diff options
Diffstat (limited to 'fs/ocfs2/cluster')
| -rw-r--r-- | fs/ocfs2/cluster/Makefile | 2 | ||||
| -rw-r--r-- | fs/ocfs2/cluster/heartbeat.c | 1171 | ||||
| -rw-r--r-- | fs/ocfs2/cluster/heartbeat.h | 7 | ||||
| -rw-r--r-- | fs/ocfs2/cluster/masklog.c | 20 | ||||
| -rw-r--r-- | fs/ocfs2/cluster/masklog.h | 152 | ||||
| -rw-r--r-- | fs/ocfs2/cluster/netdebug.c | 579 | ||||
| -rw-r--r-- | fs/ocfs2/cluster/nodemanager.c | 159 | ||||
| -rw-r--r-- | fs/ocfs2/cluster/nodemanager.h | 9 | ||||
| -rw-r--r-- | fs/ocfs2/cluster/ocfs2_nodemanager.h | 6 | ||||
| -rw-r--r-- | fs/ocfs2/cluster/quorum.c | 27 | ||||
| -rw-r--r-- | fs/ocfs2/cluster/sys.c | 4 | ||||
| -rw-r--r-- | fs/ocfs2/cluster/tcp.c | 659 | ||||
| -rw-r--r-- | fs/ocfs2/cluster/tcp.h | 34 | ||||
| -rw-r--r-- | fs/ocfs2/cluster/tcp_internal.h | 55 | ||||
| -rw-r--r-- | fs/ocfs2/cluster/ver.c | 42 | ||||
| -rw-r--r-- | fs/ocfs2/cluster/ver.h | 31 |
16 files changed, 2223 insertions, 734 deletions
diff --git a/fs/ocfs2/cluster/Makefile b/fs/ocfs2/cluster/Makefile index cdd162f1365..1aefc0350ec 100644 --- a/fs/ocfs2/cluster/Makefile +++ b/fs/ocfs2/cluster/Makefile @@ -1,4 +1,4 @@ obj-$(CONFIG_OCFS2_FS) += ocfs2_nodemanager.o ocfs2_nodemanager-objs := heartbeat.o masklog.o sys.o nodemanager.o \ - quorum.o tcp.o ver.o + quorum.o tcp.o netdebug.o diff --git a/fs/ocfs2/cluster/heartbeat.c b/fs/ocfs2/cluster/heartbeat.c index f02ccb34604..73039295d0d 100644 --- a/fs/ocfs2/cluster/heartbeat.c +++ b/fs/ocfs2/cluster/heartbeat.c @@ -33,6 +33,9 @@ #include <linux/random.h> #include <linux/crc32.h> #include <linux/time.h> +#include <linux/debugfs.h> +#include <linux/slab.h> +#include <linux/bitmap.h> #include "heartbeat.h" #include "tcp.h" @@ -60,6 +63,54 @@ static unsigned long o2hb_live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)]; static LIST_HEAD(o2hb_node_events); static DECLARE_WAIT_QUEUE_HEAD(o2hb_steady_queue); +/* + * In global heartbeat, we maintain a series of region bitmaps. + * - o2hb_region_bitmap allows us to limit the region number to max region. + * - o2hb_live_region_bitmap tracks live regions (seen steady iterations). + * - o2hb_quorum_region_bitmap tracks live regions that have seen all nodes + * heartbeat on it. + * - o2hb_failed_region_bitmap tracks the regions that have seen io timeouts. + */ +static unsigned long o2hb_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)]; +static unsigned long o2hb_live_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)]; +static unsigned long o2hb_quorum_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)]; +static unsigned long o2hb_failed_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)]; + +#define O2HB_DB_TYPE_LIVENODES 0 +#define O2HB_DB_TYPE_LIVEREGIONS 1 +#define O2HB_DB_TYPE_QUORUMREGIONS 2 +#define O2HB_DB_TYPE_FAILEDREGIONS 3 +#define O2HB_DB_TYPE_REGION_LIVENODES 4 +#define O2HB_DB_TYPE_REGION_NUMBER 5 +#define O2HB_DB_TYPE_REGION_ELAPSED_TIME 6 +#define O2HB_DB_TYPE_REGION_PINNED 7 +struct o2hb_debug_buf { + int db_type; + int db_size; + int db_len; + void *db_data; +}; + +static struct o2hb_debug_buf *o2hb_db_livenodes; +static struct o2hb_debug_buf *o2hb_db_liveregions; +static struct o2hb_debug_buf *o2hb_db_quorumregions; +static struct o2hb_debug_buf *o2hb_db_failedregions; + +#define O2HB_DEBUG_DIR "o2hb" +#define O2HB_DEBUG_LIVENODES "livenodes" +#define O2HB_DEBUG_LIVEREGIONS "live_regions" +#define O2HB_DEBUG_QUORUMREGIONS "quorum_regions" +#define O2HB_DEBUG_FAILEDREGIONS "failed_regions" +#define O2HB_DEBUG_REGION_NUMBER "num" +#define O2HB_DEBUG_REGION_ELAPSED_TIME "elapsed_time_in_ms" +#define O2HB_DEBUG_REGION_PINNED "pinned" + +static struct dentry *o2hb_debug_dir; +static struct dentry *o2hb_debug_livenodes; +static struct dentry *o2hb_debug_liveregions; +static struct dentry *o2hb_debug_quorumregions; +static struct dentry *o2hb_debug_failedregions; + static LIST_HEAD(o2hb_all_regions); static struct o2hb_callback { @@ -70,9 +121,48 @@ static struct o2hb_callback *hbcall_from_type(enum o2hb_callback_type type); #define O2HB_DEFAULT_BLOCK_BITS 9 +enum o2hb_heartbeat_modes { + O2HB_HEARTBEAT_LOCAL = 0, + O2HB_HEARTBEAT_GLOBAL, + O2HB_HEARTBEAT_NUM_MODES, +}; + +char *o2hb_heartbeat_mode_desc[O2HB_HEARTBEAT_NUM_MODES] = { + "local", /* O2HB_HEARTBEAT_LOCAL */ + "global", /* O2HB_HEARTBEAT_GLOBAL */ +}; + unsigned int o2hb_dead_threshold = O2HB_DEFAULT_DEAD_THRESHOLD; +unsigned int o2hb_heartbeat_mode = O2HB_HEARTBEAT_LOCAL; + +/* + * o2hb_dependent_users tracks the number of registered callbacks that depend + * on heartbeat. o2net and o2dlm are two entities that register this callback. + * However only o2dlm depends on the heartbeat. It does not want the heartbeat + * to stop while a dlm domain is still active. + */ +unsigned int o2hb_dependent_users; -/* Only sets a new threshold if there are no active regions. +/* + * In global heartbeat mode, all regions are pinned if there are one or more + * dependent users and the quorum region count is <= O2HB_PIN_CUT_OFF. All + * regions are unpinned if the region count exceeds the cut off or the number + * of dependent users falls to zero. + */ +#define O2HB_PIN_CUT_OFF 3 + +/* + * In local heartbeat mode, we assume the dlm domain name to be the same as + * region uuid. This is true for domains created for the file system but not + * necessarily true for userdlm domains. This is a known limitation. + * + * In global heartbeat mode, we pin/unpin all o2hb regions. This solution + * works for both file system and userdlm domains. + */ +static int o2hb_region_pin(const char *region_uuid); +static void o2hb_region_unpin(const char *region_uuid); + +/* Only sets a new threshold if there are no active regions. * * No locking or otherwise interesting code is required for reading * o2hb_dead_threshold as it can't change once regions are active and @@ -87,6 +177,22 @@ static void o2hb_dead_threshold_set(unsigned int threshold) } } +static int o2hb_global_heartbeat_mode_set(unsigned int hb_mode) +{ + int ret = -1; + + if (hb_mode < O2HB_HEARTBEAT_NUM_MODES) { + spin_lock(&o2hb_live_lock); + if (list_empty(&o2hb_all_regions)) { + o2hb_heartbeat_mode = hb_mode; + ret = 0; + } + spin_unlock(&o2hb_live_lock); + } + + return ret; +} + struct o2hb_node_event { struct list_head hn_item; enum o2hb_callback_type hn_event_type; @@ -110,7 +216,10 @@ struct o2hb_region { struct config_item hr_item; struct list_head hr_all_item; - unsigned hr_unclean_stop:1; + unsigned hr_unclean_stop:1, + hr_aborted_start:1, + hr_item_pinned:1, + hr_item_dropped:1; /* protected by the hr_callback_sem */ struct task_struct *hr_task; @@ -128,11 +237,29 @@ struct o2hb_region { struct block_device *hr_bdev; struct o2hb_disk_slot *hr_slots; + /* live node map of this region */ + unsigned long hr_live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)]; + unsigned int hr_region_num; + + struct dentry *hr_debug_dir; + struct dentry *hr_debug_livenodes; + struct dentry *hr_debug_regnum; + struct dentry *hr_debug_elapsed_time; + struct dentry *hr_debug_pinned; + struct o2hb_debug_buf *hr_db_livenodes; + struct o2hb_debug_buf *hr_db_regnum; + struct o2hb_debug_buf *hr_db_elapsed_time; + struct o2hb_debug_buf *hr_db_pinned; + /* let the person setting up hb wait for it to return until it * has reached a 'steady' state. This will be fixed when we have * a more complete api that doesn't lead to this sort of fragility. */ atomic_t hr_steady_iterations; + /* terminate o2hb thread if it does not reach steady state + * (hr_steady_iterations == 0) within hr_unsteady_iterations */ + atomic_t hr_unsteady_iterations; + char hr_dev_name[BDEVNAME_SIZE]; unsigned int hr_timeout_ms; @@ -158,20 +285,54 @@ struct o2hb_bio_wait_ctxt { static void o2hb_write_timeout(struct work_struct *work) { + int failed, quorum; + unsigned long flags; struct o2hb_region *reg = container_of(work, struct o2hb_region, hr_write_timeout_work.work); mlog(ML_ERROR, "Heartbeat write timeout to device %s after %u " "milliseconds\n", reg->hr_dev_name, - jiffies_to_msecs(jiffies - reg->hr_last_timeout_start)); + jiffies_to_msecs(jiffies - reg->hr_last_timeout_start)); + + if (o2hb_global_heartbeat_active()) { + spin_lock_irqsave(&o2hb_live_lock, flags); + if (test_bit(reg->hr_region_num, o2hb_quorum_region_bitmap)) + set_bit(reg->hr_region_num, o2hb_failed_region_bitmap); + failed = bitmap_weight(o2hb_failed_region_bitmap, + O2NM_MAX_REGIONS); + quorum = bitmap_weight(o2hb_quorum_region_bitmap, + O2NM_MAX_REGIONS); + spin_unlock_irqrestore(&o2hb_live_lock, flags); + + mlog(ML_HEARTBEAT, "Number of regions %d, failed regions %d\n", + quorum, failed); + + /* + * Fence if the number of failed regions >= half the number + * of quorum regions + */ + if ((failed << 1) < quorum) + return; + } + o2quo_disk_timeout(); } static void o2hb_arm_write_timeout(struct o2hb_region *reg) { - mlog(0, "Queue write timeout for %u ms\n", O2HB_MAX_WRITE_TIMEOUT_MS); + /* Arm writeout only after thread reaches steady state */ + if (atomic_read(®->hr_steady_iterations) != 0) + return; + + mlog(ML_HEARTBEAT, "Queue write timeout for %u ms\n", + O2HB_MAX_WRITE_TIMEOUT_MS); + if (o2hb_global_heartbeat_active()) { + spin_lock(&o2hb_live_lock); + clear_bit(reg->hr_region_num, o2hb_failed_region_bitmap); + spin_unlock(&o2hb_live_lock); + } cancel_delayed_work(®->hr_write_timeout_work); reg->hr_last_timeout_start = jiffies; schedule_delayed_work(®->hr_write_timeout_work, @@ -180,8 +341,7 @@ static void o2hb_arm_write_timeout(struct o2hb_region *reg) static void o2hb_disarm_write_timeout(struct o2hb_region *reg) { - cancel_delayed_work(®->hr_write_timeout_work); - flush_scheduled_work(); + cancel_delayed_work_sync(®->hr_write_timeout_work); } static inline void o2hb_bio_wait_init(struct o2hb_bio_wait_ctxt *wc) @@ -208,11 +368,7 @@ static inline void o2hb_bio_wait_dec(struct o2hb_bio_wait_ctxt *wc, static void o2hb_wait_on_io(struct o2hb_region *reg, struct o2hb_bio_wait_ctxt *wc) { - struct address_space *mapping = reg->hr_bdev->bd_inode->i_mapping; - - blk_run_address_space(mapping); o2hb_bio_wait_dec(wc, 1); - wait_for_completion(&wc->wc_io_complete); } @@ -257,7 +413,7 @@ static struct bio *o2hb_setup_one_bio(struct o2hb_region *reg, } /* Must put everything in 512 byte sectors for the bio... */ - bio->bi_sector = (reg->hr_start_block + cs) << (bits - 9); + bio->bi_iter.bi_sector = (reg->hr_start_block + cs) << (bits - 9); bio->bi_bdev = reg->hr_bdev; bio->bi_private = wc; bio->bi_end_io = o2hb_bio_end_io; @@ -336,7 +492,7 @@ static int o2hb_issue_node_write(struct o2hb_region *reg, } atomic_inc(&write_wc->wc_num_reqs); - submit_bio(WRITE, bio); + submit_bio(WRITE_SYNC, bio); status = 0; bail: @@ -382,27 +538,50 @@ static int o2hb_verify_crc(struct o2hb_region *reg, return read == computed; } -/* We want to make sure that nobody is heartbeating on top of us -- - * this will help detect an invalid configuration. */ -static int o2hb_check_last_timestamp(struct o2hb_region *reg) +/* + * Compare the slot data with what we wrote in the last iteration. + * If the match fails, print an appropriate error message. This is to + * detect errors like... another node hearting on the same slot, + * flaky device that is losing writes, etc. + * Returns 1 if check succeeds, 0 otherwise. + */ +static int o2hb_check_own_slot(struct o2hb_region *reg) { - int node_num, ret; struct o2hb_disk_slot *slot; struct o2hb_disk_heartbeat_block *hb_block; + char *errstr; - node_num = o2nm_this_node(); - - ret = 1; - slot = ®->hr_slots[node_num]; + slot = ®->hr_slots[o2nm_this_node()]; /* Don't check on our 1st timestamp */ - if (slot->ds_last_time) { - hb_block = slot->ds_raw_block; + if (!slot->ds_last_time) + return 0; - if (le64_to_cpu(hb_block->hb_seq) != slot->ds_last_time) - ret = 0; - } + hb_block = slot->ds_raw_block; + if (le64_to_cpu(hb_block->hb_seq) == slot->ds_last_time && + le64_to_cpu(hb_block->hb_generation) == slot->ds_last_generation && + hb_block->hb_node == slot->ds_node_num) + return 1; + +#define ERRSTR1 "Another node is heartbeating on device" +#define ERRSTR2 "Heartbeat generation mismatch on device" +#define ERRSTR3 "Heartbeat sequence mismatch on device" + + if (hb_block->hb_node != slot->ds_node_num) + errstr = ERRSTR1; + else if (le64_to_cpu(hb_block->hb_generation) != + slot->ds_last_generation) + errstr = ERRSTR2; + else + errstr = ERRSTR3; - return ret; + mlog(ML_ERROR, "%s (%s): expected(%u:0x%llx, 0x%llx), " + "ondisk(%u:0x%llx, 0x%llx)\n", errstr, reg->hr_dev_name, + slot->ds_node_num, (unsigned long long)slot->ds_last_generation, + (unsigned long long)slot->ds_last_time, hb_block->hb_node, + (unsigned long long)le64_to_cpu(hb_block->hb_generation), + (unsigned long long)le64_to_cpu(hb_block->hb_seq)); + + return 0; } static inline void o2hb_prepare_block(struct o2hb_region *reg, @@ -441,11 +620,9 @@ static void o2hb_fire_callbacks(struct o2hb_callback *hbcall, struct o2nm_node *node, int idx) { - struct list_head *iter; struct o2hb_callback_func *f; - list_for_each(iter, &hbcall->list) { - f = list_entry(iter, struct o2hb_callback_func, hc_item); + list_for_each_entry(f, &hbcall->list, hc_item) { mlog(ML_HEARTBEAT, "calling funcs %p\n", f); (f->hc_func)(node, idx, f->hc_data); } @@ -454,16 +631,9 @@ static void o2hb_fire_callbacks(struct o2hb_callback *hbcall, /* Will run the list in order until we process the passed