diff options
author | Linus Torvalds <torvalds@linux-foundation.org> | 2010-10-22 17:03:12 -0700 |
---|---|---|
committer | Linus Torvalds <torvalds@linux-foundation.org> | 2010-10-22 17:03:12 -0700 |
commit | 8abfc6e7a45eb74e51904bbae676fae008b11366 (patch) | |
tree | 57d0a24558c0693e3a52e8e756616f6c72def1e9 /drivers/block/drbd | |
parent | e9dd2b6837e26fe202708cce5ea4bb4ee3e3482e (diff) | |
parent | 6362beea8914cbd4630ccde3617d944aeca2d48f (diff) | |
download | op-kernel-dev-8abfc6e7a45eb74e51904bbae676fae008b11366.zip op-kernel-dev-8abfc6e7a45eb74e51904bbae676fae008b11366.tar.gz |
Merge branch 'for-2.6.37/drivers' of git://git.kernel.dk/linux-2.6-block
* 'for-2.6.37/drivers' of git://git.kernel.dk/linux-2.6-block: (95 commits)
cciss: fix PCI IDs for new Smart Array controllers
drbd: add race-breaker to drbd_go_diskless
drbd: use dynamic_dev_dbg to optionally log uuid changes
dynamic_debug.h: Fix dynamic_dev_dbg() macro if CONFIG_DYNAMIC_DEBUG not set
drbd: cleanup: change "<= 0" to "== 0"
drbd: relax the grace period of the md_sync timer again
drbd: add some more explicit drbd_md_sync
drbd: drop wrong debug asserts, fix recently introduced race
drbd: cleanup useless leftover warn/error printk's
drbd: add explicit drbd_md_sync to drbd_resync_finished
drbd: Do not log an ASSERT for P_OV_REQUEST packets while C_CONNECTED
drbd: fix for possible deadlock on IO error during resync
drbd: fix unlikely access after free and list corruption
drbd: fix for spurious fullsync (uuids rotated too fast)
drbd: allow for explicit resync-finished notifications
drbd: preparation commit, using full state in receive_state()
drbd: drbd_send_ack_dp must not rely on header information
drbd: Fix regression in recv_bm_rle_bits (compressed bitmap)
drbd: Fixed a stupid copy and paste error
drbd: Allow larger values for c-fill-target.
...
Fix up trivial conflict in drivers/block/ataflop.c due to BKL removal
Diffstat (limited to 'drivers/block/drbd')
-rw-r--r-- | drivers/block/drbd/drbd_actlog.c | 41 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_bitmap.c | 2 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_int.h | 216 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_main.c | 593 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_nl.c | 270 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_proc.c | 34 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_receiver.c | 946 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_req.c | 165 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_req.h | 62 | ||||
-rw-r--r-- | drivers/block/drbd/drbd_worker.c | 292 |
10 files changed, 1733 insertions, 888 deletions
diff --git a/drivers/block/drbd/drbd_actlog.c b/drivers/block/drbd/drbd_actlog.c index 9400845..ac04ef9 100644 --- a/drivers/block/drbd/drbd_actlog.c +++ b/drivers/block/drbd/drbd_actlog.c @@ -965,29 +965,30 @@ void __drbd_set_in_sync(struct drbd_conf *mdev, sector_t sector, int size, * ok, (capacity & 7) != 0 sometimes, but who cares... * we count rs_{total,left} in bits, not sectors. */ - spin_lock_irqsave(&mdev->al_lock, flags); count = drbd_bm_clear_bits(mdev, sbnr, ebnr); - if (count) { - /* we need the lock for drbd_try_clear_on_disk_bm */ - if (jiffies - mdev->rs_mark_time > HZ*10) { - /* should be rolling marks, - * but we estimate only anyways. */ - if (mdev->rs_mark_left != drbd_bm_total_weight(mdev) && + if (count && get_ldev(mdev)) { + unsigned long now = jiffies; + unsigned long last = mdev->rs_mark_time[mdev->rs_last_mark]; + int next = (mdev->rs_last_mark + 1) % DRBD_SYNC_MARKS; + if (time_after_eq(now, last + DRBD_SYNC_MARK_STEP)) { + unsigned long tw = drbd_bm_total_weight(mdev); + if (mdev->rs_mark_left[mdev->rs_last_mark] != tw && mdev->state.conn != C_PAUSED_SYNC_T && mdev->state.conn != C_PAUSED_SYNC_S) { - mdev->rs_mark_time = jiffies; - mdev->rs_mark_left = drbd_bm_total_weight(mdev); + mdev->rs_mark_time[next] = now; + mdev->rs_mark_left[next] = tw; + mdev->rs_last_mark = next; } } - if (get_ldev(mdev)) { - drbd_try_clear_on_disk_bm(mdev, sector, count, TRUE); - put_ldev(mdev); - } + spin_lock_irqsave(&mdev->al_lock, flags); + drbd_try_clear_on_disk_bm(mdev, sector, count, TRUE); + spin_unlock_irqrestore(&mdev->al_lock, flags); + /* just wake_up unconditional now, various lc_chaged(), * lc_put() in drbd_try_clear_on_disk_bm(). */ wake_up = 1; + put_ldev(mdev); } - spin_unlock_irqrestore(&mdev->al_lock, flags); if (wake_up) wake_up(&mdev->al_wait); } @@ -1118,7 +1119,7 @@ static int _is_in_al(struct drbd_conf *mdev, unsigned int enr) * @mdev: DRBD device. * @sector: The sector number. * - * This functions sleeps on al_wait. Returns 1 on success, 0 if interrupted. + * This functions sleeps on al_wait. Returns 0 on success, -EINTR if interrupted. */ int drbd_rs_begin_io(struct drbd_conf *mdev, sector_t sector) { @@ -1129,10 +1130,10 @@ int drbd_rs_begin_io(struct drbd_conf *mdev, sector_t sector) sig = wait_event_interruptible(mdev->al_wait, (bm_ext = _bme_get(mdev, enr))); if (sig) - return 0; + return -EINTR; if (test_bit(BME_LOCKED, &bm_ext->flags)) - return 1; + return 0; for (i = 0; i < AL_EXT_PER_BM_SECT; i++) { sig = wait_event_interruptible(mdev->al_wait, @@ -1145,13 +1146,11 @@ int drbd_rs_begin_io(struct drbd_conf *mdev, sector_t sector) wake_up(&mdev->al_wait); } spin_unlock_irq(&mdev->al_lock); - return 0; + return -EINTR; } } - set_bit(BME_LOCKED, &bm_ext->flags); - - return 1; + return 0; } /** diff --git a/drivers/block/drbd/drbd_bitmap.c b/drivers/block/drbd/drbd_bitmap.c index e3f88d6..fd42832 100644 --- a/drivers/block/drbd/drbd_bitmap.c +++ b/drivers/block/drbd/drbd_bitmap.c @@ -569,7 +569,7 @@ int drbd_bm_resize(struct drbd_conf *mdev, sector_t capacity, int set_new_bits) * * maybe bm_set should be atomic_t ? */ -static unsigned long _drbd_bm_total_weight(struct drbd_conf *mdev) +unsigned long _drbd_bm_total_weight(struct drbd_conf *mdev) { struct drbd_bitmap *b = mdev->bitmap; unsigned long s; diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h index 352441b..c07c370 100644 --- a/drivers/block/drbd/drbd_int.h +++ b/drivers/block/drbd/drbd_int.h @@ -337,13 +337,25 @@ static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c) * NOTE that the payload starts at a long aligned offset, * regardless of 32 or 64 bit arch! */ -struct p_header { +struct p_header80 { u32 magic; u16 command; u16 length; /* bytes of data after this header */ u8 payload[0]; } __packed; -/* 8 bytes. packet FIXED for the next century! */ + +/* Header for big packets, Used for data packets exceeding 64kB */ +struct p_header95 { + u16 magic; /* use DRBD_MAGIC_BIG here */ + u16 command; + u32 length; /* Use only 24 bits of that. Ignore the highest 8 bit. */ + u8 payload[0]; +} __packed; + +union p_header { + struct p_header80 h80; + struct p_header95 h95; +}; /* * short commands, packets without payload, plain p_header: @@ -362,12 +374,16 @@ struct p_header { */ /* these defines must not be changed without changing the protocol version */ -#define DP_HARDBARRIER 1 -#define DP_RW_SYNC 2 +#define DP_HARDBARRIER 1 /* depricated */ +#define DP_RW_SYNC 2 /* equals REQ_SYNC */ #define DP_MAY_SET_IN_SYNC 4 +#define DP_UNPLUG 8 /* equals REQ_UNPLUG */ +#define DP_FUA 16 /* equals REQ_FUA */ +#define DP_FLUSH 32 /* equals REQ_FLUSH */ +#define DP_DISCARD 64 /* equals REQ_DISCARD */ struct p_data { - struct p_header head; + union p_header head; u64 sector; /* 64 bits sector number */ u64 block_id; /* to identify the request in protocol B&C */ u32 seq_num; @@ -383,7 +399,7 @@ struct p_data { * P_DATA_REQUEST, P_RS_DATA_REQUEST */ struct p_block_ack { - struct p_header head; + struct p_header80 head; u64 sector; u64 block_id; u32 blksize; @@ -392,7 +408,7 @@ struct p_block_ack { struct p_block_req { - struct p_header head; + struct p_header80 head; u64 sector; u64 block_id; u32 blksize; @@ -409,7 +425,7 @@ struct p_block_req { */ struct p_handshake { - struct p_header head; /* 8 bytes */ + struct p_header80 head; /* 8 bytes */ u32 protocol_min; u32 feature_flags; u32 protocol_max; @@ -424,19 +440,19 @@ struct p_handshake { /* 80 bytes, FIXED for the next century */ struct p_barrier { - struct p_header head; + struct p_header80 head; u32 barrier; /* barrier number _handle_ only */ u32 pad; /* to multiple of 8 Byte */ } __packed; struct p_barrier_ack { - struct p_header head; + struct p_header80 head; u32 barrier; u32 set_size; } __packed; struct p_rs_param { - struct p_header head; + struct p_header80 head; u32 rate; /* Since protocol version 88 and higher. */ @@ -444,20 +460,31 @@ struct p_rs_param { } __packed; struct p_rs_param_89 { - struct p_header head; + struct p_header80 head; u32 rate; /* protocol version 89: */ char verify_alg[SHARED_SECRET_MAX]; char csums_alg[SHARED_SECRET_MAX]; } __packed; +struct p_rs_param_95 { + struct p_header80 head; + u32 rate; + char verify_alg[SHARED_SECRET_MAX]; + char csums_alg[SHARED_SECRET_MAX]; + u32 c_plan_ahead; + u32 c_delay_target; + u32 c_fill_target; + u32 c_max_rate; +} __packed; + enum drbd_conn_flags { CF_WANT_LOSE = 1, CF_DRY_RUN = 2, }; struct p_protocol { - struct p_header head; + struct p_header80 head; u32 protocol; u32 after_sb_0p; u32 after_sb_1p; @@ -471,17 +498,17 @@ struct p_protocol { } __packed; struct p_uuids { - struct p_header head; + struct p_header80 head; u64 uuid[UI_EXTENDED_SIZE]; } __packed; struct p_rs_uuid { - struct p_header head; + struct p_header80 head; u64 uuid; } __packed; struct p_sizes { - struct p_header head; + struct p_header80 head; u64 d_size; /* size of disk */ u64 u_size; /* user requested size */ u64 c_size; /* current exported size */ @@ -491,18 +518,18 @@ struct p_sizes { } __packed; struct p_state { - struct p_header head; + struct p_header80 head; u32 state; } __packed; struct p_req_state { - struct p_header head; + struct p_header80 head; u32 mask; u32 val; } __packed; struct p_req_state_reply { - struct p_header head; + struct p_header80 head; u32 retcode; } __packed; @@ -517,7 +544,7 @@ struct p_drbd06_param { } __packed; struct p_discard { - struct p_header head; + struct p_header80 head; u64 block_id; u32 seq_num; u32 pad; @@ -533,7 +560,7 @@ enum drbd_bitmap_code { }; struct p_compressed_bm { - struct p_header head; + struct p_header80 head; /* (encoding & 0x0f): actual encoding, see enum drbd_bitmap_code * (encoding & 0x80): polarity (set/unset) of first runlength * ((encoding >> 4) & 0x07): pad_bits, number of trailing zero bits @@ -544,10 +571,10 @@ struct p_compressed_bm { u8 code[0]; } __packed; -struct p_delay_probe { - struct p_header head; - u32 seq_num; /* sequence number to match the two probe packets */ - u32 offset; /* usecs the probe got sent after the reference time point */ +struct p_delay_probe93 { + struct p_header80 head; + u32 seq_num; /* sequence number to match the two probe packets */ + u32 offset; /* usecs the probe got sent after the reference time point */ } __packed; /* DCBP: Drbd Compressed Bitmap Packet ... */ @@ -594,7 +621,7 @@ DCBP_set_pad_bits(struct p_compressed_bm *p, int n) * so we need to use the fixed size 4KiB page size * most architechtures have used for a long time. */ -#define BM_PACKET_PAYLOAD_BYTES (4096 - sizeof(struct p_header)) +#define BM_PACKET_PAYLOAD_BYTES (4096 - sizeof(struct p_header80)) #define BM_PACKET_WORDS (BM_PACKET_PAYLOAD_BYTES/sizeof(long)) #define BM_PACKET_VLI_BYTES_MAX (4096 - sizeof(struct p_compressed_bm)) #if (PAGE_SIZE < 4096) @@ -603,13 +630,14 @@ DCBP_set_pad_bits(struct p_compressed_bm *p, int n) #endif union p_polymorph { - struct p_header header; + union p_header header; struct p_handshake handshake; struct p_data data; struct p_block_ack block_ack; struct p_barrier barrier; struct p_barrier_ack barrier_ack; struct p_rs_param_89 rs_param_89; + struct p_rs_param_95 rs_param_95; struct p_protocol protocol; struct p_sizes sizes; struct p_uuids uuids; @@ -617,6 +645,8 @@ union p_polymorph { struct p_req_state req_state; struct p_req_state_reply req_state_reply; struct p_block_req block_req; + struct p_delay_probe93 delay_probe93; + struct p_rs_uuid rs_uuid; } __packed; /**********************************************************************/ @@ -697,7 +727,7 @@ struct drbd_tl_epoch { struct list_head requests; /* requests before */ struct drbd_tl_epoch *next; /* pointer to the next barrier */ unsigned int br_number; /* the barriers identifier. */ - int n_req; /* number of requests attached before this barrier */ + int n_writes; /* number of requests attached before this barrier */ }; struct drbd_request; @@ -747,7 +777,7 @@ struct digest_info { struct drbd_epoch_entry { struct drbd_work w; struct hlist_node colision; - struct drbd_epoch *epoch; + struct drbd_epoch *epoch; /* for writes */ struct drbd_conf *mdev; struct page *pages; atomic_t pending_bios; @@ -755,7 +785,10 @@ struct drbd_epoch_entry { /* see comments on ee flag bits below */ unsigned long flags; sector_t sector; - u64 block_id; + union { + u64 block_id; + struct digest_info *digest; + }; }; /* ee flag bits. @@ -781,12 +814,16 @@ enum { * if any of those fail, we set this flag atomically * from the endio callback */ __EE_WAS_ERROR, + + /* This ee has a pointer to a digest instead of a block id */ + __EE_HAS_DIGEST, }; #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO) #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC) #define EE_IS_BARRIER (1<<__EE_IS_BARRIER) #define EE_RESUBMITTED (1<<__EE_RESUBMITTED) #define EE_WAS_ERROR (1<<__EE_WAS_ERROR) +#define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST) /* global flag bits */ enum { @@ -794,7 +831,6 @@ enum { SIGNAL_ASENDER, /* whether asender wants to be interrupted */ SEND_PING, /* whether asender should send a ping asap */ - STOP_SYNC_TIMER, /* tell timer to cancel itself */ UNPLUG_QUEUED, /* only relevant with kernel 2.4 */ UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */ MD_DIRTY, /* current uuids and flags not yet on disk */ @@ -816,6 +852,7 @@ enum { BITMAP_IO, /* suspend application io; once no more io in flight, start bitmap io */ BITMAP_IO_QUEUED, /* Started bitmap IO */ + GO_DISKLESS, /* Disk failed, local_cnt reached zero, we are going diskless */ RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */ NET_CONGESTED, /* The data socket is congested */ @@ -829,6 +866,8 @@ enum { * the peer, if it changed there as well. */ CONN_DRY_RUN, /* Expect disconnect after resync handshake. */ GOT_PING_ACK, /* set when we receive a ping_ack packet, misc wait gets woken */ + NEW_CUR_UUID, /* Create new current UUID when thawing IO */ + AL_SUSPENDED, /* Activity logging is currently suspended. */ }; struct drbd_bitmap; /* opaque for drbd_conf */ @@ -838,10 +877,6 @@ struct drbd_bitmap; /* opaque for drbd_conf */ /* THINK maybe we actually want to use the default "event/%s" worker threads * or similar in linux 2.6, which uses per cpu data and threads. - * - * To be general, this might need a spin_lock member. - * For now, please use the mdev->req_lock to protect list_head, - * see drbd_queue_work below. */ struct drbd_work_queue { struct list_head q; @@ -915,6 +950,12 @@ enum write_ordering_e { WO_bio_barrier }; +struct fifo_buffer { + int *values; + unsigned int head_index; + unsigned int size; +}; + struct drbd_conf { /* things that are stored as / read from meta data on disk */ unsigned long flags; @@ -936,9 +977,16 @@ struct drbd_conf { unsigned int ko_count; struct drbd_work resync_work, unplug_work, + go_diskless, md_sync_work; struct timer_list resync_timer; struct timer_list md_sync_timer; +#ifdef DRBD_DEBUG_MD_SYNC + struct { + unsigned int line; + const char* func; + } last_md_mark_dirty; +#endif /* Used after attach while negotiating new disk state. */ union drbd_state new_state_tmp; @@ -946,6 +994,7 @@ struct drbd_conf { union drbd_state state; wait_queue_head_t misc_wait; wait_queue_head_t state_wait; /* upon each state change. */ + wait_queue_head_t net_cnt_wait; unsigned int send_cnt; unsigned int recv_cnt; unsigned int read_cnt; @@ -974,12 +1023,16 @@ struct drbd_conf { unsigned long rs_start; /* cumulated time in PausedSyncX state [unit jiffies] */ unsigned long rs_paused; + /* skipped because csum was equal [unit BM_BLOCK_SIZE] */ + unsigned long rs_same_csum; +#define DRBD_SYNC_MARKS 8 +#define DRBD_SYNC_MARK_STEP (3*HZ) /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */ - unsigned long rs_mark_left; + unsigned long rs_mark_left[DRBD_SYNC_MARKS]; /* marks's time [unit jiffies] */ - unsigned long rs_mark_time; - /* skipped because csum was equeal [unit BM_BLOCK_SIZE] */ - unsigned long rs_same_csum; + unsigned long rs_mark_time[DRBD_SYNC_MARKS]; + /* current index into rs_mark_{left,time} */ + int rs_last_mark; /* where does the admin want us to start? (sector) */ sector_t ov_start_sector; @@ -1012,10 +1065,10 @@ struct drbd_conf { spinlock_t epoch_lock; unsigned int epochs; enum write_ordering_e write_ordering; - struct list_head active_ee; /* IO in progress */ - struct list_head sync_ee; /* IO in progress */ + struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */ + struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */ struct list_head done_ee; /* send ack */ - struct list_head read_ee; /* IO in progress */ + struct list_head read_ee; /* IO in progress (any read) */ struct list_head net_ee; /* zero-copy network send in progress */ struct hlist_head *ee_hash; /* is proteced by req_lock! */ unsigned int ee_hash_s; @@ -1026,7 +1079,8 @@ struct drbd_conf { int next_barrier_nr; struct hlist_head *app_reads_hash; /* is proteced by req_lock */ struct list_head resync_reads; - atomic_t pp_in_use; + atomic_t pp_in_use; /* allocated from page pool */ + atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */ wait_queue_head_t ee_wait; struct page *md_io_page; /* one page buffer for md_io */ struct page *md_io_tmpp; /* for logical_block_size != 512 */ @@ -1054,6 +1108,15 @@ struct drbd_conf { u64 ed_uuid; /* UUID of the exposed data */ struct mutex state_mutex; char congestion_reason; /* Why we where congested... */ + atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */ + atomic_t rs_sect_ev; /* for submitted resync data rate, both */ + int rs_last_sect_ev; /* counter to compare with */ + int rs_last_events; /* counter of read or write "events" (unit sectors) + * on the lower level device when we last looked. */ + int c_sync_rate; /* current resync rate after syncer throttle magic */ + struct fifo_buffer rs_plan_s; /* correction values of resync planer */ + int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */ + int rs_planed; /* resync sectors already planed */ }; static inline struct drbd_conf *minor_to_mdev(unsigned int minor) @@ -1138,6 +1201,8 @@ extern void drbd_free_resources(struct drbd_conf *mdev); extern void tl_release(struct drbd_conf *mdev, unsigned int barrier_nr, unsigned int set_size); extern void tl_clear(struct drbd_conf *mdev); +enum drbd_req_event; +extern void tl_restart(struct drbd_conf *mdev, enum drbd_req_event what); extern void _tl_add_barrier(struct drbd_conf *, struct drbd_tl_epoch *); extern void drbd_free_sock(struct drbd_conf *mdev); extern int drbd_send(struct drbd_conf *mdev, struct socket *sock, @@ -1150,12 +1215,12 @@ extern int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply, enum dds_f extern int _drbd_send_state(struct drbd_conf *mdev); extern int drbd_send_state(struct drbd_conf *mdev); extern int _drbd_send_cmd(struct drbd_conf *mdev, struct socket *sock, - enum drbd_packets cmd, struct p_header *h, + enum drbd_packets cmd, struct p_header80 *h, size_t size, unsigned msg_flags); #define USE_DATA_SOCKET 1 #define USE_META_SOCKET 0 extern int drbd_send_cmd(struct drbd_conf *mdev, int use_data_socket, - enum drbd_packets cmd, struct p_header *h, + enum drbd_packets cmd, struct p_header80 *h, size_t size); extern int drbd_send_cmd2(struct drbd_conf *mdev, enum drbd_packets cmd, char *data, size_t size); @@ -1167,7 +1232,7 @@ extern int drbd_send_ack(struct drbd_conf *mdev, enum drbd_packets cmd, extern int drbd_send_ack_rp(struct drbd_conf *mdev, enum drbd_packets cmd, struct p_block_req *rp); extern int drbd_send_ack_dp(struct drbd_conf *mdev, enum drbd_packets cmd, - struct p_data *dp); + struct p_data *dp, int data_size); extern int drbd_send_ack_ex(struct drbd_conf *mdev, enum drbd_packets cmd, sector_t sector, int blksize, u64 block_id); extern int drbd_send_block(struct drbd_conf *mdev, enum drbd_packets cmd, @@ -1201,7 +1266,13 @@ extern void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local) extern void drbd_md_set_flag(struct drbd_conf *mdev, int flags) __must_hold(local); extern void drbd_md_clear_flag(struct drbd_conf *mdev, int flags)__must_hold(local); extern int drbd_md_test_flag(struct drbd_backing_dev *, int); +#ifndef DRBD_DEBUG_MD_SYNC extern void drbd_md_mark_dirty(struct drbd_conf *mdev); +#else +#define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ ) +extern void drbd_md_mark_dirty_(struct drbd_conf *mdev, + unsigned int line, const char *func); +#endif extern void drbd_queue_bitmap_io(struct drbd_conf *mdev, int (*io_fn)(struct drbd_conf *), void (*done)(struct drbd_conf *, int), @@ -1209,6 +1280,7 @@ extern void drbd_queue_bitmap_io(struct drbd_conf *mdev, 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); /* Meta data layout @@ -1264,6 +1336,8 @@ struct bm_extent { * Bit 1 ==> local node thinks this block needs to be synced. */ +#define SLEEP_TIME (HZ/10) + #define BM_BLOCK_SHIFT 12 /* 4k per bit */ #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT) /* (9+3) : 512 bytes @ 8 bits; representing 16M storage @@ -1335,11 +1409,13 @@ struct bm_extent { #endif /* Sector shift value for the "hash" functions of tl_hash and ee_hash tables. - * With a value of 6 all IO in one 32K block make it to the same slot of the + * With a value of 8 all IO in one 128K block make it to the same slot of the * hash table. */ -#define HT_SHIFT 6 +#define HT_SHIFT 8 #define DRBD_MAX_SEGMENT_SIZE (1U<<(9+HT_SHIFT)) +#define DRBD_MAX_SIZE_H80_PACKET (1 << 15) /* The old