _raw_spin_lock_irqsave+0x6a/0x81 [] ? __wake_up+0x22/0x50 [] __wake_up+0x22/0x50 [] bm_async_io_complete+0x258/0x299 [drbd] but the call traces do not fit at all, all other cpus are cpu_idle. I think it may be this race: drbd_bm_write_page wait_queue_head_t io_wait; atomic_t in_flight; bm_async_io submit_bio bm_async_io_complete if (atomic_dec_and_test(in_flight)) wait_event(io_wait, atomic_read(in_flight) == 0) return wake_up(io_wait) The wake_up now accesses the wait_queue_head_t spinlock, which is no longer valid, since the stack frame of drbd_bm_write_page has been clobbered now. Fix this by using struct completion, which does both the condition test as well as the wake_up inside its spinlock, so this race cannot happen. Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_bitmap.c | 38 ++++++++++++++++++++++---------------- 1 file changed, 22 insertions(+), 16 deletions(-) (limited to 'drivers/block/drbd/drbd_bitmap.c') diff --git a/drivers/block/drbd/drbd_bitmap.c b/drivers/block/drbd/drbd_bitmap.c index 314a3632303b..25428bc28476 100644 --- a/drivers/block/drbd/drbd_bitmap.c +++ b/drivers/block/drbd/drbd_bitmap.c @@ -897,7 +897,7 @@ void drbd_bm_clear_all(struct drbd_conf *mdev) struct bm_aio_ctx { struct drbd_conf *mdev; atomic_t in_flight; - wait_queue_head_t io_wait; + struct completion done; unsigned flags; #define BM_AIO_COPY_PAGES 1 int error; @@ -948,7 +948,7 @@ static void bm_async_io_complete(struct bio *bio, int error) bio_put(bio); if (atomic_dec_and_test(&ctx->in_flight)) - wake_up(&ctx->io_wait); + complete(&ctx->done); } static void bm_page_io_async(struct bm_aio_ctx *ctx, int page_nr, int rw) __must_hold(local) @@ -1009,8 +1009,12 @@ static void bm_page_io_async(struct bm_aio_ctx *ctx, int page_nr, int rw) __must */ static int bm_rw(struct drbd_conf *mdev, int rw, unsigned lazy_writeout_upper_idx) __must_hold(local) { - struct bm_aio_ctx ctx = - { .flags = lazy_writeout_upper_idx ? BM_AIO_COPY_PAGES : 0 }; + struct bm_aio_ctx ctx = { + .mdev = mdev, + .in_flight = ATOMIC_INIT(1), + .done = COMPLETION_INITIALIZER_ONSTACK(ctx.done), + .flags = lazy_writeout_upper_idx ? BM_AIO_COPY_PAGES : 0, + }; struct drbd_bitmap *b = mdev->bitmap; int num_pages, i, count = 0; unsigned long now; @@ -1031,10 +1035,6 @@ static int bm_rw(struct drbd_conf *mdev, int rw, unsigned lazy_writeout_upper_id num_pages = b->bm_number_of_pages; now = jiffies; - ctx.mdev = mdev; - atomic_set(&ctx.in_flight, 1); /* one extra ref */ - init_waitqueue_head(&ctx.io_wait); - ctx.error = 0; /* let the layers below us try to merge these bios... */ for (i = 0; i < num_pages; i++) { @@ -1060,8 +1060,13 @@ static int bm_rw(struct drbd_conf *mdev, int rw, unsigned lazy_writeout_upper_id cond_resched(); } - atomic_dec(&ctx.in_flight); /* drop the extra ref */ - wait_event(ctx.io_wait, atomic_read(&ctx.in_flight) == 0); + /* + * We initialize ctx.in_flight to one to make sure bm_async_io_complete + * will not complete() early, and decrement / test it here. If there + * are still some bios in flight, we need to wait for them here. + */ + if (!atomic_dec_and_test(&ctx.in_flight)) + wait_for_completion(&ctx.done); dev_info(DEV, "bitmap %s of %u pages took %lu jiffies\n", rw == WRITE ? "WRITE" : "READ", count, jiffies - now); @@ -1133,19 +1138,20 @@ int drbd_bm_write_lazy(struct drbd_conf *mdev, unsigned upper_idx) __must_hold(l */ int drbd_bm_write_page(struct drbd_conf *mdev, unsigned int idx) __must_hold(local) { - struct bm_aio_ctx ctx = { .flags = BM_AIO_COPY_PAGES, }; + struct bm_aio_ctx ctx = { + .mdev = mdev, + .in_flight = ATOMIC_INIT(1), + .done = COMPLETION_INITIALIZER_ONSTACK(ctx.done), + .flags = BM_AIO_COPY_PAGES, + }; if (bm_test_page_unchanged(mdev->bitmap->bm_pages[idx])) { dynamic_dev_dbg(DEV, "skipped bm page write for idx %u\n", idx); return 0; } - ctx.mdev = mdev; - atomic_set(&ctx.in_flight, 1); - init_waitqueue_head(&ctx.io_wait); - bm_page_io_async(&ctx, idx, WRITE_SYNC); - wait_event(ctx.io_wait, atomic_read(&ctx.in_flight) == 0); + wait_for_completion(&ctx.done); if (ctx.error) drbd_chk_io_error(mdev, 1, true); -- cgit v1.2.3 From 20ceb2b22edaf51e59e76087efdc71a16a2858de Mon Sep 17 00:00:00 2001 From: Lars Ellenberg Date: Fri, 21 Jan 2011 10:56:44 +0100 Subject: drbd: describe bitmap locking for bulk operation in finer detail Now that we do no longer in-place endian-swap the bitmap, we allow selected bitmap operations (testing bits, sometimes even settting bits) during some bulk operations. This caused us to hit a lot of FIXME asserts similar to FIXME asender in drbd_bm_count_bits, bitmap locked for 'write from resync_finished' by worker Which now is nonsense: looking at the bitmap is perfectly legal as long as it is not being resized. This cosmetic patch defines some flags to describe expectations in finer detail, so the asserts in e.g. bm_change_bits_to() can be skipped if appropriate. Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_bitmap.c | 48 +++++++++++++------------------ drivers/block/drbd/drbd_int.h | 36 ++++++++++++++++++++--- drivers/block/drbd/drbd_main.c | 58 +++++++++++++++++++++++++------------- drivers/block/drbd/drbd_nl.c | 19 +++++++++---- drivers/block/drbd/drbd_receiver.c | 17 +++++++---- 5 files changed, 115 insertions(+), 63 deletions(-) (limited to 'drivers/block/drbd/drbd_bitmap.c') diff --git a/drivers/block/drbd/drbd_bitmap.c b/drivers/block/drbd/drbd_bitmap.c index 25428bc28476..b62dd5f26c5d 100644 --- a/drivers/block/drbd/drbd_bitmap.c +++ b/drivers/block/drbd/drbd_bitmap.c @@ -104,26 +104,16 @@ struct drbd_bitmap { wait_queue_head_t bm_io_wait; /* used to serialize IO of single pages */ - unsigned long bm_flags; + enum bm_flag bm_flags; /* debugging aid, in case we are still racy somewhere */ char *bm_why; struct task_struct *bm_task; }; -/* definition of bits in bm_flags */ -#define BM_LOCKED 0 -// #define BM_MD_IO_ERROR 1 unused now. -#define BM_P_VMALLOCED 2 - static int __bm_change_bits_to(struct drbd_conf *mdev, const unsigned long s, unsigned long e, int val, const enum km_type km); -static int bm_is_locked(struct drbd_bitmap *b) -{ - return test_bit(BM_LOCKED, &b->bm_flags); -} - #define bm_print_lock_info(m) __bm_print_lock_info(m, __func__) static void __bm_print_lock_info(struct drbd_conf *mdev, const char *func) { @@ -140,7 +130,7 @@ static void __bm_print_lock_info(struct drbd_conf *mdev, const char *func) b->bm_task == mdev->worker.task ? "worker" : "?"); } -void drbd_bm_lock(struct drbd_conf *mdev, char *why) +void drbd_bm_lock(struct drbd_conf *mdev, char *why, enum bm_flag flags) { struct drbd_bitmap *b = mdev->bitmap; int trylock_failed; @@ -163,8 +153,9 @@ void drbd_bm_lock(struct drbd_conf *mdev, char *why) b->bm_task == mdev->worker.task ? "worker" : "?"); mutex_lock(&b->bm_change); } - if (__test_and_set_bit(BM_LOCKED, &b->bm_flags)) + if (BM_LOCKED_MASK & b->bm_flags) dev_err(DEV, "FIXME bitmap already locked in bm_lock\n"); + b->bm_flags |= flags & BM_LOCKED_MASK; b->bm_why = why; b->bm_task = current; @@ -178,9 +169,10 @@ void drbd_bm_unlock(struct drbd_conf *mdev) return; } - if (!__test_and_clear_bit(BM_LOCKED, &mdev->bitmap->bm_flags)) + if (!(BM_LOCKED_MASK & mdev->bitmap->bm_flags)) dev_err(DEV, "FIXME bitmap not locked in bm_unlock\n"); + b->bm_flags &= ~BM_LOCKED_MASK; b->bm_why = NULL; b->bm_task = NULL; mutex_unlock(&b->bm_change); @@ -421,9 +413,9 @@ static struct page **bm_realloc_pages(struct drbd_bitmap *b, unsigned long want) } if (vmalloced) - set_bit(BM_P_VMALLOCED, &b->bm_flags); + b->bm_flags |= BM_P_VMALLOCED; else - clear_bit(BM_P_VMALLOCED, &b->bm_flags); + b->bm_flags &= ~BM_P_VMALLOCED; return new_pages; } @@ -460,7 +452,7 @@ void drbd_bm_cleanup(struct drbd_conf *mdev) { ERR_IF (!mdev->bitmap) return; bm_free_pages(mdev->bitmap->bm_pages, mdev->bitmap->bm_number_of_pages); - bm_vk_free(mdev->bitmap->bm_pages, test_bit(BM_P_VMALLOCED, &mdev->bitmap->bm_flags)); + bm_vk_free(mdev->bitmap->bm_pages, (BM_P_VMALLOCED & mdev->bitmap->bm_flags)); kfree(mdev->bitmap); mdev->bitmap = NULL; } @@ -623,7 +615,7 @@ int drbd_bm_resize(struct drbd_conf *mdev, sector_t capacity, int set_new_bits) ERR_IF(!b) return -ENOMEM; - drbd_bm_lock(mdev, "resize"); + drbd_bm_lock(mdev, "resize", BM_LOCKED_MASK); dev_info(DEV, "drbd_bm_resize called with capacity == %llu\n", (unsigned long long)capacity); @@ -631,7 +623,7 @@ int drbd_bm_resize(struct drbd_conf *mdev, sector_t capacity, int set_new_bits) if (capacity == b->bm_dev_capacity) goto out; - opages_vmalloced = test_bit(BM_P_VMALLOCED, &b->bm_flags); + opages_vmalloced = (BM_P_VMALLOCED & b->bm_flags); if (capacity == 0) { spin_lock_irq(&b->bm_lock); @@ -1030,7 +1022,7 @@ static int bm_rw(struct drbd_conf *mdev, int rw, unsigned lazy_writeout_upper_id * as we submit copies of pages anyways. */ if (!ctx.flags) - WARN_ON(!bm_is_locked(b)); + WARN_ON(!(BM_LOCKED_MASK & b->bm_flags)); num_pages = b->bm_number_of_pages; @@ -1220,7 +1212,7 @@ static unsigned long bm_find_next(struct drbd_conf *mdev, ERR_IF(!b->bm_pages) return i; spin_lock_irq(&b->bm_lock); - if (bm_is_locked(b)) + if (BM_DONT_TEST & b->bm_flags) bm_print_lock_info(mdev); i = __bm_find_next(mdev, bm_fo, find_zero_bit, KM_IRQ1); @@ -1246,13 +1238,13 @@ unsigned long drbd_bm_find_next_zero(struct drbd_conf *mdev, unsigned long bm_fo * you must take drbd_bm_lock() first */ unsigned long _drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo) { - /* WARN_ON(!bm_is_locked(mdev)); */ + /* WARN_ON(!(BM_DONT_SET & mdev->b->bm_flags)); */ return __bm_find_next(mdev, bm_fo, 0, KM_USER1); } unsigned long _drbd_bm_find_next_zero(struct drbd_conf *mdev, unsigned long bm_fo) { - /* WARN_ON(!bm_is_locked(mdev)); */ + /* WARN_ON(!