event */ static void o2hb_run_event_list(struct o2hb_node_event *queued_event) { - int empty; struct o2hb_callback *hbcall; struct o2hb_node_event *event; - spin_lock(&o2hb_live_lock); - empty = list_empty(&queued_event->hn_item); - spin_unlock(&o2hb_live_lock); - if (empty) - return; - /* Holding callback sem assures we don't alter the callback * lists when doing this, and serializes ourselves with other * processes wanting callbacks. */ @@ -505,6 +675,8 @@ static void o2hb_queue_node_event(struct o2hb_node_event *event, { assert_spin_locked(&o2hb_live_lock); + BUG_ON((!node) && (type != O2HB_NODE_DOWN_CB)); + event->hn_event_type = type; event->hn_node = node; event->hn_node_num = node_num; @@ -520,6 +692,7 @@ static void o2hb_shutdown_slot(struct o2hb_disk_slot *slot) struct o2hb_node_event event = { .hn_item = LIST_HEAD_INIT(event.hn_item), }; struct o2nm_node *node; + int queued = 0; node = o2nm_get_node_by_num(slot->ds_node_num); if (!node) @@ -537,15 +710,60 @@ static void o2hb_shutdown_slot(struct o2hb_disk_slot *slot) o2hb_queue_node_event(&event, O2HB_NODE_DOWN_CB, node, slot->ds_node_num); + queued = 1; } } spin_unlock(&o2hb_live_lock); - o2hb_run_event_list(&event); + if (queued) + o2hb_run_event_list(&event); o2nm_node_put(node); } +static void o2hb_set_quorum_device(struct o2hb_region *reg) +{ + if (!o2hb_global_heartbeat_active()) + return; + + /* Prevent race with o2hb_heartbeat_group_drop_item() */ + if (kthread_should_stop()) + return; + + /* Tag region as quorum only after thread reaches steady state */ + if (atomic_read(®->hr_steady_iterations) != 0) + return; + + spin_lock(&o2hb_live_lock); + + if (test_bit(reg->hr_region_num, o2hb_quorum_region_bitmap)) + goto unlock; + + /* + * A region can be added to the quorum only when it sees all + * live nodes heartbeat on it. In other words, the region has been + * added to all nodes. + */ + if (memcmp(reg->hr_live_node_bitmap, o2hb_live_node_bitmap, + sizeof(o2hb_live_node_bitmap))) + goto unlock; + + printk(KERN_NOTICE "o2hb: Region %s (%s) is now a quorum device\n", + config_item_name(®->hr_item), reg->hr_dev_name); + + set_bit(reg->hr_region_num, o2hb_quorum_region_bitmap); + + /* + * If global heartbeat active, unpin all regions if the + * region count > CUT_OFF + */ + if (bitmap_weight(o2hb_quorum_region_bitmap, + O2NM_MAX_REGIONS) > O2HB_PIN_CUT_OFF) + o2hb_region_unpin(NULL); +unlock: + spin_unlock(&o2hb_live_lock); +} + static int o2hb_check_slot(struct o2hb_region *reg, struct o2hb_disk_slot *slot) { @@ -557,14 +775,23 @@ static int o2hb_check_slot(struct o2hb_region *reg, u64 cputime; unsigned int dead_ms = o2hb_dead_threshold * O2HB_REGION_TIMEOUT_MS; unsigned int slot_dead_ms; + int tmp; + int queued = 0; memcpy(hb_block, slot->ds_raw_block, reg->hr_block_bytes); - /* Is this correct? Do we assume that the node doesn't exist - * if we're not configured for him? */ + /* + * If a node is no longer configured but is still in the livemap, we + * may need to clear that bit from the livemap. + */ node = o2nm_get_node_by_num(slot->ds_node_num); - if (!node) - return 0; + if (!node) { + spin_lock(&o2hb_live_lock); + tmp = test_bit(slot->ds_node_num, o2hb_live_node_bitmap); + spin_unlock(&o2hb_live_lock); + if (!tmp) + return 0; + } if (!o2hb_verify_crc(reg, hb_block)) { /* all paths from here will drop o2hb_live_lock for @@ -617,7 +844,7 @@ static int o2hb_check_slot(struct o2hb_region *reg, "seq %llu last %llu changed %u equal %u\n", slot->ds_node_num, (long long)slot->ds_last_generation, le32_to_cpu(hb_block->hb_cksum), - (unsigned long long)le64_to_cpu(hb_block->hb_seq), + (unsigned long long)le64_to_cpu(hb_block->hb_seq), (unsigned long long)slot->ds_last_time, slot->ds_changed_samples, slot->ds_equal_samples); @@ -631,14 +858,19 @@ fire_callbacks: mlog(ML_HEARTBEAT, "Node %d (id 0x%llx) joined my region\n", slot->ds_node_num, (long long)slot->ds_last_generation); + set_bit(slot->ds_node_num, reg->hr_live_node_bitmap); + /* first on the list generates a callback */ if (list_empty(&o2hb_live_slots[slot->ds_node_num])) { + mlog(ML_HEARTBEAT, "o2hb: Add node %d to live nodes " + "bitmap\n", slot->ds_node_num); set_bit(slot->ds_node_num, o2hb_live_node_bitmap); o2hb_queue_node_event(&event, O2HB_NODE_UP_CB, node, slot->ds_node_num); changed = 1; + queued = 1; } list_add_tail(&slot->ds_live_item, @@ -676,15 +908,21 @@ fire_callbacks: mlog(ML_HEARTBEAT, "Node %d left my region\n", slot->ds_node_num); + clear_bit(slot->ds_node_num, reg->hr_live_node_bitmap); + /* last off the live_slot generates a callback */ list_del_init(&slot->ds_live_item); if (list_empty(&o2hb_live_slots[slot->ds_node_num])) { + mlog(ML_HEARTBEAT, "o2hb: Remove node %d from live " + "nodes bitmap\n", slot->ds_node_num); clear_bit(slot->ds_node_num, o2hb_live_node_bitmap); - o2hb_queue_node_event(&event, O2HB_NODE_DOWN_CB, node, - slot->ds_node_num); + /* node can be null */ + o2hb_queue_node_event(&event, O2HB_NODE_DOWN_CB, + node, slot->ds_node_num); changed = 1; + queued = 1; } /* We don't clear this because the node is still @@ -700,48 +938,50 @@ fire_callbacks: out: spin_unlock(&o2hb_live_lock); - o2hb_run_event_list(&event); + if (queued) + o2hb_run_event_list(&event); - o2nm_node_put(node); + if (node) + o2nm_node_put(node); return changed; } -/* This could be faster if we just implmented a find_last_bit, but I - * don't think the circumstances warrant it. */ -static int o2hb_highest_node(unsigned long *nodes, - int numbits) +static int o2hb_highest_node(unsigned long *nodes, int numbits) { - int highest, node; - - highest = numbits; - node = -1; - while ((node = find_next_bit(nodes, numbits, node + 1)) != -1) { - if (node >= numbits) - break; - - highest = node; - } - - return highest; + return find_last_bit(nodes, numbits); } static int o2hb_do_disk_heartbeat(struct o2hb_region *reg) { - int i, ret, highest_node, change = 0; + int i, ret, highest_node; + int membership_change = 0, own_slot_ok = 0; unsigned long configured_nodes[BITS_TO_LONGS(O2NM_MAX_NODES)]; + unsigned long live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)]; struct o2hb_bio_wait_ctxt write_wc; ret = o2nm_configured_node_map(configured_nodes, sizeof(configured_nodes)); if (ret) { mlog_errno(ret); - return ret; + goto bail; + } + + /* + * If a node is not configured but is in the livemap, we still need + * to read the slot so as to be able to remove it from the livemap. + */ + o2hb_fill_node_map(live_node_bitmap, sizeof(live_node_bitmap)); + i = -1; + while ((i = find_next_bit(live_node_bitmap, + O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) { + set_bit(i, configured_nodes); } highest_node = o2hb_highest_node(configured_nodes, O2NM_MAX_NODES); if (highest_node >= O2NM_MAX_NODES) { - mlog(ML_NOTICE, "ocfs2_heartbeat: no configured nodes found!\n"); - return -EINVAL; + mlog(ML_NOTICE, "o2hb: No configured nodes found!\n"); + ret = -EINVAL; + goto bail; } /* No sense in reading the slots of nodes that don't exist @@ -751,31 +991,27 @@ static int o2hb_do_disk_heartbeat(struct o2hb_region *reg) ret = o2hb_read_slots(reg, highest_node + 1); if (ret < 0) { mlog_errno(ret); - return ret; + goto bail; } /* With an up to date view of the slots, we can check that no * other node has been improperly configured to heartbeat in * our slot. */ - if (!o2hb_check_last_timestamp(reg)) - mlog(ML_ERROR, "Device \"%s\": another node is heartbeating " - "in our slot!\n", reg->hr_dev_name); + own_slot_ok = o2hb_check_own_slot(reg); /* fill in the proper info for our next heartbeat */ o2hb_prepare_block(reg, reg->hr_generation); - /* And fire off the write. Note that we don't wait on this I/O - * until later. */ ret = o2hb_issue_node_write(reg, &write_wc); if (ret < 0) { mlog_errno(ret); - return ret; + goto bail; } i = -1; - while((i = find_next_bit(configured_nodes, O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) { - - change |= o2hb_check_slot(reg, ®->hr_slots[i]); + while((i = find_next_bit(configured_nodes, + O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) { + membership_change |= o2hb_check_slot(reg, ®->hr_slots[i]); } /* @@ -790,18 +1026,39 @@ static int o2hb_do_disk_heartbeat(struct o2hb_region *reg) * disk */ mlog(ML_ERROR, "Write error %d on device \"%s\"\n", write_wc.wc_error, reg->hr_dev_name); - return write_wc.wc_error; + ret = write_wc.wc_error; + goto bail; } - o2hb_arm_write_timeout(reg); + /* Skip disarming the timeout if own slot has stale/bad data */ + if (own_slot_ok) { + o2hb_set_quorum_device(reg); + o2hb_arm_write_timeout(reg); + } +bail: /* let the person who launched us know when things are steady */ - if (!change && (atomic_read(®->hr_steady_iterations) != 0)) { - if (atomic_dec_and_test(®->hr_steady_iterations)) + if (atomic_read(®->hr_steady_iterations) != 0) { + if (!ret && own_slot_ok && !membership_change) { + if (atomic_dec_and_test(®->hr_steady_iterations)) + wake_up(&o2hb_steady_queue); + } + } + + if (atomic_read(®->hr_steady_iterations) != 0) { + if (atomic_dec_and_test(®->hr_unsteady_iterations)) { + printk(KERN_NOTICE "o2hb: Unable to stabilize " + "heartbeart on region %s (%s)\n", + config_item_name(®->hr_item), + reg->hr_dev_name); + atomic_set(®->hr_steady_iterations, 0); + reg->hr_aborted_start = 1; wake_up(&o2hb_steady_queue); + ret = -EIO; + } } - return 0; + return ret; } /* Subtract b from a, storing the result in a. a *must* have a larger @@ -850,30 +1107,33 @@ static int o2hb_thread(void *data) mlog(ML_HEARTBEAT|ML_KTHREAD, "hb thread running\n"); - set_user_nice(current, -20); + set_user_nice(current, MIN_NICE); - while (!kthread_should_stop() && !reg->hr_unclean_stop) { + /* Pin node */ + o2nm_depend_this_node(); + + while (!kthread_should_stop() && + !reg->hr_unclean_stop && !reg->hr_aborted_start) { /* We track the time spent inside - * o2hb_do_disk_heartbeat so that we avoid more then + * o2hb_do_disk_heartbeat so that we avoid more than * hr_timeout_ms between disk writes. On busy systems * this should result in a heartbeat which is less * likely to time itself out. */ do_gettimeofday(&before_hb); - i = 0; - do { - ret = o2hb_do_disk_heartbeat(reg); - } while (ret && ++i < 2); + ret = o2hb_do_disk_heartbeat(reg); do_gettimeofday(&after_hb); elapsed_msec = o2hb_elapsed_msecs(&before_hb, &after_hb); - mlog(0, "start = %lu.%lu, end = %lu.%lu, msec = %u\n", + mlog(ML_HEARTBEAT, + "start = %lu.%lu, end = %lu.%lu, msec = %u\n", before_hb.tv_sec, (unsigned long) before_hb.tv_usec, after_hb.tv_sec, (unsigned long) after_hb.tv_usec, elapsed_msec); - if (elapsed_msec < reg->hr_timeout_ms) { + if (!kthread_should_stop() && + elapsed_msec < reg->hr_timeout_ms) { /* the kthread api has blocked signals for us so no * need to record the return value. */ msleep_interruptible(reg->hr_timeout_ms - elapsed_msec); @@ -890,22 +1150,236 @@ static int o2hb_thread(void *data) * to timeout on this region when we could just as easily * write a clear generation - thus indicating to them that * this node has left this region. - * - * XXX: Should we skip this on unclean_stop? */ - o2hb_prepare_block(reg, 0); - ret = o2hb_issue_node_write(reg, &write_wc); - if (ret == 0) { - o2hb_wait_on_io(reg, &write_wc); - } else { - mlog_errno(ret); + */ + if (!reg->hr_unclean_stop && !reg->hr_aborted_start) { + o2hb_prepare_block(reg, 0); + ret = o2hb_issue_node_write(reg, &write_wc); + if (ret == 0) + o2hb_wait_on_io(reg, &write_wc); + else + mlog_errno(ret); } - mlog(ML_HEARTBEAT|ML_KTHREAD, "hb thread exiting\n"); + /* Unpin node */ + o2nm_undepend_this_node(); + + mlog(ML_HEARTBEAT|ML_KTHREAD, "o2hb thread exiting\n"); return 0; } -void o2hb_init(void) +#ifdef CONFIG_DEBUG_FS +static int o2hb_debug_open(struct inode *inode, struct file *file) +{ + struct o2hb_debug_buf *db = inode->i_private; + struct o2hb_region *reg; + unsigned long map[BITS_TO_LONGS(O2NM_MAX_NODES)]; + unsigned long lts; + char *buf = NULL; + int i = -1; + int out = 0; + + /* max_nodes should be the largest bitmap we pass here */ + BUG_ON(sizeof(map) < db->db_size); + + buf = kmalloc(PAGE_SIZE, GFP_KERNEL); + if (!buf) + goto bail; + + switch (db->db_type) { + case O2HB_DB_TYPE_LIVENODES: + case O2HB_DB_TYPE_LIVEREGIONS: + case O2HB_DB_TYPE_QUORUMREGIONS: + case O2HB_DB_TYPE_FAILEDREGIONS: + spin_lock(&o2hb_live_lock); + memcpy(map, db->db_data, db->db_size); + spin_unlock(&o2hb_live_lock); + break; + + case O2HB_DB_TYPE_REGION_LIVENODES: + spin_lock(&o2hb_live_lock); + reg = (struct o2hb_region *)db->db_data; + memcpy(map, reg->hr_live_node_bitmap, db->db_size); + spin_unlock(&o2hb_live_lock); + break; + + case O2HB_DB_TYPE_REGION_NUMBER: + reg = (struct o2hb_region *)db->db_data; + out += snprintf(buf + out, PAGE_SIZE - out, "%d\n", + reg->hr_region_num); + goto done; + + case O2HB_DB_TYPE_REGION_ELAPSED_TIME: + reg = (struct o2hb_region *)db->db_data; + lts = reg->hr_last_timeout_start; + /* If 0, it has never been set before */ + if (lts) + lts = jiffies_to_msecs(jiffies - lts); + out += snprintf(buf + out, PAGE_SIZE - out, "%lu\n", lts); + goto done; + + case O2HB_DB_TYPE_REGION_PINNED: + reg = (struct o2hb_region *)db->db_data; + out += snprintf(buf + out, PAGE_SIZE - out, "%u\n", + !!reg->hr_item_pinned); + goto done; + + default: + goto done; + } + + while ((i = find_next_bit(map, db->db_len, i + 1)) < db->db_len) + out += snprintf(buf + out, PAGE_SIZE - out, "%d ", i); + out += snprintf(buf + out, PAGE_SIZE - out, "\n"); + +done: + i_size_write(inode, out); + + file->private_data = buf; + + return 0; +bail: + return -ENOMEM; +} + +static int o2hb_debug_release(struct inode *inode, struct file *file) +{ + kfree(file->private_data); + return 0; +} + +static ssize_t o2hb_debug_read(struct file *file, char __user *buf, + size_t nbytes, loff_t *ppos) +{ + return simple_read_from_buffer(buf, nbytes, ppos, file->private_data, + i_size_read(file->f_mapping->host)); +} +#else +static int o2hb_debug_open(struct inode *inode, struct file *file) +{ + return 0; +} +static int o2hb_debug_release(struct inode *inode, struct file *file) +{ + return 0; +} +static ssize_t o2hb_debug_read(struct file *file, char __user *buf, + size_t nbytes, loff_t *ppos) +{ + return 0; +} +#endif /* CONFIG_DEBUG_FS */ + +static const struct file_operations o2hb_debug_fops = { + .open = o2hb_debug_open, + .release = o2hb_debug_release, + .read = o2hb_debug_read, + .llseek = generic_file_llseek, +}; + +void o2hb_exit(void) +{ + kfree(o2hb_db_livenodes); + kfree(o2hb_db_liveregions); + kfree(o2hb_db_quorumregions); + kfree(o2hb_db_failedregions); + debugfs_remove(o2hb_debug_failedregions); + debugfs_remove(o2hb_debug_quorumregions); + debugfs_remove(o2hb_debug_liveregions); + debugfs_remove(o2hb_debug_livenodes); + debugfs_remove(o2hb_debug_dir); +} + +static struct dentry *o2hb_debug_create(const char *name, struct dentry *dir, + struct o2hb_debug_buf **db, int db_len, + int type, int size, int len, void *data) +{ + *db = kmalloc(db_len, GFP_KERNEL); + if (!