header only allows packets up to 32Kib data */ + /* Number of elements in the app_reads_hash */ #define APP_R_HSIZE 15 @@ -1369,6 +1445,7 @@ extern unsigned long drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_ /* 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); extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf *mdev, unsigned long bm_fo); +extern unsigned long _drbd_bm_total_weight(struct drbd_conf *mdev); extern unsigned long drbd_bm_total_weight(struct drbd_conf *mdev); extern int drbd_bm_rs_done(struct drbd_conf *mdev); /* for receive_bitmap */ @@ -1421,7 +1498,8 @@ extern void resync_after_online_grow(struct drbd_conf *); extern void drbd_setup_queue_param(struct drbd_conf *mdev, unsigned int) __must_hold(local); extern int drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role, int force); -enum drbd_disk_state drbd_try_outdate_peer(struct drbd_conf *mdev); +extern enum drbd_disk_state drbd_try_outdate_peer(struct drbd_conf *mdev); +extern void drbd_try_outdate_peer_async(struct drbd_conf *mdev); extern int drbd_khelper(struct drbd_conf *mdev, char *cmd); /* drbd_worker.c */ @@ -1467,10 +1545,12 @@ extern int w_send_barrier(struct drbd_conf *, struct drbd_work *, int); extern int w_send_read_req(struct drbd_conf *, struct drbd_work *, int); extern int w_prev_work_done(struct drbd_conf *, struct drbd_work *, int); extern int w_e_reissue(struct drbd_conf *, struct drbd_work *, int); +extern int w_restart_disk_io(struct drbd_conf *, struct drbd_work *, int); extern void resync_timer_fn(unsigned long data); /* drbd_receiver.c */ +extern int drbd_rs_should_slow_down(struct drbd_conf *mdev); extern int drbd_submit_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e, const unsigned rw, const int fault_type); extern int drbd_release_ee(struct drbd_conf *mdev, struct list_head *list); @@ -1479,7 +1559,10 @@ extern struct drbd_epoch_entry *drbd_alloc_ee(struct drbd_conf *mdev, sector_t sector, unsigned int data_size, gfp_t gfp_mask) __must_hold(local); -extern void drbd_free_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e); +extern void drbd_free_some_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e, + int is_net); +#define drbd_free_ee(m,e) drbd_free_some_ee(m, e, 0) +#define drbd_free_net_ee(m,e) drbd_free_some_ee(m, e, 1) extern void drbd_wait_ee_list_empty(struct drbd_conf *mdev, struct list_head *head); extern void _drbd_wait_ee_list_empty(struct drbd_conf *mdev, @@ -1487,6 +1570,7 @@ extern void _drbd_wait_ee_list_empty(struct drbd_conf *mdev, extern void drbd_set_recv_tcq(struct drbd_conf *mdev, int tcq_enabled); extern void _drbd_clear_done_ee(struct drbd_conf *mdev, struct list_head *to_be_freed); extern void drbd_flush_workqueue(struct drbd_conf *mdev); +extern void drbd_free_tl_hash(struct drbd_conf *mdev); /* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to * mess with get_fs/set_fs, we know we are KERNEL_DS always. */ @@ -1600,6 +1684,8 @@ void drbd_bcast_ee(struct drbd_conf *mdev, #define susp_MASK 1 #define user_isp_MASK 1 #define aftr_isp_MASK 1 +#define susp_nod_MASK 1 +#define susp_fen_MASK 1 #define NS(T, S) \ ({ union drbd_state mask; mask.i = 0; mask.T = T##_MASK; mask; }), \ @@ -1856,13 +1942,6 @@ static inline sector_t drbd_md_ss__(struct drbd_conf *mdev, } static inline void -_drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w) -{ - list_add_tail(&w->list, &q->q); - up(&q->s); -} - -static inline void drbd_queue_work_front(struct drbd_work_queue *q, struct drbd_work *w) { unsigned long flags; @@ -1899,19 +1978,19 @@ static inline void request_ping(struct drbd_conf *mdev) static inline int drbd_send_short_cmd(struct drbd_conf *mdev, enum drbd_packets cmd) { - struct p_header h; + struct p_header80 h; return drbd_send_cmd(mdev, USE_DATA_SOCKET, cmd, &h, sizeof(h)); } static inline int drbd_send_ping(struct drbd_conf *mdev) { - struct p_header h; + struct p_header80 h; return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING, &h, sizeof(h)); } static inline int drbd_send_ping_ack(struct drbd_conf *mdev) { - struct p_header h; + struct p_header80 h; return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING_ACK, &h, sizeof(h)); } @@ -2013,7 +2092,7 @@ static inline void inc_unacked(struct drbd_conf *mdev) static inline void put_net_conf(struct drbd_conf *mdev) { if (atomic_dec_and_test(&mdev->net_cnt)) - wake_up(&mdev->misc_wait); + wake_up(&mdev->net_cnt_wait); } /** @@ -2044,10 +2123,14 @@ static inline int get_net_conf(struct drbd_conf *mdev) static inline void put_ldev(struct drbd_conf *mdev) { + int i = atomic_dec_return(&mdev->local_cnt); __release(local); - if (atomic_dec_and_test(&mdev->local_cnt)) + D_ASSERT(i >= 0); + if (i == 0) { + if (mdev->state.disk == D_FAILED) + drbd_go_diskless(mdev); wake_up(&mdev->misc_wait); - D_ASSERT(atomic_read(&mdev->local_cnt) >= 0); + } } #ifndef __CHECKER__ @@ -2179,11 +2262,16 @@ static inline int drbd_state_is_stable(union drbd_state s) return 1; } +static inline int is_susp(union drbd_state s) +{ + return s.susp || s.susp_nod || s.susp_fen; +} + static inline int __inc_ap_bio_cond(struct drbd_conf *mdev) { int mxb = drbd_get_max_buffers(mdev); - if (mdev->state.susp) + if (is_susp(mdev->state)) return 0; if (test_bit(SUSPEND_IO, &mdev->flags)) return 0; diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index e4b5611..c5dfe64 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -78,6 +78,7 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, static int w_md_sync(struct drbd_conf *mdev, struct drbd_work *w, int unused); static void md_sync_timer_fn(unsigned long data); static int w_bitmap_io(struct drbd_conf *mdev, struct drbd_work *w, int unused); +static int w_go_diskless(struct drbd_conf *mdev, struct drbd_work *w, int unused); MODULE_AUTHOR("Philipp Reisner <phil@linbit.com>, " "Lars Ellenberg <lars@linbit.com>"); @@ -200,7 +201,7 @@ static int tl_init(struct drbd_conf *mdev) INIT_LIST_HEAD(&b->w.list); b->next = NULL; b->br_number = 4711; - b->n_req = 0; + b->n_writes = 0; b->w.cb = NULL; /* if this is != NULL, we need to dec_ap_pending in tl_clear */ mdev->oldest_tle = b; @@ -241,7 +242,7 @@ void _tl_add_barrier(struct drbd_conf *mdev, struct drbd_tl_epoch *new) INIT_LIST_HEAD(&new->w.list); new->w.cb = NULL; /* if this is != NULL, we need to dec_ap_pending in tl_clear */ new->next = NULL; - new->n_req = 0; + new->n_writes = 0; newest_before = mdev->newest_tle; /* never send a barrier number == 0, because that is special-cased @@ -285,9 +286,9 @@ void tl_release(struct drbd_conf *mdev, unsigned int barrier_nr, barrier_nr, b->br_number); goto bail; } - if (b->n_req != set_size) { - dev_err(DEV, "BAD! BarrierAck #%u received with n_req=%u, expected n_req=%u!\n", - barrier_nr, set_size, b->n_req); + if (b->n_writes != set_size) { + dev_err(DEV, "BAD! BarrierAck #%u received with n_writes=%u, expected n_writes=%u!\n", + barrier_nr, set_size, b->n_writes); goto bail; } @@ -334,6 +335,82 @@ bail: drbd_force_state(mdev, NS(conn, C_PROTOCOL_ERROR)); } +/** + * _tl_restart() - Walks the transfer log, and applies an action to all requests + * @mdev: DRBD device. + * @what: The action/event to perform with all request objects + * + * @what might be one of connection_lost_while_pending, resend, fail_frozen_disk_io, + * restart_frozen_disk_io. + */ +static void _tl_restart(struct drbd_conf *mdev, enum drbd_req_event what) +{ + struct drbd_tl_epoch *b, *tmp, **pn; + struct list_head *le, *tle, carry_reads; + struct drbd_request *req; + int rv, n_writes, n_reads; + + b = mdev->oldest_tle; + pn = &mdev->oldest_tle; + while (b) { + n_writes = 0; + n_reads = 0; + INIT_LIST_HEAD(&carry_reads); + list_for_each_safe(le, tle, &b->requests) { + req = list_entry(le, struct drbd_request, tl_requests); + rv = _req_mod(req, what); + + n_writes += (rv & MR_WRITE) >> MR_WRITE_SHIFT; + n_reads += (rv & MR_READ) >> MR_READ_SHIFT; + } + tmp = b->next; + + if (n_writes) { + if (what == resend) { + b->n_writes = n_writes; + if (b->w.cb == NULL) { + b->w.cb = w_send_barrier; + inc_ap_pending(mdev); + set_bit(CREATE_BARRIER, &mdev->flags); + } + + drbd_queue_work(&mdev->data.work, &b->w); + } + pn = &b->next; + } else { + if (n_reads) + list_add(&carry_reads, &b->requests); + /* there could still be requests on that ring list, + * in case local io is still pending */ + list_del(&b->requests); + + /* dec_ap_pending corresponding to queue_barrier. + * the newest barrier may not have been queued yet, + * in which case w.cb is still NULL. */ + if (b->w.cb != NULL) + dec_ap_pending(mdev); + + if (b == mdev->newest_tle) { + /* recycle, but reinit! */ + D_ASSERT(tmp == NULL); + INIT_LIST_HEAD(&b->requests); + list_splice(&carry_reads, &b->requests); + INIT_LIST_HEAD(&b->w.list); + b->w.cb = NULL; + b->br_number = net_random(); + b->n_writes = 0; + + *pn = b; + break; + } + *pn = tmp; + kfree(b); + } + b = tmp; + list_splice(&carry_reads, &b->requests); + } +} + /** * tl_clear() - Clears all requests and &struct drbd_tl_epoch objects out of the TL @@ -345,48 +422,12 @@ bail: */ void tl_clear(struct drbd_conf *mdev) { - struct drbd_tl_epoch *b, *tmp; struct list_head *le, *tle; struct drbd_request *r; - int new_initial_bnr = net_random(); spin_lock_irq(&mdev->req_lock); - b = mdev->oldest_tle; - while (b) { - list_for_each_safe(le, tle, &b->requests) { - r = list_entry(le, struct drbd_request, tl_requests); - /* It would be nice to complete outside of spinlock. - * But this is easier for now. */ - _req_mod(r, connection_lost_while_pending); - } - tmp = b->next; - - /* there could still be requests on that ring list, - * in case local io is still pending */ - list_del(&b->requests); - - /* dec_ap_pending corresponding to queue_barrier. - * the newest barrier may not have been queued yet, - * in which case w.cb is still NULL. */ - if (b->w.cb != NULL) - dec_ap_pending(mdev); - - if (b == mdev->newest_tle) { - /* recycle, but reinit! */ - D_ASSERT(tmp == NULL); - INIT_LIST_HEAD(&b->requests); - INIT_LIST_HEAD(&b->w.list); - b->w.cb = NULL; - b->br_number = new_initial_bnr; - b->n_req = 0; - - mdev->oldest_tle = b; - break; - } - kfree(b); - b = tmp; - } + _tl_restart(mdev, connection_lost_while_pending); /* we expect this list to be empty. */ D_ASSERT(list_empty(&mdev->out_of_sequence_requests)); @@ -402,6 +443,15 @@ void tl_clear(struct drbd_conf *mdev) /* ensure bit indicating barrier is required is clear */ clear_bit(CREATE_BARRIER, &mdev->flags); + memset(mdev->app_reads_hash, 0, APP_R_HSIZE*sizeof(void *)); + + spin_unlock_irq(&mdev->req_lock); +} + +void tl_restart(struct drbd_conf *mdev, enum drbd_req_event what) +{ + spin_lock_irq(&mdev->req_lock); + _tl_restart(mdev, what); spin_unlock_irq(&mdev->req_lock); } @@ -456,7 +506,7 @@ static int is_valid_state(struct drbd_conf *mdev, union drbd_state ns); static int is_valid_state_transition(struct drbd_conf *, union drbd_state, union drbd_state); static union drbd_state sanitize_state(struct drbd_conf *mdev, union drbd_state os, - union drbd_state ns, int *warn_sync_abort); + union drbd_state ns, const char **warn_sync_abort); int drbd_send_state_req(struct drbd_conf *, union drbd_state, union drbd_state); @@ -606,7 +656,7 @@ static void print_st(struct drbd_conf *mdev, char *name, union drbd_state ns) drbd_role_str(ns.peer), drbd_disk_str(ns.disk), drbd_disk_str(ns.pdsk), - ns.susp ? 's' : 'r', + is_susp(ns) ? 's' : 'r', ns.aftr_isp ? 'a' : '-', ns.peer_isp ? 'p' : '-', ns.user_isp ? 'u' : '-' @@ -764,7 +814,7 @@ static int is_valid_state_transition(struct drbd_conf *mdev, * to D_UNKNOWN. This rule and many more along those lines are in this function. */ static union drbd_state sanitize_state(struct drbd_conf *mdev, union drbd_state os, - union drbd_state ns, int *warn_sync_abort) + union drbd_state ns, const char **warn_sync_abort) { enum drbd_fencing_p fp; @@ -779,9 +829,10 @@ static union drbd_state sanitize_state(struct drbd_conf *mdev, union drbd_state os.conn <= C_DISCONNECTING) ns.conn = os.conn; - /* After a network error (+C_TEAR_DOWN) only C_UNCONNECTED or C_DISCONNECTING can follow */ + /* After a network error (+C_TEAR_DOWN) only C_UNCONNECTED or C_DISCONNECTING can follow. + * If you try to go into some Sync* state, that shall fail (elsewhere). */ if (os.conn >= C_TIMEOUT && os.conn <= C_TEAR_DOWN && - ns.conn != C_UNCONNECTED && ns.conn != C_DISCONNECTING) + ns.conn != C_UNCONNECTED && ns.conn != C_DISCONNECTING && ns.conn <= C_TEAR_DOWN) ns.conn = os.conn; /* After C_DISCONNECTING only C_STANDALONE may follow */ @@ -799,14 +850,13 @@ static union drbd_state sanitize_state(struct drbd_conf *mdev, union drbd_state if (ns.conn == C_STANDALONE && ns.disk == D_DISKLESS && ns.role == R_SECONDARY) ns.aftr_isp = 0; - if (ns.conn <= C_DISCONNECTING && ns.disk == D_DISKLESS) - ns.pdsk = D_UNKNOWN; - /* Abort resync if a disk fails/detaches */ if (os.conn > C_CONNECTED && ns.conn > C_CONNECTED && (ns.disk <= D_FAILED || ns.pdsk <= D_FAILED)) { if (warn_sync_abort) - *warn_sync_abort = 1; + *warn_sync_abort = + os.conn == C_VERIFY_S || os.conn == C_VERIFY_T ? + "Online-verify" : "Resync"; ns.conn = C_CONNECTED; } @@ -877,7 +927,12 @@ static union drbd_state sanitize_state(struct drbd_conf *mdev, union drbd_state if (fp == FP_STONITH && (ns.role == R_PRIMARY && ns.conn < C_CONNECTED && ns.pdsk > D_OUTDATED) && !(os.role == R_PRIMARY && os.conn < C_CONNECTED && os.pdsk > D_OUTDATED)) - ns.susp = 1; + ns.susp_fen = 1; /* Suspend IO while fence-peer handler runs (peer lost) */ + + if (mdev->sync_conf.on_no_data == OND_SUSPEND_IO && + (ns.role == R_PRIMARY && ns.disk < D_UP_TO_DATE && ns.pdsk < D_UP_TO_DATE) && + !(os.role == R_PRIMARY && os.disk < D_UP_TO_DATE && os.pdsk < D_UP_TO_DATE)) + ns.susp_nod = 1; /* Suspend IO while no data available (no accessible data available) */ if (ns.aftr_isp || ns.peer_isp || ns.user_isp) { if (ns.conn == C_SYNC_SOURCE) @@ -913,6 +968,12 @@ static void set_ov_position(struct drbd_conf *mdev, enum drbd_conns cs) } } +static void drbd_resume_al(struct drbd_conf *mdev) +{ + if (test_and_clear_bit(AL_SUSPENDED, &mdev->flags)) + dev_info(DEV, "Resumed AL updates\n"); +} + /** * __drbd_set_state() - Set a new DRBD state * @mdev: DRBD device. @@ -928,7 +989,7 @@ int __drbd_set_state(struct drbd_conf *mdev, { union drbd_state os; int rv = SS_SUCCESS; - int warn_sync_abort = 0; + const char *warn_sync_abort = NULL; struct after_state_chg_work *ascw; os = mdev->state; @@ -947,14 +1008,8 @@ int __drbd_set_state(struct drbd_conf *mdev, /* If the old state was illegal as well, then let this happen...*/ - if (is_valid_state(mdev, os) == rv) { - dev_err(DEV, "Considering state change from bad state. " - "Error would be: '%s'\n", - drbd_set_st_err_str(rv)); - print_st(mdev, "old", os); - print_st(mdev, "new", ns); + if (is_valid_state(mdev, os) == rv) rv = is_valid_state_transition(mdev, ns, os); - } } else rv = is_valid_state_transition(mdev, ns, os); } @@ -966,7 +1021,7 @@ int __drbd_set_state(struct drbd_conf *mdev, } if (warn_sync_abort) - dev_warn(DEV, "Resync aborted.\n"); + dev_warn(DEV, "%s aborted.\n", warn_sync_abort); { char *pbp, pb[300]; @@ -977,7 +1032,10 @@ int __drbd_set_state(struct drbd_conf *mdev, PSC(conn); PSC(disk); PSC(pdsk); - PSC(susp); + if (is_susp(ns) != is_susp(os)) + pbp += sprintf(pbp, "susp( %s -> %s ) ", + drbd_susp_str(is_susp(os)), + drbd_susp_str(is_susp(ns))); PSC(aftr_isp); PSC(peer_isp); PSC(user_isp); @@ -1002,12 +1060,6 @@ int __drbd_set_state(struct drbd_conf *mdev, wake_up(&mdev->misc_wait); wake_up(&mdev->state_wait); - /* post-state-change actions */ - if (os.conn >= C_SYNC_SOURCE && ns.conn <= C_CONNECTED) { - set_bit(STOP_SYNC_TIMER, &mdev->flags); - mod_timer(&mdev->resync_timer, jiffies); - } - /* aborted verify run. log the last position */ if ((os.conn == C_VERIFY_S || os.conn == C_VERIFY_T) && ns.conn < C_CONNECTED) { @@ -1020,41 +1072,42 @@ int __drbd_set_state(struct drbd_conf *mdev, if ((os.conn == C_PAUSED_SYNC_T || os.conn == C_PAUSED_SYNC_S) && (ns.conn == C_SYNC_TARGET || ns.conn == C_SYNC_SOURCE)) { dev_info(DEV, "Syncer continues.\n"); - mdev->rs_paused += (long)jiffies-(long)mdev->rs_mark_time; - if (ns.conn == C_SYNC_TARGET) { - if (!test_and_clear_bit(STOP_SYNC_TIMER, &mdev->flags)) - mod_timer(&mdev->resync_timer, jiffies); - /* This if (!test_bit) is only needed for the case - that a device that has ceased to used its timer, - i.e. it is already in drbd_resync_finished() gets - paused and resumed. */ - } + mdev->rs_paused += (long)jiffies + -(long)mdev->rs_mark_time[mdev->rs_last_mark]; + if (ns.conn == C_SYNC_TARGET) + mod_timer(&mdev->resync_timer, jiffies); } if ((os.conn == C_SYNC_TARGET || os.conn == C_SYNC_SOURCE) && (ns.conn == C_PAUSED_SYNC_T || ns.conn == C_PAUSED_SYNC_S)) { dev_info(DEV, "Resync suspended\n"); - mdev->rs_mark_time = jiffies; - if (ns.conn == C_PAUSED_SYNC_T) - set_bit(STOP_SYNC_TIMER, &mdev->flags); + mdev->rs_mark_time[mdev->rs_last_mark] = jiffies; } if (os.conn == C_CONNECTED && (ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T)) { + unsigned long now = jiffies; + int i; + mdev->ov_position = 0; - mdev->rs_total = - mdev->rs_mark_left = drbd_bm_bits(mdev); + mdev->rs_total = drbd_bm_bits(mdev); if (mdev->agreed_pro_version >= 90) set_ov_position(mdev, ns.conn); else mdev->ov_start_sector = 0; mdev->ov_left = mdev->rs_total - BM_SECT_TO_BIT(mdev->ov_position); - mdev->rs_start = - mdev->rs_mark_time = jiffies; + mdev->rs_start = now; + mdev->rs_last_events = 0; + mdev->rs_last_sect_ev = 0; mdev->ov_last_oos_size = 0; mdev->ov_last_oos_start = 0; + for (i = 0; i < DRBD_SYNC_MARKS; i++) { + mdev->rs_mark_left[i] = mdev->rs_total; + mdev->rs_mark_time[i] = now; + } + if (ns.conn == C_VERIFY_S) { dev_info(DEV, "Starting Online Verify from sector %llu\n", (unsigned long long)mdev->ov_position); @@ -1107,6 +1160,10 @@ int __drbd_set_state(struct drbd_conf *mdev, ns.conn <= C_TEAR_DOWN && ns.conn >= C_TIMEOUT) drbd_thread_restart_nowait(&mdev->receiver); + /* Resume AL writing if we get a connection */ + if (os.conn < C_CONNECTED && ns.conn >= C_CONNECTED) + drbd_resume_al(mdev); + ascw = kmalloc(sizeof(*ascw), GFP_ATOMIC); if (ascw) { ascw->os = os; @@ -1165,6 +1222,8 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, union drbd_state ns, enum chg_state_flags flags) { enum drbd_fencing_p fp; + enum drbd_req_event what = nothing; + union drbd_state nsm = (union drbd_state){ .i = -1 }; if (os.conn != C_CONNECTED && ns.conn == C_CONNECTED) { clear_bit(CRASHED_PRIMARY, &mdev->flags); @@ -1188,17 +1247,49 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, /* Here we have the actions that are performed after a state change. This function might sleep */ - if (fp == FP_STONITH && ns.susp) { - /* case1: The outdate peer handler is successful: - * case2: The connection was established again: */ - if ((os.pdsk > D_OUTDATED && ns.pdsk <= D_OUTDATED) || - (os.conn < C_CONNECTED && ns.conn >= C_CONNECTED)) { + nsm.i = -1; + if (ns.susp_nod) { + if (os.conn < C_CONNECTED && ns.conn >= C_CONNECTED) { + if (ns.conn == C_CONNECTED) + what = resend, nsm.susp_nod = 0; + else /* ns.conn > C_CONNECTED */ + dev_err(DEV, "Unexpected Resynd going on!\n"); + } + + if (os.disk == D_ATTACHING && ns.disk > D_ATTACHING) + what = restart_frozen_disk_io, nsm.susp_nod = 0; + + } + + if (ns.susp_fen) { + /* case1: The outdate peer handler is successful: */ + if (os.pdsk > D_OUTDATED && ns.pdsk <= D_OUTDATED) { tl_clear(mdev); + if (test_bit(NEW_CUR_UUID, &mdev->flags)) { + drbd_uuid_new_current(mdev); + clear_bit(NEW_CUR_UUID, &mdev->flags); + drbd_md_sync(mdev); + } spin_lock_irq(&mdev->req_lock); - _drbd_set_state(_NS(mdev, susp, 0), CS_VERBOSE, NULL); + _drbd_set_state(_NS(mdev, susp_fen, 0), CS_VERBOSE, NULL); spin_unlock_irq(&mdev->req_lock); } + /* case2: The connection was established again: */ + if (os.conn < C_CONNECTED && ns.conn >= C_CONNECTED) { + clear_bit(NEW_CUR_UUID, &mdev->flags); + what = resend; + nsm.susp_fen = 0; + } + } + + if (what != nothing) { + spin_lock_irq(&mdev->req_lock); + _tl_restart(mdev, what); + nsm.i &= mdev->state.i; + _drbd_set_state(mdev, nsm, CS_VERBOSE, NULL); + spin_unlock_irq(&mdev->req_lock); } + /* Do not change the order of the if above and the two below... */ if (os.pdsk == D_DISKLESS && ns.pdsk > D_DISKLESS) { /* attach on the peer */ drbd_send_uuids(mdev); @@ -1217,16 +1308,22 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, if (get_ldev(mdev)) { if ((ns.role == R_PRIMARY || ns.peer == R_PRIMARY) && mdev->ldev->md.uuid[UI_BITMAP] == 0 && ns.disk >= D_UP_TO_DATE) { - drbd_uuid_new_current(mdev); - drbd_send_uuids(mdev); + if (is_susp(mdev->state)) { + set_bit(NEW_CUR_UUID, &mdev->flags); + } else { + drbd_uuid_new_current(mdev); + drbd_send_uuids(mdev); + } } put_ldev(mdev); } } if (ns.pdsk < D_INCONSISTENT && get_ldev(mdev)) { - if (ns.peer == R_PRIMARY && mdev->ldev->md.uuid[UI_BITMAP] == 0) + if (ns.peer == R_PRIMARY && mdev->ldev->md.uuid[UI_BITMAP] == 0) { drbd_uuid_new_current(mdev); + drbd_send_uuids(mdev); + } /* D_DISKLESS Peer becomes secondary */ if (os.peer == R_PRIMARY && ns.peer == R_SECONDARY) @@ -1268,42 +1365,51 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, os.disk > D_INCONSISTENT && ns.disk == D_INCONSISTENT) drbd_queue_bitmap_io(mdev, &drbd_bmio_set_n_write, NULL, "set_n_write from invalidate"); + /* first half of local IO error */ if (os.disk > D_FAILED && ns.disk == D_FAILED) { - enum drbd_io_error_p eh; + enum drbd_io_error_p eh = EP_PASS_ON; + + if (drbd_send_state(mdev)) + dev_warn(DEV, "Notified peer that my disk is broken.