(BM_DONT_SET & mdev->b->bm_flags)); */ return __bm_find_next(mdev, bm_fo, 1, KM_USER1); } @@ -1322,7 +1314,7 @@ static int bm_change_bits_to(struct drbd_conf *mdev, const unsigned long s, ERR_IF(!b->bm_pages) return 0; spin_lock_irqsave(&b->bm_lock, flags); - if (bm_is_locked(b)) + if ((val ? BM_DONT_SET : BM_DONT_CLEAR) & b->bm_flags) bm_print_lock_info(mdev); c = __bm_change_bits_to(mdev, s, e, val, KM_IRQ1); @@ -1439,7 +1431,7 @@ int drbd_bm_test_bit(struct drbd_conf *mdev, const unsigned long bitnr) ERR_IF(!b->bm_pages) return 0; spin_lock_irqsave(&b->bm_lock, flags); - if (bm_is_locked(b)) + if (BM_DONT_TEST & b->bm_flags) bm_print_lock_info(mdev); if (bitnr < b->bm_bits) { p_addr = bm_map_pidx(b, bm_bit_to_page_idx(b, bitnr)); @@ -1474,7 +1466,7 @@ int drbd_bm_count_bits(struct drbd_conf *mdev, const unsigned long s, const unsi ERR_IF(!b->bm_pages) return 1; spin_lock_irqsave(&b->bm_lock, flags); - if (bm_is_locked(b)) + if (BM_DONT_TEST & b->bm_flags) bm_print_lock_info(mdev); for (bitnr = s; bitnr <= e; bitnr++) { unsigned int idx = bm_bit_to_page_idx(b, bitnr); @@ -1522,7 +1514,7 @@ int drbd_bm_e_weight(struct drbd_conf *mdev, unsigned long enr) ERR_IF(!b->bm_pages) return 0; spin_lock_irqsave(&b->bm_lock, flags); - if (bm_is_locked(b)) + if (BM_DONT_TEST & b->bm_flags) bm_print_lock_info(mdev); s = S2W(enr); @@ -1555,7 +1547,7 @@ unsigned long drbd_bm_ALe_set_all(struct drbd_conf *mdev, unsigned long al_enr) ERR_IF(!b->bm_pages) return 0; spin_lock_irq(&b->bm_lock); - if (bm_is_locked(b)) + if (BM_DONT_SET & b->bm_flags) bm_print_lock_info(mdev); weight = b->bm_set; diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h index 0a9059eb94db..267d9897ca8c 100644 --- a/drivers/block/drbd/drbd_int.h +++ b/drivers/block/drbd/drbd_int.h @@ -855,6 +855,32 @@ enum { struct drbd_bitmap; /* opaque for drbd_conf */ +/* definition of bits in bm_flags to be used in drbd_bm_lock + * and drbd_bitmap_io and friends. */ +enum bm_flag { + /* do we need to kfree, or vfree bm_pages? */ + BM_P_VMALLOCED = 0x10000, /* internal use only, will be masked out */ + + /* currently locked for bulk operation */ + BM_LOCKED_MASK = 0x7, + + /* in detail, that is: */ + BM_DONT_CLEAR = 0x1, + BM_DONT_SET = 0x2, + BM_DONT_TEST = 0x4, + + /* (test bit, count bit) allowed (common case) */ + BM_LOCKED_TEST_ALLOWED = 0x3, + + /* testing bits, as well as setting new bits allowed, but clearing bits + * would be unexpected. Used during bitmap receive. Setting new bits + * requires sending of "out-of-sync" information, though. */ + BM_LOCKED_SET_ALLOWED = 0x1, + + /* clear is not expected while bitmap is locked for bulk operation */ +}; + + /* TODO sort members for performance * MAYBE group them further */ @@ -920,6 +946,7 @@ struct drbd_md_io { struct bm_io_work { struct drbd_work w; char *why; + enum bm_flag flags; int (*io_fn)(struct drbd_conf *mdev); void (*done)(struct drbd_conf *mdev, int rv); }; @@ -1242,7 +1269,6 @@ extern void drbd_free_bc(struct drbd_backing_dev *ldev); extern void drbd_mdev_cleanup(struct drbd_conf *mdev); void drbd_print_uuids(struct drbd_conf *mdev, const char *text); -/* drbd_meta-data.c (still in drbd_main.c) */ extern void drbd_md_sync(struct drbd_conf *mdev); extern int drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev); extern void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local); @@ -1263,10 +1289,12 @@ extern void drbd_md_mark_dirty_(struct drbd_conf *mdev, extern void drbd_queue_bitmap_io(struct drbd_conf *mdev, int (*io_fn)(struct drbd_conf *), void (*done)(struct drbd_conf *, int), - char *why); + char *why, enum bm_flag flags); +extern int drbd_bitmap_io(struct drbd_conf *mdev, + int (*io_fn)(struct drbd_conf *), + char *why, enum bm_flag flags); extern int drbd_bmio_set_n_write(struct drbd_conf *mdev); extern int drbd_bmio_clear_n_write(struct drbd_conf *mdev); -extern int drbd_bitmap_io(struct drbd_conf *mdev, int (*io_fn)(struct drbd_conf *), char *why); extern void drbd_go_diskless(struct drbd_conf *mdev); extern void drbd_ldev_destroy(struct drbd_conf *mdev); @@ -1452,7 +1480,7 @@ extern void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset, extern void drbd_bm_get_lel(struct drbd_conf *mdev, size_t offset, size_t number, unsigned long *buffer); -extern void drbd_bm_lock(struct drbd_conf *mdev, char *why); +extern void drbd_bm_lock(struct drbd_conf *mdev, char *why, enum bm_flag flags); extern void drbd_bm_unlock(struct drbd_conf *mdev); /* drbd_main.c */ diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index b68332a0e73e..a9e9b496e73b 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -1320,7 +1320,9 @@ static void abw_start_sync(struct drbd_conf *mdev, int rv) } } -int drbd_bitmap_io_from_worker(struct drbd_conf *mdev, int (*io_fn)(struct drbd_conf *), char *why) +int drbd_bitmap_io_from_worker(struct drbd_conf *mdev, + int (*io_fn)(struct drbd_conf *), + char *why, enum bm_flag flags) { int rv; @@ -1328,10 +1330,8 @@ int drbd_bitmap_io_from_worker(struct drbd_conf *mdev, int (*io_fn)(struct drbd_ /* open coded non-blocking drbd_suspend_io(mdev); */ set_bit(SUSPEND_IO, &mdev->flags); - if (!is_susp(mdev->state)) - D_ASSERT(atomic_read(&mdev->ap_bio_cnt) == 0); - drbd_bm_lock(mdev, why); + drbd_bm_lock(mdev, why, flags); rv = io_fn(mdev); drbd_bm_unlock(mdev); @@ -1438,7 +1438,8 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, if (os.conn != C_WF_BITMAP_S && ns.conn == C_WF_BITMAP_S && mdev->state.conn == C_WF_BITMAP_S) drbd_queue_bitmap_io(mdev, &drbd_send_bitmap, NULL, - "send_bitmap (WFBitMapS)"); + "send_bitmap (WFBitMapS)", + BM_LOCKED_TEST_ALLOWED); /* Lost contact to peer's copy of the data */ if ((os.pdsk >= D_INCONSISTENT && @@ -1469,7 +1470,11 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, /* D_DISKLESS Peer becomes secondary */ if (os.peer == R_PRIMARY && ns.peer == R_SECONDARY) - drbd_bitmap_io_from_worker(mdev, &drbd_bm_write, "demote diskless peer"); + /* We may still be Primary ourselves. + * No harm done if the bitmap still changes, + * redirtied pages will follow later. */ + drbd_bitmap_io_from_worker(mdev, &drbd_bm_write, + "demote diskless peer", BM_LOCKED_SET_ALLOWED); put_ldev(mdev); } @@ -1478,7 +1483,10 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, * if there is a resync going on still */ if (os.role == R_PRIMARY && ns.role == R_SECONDARY && mdev->state.conn <= C_CONNECTED && get_ldev(mdev)) { - drbd_bitmap_io_from_worker(mdev, &drbd_bm_write, "demote"); + /* No changes to the bitmap expected this time, so assert that, + * even though no harm was done if it did change. */ + drbd_bitmap_io_from_worker(mdev, &drbd_bm_write, + "demote", BM_LOCKED_TEST_ALLOWED); put_ldev(mdev); } @@ -1512,12 +1520,17 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, /* We are in the progress to start a full sync... */ if ((os.conn != C_STARTING_SYNC_T && ns.conn == C_STARTING_SYNC_T) || (os.conn != C_STARTING_SYNC_S && ns.conn == C_STARTING_SYNC_S)) - drbd_queue_bitmap_io(mdev, &drbd_bmio_set_n_write, &abw_start_sync, "set_n_write from StartingSync"); + /* no other bitmap changes expected during this phase */ + drbd_queue_bitmap_io(mdev, + &drbd_bmio_set_n_write, &abw_start_sync, + "set_n_write from StartingSync", BM_LOCKED_TEST_ALLOWED); /* We are invalidating our self... */ if (os.conn < C_CONNECTED && ns.conn < C_CONNECTED && os.disk > D_INCONSISTENT && ns.disk == D_INCONSISTENT) - drbd_queue_bitmap_io(mdev, &drbd_bmio_set_n_write, NULL, "set_n_write from invalidate"); + /* other bitmap operation expected during this phase */ + drbd_queue_bitmap_io(mdev, &drbd_bmio_set_n_write, NULL, + "set_n_write from invalidate", BM_LOCKED_MASK); /* first half of local IO error, failure to attach, * or administrative detach */ @@ -1599,14 +1612,14 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, /* This triggers bitmap writeout of potentially still unwritten pages * if the resync finished cleanly, or aborted because of peer disk - * failure. Resync aborted because of connection failure does bitmap - * writeout from drbd_disconnect. + * failure, or because of connection loss. * For resync aborted because of local disk failure, we cannot do * any bitmap writeout anymore. + * No harm done if some bits change during this phase. */ - if (os.conn > C_CONNECTED && ns.conn == C_CONNECTED && - mdev->state.conn == C_CONNECTED && get_ldev(mdev)) { - drbd_queue_bitmap_io(mdev, &drbd_bm_write, NULL, "write from resync_finished"); + if (os.conn > C_CONNECTED && ns.conn <= C_CONNECTED && get_ldev(mdev)) { + drbd_queue_bitmap_io(mdev, &drbd_bm_write, NULL, + "write from resync_finished", BM_LOCKED_SET_ALLOWED); put_ldev(mdev); } @@ -3929,7 +3942,7 @@ static int w_bitmap_io(struct drbd_conf *mdev, struct drbd_work *w, int unused) D_ASSERT(atomic_read(&mdev->ap_bio_cnt) == 0); if (get_ldev(mdev)) { - drbd_bm_lock(mdev, work->why); + drbd_bm_lock(mdev, work->why, work->flags); rv = work->io_fn(mdev); drbd_bm_unlock(mdev); put_ldev(mdev); @@ -3944,6 +3957,7 @@ static int w_bitmap_io(struct drbd_conf *mdev, struct drbd_work *w, int unused) clear_bit(BITMAP_IO_QUEUED, &mdev->flags); work->why = NULL; + work->flags = 0; return 1; } @@ -3998,7 +4012,7 @@ void drbd_go_diskless(struct drbd_conf *mdev) void drbd_queue_bitmap_io(struct drbd_conf *mdev, int (*io_fn)(struct drbd_conf *), void (*done)(struct drbd_conf *, int), - char *why) + char *why, enum bm_flag flags) { D_ASSERT(current == mdev->worker.task); @@ -4012,6 +4026,7 @@ void drbd_queue_bitmap_io(struct drbd_conf *mdev, mdev->bm_io_work.io_fn = io_fn; mdev->bm_io_work.done = done; mdev->bm_io_work.why = why; + mdev->bm_io_work.flags = flags; spin_lock_irq(&mdev->req_lock); set_bit(BITMAP_IO, &mdev->flags); @@ -4031,19 +4046,22 @@ void drbd_queue_bitmap_io(struct drbd_conf *mdev, * freezes application IO while that the actual IO operations runs. This * functions MAY NOT be