*db) + return NULL; + + (*db)->db_type = type; + (*db)->db_size = size; + (*db)->db_len = len; + (*db)->db_data = data; + + return debugfs_create_file(name, S_IFREG|S_IRUSR, dir, *db, + &o2hb_debug_fops); +} + +static int o2hb_debug_init(void) +{ + int ret = -ENOMEM; + + o2hb_debug_dir = debugfs_create_dir(O2HB_DEBUG_DIR, NULL); + if (!o2hb_debug_dir) { + mlog_errno(ret); + goto bail; + } + + o2hb_debug_livenodes = o2hb_debug_create(O2HB_DEBUG_LIVENODES, + o2hb_debug_dir, + &o2hb_db_livenodes, + sizeof(*o2hb_db_livenodes), + O2HB_DB_TYPE_LIVENODES, + sizeof(o2hb_live_node_bitmap), + O2NM_MAX_NODES, + o2hb_live_node_bitmap); + if (!o2hb_debug_livenodes) { + mlog_errno(ret); + goto bail; + } + + o2hb_debug_liveregions = o2hb_debug_create(O2HB_DEBUG_LIVEREGIONS, + o2hb_debug_dir, + &o2hb_db_liveregions, + sizeof(*o2hb_db_liveregions), + O2HB_DB_TYPE_LIVEREGIONS, + sizeof(o2hb_live_region_bitmap), + O2NM_MAX_REGIONS, + o2hb_live_region_bitmap); + if (!o2hb_debug_liveregions) { + mlog_errno(ret); + goto bail; + } + + o2hb_debug_quorumregions = + o2hb_debug_create(O2HB_DEBUG_QUORUMREGIONS, + o2hb_debug_dir, + &o2hb_db_quorumregions, + sizeof(*o2hb_db_quorumregions), + O2HB_DB_TYPE_QUORUMREGIONS, + sizeof(o2hb_quorum_region_bitmap), + O2NM_MAX_REGIONS, + o2hb_quorum_region_bitmap); + if (!o2hb_debug_quorumregions) { + mlog_errno(ret); + goto bail; + } + + o2hb_debug_failedregions = + o2hb_debug_create(O2HB_DEBUG_FAILEDREGIONS, + o2hb_debug_dir, + &o2hb_db_failedregions, + sizeof(*o2hb_db_failedregions), + O2HB_DB_TYPE_FAILEDREGIONS, + sizeof(o2hb_failed_region_bitmap), + O2NM_MAX_REGIONS, + o2hb_failed_region_bitmap); + if (!o2hb_debug_failedregions) { + mlog_errno(ret); + goto bail; + } + + ret = 0; +bail: + if (ret) + o2hb_exit(); + + return ret; +} + +int o2hb_init(void) { int i; @@ -918,6 +1392,14 @@ void o2hb_init(void) INIT_LIST_HEAD(&o2hb_node_events); memset(o2hb_live_node_bitmap, 0, sizeof(o2hb_live_node_bitmap)); + memset(o2hb_region_bitmap, 0, sizeof(o2hb_region_bitmap)); + memset(o2hb_live_region_bitmap, 0, sizeof(o2hb_live_region_bitmap)); + memset(o2hb_quorum_region_bitmap, 0, sizeof(o2hb_quorum_region_bitmap)); + memset(o2hb_failed_region_bitmap, 0, sizeof(o2hb_failed_region_bitmap)); + + o2hb_dependent_users = 0; + + return o2hb_debug_init(); } /* if we're already in a callback then we're already serialized by the sem */ @@ -963,8 +1445,9 @@ static void o2hb_region_release(struct config_item *item) struct page *page; struct o2hb_region *reg = to_o2hb_region(item); - if (reg->hr_tmp_block) - kfree(reg->hr_tmp_block); + mlog(ML_HEARTBEAT, "hb region release (%s)\n", reg->hr_dev_name); + + kfree(reg->hr_tmp_block); if (reg->hr_slot_data) { for (i = 0; i < reg->hr_num_pages; i++) { @@ -976,10 +1459,17 @@ static void o2hb_region_release(struct config_item *item) } if (reg->hr_bdev) - blkdev_put(reg->hr_bdev); + blkdev_put(reg->hr_bdev, FMODE_READ|FMODE_WRITE); + + kfree(reg->hr_slots); - if (reg->hr_slots) - kfree(reg->hr_slots); + kfree(reg->hr_db_regnum); + kfree(reg->hr_db_livenodes); + debugfs_remove(reg->hr_debug_livenodes); + debugfs_remove(reg->hr_debug_regnum); + debugfs_remove(reg->hr_debug_elapsed_time); + debugfs_remove(reg->hr_debug_pinned); + debugfs_remove(reg->hr_debug_dir); spin_lock(&o2hb_live_lock); list_del(®->hr_all_item); @@ -1197,8 +1687,6 @@ static int o2hb_populate_slot_data(struct o2hb_region *reg) struct o2hb_disk_slot *slot; struct o2hb_disk_heartbeat_block *hb_block; - mlog_entry_void(); - ret = o2hb_read_slots(reg, reg->hr_blocks); if (ret) { mlog_errno(ret); @@ -1220,7 +1708,6 @@ static int o2hb_populate_slot_data(struct o2hb_region *reg) } out: - mlog_exit(ret); return ret; } @@ -1233,9 +1720,10 @@ static ssize_t o2hb_region_dev_write(struct o2hb_region *reg, long fd; int sectsize; char *p = (char *)page; - struct file *filp = NULL; - struct inode *inode = NULL; + struct fd f; + struct inode *inode; ssize_t ret = -EINVAL; + int live_threshold; if (reg->hr_bdev) goto out; @@ -1252,38 +1740,38 @@ static ssize_t o2hb_region_dev_write(struct o2hb_region *reg, if (fd < 0 || fd >= INT_MAX) goto out; - filp = fget(fd); - if (filp == NULL) + f = fdget(fd); + if (f.file == NULL) goto out; if (reg->hr_blocks == 0 || reg->hr_start_block == 0 || reg->hr_block_bytes == 0) - goto out; + goto out2; - inode = igrab(filp->f_mapping->host); + inode = igrab(f.file->f_mapping->host); if (inode == NULL) - goto out; + goto out2; if (!S_ISBLK(inode->i_mode)) - goto out; + goto out3; - reg->hr_bdev = I_BDEV(filp->f_mapping->host); - ret = blkdev_get(reg->hr_bdev, FMODE_WRITE | FMODE_READ, 0); + reg->hr_bdev = I_BDEV(f.file->f_mapping->host); + ret = blkdev_get(reg->hr_bdev, FMODE_WRITE | FMODE_READ, NULL); if (ret) { reg->hr_bdev = NULL; - goto out; + goto out3; } inode = NULL; bdevname(reg->hr_bdev, reg->hr_dev_name); - sectsize = bdev_hardsect_size(reg->hr_bdev); + sectsize = bdev_logical_block_size(reg->hr_bdev); if (sectsize != reg->hr_block_bytes) { mlog(ML_ERROR, "blocksize %u incorrect for device, expected %d", reg->hr_block_bytes, sectsize); ret = -EINVAL; - goto out; + goto out3; } o2hb_init_region_params(reg); @@ -1297,13 +1785,13 @@ static ssize_t o2hb_region_dev_write(struct o2hb_region *reg, ret = o2hb_map_slot_data(reg); if (ret) { mlog_errno(ret); - goto out; + goto out3; } ret = o2hb_populate_slot_data(reg); if (ret) { mlog_errno(ret); - goto out; + goto out3; } INIT_DELAYED_WORK(®->hr_write_timeout_work, o2hb_write_timeout); @@ -1312,15 +1800,28 @@ static ssize_t o2hb_region_dev_write(struct o2hb_region *reg, * A node is considered live after it has beat LIVE_THRESHOLD * times. We're not steady until we've given them a chance * _after_ our first read. + * The default threshold is bare minimum so as to limit the delay + * during mounts. For global heartbeat, the threshold doubled for the + * first region. */ - atomic_set(®->hr_steady_iterations, O2HB_LIVE_THRESHOLD + 1); + live_threshold = O2HB_LIVE_THRESHOLD; + if (o2hb_global_heartbeat_active()) { + spin_lock(&o2hb_live_lock); + if (bitmap_weight(o2hb_region_bitmap, O2NM_MAX_REGIONS) == 1) + live_threshold <<= 1; + spin_unlock(&o2hb_live_lock); + } + ++live_threshold; + atomic_set(®->hr_steady_iterations, live_threshold); + /* unsteady_iterations is double the steady_iterations */ + atomic_set(®->hr_unsteady_iterations, (live_threshold << 1)); hb_task = kthread_run(o2hb_thread, reg, "o2hb-%s", reg->hr_item.ci_name); if (IS_ERR(hb_task)) { ret = PTR_ERR(hb_task); mlog_errno(ret); - goto out; + goto out3; } spin_lock(&o2hb_live_lock); @@ -1330,20 +1831,20 @@ static ssize_t o2hb_region_dev_write(struct o2hb_region *reg, ret = wait_event_interruptible(o2hb_steady_queue, atomic_read(®->hr_steady_iterations) == 0); if (ret) { - /* We got interrupted (hello ptrace!). Clean up */ - spin_lock(&o2hb_live_lock); - hb_task = reg->hr_task; - reg->hr_task = NULL; - spin_unlock(&o2hb_live_lock); + atomic_set(®->hr_steady_iterations, 0); + reg->hr_aborted_start = 1; + } - if (hb_task) - kthread_stop(hb_task); - goto out; + if (reg->hr_aborted_start) { + ret = -EIO; + goto out3; } /* Ok, we were woken. Make sure it wasn't by drop_item() */ spin_lock(&o2hb_live_lock); hb_task = reg->hr_task; + if (o2hb_global_heartbeat_active()) + set_bit(reg->hr_region_num, o2hb_live_region_bitmap); spin_unlock(&o2hb_live_lock); if (hb_task) @@ -1351,14 +1852,18 @@ static ssize_t o2hb_region_dev_write(struct o2hb_region *reg, else ret = -EIO; + if (hb_task && o2hb_global_heartbeat_active()) + printk(KERN_NOTICE "o2hb: Heartbeat started on region %s (%s)\n", + config_item_name(®->hr_item), reg->hr_dev_name); + +out3: + iput(inode); +out2: + fdput(f); out: - if (filp) - fput(filp); - if (inode) - iput(inode); if (ret < 0) { if (reg->hr_bdev) { - blkdev_put(reg->hr_bdev); + blkdev_put(reg->hr_bdev, FMODE_READ|FMODE_WRITE); reg->hr_bdev = NULL; } } @@ -1489,28 +1994,113 @@ static struct o2hb_heartbeat_group *to_o2hb_heartbeat_group(struct config_group : NULL; } +static int o2hb_debug_region_init(struct o2hb_region *reg, struct dentry *dir) +{ + int ret = -ENOMEM; + + reg->hr_debug_dir = + debugfs_create_dir(config_item_name(®->hr_item), dir); + if (!reg->hr_debug_dir) { + mlog_errno(ret); + goto bail; + } + + reg->hr_debug_livenodes = + o2hb_debug_create(O2HB_DEBUG_LIVENODES, + reg->hr_debug_dir, + &(reg->hr_db_livenodes), + sizeof(*(reg->hr_db_livenodes)), + O2HB_DB_TYPE_REGION_LIVENODES, + sizeof(reg->hr_live_node_bitmap), + O2NM_MAX_NODES, reg); + if (!reg->hr_debug_livenodes) { + mlog_errno(ret); + goto bail; + } + + reg->hr_debug_regnum = + o2hb_debug_create(O2HB_DEBUG_REGION_NUMBER, + reg->hr_debug_dir, + &(reg->hr_db_regnum), + sizeof(*(reg->hr_db_regnum)), + O2HB_DB_TYPE_REGION_NUMBER, + 0, O2NM_MAX_NODES, reg); + if (!reg->hr_debug_regnum) { + mlog_errno(ret); + goto bail; + } + + reg->hr_debug_elapsed_time = + o2hb_debug_create(O2HB_DEBUG_REGION_ELAPSED_TIME, + reg->hr_debug_dir, + &(reg->hr_db_elapsed_time), + sizeof(*(reg->hr_db_elapsed_time)), + O2HB_DB_TYPE_REGION_ELAPSED_TIME, + 0, 0, reg); + if (!reg->hr_debug_elapsed_time) { + mlog_errno(ret); + goto bail; + } + + reg->hr_debug_pinned = + o2hb_debug_create(O2HB_DEBUG_REGION_PINNED, + reg->hr_debug_dir, + &(reg->hr_db_pinned), + sizeof(*(reg->hr_db_pinned)), + O2HB_DB_TYPE_REGION_PINNED, + 0, 0, reg); + if (!reg->hr_debug_pinned) { + mlog_errno(ret); + goto bail; + } + + ret = 0; +bail: + return ret; +} + static struct config_item *o2hb_heartbeat_group_make_item(struct config_group *group, const char *name) { struct o2hb_region *reg = NULL; - struct config_item *ret = NULL; + int ret; reg = kzalloc(sizeof(struct o2hb_region), GFP_KERNEL); if (reg == NULL) - goto out; /* ENOMEM */ - - config_item_init_type_name(®->hr_item, name, &o2hb_region_type); + return ERR_PTR(-ENOMEM); - ret = ®->hr_item; + if (strlen(name) > O2HB_MAX_REGION_NAME_LEN) { + ret = -ENAMETOOLONG; + goto free; + } spin_lock(&o2hb_live_lock); + reg->hr_region_num = 0; + if (o2hb_global_heartbeat_active()) { + reg->hr_region_num = find_first_zero_bit(o2hb_region_bitmap, + O2NM_MAX_REGIONS); + if (reg->hr_region_num >= O2NM_MAX_REGIONS) { + spin_unlock(&o2hb_live_lock); + ret = -EFBIG; + goto free; + } + set_bit(reg->hr_region_num, o2hb_region_bitmap); + } list_add_tail(®->hr_all_item, &o2hb_all_regions); spin_unlock(&o2hb_live_lock); -out: - if (ret == NULL) - kfree(reg); - return ret; + config_item_init_type_name(®->hr_item, name, &o2hb_region_type); + + ret = o2hb_debug_region_init(reg, o2hb_debug_dir); + if (ret) { + config_item_put(®->hr_item); + goto free; + } + + return ®->hr_item; +free: + kfree(reg); + return ERR_PTR(ret); } static void o2hb_heartbeat_group_drop_item(struct config_group *group, @@ -1518,26 +2108,62 @@ static void o2hb_heartbeat_group_drop_item(struct config_group *group, { struct task_struct *hb_task; struct o2hb_region *reg = to_o2hb_region(item); + int quorum_region = 0; /* stop the thread when the user removes the region dir */ spin_lock(&o2hb_live_lock); hb_task = reg->hr_task; reg->hr_task = NULL; + reg->hr_item_dropped = 1; spin_unlock(&o2hb_live_lock); if (hb_task) kthread_stop(hb_task); + if (o2hb_global_heartbeat_active()) { + spin_lock(&o2hb_live_lock); + clear_bit(reg->hr_region_num, o2hb_region_bitmap); + clear_bit(reg->hr_region_num, o2hb_live_region_bitmap); + if (test_bit(reg->hr_region_num, o2hb_quorum_region_bitmap)) + quorum_region = 1; + clear_bit(reg->hr_region_num, o2hb_quorum_region_bitmap); + spin_unlock(&o2hb_live_lock); + printk(KERN_NOTICE "o2hb: Heartbeat %s on region %s (%s)\n", + ((atomic_read(®->hr_steady_iterations) == 0) ? + "stopped" : "start aborted"), config_item_name(item), + reg->hr_dev_name); + } + /* * If we're racing a dev_write(), we need to wake them. They will * check reg->hr_task */ if (atomic_read(®->hr_steady_iterations) != 