\n"); + else + dev_err(DEV, "Sending state for drbd_io_error() failed\n"); + + drbd_rs_cancel_all(mdev); - eh = EP_PASS_ON; if (get_ldev_if_state(mdev, D_FAILED)) { eh = mdev->ldev->dc.on_io_error; put_ldev(mdev); } + if (eh == EP_CALL_HELPER) + drbd_khelper(mdev, "local-io-error"); + } - drbd_rs_cancel_all(mdev); - /* since get_ldev() only works as long as disk>=D_INCONSISTENT, - and it is D_DISKLESS here, local_cnt can only go down, it can - not increase... It will reach zero */ - wait_event(mdev->misc_wait, !atomic_read(&mdev->local_cnt)); + + /* second half of local IO error handling, + * after local_cnt references have reached zero: */ + if (os.disk == D_FAILED && ns.disk == D_DISKLESS) { mdev->rs_total = 0; mdev->rs_failed = 0; atomic_set(&mdev->rs_pending_cnt, 0); - - spin_lock_irq(&mdev->req_lock); - _drbd_set_state(_NS(mdev, disk, D_DISKLESS), CS_HARD, NULL); - spin_unlock_irq(&mdev->req_lock); - - if (eh == EP_CALL_HELPER) - drbd_khelper(mdev, "local-io-error"); } if (os.disk > D_DISKLESS && ns.disk == D_DISKLESS) { + /* We must still be diskless, + * re-attach has to be serialized with this! */ + if (mdev->state.disk != D_DISKLESS) + dev_err(DEV, + "ASSERT FAILED: disk is %s while going diskless\n", + drbd_disk_str(mdev->state.disk)); + + /* we cannot assert local_cnt == 0 here, as get_ldev_if_state + * will inc/dec it frequently. Since we became D_DISKLESS, no + * one has touched the protected members anymore, though, so we + * are safe to free them here. */ + if (drbd_send_state(mdev)) + dev_warn(DEV, "Notified peer that I detached my disk.\n"); + else + dev_err(DEV, "Sending state for detach failed\n"); - if (os.disk == D_FAILED) /* && ns.disk == D_DISKLESS*/ { - if (drbd_send_state(mdev)) - dev_warn(DEV, "Notified peer that my disk is broken.\n"); - else - dev_err(DEV, "Sending state in drbd_io_error() failed\n"); - } - - wait_event(mdev->misc_wait, !atomic_read(&mdev->local_cnt)); lc_destroy(mdev->resync); mdev->resync = NULL; lc_destroy(mdev->act_log); @@ -1312,8 +1418,10 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, drbd_free_bc(mdev->ldev); mdev->ldev = NULL;); - if (mdev->md_io_tmpp) + if (mdev->md_io_tmpp) { __free_page(mdev->md_io_tmpp); + mdev->md_io_tmpp = NULL; + } } /* Disks got bigger while they were detached */ @@ -1329,6 +1437,15 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, (os.user_isp && !ns.user_isp)) resume_next_sg(mdev); + /* sync target done with resync. Explicitly notify peer, even though + * it should (at least for non-empty resyncs) already know itself. */ + if (os.disk < D_UP_TO_DATE && os.conn >= C_SYNC_SOURCE && ns.conn == C_CONNECTED) + drbd_send_state(mdev); + + /* free tl_hash if we Got thawed and are C_STANDALONE */ + if (ns.conn == C_STANDALONE && !is_susp(ns) && mdev->tl_hash) + drbd_free_tl_hash(mdev); + /* Upon network connection, we need to start the receiver */ if (os.conn == C_STANDALONE && ns.conn == C_UNCONNECTED) drbd_thread_start(&mdev->receiver); @@ -1555,7 +1672,7 @@ void drbd_thread_current_set_cpu(struct drbd_conf *mdev) /* the appropriate socket mutex must be held already */ int _drbd_send_cmd(struct drbd_conf *mdev, struct socket *sock, - enum drbd_packets cmd, struct p_header *h, + enum drbd_packets cmd, struct p_header80 *h, size_t size, unsigned msg_flags) { int sent, ok; @@ -1565,7 +1682,7 @@ int _drbd_send_cmd(struct drbd_conf *mdev, struct socket *sock, h->magic = BE_DRBD_MAGIC; h->command = cpu_to_be16(cmd); - h->length = cpu_to_be16(size-sizeof(struct p_header)); + h->length = cpu_to_be16(size-sizeof(struct p_header80)); sent = drbd_send(mdev, sock, h, size, msg_flags); @@ -1580,7 +1697,7 @@ int _drbd_send_cmd(struct drbd_conf *mdev, struct socket *sock, * when we hold the appropriate socket mutex. */ int drbd_send_cmd(struct drbd_conf *mdev, int use_data_socket, - enum drbd_packets cmd, struct p_header *h, size_t size) + enum drbd_packets cmd, struct p_header80 *h, size_t size) { int ok = 0; struct socket *sock; @@ -1608,7 +1725,7 @@ int drbd_send_cmd(struct drbd_conf *mdev, int use_data_socket, int drbd_send_cmd2(struct drbd_conf *mdev, enum drbd_packets cmd, char *data, size_t size) { - struct p_header h; + struct p_header80 h; int ok; h.magic = BE_DRBD_MAGIC; @@ -1630,7 +1747,7 @@ int drbd_send_cmd2(struct drbd_conf *mdev, enum drbd_packets cmd, char *data, int drbd_send_sync_param(struct drbd_conf *mdev, struct syncer_conf *sc) { - struct p_rs_param_89 *p; + struct p_rs_param_95 *p; struct socket *sock; int size, rv; const int apv = mdev->agreed_pro_version; @@ -1638,7 +1755,8 @@ int drbd_send_sync_param(struct drbd_conf *mdev, struct syncer_conf *sc) size = apv <= 87 ? sizeof(struct p_rs_param) : apv == 88 ? sizeof(struct p_rs_param) + strlen(mdev->sync_conf.verify_alg) + 1 - : /* 89 */ sizeof(struct p_rs_param_89); + : apv <= 94 ? sizeof(struct p_rs_param_89) + : /* apv >= 95 */ sizeof(struct p_rs_param_95); /* used from admin command context and receiver/worker context. * to avoid kmalloc, grab the socket right here, @@ -1649,12 +1767,16 @@ int drbd_send_sync_param(struct drbd_conf *mdev, struct syncer_conf *sc) if (likely(sock != NULL)) { enum drbd_packets cmd = apv >= 89 ? P_SYNC_PARAM89 : P_SYNC_PARAM; - p = &mdev->data.sbuf.rs_param_89; + p = &mdev->data.sbuf.rs_param_95; /* initialize verify_alg and csums_alg */ memset(p->verify_alg, 0, 2 * SHARED_SECRET_MAX); p->rate = cpu_to_be32(sc->rate); + p->c_plan_ahead = cpu_to_be32(sc->c_plan_ahead); + p->c_delay_target = cpu_to_be32(sc->c_delay_target); + p->c_fill_target = cpu_to_be32(sc->c_fill_target); + p->c_max_rate = cpu_to_be32(sc->c_max_rate); if (apv >= 88) strcpy(p->verify_alg, mdev->sync_conf.verify_alg); @@ -1710,7 +1832,7 @@ int drbd_send_protocol(struct drbd_conf *mdev) strcpy(p->integrity_alg, mdev->net_conf->integrity_alg); rv = drbd_send_cmd(mdev, USE_DATA_SOCKET, P_PROTOCOL, - (struct p_header *)p, size); + (struct p_header80 *)p, size); kfree(p); return rv; } @@ -1736,7 +1858,7 @@ int _drbd_send_uuids(struct drbd_conf *mdev, u64 uuid_flags) put_ldev(mdev); return drbd_send_cmd(mdev, USE_DATA_SOCKET, P_UUIDS, - (struct p_header *)&p, sizeof(p)); + (struct p_header80 *)&p, sizeof(p)); } int drbd_send_uuids(struct drbd_conf *mdev) @@ -1757,7 +1879,7 @@ int drbd_send_sync_uuid(struct drbd_conf *mdev, u64 val) p.uuid = cpu_to_be64(val); return drbd_send_cmd(mdev, USE_DATA_SOCKET, P_SYNC_UUID, - (struct p_header *)&p, sizeof(p)); + (struct p_header80 *)&p, sizeof(p)); } int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply, enum dds_flags flags) @@ -1787,7 +1909,7 @@ int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply, enum dds_flags fl p.dds_flags = cpu_to_be16(flags); ok = drbd_send_cmd(mdev, USE_DATA_SOCKET, P_SIZES, - (struct p_header *)&p, sizeof(p)); + (struct p_header80 *)&p, sizeof(p)); return ok; } @@ -1812,7 +1934,7 @@ int drbd_send_state(struct drbd_conf *mdev) if (likely(sock != NULL)) { ok = _drbd_send_cmd(mdev, sock, P_STATE, - (struct p_header *)&p, sizeof(p), 0); + (struct p_header80 *)&p, sizeof(p), 0); } mutex_unlock(&mdev->data.mutex); @@ -1830,7 +1952,7 @@ int drbd_send_state_req(struct drbd_conf *mdev, p.val = cpu_to_be32(val.i); return drbd_send_cmd(mdev, USE_DATA_SOCKET, P_STATE_CHG_REQ, - (struct p_header *)&p, sizeof(p)); + (struct p_header80 *)&p, sizeof(p)); } int drbd_send_sr_reply(struct drbd_conf *mdev, int retcode) @@ -1840,7 +1962,7 @@ int drbd_send_sr_reply(struct drbd_conf *mdev, int retcode) p.retcode = cpu_to_be32(retcode); return drbd_send_cmd(mdev, USE_META_SOCKET, P_STATE_CHG_REPLY, - (struct p_header *)&p, sizeof(p)); + (struct p_header80 *)&p, sizeof(p)); } int fill_bitmap_rle_bits(struct drbd_conf *mdev, @@ -1939,7 +2061,7 @@ int fill_bitmap_rle_bits(struct drbd_conf *mdev, enum { OK, FAILED, DONE } send_bitmap_rle_or_plain(struct drbd_conf *mdev, - struct p_header *h, struct bm_xfer_ctx *c) + struct p_header80 *h, struct bm_xfer_ctx *c) { struct p_compressed_bm *p = (void*)h; unsigned long num_words; @@ -1969,12 +2091,12 @@ send_bitmap_rle_or_plain(struct drbd_conf *mdev, if (len) drbd_bm_get_lel(mdev, c->word_offset, num_words, (unsigned long*)h->payload); ok = _drbd_send_cmd(mdev, mdev->data.socket, P_BITMAP, - h, sizeof(struct p_header) + len, 0); + h, sizeof(struct p_header80) + len, 0); c->word_offset += num_words; c->bit_offset = c->word_offset * BITS_PER_LONG; c->packets[1]++; - c->bytes[1] += sizeof(struct p_header) + len; + c->bytes[1] += sizeof(struct p_header80) + len; if (c->bit_offset > c->bm_bits) c->bit_offset = c->bm_bits; @@ -1990,14 +2112,14 @@ send_bitmap_rle_or_plain(struct drbd_conf *mdev, int _drbd_send_bitmap(struct drbd_conf *mdev) { struct bm_xfer_ctx c; - struct p_header *p; + struct p_header80 *p; int ret; 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_header *) __get_free_page(GFP_NOIO); + 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; @@ -2055,7 +2177,7 @@ int drbd_send_b_ack(struct drbd_conf *mdev, u32 barrier_nr, u32 set_size) if (mdev->state.conn < C_CONNECTED) return FALSE; ok = drbd_send_cmd(mdev, USE_META_SOCKET, P_BARRIER_ACK, - (struct p_header *)&p, sizeof(p)); + (struct p_header80 *)&p, sizeof(p)); return ok; } @@ -2083,17 +2205,18 @@ static int _drbd_send_ack(struct drbd_conf *mdev, enum drbd_packets cmd, if (!mdev->meta.socket || mdev->state.conn < C_CONNECTED) return FALSE; ok = drbd_send_cmd(mdev, USE_META_SOCKET, cmd, - (struct p_header *)&p, sizeof(p)); + (struct p_header80 *)&p, sizeof(p)); return ok; } +/* dp->sector and dp->block_id already/still in network byte order, + * data_size is payload size according to dp->head, + * and may need to be corrected for digest size. */ int drbd_send_ack_dp(struct drbd_conf *mdev, enum drbd_packets cmd, - struct p_data *dp) + struct p_data *dp, int data_size) { - const int header_size = sizeof(struct p_data) - - sizeof(struct p_header); - int data_size = ((struct p_header *)dp)->length - header_size; - + data_size -= (mdev->agreed_pro_version >= 87 && mdev->integrity_r_tfm) ? + crypto_hash_digestsize(mdev->integrity_r_tfm) : 0; return _drbd_send_ack(mdev, cmd, dp->sector, cpu_to_be32(data_size), dp->block_id); } @@ -2141,7 +2264,7 @@ int drbd_send_drequest(struct drbd_conf *mdev, int cmd, p.blksize = cpu_to_be32(size); ok = drbd_send_cmd(mdev, USE_DATA_SOCKET, cmd, - (struct p_header *)&p, sizeof(p)); + (struct p_header80 *)&p, sizeof(p)); return ok; } @@ -2159,7 +2282,7 @@ int drbd_send_drequest_csum(struct drbd_conf *mdev, p.head.magic = BE_DRBD_MAGIC; p.head.command = cpu_to_be16(cmd); - p.head.length = cpu_to_be16(sizeof(p) - sizeof(struct p_header) + digest_size); + p.head.length = cpu_to_be16(sizeof(p) - sizeof(struct p_header80) + digest_size); mutex_lock(&mdev->data.mutex); @@ -2181,7 +2304,7 @@ int drbd_send_ov_request(struct drbd_conf *mdev, sector_t sector, int size) p.blksize = cpu_to_be32(size); ok = drbd_send_cmd(mdev, USE_DATA_SOCKET, P_OV_REQUEST, - (struct p_header *)&p, sizeof(p)); + (struct p_header80 *)&p, sizeof(p)); return ok; } @@ -2333,6 +2456,18 @@ static int _drbd_send_zc_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e) return 1; } +static u32 bio_flags_to_wire(struct drbd_conf *mdev, unsigned long bi_rw) +{ + if (mdev->agreed_pro_version >= 95) + return (bi_rw & REQ_SYNC ? DP_RW_SYNC : 0) | + (bi_rw & REQ_UNPLUG ? DP_UNPLUG : 0) | + (bi_rw & REQ_FUA ? DP_FUA : 0) | + (bi_rw & REQ_FLUSH ? DP_FLUSH : 0) | + (bi_rw & REQ_DISCARD ? DP_DISCARD : 0); + else + return bi_rw & (REQ_SYNC | REQ_UNPLUG) ? DP_RW_SYNC : 0; +} + /* Used to send write requests * R_PRIMARY -> Peer (P_DATA) */ @@ -2350,30 +2485,25 @@ int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req) dgs = (mdev->agreed_pro_version >= 87 && mdev->integrity_w_tfm) ? crypto_hash_digestsize(mdev->integrity_w_tfm) : 0; - p.head.magic = BE_DRBD_MAGIC; - p.head.command = cpu_to_be16(P_DATA); - p.head.length = - cpu_to_be16(sizeof(p) - sizeof(struct p_header) + dgs + req->size); + if (req->size <= DRBD_MAX_SIZE_H80_PACKET) { + p.head.h80.magic = BE_DRBD_MAGIC; + p.head.h80.command = cpu_to_be16(P_DATA); + p.head.h80.length = + cpu_to_be16(sizeof(p) - sizeof(union p_header) + dgs + req->size); + } else { + p.head.h95.magic = BE_DRBD_MAGIC_BIG; + p.head.h95.command = cpu_to_be16(P_DATA); + p.head.h95.length = + cpu_to_be32(sizeof(p) - sizeof(union p_header) + dgs + req->size); + } p.sector = cpu_to_be64(req->sector); p.block_id = (unsigned long)req; p.seq_num = cpu_to_be32(req->seq_num = atomic_add_return(1, &mdev->packet_seq)); - dp_flags = 0; - /* NOTE: no need to check if barriers supported here as we would - * not pass the test in make_request_common in that case - */ - if (req->master_bio->bi_rw & REQ_HARDBARRIER) { - dev_err(DEV, "ASSERT FAILED would have set DP_HARDBARRIER\n"); - /* dp_flags |= DP_HARDBARRIER; */ - } - if (req->master_bio->bi_rw & REQ_SYNC) - dp_flags |= DP_RW_SYNC; - /* for now handle SYNCIO and UNPLUG - * as if they still were one and the same flag */ - if (req->master_bio->bi_rw & REQ_UNPLUG) - dp_flags |= DP_RW_SYNC; + dp_flags = bio_flags_to_wire(mdev, req->master_bio->bi_rw); + if (mdev->state.conn >= C_SYNC_SOURCE && mdev->state.conn <= C_PAUSED_SYNC_T) dp_flags |= DP_MAY_SET_IN_SYNC; @@ -2414,10 +2544,17 @@ int drbd_send_block(struct drbd_conf *mdev, enum drbd_packets cmd, dgs = (mdev->agreed_pro_version >= 87 && mdev->integrity_w_tfm) ? crypto_hash_digestsize(mdev->integrity_w_tfm) : 0; - p.head.magic = BE_DRBD_MAGIC; - p.head.command = cpu_to_be16(cmd); - p.head.length = - cpu_to_be16(sizeof(p) - sizeof(struct p_header) + dgs + e->size); + if (e->size <= DRBD_MAX_SIZE_H80_PACKET) { + p.head.h80.magic = BE_DRBD_MAGIC; + p.head.h80.command = cpu_to_be16(cmd); + p.head.h80.length = + cpu_to_be16(sizeof(p) - sizeof(struct p_header80) + dgs + e->size); + } else { + p.head.h95.magic = BE_DRBD_MAGIC_BIG; + p.head.h95.command = cpu_to_be16(cmd); + p.head.h95.length = + cpu_to_be32(sizeof(p) - sizeof(struct p_header80) + dgs + e->size); + } p.sector = cpu_to_be64(e->sector); p.block_id = e->block_id; @@ -2430,8 +2567,7 @@ int drbd_send_block(struct drbd_conf *mdev, enum drbd_packets cmd, if (!drbd_get_data_sock(mdev)) return 0; - ok = sizeof(p) == drbd_send(mdev, mdev->data.socket, &p, - sizeof(p), dgs ? MSG_MORE : 0); + ok = sizeof(p) == drbd_send(mdev, mdev->data.socket, &p, sizeof(p), dgs ? MSG_MORE : 0); if (ok && dgs) { dgb = mdev->int_dig_out; drbd_csum_ee(mdev, mdev->integrity_w_tfm, e, dgb); @@ -2606,7 +2742,13 @@ static void drbd_set_defaults(struct drbd_conf *mdev) /* .verify_alg = */ {}, 0, /* .cpu_mask = */ {}, 0, /* .csums_alg = */ {}, 0, - /* .use_rle = */ 0 + /* .use_rle = */ 0, + /* .on_no_data = */ DRBD_ON_NO_DATA_DEF, + /* .c_plan_ahead = */ DRBD_C_PLAN_AHEAD_DEF, + /* .c_delay_target = */ DRBD_C_DELAY_TARGET_DEF, + /* .c_fill_target = */ DRBD_C_FILL_TARGET_DEF, + /* .c_max_rate = */ DRBD_C_MAX_RATE_DEF, + /* .c_min_rate = */ DRBD_C_MIN_RATE_DEF }; /* Have to use that way, because the layout differs between @@ -2617,7 +2759,9 @@ static void drbd_set_defaults(struct drbd_conf *mdev) .conn = C_STANDALONE, .disk = D_DISKLESS, .pdsk = D_UNKNOWN, - .susp = 0 + .susp = 0, + .susp_nod = 0, + .susp_fen = 0 } }; } @@ -2641,6 +2785,9 @@ void drbd_init_set_defaults(struct drbd_conf *mdev) atomic_set(&mdev->net_cnt, 0); atomic_set(&mdev->packet_seq, 0); atomic_set(&mdev->pp_in_use, 0); + atomic_set(&mdev->pp_in_use_by_net, 0); + atomic_set(&mdev->rs_sect_in, 0); + atomic_set(&mdev->rs_sect_ev, 0); mutex_init(&mdev->md_io_mutex); mutex_init(&mdev->data.mutex); @@ -2667,11 +2814,13 @@ void drbd_init_set_defaults(struct drbd_conf *mdev) INIT_LIST_HEAD(&mdev->meta.work.q); INIT_LIST_HEAD(&mdev->resync_work.list); INIT_LIST_HEAD(&mdev->unplug_work.list); + INIT_LIST_HEAD(&mdev->go_diskless.list); INIT_LIST_HEAD(&mdev->md_sync_work.list); INIT_LIST_HEAD(&mdev->bm_io_work.w.list); mdev->resync_work.cb = w_resync_inactive; mdev->unplug_work.cb = w_send_write_hint; + mdev->go_diskless.cb = w_go_diskless; mdev->md_sync_work.cb = w_md_sync; mdev->bm_io_work.w.cb = w_bitmap_io; init_timer(&mdev->resync_timer); @@ -2683,6 +2832,7 @@ void drbd_init_set_defaults(struct drbd_conf *mdev) init_waitqueue_head(&mdev->misc_wait); init_waitqueue_head(&mdev->state_wait); + init_waitqueue_head(&mdev->net_cnt_wait); init_waitqueue_head(&mdev->ee_wait); init_waitqueue_head(&mdev->al_wait); init_waitqueue_head(&mdev->seq_wait); @@ -2698,6 +2848,7 @@ void drbd_init_set_defaults(struct drbd_conf *mdev) void drbd_mdev_cleanup(struct drbd_conf *mdev) { + int i; if (mdev->receiver.t_state != None) dev_err(DEV, "ASSERT FAILED: receiver t_state == %d expected 0.\n", mdev->receiver.t_state); @@ -2714,9 +2865,13 @@ void drbd_mdev_cleanup(struct drbd_conf *mdev) mdev->p_size = mdev->rs_start = mdev->rs_total = - mdev->rs_failed = - mdev->rs_mark_left = - mdev->rs_mark_time = 0; + mdev->rs_failed = 0; + mdev->rs_last_events = 0; + mdev->rs_last_sect_ev = 0; + for (i = 0; i < DRBD_SYNC_MARKS; i++) { + mdev->rs_mark_left[i] = 0; + mdev->rs_mark_time[i] = 0; + } D_ASSERT(mdev->net_conf == NULL); drbd_set_my_capacity(mdev, 0); @@ -2727,6 +2882,7 @@ void drbd_mdev_cleanup(struct drbd_conf *mdev) } drbd_free_resources(mdev); + clear_bit(AL_SUSPENDED, &mdev->flags); /* * currently we drbd_init_ee only on module load, so @@ -2742,6 +2898,7 @@ void drbd_mdev_cleanup(struct drbd_conf *mdev) D_ASSERT(list_empty(&mdev->meta.work.q)); D_ASSERT(list_empty(&mdev->resync_work.list)); D_ASSERT(list_empty(&mdev->unplug_work.list)); + D_ASSERT(list_empty(&mdev->go_diskless.list)); } @@ -3281,9 +3438,10 @@ void drbd_md_sync(struct drbd_conf *mdev) sector_t sector; int i; + del_timer(&mdev->md_sync_timer); + /* timer may be rearmed by drbd_md_mark_dirty() now. */ if (!test_and_clear_bit(MD_DIRTY, &mdev->flags)) return; - del_timer(&mdev->md_sync_timer); /* We use here D_FAILED and not D_ATTACHING because we try to write * metadata even if we detach due to a disk failure! */ @@ -3311,12 +3469,9 @@ void drbd_md_sync(struct drbd_conf *mdev) D_ASSERT(drbd_md_ss__(mdev, mdev->ldev) == mdev->ldev->md.md_offset); sector = mdev->ldev->md.md_offset; - if (drbd_md_sync_page_io(mdev, mdev->ldev, sector, WRITE)) { - clear_bit(MD_DIRTY, &mdev->flags); - } else { + 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); } @@ -3403,6 +3558,28 @@ int drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev) return rv; } +static void debug_drbd_uuid(struct drbd_conf *mdev, enum drbd_uuid_index index) +{ + static char *uuid_str[UI_EXTENDED_SIZE] = { + [UI_CURRENT] = "CURRENT", + [UI_BITMAP] = "BITMAP", + [UI_HISTORY_START] = "HISTORY_START", + [UI_HISTORY_END] = "HISTORY_END", + [UI_SIZE] = "SIZE", + [UI_FLAGS] = "FLAGS", + }; + + if (index >= UI_EXTENDED_SIZE) { + dev_warn(DEV, " uuid_index >= EXTENDED_SIZE\n"); + return; + } + + dynamic_dev_dbg(DEV, " uuid[%s] now %016llX\n", + uuid_str[index], + (unsigned long long)mdev->ldev->md.uuid[index]); +} + + /** * drbd_md_mark_dirty() - Mark meta data super block as dirty * @mdev: DRBD device. @@ -3411,19 +3588,31 @@ int drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev) * the meta-data super block. This function sets MD_DIRTY, and starts a * timer that ensures that within five seconds you have to call drbd_md_sync(). */ +#ifdef DEBUG +void drbd_md_mark_dirty_(struct drbd_conf *mdev, unsigned int line, const char *func) +{ + if (!test_and_set_bit(MD_DIRTY, &mdev->flags)) { + mod_timer(&mdev->md_sync_timer, jiffies + HZ); + mdev->last_md_mark_dirty.line = line; + mdev->last_md_mark_dirty.func = func; + } +} +#else void drbd_md_mark_dirty(struct drbd_conf *mdev) { - set_bit(MD_DIRTY, &mdev->flags); - mod_timer(&mdev->md_sync_timer, jiffies + 5*HZ); + if (!test_and_set_bit(MD_DIRTY, &mdev->flags)) + mod_timer(&mdev->md_sync_timer, jiffies + 5*HZ); } - +#endif static void drbd_uuid_move_history(struct drbd_conf *mdev) __must_hold(local) { int i; - for (i = UI_HISTORY_START; i < UI_HISTORY_END; i++) + for (i = UI_HISTORY_START; i < UI_HISTORY_END; i++) { mdev->ldev->md.uuid[i+1] = mdev->ldev->md.uuid[i]; + debug_drbd_uuid(mdev, i+1); + } } void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local) @@ -3438,6 +3627,7 @@ void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local) } mdev->ldev->md.uuid[idx] = val; + debug_drbd_uuid(mdev, idx); drbd_md_mark_dirty(mdev); } @@ -3447,6 +3637,7 @@ void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local) if (mdev->ldev->md.uuid[idx]) { drbd_uuid_move_history(mdev); mdev->ldev->md.uuid[UI_HISTORY_START] = mdev->ldev->md.uuid[idx]; + debug_drbd_uuid(mdev, UI_HISTORY_START); } _drbd_uuid_set(mdev, idx, val); } @@ -3465,6 +3656,7 @@ void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local) dev_info(DEV, "Creating new current UUID\n"); D_ASSERT(mdev->ldev->md.uuid[UI_BITMAP] == 0); mdev->ldev->md.uuid[UI_BITMAP] = mdev->ldev->md.uuid[UI_CURRENT]; + debug_drbd_uuid(mdev, UI_BITMAP); get_random_bytes(&val, sizeof(u64)); _drbd_uuid_set(mdev, UI_CURRENT, val); @@ -3479,6 +3671,8 @@ void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local) drbd_uuid_move_history(mdev); mdev->ldev->md.uuid[UI_HISTORY_START] = mdev->ldev->md.uuid[UI_BITMAP]; mdev->ldev->md.uuid[UI_BITMAP] = 0; + debug_drbd_uuid(mdev, UI_HISTORY_START); + debug_drbd_uuid(mdev, UI_BITMAP); } else { if (mdev->ldev->md.uuid[UI_BITMAP]) dev_warn(DEV, "bm UUID already set"); @@ -3486,6 +3680,7 @@ void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local) mdev->ldev->md.uuid[UI_BITMAP] = val; mdev->ldev->md.uuid[UI_BITMAP] &= ~((u64)1); + debug_drbd_uuid(mdev, UI_BITMAP); } drbd_md_mark_dirty(mdev); } @@ -3528,6 +3723,7 @@ int drbd_bmio_clear_n_write(struct drbd_conf *mdev) { int rv = -EIO; + drbd_resume_al(mdev); if (get_ldev_if_state(mdev, D_ATTACHING)) { drbd_bm_clear_all(mdev); rv = drbd_bm_write(mdev); @@ -3560,6 +3756,32 @@ static int w_bitmap_io(struct drbd_conf *mdev, struct drbd_work *w, int unused) return 1; } +static int w_go_diskless(struct drbd_conf *mdev, struct drbd_work *w, int unused) +{ + D_ASSERT(mdev->state.disk == D_FAILED); + /* we cannot assert local_cnt == 0 here, as get_ldev_if_state will + * inc/dec it frequently. Once we are D_DISKLESS, no one will touch + * the protected members anymore, though, so in the after_state_ch work + * it will be safe to free them. */ + drbd_force_state(mdev, NS(disk, D_DISKLESS)); + /* We need to wait for return of references checked out while we still + * have been D_FAILED, though (drbd_md_sync, bitmap io). */ + wait_event(mdev->misc_wait, !atomic_read(&mdev->local_cnt)); + + clear_bit(GO_DISKLESS, &mdev->flags); + return 1; +} + +void drbd_go_diskless(struct drbd_conf *mdev) +{ + D_ASSERT(mdev->state.disk == D_FAILED); + if (!test_and_set_bit(GO_DISKLESS, &mdev->flags)) + drbd_queue_work(&mdev->data.work, &mdev->go_diskless); + /* don't drbd_queue_work_front, + * we need to serialize with the after_state_ch work + * of the -> D_FAILED transition. */ +} + /** * drbd_queue_bitmap_io() - Queues an IO operation on the whole bitmap * @mdev: DRBD device. @@ -3656,8 +3878,11 @@ static void md_sync_timer_fn(unsigned long data) static int w_md_sync(struct drbd_conf *mdev, struct drbd_work *w, int unused) { dev_warn(DEV, "md_sync_timer expired! Worker calls drbd_md_sync().