called from worker context. */ -int drbd_bitmap_io(struct drbd_conf *mdev, int (*io_fn)(struct drbd_conf *), char *why) +int drbd_bitmap_io(struct drbd_conf *mdev, int (*io_fn)(struct drbd_conf *), + char *why, enum bm_flag flags) { int rv; D_ASSERT(current != mdev->worker.task); - drbd_suspend_io(mdev); + if ((flags & BM_LOCKED_SET_ALLOWED) == 0) + drbd_suspend_io(mdev); - drbd_bm_lock(mdev, why); + drbd_bm_lock(mdev, why, flags); rv = io_fn(mdev); drbd_bm_unlock(mdev); - drbd_resume_io(mdev); + if ((flags & BM_LOCKED_SET_ALLOWED) == 0) + drbd_resume_io(mdev); return rv; } diff --git a/drivers/block/drbd/drbd_nl.c b/drivers/block/drbd/drbd_nl.c index ce6f2fe80852..bc0bcb964603 100644 --- a/drivers/block/drbd/drbd_nl.c +++ b/drivers/block/drbd/drbd_nl.c @@ -648,7 +648,9 @@ enum determine_dev_size drbd_determin_dev_size(struct drbd_conf *mdev, enum dds_ dev_info(DEV, "Writing the whole bitmap, %s\n", la_size_changed && md_moved ? "size changed and md moved" : la_size_changed ? "size changed" : "md moved"); - err = drbd_bitmap_io(mdev, &drbd_bm_write, "size changed"); /* does drbd_resume_io() ! */ + /* next line implicitly does drbd_suspend_io()+drbd_resume_io() */ + err = drbd_bitmap_io(mdev, &drbd_bm_write, + "size changed", BM_LOCKED_MASK); if (err) { rv = dev_size_error; goto out; @@ -1160,12 +1162,14 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp if (drbd_md_test_flag(mdev->ldev, MDF_FULL_SYNC)) { dev_info(DEV, "Assuming that all blocks are out of sync " "(aka FullSync)\n"); - if (drbd_bitmap_io(mdev, &drbd_bmio_set_n_write, "set_n_write from attaching")) { + if (drbd_bitmap_io(mdev, &drbd_bmio_set_n_write, + "set_n_write from attaching", BM_LOCKED_MASK)) { retcode = ERR_IO_MD_DISK; goto force_diskless_dec; } } else { - if (drbd_bitmap_io(mdev, &drbd_bm_read, "read from attaching") < 0) { + if (drbd_bitmap_io(mdev, &drbd_bm_read, + "read from attaching", BM_LOCKED_MASK) < 0) { retcode = ERR_IO_MD_DISK; goto force_diskless_dec; } @@ -1173,7 +1177,8 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp if (cp_discovered) { drbd_al_apply_to_bm(mdev); - if (drbd_bitmap_io(mdev, &drbd_bm_write, "crashed primary apply AL")) { + if (drbd_bitmap_io(mdev, &drbd_bm_write, + "crashed primary apply AL", BM_LOCKED_MASK)) { retcode = ERR_IO_MD_DISK; goto force_diskless_dec; } @@ -1925,7 +1930,8 @@ static int drbd_nl_invalidate_peer(struct drbd_conf *mdev, struct drbd_nl_cfg_re retcode = drbd_request_state(mdev, NS(pdsk, D_INCONSISTENT)); if (retcode >= SS_SUCCESS) { if (drbd_bitmap_io(mdev, &drbd_bmio_set_susp_al, - "set_n_write from invalidate_peer")) + "set_n_write from invalidate_peer", + BM_LOCKED_SET_ALLOWED)) retcode = ERR_IO_MD_DISK; } } else @@ -2143,7 +2149,8 @@ static int drbd_nl_new_c_uuid(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nl drbd_uuid_new_current(mdev); /* New current, previous to UI_BITMAP */ if (args.clear_bm) { - err = drbd_bitmap_io(mdev, &drbd_bmio_clear_n_write, "clear_n_write from new_c_uuid"); + err = drbd_bitmap_io(mdev, &drbd_bmio_clear_n_write, + "clear_n_write from new_c_uuid", BM_LOCKED_MASK); if (err) { dev_err(DEV, "Writing bitmap failed with %d\n",err); retcode = ERR_IO_MD_DISK; diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c index e5686a81f42c..e13134f83fae 100644 --- a/drivers/block/drbd/drbd_receiver.c +++ b/drivers/block/drbd/drbd_receiver.c @@ -2599,7 +2599,8 @@ static enum drbd_conns drbd_sync_handshake(struct drbd_conf *mdev, enum drbd_rol if (abs(hg) >= 2) { dev_info(DEV, "Writing the whole bitmap, full sync required after drbd_sync_handshake.