0) { + reg->hr_aborted_start = 1; atomic_set(®->hr_steady_iterations, 0); wake_up(&o2hb_steady_queue); } config_item_put(item); + + if (!o2hb_global_heartbeat_active() || !quorum_region) + return; + + /* + * If global heartbeat active and there are dependent users, + * pin all regions if quorum region count <= CUT_OFF + */ + spin_lock(&o2hb_live_lock); + + if (!o2hb_dependent_users) + goto unlock; + + if (bitmap_weight(o2hb_quorum_region_bitmap, + O2NM_MAX_REGIONS) <= O2HB_PIN_CUT_OFF) + o2hb_region_pin(NULL); + +unlock: + spin_unlock(&o2hb_live_lock); } struct o2hb_heartbeat_group_attribute { @@ -1597,6 +2223,41 @@ static ssize_t o2hb_heartbeat_group_threshold_store(struct o2hb_heartbeat_group return count; } +static +ssize_t o2hb_heartbeat_group_mode_show(struct o2hb_heartbeat_group *group, + char *page) +{ + return sprintf(page, "%s\n", + o2hb_heartbeat_mode_desc[o2hb_heartbeat_mode]); +} + +static +ssize_t o2hb_heartbeat_group_mode_store(struct o2hb_heartbeat_group *group, + const char *page, size_t count) +{ + unsigned int i; + int ret; + size_t len; + + len = (page[count - 1] == '\n') ? count - 1 : count; + if (!len) + return -EINVAL; + + for (i = 0; i < O2HB_HEARTBEAT_NUM_MODES; ++i) { + if (strnicmp(page, o2hb_heartbeat_mode_desc[i], len)) + continue; + + ret = o2hb_global_heartbeat_mode_set(i); + if (!ret) + printk(KERN_NOTICE "o2hb: Heartbeat mode set to %s\n", + o2hb_heartbeat_mode_desc[i]); + return count; + } + + return -EINVAL; + +} + static struct o2hb_heartbeat_group_attribute o2hb_heartbeat_group_attr_threshold = { .attr = { .ca_owner = THIS_MODULE, .ca_name = "dead_threshold", @@ -1605,12 +2266,21 @@ static struct o2hb_heartbeat_group_attribute o2hb_heartbeat_group_attr_threshold .store = o2hb_heartbeat_group_threshold_store, }; +static struct o2hb_heartbeat_group_attribute o2hb_heartbeat_group_attr_mode = { + .attr = { .ca_owner = THIS_MODULE, + .ca_name = "mode", + .ca_mode = S_IRUGO | S_IWUSR }, + .show = o2hb_heartbeat_group_mode_show, + .store = o2hb_heartbeat_group_mode_store, +}; + static struct configfs_attribute *o2hb_heartbeat_group_attrs[] = { &o2hb_heartbeat_group_attr_threshold.attr, + &o2hb_heartbeat_group_attr_mode.attr, NULL, }; -static struct configfs_item_operations o2hb_hearbeat_group_item_ops = { +static struct configfs_item_operations o2hb_heartbeat_group_item_ops = { .show_attribute = o2hb_heartbeat_group_show, .store_attribute = o2hb_heartbeat_group_store, }; @@ -1622,7 +2292,7 @@ static struct configfs_group_operations o2hb_heartbeat_group_group_ops = { static struct config_item_type o2hb_heartbeat_group_type = { .ct_group_ops = &o2hb_heartbeat_group_group_ops, - .ct_item_ops = &o2hb_hearbeat_group_item_ops, + .ct_item_ops = &o2hb_heartbeat_group_item_ops, .ct_attrs = o2hb_heartbeat_group_attrs, .ct_owner = THIS_MODULE, }; @@ -1654,7 +2324,7 @@ void o2hb_free_hb_set(struct config_group *group) kfree(hs); } -/* hb callback registration and issueing */ +/* hb callback registration and issuing */ static struct o2hb_callback *hbcall_from_type(enum o2hb_callback_type type) { @@ -1679,70 +2349,150 @@ void o2hb_setup_callback(struct o2hb_callback_func *hc, } EXPORT_SYMBOL_GPL(o2hb_setup_callback); -static struct o2hb_region *o2hb_find_region(const char *region_uuid) +/* + * In local heartbeat mode, region_uuid passed matches the dlm domain name. + * In global heartbeat mode, region_uuid passed is NULL. + * + * In local, we only pin the matching region. In global we pin all the active + * regions. + */ +static int o2hb_region_pin(const char *region_uuid) { - struct o2hb_region *p, *reg = NULL; + int ret = 0, found = 0; + struct o2hb_region *reg; + char *uuid; assert_spin_locked(&o2hb_live_lock); - list_for_each_entry(p, &o2hb_all_regions, hr_all_item) { - if (!strcmp(region_uuid, config_item_name(&p->hr_item))) { - reg = p; - break; + list_for_each_entry(reg, &o2hb_all_regions, hr_all_item) { + if (reg->hr_item_dropped) + continue; + + uuid = config_item_name(®->hr_item); + + /* local heartbeat */ + if (region_uuid) { + if (strcmp(region_uuid, uuid)) + continue; + found = 1; } + + if (reg->hr_item_pinned || reg->hr_item_dropped) + goto skip_pin; + + /* Ignore ENOENT only for local hb (userdlm domain) */ + ret = o2nm_depend_item(®->hr_item); + if (!ret) { + mlog(ML_CLUSTER, "Pin region %s\n", uuid); + reg->hr_item_pinned = 1; + } else { + if (ret == -ENOENT && found) + ret = 0; + else { + mlog(ML_ERROR, "Pin region %s fails with %d\n", + uuid, ret); + break; + } + } +skip_pin: + if (found) + break; } - return reg; + return ret; } -static int o2hb_region_get(const char *region_uuid) +/* + * In local heartbeat mode, region_uuid passed matches the dlm domain name. + * In global heartbeat mode, region_uuid passed is NULL. + * + * In local, we only unpin the matching region. In global we unpin all the + * active regions. + */ +static void o2hb_region_unpin(const char *region_uuid) { - int ret = 0; struct o2hb_region *reg; + char *uuid; + int found = 0; - spin_lock(&o2hb_live_lock); - - reg = o2hb_find_region(region_uuid); - if (!reg) - ret = -ENOENT; - spin_unlock(&o2hb_live_lock); - - if (ret) - goto out; + assert_spin_locked(&o2hb_live_lock); - ret = o2nm_depend_this_node(); - if (ret) - goto out; + list_for_each_entry(reg, &o2hb_all_regions, hr_all_item) { + if (reg->hr_item_dropped) + continue; - ret = o2nm_depend_item(®->hr_item); - if (ret) - o2nm_undepend_this_node(); + uuid = config_item_name(®->hr_item); + if (region_uuid) { + if (strcmp(region_uuid, uuid)) + continue; + found = 1; + } -out: - return ret; + if (reg->hr_item_pinned) { + mlog(ML_CLUSTER, "Unpin region %s\n", uuid); + o2nm_undepend_item(®->hr_item); + reg->hr_item_pinned = 0; + } + if (found) + break; + } } -static void o2hb_region_put(const char *region_uuid) +static int o2hb_region_inc_user(const char *region_uuid) { - struct o2hb_region *reg; + int ret = 0; spin_lock(&o2hb_live_lock); - reg = o2hb_find_region(region_uuid); + /* local heartbeat */ + if (!o2hb_global_heartbeat_active()) { + ret = o2hb_region_pin(region_uuid); + goto unlock; + } + + /* + * if global heartbeat active and this is the first dependent user, + * pin all regions if quorum region count <= CUT_OFF + */ + o2hb_dependent_users++; + if (o2hb_dependent_users > 1) + goto unlock; + + if (bitmap_weight(o2hb_quorum_region_bitmap, + O2NM_MAX_REGIONS) <= O2HB_PIN_CUT_OFF) + ret = o2hb_region_pin(NULL); +unlock: spin_unlock(&o2hb_live_lock); + return ret; +} - if (reg) { - o2nm_undepend_item(®->hr_item); - o2nm_undepend_this_node(); +void o2hb_region_dec_user(const char *region_uuid) +{ + spin_lock(&o2hb_live_lock); + + /* local heartbeat */ + if (!o2hb_global_heartbeat_active()) { + o2hb_region_unpin(region_uuid); + goto unlock; } + + /* + * if global heartbeat active and there are no dependent users, + * unpin all quorum regions + */ + o2hb_dependent_users--; + if (!o2hb_dependent_users) + o2hb_region_unpin(NULL); + +unlock: + spin_unlock(&o2hb_live_lock); } int o2hb_register_callback(const char *region_uuid, struct o2hb_callback_func *hc) { - struct o2hb_callback_func *tmp; - struct list_head *iter; + struct o2hb_callback_func *f; struct o2hb_callback *hbcall; int ret; @@ -1756,17 +2506,18 @@ int o2hb_register_callback(const char *region_uuid, } if (region_uuid) { - ret = o2hb_region_get(region_uuid); - if (ret) + ret = o2hb_region_inc_user(region_uuid); + if (ret) { + mlog_errno(ret); goto out; + } } down_write(&o2hb_callback_sem); - list_for_each(iter, &hbcall->list) { - tmp = list_entry(iter, struct o2hb_callback_func, hc_item); - if (hc->hc_priority < tmp->hc_priority) { - list_add_tail(&hc->hc_item, iter); + list_for_each_entry(f, &hbcall->list, hc_item) { + if (hc->hc_priority < f->hc_priority) { + list_add_tail(&hc->hc_item, &f->hc_item); break; } } @@ -1776,7 +2527,7 @@ int o2hb_register_callback(const char *region_uuid, up_write(&o2hb_callback_sem); ret = 0; out: - mlog(ML_HEARTBEAT, "returning %d on behalf of %p for funcs %p\n", + mlog(ML_CLUSTER, "returning %d on behalf of %p for funcs %p\n", ret, __builtin_return_address(0), hc); return ret; } @@ -1787,7 +2538,7 @@ void o2hb_unregister_callback(const char *region_uuid, { BUG_ON(hc->hc_magic != O2HB_CB_MAGIC); - mlog(ML_HEARTBEAT, "on behalf of %p for funcs %p\n", + mlog(ML_CLUSTER, "on behalf of %p for funcs %p\n", __builtin_return_address(0), hc); /* XXX Can this happen _with_ a region reference? */ @@ -1795,7 +2546,7 @@ void o2hb_unregister_callback(const char *region_uuid, return; if (region_uuid) - o2hb_region_put(region_uuid); + o2hb_region_dec_user(region_uuid); down_write(&o2hb_callback_sem); @@ -1872,3 +2623,37 @@ void o2hb_stop_all_regions(void) spin_unlock(&o2hb_live_lock); } EXPORT_SYMBOL_GPL(o2hb_stop_all_regions); + +int o2hb_get_all_regions(char *region_uuids, u8 max_regions) +{ + struct o2hb_region *reg; + int numregs = 0; + char *p; + + spin_lock(&o2hb_live_lock); + + p = region_uuids; + list_for_each_entry(reg, &o2hb_all_regions, hr_all_item) { + if (reg->hr_item_dropped) + continue; + + mlog(0, "Region: %s\n", config_item_name(®->hr_item)); + if (numregs < max_regions) { + memcpy(p, config_item_name(®->hr_item), + O2HB_MAX_REGION_NAME_LEN); + p += O2HB_MAX_REGION_NAME_LEN; + } + numregs++; + } + + spin_unlock(&o2hb_live_lock); + + return numregs; +} +EXPORT_SYMBOL_GPL(o2hb_get_all_regions); + +int o2hb_global_heartbeat_active(void) +{ + return (o2hb_heartbeat_mode == O2HB_HEARTBEAT_GLOBAL); +} +EXPORT_SYMBOL(o2hb_global_heartbeat_active); diff --git a/fs/ocfs2/cluster/heartbeat.h b/fs/ocfs2/cluster/heartbeat.h index e511339886b..00ad8e8fea5 100644 --- a/fs/ocfs2/cluster/heartbeat.h +++ b/fs/ocfs2/cluster/heartbeat.h @@ -31,6 +31,8 @@ #define O2HB_REGION_TIMEOUT_MS 2000 +#define O2HB_MAX_REGION_NAME_LEN 32 + /* number of changes to be seen as live */ #define O2HB_LIVE_THRESHOLD 2 /* number of equal samples to be seen as dead */ @@ -75,10 +77,13 @@ void o2hb_unregister_callback(const char *region_uuid, struct o2hb_callback_func *hc); void o2hb_fill_node_map(unsigned long *map, unsigned bytes); -void o2hb_init(void); +void o2hb_exit(void); +int o2hb_init(void); int o2hb_check_node_heartbeating(u8 node_num); int o2hb_check_node_heartbeating_from_callback(u8 node_num); int o2hb_check_local_node_heartbeating(void); void o2hb_stop_all_regions(void); +int o2hb_get_all_regions(char *region_uuids, u8 numregions); +int o2hb_global_heartbeat_active(void); #endif /* O2CLUSTER_HEARTBEAT_H */ diff --git a/fs/ocfs2/cluster/masklog.c b/fs/ocfs2/cluster/masklog.c index 23c732f2752..07ac24fd925 100644 --- a/fs/ocfs2/cluster/masklog.c +++ b/fs/ocfs2/cluster/masklog.c @@ -30,7 +30,7 @@ struct mlog_bits mlog_and_bits = MLOG_BITS_RHS(MLOG_INITIAL_AND_MASK); EXPORT_SYMBOL_GPL(mlog_and_bits); -struct mlog_bits mlog_not_bits = MLOG_BITS_RHS(MLOG_INITIAL_NOT_MASK); +struct mlog_bits mlog_not_bits = MLOG_BITS_RHS(0); EXPORT_SYMBOL_GPL(mlog_not_bits); static ssize_t mlog_mask_show(u64 mask, char *buf) @@ -80,8 +80,6 @@ struct mlog_attribute { } static struct mlog_attribute mlog_attrs[MLOG_MAX_BITS] = { - define_mask(ENTRY), - define_mask(EXIT), define_mask(TCP), define_mask(MSG), define_mask(SOCKET), @@ -93,22 +91,12 @@ static struct mlog_attribute mlog_attrs[MLOG_MAX_BITS] = { define_mask(DLM_THREAD), define_mask(DLM_MASTER), define_mask(DLM_RECOVERY), - define_mask(AIO), - define_mask(JOURNAL), - define_mask(DISK_ALLOC), - define_mask(SUPER), - define_mask(FILE_IO), - define_mask(EXTENT_MAP), define_mask(DLM_GLUE), - define_mask(BH_IO), - define_mask(UPTODATE), - define_mask(NAMEI), - define_mask(INODE), define_mask(VOTE), - define_mask(DCACHE), define_mask(CONN), define_mask(QUORUM), - define_mask(EXPORT), + define_mask(BASTS), + define_mask(CLUSTER), define_mask(ERROR), define_mask(NOTICE), define_mask(KTHREAD), @@ -132,7 +120,7 @@ static ssize_t mlog_store(struct kobject *obj, struct attribute *attr, return mlog_mask_store(mlog_attr->mask, buf, count); } -static struct sysfs_ops mlog_attr_ops = { +static const struct sysfs_ops mlog_attr_ops = { .show = mlog_show, .store = mlog_store, }; diff --git a/fs/ocfs2/cluster/masklog.h b/fs/ocfs2/cluster/masklog.h index 597e064bb94..2260fb9e650 100644 --- a/fs/ocfs2/cluster/masklog.h +++ b/fs/ocfs2/cluster/masklog.h @@ -48,77 +48,64 @@ * only emit the appropriage printk() when the caller passes in a constant * mask, as is almost always the case. * - * All this bitmask nonsense is hidden from the /proc interface so that Joel - * doesn't have an aneurism. Reading the file gives a straight forward - * indication of which bits are on or off: - * ENTRY off - * EXIT off + * All this bitmask nonsense is managed from the files under + * /sys/fs/o2cb/logmask/. Reading the files gives a straightforward + * indication of which bits are allowed (allow) or denied (off/deny). + * ENTRY deny + * EXIT deny * TCP off * MSG off * SOCKET off - * ERROR off - * NOTICE on + * ERROR allow + * NOTICE allow * * Writing changes the state of a given bit and requires a strictly formatted * single write() call: * - * write(fd, "ENTRY on", 8); + * write(fd, "allow", 5); * - * would turn the entry bit on. "1" is also accepted in the place of "on", and - * "off" and "0" behave as expected. + * Echoing allow/deny/off string into the logmask files can flip the bits + * on or off as expected; here is the bash script for example: * - * Some trivial shell can flip all the