\n"); +#ifdef DEBUG + dev_warn(DEV, "last md_mark_dirty: %s:%u\n", + mdev->last_md_mark_dirty.func, mdev->last_md_mark_dirty.line); +#endif drbd_md_sync(mdev); - return 1; } diff --git a/drivers/block/drbd/drbd_nl.c b/drivers/block/drbd/drbd_nl.c index 73131c5..87925e9 100644 --- a/drivers/block/drbd/drbd_nl.c +++ b/drivers/block/drbd/drbd_nl.c @@ -33,10 +33,13 @@ #include <linux/blkpg.h> #include <linux/cpumask.h> #include "drbd_int.h" +#include "drbd_req.h" #include "drbd_wrappers.h" #include <asm/unaligned.h> #include <linux/drbd_tag_magic.h> #include <linux/drbd_limits.h> +#include <linux/compiler.h> +#include <linux/kthread.h> static unsigned short *tl_add_blob(unsigned short *, enum drbd_tags, const void *, int); static unsigned short *tl_add_str(unsigned short *, enum drbd_tags, const char *); @@ -169,6 +172,10 @@ int drbd_khelper(struct drbd_conf *mdev, char *cmd) put_net_conf(mdev); } + /* The helper may take some time. + * write out any unsynced meta data changes now */ + drbd_md_sync(mdev); + dev_info(DEV, "helper command: %s %s %s\n", usermode_helper, cmd, mb); drbd_bcast_ev_helper(mdev, cmd); @@ -202,12 +209,10 @@ enum drbd_disk_state drbd_try_outdate_peer(struct drbd_conf *mdev) put_ldev(mdev); } else { dev_warn(DEV, "Not fencing peer, I'm not even Consistent myself.\n"); - return mdev->state.pdsk; + nps = mdev->state.pdsk; + goto out; } - if (fp == FP_STONITH) - _drbd_request_state(mdev, NS(susp, 1), CS_WAIT_COMPLETE); - r = drbd_khelper(mdev, "fence-peer"); switch ((r>>8) & 0xff) { @@ -252,9 +257,36 @@ enum drbd_disk_state drbd_try_outdate_peer(struct drbd_conf *mdev) dev_info(DEV, "fence-peer helper returned %d (%s)\n", (r>>8) & 0xff, ex_to_string); + +out: + if (mdev->state.susp_fen && nps >= D_UNKNOWN) { + /* The handler was not successful... unfreeze here, the + state engine can not unfreeze... */ + _drbd_request_state(mdev, NS(susp_fen, 0), CS_VERBOSE); + } + return nps; } +static int _try_outdate_peer_async(void *data) +{ + struct drbd_conf *mdev = (struct drbd_conf *)data; + enum drbd_disk_state nps; + + nps = drbd_try_outdate_peer(mdev); + drbd_request_state(mdev, NS(pdsk, nps)); + + return 0; +} + +void drbd_try_outdate_peer_async(struct drbd_conf *mdev) +{ + struct task_struct *opa; + + opa = kthread_run(_try_outdate_peer_async, mdev, "drbd%d_a_helper", mdev_to_minor(mdev)); + if (IS_ERR(opa)) + dev_err(DEV, "out of mem, failed to invoke fence-peer helper\n"); +} int drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role, int force) { @@ -394,6 +426,39 @@ int drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role, int force) return r; } +static struct drbd_conf *ensure_mdev(int minor, int create) +{ + struct drbd_conf *mdev; + + if (minor >= minor_count) + return NULL; + + mdev = minor_to_mdev(minor); + + if (!mdev && create) { + struct gendisk *disk = NULL; + mdev = drbd_new_device(minor); + + spin_lock_irq(&drbd_pp_lock); + if (minor_table[minor] == NULL) { + minor_table[minor] = mdev; + disk = mdev->vdisk; + mdev = NULL; + } /* else: we lost the race */ + spin_unlock_irq(&drbd_pp_lock); + + if (disk) /* we won the race above */ + /* in case we ever add a drbd_delete_device(), + * don't forget the del_gendisk! */ + add_disk(disk); + else /* we lost the race above */ + drbd_free_mdev(mdev); + + mdev = minor_to_mdev(minor); + } + + return mdev; +} static int drbd_nl_primary(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp, struct drbd_nl_cfg_reply *reply) @@ -494,6 +559,8 @@ char *ppsize(char *buf, unsigned long long size) void drbd_suspend_io(struct drbd_conf *mdev) { set_bit(SUSPEND_IO, &mdev->flags); + if (is_susp(mdev->state)) + return; wait_event(mdev->misc_wait, !atomic_read(&mdev->ap_bio_cnt)); } @@ -713,9 +780,6 @@ void drbd_setup_queue_param(struct drbd_conf *mdev, unsigned int max_seg_s) __mu blk_queue_segment_boundary(q, PAGE_SIZE-1); blk_stack_limits(&q->limits, &b->limits, 0); - if (b->merge_bvec_fn) - dev_warn(DEV, "Backing device's merge_bvec_fn() = %p\n", - b->merge_bvec_fn); dev_info(DEV, "max_segment_size ( = BIO size ) = %u\n", queue_max_segment_size(q)); if (q->backing_dev_info.ra_pages != b->backing_dev_info.ra_pages) { @@ -729,14 +793,16 @@ void drbd_setup_queue_param(struct drbd_conf *mdev, unsigned int max_seg_s) __mu /* serialize deconfig (worker exiting, doing cleanup) * and reconfig (drbdsetup disk, drbdsetup net) * - * wait for a potentially exiting worker, then restart it, - * or start a new one. + * Wait for a potentially exiting worker, then restart it, + * or start a new one. Flush any pending work, there may still be an + * after_state_change queued. */ static void drbd_reconfig_start(struct drbd_conf *mdev) { wait_event(mdev->state_wait, !test_and_set_bit(CONFIG_PENDING, &mdev->flags)); wait_event(mdev->state_wait, !test_bit(DEVICE_DYING, &mdev->flags)); drbd_thread_start(&mdev->worker); + drbd_flush_workqueue(mdev); } /* if still unconfigured, stops worker again. @@ -756,6 +822,29 @@ static void drbd_reconfig_done(struct drbd_conf *mdev) wake_up(&mdev->state_wait); } +/* Make sure IO is suspended before calling this function(). */ +static void drbd_suspend_al(struct drbd_conf *mdev) +{ + int s = 0; + + if (lc_try_lock(mdev->act_log)) { + drbd_al_shrink(mdev); + lc_unlock(mdev->act_log); + } else { + dev_warn(DEV, "Failed to lock al in drbd_suspend_al()\n"); + return; + } + + spin_lock_irq(&mdev->req_lock); + if (mdev->state.conn < C_CONNECTED) + s = !test_and_set_bit(AL_SUSPENDED, &mdev->flags); + + spin_unlock_irq(&mdev->req_lock); + + if (s) + dev_info(DEV, "Suspended AL updates\n"); +} + /* does always return 0; * interesting return code is in reply->ret_code */ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp, @@ -769,6 +858,7 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp struct inode *inode, *inode2; struct lru_cache *resync_lru = NULL; union drbd_state ns, os; + unsigned int max_seg_s; int rv; int cp_discovered = 0; int logical_block_size; @@ -803,6 +893,15 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp goto fail; } + if (get_net_conf(mdev)) { + int prot = mdev->net_conf->wire_protocol; + put_net_conf(mdev); + if (nbc->dc.fencing == FP_STONITH && prot == DRBD_PROT_A) { + retcode = ERR_STONITH_AND_PROT_A; + goto fail; + } + } + nbc->lo_file = filp_open(nbc->dc.backing_dev, O_RDWR, 0); if (IS_ERR(nbc->lo_file)) { dev_err(DEV, "open(\"%s\") failed with %ld\n", nbc->dc.backing_dev, @@ -924,7 +1023,7 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp drbd_suspend_io(mdev); /* also wait for the last barrier ack. */ - wait_event(mdev->misc_wait, !atomic_read(&mdev->ap_pending_cnt)); + wait_event(mdev->misc_wait, !atomic_read(&mdev->ap_pending_cnt) || is_susp(mdev->state)); /* and for any other previously queued work */ drbd_flush_workqueue(mdev); @@ -1021,7 +1120,8 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp else clear_bit(CRASHED_PRIMARY, &mdev->flags); - if (drbd_md_test_flag(mdev->ldev, MDF_PRIMARY_IND)) { + if (drbd_md_test_flag(mdev->ldev, MDF_PRIMARY_IND) && + !(mdev->state.role == R_PRIMARY && mdev->state.susp_nod)) { set_bit(CRASHED_PRIMARY, &mdev->flags); cp_discovered = 1; } @@ -1031,7 +1131,20 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp mdev->read_cnt = 0; mdev->writ_cnt = 0; - drbd_setup_queue_param(mdev, DRBD_MAX_SEGMENT_SIZE); + max_seg_s = DRBD_MAX_SEGMENT_SIZE; + if (mdev->state.conn == C_CONNECTED) { + /* We are Primary, Connected, and now attach a new local + * backing store. We must not increase the user visible maximum + * bio size on this device to something the peer may not be + * able to handle. */ + if (mdev->agreed_pro_version < 94) + max_seg_s = queue_max_segment_size(mdev->rq_queue); + else if (mdev->agreed_pro_version == 94) + max_seg_s = DRBD_MAX_SIZE_H80_PACKET; + /* else: drbd 8.3.9 and later, stay with default */ + } + + drbd_setup_queue_param(mdev, max_seg_s); /* If I am currently not R_PRIMARY, * but meta data primary indicator is set, @@ -1079,6 +1192,9 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp drbd_al_to_on_disk_bm(mdev); } + if (_drbd_bm_total_weight(mdev) == drbd_bm_bits(mdev)) + drbd_suspend_al(mdev); /* IO is still suspended here... */ + spin_lock_irq(&mdev->req_lock); os = mdev->state; ns.i = os.i; @@ -1235,7 +1351,16 @@ static int drbd_nl_net_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp, && (new_conf->wire_protocol != DRBD_PROT_C)) { retcode = ERR_NOT_PROTO_C; goto fail; - }; + } + + if (get_ldev(mdev)) { + enum drbd_fencing_p fp = mdev->ldev->dc.fencing; + put_ldev(mdev); + if (new_conf->wire_protocol == DRBD_PROT_A && fp == FP_STONITH) { + retcode = ERR_STONITH_AND_PROT_A; + goto fail; + } + } if (mdev->state.role == R_PRIMARY && new_conf->want_lose) { retcode = ERR_DISCARD; @@ -1350,6 +1475,7 @@ static int drbd_nl_net_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp, } } + drbd_flush_workqueue(mdev); spin_lock_irq(&mdev->req_lock); if (mdev->net_conf != NULL) { retcode = ERR_NET_CONFIGURED; @@ -1388,10 +1514,9 @@ static int drbd_nl_net_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp, mdev->int_dig_out=int_dig_out; mdev->int_dig_in=int_dig_in; mdev->int_dig_vv=int_dig_vv; + retcode = _drbd_set_state(_NS(mdev, conn, C_UNCONNECTED), CS_VERBOSE, NULL); spin_unlock_irq(&mdev->req_lock); - retcode = _drbd_request_state(mdev, NS(conn, C_UNCONNECTED), CS_VERBOSE); - kobject_uevent(&disk_to_dev(mdev->vdisk)->kobj, KOBJ_CHANGE); reply->ret_code = retcode; drbd_reconfig_done(mdev); @@ -1546,6 +1671,8 @@ static int drbd_nl_syncer_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *n struct crypto_hash *csums_tfm = NULL; struct syncer_conf sc; cpumask_var_t new_cpu_mask; + int *rs_plan_s = NULL; + int fifo_size; if (!zalloc_cpumask_var(&new_cpu_mask, GFP_KERNEL)) { retcode = ERR_NOMEM; @@ -1557,6 +1684,12 @@ static int drbd_nl_syncer_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *n sc.rate = DRBD_RATE_DEF; sc.after = DRBD_AFTER_DEF; sc.al_extents = DRBD_AL_EXTENTS_DEF; + sc.on_no_data = DRBD_ON_NO_DATA_DEF; + sc.c_plan_ahead = DRBD_C_PLAN_AHEAD_DEF; + sc.c_delay_target = DRBD_C_DELAY_TARGET_DEF; + sc.c_fill_target = DRBD_C_FILL_TARGET_DEF; + sc.c_max_rate = DRBD_C_MAX_RATE_DEF; + sc.c_min_rate = DRBD_C_MIN_RATE_DEF; } else memcpy(&sc, &mdev->sync_conf, sizeof(struct syncer_conf)); @@ -1634,6 +1767,12 @@ static int drbd_nl_syncer_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *n } #undef AL_MAX + /* to avoid spurious errors when configuring minors before configuring + * the minors they depend on: if necessary, first create the minor we + * depend on */ + if (sc.after >= 0) + ensure_mdev(sc.after, 1); + /* most sanity checks done, try to assign the new sync-after * dependency. need to hold the global lock in there, * to avoid a race in the dependency loop check. */ @@ -1641,6 +1780,16 @@ static int drbd_nl_syncer_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *n if (retcode != NO_ERROR) goto fail; + fifo_size = (sc.c_plan_ahead * 10 * SLEEP_TIME) / HZ; + if (fifo_size != mdev->rs_plan_s.size && fifo_size > 0) { + rs_plan_s = kzalloc(sizeof(int) * fifo_size, GFP_KERNEL); + if (!rs_plan_s) { + dev_err(DEV, "kmalloc of fifo_buffer failed"); + retcode = ERR_NOMEM; + goto fail; + } + } + /* ok, assign the rest of it as well. * lock against receive_SyncParam() */ spin_lock(&mdev->peer_seq_lock); @@ -1657,6 +1806,15 @@ static int drbd_nl_syncer_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *n mdev->verify_tfm = verify_tfm; verify_tfm = NULL; } + + if (fifo_size != mdev->rs_plan_s.size) { + kfree(mdev->rs_plan_s.values); + mdev->rs_plan_s.values = rs_plan_s; + mdev->rs_plan_s.size = fifo_size; + mdev->rs_planed = 0; + rs_plan_s = NULL; + } + spin_unlock(&mdev->peer_seq_lock); if (get_ldev(mdev)) { @@ -1688,6 +1846,7 @@ static int drbd_nl_syncer_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *n kobject_uevent(&disk_to_dev(mdev->vdisk)->kobj, KOBJ_CHANGE); fail: + kfree(rs_plan_s); free_cpumask_var(new_cpu_mask); crypto_free_hash(csums_tfm); crypto_free_hash(verify_tfm); @@ -1721,12 +1880,38 @@ static int drbd_nl_invalidate(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nl return 0; } +static int drbd_bmio_set_susp_al(struct drbd_conf *mdev) +{ + int rv; + + rv = drbd_bmio_set_n_write(mdev); + drbd_suspend_al(mdev); + return rv; +} + static int drbd_nl_invalidate_peer(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp, struct drbd_nl_cfg_reply *reply) { + int retcode; - reply->ret_code = drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_S)); + retcode = _drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_S), CS_ORDERED); + + if (retcode < SS_SUCCESS) { + if (retcode == SS_NEED_CONNECTION && mdev->state.role == R_PRIMARY) { + /* The peer will get a resync upon connect anyways. Just make that + into a full resync. */ + retcode = drbd_request_state(mdev, NS(pdsk, D_INCONSISTENT)); + if (retcode >= SS_SUCCESS) { + /* open coded drbd_bitmap_io() */ + if (drbd_bitmap_io(mdev, &drbd_bmio_set_susp_al, + "set_n_write from invalidate_peer")) + retcode = ERR_IO_MD_DISK; + } + } else + retcode = drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_S)); + } + reply->ret_code = retcode; return 0; } @@ -1765,7 +1950,21 @@ static int drbd_nl_suspend_io(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nl static int drbd_nl_resume_io(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp, struct drbd_nl_cfg_reply *reply) { - reply->ret_code = drbd_request_state(mdev, NS(susp, 0)); + if (test_bit(NEW_CUR_UUID, &mdev->flags)) { + drbd_uuid_new_current(mdev); + clear_bit(NEW_CUR_UUID, &mdev->flags); + drbd_md_sync(mdev); + } + drbd_suspend_io(mdev); + reply->ret_code = drbd_request_state(mdev, NS3(susp, 0, susp_nod, 0, susp_fen, 0)); + if (reply->ret_code == SS_SUCCESS) { + if (mdev->state.conn < C_CONNECTED) + tl_clear(mdev); + if (mdev->state.disk == D_DISKLESS || mdev->state.disk == D_FAILED) + tl_restart(mdev, fail_frozen_disk_io); + } + drbd_resume_io(mdev); + return 0; } @@ -1941,40 +2140,6 @@ out: return 0; } -static struct drbd_conf *ensure_mdev(struct drbd_nl_cfg_req *nlp) -{ - struct drbd_conf *mdev; - - if (nlp->drbd_minor >= minor_count) - return NULL; - - mdev = minor_to_mdev(nlp->drbd_minor); - - if (!mdev && (nlp->flags & DRBD_NL_CREATE_DEVICE)) { - struct gendisk *disk = NULL; - mdev = drbd_new_device(nlp->drbd_minor); - - spin_lock_irq(&drbd_pp_lock); - if (minor_table[nlp->drbd_minor] == NULL) { - minor_table[nlp->drbd_minor] = mdev; - disk = mdev->vdisk; - mdev = NULL; - } /* else: we lost the race */ - spin_unlock_irq(&drbd_pp_lock); - - if (disk) /* we won the race above */ - /* in case we ever add a drbd_delete_device(), - * don't forget the del_gendisk! */ - add_disk(disk); - else /* we lost the race above */ - drbd_free_mdev(mdev); - - mdev = minor_to_mdev(nlp->drbd_minor); - } - - return mdev; -} - struct cn_handler_struct { int (*function)(struct drbd_conf *, struct drbd_nl_cfg_req *, @@ -2035,7 +2200,8 @@ static void drbd_connector_callback(struct cn_msg *req, struct netlink_skb_parms goto fail; } - mdev = ensure_mdev(nlp); + mdev = ensure_mdev(nlp->drbd_minor, + (nlp->flags & DRBD_NL_CREATE_DEVICE)); if (!mdev) { retcode = ERR_MINOR_INVALID; goto fail; diff --git a/drivers/block/drbd/drbd_proc.c b/drivers/block/drbd/drbd_proc.c index be3374b..ad325c5 100644 --- a/drivers/block/drbd/drbd_proc.c +++ b/drivers/block/drbd/drbd_proc.c @@ -57,6 +57,7 @@ static void drbd_syncer_progress(struct drbd_conf *mdev, struct seq_file *seq) unsigned long db, dt, dbdt, rt, rs_left; unsigned int res; int i, x, y; + int stalled = 0; drbd_get_syncer_progress(mdev, &rs_left, &res); @@ -90,18 +91,17 @@ static void drbd_syncer_progress(struct drbd_conf *mdev, struct seq_file *seq) * db: blocks written from mark until now * rt: remaining time */ - dt = (jiffies - mdev->rs_mark_time) / HZ; - - if (dt > 20) { - /* if we made no update to rs_mark_time for too long, - * we are stalled. show that. */ - seq_printf(seq, "stalled\n"); - return; - } + /* Rolling marks. last_mark+1 may just now be modified. last_mark+2 is + * at least (DRBD_SYNC_MARKS-2)*DRBD_SYNC_MARK_STEP old, and has at + * least DRBD_SYNC_MARK_STEP time before it will be modified. */ + i = (mdev->rs_last_mark + 2) % DRBD_SYNC_MARKS; + dt = (jiffies - mdev->rs_mark_time[i]) / HZ; + if (dt > (DRBD_SYNC_MARK_STEP * DRBD_SYNC_MARKS)) + stalled = 1; if (!dt) dt++; - db = mdev->rs_mark_left - rs_left; + db = mdev->rs_mark_left[i] - rs_left; rt = (dt * (rs_left / (db/100+1)))/100; /* seconds */ seq_printf(seq, "finish: %lu:%02lu:%02lu", @@ -118,7 +118,7 @@ static void drbd_syncer_progress(struct drbd_conf *mdev, struct seq_file *seq) /* mean speed since syncer started * we do account for PausedSync periods */ dt = (jiffies - mdev->rs_start - mdev->rs_paused) / HZ; - if (dt <= 0) + if (dt == 0) dt = 1; db = mdev->rs_total - rs_left; dbdt = Bit2KB(db/dt); @@ -128,7 +128,14 @@ static void drbd_syncer_progress(struct drbd_conf *mdev, struct seq_file *seq) else seq_printf(seq, " (%ld)", dbdt); - seq_printf(seq, " K/sec\n"); + if (mdev->state.conn == C_SYNC_TARGET) { + if (mdev->c_sync_rate > 1000) + seq_printf(seq, " want: %d,%03d", + mdev->c_sync_rate / 1000, mdev->c_sync_rate % 1000); + else + seq_printf(seq, " want: %d", mdev->c_sync_rate); + } + seq_printf(seq, " K/sec%s\n", stalled ? " (stalled)" : ""); } static void resync_dump_detail(struct seq_file *seq, struct lc_element *e) @@ -196,7 +203,7 @@ static int drbd_seq_show(struct seq_file *seq, void *v) seq_printf(seq, "%2d: cs:Unconfigured\n", i); } else { seq_printf(seq, - "%2d: cs:%s ro:%s/%s ds:%s/%s %c %c%c%c%c%c\n" + "%2d: cs:%s ro:%s/%s ds:%s/%s %c %c%c%c%c%c%c\n" " ns:%u nr:%u dw:%u dr:%u al:%u bm:%u " "lo:%d pe:%d ua:%d ap:%d ep:%d wo:%c", i, sn, @@ -206,11 +213,12 @@ static int drbd_seq_show(struct seq_file *seq, void *v) drbd_disk_str(mdev->state.pdsk), (mdev->net_conf == NULL ? ' ' : (mdev->net_conf->wire_protocol - DRBD_PROT_A+'A')), - mdev->state.susp ? 's' : 'r', + is_susp(mdev->state) ? 's' : 'r', mdev->state.aftr_isp ? 'a' : '-', mdev->state.peer_isp ? 'p' : '-', mdev->state.user_isp ? 'u' : '-', mdev->congestion_reason ?: '-', + test_bit(AL_SUSPENDED, &mdev->flags) ? 's' : '-', mdev->send_cnt/2, mdev->recv_cnt/2, mdev->writ_cnt/2, diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c index 484ecbb..760ae0d 100644 --- a/drivers/block/drbd/drbd_receiver.c +++ b/drivers/block/drbd/drbd_receiver.c @@ -241,7 +241,7 @@ static void drbd_kick_lo_and_reclaim_net(struct drbd_conf *mdev) spin_unlock_irq(&mdev->req_lock); list_for_each_entry_safe(e, t, &reclaimed, w.list) - drbd_free_ee(mdev, e); + drbd_free_net_ee(mdev, e); } /** @@ -298,9 +298,11 @@ static struct page *drbd_pp_alloc(struct drbd_conf *mdev, unsigned number, bool * Is also used from inside an other spin_lock_irq(&mdev->req_lock); * Either links the page chain back to the global pool, * or returns all pages to the system. */ -static void drbd_pp_free(struct drbd_conf *mdev, struct page *page) +static void drbd_pp_free(struct drbd_conf *mdev, struct page *page, int is_net) { + atomic_t *a = is_net ? &mdev->pp_in_use_by_net : &mdev->pp_in_use; int i; + if (drbd_pp_vacant > (DRBD_MAX_SEGMENT_SIZE/PAGE_SIZE)*minor_count) i = page_chain_free(page); else { @@ -311,10 +313,10 @@ static void drbd_pp_free(struct drbd_conf *mdev, struct page *page) drbd_pp_vacant += i; spin_unlock(&drbd_pp_lock); } - atomic_sub(i, &mdev->pp_in_use); - i = atomic_read(&mdev->pp_in_use); + i = atomic_sub_return(i, a); if (i < 0) - dev_warn(DEV, "ASSERTION FAILED: pp_in_use: %d < 0\n", i); + dev_warn(DEV, "ASSERTION FAILED: %s: %d < 0\n", + is_net ? "pp_in_use_by_net" : "pp_in_use", i); wake_up(&drbd_pp_wait); } @@ -365,7 +367,6 @@ struct drbd_epoch_entry *drbd_alloc_ee(struct drbd_conf *mdev, e->size = data_size; e->flags = 0; e->sector = sector; - e->sector = sector; e->block_id = id; return e; @@ -375,9 +376,11 @@ struct drbd_epoch_entry *drbd_alloc_ee(struct drbd_conf *mdev, return NULL; } -void drbd_free_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e) +void drbd_free_some_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e, int is_net) { - drbd_pp_free(mdev, e->pages); + if (e->flags & EE_HAS_DIGEST) + kfree(e->digest); + drbd_pp_free(mdev, e->pages, is_net); D_ASSERT(atomic_read(&e->pending_bios) == 0); D_ASSERT(hlist_unhashed(&e->colision)); mempool_free(e, drbd_ee_mempool); @@ -388,13 +391,14 @@ int drbd_release_ee(struct drbd_conf *mdev, struct list_head *list) LIST_HEAD(work_list); struct drbd_epoch_entry *e, *t; int count = 0; + int is_net = list == &mdev->net_ee; spin_lock_irq(&mdev->req_lock); list_splice_init(list, &work_list); spin_unlock_irq(&mdev->req_lock); list_for_each_entry_safe(e, t, &work_list, w.list) { - drbd_free_ee(mdev, e); + drbd_free_some_ee(mdev, e, is_net); count++; } return count; @@ -423,7 +427,7 @@ static int drbd_process_done_ee(struct drbd_conf *mdev) spin_unlock_irq(&mdev->req_lock); list_for_each_entry_safe(e, t, &reclaimed, w.list) - drbd_free_ee(mdev, e); + drbd_free_net_ee(mdev, e); /* possible callbacks here: * e_end_block, and e_end_resync_block, e_send_discard_ack. @@ -719,14 +723,14 @@ out: static int drbd_send_fp(struct drbd_conf *mdev, struct socket *sock, enum drbd_packets cmd) { - struct p_header *h = (struct p_header *) &mdev->data.sbuf.header; + struct p_header80 *h = &mdev->data.sbuf.header.h80; return _drbd_send_cmd(mdev, sock, cmd, h, sizeof(*h), 0); } static enum drbd_packets drbd_recv_fp(struct drbd_conf *mdev, struct socket *sock) { - struct p_header *h = (struct p_header *) &mdev->data.sbuf.header; + struct p_header80 *h = &mdev->data.rbuf.header.h80; int rr; rr = drbd_recv_short(mdev, sock, h, sizeof(*h), 0); @@ -776,9 +780,6 @@ static int drbd_connect(struct drbd_conf *mdev) D_ASSERT(!mdev->data.socket); - if (test_and_clear_bit(CREATE_BARRIER, &mdev->flags)) - dev_err(DEV, "CREATE_BARRIER flag was set in drbd_connect - now cleared!