\n"); - if (drbd_bitmap_io(mdev, &drbd_bmio_set_n_write, "set_n_write from sync_handshake")) + if (drbd_bitmap_io(mdev, &drbd_bmio_set_n_write, "set_n_write from sync_handshake", + BM_LOCKED_SET_ALLOWED)) return C_MASK; } @@ -3053,7 +3054,8 @@ static int receive_uuids(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned if (skip_initial_sync) { dev_info(DEV, "Accepted new current UUID, preparing to skip initial sync\n"); drbd_bitmap_io(mdev, &drbd_bmio_clear_n_write, - "clear_n_write from receive_uuids"); + "clear_n_write from receive_uuids", + BM_LOCKED_TEST_ALLOWED); _drbd_uuid_set(mdev, UI_CURRENT, p_uuid[UI_CURRENT]); _drbd_uuid_set(mdev, UI_BITMAP, 0); _drbd_set_state(_NS2(mdev, disk, D_UP_TO_DATE, pdsk, D_UP_TO_DATE), @@ -3494,7 +3496,9 @@ static int receive_bitmap(struct drbd_conf *mdev, enum drbd_packets cmd, unsigne int ok = false; struct p_header80 *h = &mdev->data.rbuf.header.h80; - /* drbd_bm_lock(mdev, "receive bitmap"); By intention no bm_lock */ + drbd_bm_lock(mdev, "receive bitmap", BM_LOCKED_SET_ALLOWED); + /* you are supposed to send additional out-of-sync information + * if you actually set bits during this phase */ /* maybe we should use some per thread scratch page, * and allocate that during initial device creation? */ @@ -3568,7 +3572,7 @@ static int receive_bitmap(struct drbd_conf *mdev, enum drbd_packets cmd, unsigne ok = true; out: - /* drbd_bm_unlock(mdev); by intention no lock */ + drbd_bm_unlock(mdev); if (ok && mdev->state.conn == C_WF_BITMAP_S) drbd_start_resync(mdev, C_SYNC_SOURCE); free_page((unsigned long) buffer); @@ -3817,7 +3821,6 @@ static void drbd_disconnect(struct drbd_conf *mdev) fp = FP_DONT_CARE; if (get_ldev(mdev)) { - drbd_bitmap_io(mdev, &drbd_bm_write, "write from disconnect"); fp = mdev->ldev->dc.fencing; put_ldev(mdev); } @@ -3846,6 +3849,10 @@ static void drbd_disconnect(struct drbd_conf *mdev) drbd_request_state(mdev, NS(conn, C_STANDALONE)); } + /* serialize with bitmap writeout triggered by the state change, + * if any. */ + wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags)); + /* tcp_close and release of sendpage pages can be deferred. I don't * want to use SO_LINGER, because apparently it can be deferred for * more than 20 seconds (longest time I checked). -- cgit v1.2.3 From f0ff1357ce391265edbf844792da7da9a694f4bd Mon Sep 17 00:00:00 2001 From: Stephen Rothwell Date: Thu, 17 Mar 2011 15:02:51 +0100 Subject: drbd: need include for bitops functions declarations Signed-off-by: Stephen Rothwell Signed-off-by: Jens Axboe --- drivers/block/drbd/drbd_bitmap.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'drivers/block/drbd/drbd_bitmap.c') diff --git a/drivers/block/drbd/drbd_bitmap.c b/drivers/block/drbd/drbd_bitmap.c index b62dd5f26c5d..f0ae63d2df65 100644 --- a/drivers/block/drbd/drbd_bitmap.c +++ b/drivers/block/drbd/drbd_bitmap.c @@ -28,6 +28,9 @@ #include #include #include + +#include + #include "drbd_int.h" -- cgit v1.2.3