bits on or off: + * log_mask="/sys/fs/o2cb/log_mask" + * for node in ENTRY EXIT TCP MSG SOCKET ERROR NOTICE; do + * echo allow >"$log_mask"/"$node" + * done * - * log_mask="/proc/fs/ocfs2_nodemanager/log_mask" - * cat $log_mask | ( - * while read bit status; do - * # $1 is "on" or "off", say - * echo "$bit $1" > $log_mask - * done - * ) + * The debugfs.ocfs2 tool can also flip the bits with the -l option: + * + * debugfs.ocfs2 -l TCP allow */ /* for task_struct */ #include <linux/sched.h> /* bits that are frequently given and infrequently matched in the low word */ -/* NOTE: If you add a flag, you need to also update mlog.c! */ -#define ML_ENTRY 0x0000000000000001ULL /* func call entry */ -#define ML_EXIT 0x0000000000000002ULL /* func call exit */ -#define ML_TCP 0x0000000000000004ULL /* net cluster/tcp.c */ -#define ML_MSG 0x0000000000000008ULL /* net network messages */ -#define ML_SOCKET 0x0000000000000010ULL /* net socket lifetime */ -#define ML_HEARTBEAT 0x0000000000000020ULL /* hb all heartbeat tracking */ -#define ML_HB_BIO 0x0000000000000040ULL /* hb io tracing */ -#define ML_DLMFS 0x0000000000000080ULL /* dlm user dlmfs */ -#define ML_DLM 0x0000000000000100ULL /* dlm general debugging */ -#define ML_DLM_DOMAIN 0x0000000000000200ULL /* dlm domain debugging */ -#define ML_DLM_THREAD 0x0000000000000400ULL /* dlm domain thread */ -#define ML_DLM_MASTER 0x0000000000000800ULL /* dlm master functions */ -#define ML_DLM_RECOVERY 0x0000000000001000ULL /* dlm master functions */ -#define ML_AIO 0x0000000000002000ULL /* ocfs2 aio read and write */ -#define ML_JOURNAL 0x0000000000004000ULL /* ocfs2 journalling functions */ -#define ML_DISK_ALLOC 0x0000000000008000ULL /* ocfs2 disk allocation */ -#define ML_SUPER 0x0000000000010000ULL /* ocfs2 mount / umount */ -#define ML_FILE_IO 0x0000000000020000ULL /* ocfs2 file I/O */ -#define ML_EXTENT_MAP 0x0000000000040000ULL /* ocfs2 extent map caching */ -#define ML_DLM_GLUE 0x0000000000080000ULL /* ocfs2 dlm glue layer */ -#define ML_BH_IO 0x0000000000100000ULL /* ocfs2 buffer I/O */ -#define ML_UPTODATE 0x0000000000200000ULL /* ocfs2 caching sequence #'s */ -#define ML_NAMEI 0x0000000000400000ULL /* ocfs2 directory / namespace */ -#define ML_INODE 0x0000000000800000ULL /* ocfs2 inode manipulation */ -#define ML_VOTE 0x0000000001000000ULL /* ocfs2 node messaging */ -#define ML_DCACHE 0x0000000002000000ULL /* ocfs2 dcache operations */ -#define ML_CONN 0x0000000004000000ULL /* net connection management */ -#define ML_QUORUM 0x0000000008000000ULL /* net connection quorum */ -#define ML_EXPORT 0x0000000010000000ULL /* ocfs2 export operations */ +/* NOTE: If you add a flag, you need to also update masklog.c! */ +#define ML_TCP 0x0000000000000001ULL /* net cluster/tcp.c */ +#define ML_MSG 0x0000000000000002ULL /* net network messages */ +#define ML_SOCKET 0x0000000000000004ULL /* net socket lifetime */ +#define ML_HEARTBEAT 0x0000000000000008ULL /* hb all heartbeat tracking */ +#define ML_HB_BIO 0x0000000000000010ULL /* hb io tracing */ +#define ML_DLMFS 0x0000000000000020ULL /* dlm user dlmfs */ +#define ML_DLM 0x0000000000000040ULL /* dlm general debugging */ +#define ML_DLM_DOMAIN 0x0000000000000080ULL /* dlm domain debugging */ +#define ML_DLM_THREAD 0x0000000000000100ULL /* dlm domain thread */ +#define ML_DLM_MASTER 0x0000000000000200ULL /* dlm master functions */ +#define ML_DLM_RECOVERY 0x0000000000000400ULL /* dlm master functions */ +#define ML_DLM_GLUE 0x0000000000000800ULL /* ocfs2 dlm glue layer */ +#define ML_VOTE 0x0000000000001000ULL /* ocfs2 node messaging */ +#define ML_CONN 0x0000000000002000ULL /* net connection management */ +#define ML_QUORUM 0x0000000000004000ULL /* net connection quorum */ +#define ML_BASTS 0x0000000000008000ULL /* dlmglue asts and basts */ +#define ML_CLUSTER 0x0000000000010000ULL /* cluster stack */ + /* bits that are infrequently given and frequently matched in the high word */ -#define ML_ERROR 0x0000000100000000ULL /* sent to KERN_ERR */ -#define ML_NOTICE 0x0000000200000000ULL /* setn to KERN_NOTICE */ -#define ML_KTHREAD 0x0000000400000000ULL /* kernel thread activity */ +#define ML_ERROR 0x1000000000000000ULL /* sent to KERN_ERR */ +#define ML_NOTICE 0x2000000000000000ULL /* setn to KERN_NOTICE */ +#define ML_KTHREAD 0x4000000000000000ULL /* kernel thread activity */ #define MLOG_INITIAL_AND_MASK (ML_ERROR|ML_NOTICE) -#define MLOG_INITIAL_NOT_MASK (ML_ENTRY|ML_EXIT) #ifndef MLOG_MASK_PREFIX #define MLOG_MASK_PREFIX 0 #endif @@ -192,9 +179,9 @@ extern struct mlog_bits mlog_and_bits, mlog_not_bits; * previous token if args expands to nothing. */ #define __mlog_printk(level, fmt, args...) \ - printk(level "(%u,%lu):%s:%d " fmt, task_pid_nr(current), \ - __mlog_cpu_guess, __PRETTY_FUNCTION__, __LINE__ , \ - ##args) + printk(level "(%s,%u,%lu):%s:%d " fmt, current->comm, \ + task_pid_nr(current), __mlog_cpu_guess, \ + __PRETTY_FUNCTION__, __LINE__ , ##args) #define mlog(mask, fmt, args...) do { \ u64 __m = MLOG_MASK_PREFIX | (mask); \ @@ -212,62 +199,11 @@ extern struct mlog_bits mlog_and_bits, mlog_not_bits; #define mlog_errno(st) do { \ int _st = (st); \ if (_st != -ERESTARTSYS && _st != -EINTR && \ - _st != AOP_TRUNCATED_PAGE && _st != -ENOSPC) \ + _st != AOP_TRUNCATED_PAGE && _st != -ENOSPC && \ + _st != -EDQUOT) \ mlog(ML_ERROR, "status = %lld\n", (long long)_st); \ } while (0) -#if defined(CONFIG_OCFS2_DEBUG_MASKLOG) -#define mlog_entry(fmt, args...) do { \ - mlog(ML_ENTRY, "ENTRY:" fmt , ##args); \ -} while (0) - -#define mlog_entry_void() do { \ - mlog(ML_ENTRY, "ENTRY:\n"); \ -} while (0) - -/* - * We disable this for sparse. - */ -#if !defined(__CHECKER__) -#define mlog_exit(st) do { \ - if (__builtin_types_compatible_p(typeof(st), unsigned long)) \ - mlog(ML_EXIT, "EXIT: %lu\n", (unsigned long) (st)); \ - else if (__builtin_types_compatible_p(typeof(st), signed long)) \ - mlog(ML_EXIT, "EXIT: %ld\n", (signed long) (st)); \ - else if (__builtin_types_compatible_p(typeof(st), unsigned int) \ - || __builtin_types_compatible_p(typeof(st), unsigned short) \ - || __builtin_types_compatible_p(typeof(st), unsigned char)) \ - mlog(ML_EXIT, "EXIT: %u\n", (unsigned int) (st)); \ - else if (__builtin_types_compatible_p(typeof(st), signed int) \ - || __builtin_types_compatible_p(typeof(st), signed short) \ - || __builtin_types_compatible_p(typeof(st), signed char)) \ - mlog(ML_EXIT, "EXIT: %d\n", (signed int) (st)); \ - else if (__builtin_types_compatible_p(typeof(st), long long)) \ - mlog(ML_EXIT, "EXIT: %lld\n", (long long) (st)); \ - else \ - mlog(ML_EXIT, "EXIT: %llu\n", (unsigned long long) (st)); \ -} while (0) -#else -#define mlog_exit(st) do { \ - mlog(ML_EXIT, "EXIT: %lld\n", (long long) (st)); \ -} while (0) -#endif - -#define mlog_exit_ptr(ptr) do { \ - mlog(ML_EXIT, "EXIT: %p\n", ptr); \ -} while (0) - -#define mlog_exit_void() do { \ - mlog(ML_EXIT, "EXIT\n"); \ -} while (0) -#else -#define mlog_entry(...) do { } while (0) -#define mlog_entry_void(...) do { } while (0) -#define mlog_exit(...) do { } while (0) -#define mlog_exit_ptr(...) do { } while (0) -#define mlog_exit_void(...) do { } while (0) -#endif /* defined(CONFIG_OCFS2_DEBUG_MASKLOG) */ - #define mlog_bug_on_msg(cond, fmt, args...) do { \ if (cond) { \ mlog(ML_ERROR, "bug expression: " #cond "\n"); \ diff --git a/fs/ocfs2/cluster/netdebug.c b/fs/ocfs2/cluster/netdebug.c new file mode 100644 index 00000000000..73ba81928bc --- /dev/null +++ b/fs/ocfs2/cluster/netdebug.c @@ -0,0 +1,579 @@ +/* -*- mode: c; c-basic-offset: 8; -*- + * vim: noexpandtab sw=8 ts=8 sts=0: + * + * netdebug.c + * + * debug functionality for o2net + * + * Copyright (C) 2005, 2008 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License as published by the Free Software Foundation; either + * version 2 of the License, or (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + * + */ + +#ifdef CONFIG_DEBUG_FS + +#include <linux/module.h> +#include <linux/types.h> +#include <linux/slab.h> +#include <linux/idr.h> +#include <linux/kref.h> +#include <linux/seq_file.h> +#include <linux/debugfs.h> + +#include <linux/uaccess.h> + +#include "tcp.h" +#include "nodemanager.h" +#define MLOG_MASK_PREFIX ML_TCP +#include "masklog.h" + +#include "tcp_internal.h" + +#define O2NET_DEBUG_DIR "o2net" +#define SC_DEBUG_NAME "sock_containers" +#define NST_DEBUG_NAME "send_tracking" +#define STATS_DEBUG_NAME "stats" +#define NODES_DEBUG_NAME "connected_nodes" + +#define SHOW_SOCK_CONTAINERS 0 +#define SHOW_SOCK_STATS 1 + +static struct dentry *o2net_dentry; +static struct dentry *sc_dentry; +static struct dentry *nst_dentry; +static struct dentry *stats_dentry; +static struct dentry *nodes_dentry; + +static DEFINE_SPINLOCK(o2net_debug_lock); + +static LIST_HEAD(sock_containers); +static LIST_HEAD(send_tracking); + +void o2net_debug_add_nst(struct o2net_send_tracking *nst) +{ + spin_lock(&o2net_debug_lock); + list_add(&nst->st_net_debug_item, &send_tracking); + spin_unlock(&o2net_debug_lock); +} + +void o2net_debug_del_nst(struct o2net_send_tracking *nst) +{ + spin_lock(&o2net_debug_lock); + if (!list_empty(&nst->st_net_debug_item)) + list_del_init(&nst->st_net_debug_item); + spin_unlock(&o2net_debug_lock); +} + +static struct o2net_send_tracking + *next_nst(struct o2net_send_tracking *nst_start) +{ + struct o2net_send_tracking *nst, *ret = NULL; + + assert_spin_locked(&o2net_debug_lock); + + list_for_each_entry(nst, &nst_start->st_net_debug_item, + st_net_debug_item) { + /* discover the head of the list */ + if (&nst->st_net_debug_item == &send_tracking) + break; + + /* use st_task to detect real nsts in the list */ + if (nst->st_task != NULL) { + ret = nst; + break; + } + } + + return ret; +} + +static void *nst_seq_start(struct seq_file *seq, loff_t *pos) +{ + struct o2net_send_tracking *nst, *dummy_nst = seq->private; + + spin_lock(&o2net_debug_lock); + nst = next_nst(dummy_nst); + spin_unlock(&o2net_debug_lock); + + return nst; +} + +static void *nst_seq_next(struct seq_file *seq, void *v, loff_t *pos) +{ + struct o2net_send_tracking *nst, *dummy_nst = seq->private; + + spin_lock(&o2net_debug_lock); + nst = next_nst(dummy_nst); + list_del_init(&dummy_nst->st_net_debug_item); + if (nst) + list_add(&dummy_nst->st_net_debug_item, + &nst->st_net_debug_item); + spin_unlock(&o2net_debug_lock); + + return nst; /* unused, just needs to be null when done */ +} + +static int nst_seq_show(struct seq_file *seq, void *v) +{ + struct o2net_send_tracking *nst, *dummy_nst = seq->private; + ktime_t now; + s64 sock, send, status; + + spin_lock(&o2net_debug_lock); + nst = next_nst(dummy_nst); + if (!nst) + goto out; + + now = ktime_get(); + sock = ktime_to_us(ktime_sub(now, nst->st_sock_time)); + send = ktime_to_us(ktime_sub(now, nst->st_send_time)); + status = ktime_to_us(ktime_sub(now, nst->st_status_time)); + + /* get_task_comm isn't exported. oh well. */ + seq_printf(seq, "%p:\n" + " pid: %lu\n" + " tgid: %lu\n" + " process name: %s\n" + " node: %u\n" + " sc: %p\n" + " message id: %d\n" + " message type: %u\n" + " message key: 0x%08x\n" + " sock acquiry: %lld usecs ago\n" + " send start: %lld usecs ago\n" + " wait start: %lld usecs ago\n", + nst, (unsigned long)task_pid_nr(nst->st_task), + (unsigned long)nst->st_task->tgid, + nst->st_task->comm, nst->st_node, + nst->st_sc, nst->st_id, nst->st_msg_type, + nst->st_msg_key, + (long long)sock, + (long long)send, + (long long)status); + +out: + spin_unlock(&o2net_debug_lock); + + return 0; +} + +static void nst_seq_stop(struct seq_file *seq, void *v) +{ +} + +static const struct seq_operations nst_seq_ops = { + .start = nst_seq_start, + .next = nst_seq_next, + .stop = nst_seq_stop, + .show = nst_seq_show, +}; + +static int nst_fop_open(struct inode *inode, struct file *file) +{ + struct o2net_send_tracking *dummy_nst; + struct seq_file *seq; + int ret; + + dummy_nst = kmalloc(sizeof(struct o2net_send_tracking), GFP_KERNEL); + if (dummy_nst == NULL) { + ret = -ENOMEM; + goto out; + } + dummy_nst->st_task = NULL; + + ret = seq_open(file, &nst_seq_ops); + if (ret) + goto out; + + seq = file->private_data; + seq->private = dummy_nst; + o2net_debug_add_nst(dummy_nst); + + dummy_nst = NULL; + +out: + kfree(dummy_nst); + return ret; +} + +static int nst_fop_release(struct inode *inode, struct file *file) +{ + struct seq_file *seq = file->private_data; + struct o2net_send_tracking *dummy_nst = seq->private; + + o2net_debug_del_nst(dummy_nst); + return seq_release_private(inode, file); +} + +static const struct file_operations nst_seq_fops = { + .open = nst_fop_open, + .read = seq_read, + .llseek = seq_lseek, + .release = nst_fop_release, +}; + +void o2net_debug_add_sc(struct o2net_sock_container *sc) +{ + spin_lock(&o2net_debug_lock); + list_add(&sc->sc_net_debug_item, &sock_containers); + spin_unlock(&o2net_debug_lock); +} + +void o2net_debug_del_sc(struct o2net_sock_container *sc) +{ + spin_lock(&o2net_debug_lock); + list_del_init(&sc->sc_net_debug_item); + spin_unlock(&o2net_debug_lock); +} + +struct o2net_sock_debug { + int dbg_ctxt; + struct o2net_sock_container *dbg_sock; +}; + +static struct o2net_sock_container + *next_sc(struct o2net_sock_container *sc_start) +{ + struct o2net_sock_container *sc, *ret = NULL; + + assert_spin_locked(&o2net_debug_lock); + + list_for_each_entry(sc, &sc_start->sc_net_debug_item, + sc_net_debug_item) { + /* discover the head of the list