\n"); - if (drbd_request_state(mdev, NS(conn, C_WF_CONNECTION)) < SS_SUCCESS) return -2; @@ -927,6 +928,11 @@ retry: drbd_thread_start(&mdev->asender); + if (mdev->agreed_pro_version < 95 && get_ldev(mdev)) { + drbd_setup_queue_param(mdev, DRBD_MAX_SIZE_H80_PACKET); + put_ldev(mdev); + } + if (!drbd_send_protocol(mdev)) return -1; drbd_send_sync_param(mdev, &mdev->sync_conf); @@ -946,22 +952,28 @@ out_release_sockets: return -1; } -static int drbd_recv_header(struct drbd_conf *mdev, struct p_header *h) +static int drbd_recv_header(struct drbd_conf *mdev, enum drbd_packets *cmd, unsigned int *packet_size) { + union p_header *h = &mdev->data.rbuf.header; int r; 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; - }; - h->command = be16_to_cpu(h->command); - h->length = be16_to_cpu(h->length); - if (unlikely(h->magic != BE_DRBD_MAGIC)) { - dev_err(DEV, "magic?? on data m: 0x%lx c: %d l: %d\n", - (long)be32_to_cpu(h->magic), - h->command, h->length); + } + + if (likely(h->h80.magic == BE_DRBD_MAGIC)) { + *cmd = be16_to_cpu(h->h80.command); + *packet_size = be16_to_cpu(h->h80.length); + } else if (h->h95.magic == BE_DRBD_MAGIC_BIG) { + *cmd = be16_to_cpu(h->h95.command); + *packet_size = be32_to_cpu(h->h95.length); + } else { + dev_err(DEV, "magic?? on data m: 0x%08x c: %d l: %d\n", + be32_to_cpu(h->h80.magic), + be16_to_cpu(h->h80.command), + be16_to_cpu(h->h80.length)); return FALSE; } mdev->last_received = jiffies; @@ -1268,17 +1280,12 @@ int w_e_reissue(struct drbd_conf *mdev, struct drbd_work *w, int cancel) __relea return 1; } -static int receive_Barrier(struct drbd_conf *mdev, struct p_header *h) +static int receive_Barrier(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned int data_size) { int rv, issue_flush; - struct p_barrier *p = (struct p_barrier *)h; + struct p_barrier *p = &mdev->data.rbuf.barrier; struct drbd_epoch *epoch; - ERR_IF(h->length != (sizeof(*p)-sizeof(*h))) return FALSE; - - rv = drbd_recv(mdev, h->payload, h->length); - ERR_IF(rv != h->length) return FALSE; - inc_unacked(mdev); if (mdev->net_conf->wire_protocol != DRBD_PROT_C) @@ -1457,7 +1464,7 @@ static int drbd_drain_block(struct drbd_conf *mdev, int data_size) data_size -= rr; } kunmap(page); - drbd_pp_free(mdev, page); + drbd_pp_free(mdev, page, 0); return rv; } @@ -1562,30 +1569,29 @@ static int recv_resync_read(struct drbd_conf *mdev, sector_t sector, int data_si list_add(&e->w.list, &mdev->sync_ee); spin_unlock_irq(&mdev->req_lock); + atomic_add(data_size >> 9, &mdev->rs_sect_ev); if (drbd_submit_ee(mdev, e, WRITE, DRBD_FAULT_RS_WR) == 0) return TRUE; + /* drbd_submit_ee currently fails for one reason only: + * not being able to allocate enough bios. + * Is dropping the connection going to help? */ + spin_lock_irq(&mdev->req_lock); + list_del(&e->w.list); + spin_unlock_irq(&mdev->req_lock); + drbd_free_ee(mdev, e); fail: put_ldev(mdev); return FALSE; } -static int receive_DataReply(struct drbd_conf *mdev, struct p_header *h) +static int receive_DataReply(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned int data_size) { struct drbd_request *req; sector_t sector; - unsigned int header_size, data_size; int ok; - struct p_data *p = (struct p_data *)h; - - header_size = sizeof(*p) - sizeof(*h); - data_size = h->length - header_size; - - ERR_IF(data_size == 0) return FALSE; - - if (drbd_recv(mdev, h->payload, header_size) != header_size) - return FALSE; + struct p_data *p = &mdev->data.rbuf.data; sector = be64_to_cpu(p->sector); @@ -1611,20 +1617,11 @@ static int receive_DataReply(struct drbd_conf *mdev, struct p_header *h) return ok; } -static int receive_RSDataReply(struct drbd_conf *mdev, struct p_header *h) +static int receive_RSDataReply(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned int data_size) { sector_t sector; - unsigned int header_size, data_size; int ok; - struct p_data *p = (struct p_data *)h; - - header_size = sizeof(*p) - sizeof(*h); - data_size = h->length - header_size; - - ERR_IF(data_size == 0) return FALSE; - - if (drbd_recv(mdev, h->payload, header_size) != header_size) - return FALSE; + struct p_data *p = &mdev->data.rbuf.data; sector = be64_to_cpu(p->sector); D_ASSERT(p->block_id == ID_SYNCER); @@ -1640,9 +1637,11 @@ static int receive_RSDataReply(struct drbd_conf *mdev, struct p_header *h) ok = drbd_drain_block(mdev, data_size); - drbd_send_ack_dp(mdev, P_NEG_ACK, p); + drbd_send_ack_dp(mdev, P_NEG_ACK, p, data_size); } + atomic_add(data_size >> 9, &mdev->rs_sect_in); + return ok; } @@ -1765,24 +1764,27 @@ static int drbd_wait_peer_seq(struct drbd_conf *mdev, const u32 packet_seq) return ret; } +static unsigned long write_flags_to_bio(struct drbd_conf *mdev, u32 dpf) +{ + if (mdev->agreed_pro_version >= 95) + return (dpf & DP_RW_SYNC ? REQ_SYNC : 0) | + (dpf & DP_UNPLUG ? REQ_UNPLUG : 0) | + (dpf & DP_FUA ? REQ_FUA : 0) | + (dpf & DP_FLUSH ? REQ_FUA : 0) | + (dpf & DP_DISCARD ? REQ_DISCARD : 0); + else + return dpf & DP_RW_SYNC ? (REQ_SYNC | REQ_UNPLUG) : 0; +} + /* mirrored write */ -static int receive_Data(struct drbd_conf *mdev, struct p_header *h) +static int receive_Data(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned int data_size) { sector_t sector; struct drbd_epoch_entry *e; - struct p_data *p = (struct p_data *)h; - int header_size, data_size; + struct p_data *p = &mdev->data.rbuf.data; int rw = WRITE; u32 dp_flags; - header_size = sizeof(*p) - sizeof(*h); - data_size = h->length - header_size; - - ERR_IF(data_size == 0) return FALSE; - - if (drbd_recv(mdev, h->payload, header_size) != header_size) - return FALSE; - if (!get_ldev(mdev)) { if (__ratelimit(&drbd_ratelimit_state)) dev_err(DEV, "Can not write mirrored data block " @@ -1792,7 +1794,7 @@ static int receive_Data(struct drbd_conf *mdev, struct p_header *h) mdev->peer_seq++; spin_unlock(&mdev->peer_seq_lock); - drbd_send_ack_dp(mdev, P_NEG_ACK, p); + drbd_send_ack_dp(mdev, P_NEG_ACK, p, data_size); atomic_inc(&mdev->current_epoch->epoch_size); return drbd_drain_block(mdev, data_size); } @@ -1839,12 +1841,8 @@ static int receive_Data(struct drbd_conf *mdev, struct p_header *h) spin_unlock(&mdev->epoch_lock); dp_flags = be32_to_cpu(p->dp_flags); - if (dp_flags & DP_HARDBARRIER) { - dev_err(DEV, "ASSERT FAILED would have submitted barrier request\n"); - /* rw |= REQ_HARDBARRIER; */ - } - if (dp_flags & DP_RW_SYNC) - rw |= REQ_SYNC | REQ_UNPLUG; + rw |= write_flags_to_bio(mdev, dp_flags); + if (dp_flags & DP_MAY_SET_IN_SYNC) e->flags |= EE_MAY_SET_IN_SYNC; @@ -2007,6 +2005,16 @@ static int receive_Data(struct drbd_conf *mdev, struct p_header *h) if (drbd_submit_ee(mdev, e, rw, DRBD_FAULT_DT_WR) == 0) return TRUE; + /* drbd_submit_ee currently fails for one reason only: + * not being able to allocate enough bios. + * Is dropping the connection going to help? */ + spin_lock_irq(&mdev->req_lock); + list_del(&e->w.list); + hlist_del_init(&e->colision); + spin_unlock_irq(&mdev->req_lock); + if (e->flags & EE_CALL_AL_COMPLETE_IO) + drbd_al_complete_io(mdev, e->sector); + out_interrupted: /* yes, the epoch_size now is imbalanced. * but we drop the connection anyways, so we don't have a chance to @@ -2016,20 +2024,64 @@ out_interrupted: return FALSE; } -static int receive_DataRequest(struct drbd_conf *mdev, struct p_header *h) +/* We may throttle resync, if the lower device seems to be busy, + * and current sync rate is above c_min_rate. + * + * To decide whether or not the lower device is busy, we use a scheme similar + * to MD RAID is_mddev_idle(): if the partition stats reveal "significant" + * (more than 64 sectors) of activity we cannot account for with our own resync + * activity, it obviously is "busy". + * + * The current sync rate used here uses only the most recent two step marks, + * to have a short time average so we can react faster. + */ +int drbd_rs_should_slow_down(struct drbd_conf *mdev) +{ + struct gendisk *disk = mdev->ldev->backing_bdev->bd_contains->bd_disk; + unsigned long db, dt, dbdt; + int curr_events; + int throttle = 0; + + /* feature disabled? */ + if (mdev->sync_conf.c_min_rate == 0) + return 0; + + curr_events = (int)part_stat_read(&disk->part0, sectors[0]) + + (int)part_stat_read(&disk->part0, sectors[1]) - + atomic_read(&mdev->rs_sect_ev); + if (!mdev->rs_last_events || curr_events - mdev->rs_last_events > 64) { + unsigned long rs_left; + int i; + + mdev->rs_last_events = curr_events; + + /* sync speed average over the last 2*DRBD_SYNC_MARK_STEP, + * approx. */ + i = (mdev->rs_last_mark + DRBD_SYNC_MARKS-2) % DRBD_SYNC_MARKS; + rs_left = drbd_bm_total_weight(mdev) - mdev->rs_failed; + + dt = ((long)jiffies - (long)mdev->rs_mark_time[i]) / HZ; + if (!dt) + dt++; + db = mdev->rs_mark_left[i] - rs_left; + dbdt = Bit2KB(db/dt); + + if (dbdt > mdev->sync_conf.c_min_rate) + throttle = 1; + } + return throttle; +} + + +static int receive_DataRequest(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned int digest_size) { sector_t sector; const sector_t capacity = drbd_get_capacity(mdev->this_bdev); struct drbd_epoch_entry *e; struct digest_info *di = NULL; - int size, digest_size; + int size, verb; unsigned int fault_type; - struct p_block_req *p = - (struct p_block_req *)h; - const int brps = sizeof(*p)-sizeof(*h); - - if (drbd_recv(mdev, h->payload, brps) != brps) - return FALSE; + struct p_block_req *p = &mdev->data.rbuf.block_req; sector = be64_to_cpu(p->sector); size = be32_to_cpu(p->blksize); @@ -2046,12 +2098,31 @@ static int receive_DataRequest(struct drbd_conf *mdev, struct p_header *h) } if (!get_ldev_if_state(mdev, D_UP_TO_DATE)) { - if (__ratelimit(&drbd_ratelimit_state)) + verb = 1; + switch (cmd) { + case P_DATA_REQUEST: + drbd_send_ack_rp(mdev, P_NEG_DREPLY, p); + break; + case P_RS_DATA_REQUEST: + case P_CSUM_RS_REQUEST: + case P_OV_REQUEST: + drbd_send_ack_rp(mdev, P_NEG_RS_DREPLY , p); + break; + case P_OV_REPLY: + verb = 0; + dec_rs_pending(mdev); + drbd_send_ack_ex(mdev, P_OV_RESULT, sector, size, ID_IN_SYNC); + break; + default: + dev_err(DEV, "unexpected command (%s) in receive_DataRequest\n", + cmdname(cmd)); + } + if (verb && __ratelimit(&drbd_ratelimit_state)) dev_err(DEV, "Can not satisfy peer's read request, " "no local data.\n"); - drbd_send_ack_rp(mdev, h->command == P_DATA_REQUEST ? P_NEG_DREPLY : - P_NEG_RS_DREPLY , p); - return drbd_drain_block(mdev, h->length - brps); + + /* drain possibly payload */ + return drbd_drain_block(mdev, digest_size); } /* GFP_NOIO, because we must not cause arbitrary write-out: in a DRBD @@ -2063,31 +2134,21 @@ static int receive_DataRequest(struct drbd_conf *mdev, struct p_header *h) return FALSE; } - switch (h->command) { + switch (cmd) { case P_DATA_REQUEST: e->w.cb = w_e_end_data_req; fault_type = DRBD_FAULT_DT_RD; - break; + /* application IO, don't drbd_rs_begin_io */ + goto submit; + case P_RS_DATA_REQUEST: e->w.cb = w_e_end_rsdata_req; fault_type = DRBD_FAULT_RS_RD; - /* Eventually this should become asynchronously. Currently it - * blocks the whole receiver just to delay the reading of a - * resync data block. - * the drbd_work_queue mechanism is made for this... - */ - if (!drbd_rs_begin_io(mdev, sector)) { - /* we have been interrupted, - * probably connection lost! */ - D_ASSERT(signal_pending(current)); - goto out_free_e; - } break; case P_OV_REPLY: case P_CSUM_RS_REQUEST: fault_type = DRBD_FAULT_RS_RD; - digest_size = h->length - brps ; di = kmalloc(sizeof(*di) + digest_size, GFP_NOIO); if (!di) goto out_free_e; @@ -2095,31 +2156,25 @@ static int receive_DataRequest(struct drbd_conf *mdev, struct p_header *h) di->digest_size = digest_size; di->digest = (((char *)di)+sizeof(struct digest_info)); + e->digest = di; + e->flags |= EE_HAS_DIGEST; + if (drbd_recv(mdev, di->digest, digest_size) != digest_size) goto out_free_e; - e->block_id = (u64)(unsigned long)di; - if (h->command == P_CSUM_RS_REQUEST) { + if (cmd == P_CSUM_RS_REQUEST) { D_ASSERT(mdev->agreed_pro_version >= 89); e->w.cb = w_e_end_csum_rs_req; - } else if (h->command == P_OV_REPLY) { + } else if (cmd == P_OV_REPLY) { e->w.cb = w_e_end_ov_reply; dec_rs_pending(mdev); - break; - } - - if (!drbd_rs_begin_io(mdev, sector)) { - /* we have been interrupted, probably connection lost! */ - D_ASSERT(signal_pending(current)); - goto out_free_e; + /* drbd_rs_begin_io done when we sent this request, + * but accounting still needs to be done. */ + goto submit_for_resync; } break; case P_OV_REQUEST: - if (mdev->state.conn >= C_CONNECTED && - mdev->state.conn != C_VERIFY_T) - dev_warn(DEV, "ASSERT FAILED: got P_OV_REQUEST while being %s\n", - drbd_conn_str(mdev->state.conn)); if (mdev->ov_start_sector == ~(sector_t)0 && mdev->agreed_pro_version >= 90) { mdev->ov_start_sector = sector; @@ -2130,37 +2185,63 @@ static int receive_DataRequest(struct drbd_conf *mdev, struct p_header *h) } e->w.cb = w_e_end_ov_req; fault_type = DRBD_FAULT_RS_RD; - /* Eventually this should become asynchronous. Currently it - * blocks the whole receiver just to delay the reading of a - * resync data block. - * the drbd_work_queue mechanism is made for this... - */ - if (!drbd_rs_begin_io(mdev, sector)) { - /* we have been interrupted, - * probably connection lost! */ - D_ASSERT(signal_pending(current)); - goto out_free_e; - } break; - default: dev_err(DEV, "unexpected command (%s) in receive_DataRequest\n", - cmdname(h->command)); + cmdname(cmd)); fault_type = DRBD_FAULT_MAX; + goto out_free_e; } - spin_lock_irq(&mdev->req_lock); - list_add(&e->w.list, &mdev->read_ee); - spin_unlock_irq(&mdev->req_lock); + /* Throttle, drbd_rs_begin_io and submit should become asynchronous + * wrt the receiver, but it is not as straightforward as it may seem. + * Various places in the resync start and stop logic assume resync + * requests are processed in order, requeuing this on the worker thread + * introduces a bunch of new code for synchronization between threads. + * + * Unlimited throttling before drbd_rs_begin_io may stall the resync + * "forever", throttling after drbd_rs_begin_io will lock that extent + * for application writes for the same time. For now, just throttle + * here, where the rest of the code expects the receiver to sleep for + * a while, anyways. + */ + + /* Throttle before drbd_rs_begin_io, as that locks out application IO; + * this defers syncer requests for some time, before letting at least + * on request through. The resync controller on the receiving side + * will adapt to the incoming rate accordingly. + * + * We cannot throttle here if remote is Primary/SyncTarget: + * we would also throttle its application reads. + * In that case, throttling is done on the SyncTarget only. + */ + if (mdev->state.peer != R_PRIMARY && drbd_rs_should_slow_down(mdev)) + msleep(100); + if (drbd_rs_begin_io(mdev, e->sector)) + goto out_free_e; +submit_for_resync: + atomic_add(size >> 9, &mdev->rs_sect_ev); + +submit: inc_unacked(mdev); + spin_lock_irq(&mdev->req_lock); + list_add_tail(&e->w.list, &mdev->read_ee); + spin_unlock_irq(&mdev->req_lock); if (drbd_submit_ee(mdev, e, READ, fault_type) == 0) return TRUE; + /* drbd_submit_ee currently fails for one reason only: + * not being able to allocate enough bios. + * Is dropping the connection going to help? */ + spin_lock_irq(&mdev->req_lock); + list_del(&e->w.list); + spin_unlock_irq(&mdev->req_lock); + /* no drbd_rs_complete_io(), we are dropping the connection anyways */ + out_free_e: - kfree(di); put_ldev(mdev); drbd_free_ee(mdev, e); return FALSE; @@ -2699,20 +2780,13 @@ static int cmp_after_sb(enum drbd_after_sb_p peer, enum drbd_after_sb_p self) return 1; } -static int receive_protocol(struct drbd_conf *mdev, struct p_header *h) +static int receive_protocol(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned int data_size) { - struct p_protocol *p = (struct p_protocol *)h; - int header_size, data_size; + struct p_protocol *p = &mdev->data.rbuf.protocol; int p_proto, p_after_sb_0p, p_after_sb_1p, p_after_sb_2p; int p_want_lose, p_two_primaries, cf; char p_integrity_alg[SHARED_SECRET_MAX] = ""; - header_size = sizeof(*p) - sizeof(*h); - data_size = h->length - header_size; - - if (drbd_recv(mdev, h->payload, header_size) != header_size) - return FALSE; - p_proto = be32_to_cpu(p->protocol); p_after_sb_0p = be32_to_cpu(p->after_sb_0p); p_after_sb_1p = be32_to_cpu(p->after_sb_1p); @@ -2805,39 +2879,46 @@ struct crypto_hash *drbd_crypto_alloc_digest_safe(const struct drbd_conf *mdev, return tfm; } -static int receive_SyncParam(struct drbd_conf *mdev, struct p_header *h) +static int receive_SyncParam(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned int packet_size) { int ok = TRUE; - struct p_rs_param_89 *p = (struct p_rs_param_89 *)h; + 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; struct crypto_hash *csums_tfm = NULL; const int apv = mdev->agreed_pro_version; + int *rs_plan_s = NULL; + int fifo_size = 0; exp_max_sz = apv <= 87 ? sizeof(struct p_rs_param) : apv == 88 ? sizeof(struct p_rs_param) + SHARED_SECRET_MAX - : /* 89 */ sizeof(struct p_rs_param_89); + : apv <= 94 ? sizeof(struct p_rs_param_89) + : /* apv >= 95 */ sizeof(struct p_rs_param_95); - if (h->length > exp_max_sz) { + if (packet_size > exp_max_sz) { dev_err(DEV, "SyncParam packet too long: received %u, expected <= %u bytes\n", - h->length, exp_max_sz); + packet_size, exp_max_sz); return FALSE; } if (apv <= 88) { - header_size = sizeof(struct p_rs_param) - sizeof(*h); - data_size = h->length - header_size; - } else /* apv >= 89 */ { - header_size = sizeof(struct p_rs_param_89) - sizeof(*h); - data_size = h->length - header_size; + header_size = sizeof(struct p_rs_param) - sizeof(struct p_header80); + data_size = packet_size - header_size; + } else if (apv <= 94) { + header_size = sizeof(struct p_rs_param_89) - sizeof(struct p_header80); + data_size = packet_size - header_size; + D_ASSERT(data_size == 0); + } else { + header_size = sizeof(struct p_rs_param_95) - sizeof(struct p_header80); + data_size = packet_size - header_size; D_ASSERT(data_size == 0); } /* initialize verify_alg and csums_alg */ memset(p->verify_alg, 0, 2 * SHARED_SECRET_MAX); - if (drbd_recv(mdev, h->payload, header_size) != header_size) + if (drbd_recv(mdev, &p->head.payload, header_size) != header_size) return FALSE; mdev->sync_conf.rate = be32_to_cpu(p->rate); @@ -2896,6 +2977,22 @@ static int receive_SyncParam(struct drbd_conf *mdev, struct p_header *h) } } + if (apv > 94) { + mdev->sync_conf.rate = be32_to_cpu(p->rate); + mdev->sync_conf.c_plan_ahead = be32_to_cpu(p->c_plan_ahead); + mdev->sync_conf.c_delay_target = be32_to_cpu(p->c_delay_target); + mdev->sync_conf.c_fill_target = be32_to_cpu(p->c_fill_target); + mdev->sync_conf.c_max_rate = be32_to_cpu(p->c_max_rate); + + fifo_size = (mdev->sync_conf.c_plan_ahead * 10 * SLEEP_TIME) / HZ; + if (fifo_size != mdev->rs_plan_s.size && fifo_size > 0) { + rs_plan_s = kzalloc(sizeof(int) * fifo_size, GFP_KERNEL); + if (!rs_plan_s) { + dev_err(DEV, "kmalloc of fifo_buffer failed"); + goto disconnect; + } + } + } spin_lock(&mdev->peer_seq_lock); /* lock against drbd_nl_syncer_conf() */ @@ -2913,6 +3010,12 @@ static int receive_SyncParam(struct drbd_conf *mdev, struct p_header *h) mdev->csums_tfm = csums_tfm; dev_info(DEV, "using csums-alg: \"%s\"\n", p->csums_alg); } + if (fifo_size != mdev->rs_plan_s.size) { + kfree(mdev->rs_plan_s.values); + mdev->rs_plan_s.values = rs_plan_s; + mdev->rs_plan_s.size = fifo_size; + mdev->rs_planed = 0; + } spin_unlock(&mdev->peer_seq_lock); } @@ -2946,19 +3049,15 @@ static void warn_if_differ_considerably(struct drbd_conf *mdev, (unsigned long long)a, (unsigned long long)b); } -static int receive_sizes(struct drbd_conf *mdev, struct p_header *h) +static int receive_sizes(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned int data_size) { - struct p_sizes *p = (struct p_sizes *)h; + struct p_sizes *p = &mdev->data.rbuf.sizes; enum determine_dev_size dd = unchanged; unsigned int max_seg_s; sector_t p_size, p_usize, my_usize; int ldsc = 0; /* local disk size changed */ enum dds_flags ddsf; - ERR_IF(h->length != (sizeof(*p)-sizeof(*h))) return FALSE; - if (drbd_recv(mdev, h->payload, h->length) != h->length) - return FALSE; - p_size = be64_to_cpu(p->d_size); p_usize = be64_to_cpu(p->u_size); @@ -3028,6 +3127,8 @@ static int receive_sizes(struct drbd_conf *mdev, struct p_header *h) if (mdev->agreed_pro_version < 94) max_seg_s = be32_to_cpu(p->max_segment_size); + else if (mdev->agreed_pro_version == 94) + max_seg_s = DRBD_MAX_SIZE_H80_PACKET; else /* drbd 8.3.8 onwards */ max_seg_s = DRBD_MAX_SEGMENT_SIZE; @@ -3061,16 +3162,12 @@ static int receive_sizes(struct drbd_conf *mdev, struct p_header *h) return TRUE; } -static int receive_uuids(struct drbd_conf *mdev, struct p_header *h) +static int receive_uuids(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned int data_size) { - struct p_uuids *p = (struct p_uuids *)h; + struct p_uuids *p = &mdev->data.rbuf.uuids; u64 *p_uuid; int i; - ERR_IF(h->length != (sizeof(*p)-sizeof(*h))) return FALSE; - if (drbd_recv(mdev, h->payload, h->length) != h->length) - return FALSE; - p_uuid = kmalloc(sizeof(u64)*UI_EXTENDED_SIZE, GFP_NOIO); for (i = UI_CURRENT; i < UI_EXTENDED_SIZE; i++) @@ -3106,6 +3203,11 @@ static int receive_uuids(struct drbd_conf *mdev, struct p_header *h) drbd_md_sync(mdev); } put_ldev(mdev); + } else if (mdev->state.disk < D_INCONSISTENT && + mdev->state.role == R_PRIMARY) { + /* I