miscast as a sc */ + if (&sc->sc_net_debug_item == &sock_containers) + break; + + /* use sc_page to detect real scs in the list */ + if (sc->sc_page != NULL) { + ret = sc; + break; + } + } + + return ret; +} + +static void *sc_seq_start(struct seq_file *seq, loff_t *pos) +{ + struct o2net_sock_debug *sd = seq->private; + struct o2net_sock_container *sc, *dummy_sc = sd->dbg_sock; + + spin_lock(&o2net_debug_lock); + sc = next_sc(dummy_sc); + spin_unlock(&o2net_debug_lock); + + return sc; +} + +static void *sc_seq_next(struct seq_file *seq, void *v, loff_t *pos) +{ + struct o2net_sock_debug *sd = seq->private; + struct o2net_sock_container *sc, *dummy_sc = sd->dbg_sock; + + spin_lock(&o2net_debug_lock); + sc = next_sc(dummy_sc); + list_del_init(&dummy_sc->sc_net_debug_item); + if (sc) + list_add(&dummy_sc->sc_net_debug_item, &sc->sc_net_debug_item); + spin_unlock(&o2net_debug_lock); + + return sc; /* unused, just needs to be null when done */ +} + +#ifdef CONFIG_OCFS2_FS_STATS +# define sc_send_count(_s) ((_s)->sc_send_count) +# define sc_recv_count(_s) ((_s)->sc_recv_count) +# define sc_tv_acquiry_total_ns(_s) (ktime_to_ns((_s)->sc_tv_acquiry_total)) +# define sc_tv_send_total_ns(_s) (ktime_to_ns((_s)->sc_tv_send_total)) +# define sc_tv_status_total_ns(_s) (ktime_to_ns((_s)->sc_tv_status_total)) +# define sc_tv_process_total_ns(_s) (ktime_to_ns((_s)->sc_tv_process_total)) +#else +# define sc_send_count(_s) (0U) +# define sc_recv_count(_s) (0U) +# define sc_tv_acquiry_total_ns(_s) (0LL) +# define sc_tv_send_total_ns(_s) (0LL) +# define sc_tv_status_total_ns(_s) (0LL) +# define sc_tv_process_total_ns(_s) (0LL) +#endif + +/* So that debugfs.ocfs2 can determine which format is being used */ +#define O2NET_STATS_STR_VERSION 1 +static void sc_show_sock_stats(struct seq_file *seq, + struct o2net_sock_container *sc) +{ + if (!sc) + return; + + seq_printf(seq, "%d,%u,%lu,%lld,%lld,%lld,%lu,%lld\n", O2NET_STATS_STR_VERSION, + sc->sc_node->nd_num, (unsigned long)sc_send_count(sc), + (long long)sc_tv_acquiry_total_ns(sc), + (long long)sc_tv_send_total_ns(sc), + (long long)sc_tv_status_total_ns(sc), + (unsigned long)sc_recv_count(sc), + (long long)sc_tv_process_total_ns(sc)); +} + +static void sc_show_sock_container(struct seq_file *seq, + struct o2net_sock_container *sc) +{ + struct inet_sock *inet = NULL; + __be32 saddr = 0, daddr = 0; + __be16 sport = 0, dport = 0; + + if (!sc) + return; + + if (sc->sc_sock) { + inet = inet_sk(sc->sc_sock->sk); + /* the stack's structs aren't sparse endian clean */ + saddr = (__force __be32)inet->inet_saddr; + daddr = (__force __be32)inet->inet_daddr; + sport = (__force __be16)inet->inet_sport; + dport = (__force __be16)inet->inet_dport; + } + + /* XXX sigh, inet-> doesn't have sparse annotation so any + * use of it here generates a warning with -Wbitwise */ + seq_printf(seq, "%p:\n" + " krefs: %d\n" + " sock: %pI4:%u -> " + "%pI4:%u\n" + " remote node: %s\n" + " page off: %zu\n" + " handshake ok: %u\n" + " timer: %lld usecs\n" + " data ready: %lld usecs\n" + " advance start: %lld usecs\n" + " advance stop: %lld usecs\n" + " func start: %lld usecs\n" + " func stop: %lld usecs\n" + " func key: 0x%08x\n" + " func type: %u\n", + sc, + atomic_read(&sc->sc_kref.refcount), + &saddr, inet ? ntohs(sport) : 0, + &daddr, inet ? ntohs(dport) : 0, + sc->sc_node->nd_name, + sc->sc_page_off, + sc->sc_handshake_ok, + (long long)ktime_to_us(sc->sc_tv_timer), + (long long)ktime_to_us(sc->sc_tv_data_ready), + (long long)ktime_to_us(sc->sc_tv_advance_start), + (long long)ktime_to_us(sc->sc_tv_advance_stop), + (long long)ktime_to_us(sc->sc_tv_func_start), + (long long)ktime_to_us(sc->sc_tv_func_stop), + sc->sc_msg_key, + sc->sc_msg_type); +} + +static int sc_seq_show(struct seq_file *seq, void *v) +{ + struct o2net_sock_debug *sd = seq->private; + struct o2net_sock_container *sc, *dummy_sc = sd->dbg_sock; + + spin_lock(&o2net_debug_lock); + sc = next_sc(dummy_sc); + + if (sc) { + if (sd->dbg_ctxt == SHOW_SOCK_CONTAINERS) + sc_show_sock_container(seq, sc); + else + sc_show_sock_stats(seq, sc); + } + + spin_unlock(&o2net_debug_lock); + + return 0; +} + +static void sc_seq_stop(struct seq_file *seq, void *v) +{ +} + +static const struct seq_operations sc_seq_ops = { + .start = sc_seq_start, + .next = sc_seq_next, + .stop = sc_seq_stop, + .show = sc_seq_show, +}; + +static int sc_common_open(struct file *file, struct o2net_sock_debug *sd) +{ + struct o2net_sock_container *dummy_sc; + struct seq_file *seq; + int ret; + + dummy_sc = kmalloc(sizeof(struct o2net_sock_container), GFP_KERNEL); + if (dummy_sc == NULL) { + ret = -ENOMEM; + goto out; + } + dummy_sc->sc_page = NULL; + + ret = seq_open(file, &sc_seq_ops); + if (ret) + goto out; + + seq = file->private_data; + seq->private = sd; + sd->dbg_sock = dummy_sc; + o2net_debug_add_sc(dummy_sc); + + dummy_sc = NULL; + +out: + kfree(dummy_sc); + return ret; +} + +static int sc_fop_release(struct inode *inode, struct file *file) +{ + struct seq_file *seq = file->private_data; + struct o2net_sock_debug *sd = seq->private; + struct o2net_sock_container *dummy_sc = sd->dbg_sock; + + o2net_debug_del_sc(dummy_sc); + return seq_release_private(inode, file); +} + +static int stats_fop_open(struct inode *inode, struct file *file) +{ + struct o2net_sock_debug *sd; + + sd = kmalloc(sizeof(struct o2net_sock_debug), GFP_KERNEL); + if (sd == NULL) + return -ENOMEM; + + sd->dbg_ctxt = SHOW_SOCK_STATS; + sd->dbg_sock = NULL; + + return sc_common_open(file, sd); +} + +static const struct file_operations stats_seq_fops = { + .open = stats_fop_open, + .read = seq_read, + .llseek = seq_lseek, + .release = sc_fop_release, +}; + +static int sc_fop_open(struct inode *inode, struct file *file) +{ + struct o2net_sock_debug *sd; + + sd = kmalloc(sizeof(struct o2net_sock_debug), GFP_KERNEL); + if (sd == NULL) + return -ENOMEM; + + sd->dbg_ctxt = SHOW_SOCK_CONTAINERS; + sd->dbg_sock = NULL; + + return sc_common_open(file, sd); +} + +static const struct file_operations sc_seq_fops = { + .open = sc_fop_open, + .read = seq_read, + .llseek = seq_lseek, + .release = sc_fop_release, +}; + +static int o2net_fill_bitmap(char *buf, int len) +{ + unsigned long map[BITS_TO_LONGS(O2NM_MAX_NODES)]; + int i = -1, out = 0; + + o2net_fill_node_map(map, sizeof(map)); + + while ((i = find_next_bit(map, O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) + out += snprintf(buf + out, PAGE_SIZE - out, "%d ", i); + out += snprintf(buf + out, PAGE_SIZE - out, "\n"); + + return out; +} + +static int nodes_fop_open(struct inode *inode, struct file *file) +{ + char *buf; + + buf = kmalloc(PAGE_SIZE, GFP_KERNEL); + if (!buf) + return -ENOMEM; + + i_size_write(inode, o2net_fill_bitmap(buf, PAGE_SIZE)); + + file->private_data = buf; + + return 0; +} + +static int o2net_debug_release(struct inode *inode, struct file *file) +{ + kfree(file->private_data); + return 0; +} + +static ssize_t o2net_debug_read(struct file *file, char __user *buf, + size_t nbytes, loff_t *ppos) +{ + return simple_read_from_buffer(buf, nbytes, ppos, file->private_data, + i_size_read(file->f_mapping->host)); +} + +static const struct file_operations nodes_fops = { + .open = nodes_fop_open, + .release = o2net_debug_release, + .read = o2net_debug_read, + .llseek = generic_file_llseek, +}; + +void o2net_debugfs_exit(void) +{ + debugfs_remove(nodes_dentry); + debugfs_remove(stats_dentry); + debugfs_remove(sc_dentry); + debugfs_remove(nst_dentry); + debugfs_remove(o2net_dentry); +} + +int o2net_debugfs_init(void) +{ + umode_t mode = S_IFREG|S_IRUSR; + + o2net_dentry = debugfs_create_dir(O2NET_DEBUG_DIR, NULL); + if (o2net_dentry) + nst_dentry = debugfs_create_file(NST_DEBUG_NAME, mode, + o2net_dentry, NULL, &nst_seq_fops); + if (nst_dentry) + sc_dentry = debugfs_create_file(SC_DEBUG_NAME, mode, + o2net_dentry, NULL, &sc_seq_fops); + if (sc_dentry) + stats_dentry = debugfs_create_file(STATS_DEBUG_NAME, mode, + o2net_dentry, NULL, &stats_seq_fops); + if (stats_dentry) + nodes_dentry = debugfs_create_file(NODES_DEBUG_NAME, mode, + o2net_dentry, NULL, &nodes_fops); + if (nodes_dentry) + return 0; + + o2net_debugfs_exit(); + mlog_errno(-ENOMEM); + return -ENOMEM; +} + +#endif /* CONFIG_DEBUG_FS */ diff --git a/fs/ocfs2/cluster/nodemanager.c b/fs/ocfs2/cluster/nodemanager.c index 709fba25bf7..441c84e169e 100644 --- a/fs/ocfs2/cluster/nodemanager.c +++ b/fs/ocfs2/cluster/nodemanager.c @@ -19,9 +19,9 @@ * Boston, MA 021110-1307, USA. */ +#include <linux/slab.h> #include <linux/kernel.h> #include <linux/module.h> -#include <linux/sysctl.h> #include <linux/configfs.h> #include "tcp.h" @@ -29,73 +29,17 @@ #include "heartbeat.h" #include "masklog.h" #include "sys.h" -#include "ver.h" /* for now we operate under the assertion that there can be only one * cluster active at a time. Changing this will require trickling * cluster references throughout where nodes are looked up */ struct o2nm_cluster *o2nm_single_cluster = NULL; -#define OCFS2_MAX_HB_CTL_PATH 256 -static char ocfs2_hb_ctl_path[OCFS2_MAX_HB_CTL_PATH] = "/sbin/ocfs2_hb_ctl"; - -static ctl_table ocfs2_nm_table[] = { - { - .ctl_name = 1, - .procname = "hb_ctl_path", - .data = ocfs2_hb_ctl_path, - .maxlen = OCFS2_MAX_HB_CTL_PATH, - .mode = 0644, - .proc_handler = &proc_dostring, - .strategy = &sysctl_string, - }, - { .ctl_name = 0 } -}; - -static ctl_table ocfs2_mod_table[] = { - { - .ctl_name = FS_OCFS2_NM, - .procname = "nm", - .data = NULL, - .maxlen = 0, - .mode = 0555, - .child = ocfs2_nm_table - }, - { .ctl_name = 0} -}; - -static ctl_table ocfs2_kern_table[] = { - { - .ctl_name = FS_OCFS2, - .procname = "ocfs2", - .data = NULL, - .maxlen = 0, - .mode = 0555, - .child = ocfs2_mod_table - }, - { .ctl_name = 0} -}; - -static ctl_table ocfs2_root_table[] = { - { - .ctl_name = CTL_FS, - .procname = "fs", - .data = NULL, - .maxlen = 0, - .mode = 0555, - .child = ocfs2_kern_table - }, - { .ctl_name = 0 } +char *o2nm_fence_method_desc[O2NM_FENCE_METHODS] = { + "reset", /* O2NM_FENCE_RESET */ + "panic", /* O2NM_FENCE_PANIC */ }; -static struct ctl_table_header *ocfs2_table_header = NULL; - -const char *o2nm_get_hb_ctl_path(void) -{ - return ocfs2_hb_ctl_path; -} -EXPORT_SYMBOL_GPL(o2nm_get_hb_ctl_path); - struct o2nm_node *o2nm_get_node_by_num(u8 node_num) { struct o2nm_node *node = NULL; @@ -310,7 +254,7 @@ static ssize_t o2nm_node_ipv4_port_write(struct o2nm_node *node, static ssize_t o2nm_node_ipv4_address_read(struct o2nm_node *node, char *page) { - return sprintf(page, "%u.%u.%u.%u\n", NIPQUAD(node->nd_ipv4_address)); + return sprintf(page, "%pI4\n", &node->nd_ipv4_address); } static ssize_t o2nm_node_ipv4_address_write(struct o2nm_node *node, @@ -639,6 +583,43 @@ static ssize_t o2nm_cluster_attr_reconnect_delay_ms_write( return o2nm_cluster_attr_write(page, count, &cluster->cl_reconnect_delay_ms); } + +static ssize_t o2nm_cluster_attr_fence_method_read( + struct o2nm_cluster *cluster, char *page) +{ + ssize_t ret = 0; + + if (cluster) + ret = sprintf(page, "%s\n", + o2nm_fence_method_desc[cluster->cl_fence_method]); + return ret; +} + +static ssize_t o2nm_cluster_attr_fence_method_write( + struct o2nm_cluster *cluster, const char *page, size_t count) +{ + unsigned int i; + + if (page[count - 1] != '\n') + goto bail; + + for (i = 0; i < O2NM_FENCE_METHODS; ++i) { + if (count != strlen(o2nm_fence_method_desc[i]) + 1) + continue; + if (strncasecmp(page, o2nm_fence_method_desc[i], count - 1)) + continue; + if (cluster->cl_fence_method != i) { + printk(KERN_INFO "ocfs2: Changing fence method to %s\n", + o2nm_fence_method_desc[i]); + cluster->cl_fence_method = i; + } + return count; + } + +bail: + return -EINVAL; +} + static struct o2nm_cluster_attribute o2nm_cluster_attr_idle_timeout_ms = { .attr = { .ca_owner = THIS_MODULE, .ca_name = "idle_timeout_ms", @@ -663,10 +644,19 @@ static struct o2nm_cluster_attribute o2nm_cluster_attr_reconnect_delay_ms = { .store = o2nm_cluster_attr_reconnect_delay_ms_write, }; +static struct o2nm_cluster_attribute o2nm_cluster_attr_fence_method = { + .attr = { .ca_owner = THIS_MODULE, + .ca_name = "fence_method", + .ca_mode = S_IRUGO | S_IWUSR }, + .show = o2nm_cluster_attr_fence_method_read, + .store = o2nm_cluster_attr_fence_method_write, +}; + static struct configfs_attribute *o2nm_cluster_attrs[] = { &o2nm_cluster_attr_idle_timeout_ms.attr, &o2nm_cluster_attr_keepalive_delay_ms.attr, &o2nm_cluster_attr_reconnect_delay_ms.attr, + &o2nm_cluster_attr_fence_method.attr, NULL, }; static ssize_t o2nm_cluster_show(struct config_item *item, @@ -708,26 +698,21 @@ static struct config_item *o2nm_node_group_make_item(struct config_group *group, const char *name) { struct o2nm_node *node = NULL; - struct config_item *ret = NULL; if (strlen(name) > O2NM_MAX_NAME_LEN) - goto out; /* ENAMETOOLONG */ + return ERR_PTR(-ENAMETOOLONG); node = kzalloc(sizeof(struct o2nm_node), GFP_KERNEL); if (node == NULL) - goto out; /* ENOMEM */ + return ERR_PTR(-ENOMEM); strcpy(node->nd_name, name); /* use item.ci_namebuf instead? */ config_item_init_type_name(&node->nd_item, name, &o2nm_node_type); spin_lock_init(&node->nd_lock); - ret = &node->nd_item; + mlog(ML_CLUSTER, "o2nm: Registering node %s\n", name); -out: - if (ret == NULL) - kfree(node); - - return ret; + return &node->nd_item; } static void o2nm_node_group_drop_item(struct config_group *group, @@ -760,6 +745,9 @@ static void