am a diskless primary, the peer just created a new current UUID + for me. */ + drbd_set_ed_uuid(mdev, p_uuid[UI_CURRENT]); } /* Before we test for the disk state, we should wait until an eventually @@ -3149,16 +3251,12 @@ static union drbd_state convert_state(union drbd_state ps) return ms; } -static int receive_req_state(struct drbd_conf *mdev, struct p_header *h) +static int receive_req_state(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned int data_size) { - struct p_req_state *p = (struct p_req_state *)h; + struct p_req_state *p = &mdev->data.rbuf.req_state; union drbd_state mask, val; int rv; - ERR_IF(h->length != (sizeof(*p)-sizeof(*h))) return FALSE; - if (drbd_recv(mdev, h->payload, h->length) != h->length) - return FALSE; - mask.i = be32_to_cpu(p->mask); val.i = be32_to_cpu(p->val); @@ -3179,20 +3277,14 @@ static int receive_req_state(struct drbd_conf *mdev, struct p_header *h) return TRUE; } -static int receive_state(struct drbd_conf *mdev, struct p_header *h) +static int receive_state(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned int data_size) { - struct p_state *p = (struct p_state *)h; - enum drbd_conns nconn, oconn; - union drbd_state ns, peer_state; + struct p_state *p = &mdev->data.rbuf.state; + union drbd_state os, ns, peer_state; enum drbd_disk_state real_peer_disk; + enum chg_state_flags cs_flags; int rv; - ERR_IF(h->length != (sizeof(*p)-sizeof(*h))) - return FALSE; - - if (drbd_recv(mdev, h->payload, h->length) != h->length) - return FALSE; - peer_state.i = be32_to_cpu(p->state); real_peer_disk = peer_state.disk; @@ -3203,38 +3295,72 @@ static int receive_state(struct drbd_conf *mdev, struct p_header *h) spin_lock_irq(&mdev->req_lock); retry: - oconn = nconn = mdev->state.conn; + os = ns = mdev->state; spin_unlock_irq(&mdev->req_lock); - if (nconn == C_WF_REPORT_PARAMS) - nconn = C_CONNECTED; + /* peer says his disk is uptodate, while we think it is inconsistent, + * and this happens while we think we have a sync going on. */ + if (os.pdsk == D_INCONSISTENT && real_peer_disk == D_UP_TO_DATE && + os.conn > C_CONNECTED && os.disk == D_UP_TO_DATE) { + /* If we are (becoming) SyncSource, but peer is still in sync + * preparation, ignore its uptodate-ness to avoid flapping, it + * will change to inconsistent once the peer reaches active + * syncing states. + * It may have changed syncer-paused flags, however, so we + * cannot ignore this completely. */ + if (peer_state.conn > C_CONNECTED && + peer_state.conn < C_SYNC_SOURCE) + real_peer_disk = D_INCONSISTENT; + + /* if peer_state changes to connected at the same time, + * it explicitly notifies us that it finished resync. + * Maybe we should finish it up, too? */ + else if (os.conn >= C_SYNC_SOURCE && + peer_state.conn == C_CONNECTED) { + if (drbd_bm_total_weight(mdev) <= mdev->rs_failed) + drbd_resync_finished(mdev); + return TRUE; + } + } + + /* peer says his disk is inconsistent, while we think it is uptodate, + * and this happens while the peer still thinks we have a sync going on, + * but we think we are already done with the sync. + * We ignore this to avoid flapping pdsk. + * This should not happen, if the peer is a recent version of drbd. */ + if (os.pdsk == D_UP_TO_DATE && real_peer_disk == D_INCONSISTENT && + os.conn == C_CONNECTED && peer_state.conn > C_SYNC_SOURCE) + real_peer_disk = D_UP_TO_DATE; + + if (ns.conn == C_WF_REPORT_PARAMS) + ns.conn = C_CONNECTED; if (mdev->p_uuid && peer_state.disk >= D_NEGOTIATING && get_ldev_if_state(mdev, D_NEGOTIATING)) { int cr; /* consider resync */ /* if we established a new connection */ - cr = (oconn < C_CONNECTED); + cr = (os.conn < C_CONNECTED); /* if we had an established connection * and one of the nodes newly attaches a disk */ - cr |= (oconn == C_CONNECTED && + cr |= (os.conn == C_CONNECTED && (peer_state.disk == D_NEGOTIATING || - mdev->state.disk == D_NEGOTIATING)); + os.disk == D_NEGOTIATING)); /* if we have both been inconsistent, and the peer has been * forced to be UpToDate with --overwrite-data */ cr |= test_bit(CONSIDER_RESYNC, &mdev->flags); /* if we had been plain connected, and the admin requested to * start a sync by "invalidate" or "invalidate-remote" */ - cr |= (oconn == C_CONNECTED && + cr |= (os.conn == C_CONNECTED && (peer_state.conn >= C_STARTING_SYNC_S && peer_state.conn <= C_WF_BITMAP_T)); if (cr) - nconn = drbd_sync_handshake(mdev, peer_state.role, real_peer_disk); + ns.conn = drbd_sync_handshake(mdev, peer_state.role, real_peer_disk); put_ldev(mdev); - if (nconn == C_MASK) { - nconn = C_CONNECTED; + if (ns.conn == C_MASK) { + ns.conn = C_CONNECTED; if (mdev->state.disk == D_NEGOTIATING) { drbd_force_state(mdev, NS(disk, D_DISKLESS)); } else if (peer_state.disk == D_NEGOTIATING) { @@ -3244,7 +3370,7 @@ static int receive_state(struct drbd_conf *mdev, struct p_header *h) } else { if (test_and_clear_bit(CONN_DRY_RUN, &mdev->flags)) return FALSE; - D_ASSERT(oconn == C_WF_REPORT_PARAMS); + D_ASSERT(os.conn == C_WF_REPORT_PARAMS); drbd_force_state(mdev, NS(conn, C_DISCONNECTING)); return FALSE; } @@ -3252,18 +3378,28 @@ static int receive_state(struct drbd_conf *mdev, struct p_header *h) } spin_lock_irq(&mdev->req_lock); - if (mdev->state.conn != oconn) + if (mdev->state.i != os.i) goto retry; clear_bit(CONSIDER_RESYNC, &mdev->flags); - ns.i = mdev->state.i; - ns.conn = nconn; ns.peer = peer_state.role; ns.pdsk = real_peer_disk; ns.peer_isp = (peer_state.aftr_isp | peer_state.user_isp); - if ((nconn == C_CONNECTED || nconn == C_WF_BITMAP_S) && ns.disk == D_NEGOTIATING) + if ((ns.conn == C_CONNECTED || ns.conn == C_WF_BITMAP_S) && ns.disk == D_NEGOTIATING) ns.disk = mdev->new_state_tmp.disk; - - rv = _drbd_set_state(mdev, ns, CS_VERBOSE | CS_HARD, NULL); + cs_flags = CS_VERBOSE + (os.conn < C_CONNECTED && ns.conn >= C_CONNECTED ? 0 : CS_HARD); + if (ns.pdsk == D_CONSISTENT && is_susp(ns) && ns.conn == C_CONNECTED && os.conn < C_CONNECTED && + test_bit(NEW_CUR_UUID, &mdev->flags)) { + /* Do not allow tl_restart(resend) for a rebooted peer. We can only allow this + for temporal network outages! */ + spin_unlock_irq(&mdev->req_lock); + dev_err(DEV, "Aborting Connect, can not thaw IO with an only Consistent peer\n"); + tl_clear(mdev); + 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; + } + rv = _drbd_set_state(mdev, ns, cs_flags, NULL); ns = mdev->state; spin_unlock_irq(&mdev->req_lock); @@ -3272,8 +3408,8 @@ static int receive_state(struct drbd_conf *mdev, struct p_header *h) return FALSE; } - if (oconn > C_WF_REPORT_PARAMS) { - if (nconn > C_CONNECTED && peer_state.conn <= C_CONNECTED && + if (os.conn > C_WF_REPORT_PARAMS) { + if (ns.conn > C_CONNECTED && peer_state.conn <= C_CONNECTED && peer_state.disk != D_NEGOTIATING ) { /* we want resync, peer has not yet decided to sync... */ /* Nowadays only used when forcing a node into primary role and @@ -3290,9 +3426,9 @@ static int receive_state(struct drbd_conf *mdev, struct p_header *h) return TRUE; } -static int receive_sync_uuid(struct drbd_conf *mdev, struct p_header *h) +static int receive_sync_uuid(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned int data_size) { - struct p_rs_uuid *p = (struct p_rs_uuid *)h; + struct p_rs_uuid *p = &mdev->data.rbuf.rs_uuid; wait_event(mdev->misc_wait, mdev->state.conn == C_WF_SYNC_UUID || @@ -3301,10 +3437,6 @@ static int receive_sync_uuid(struct drbd_conf *mdev, struct p_header *h) /* D_ASSERT( mdev->state.conn == C_WF_SYNC_UUID ); */ - ERR_IF(h->length != (sizeof(*p)-sizeof(*h))) return FALSE; - if (drbd_recv(mdev, h->payload, h->length) != h->length) - return FALSE; - /* Here the _drbd_uuid_ functions are right, current should _not_ be rotated into the history */ if (get_ldev_if_state(mdev, D_NEGOTIATING)) { @@ -3323,14 +3455,14 @@ static int receive_sync_uuid(struct drbd_conf *mdev, struct p_header *h) enum receive_bitmap_ret { OK, DONE, FAILED }; static enum receive_bitmap_ret -receive_bitmap_plain(struct drbd_conf *mdev, struct p_header *h, - unsigned long *buffer, struct bm_xfer_ctx *c) +receive_bitmap_plain(struct drbd_conf *mdev, unsigned int data_size, + unsigned long *buffer, struct bm_xfer_ctx *c) { unsigned num_words = min_t(size_t, BM_PACKET_WORDS, c->bm_words - c->word_offset); unsigned want = num_words * sizeof(long); - if (want != h->length) { - dev_err(DEV, "%s:want (%u) != h->length (%u)\n", __func__, want, h->length); + if (want != data_size) { + dev_err(DEV, "%s:want (%u) != data_size (%u)\n", __func__, want, data_size); return FAILED; } if (want == 0) @@ -3359,7 +3491,7 @@ recv_bm_rle_bits(struct drbd_conf *mdev, u64 tmp; unsigned long s = c->bit_offset; unsigned long e; - int len = p->head.length - (sizeof(*p) - sizeof(p->head)); + int len = be16_to_cpu(p->head.length) - (sizeof(*p) - sizeof(p->head)); int toggle = DCBP_get_start(p); int have; int bits; @@ -3428,7 +3560,7 @@ void INFO_bm_xfer_stats(struct drbd_conf *mdev, const char *direction, struct bm_xfer_ctx *c) { /* what would it take to transfer it "plaintext" */ - unsigned plain = sizeof(struct p_header) * + unsigned plain = sizeof(struct p_header80) * ((c->bm_words+BM_PACKET_WORDS-1)/BM_PACKET_WORDS+1) + c->bm_words * sizeof(long); unsigned total = c->bytes[0] + c->bytes[1]; @@ -3466,12 +3598,13 @@ void INFO_bm_xfer_stats(struct drbd_conf *mdev, in order to be agnostic to the 32 vs 64 bits issue. returns 0 on failure, 1 if we successfully received it. */ -static int receive_bitmap(struct drbd_conf *mdev, struct p_header *h) +static int receive_bitmap(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned int data_size) { struct bm_xfer_ctx c; void *buffer; enum receive_bitmap_ret ret; int ok = FALSE; + struct p_header80 *h = &mdev->data.rbuf.header.h80; wait_event(mdev->misc_wait, !atomic_read(&mdev->ap_bio_cnt)); @@ -3491,39 +3624,39 @@ static int receive_bitmap(struct drbd_conf *mdev, struct p_header *h) }; do { - if (h->command == P_BITMAP) { - ret = receive_bitmap_plain(mdev, h, buffer, &c); - } else if (h->command == P_COMPRESSED_BITMAP) { + if (cmd == P_BITMAP) { + ret = receive_bitmap_plain(mdev, data_size, buffer, &c); + } else if (cmd == P_COMPRESSED_BITMAP) { /* MAYBE: sanity check that we speak proto >= 90, * and the feature is enabled! */ struct p_compressed_bm *p; - if (h->length > BM_PACKET_PAYLOAD_BYTES) { + if (data_size > BM_PACKET_PAYLOAD_BYTES) { dev_err(DEV, "ReportCBitmap packet too large\n"); goto out; } /* use the page buff */ p = buffer; memcpy(p, h, sizeof(*h)); - if (drbd_recv(mdev, p->head.payload, h->length) != h->length) + if (drbd_recv(mdev, p->head.payload, data_size) != data_size) goto out; - if (p->head.length <= (sizeof(*p) - sizeof(p->head))) { - dev_err(DEV, "ReportCBitmap packet too small (l:%u)\n", p->head.length); + if (data_size <= (sizeof(*p) - sizeof(p->head))) { + dev_err(DEV, "ReportCBitmap packet too small (l:%u)\n", data_size); return FAILED; } ret = decode_bitmap_c(mdev, p, &c); } else { - dev_warn(DEV, "receive_bitmap: h->command neither ReportBitMap nor ReportCBitMap (is 0x%x)", h->command); + dev_warn(DEV, "receive_bitmap: cmd neither ReportBitMap nor ReportCBitMap (is 0x%x)", cmd); goto out; } - c.packets[h->command == P_BITMAP]++; - c.bytes[h->command == P_BITMAP] += sizeof(struct p_header) + h->length; + c.packets[cmd == P_BITMAP]++; + c.bytes[cmd == P_BITMAP] += sizeof(struct p_header80) + data_size; if (ret != OK) break; - if (!drbd_recv_header(mdev, h)) + if (!drbd_recv_header(mdev, &cmd, &data_size)) goto out; } while (ret == OK); if (ret == FAILED) @@ -3554,17 +3687,16 @@ static int receive_bitmap(struct drbd_conf *mdev, struct p_header *h) return ok; } -static int receive_skip_(struct drbd_conf *mdev, struct p_header *h, int silent) +static int receive_skip(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned int data_size) { /* TODO zero copy sink :) */ static char sink[128]; int size, want, r; - if (!silent) - dev_warn(DEV, "skipping unknown optional packet type %d, l: %d!\n", - h->command, h->length); + dev_warn(DEV, "skipping unknown optional packet type %d, l: %d!\n", + cmd, data_size); - size = h->length; + size = data_size; while (size > 0) { want = min_t(int, size, sizeof(sink)); r = drbd_recv(mdev, sink, want); @@ -3574,17 +3706,7 @@ static int receive_skip_(struct drbd_conf *mdev, struct p_header *h, int silent) return size == 0; } -static int receive_skip(struct drbd_conf *mdev, struct p_header *h) -{ - return receive_skip_(mdev, h, 0); -} - -static int receive_skip_silent(struct drbd_conf *mdev, struct p_header *h) -{ - return receive_skip_(mdev, h, 1); -} - -static int receive_UnplugRemote(struct drbd_conf *mdev, struct p_header *h) +static int receive_UnplugRemote(struct drbd_conf *mdev, enum drbd_packets cmd, unsigned int data_size) { if (mdev->state.disk >= D_INCONSISTENT) drbd_kick_lo(mdev); @@ -3596,108 +3718,94 @@ static int receive_UnplugRemote(struct drbd_conf *mdev, struct p_header *h) return TRUE; } -typedef int (*drbd_cmd_handler_f)(struct drbd_conf *, struct p_header *); - -static drbd_cmd_handler_f drbd_default_handler[] = { - [P_DATA] = receive_Data, - [P_DATA_REPLY] = receive_DataReply, - [P_RS_DATA_REPLY] = receive_RSDataReply, - [P_BARRIER] = receive_Barrier, - [P_BITMAP] = receive_bitmap, - [P_COMPRESSED_BITMAP] = receive_bitmap, - [P_UNPLUG_REMOTE] = receive_UnplugRemote, - [P_DATA_REQUEST] = receive_DataRequest, - [P_RS_DATA_REQUEST] = receive_DataRequest, - [P_SYNC_PARAM] = receive_SyncParam, - [P_SYNC_PARAM89] = receive_SyncParam, - [P_PROTOCOL] = receive_protocol, - [P_UUIDS] = receive_uuids, - [P_SIZES] = receive_sizes, - [P_STATE] = receive_state, - [P_STATE_CHG_REQ] = receive_req_state, - [P_SYNC_UUID] = receive_sync_uuid, - [P_OV_REQUEST] = receive_DataRequest, - [P_OV_REPLY] = receive_DataRequest, - [P_CSUM_RS_REQUEST] = receive_DataRequest, - [P_DELAY_PROBE] = receive_skip_silent, +typedef int (*drbd_cmd_handler_f)(struct drbd_conf *, enum drbd_packets cmd, unsigned int to_receive); + +struct data_cmd { + int expect_payload; + size_t pkt_size; + drbd_cmd_handler_f function; +}; + +static struct data_cmd drbd_cmd_handler[] = { + [P_DATA] = { 1, sizeof(struct p_data), receive_Data }, + [P_DATA_REPLY] = { 1, sizeof(struct p_data), receive_DataReply }, + [P_RS_DATA_REPLY] = { 1, sizeof(struct p_data), receive_RSDataReply } , + [P_BARRIER] = { 0, sizeof(struct p_barrier), receive_Barrier } , + [P_BITMAP] = { 1, sizeof(struct p_header80), receive_bitmap } , + [P_COMPRESSED_BITMAP] = { 1, sizeof(struct p_header80), receive_bitmap } , + [P_UNPLUG_REMOTE] = { 0, sizeof(struct p_header80), receive_UnplugRemote }, + [P_DATA_REQUEST] = { 0, sizeof(struct p_block_req), receive_DataRequest }, + [P_RS_DATA_REQUEST] = { 0, sizeof(struct p_block_req), receive_DataRequest }, + [P_SYNC_PARAM] = { 1, sizeof(struct p_header80), receive_SyncParam }, + [P_SYNC_PARAM89] = { 1, sizeof(struct p_header80), receive_SyncParam }, + [P_PROTOCOL] = { 1, sizeof(struct p_protocol), receive_protocol }, + [P_UUIDS] = { 0, sizeof(struct p_uuids), receive_uuids }, + [P_SIZES] = { 0, sizeof(struct p_sizes), receive_sizes }, + [P_STATE] = { 0, sizeof(struct p_state), receive_state }, + [P_STATE_CHG_REQ] = { 0, sizeof(struct p_req_state), receive_req_state }, + [P_SYNC_UUID] = { 0, sizeof(struct p_rs_uuid), receive_sync_uuid }, + [P_OV_REQUEST] = { 0, sizeof(struct p_block_req), receive_DataRequest }, + [P_OV_REPLY] = { 1, sizeof(struct p_block_req), receive_DataRequest }, + [P_CSUM_RS_REQUEST] = { 1, sizeof(struct p_block_req), receive_DataRequest }, + [P_DELAY_PROBE] = { 0, sizeof(struct p_delay_probe93), receive_skip }, /* anything missing from this table is in * the asender_tbl, see get_asender_cmd */ - [P_MAX_CMD] = NULL, + [P_MAX_CMD] = { 0, 0, NULL }, }; -static drbd_cmd_handler_f *drbd_cmd_handler = drbd_default_handler; -static drbd_cmd_handler_f *drbd_opt_cmd_handler; +/* All handler functions that expect a sub-header get that sub-heder in + mdev->data.rbuf.header.head.payload. + + Usually in mdev->data.rbuf.header.head the callback can find the usual + p_header, but they may not rely on that. Since there is also p_header95 ! + */ static void drbdd(struct drbd_conf *mdev) { - drbd_cmd_handler_f handler; - struct p_header *header = &mdev->data.rbuf.header; + union p_header *header = &mdev->data.rbuf.header; + unsigned int packet_size; + enum drbd_packets cmd; + size_t shs; /* sub header size */ + int rv; while (get_t_state(&mdev->receiver) == Running) { drbd_thread_current_set_cpu(mdev); - if (!drbd_recv_header(mdev, header)) { - drbd_force_state(mdev, NS(conn, C_PROTOCOL_ERROR)); - break; - } + if (!drbd_recv_header(mdev, &cmd, &packet_size)) + goto err_out; - if (header->command < P_MAX_CMD) - handler = drbd_cmd_handler[header->command]; - else if (P_MAY_IGNORE < header->command - && header->command < P_MAX_OPT_CMD) - handler = drbd_opt_cmd_handler[header->command-P_MAY_IGNORE]; - else if (header->command > P_MAX_OPT_CMD) - handler = receive_skip; - else - handler = NULL; + if (unlikely(cmd >= P_MAX_CMD || !drbd_cmd_handler[cmd].function)) { + dev_err(DEV, "unknown packet type %d, l: %d!\n", cmd, packet_size); + goto err_out; + } - if (unlikely(!handler)) { - dev_err(DEV, "unknown packet type %d, l: %d!\n", - header->command, header->length); - drbd_force_state(mdev, NS(conn, C_PROTOCOL_ERROR)); - break; + shs = drbd_cmd_handler[cmd].pkt_size - sizeof(union p_header); + rv = drbd_recv(mdev, &header->h80.payload, shs); + if (unlikely(rv != shs)) { + dev_err(DEV, "short read while reading sub header: rv=%d\n", rv); + goto err_out; } - if (unlikely(!handler(mdev, header))) { - dev_err(DEV, "error receiving %s, l: %d!\n", - cmdname(header->command), header->length); - drbd_force_state(mdev, NS(conn, C_PROTOCOL_ERROR)); - break; + + if (packet_size - shs > 0 && !drbd_cmd_handler[cmd].expect_payload) { + dev_err(DEV, "No payload expected %s l:%d\n", cmdname(cmd), packet_size); + goto err_out; } - } -} -static void drbd_fail_pending_reads(struct drbd_conf *mdev) -{ - struct hlist_head *slot; - struct hlist_node *pos; - struct hlist_node *tmp; - struct drbd_request *req; - int i; + rv = drbd_cmd_handler[cmd].function(mdev, cmd, packet_size - shs); - /* - * Application READ requests - */ - spin_lock_irq(&mdev->req_lock); - for (i = 0; i < APP_R_HSIZE; i++) { - slot = mdev->app_reads_hash+i; - hlist_for_each_entry_safe(req, pos, tmp, slot, colision) { - /* it may (but should not any longer!) - * be on the work queue; if that assert triggers, - * we need to also grab the - * spin_lock_irq(&mdev->data.work.q_lock); - * and list_del_init here. */ - D_ASSERT(list_empty(&req->w.list)); - /* It would be nice to complete outside of spinlock. - * But this is easier for now. */ - _req_mod(req, connection_lost_while_pending); + if (unlikely(!rv)) { + dev_err(DEV, "error receiving %s, l: %d!\n", + cmdname(cmd), packet_size); + goto err_out; } } - for (i = 0; i < APP_R_HSIZE; i++) - if (!hlist_empty(mdev->app_reads_hash+i)) - dev_warn(DEV, "ASSERT FAILED: app_reads_hash[%d].first: " - "%p, should be NULL\n", i, mdev->app_reads_hash[i].first); - memset(mdev->app_reads_hash, 0, APP_R_HSIZE*sizeof(void *)); - spin_unlock_irq(&mdev->req_lock); + if (0) { + err_out: + drbd_force_state(mdev, NS(conn, C_PROTOCOL_ERROR)); + } + /* If we leave here, we probably want to update at least the + * "Connected" indicator on stable storage. Do so explicitly here. */ + drbd_md_sync(mdev); } void drbd_flush_workqueue(struct drbd_conf *mdev) @@ -3710,6 +3818,36 @@ void drbd_flush_workqueue(struct drbd_conf *mdev) wait_for_completion(&barr.done); } +void drbd_free_tl_hash(struct drbd_conf *mdev) +{ + struct hlist_head *h; + + spin_lock_irq(&mdev->req_lock); + + if (!mdev->tl_hash || mdev->state.conn != C_STANDALONE) { + spin_unlock_irq(&mdev->req_lock); + return; + } + /* paranoia code */ + for (h = mdev->ee_hash; h < mdev->ee_hash + mdev->ee_hash_s; h++) + if (h->first) + dev_err(DEV, "ASSERT FAILED ee_hash[%u].first == %p, expected NULL\n", + (int)(h - mdev->ee_hash), h->first); + kfree(mdev->ee_hash); + mdev->ee_hash = NULL; + mdev->ee_hash_s = 0; + + /* paranoia code */ + for (h = mdev->tl_hash; h < mdev->tl_hash + mdev->tl_hash_s; h++) + if (h->first) + dev_err(DEV, "ASSERT FAILED tl_hash[%u] == %p, expected NULL\n", + (int)(h - mdev->tl_hash), h->first); + kfree(mdev->tl_hash); + mdev->tl_hash = NULL; + mdev->tl_hash_s = 0; + spin_unlock_irq(&mdev->req_lock); +} + static void drbd_disconnect(struct drbd_conf *mdev) { enum drbd_fencing_p fp; @@ -3727,6 +3865,7 @@ static void drbd_disconnect(struct drbd_conf *mdev) drbd_thread_stop(&mdev->asender); drbd_free_sock(mdev); + /* wait for current activity to cease. */ spin_lock_irq(&mdev->req_lock); _drbd_wait_ee_list_empty(mdev, &mdev->active_ee); _drbd_wait_ee_list_empty(mdev, &mdev->sync_ee); @@ -3751,7 +3890,6 @@ static void drbd_disconnect(struct drbd_conf *mdev) /* make sure syncer is stopped and w_resume_next_sg queued */ del_timer_sync(&mdev->resync_timer); - set_bit(STOP_SYNC_TIMER, &mdev->flags); resync_timer_fn((unsigned long)mdev); /* wait for all w_e_end_data_req, w_e_end_rsdata_req, w_send_barrier, @@ -3766,11 +3904,9 @@ static void drbd_disconnect(struct drbd_conf *mdev) kfree(mdev->p_uuid); mdev->p_uuid = NULL; - if (!mdev->state.susp) + if (!is_susp(mdev->state)) tl_clear(mdev); - drbd_fail_pending_reads(mdev); - dev_info(DEV, "Connection closed\n"); drbd_md_sync(mdev); @@ -3781,12 +3917,8 @@ static void drbd_disconnect(struct drbd_conf *mdev) put_ldev(mdev); } - if (mdev->state.role == R_PRIMARY) { - if (fp >= FP_RESOURCE && mdev->state.pdsk >= D_UNKNOWN) { - enum drbd_disk_state nps = drbd_try_outdate_peer(mdev); - drbd_request_state(mdev, NS(pdsk, nps)); - } - } + if (mdev->state.role == R_PRIMARY && fp >= FP_RESOURCE && mdev->state.pdsk >= D_UNKNOWN) + drbd_try_outdate_peer_async(mdev); spin_lock_irq(&mdev->req_lock); os = mdev->state; @@ -3799,32 +3931,14 @@ static void drbd_disconnect(struct drbd_conf *mdev) spin_unlock_irq(&mdev->req_lock); if (os.conn == C_DISCONNECTING) { - struct hlist_head *h; - wait_event(mdev->misc_wait, atomic_read(&mdev->net_cnt) == 0); + wait_event(mdev->net_cnt_wait, atomic_read(&mdev->net_cnt) == 0); - /* we must not free the tl_hash - * while application io is still on the fly */ - wait_event(mdev->misc_wait, atomic_read(&mdev->ap_bio_cnt) == 0); - - spin_lock_irq(&mdev->req_lock); - /* paranoia code */ - for (h = mdev->ee_hash; h < mdev->ee_hash + mdev->ee_hash_s; h++) - if (h->first) - dev_err(DEV, "ASSERT FAILED ee_hash[%u].first == %p, expected NULL\n", - (int)(h - mdev->ee_hash), h->first); - kfree(mdev->ee_hash); - mdev->ee_hash = NULL; - mdev->ee_hash_s = 0; - - /* paranoia code */ - for (h = mdev->tl_hash; h < mdev->tl_hash + mdev->tl_hash_s; h++) - if (h->first) - dev_err(DEV, "ASSERT FAILED tl_hash[%u] == %p, expected NULL\n", - (int)(h - mdev->tl_hash), h->first); - kfree(mdev->tl_hash); - mdev->tl_hash = NULL; - mdev->tl_hash_s = 0; - spin_unlock_irq(&mdev->req_lock); + if (!is_susp(mdev->state)) { + /* we must not free the tl_hash + * while application io is still on the fly */ + wait_event(mdev->misc_wait, !atomic_read(&mdev->ap_bio_cnt)); + drbd_free_tl_hash(mdev); + } crypto_free_hash(mdev->cram_hmac_tfm); mdev->cram_hmac_tfm = NULL; @@ -3844,6 +3958,9 @@ static void drbd_disconnect(struct drbd_conf *mdev) i = drbd_release_ee(mdev, &mdev->net_ee); if (i) dev_info(DEV, "net_ee not empty, killed %u entries\n", i); + i = atomic_read(&mdev->pp_in_use_by_net); + if (i) + dev_info(DEV, "pp_in_use_by_net = %d, expected 0\n", i); i = atomic_read(&mdev->pp_in_use); if (i) dev_info(DEV, "pp_in_use = %d, expected 0\n", i); @@ -3887,7 +4004,7 @@ static int drbd_send_handshake(struct drbd_conf *mdev) p->protocol_min = cpu_to_be32(PRO_VERSION_MIN); p->protocol_max = cpu_to_be32(PRO_VERSION_MAX); ok = _drbd_send_cmd( mdev, mdev->data.socket, P_HAND_SHAKE, - (struct p_header *)p, sizeof(*p), 0 ); + (struct p_header80 *)p, sizeof(*p), 0 ); mutex_unlock(&mdev->data.mutex); return ok; } @@ -3903,27 +4020,28 @@ static int drbd_do_handshake(struct drbd_conf *mdev) { /* ASSERT current == mdev->receiver ... */ struct p_handshake *p = &mdev->data.rbuf.handshake; - const int expect = sizeof(struct p_handshake) - -sizeof(struct p_header); + const int expect = sizeof(struct p_handshake) - sizeof(struct p_header80); + unsigned int length; + enum drbd_packets cmd; int rv; rv = drbd_send_handshake(mdev); if (!rv) return 0; - rv = drbd_recv_header(mdev, &p->head); + rv = drbd_recv_header(mdev, &cmd, &length); if (!rv) return 0; - if (p->head.command != P_HAND_SHAKE) { + if (cmd != P_HAND_SHAKE) { dev_err(DEV, "expected HandShake packet, received: %s (0x%04x)\n", - cmdname(p->head.command), p->head.command); + cmdname(cmd), cmd); return -1; } - if (p->head.length != expect) { + if (length != expect) { dev_err(DEV, "expected HandShake length: %u, received: %u\n", - expect, p->head.length); + expect, length); return -1; } @@ -3981,10 +4099,11 @@ static int drbd_do_auth(struct drbd_conf *mdev) char *response = NULL; char *right_response = NULL; char *peers_ch = NULL; - struct p_header p; unsigned int key_len = strlen(mdev->net_conf->shared_secret); unsigned int resp_size; struct hash_desc desc; + enum drbd_packets cmd; + unsigned int length; int rv; desc.tfm = mdev->cram_hmac_tfm; @@ -4004,33 +4123,33 @@ static int drbd_do_auth(struct drbd_conf *mdev) if (!rv) goto fail; - rv = drbd_recv_header(mdev, &p); + rv = drbd_recv_header(mdev, &cmd, &length); if (!rv) goto fail; - if (p.command != P_AUTH_CHALLENGE) { + if (cmd != P_AUTH_CHALLENGE) { dev_err(DEV, "expected AuthChallenge packet, received: %s (0x%04x)\n", - cmdname(p.command), p.command); + cmdname(cmd), cmd); rv = 0; goto fail; } - if (p.length > CHALLENGE_LEN*2) { + if (length > CHALLENGE_LEN * 2) { dev_err(DEV, "expected AuthChallenge payload too big.\n"); rv = -1; goto fail; } - peers_ch = kmalloc(p.length, GFP_NOIO); + peers_ch = kmalloc(length, GFP_NOIO); if (peers_ch == NULL) { dev_err(DEV, "kmalloc of peers_ch failed\n"); rv = -1; goto fail; } - rv = drbd_recv(mdev, peers_ch, p.length); + rv = drbd_recv(mdev, peers_ch, length); - if (rv != p.length) { + if (rv != length) { dev_err(DEV, "short read AuthChallenge: l=%u\n", rv); rv = 0; goto fail; @@ -4045,7 +4164,7 @@ static int drbd_do_auth(struct drbd_conf *mdev) } sg_init_table(&sg, 1); - sg_set_buf(&sg, peers_ch, p.length); + sg_set_buf(&sg, peers_ch, length); rv = crypto_hash_digest(&desc, &sg, sg.length, response); if (rv) { @@ -4058,18 +4177,18 @@ static int drbd_do_auth(struct drbd_conf *mdev) if (!rv) goto fail; - rv = drbd_recv_header(mdev, &p); + rv = drbd_recv_header(mdev, &cmd, &length); if (!rv) goto fail; - if (p.command != P_AUTH_RESPONSE) { + if (cmd != P_AUTH_RESPONSE) { dev_err(DEV, "expected AuthResponse packet, received: %s (0x%04x)\n", - cmdname(p.command), p.command); + cmdname(cmd), cmd); rv = 0; goto fail; } - if (p.length != resp_size) { + if (length != resp_size) { dev_err(DEV, "expected AuthResponse payload of wrong size\n"); rv = 0; goto fail; @@ -4154,7 +4273,7 @@ int drbdd_init(struct drbd_thread *thi) /* ********* acknowledge sender ******** */ -static int got_RqSReply(struct drbd_conf *mdev, struct p_header *h) +static int got_RqSReply(struct drbd_conf *mdev, struct p_header80 *h) { struct p_req_state_reply *p = (struct p_req_state_reply *)h; @@ -4172,13 +4291,13 @@ static int got_RqSReply(struct drbd_conf *mdev, struct p_header *h) return TRUE; } -static int got_Ping(struct drbd_conf *mdev, struct p_header *h) +static int got_Ping(struct drbd_conf *mdev, struct p_header80 *h) { return drbd_send_ping_ack(mdev); } -static int got_PingAck(struct drbd_conf *mdev, struct p_header *h) +static int got_PingAck(struct drbd_conf *mdev, struct p_header80 *h) { /* restore idle timeout */ mdev->meta.socket->sk->sk_rcvtimeo = mdev->net_conf->ping_int*HZ; @@ -4188,7 +4307,7 @@ static int got_PingAck(struct drbd_conf *mdev, struct p_header *h) return TRUE; } -static int got_IsInSync(struct drbd_conf *mdev, struct p_header *h) +static int got_IsInSync(struct drbd_conf *mdev, struct p_header80 *h) { struct p_block_ack *p = (struct p_block_ack *)h; sector_t sector = be64_to_cpu(p->sector); @@ -4198,11 +4317,15 @@ static int got_IsInSync(struct drbd_conf *mdev, struct p_header *h) update_peer_seq(mdev, be32_to_cpu(p->seq_num)); - drbd_rs_complete_io(mdev, sector); - drbd_set_in_sync(mdev, sector, blksize); - /* rs_same_csums is supposed to count in units of BM_BLOCK_SIZE */ - mdev->rs_same_csum += (blksize >> BM_BLOCK_SHIFT); + if (get_ldev(mdev)) { + drbd_rs_complete_io(mdev, sector); + drbd_set_in_sync(mdev, sector, blksize); + /* rs_same_csums is supposed to count in units of BM_BLOCK_SIZE */ + mdev->rs_same_csum += (blksize >> BM_BLOCK_SHIFT); + put_ldev(mdev); + } dec_rs_pending(mdev); + atomic_add(blksize >> 9, &mdev->rs_sect_in); return TRUE; } @@ -4258,7 +4381,7 @@ static int validate_req_change_req_state(struct drbd_conf *mdev, return TRUE; } -static int got_BlockAck(struct drbd_conf *mdev, struct p_header *h) +static int got_BlockAck(struct drbd_conf *mdev, struct p_header80 *h) { struct p_block_ack *p = (struct p_block_ack *)h; sector_t sector = be64_to_cpu(p->sector); @@ -4298,7 +4421,7 @@ static int got_BlockAck(struct drbd_conf *mdev, struct p_header *h) _ack_id_to_req, __func__ , what); } -static int got_NegAck(struct drbd_conf *mdev, struct p_header *h) +static int got_NegAck(struct drbd_conf *mdev, struct p_header80 *h) { struct p_block_ack *p = (struct p_block_ack *)h; sector_t sector = be64_to_cpu(p->sector); @@ -4318,7 +4441,7 @@ static int got_NegAck(struct drbd_conf *mdev, struct p_header *h) _ack_id_to_req, __func__ , neg_acked); } -static int got_NegDReply(struct drbd_conf *mdev, struct p_header *h) +static int got_NegDReply(struct drbd_conf *mdev, struct p_header80 *h) { struct p_block_ack *p = (struct p_block_ack *)h; sector_t sector = be64_to_cpu(p->sector); @@ -4331,7 +4454,7 @@ static int got_NegDReply(struct drbd_conf *mdev, struct p_header *h) _ar_id_to_req, __func__ , neg_acked); } -static int got_NegRSDReply(struct drbd_conf *mdev, struct p_header *h) +static int got_NegRSDReply(struct drbd_conf *mdev, struct p_header80 *h) { sector_t sector; int size; @@ -4353,7 +4476,7 @@ static int got_NegRSDReply(struct drbd_conf *mdev, struct p_header *h) return TRUE; } -static int got_BarrierAck(struct drbd_conf *mdev, struct p_header *h) +static int got_BarrierAck(struct drbd_conf *mdev, struct p_header80 *h) { struct p_barrier_ack *p = (struct p_barrier_ack *)h; @@ -4362,7 +4485,7 @@ static int got_BarrierAck(struct drbd_conf *mdev, struct p_header *h) return TRUE; } -static int got_OVResult(struct drbd_conf *mdev, struct p_header *h) +static int got_OVResult(struct drbd_conf *mdev, struct p_header80 *h) { struct p_block_ack *p = (struct p_block_ack *)h; struct drbd_work *w; @@ -4379,6 +4502,9 @@ static int got_OVResult(struct drbd_conf *mdev, struct p_header *h) else ov_oos_print(mdev); + if (!get_ldev(mdev)) + return TRUE; + drbd_rs_complete_io(mdev, sector); dec_rs_pending(mdev); @@ -4393,18 +4519,18 @@ static int got_OVResult(struct drbd_conf *mdev, struct p_header *h) drbd_resync_finished(mdev); } } + put_ldev(mdev); return TRUE; } -static int got_something_to_ignore_m(struct drbd_conf *mdev, struct p_header *h) +static int got_skip(struct drbd_conf *mdev, struct p_header80 *h) { - /* IGNORE */ return TRUE; } struct asender_cmd { size_t pkt_size; - int (*process)(struct drbd_conf *mdev, struct p_header *h); + int (*process)(struct drbd_conf *mdev, struct p_header80 *h); }; static struct asender_cmd *get_asender_cmd(int cmd) @@ -4413,8 +4539,8 @@ static struct asender_cmd *get_asender_cmd(int cmd) /* anything missing from this table is in * the drbd_cmd_handler (drbd_default_handler) table, * see the beginning of drbdd() */ - [P_PING] = { sizeof(struct p_header), got_Ping }, - [P_PING_ACK] = { sizeof(struct p_header), got_PingAck }, + [P_PING] = { sizeof(struct p_header80), got_Ping }, + [P_PING_ACK] = { sizeof(struct p_header80), got_PingAck }, [P_RECV_ACK] = { sizeof(struct p_block_ack), got_BlockAck }, [P_WRITE_ACK] = { sizeof(struct p_block_ack), got_BlockAck }, [P_RS_WRITE_ACK] = { sizeof(struct p_block_ack), got_BlockAck }, @@ -4426,7 +4552,7 @@ static struct asender_cmd *get_asender_cmd(int cmd) [P_BARRIER_ACK] = { sizeof(struct p_barrier_ack), got_BarrierAck }, [P_STATE_CHG_REPLY] = { sizeof(struct p_req_state_reply), got_RqSReply }, [P_RS_IS_IN_SYNC] = { sizeof(struct p_block_ack), got_IsInSync }, - [P_DELAY_PROBE] = { sizeof(struct p_delay_probe), got_something_to_ignore_m }, + [P_DELAY_PROBE] = { sizeof(struct p_delay_probe93), got_skip }, [P_MAX_CMD] = { 0, NULL }, }; if (cmd > P_MAX_CMD || asender_tbl[cmd].process == NULL) @@ -4437,13 +4563,13 @@ static struct asender_cmd *get_asender_cmd(int cmd) int drbd_asender(struct drbd_thread *thi) { struct drbd_conf *mdev = thi->mdev; - struct p_header *h = &mdev->meta.rbuf.header; + struct p_header80 *h = &mdev->meta.rbuf.header.h80; struct asender_cmd *cmd = NULL; int rv, len; void *buf = h; int received = 0; - int expect = sizeof(struct p_header); + int expect = sizeof(struct p_header80); int empty; sprintf(current->comm, "drbd%d_asender", mdev_to_minor(mdev)); @@ -4467,10 +4593,8 @@ int drbd_asender(struct drbd_thread *thi) while (1) { clear_bit(SIGNAL_ASENDER, &mdev->flags); flush_signals(current); - if (!drbd_process_done_ee(mdev)) { - dev_err(DEV, "process_done_ee() = NOT_OK\n"); + if (!drbd_process_done_ee(mdev)) goto reconnect; - } /* to avoid race with newly queued ACKs */ set_bit(SIGNAL_ASENDER, &mdev->flags); spin_lock_irq(&mdev->req_lock); @@ -4529,21 +4653,23 @@ int drbd_asender(struct drbd_thread *thi) if (received == expect && cmd == NULL) { if (unlikely(h->magic != BE_DRBD_MAGIC)) { - dev_err(DEV, "magic?? on meta m: 0x%lx c: %d l: %d\n", - (long)be32_to_cpu(h->magic), - h->command, h->length); + dev_err(DEV, "magic?? on meta m: 0x%08x c: %d l: %d\n", + be32_to_cpu(h->magic), + be16_to_cpu(h->command), + be16_to_cpu(h->length)); goto reconnect; } cmd = get_asender_cmd(be16_to_cpu(h->command)); len = be16_to_cpu(h->length); if (unlikely(cmd == NULL)) { - dev_err(DEV, "unknown command?? on meta m: 0x%lx c: %d l: %d\n", - (long)be32_to_cpu(h->magic), - h->command, h->length); + dev_err(DEV, "unknown command?? on meta m: 0x%08x c: %d l: %d\n", + be32_to_cpu(h->magic), + be16_to_cpu(h->command), + be16_to_cpu(h->length)); goto disconnect; } expect = cmd->pkt_size; - ERR_IF(len != expect-sizeof(struct p_header)) + ERR_IF(len != expect-sizeof(struct p_header80)) goto reconnect; } if (received == expect) { @@ -4553,7 +4679,7 @@ int drbd_asender(struct drbd_thread *thi) buf = h; received = 0; - expect = sizeof(struct p_header); + expect = sizeof(struct p_header80); cmd = NULL; } } @@ -4561,10 +4687,12 @@ int drbd_asender(struct drbd_thread *thi) if (0) { reconnect: drbd_force_state(mdev, NS(conn, C_NETWORK_FAILURE)); + drbd_md_sync(mdev); } if (0) { disconnect: drbd_force_state(mdev, NS(conn, C_DISCONNECTING)); + drbd_md_sync(mdev); } clear_bit(SIGNAL_ASENDER, &mdev->flags); diff --git a/drivers/block/drbd/drbd_req.c b/drivers/block/drbd/drbd_req.c index f761d98..9e91a25 100644 --- a/drivers/block/drbd/drbd_req.c +++ b/drivers/block/drbd/drbd_req.c @@ -59,17 +59,19 @@ static void _drbd_end_io_acct(struct drbd_conf *mdev, struct drbd_request *req) static void _req_is_done(struct drbd_conf *mdev, struct drbd_request *req, const int rw) { const unsigned long s = req->rq_state; + + /* remove it from the transfer log. + * well, only if it had been there in the first + * place... if it had not (local only or conflicting + * and never sent), it should still be "empty" as + * initialized in drbd_req_new(), so we can list_del() it + * here unconditionally */ + list_del(&req->tl_requests); + /* if it was a write, we may have to set the corresponding * bit(s) out-of-sync first. If it had a local part, we need to * release the reference to the activity log. */ if (rw == WRITE) { - /* remove it from the transfer log. - * well, only if it had been there in the first - * place... if it had not (local only or conflicting - * and never sent), it should still be "empty" as - * initialized in drbd_req_new(), so we can list_del() it - * here unconditionally */ - list_del(&req->tl_requests); /* Set out-of-sync unless both OK flags are set * (local only or remote failed). * Other places where we set out-of-sync: @@ -92,7 +94,8 @@ static void _req_is_done(struct drbd_conf *mdev, struct drbd_request *req, const */ if (s & RQ_LOCAL_MASK) { if (get_ldev_if_state(mdev, D_FAILED)) { - drbd_al_complete_io(mdev, req->sector); + if (s & RQ_IN_ACT_LOG) + drbd_al_complete_io(mdev, req->sector); put_ldev(mdev); } else if (__ratelimit(&drbd_ratelimit_state)) { dev_warn(DEV, "Should have called drbd_al_complete_io(, %llu), " @@ -280,6 +283,14 @@ void _req_may_be_done(struct drbd_request *req, struct bio_and_error *m) * protocol A or B, barrier ack still pending... */ } +static void _req_may_be_done_not_susp(struct drbd_request *req, struct bio_and_error *m) +{ + struct drbd_conf *mdev = req->mdev; + + if (!is_susp(mdev->state)) + _req_may_be_done(req, m); +} + /* * checks whether there was an overlapping request * or ee already registered. @@ -380,10 +391,11 @@ out_conflict: * and it enforces that we have to think in a very structured manner * about the "events" that may happen to a request during its life time ... */ -void __req_mod(struct drbd_request *req, enum drbd_req_event what, +int __req_mod(struct drbd_request *req, enum drbd_req_event what, struct bio_and_error *m) { struct drbd_conf *mdev = req->mdev; + int rv = 0; m->bio = NULL; switch (what) { @@ -420,7 +432,7 @@ void __req_mod(struct drbd_request *req, enum drbd_req_event what, req->rq_state |= (RQ_LOCAL_COMPLETED|RQ_LOCAL_OK); req->rq_state &= ~RQ_LOCAL_PENDING; - _req_may_be_done(req, m); + _req_may_be_done_not_susp(req, m); put_ldev(mdev); break; @@ -429,7 +441,7 @@ void __req_mod(struct drbd_request *req, enum drbd_req_event what, req->rq_state &= ~RQ_LOCAL_PENDING; __drbd_chk_io_error(mdev, FALSE); - _req_may_be_done(req, m); + _req_may_be_done_not_susp(req, m); put_ldev(mdev); break; @@ -437,7 +449,7 @@ void __req_mod(struct drbd_request *req, enum drbd_req_event what, /* it is legal to fail READA */ req->rq_state |= RQ_LOCAL_COMPLETED; req->rq_state &= ~RQ_LOCAL_PENDING; - _req_may_be_done(req, m); + _req_may_be_done_not_susp(req, m); put_ldev(mdev); break; @@ -455,7 +467,7 @@ void __req_mod(struct drbd_request *req, enum drbd_req_event what, /* no point in retrying if there is no good remote data, * or we have no connection. */ if (mdev->state.pdsk != D_UP_TO_DATE) { - _req_may_be_done(req, m); + _req_may_be_done_not_susp(req, m); break; } @@ -517,11 +529,9 @@ void __req_mod(struct drbd_request *req, enum drbd_req_event what, D_ASSERT(test_bit(CREATE_BARRIER, &mdev->flags) == 0); req->epoch = mdev->newest_tle->br_number; - list_add_tail(&req->tl_requests, - &mdev->newest_tle->requests); /* increment size of current epoch */ - mdev->newest_tle->n_req++; + mdev->newest_tle->n_writes++; /* queue work item to send data */ D_ASSERT(req->rq_state & RQ_NET_PENDING); @@ -530,7 +540,7 @@ void __req_mod(struct drbd_request *req, enum drbd_req_event what, drbd_queue_work(&mdev->data.work, &req->w); /* close the epoch, in case it outgrew the limit */ - if (mdev->newest_tle->n_req >= mdev->net_conf->max_epoch_size) + if (mdev->newest_tle->n_writes >= mdev->net_conf->max_epoch_size) queue_barrier(mdev); break; @@ -543,7 +553,7 @@ void __req_mod(struct drbd_request *req, enum drbd_req_event what, req->rq_state &= ~RQ_NET_QUEUED; /* if we did it right, tl_clear should be scheduled only after * this, so this should not be necessary! */ - _req_may_be_done(req, m); + _req_may_be_done_not_susp(req, m); break; case handed_over_to_network: @@ -568,7 +578,7 @@ void __req_mod(struct drbd_request *req, enum drbd_req_event what, * "completed_ok" events came in, once we return from * _drbd_send_zc_bio (drbd_send_dblock), we have to check * whether it is done already, and end it. */ - _req_may_be_done(req, m); + _req_may_be_done_not_susp(req, m); break; case read_retry_remote_canceled: @@ -584,7 +594,7 @@ void __req_mod(struct drbd_request *req, enum drbd_req_event what, /* if it is still queued, we may not complete it here. * it will be canceled soon. */ if (!(req->rq_state & RQ_NET_QUEUED)) - _req_may_be_done(req, m); + _req_may_be_done(req, m); /* Allowed while state.susp */ break; case write_acked_by_peer_and_sis: @@ -619,7 +629,7 @@ void __req_mod(struct drbd_request *req, enum drbd_req_event what, D_ASSERT(req->rq_state & RQ_NET_PENDING); dec_ap_pending(mdev); req->rq_state &= ~RQ_NET_PENDING; - _req_may_be_done(req, m); + _req_may_be_done_not_susp(req, m); break; case neg_acked: @@ -629,11 +639,50 @@ void __req_mod(struct drbd_request *req, enum drbd_req_event what, req->rq_state &= ~(RQ_NET_OK|RQ_NET_PENDING); req->rq_state |= RQ_NET_DONE; - _req_may_be_done(req, m); + _req_may_be_done_not_susp(req, m); /* else: done by handed_over_to_network */ break; + case fail_frozen_disk_io: + if (!(req->rq_state & RQ_LOCAL_COMPLETED)) + break; + + _req_may_be_done(req, m); /* Allowed while state.susp */ + break; + + case restart_frozen_disk_io: + if (!(req->rq_state & RQ_LOCAL_COMPLETED)) + break; + + req->rq_state &= ~RQ_LOCAL_COMPLETED; + + rv = MR_READ; + if (bio_data_dir(req->master_bio) == WRITE) + rv = MR_WRITE; + + get_ldev(mdev); + req->w.cb = w_restart_disk_io; + drbd_queue_work(&mdev->data.work, &req->w); + break; + + case resend: + /* If RQ_NET_OK is already set, we got a P_WRITE_ACK or P_RECV_ACK + before the connection loss (B&C only); only P_BARRIER_ACK was missing. + Trowing them out of the TL here by pretending we got a BARRIER_ACK + We ensure that the peer was not rebooted */ + if (!(req->rq_state & RQ_NET_OK)) { + if (req->w.cb) { + drbd_queue_work(&mdev->data.work, &req->w); + rv = req->rq_state & RQ_WRITE ? MR_WRITE : MR_READ; + } + break; + } + /* else, fall through to barrier_acked */ + case barrier_acked: + if (!(req->rq_state & RQ_WRITE)) + break; + if (req->rq_state & RQ_NET_PENDING) { /* barrier came in before all requests have been acked. * this is bad, because if the connection is lost now, @@ -643,7 +692,7 @@ void __req_mod(struct drbd_request *req, enum drbd_req_event what, } D_ASSERT(req->rq_state & RQ_NET_SENT); req->rq_state |= RQ_NET_DONE; - _req_may_be_done(req, m); + _req_may_be_done(req, m); /* Allowed while state.susp */ break; case data_received: @@ -651,9 +700,11 @@ void __req_mod(struct drbd_request *req, enum drbd_req_event what, dec_ap_pending(mdev); req->rq_state &= ~RQ_NET_PENDING; req->rq_state |= (RQ_NET_OK|RQ_NET_DONE); - _req_may_be_done(req, m); + _req_may_be_done_not_susp(req, m); break; }; + + return rv; } /* we may do a local read if: @@ -752,14 +803,16 @@ static int drbd_make_request_common(struct drbd_conf *mdev, struct bio *bio) * resync extent to finish, and, if necessary, pulls in the target * extent into the activity log, which involves further disk io because * of transactional on-disk meta data updates. */ - if (rw == WRITE && local) + if (rw == WRITE && local && !test_bit(AL_SUSPENDED, &mdev->flags)) { + req->rq_state |= RQ_IN_ACT_LOG; drbd_al_begin_io(mdev, sector); + } remote = remote && (mdev->state.pdsk == D_UP_TO_DATE || (mdev->state.pdsk == D_INCONSISTENT && mdev->state.conn >= C_CONNECTED)); - if (!(local || remote) && !mdev->state.susp) { + if (!(local || remote) && !is_susp(mdev->state)) { dev_err(DEV, "IO ERROR: neither local nor remote disk\n"); goto fail_free_complete; } @@ -785,7 +838,7 @@ allocate_barrier: /* GOOD, everything prepared, grab the spin_lock */ spin_lock_irq(&mdev->req_lock); - if (mdev->state.susp) { + if (is_susp(mdev->state)) { /* If we got suspended, use the retry mechanism of generic_make_request() to restart processing of this bio. In the next call to drbd_make_request_26 @@ -867,30 +920,10 @@ allocate_barrier: /* check this request on the collision detection hash tables. * if we have a conflict, just complete it here. * THINK do we want to check reads, too? (I don't think so...) */ - if (rw == WRITE && _req_conflicts(req)) { - /* this is a conflicting request. - * even though it may have been only _partially_ - * overlapping with one of the currently pending requests, - * without even submitting or sending it, we will - * pretend that it was successfully served right now. - */ - if (local) { - bio_put(req->private_bio); - req->private_bio = NULL; - drbd_al_complete_io(mdev, req->sector); - put_ldev(mdev); - local = 0; - } - if (remote) - dec_ap_pending(mdev); - _drbd_end_io_acct(mdev, req); - /* THINK: do we want to fail it (-EIO), or pretend success? */ - bio_endio(req->master_bio, 0); - req->master_bio = NULL; - dec_ap_bio(mdev); - drbd_req_free(req); - remote = 0; - } + if (rw == WRITE && _req_conflicts(req)) + goto fail_conflicting; + + list_add_tail(&req->tl_requests, &mdev->newest_tle->requests); /* NOTE remote first: to get the concurrent write detection right, * we must register the request before start of local IO. */ @@ -923,6 +956,21 @@ allocate_barrier: return 0; +fail_conflicting: + /* this is a conflicting request. + * even though it may have been only _partially_ + * overlapping with one of the currently pending requests, + * without even submitting or sending it, we will + * pretend that it was successfully served right now. + */ + _drbd_end_io_acct(mdev, req); + spin_unlock_irq(&mdev->req_lock); + if (remote) + dec_ap_pending(mdev); + /* THINK: do we want to fail it (-EIO), or pretend success? + * this pretends success. */ + err = 0; + fail_free_complete: if (rw == WRITE && local) drbd_al_complete_io(mdev, sector); @@ -961,21 +1009,6 @@ static int drbd_fail_request_early(struct drbd_conf *mdev, int is_write) return 1; } - /* - * Paranoia: we might have been primary, but sync target, or - * even diskless, then lost the connection. - * This should have been handled (panic? suspend?) somewhere - * else. But maybe it was not, so check again here. - * Caution: as long as we do not have a read/write lock on mdev, - * to serialize state changes, this is racy, since we may lose - * the connection *after* we test for the cstate. - */ - if (mdev->state.disk < D_UP_TO_DATE && mdev->state.pdsk < D_UP_TO_DATE) { - if (__ratelimit(&drbd_ratelimit_state)) - dev_err(DEV, "Sorry, I have no access to good data anymore.\n"); - return 1; - } - return 0; } diff --git a/drivers/block/drbd/drbd_req.h b/drivers/block/drbd/drbd_req.h index 02d575d..181ea03 100644 --- a/drivers/block/drbd/drbd_req.h +++ b/drivers/block/drbd/drbd_req.h @@ -104,6 +104,9 @@ enum drbd_req_event { read_ahead_completed_with_error, write_completed_with_error, completed_ok, + resend, + fail_frozen_disk_io, + restart_frozen_disk_io, nothing, /* for tracing only */ }; @@ -183,6 +186,12 @@ enum drbd_req_state_bits { /* keep this last, its for the RQ_NET_MASK */ __RQ_NET_MAX, + + /* Set when this is a write, clear for a read */ + __RQ_WRITE, + + /* Should call drbd_al_complete_io() for this request... */ + __RQ_IN_ACT_LOG, }; #define RQ_LOCAL_PENDING (1UL << __RQ_LOCAL_PENDING) @@ -201,6 +210,16 @@ enum drbd_req_state_bits { /* 0x1f8 */ #define RQ_NET_MASK (((1UL << __RQ_NET_MAX)-1) & ~RQ_LOCAL_MASK) +#define RQ_WRITE (1UL << __RQ_WRITE) +#define RQ_IN_ACT_LOG (1UL << __RQ_IN_ACT_LOG) + +/* For waking up the frozen transfer log mod_req() has to return if the request + should be counted in the epoch object*/ +#define MR_WRITE_SHIFT 0 +#define MR_WRITE (1 << MR_WRITE_SHIFT) +#define MR_READ_SHIFT 1 +#define MR_READ (1 << MR_READ_SHIFT) + /* epoch entries */ static inline struct hlist_head *ee_hash_slot(struct drbd_conf *mdev, sector_t sector) @@ -244,30 +263,36 @@ static inline struct drbd_request *_ar_id_to_req(struct drbd_conf *mdev, return NULL; } +static inline void drbd_req_make_private_bio(struct drbd_request *req, struct bio *bio_src) +{ + struct bio *bio; + bio = bio_clone(bio_src, GFP_NOIO); /* XXX cannot fail?? */ + + req->private_bio = bio; + + bio->bi_private = req; + bio->bi_end_io = drbd_endio_pri; + bio->bi_next = NULL; +} + static inline struct drbd_request *drbd_req_new(struct drbd_conf *mdev, struct bio *bio_src) { - struct bio *bio; struct drbd_request *req = mempool_alloc(drbd_request_mempool, GFP_NOIO); if (likely(req)) { - bio = bio_clone(bio_src, GFP_NOIO); /* XXX cannot fail?? */ + drbd_req_make_private_bio(req, bio_src); - req->rq_state = 0; + req->rq_state = bio_data_dir(bio_src) == WRITE ? RQ_WRITE : 0; req->mdev = mdev; req->master_bio = bio_src; - req->private_bio = bio; req->epoch = 0; - req->sector = bio->bi_sector; - req->size = bio->bi_size; + req->sector = bio_src->bi_sector; + req->size = bio_src->bi_size; req->start_time = jiffies; INIT_HLIST_NODE(&req->colision); INIT_LIST_HEAD(&req->tl_requests); INIT_LIST_HEAD(&req->w.list); - - bio->bi_private = req; - bio->bi_end_io = drbd_endio_pri; - bio->bi_next = NULL; } return req; } @@ -292,36 +317,43 @@ struct bio_and_error { extern void _req_may_be_done(struct drbd_request *req, struct bio_and_error *m); -extern void __req_mod(struct drbd_request *req, enum drbd_req_event what, +extern int __req_mod(struct drbd_request *req, enum drbd_req_event what, struct bio_and_error *m); extern void complete_master_bio(struct drbd_conf *mdev, struct bio_and_error *m); /* use this if you don't want to deal with calling complete_master_bio() * outside the spinlock, e.g. when walking some list on cleanup. */ -static inline void _req_mod(struct drbd_request *req, enum drbd_req_event what) +static inline int _req_mod(struct drbd_request *req, enum drbd_req_event what) { struct drbd_conf *mdev = req->mdev; struct bio_and_error m; + int rv; /* __req_mod possibly frees req, do not touch req after that! */ - __req_mod(req, what, &m); + rv = __req_mod(req, what, &m); if (m.bio) complete_master_bio(mdev, &m); + + return rv; } /* completion of master bio is outside of spinlock. * If you need it irqsave, do it your self! */ -static inline void req_mod(struct drbd_request *req, +static inline int req_mod(struct drbd_request *req, enum drbd_req_event what) { struct drbd_conf *mdev = req->mdev; struct bio_and_error m; + int rv; + spin_lock_irq(&mdev->req_lock); - __req_mod(req, what, &m); + rv = __req_mod(req, what, &m); spin_unlock_irq(&mdev->req_lock); if (m.bio) complete_master_bio(mdev, &m); + + return rv; } #endif diff --git a/drivers/block/drbd/drbd_worker.c b/drivers/block/drbd/drbd_worker.c index ca4a16c..108d580 100644 --- a/drivers/block/drbd/drbd_worker.c +++ b/drivers/block/drbd/drbd_worker.c @@ -39,8 +39,6 @@ #include "drbd_int.h" #include "drbd_req.h" -#define SLEEP_TIME (HZ/10) - static int w_make_ov_request(struct drbd_conf *mdev, struct drbd_work *w, int cancel); @@ -217,10 +215,8 @@ void drbd_endio_sec(struct bio *bio, int error) */ void drbd_endio_pri(struct bio *bio, int error) { - unsigned long flags; struct drbd_request *req = bio->bi_private; struct drbd_conf *mdev = req->mdev; - struct bio_and_error m; enum drbd_req_event what; int uptodate = bio_flagged(bio, BIO_UPTODATE); @@ -246,12 +242,7 @@ void drbd_endio_pri(struct bio *bio, int error) bio_put(req->private_bio); req->private_bio = ERR_PTR(error); - spin_lock_irqsave(&mdev->req_lock, flags); - __req_mod(req, what, &m); - spin_unlock_irqrestore(&mdev->req_lock, flags); - - if (m.bio) - complete_master_bio(mdev, &m); + req_mod(req, what); } int w_read_retry_remote(struct drbd_conf *mdev, struct drbd_work *w, int cancel) @@ -376,54 +367,145 @@ static int read_for_csum(struct drbd_conf *mdev, sector_t sector, int size) struct drbd_epoch_entry *e; if (!get_ldev(mdev)) - return 0; + return -EIO; + + if (drbd_rs_should_slow_down(mdev)) + goto defer; /* GFP_TRY, because if there is no memory available right now, this may * be rescheduled for later. It is "only" background resync, after all. */ e = drbd_alloc_ee(mdev, DRBD_MAGIC+0xbeef, sector, size, GFP_TRY); if (!e) - goto fail; + goto defer; + e->w.cb = w_e_send_csum; spin_lock_irq(&mdev->req_lock); list_add(&e->w.list, &mdev->read_ee); spin_unlock_irq(&mdev->req_lock); - e->w.cb = w_e_send_csum; + atomic_add(size >> 9, &mdev->rs_sect_ev); if (drbd_submit_ee(mdev, e, READ, DRBD_FAULT_RS_RD) == 0) - return 1; + return 0; + + /* drbd_submit_ee currently fails for one reason only: + * not being able to allocate enough bios. + * Is dropping the connection going to help? */ + spin_lock_irq(&mdev->req_lock); + list_del(&e->w.list); + spin_unlock_irq(&mdev->req_lock); drbd_free_ee(mdev, e); -fail: +defer: put_ldev(mdev); - return 2; + return -EAGAIN; } void resync_timer_fn(unsigned long data) { - unsigned long flags; struct drbd_conf *mdev = (struct drbd_conf *) data; int queue; - spin_lock_irqsave(&mdev->req_lock, flags); - - if (likely(!test_and_clear_bit(STOP_SYNC_TIMER, &mdev->flags))) { - queue = 1; - if (mdev->state.conn == C_VERIFY_S) - mdev->resync_work.cb = w_make_ov_request; - else - mdev->resync_work.cb = w_make_resync_request; - } else { + queue = 1; + switch (mdev->state.conn) { + case C_VERIFY_S: + mdev->resync_work.cb = w_make_ov_request; + break; + case C_SYNC_TARGET: + mdev->resync_work.cb = w_make_resync_request; + break; + default: queue = 0; mdev->resync_work.cb = w_resync_inactive; } - spin_unlock_irqrestore(&mdev->req_lock, flags); - /* harmless race: list_empty outside data.work.q_lock */ if (list_empty(&mdev->resync_work.list) && queue) drbd_queue_work(&mdev->data.work, &mdev->resync_work); } +static void fifo_set(struct fifo_buffer *fb, int value) +{ + int i; + + for (i = 0; i < fb->size; i++) + fb->values[i] = value; +} + +static int fifo_push(struct fifo_buffer *fb, int value) +{ + int ov; + + ov = fb->values[fb->head_index]; + fb->values[fb->head_index++] = value; + + if (fb->head_index >= fb->size) + fb->head_index = 0; + + return ov; +} + +static void fifo_add_val(struct fifo_buffer *fb, int value) +{ + int i; + + for (i = 0; i < fb->size; i++) + fb->values[i] += value; +} + +int drbd_rs_controller(struct drbd_conf *mdev) +{ + unsigned int sect_in; /* Number of sectors that came in since the last turn */ + unsigned int want; /* The number of sectors we want in the proxy */ + int req_sect; /* Number of sectors to request in this turn */ + int correction; /* Number of sectors more we need in the proxy*/ + int cps; /* correction per invocation of drbd_rs_controller() */ + int steps; /* Number of time steps to plan ahead */ + int curr_corr; + int max_sect; + + sect_in = atomic_xchg(&mdev->rs_sect_in, 0); /* Number of sectors that came in */ + mdev->rs_in_flight -= sect_in; + + spin_lock(&mdev->peer_seq_lock); /* get an atomic view on mdev->rs_plan_s */ + + steps = mdev->rs_plan_s.size; /* (mdev->sync_conf.c_plan_ahead * 10 * SLEEP_TIME) / HZ; */ + + if (mdev->rs_in_flight + sect_in == 0) { /* At start of resync */ + want = ((mdev->sync_conf.rate * 2 * SLEEP_TIME) / HZ) * steps; + } else { /* normal path */ + want = mdev->sync_conf.c_fill_target ? mdev->sync_conf.c_fill_target : + sect_in * mdev->sync_conf.c_delay_target * HZ / (SLEEP_TIME * 10); + } + + correction = want - mdev->rs_in_flight - mdev->rs_planed; + + /* Plan ahead */ + cps = correction / steps; + fifo_add_val(&mdev->rs_plan_s, cps); + mdev->rs_planed += cps * steps; + + /* What we do in this step */ + curr_corr = fifo_push(&mdev->rs_plan_s, 0); + spin_unlock(&mdev->peer_seq_lock); + mdev->rs_planed -= curr_corr; + + req_sect = sect_in + curr_corr; + if (req_sect < 0) + req_sect = 0; + + max_sect = (mdev->sync_conf.c_max_rate * 2 * SLEEP_TIME) / HZ; + if (req_sect > max_sect) + req_sect = max_sect; + + /* + dev_warn(DEV, "si=%u if=%d wa=%u co=%d st=%d cps=%d pl=%d cc=%d rs=%d\n", + sect_in, mdev->rs_in_flight, want, correction, + steps, cps, mdev->rs_planed, curr_corr, req_sect); + */ + + return req_sect; +} + int w_make_resync_request(struct drbd_conf *mdev, struct drbd_work *w, int cancel) { @@ -431,8 +513,9 @@ int w_make_resync_request(struct drbd_conf *mdev, sector_t sector; const sector_t capacity = drbd_get_capacity(mdev->this_bdev); int max_segment_size; - int number, i, size, pe, mx; + int number, rollback_i, size, pe, mx; int align, queued, sndbuf; + int i = 0; if (unlikely(cancel)) return 1; @@ -446,6 +529,12 @@ int w_make_resync_request(struct drbd_conf *mdev, dev_err(DEV, "%s in w_make_resync_request\n", drbd_conn_str(mdev->state.conn)); + if (mdev->rs_total == 0) { + /* empty resync? */ + drbd_resync_finished(mdev); + return 1; + } + if (!get_ldev(mdev)) { /* Since we only need to access mdev->rsync a get_ldev_if_state(mdev,D_FAILED) would be sufficient, but @@ -458,11 +547,25 @@ int w_make_resync_request(struct drbd_conf *mdev, /* starting with drbd 8.3.8, we can handle multi-bio EEs, * if it should be necessary */ - max_segment_size = mdev->agreed_pro_version < 94 ? - queue_max_segment_size(mdev->rq_queue) : DRBD_MAX_SEGMENT_SIZE; + max_segment_size = + mdev->agreed_pro_version < 94 ? queue_max_segment_size(mdev->rq_queue) : + mdev->agreed_pro_version < 95 ? DRBD_MAX_SIZE_H80_PACKET : DRBD_MAX_SEGMENT_SIZE; - number = SLEEP_TIME * mdev->sync_conf.rate / ((BM_BLOCK_SIZE / 1024) * HZ); - pe = atomic_read(&mdev->rs_pending_cnt); + if (mdev->rs_plan_s.size) { /* mdev->sync_conf.c_plan_ahead */ + number = drbd_rs_controller(mdev) >> (BM_BLOCK_SHIFT - 9); + mdev->c_sync_rate = number * HZ * (BM_BLOCK_SIZE / 1024) / SLEEP_TIME; + } else { + mdev->c_sync_rate = mdev->sync_conf.rate; + number = SLEEP_TIME * mdev->c_sync_rate / ((BM_BLOCK_SIZE / 1024) * HZ); + } + + /* Throttle resync on lower level disk activity, which may also be + * caused by application IO on Primary/SyncTarget. + * Keep this after the call to drbd_rs_controller, as that assumes + * to be called as precisely as possible every SLEEP_TIME, + * and would be confused otherwise. */ + if (drbd_rs_should_slow_down(mdev)) + goto requeue; mutex_lock(&mdev->data.mutex); if (mdev->data.socket) @@ -476,6 +579,7 @@ int w_make_resync_request(struct drbd_conf *mdev, mx = number; /* Limit the number of pending RS requests to no more than the peer's receive buffer */ + pe = atomic_read(&mdev->rs_pending_cnt); if ((pe + number) > mx) { number = mx - pe; } @@ -526,6 +630,7 @@ next_sector: * be prepared for all stripe sizes of software RAIDs. */ align = 1; + rollback_i = i; for (;;) { if (size + BM_BLOCK_SIZE > max_segment_size) break; @@ -561,14 +666,19 @@ next_sector: size = (capacity-sector)<<9; if (mdev->agreed_pro_version >= 89 && mdev->csums_tfm) { switch (read_for_csum(mdev, sector, size)) { - case 0: /* Disk failure*/ + case -EIO: /* Disk failure */ put_ldev(mdev); return 0; - case 2: /* Allocation failed */ + case -EAGAIN: /* allocation failed, or ldev busy */ drbd_rs_complete_io(mdev, sector); mdev->bm_resync_fo = BM_SECT_TO_BIT(sector); + i = rollback_i; goto requeue; - /* case 1: everything ok */ + case 0: + /* everything ok */ + break; + default: + BUG(); } } else { inc_rs_pending(mdev); @@ -595,6 +705,7 @@ next_sector: } requeue: + mdev->rs_in_flight += (i << (BM_BLOCK_SHIFT - 9)); mod_timer(&mdev->resync_timer, jiffies + SLEEP_TIME); put_ldev(mdev); return 1; @@ -670,6 +781,14 @@ static int w_resync_finished(struct drbd_conf *mdev, struct drbd_work *w, int ca return 1; } +static void ping_peer(struct drbd_conf *mdev) +{ + clear_bit(GOT_PING_ACK, &mdev->flags); + request_ping(mdev); + wait_event(mdev->misc_wait, + test_bit(GOT_PING_ACK, &mdev->flags) || mdev->state.conn < C_CONNECTED); +} + int drbd_resync_finished(struct drbd_conf *mdev) { unsigned long db, dt, dbdt; @@ -709,6 +828,8 @@ int drbd_resync_finished(struct drbd_conf *mdev) if (!get_ldev(mdev)) goto out; + ping_peer(mdev); + spin_lock_irq(&mdev->req_lock); os = mdev->state; @@ -801,6 +922,8 @@ out: mdev->rs_paused = 0; mdev->ov_start_sector = 0; + drbd_md_sync(mdev); + if (test_and_clear_bit(WRITE_BM_AFTER_RESYNC, &mdev->flags)) { dev_warn(DEV, "Writing the whole bitmap, due to failed kmalloc\n"); drbd_queue_bitmap_io(mdev, &drbd_bm_write, NULL, "write from resync_finished"); @@ -817,9 +940,13 @@ static void move_to_net_ee_or_free(struct drbd_conf *mdev, struct drbd_epoch_ent { if (drbd_ee_has_active_page(e)) { /* This might happen if sendpage() has not finished */ + int i = (e->size + PAGE_SIZE -1) >> PAGE_SHIFT; + atomic_add(i, &mdev->pp_in_use_by_net); + atomic_sub(i, &mdev->pp_in_use); spin_lock_irq(&mdev->req_lock); list_add_tail(&e->w.list, &mdev->net_ee); spin_unlock_irq(&mdev->req_lock); + wake_up(&drbd_pp_wait); } else drbd_free_ee(mdev, e); } @@ -926,9 +1053,12 @@ int w_e_end_csum_rs_req(struct drbd_conf *mdev, struct drbd_work *w, int cancel) return 1; } - drbd_rs_complete_io(mdev, e->sector); + if (get_ldev(mdev)) { + drbd_rs_complete_io(mdev, e->sector); + put_ldev(mdev); + } - di = (struct digest_info *)(unsigned long)e->block_id; + di = e->digest; if (likely((e->flags & EE_WAS_ERROR) == 0)) { /* quick hack to try to avoid a race against reconfiguration. @@ -952,7 +1082,9 @@ int w_e_end_csum_rs_req(struct drbd_conf *mdev, struct drbd_work *w, int cancel) ok = drbd_send_ack(mdev, P_RS_IS_IN_SYNC, e); } else { inc_rs_pending(mdev); - e->block_id = ID_SYNCER; + e->block_id = ID_SYNCER; /* By setting block_id, digest pointer becomes invalid! */ + e->flags &= ~EE_HAS_DIGEST; /* This e no longer has a digest pointer */ + kfree(di); ok = drbd_send_block(mdev, P_RS_DATA_REPLY, e); } } else { @@ -962,9 +1094,6 @@ int w_e_end_csum_rs_req(struct drbd_conf *mdev, struct drbd_work *w, int cancel) } dec_unacked(mdev); - - kfree(di); - move_to_net_ee_or_free(mdev, e); if (unlikely(!ok)) @@ -1034,9 +1163,12 @@ int w_e_end_ov_reply(struct drbd_conf *mdev, struct drbd_work *w, int cancel) /* after "cancel", because after drbd_disconnect/drbd_rs_cancel_all * the resync lru has been cleaned up already */ - drbd_rs_complete_io(mdev, e->sector); + if (get_ldev(mdev)) { + drbd_rs_complete_io(mdev, e->sector); + put_ldev(mdev); + } - di = (struct digest_info *)(unsigned long)e->block_id; + di = e->digest; if (likely((e->flags & EE_WAS_ERROR) == 0)) { digest_size = crypto_hash_digestsize(mdev->verify_tfm); @@ -1055,9 +1187,6 @@ int w_e_end_ov_reply(struct drbd_conf *mdev, struct drbd_work *w, int cancel) } dec_unacked(mdev); - - kfree(di); - if (!eq) drbd_ov_oos_found(mdev, e->sector, e->size); else @@ -1108,7 +1237,7 @@ int w_send_barrier(struct drbd_conf *mdev, struct drbd_work *w, int cancel) * dec_ap_pending will be done in got_BarrierAck * or (on connection loss) in w_clear_epoch. */ ok = _drbd_send_cmd(mdev, mdev->data.socket, P_BARRIER, - (struct p_header *)p, sizeof(*p), 0); + (struct p_header80 *)p, sizeof(*p), 0); drbd_put_data_sock(mdev); return ok; @@ -1173,6 +1302,24 @@ int w_send_read_req(struct drbd_conf *mdev, struct drbd_work *w, int cancel) return ok; } +int w_restart_disk_io(struct drbd_conf *mdev, struct drbd_work *w, int cancel) +{ + struct drbd_request *req = container_of(w, struct drbd_request, w); + + if (bio_data_dir(req->master_bio) == WRITE && req->rq_state & RQ_IN_ACT_LOG) + drbd_al_begin_io(mdev, req->sector); + /* Calling drbd_al_begin_io() out of the worker might deadlocks + theoretically. Practically it can not deadlock, since this is + only used when unfreezing IOs. All the extents of the requests + that made it into the TL are already active */ + + drbd_req_make_private_bio(req, req->master_bio); + req->private_bio->bi_bdev = mdev->ldev->backing_bdev; + generic_make_request(req->private_bio); + + return 1; +} + static int _drbd_may_sync_now(struct drbd_conf *mdev) { struct drbd_conf *odev = mdev; @@ -1298,14 +1445,6 @@ int drbd_alter_sa(struct drbd_conf *mdev, int na) return retcode; } -static void ping_peer(struct drbd_conf *mdev) -{ - clear_bit(GOT_PING_ACK, &mdev->flags); - request_ping(mdev); - wait_event(mdev->misc_wait, - test_bit(GOT_PING_ACK, &mdev->flags) || mdev->state.conn < C_CONNECTED); -} - /** * drbd_start_resync() - Start the resync process * @mdev: DRBD device. @@ -1379,13 +1518,21 @@ void drbd_start_resync(struct drbd_conf *mdev, enum drbd_conns side) r = SS_UNKNOWN_ERROR; if (r == SS_SUCCESS) { - mdev->rs_total = - mdev->rs_mark_left = drbd_bm_total_weight(mdev); + unsigned long tw = drbd_bm_total_weight(mdev); + unsigned long now = jiffies; + int i; + mdev->rs_failed = 0; mdev->rs_paused = 0; - mdev->rs_start = - mdev->rs_mark_time = jiffies; mdev->rs_same_csum = 0; + mdev->rs_last_events = 0; + mdev->rs_last_sect_ev = 0; + mdev->rs_total = tw; + mdev->rs_start = now; + for (i = 0; i < DRBD_SYNC_MARKS; i++) { + mdev->rs_mark_left[i] = tw; + mdev->rs_mark_time[i] = now; + } _drbd_pause_after(mdev); } write_unlock_irq(&global_state_lock); @@ -1397,12 +1544,31 @@ void drbd_start_resync(struct drbd_conf *mdev, enum drbd_conns side) (unsigned long) mdev->rs_total << (BM_BLOCK_SHIFT-10), (unsigned long) mdev->rs_total); - if (mdev->rs_total == 0) { - /* Peer still reachable? Beware of failing before-resync-target handlers! */ - ping_peer(mdev); + if (mdev->agreed_pro_version < 95 && mdev->rs_total == 0) { + /* This still has a race (about when exactly the peers + * detect connection loss) that can lead to a full sync + * on next handshake. In 8.3.9 we fixed this with explicit + * resync-finished notifications, but the fix + * introduces a protocol change. Sleeping for some + * time longer than the ping interval + timeout on the + * SyncSource, to give the SyncTarget the chance to + * detect connection loss, then waiting for a ping + * response (implicit in drbd_resync_finished) reduces + * the race considerably, but does not solve it. */ + if (side == C_SYNC_SOURCE) + schedule_timeout_interruptible( + mdev->net_conf->ping_int * HZ + + mdev->net_conf->ping_timeo*HZ/9); drbd_resync_finished(mdev); } + atomic_set(&mdev->rs_sect_in, 0); + atomic_set(&mdev->rs_sect_ev, 0); + mdev->rs_in_flight = 0; + mdev->rs_planed = 0; + spin_lock(&mdev->peer_seq_lock); + fifo_set(&mdev->rs_plan_s, 0); + spin_unlock(&mdev->peer_seq_lock); /* ns.conn may already be != mdev->state.conn, * we may have been paused in between, or become paused until * the timer triggers. |