o2nm_node_group_drop_item(struct config_group *group, } write_unlock(&cluster->cl_nodes_lock); + mlog(ML_CLUSTER, "o2nm: Unregistered node %s\n", + config_item_name(&node->nd_item)); + config_item_put(item); } @@ -822,7 +810,7 @@ static struct config_group *o2nm_cluster_group_make_group(struct config_group *g /* this runs under the parent dir's i_mutex; there can be only * one caller in here at a time */ if (o2nm_single_cluster) - goto out; /* ENOSPC */ + return ERR_PTR(-ENOSPC); cluster = kzalloc(sizeof(struct o2nm_cluster), GFP_KERNEL); ns = kzalloc(sizeof(struct o2nm_node_group), GFP_KERNEL); @@ -845,6 +833,7 @@ static struct config_group *o2nm_cluster_group_make_group(struct config_group *g cluster->cl_reconnect_delay_ms = O2NET_RECONNECT_DELAY_MS_DEFAULT; cluster->cl_idle_timeout_ms = O2NET_IDLE_TIMEOUT_MS_DEFAULT; cluster->cl_keepalive_delay_ms = O2NET_KEEPALIVE_DELAY_MS_DEFAULT; + cluster->cl_fence_method = O2NM_FENCE_RESET; ret = &cluster->cl_group; o2nm_single_cluster = cluster; @@ -855,6 +844,7 @@ out: kfree(ns); o2hb_free_hb_set(o2hb_group); kfree(defs); + ret = ERR_PTR(-ENOMEM); } return ret; @@ -941,36 +931,30 @@ void o2nm_undepend_this_node(void) static void __exit exit_o2nm(void) { - if (ocfs2_table_header) - unregister_sysctl_table(ocfs2_table_header); - /* XXX sync with hb callbacks and shut down hb? */ o2net_unregister_hb_callbacks(); configfs_unregister_subsystem(&o2nm_cluster_group.cs_subsys); o2cb_sys_shutdown(); o2net_exit(); + o2hb_exit(); } static int __init init_o2nm(void) { int ret = -1; - cluster_print_version(); - - o2hb_init(); - o2net_init(); + ret = o2hb_init(); + if (ret) + goto out; - ocfs2_table_header = register_sysctl_table(ocfs2_root_table); - if (!ocfs2_table_header) { - printk(KERN_ERR "nodemanager: unable to register sysctl\n"); - ret = -ENOMEM; /* or something. */ - goto out_o2net; - } + ret = o2net_init(); + if (ret) + goto out_o2hb; ret = o2net_register_hb_callbacks(); if (ret) - goto out_sysctl; + goto out_o2net; config_group_init(&o2nm_cluster_group.cs_subsys.su_group); mutex_init(&o2nm_cluster_group.cs_subsys.su_mutex); @@ -987,16 +971,17 @@ static int __init init_o2nm(void) configfs_unregister_subsystem(&o2nm_cluster_group.cs_subsys); out_callbacks: o2net_unregister_hb_callbacks(); -out_sysctl: - unregister_sysctl_table(ocfs2_table_header); out_o2net: o2net_exit(); +out_o2hb: + o2hb_exit(); out: return ret; } MODULE_AUTHOR("Oracle"); MODULE_LICENSE("GPL"); +MODULE_DESCRIPTION("OCFS2 cluster management"); module_init(init_o2nm) module_exit(exit_o2nm) diff --git a/fs/ocfs2/cluster/nodemanager.h b/fs/ocfs2/cluster/nodemanager.h index 7c860361b8d..09ea2d388bb 100644 --- a/fs/ocfs2/cluster/nodemanager.h +++ b/fs/ocfs2/cluster/nodemanager.h @@ -33,9 +33,11 @@ #include <linux/configfs.h> #include <linux/rbtree.h> -#define FS_OCFS2_NM 1 - -const char *o2nm_get_hb_ctl_path(void); +enum o2nm_fence_method { + O2NM_FENCE_RESET = 0, + O2NM_FENCE_PANIC, + O2NM_FENCE_METHODS, /* Number of fence methods */ +}; struct o2nm_node { spinlock_t nd_lock; @@ -62,6 +64,7 @@ struct o2nm_cluster { unsigned int cl_idle_timeout_ms; unsigned int cl_keepalive_delay_ms; unsigned int cl_reconnect_delay_ms; + enum o2nm_fence_method cl_fence_method; /* this bitmap is part of a hack for disk bitmap.. will go eventually. - zab */ unsigned long cl_nodes_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)]; diff --git a/fs/ocfs2/cluster/ocfs2_nodemanager.h b/fs/ocfs2/cluster/ocfs2_nodemanager.h index 5b9854bad57..49b594325be 100644 --- a/fs/ocfs2/cluster/ocfs2_nodemanager.h +++ b/fs/ocfs2/cluster/ocfs2_nodemanager.h @@ -36,4 +36,10 @@ /* host name, group name, cluster name all 64 bytes */ #define O2NM_MAX_NAME_LEN 64 // __NEW_UTS_LEN +/* + * Maximum number of global heartbeat regions allowed. + * **CAUTION** Changing this number will break dlm compatibility. + */ +#define O2NM_MAX_REGIONS 32 + #endif /* _OCFS2_NODEMANAGER_H */ diff --git a/fs/ocfs2/cluster/quorum.c b/fs/ocfs2/cluster/quorum.c index bbacf7da48a..1ec141e758d 100644 --- a/fs/ocfs2/cluster/quorum.c +++ b/fs/ocfs2/cluster/quorum.c @@ -44,7 +44,6 @@ * and if they're the last, they fire off the decision. */ #include <linux/kernel.h> -#include <linux/slab.h> #include <linux/workqueue.h> #include <linux/reboot.h> @@ -74,11 +73,23 @@ static void o2quo_fence_self(void) * threads can still schedule, etc, etc */ o2hb_stop_all_regions(); - printk("ocfs2 is very sorry to be fencing this system by restarting\n"); - emergency_restart(); + switch (o2nm_single_cluster->cl_fence_method) { + case O2NM_FENCE_PANIC: + panic("*** ocfs2 is very sorry to be fencing this system by " + "panicing ***\n"); + break; + default: + WARN_ON(o2nm_single_cluster->cl_fence_method >= + O2NM_FENCE_METHODS); + case O2NM_FENCE_RESET: + printk(KERN_ERR "*** ocfs2 is very sorry to be fencing this " + "system by restarting ***\n"); + emergency_restart(); + break; + }; } -/* Indicate that a timeout occured on a hearbeat region write. The +/* Indicate that a timeout occurred on a hearbeat region write. The * other nodes in the cluster may consider us dead at that time so we * want to "fence" ourselves so that we don't scribble on the disk * after they think they've recovered us. This can't solve all @@ -250,10 +261,10 @@ void o2quo_hb_still_up(u8 node) spin_unlock(&qs->qs_lock); } -/* This is analagous to hb_up. as a node's connection comes up we delay the +/* This is analogous to hb_up. as a node's connection comes up we delay the * quorum decision until we see it heartbeating. the hold will be droped in * hb_up or hb_down. it might be perpetuated by con_err until hb_down. if - * it's already heartbeating we we might be dropping a hold that conn_up got. + * it's already heartbeating we might be dropping a hold that conn_up got. * */ void o2quo_conn_up(u8 node) { @@ -314,5 +325,7 @@ void o2quo_init(void) void o2quo_exit(void) { - flush_scheduled_work(); + struct o2quo_state *qs = &o2quo_state; + + flush_work(&qs->qs_work); } diff --git a/fs/ocfs2/cluster/sys.c b/fs/ocfs2/cluster/sys.c index 0c095ce7723..b7f57271d49 100644 --- a/fs/ocfs2/cluster/sys.c +++ b/fs/ocfs2/cluster/sys.c @@ -41,7 +41,7 @@ static ssize_t version_show(struct kobject *kobj, struct kobj_attribute *attr, return snprintf(buf, PAGE_SIZE, "%u\n", O2NM_API_VERSION); } static struct kobj_attribute attr_version = - __ATTR(interface_revision, S_IFREG | S_IRUGO, version_show, NULL); + __ATTR(interface_revision, S_IRUGO, version_show, NULL); static struct attribute *o2cb_attrs[] = { &attr_version.attr, @@ -64,7 +64,7 @@ int o2cb_sys_init(void) { int ret; - o2cb_kset = kset_create_and_add("o2cb", NULL, NULL); + o2cb_kset = kset_create_and_add("o2cb", NULL, fs_kobj); if (!o2cb_kset) return -ENOMEM; diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c index b8057c51b20..681691bc233 100644 --- a/fs/ocfs2/cluster/tcp.c +++ b/fs/ocfs2/cluster/tcp.c @@ -59,6 +59,7 @@ #include <linux/idr.h> #include <linux/kref.h> #include <linux/net.h> +#include <linux/export.h> #include <net/tcp.h> #include <asm/uaccess.h> @@ -72,9 +73,9 @@ #include "tcp_internal.h" -#define SC_NODEF_FMT "node %s (num %u) at %u.%u.%u.%u:%u" +#define SC_NODEF_FMT "node %s (num %u) at %pI4:%u" #define SC_NODEF_ARGS(sc) sc->sc_node->nd_name, sc->sc_node->nd_num, \ - NIPQUAD(sc->sc_node->nd_ipv4_address), \ + &sc->sc_node->nd_ipv4_address, \ ntohs(sc->sc_node->nd_ipv4_port) /* @@ -107,7 +108,7 @@ static struct rb_root o2net_handler_tree = RB_ROOT; static struct o2net_node o2net_nodes[O2NM_MAX_NODES]; /* XXX someday we'll need better accounting */ -static struct socket *o2net_listen_sock = NULL; +static struct socket *o2net_listen_sock; /* * listen work is only queued by the listening socket callbacks on the @@ -136,29 +137,142 @@ static int o2net_sys_err_translations[O2NET_ERR_MAX] = static void o2net_sc_connect_completed(struct work_struct *work); static void o2net_rx_until_empty(struct work_struct *work); static void o2net_shutdown_sc(struct work_struct *work); -static void o2net_listen_data_ready(struct sock *sk, int bytes); +static void o2net_listen_data_ready(struct sock *sk); static void o2net_sc_send_keep_req(struct work_struct *work); static void o2net_idle_timer(unsigned long data); static void o2net_sc_postpone_idle(struct o2net_sock_container *sc); static void o2net_sc_reset_idle_timer(struct o2net_sock_container *sc); -/* - * FIXME: These should use to_o2nm_cluster_from_node(), but we end up - * losing our parent link to the cluster during shutdown. This can be - * solved by adding a pre-removal callback to configfs, or passing - * around the cluster with the node. -jeffm - */ -static inline int o2net_reconnect_delay(struct o2nm_node *node) +#ifdef CONFIG_DEBUG_FS +static void o2net_init_nst(struct o2net_send_tracking *nst, u32 msgtype, + u32 msgkey, struct task_struct *task, u8 node) +{ + INIT_LIST_HEAD(&nst->st_net_debug_item); + nst->st_task = task; + nst->st_msg_type = msgtype; + nst->st_msg_key = msgkey; + nst->st_node = node; +} + +static inline void o2net_set_nst_sock_time(struct o2net_send_tracking *nst) +{ + nst->st_sock_time = ktime_get(); +} + +static inline void o2net_set_nst_send_time(struct o2net_send_tracking *nst) +{ + nst->st_send_time = ktime_get(); +} + +static inline void o2net_set_nst_status_time(struct o2net_send_tracking *nst) +{ + nst->st_status_time = ktime_get(); +} + +static inline void o2net_set_nst_sock_container(struct o2net_send_tracking *nst, + struct o2net_sock_container *sc) +{ + nst->st_sc = sc; +} + +static inline void o2net_set_nst_msg_id(struct o2net_send_tracking *nst, + u32 msg_id) +{ + nst->st_id = msg_id; +} + +static inline void o2net_set_sock_timer(struct o2net_sock_container *sc) +{ + sc->sc_tv_timer = ktime_get(); +} + +static inline void o2net_set_data_ready_time(struct o2net_sock_container *sc) +{ + sc->sc_tv_data_ready = ktime_get(); +} + +static inline void o2net_set_advance_start_time(struct o2net_sock_container *sc) +{ + sc->sc_tv_advance_start = ktime_get(); +} + +static inline void o2net_set_advance_stop_time(struct o2net_sock_container *sc) +{ + sc->sc_tv_advance_stop = ktime_get(); +} + +static inline void o2net_set_func_start_time(struct o2net_sock_container *sc) +{ + sc->sc_tv_func_start = ktime_get(); +} + +static inline void o2net_set_func_stop_time(struct o2net_sock_container *sc) +{ + sc->sc_tv_func_stop = ktime_get(); +} + +#else /* CONFIG_DEBUG_FS */ +# define o2net_init_nst(a, b, c, d, e) +# define o2net_set_nst_sock_time(a) +# define o2net_set_nst_send_time(a) +# define o2net_set_nst_status_time(a) +# define o2net_set_nst_sock_container(a, b) +# define o2net_set_nst_msg_id(a, b) +# define o2net_set_sock_timer(a) +# define o2net_set_data_ready_time(a) +# define o2net_set_advance_start_time(a) +# define o2net_set_advance_stop_time(a) +# define o2net_set_func_start_time(a) +# define o2net_set_func_stop_time(a) +#endif /* CONFIG_DEBUG_FS */ + +#ifdef CONFIG_OCFS2_FS_STATS +static ktime_t o2net_get_func_run_time(struct o2net_sock_container *sc) +{ + return ktime_sub(sc->sc_tv_func_stop, sc->sc_tv_func_start); +} + +static void o2net_update_send_stats(struct o2net_send_tracking *nst, + struct o2net_sock_container *sc) +{ + sc->sc_tv_status_total = ktime_add(sc->sc_tv_status_total, + ktime_sub(ktime_get(), + nst->st_status_time)); + sc->sc_tv_send_total = ktime_add(sc->sc_tv_send_total, + ktime_sub(nst->st_status_time, + nst->st_send_time)); + sc->sc_tv_acquiry_total = ktime_add(sc->sc_tv_acquiry_total, + ktime_sub(nst->st_send_time, + nst->st_sock_time)); + sc->sc_send_count++; +} + +static void o2net_update_recv_stats(struct o2net_sock_container *sc) +{ + sc->sc_tv_process_total = ktime_add(sc->sc_tv_process_total, + o2net_get_func_run_time(sc)); + sc->sc_recv_count++; +} + +#else + +# define o2net_update_send_stats(a, b) + +# define o2net_update_recv_stats(sc) + +#endif /* CONFIG_OCFS2_FS_STATS */ + +static inline unsigned int o2net_reconnect_delay(void) { return o2nm_single_cluster->cl_reconnect_delay_ms; } -static inline int o2net_keepalive_delay(struct o2nm_node *node) +static inline unsigned int o2net_keepalive_delay(void) { return o2nm_single_cluster->cl_keepalive_delay_ms; } -static inline int o2net_idle_timeout(struct o2nm_node *node) +static inline unsigned int o2net_idle_timeout(void) { return o2nm_single_cluster->cl_idle_timeout_ms; } @@ -190,28 +304,22 @@ static u8 o2net_num_from_nn(struct o2net_node *nn) static int o2net_prep_nsw(struct o2net_node *nn, struct o2net_status_wait *nsw) { - int ret = 0; - - do { - if (!idr_pre_get(&nn->nn_status_idr, GFP_ATOMIC)) { - ret = -EAGAIN; - break; - } - spin_lock(&nn->nn_lock); - ret = idr_get_new(&nn->nn_status_idr, nsw, &nsw->ns_id); - if (ret == 0) - list_add_tail(&nsw->ns_node_item, - &nn->nn_status_list); - spin_unlock(&nn->nn_lock); - } while (ret == -EAGAIN); + int ret; - if (ret == 0) { - init_waitqueue_head(&nsw->ns_wq); - nsw->ns_sys_status = O2NET_ERR_NONE; - nsw->ns_status = 0; + spin_lock(&nn->nn_lock); + ret = idr_alloc(&nn->nn_status_idr, nsw, 0, 0, GFP_ATOMIC); + if (ret >= 0) { + nsw->ns_id = ret; + list_add_tail(&nsw->ns_node_item, &nn->nn_status_list); } + spin_unlock(&nn->nn_lock); + if (ret < 0) + return ret; - return ret; + init_waitqueue_head(&nsw->ns_wq); + nsw->ns_sys_status = O2NET_ERR_NONE; + nsw->ns_status = 0; + return 0; } static void o2net_complete_nsw_locked(struct o2net_node *nn, @@ -293,9 +401,14 @@ static void sc_kref_release(struct kref *kref) sc->sc_sock = NULL; } + o2nm_undepend_item(&sc->sc_node->nd_item); o2nm_node_put(sc->sc_node); sc->sc_node = NULL; + o2net_debug_del_sc(sc); + + if (sc->sc_page) + __free_page(sc->sc_page); kfree(sc); } @@ -313,6 +426,7 @@ static struct o2net_sock_container *sc_alloc(struct o2nm_node *node) { struct o2net_sock_container *sc, *ret = NULL; struct page *page = NULL; + int status = 0; page = alloc_page(GFP_NOFS); sc = kzalloc(sizeof(*sc), GFP_NOFS); @@ -323,6 +437,13 @@ static struct o2net_sock_container *sc_alloc(struct o2nm_node *node) o2nm_node_get(node); sc->sc_node = node; + /* pin the node item of the remote node */ + status = o2nm_depend_item(&node->nd_item); + if (status) { + mlog_errno(status); + o2nm_node_put(node); + goto out; + } INIT_WORK(&sc->sc_connect_work, o2net_sc_connect_completed); INIT_WORK(&sc->sc_rx_work, o2net_rx_until_empty); INIT_WORK(&sc->sc_shutdown_work, o2net_shutdown_sc); @@ -336,6 +457,7 @@ static struct o2net_sock_container *sc_alloc(struct o2nm_node *node) ret = sc; sc->sc_page = page; + o2net_debug_add_sc(sc); sc = NULL; page = NULL; @@ -399,8 +521,6 @@ static void o2net_set_nn_state(struct o2net_node *nn, mlog_bug_on_msg(err && valid, "err %d valid %u\n", err, valid); mlog_bug_on_msg(valid && !sc, "valid %u sc %p\n", valid, sc); - /* we won't reconnect after our valid conn goes away for - * this hb iteration.. here so it shows up in the logs */ if (was_valid && !valid && err == 0) err = -ENOTCONN; @@ -423,22 +543,18 @@ static void o2net_set_nn_state(struct o2net_node *nn, } if (was_valid && !valid) { - printk(KERN_INFO "o2net: no longer connected to " - SC_NODEF_FMT "\n", SC_NODEF_ARGS(old_sc)); + if (old_sc) + printk(KERN_NOTICE "o2net: No longer connected to " + SC_NODEF_FMT "\n", SC_NODEF_ARGS(old_sc)); o2net_complete_nodes_nsw(nn); } if (!was_valid && valid) { o2quo_conn_up(o2net_num_from_nn(nn)); - /* this is a bit of a hack. we only try reconnecting - * when heartbeating starts until we get a connection. - * if that connection then dies we don't try reconnecting. - * the only way to start connecting again is to down - * heartbeat and bring it back up. */ cancel_delayed_work(&nn->nn_connect_expired); - printk(KERN_INFO "o2net: %s " SC_NODEF_FMT "\n", + printk(KERN_NOTICE "o2net: %s " SC_NODEF_FMT "\n", o2nm_this_node() > sc->sc_node->nd_num ? - "connected to" : "accepted connection from", + "Connected to" : "Accepted connection from", SC_NODEF_ARGS(sc)); } @@ -448,15 +564,27 @@ static void o2net_set_nn_state(struct o2net_node *nn, * the work queue actually being up. */ if (!valid && o2net_wq) { unsigned long delay; - /* delay if we're withing a RECONNECT_DELAY of the + /* delay if we're within a RECONNECT_DELAY of the * last attempt */ delay = (nn->nn_last_connect_attempt + - msecs_to_jiffies(o2net_reconnect_delay(NULL))) + msecs_to_jiffies(o2net_reconnect_delay())) - jiffies; - if (delay > msecs_to_jiffies(o2net_reconnect_delay(NULL))) + if (delay > msecs_to_jiffies(o2net_reconnect_delay())) delay = 0; mlog(ML_CONN, "queueing conn attempt in %lu jiffies\n", delay); queue_delayed_work(o2net_wq, &nn->nn_connect_work, delay); + + /* + * Delay the expired work after idle timeout. + * + * We might have lots of failed connection attempts that run + * through here but we only cancel the connect_expired work when + * a connection attempt succeeds. So only the first enqueue of + * the connect_expired work will do anything. The rest will see + * that it's already queued and do nothing. + */ + delay += msecs_to_jiffies(o2net_idle_timeout()); + queue_delayed_work(o2net_wq, &nn->nn_connect_expired, delay); } /* keep track of the nn's sc ref for the caller */ @@ -469,15 +597,15 @@ static void o2net_set_nn_state(struct o2net_node *nn, } /* see o2net_register_callbacks() */ -static void o2net_data_ready(struct sock *sk, int bytes) +static void o2net_data_ready(struct sock *sk) { - void (*ready)(struct sock *sk, int bytes); + void (*ready)(struct sock *sk); read_lock(&sk->sk_callback_lock); if (sk->sk_user_data) { struct o2net_sock_container *sc = sk->sk_user_data; sclog(sc, "data_ready hit\n"); - do_gettimeofday(&sc->sc_tv_data_ready); + o2net_set_data_ready_time(sc); o2net_sc_queue_work(sc, &sc->sc_rx_work); ready = sc->sc_data_ready; } else { @@ -485,7 +613,7 @@ static void o2net_data_ready(struct sock *sk, int bytes) } read_unlock(&sk->sk_callback_lock); - ready(sk, bytes); + ready(sk); } /* see o2net_register_callbacks() */ @@ -506,16 +634,19 @@ static void o2net_state_change(struct sock *sk) state_change = sc->sc_state_change; switch(sk->sk_state) { - /* ignore connecting sockets as they make progress */ - case TCP_SYN_SENT: - case TCP_SYN_RECV: - break; - case TCP_ESTABLISHED: - o2net_sc_queue_work(sc, &sc->sc_connect_work); - break; - default: - o2net_sc_queue_work(sc, &sc->sc_shutdown_work); - break; + /* ignore connecting sockets as they make progress */ + case TCP_SYN_SENT: + case TCP_SYN_RECV: + break; + case TCP_ESTABLISHED: + o2net_sc_queue_work(sc, &sc->sc_connect_work); + break; + default: + printk(KERN_INFO "o2net: Connection to " SC_NODEF_FMT + " shutdown, state %d\n", + SC_NODEF_ARGS(sc), sk->sk_state); + o2net_sc_queue_work(sc, &sc->sc_shutdown_work); + break; } out: read_unlock(&sk->sk_callback_lock); @@ -635,32 +766,32 @@ static struct o2net_msg_handler * o2net_handler_tree_lookup(u32 msg_type, u32 key, struct rb_node ***ret_p, struct rb_node **ret_parent) { - struct rb_node **p = &o2net_handler_tree.rb_node; - struct rb_node *parent = NULL; + struct rb_node **p = &o2net_handler_tree.rb_node; + struct rb_node *parent = NULL; struct o2net_msg_handler *nmh, *ret = NULL; int cmp; - while (*p) { - parent = *p; - nmh = rb_entry(parent, struct o2net_msg_handler, nh_node); + while (*p) { + parent = *p; + nmh = rb_entry(parent, struct o2net_msg_handler, nh_node); cmp = o2net_handler_cmp(nmh, msg_type, key); - if (cmp < 0) - p = &(*p)->rb_left; - else if (cmp > 0) - p = &(*p)->rb_right; - else { + if (cmp < 0) + p = &(*p)->rb_left; + else if (cmp > 0) + p = &(*p)->rb_right; + else { ret = nmh; - break; + break; } - } + } - if (ret_p != NULL) - *ret_p = p; - if (ret_parent != NULL) - *ret_parent = parent; + if (ret_p != NULL) + *ret_p = p; + if (ret_parent != NULL) + *ret_parent = parent; - return ret; + return ret; } static void o2net_handler_kref_release(struct kref *kref) @@ -737,7 +868,7 @@ int o2net_register_handler(u32 msg_type, u32 key, u32 max_len, /* we've had some trouble with handlers seemingly vanishing. */ mlog_bug_on_msg(o2net_handler_tree_lookup(msg_type, key, &p, &parent) == NULL, - "couldn't find handler we *just* registerd " + "couldn't find handler we *just* registered " "for type %u key %08x\n", msg_type, key); } write_unlock(&o2net_handler_lock); @@ -785,57 +916,30 @@ static struct o2net_msg_handler *o2net_handler_get(u32 msg_type, u32 key) static int o2net_recv_tcp_msg(struct socket *sock, void *data, size_t len) { - int ret; - mm_segment_t oldfs; - struct kvec vec = { - .iov_len = len, - .iov_base = data, - }; - struct msghdr msg = { - .msg_iovlen = 1, - .msg_iov = (struct iovec *)&vec, - .msg_flags = MSG_DONTWAIT, - }; - - oldfs = get_fs(); - set_fs(get_ds()); - ret = sock_recvmsg(sock, &msg, len, msg.msg_flags); - set_fs(oldfs); - - return ret; + struct kvec vec = { .iov_len = len, .iov_base = data, }; + struct msghdr msg = { .msg_flags = MSG_DONTWAIT, }; + return kernel_recvmsg(sock, &msg, &vec, 1, len, msg.msg_flags); } static int o2net_send_tcp_msg(struct socket *sock, struct kvec *vec, size_t veclen, size_t total) { int ret; - mm_segment_t oldfs; - struct msghdr msg = { - .msg_iov = (struct iovec *)vec, - .msg_iovlen = veclen, - }; + struct msghdr msg; if (sock == NULL) { ret = -EINVAL; goto out; } - oldfs = get_fs(); - set_fs(get_ds()); - ret = sock_sendmsg(sock, &msg, total); - set_fs(oldfs); - if (ret != total) { - mlog(ML_ERROR, "sendmsg returned %d instead of %zu\n", ret, - total); - if (ret >= 0) - ret = -EPIPE; /* should be smarter, I bet */ - goto out; - } - - ret = 0; + ret = kernel_sendmsg(sock, &msg, vec, veclen, total); + if (likely(ret == total)) + return 0; + mlog(ML_ERROR, "sendmsg returned %d instead of %zu\n", ret, total); + if (ret >= 0) + ret = -EPIPE; /* should be smarter, I bet */ out: - if (ret < 0) - mlog(0, "returning error: %d\n", ret); + mlog(0, "returning error: %d\n", ret); return ret; } @@ -861,7 +965,7 @@ static void o2net_sendpage(struct o2net_sock_container *sc, cond_resched(); continue; } - mlog(ML_ERROR, "sendpage of size %zu to " SC_NODEF_FMT + mlog(ML_ERROR, "sendpage of size %zu to " SC_NODEF_FMT " failed with %zd\n", size, SC_NODEF_ARGS(sc), ret); o2net_ensure_shutdown(nn, sc, 0); break; @@ -902,10 +1006,29 @@ static int o2net_tx_can_proceed(struct o2net_node *nn, return ret; } +/* Get a map of all nodes to which this node is currently connected to */ +void o2net_fill_node_map(unsigned long *map, unsigned bytes) +{ + struct o2net_sock_container *sc; + int node, ret; + + BUG_ON(bytes < (BITS_TO_LONGS(O2NM_MAX_NODES) * sizeof(unsigned long))); + + memset(map, 0, bytes); + for (node = 0; node < O2NM_MAX_NODES; ++node) { + o2net_tx_can_proceed(o2net_nn_from_num(node), &sc, &ret); + if (!ret) { + set_bit(node, map); + sc_put(sc); + } + } +} +EXPORT_SYMBOL_GPL(o2net_fill_node_map); + int o2net_send_message_vec(u32 msg_type, u32 key, struct kvec *caller_vec, size_t caller_veclen, u8 target_node, int *status) { - int ret, error = 0; + int ret = 0; struct o2net_msg *msg = NULL; size_t veclen, caller_bytes = 0; struct kvec *vec = NULL; @@ -914,6 +1037,9 @@ int o2net_send_message_vec(u32 msg_type, u32 key, struct kvec *caller_vec, struct o2net_status_wait nsw = { .ns_node_item = LIST_HEAD_INIT(nsw.ns_node_item), }; + struct o2net_send_tracking nst; + + o2net_init_nst(&nst, msg_type, key, current, target_node); if (o2net_wq == NULL) { mlog(0, "attempt to tx without o2netd running\n"); @@ -939,13 +1065,16 @@ int o2net_send_message_vec(u32 msg_type, u32 key, struct kvec *caller_vec, goto out; } - ret = wait_event_interruptible(nn->nn_sc_wq, - o2net_tx_can_proceed(nn, &sc, &error)); - if (!ret && error) - ret = error; + o2net_debug_add_nst(&nst); + + o2net_set_nst_sock_time(&nst); + + wait_event(nn->nn_sc_wq, o2net_tx_can_proceed(nn, &sc, &ret)); if (ret) goto out; + o2net_set_nst_sock_container(&nst, sc); + veclen = caller_veclen + 1; vec = kmalloc(sizeof(struct kvec) * veclen, GFP_ATOMIC); if (vec == NULL) { @@ -972,6 +1101,9 @@ int o2net_send_message_vec(u32 msg_type, u32 key, struct kvec *caller_vec, goto out; msg->msg_num = cpu_to_be32(nsw.ns_id); + o2net_set_nst_msg_id(&nst, nsw.ns_id); + + o2net_set_nst_send_time(&nst); /* finally, convert the message header to network byte-order * and send */ @@ -986,8 +1118,11 @@ int o2net_send_message_vec(u32 msg_type, u32 key, struct kvec *caller_vec, } /* wait on other node's handler */ + o2net_set_nst_status_time(&nst); wait_event(nsw.ns_wq, o2net_nsw_completed(nn, &nsw)); + o2net_update_send_stats(&nst, sc); + /* Note that we avoid overwriting the callers status return * variable if a system error was reported on the other * side. Callers beware. */ @@ -998,12 +1133,11 @@ int o2net_send_message_vec(u32 msg_type, u32 key, struct kvec *caller_vec, mlog(0, "woken, returning system status %d, user status %d\n", ret, nsw.ns_status); out: + o2net_debug_del_nst(&nst); /* must be before dropping sc and node */ if (sc) sc_put(sc); - if (vec) - kfree(vec); - if (msg) - kfree(msg); + kfree(vec); + kfree(msg); o2net_complete_nsw(nn, &nsw, 0, 0, 0); return ret; } @@ -1100,13 +1234,15 @@ static int o2net_process_message(struct o2net_sock_container *sc, if (syserr != O2NET_ERR_NONE) goto out_respond; - do_gettimeofday(&sc->sc_tv_func_start); + o2net_set_func_start_time(sc); sc->sc_msg_key = be32_to_cpu(hdr->key); sc->sc_msg_type = be16_to_cpu(hdr->msg_type); handler_status = (nmh->nh_func)(hdr, sizeof(struct o2net_msg) + be16_to_cpu(hdr->data_len), nmh->nh_func_data, &ret_data); - do_gettimeofday(&sc->sc_tv_func_stop); + o2net_set_func_stop_time(sc); + + o2net_update_recv_stats(sc); out_respond: /* this destroys the hdr, so don't use it after this */ @@ -1137,11 +1273,11 @@ static int o2net_check_handshake(struct o2net_sock_container *sc) struct o2net_node *nn = o2net_nn_from_num(sc->sc_node->nd_num); if (hand->protocol_version != cpu_to_be64(O2NET_PROTOCOL_VERSION)) { - mlog(ML_NOTICE, SC_NODEF_FMT " advertised net protocol " - "version %llu but %llu is required, disconnecting\n", - SC_NODEF_ARGS(sc), - (unsigned long long)be64_to_cpu(hand->protocol_version), - O2NET_PROTOCOL_VERSION); + printk(KERN_NOTICE "o2net: " SC_NODEF_FMT " Advertised net " + "protocol version %llu but %llu is required. " + "Disconnecting.\n", SC_NODEF_ARGS(sc), + (unsigned long long)be64_to_cpu(hand->protocol_version), + O2NET_PROTOCOL_VERSION); /* don't bother reconnecting if its the wrong version. */ o2net_ensure_shutdown(nn, sc, -ENOTCONN); @@ -1154,34 +1290,34 @@ static int o2net_check_handshake(struct o2net_sock_container *sc) * but isn't. This can ultimately cause corruption. */ if (be32_to_cpu(hand->o2net_idle_timeout_ms) != - o2net_idle_timeout(sc->sc_node)) { - mlog(ML_NOTICE, SC_NODEF_FMT " uses a network idle timeout of " - "%u ms, but we use %u ms locally. disconnecting\n", - SC_NODEF_ARGS(sc), - be32_to_cpu(hand->o2net_idle_timeout_ms), - o2net_idle_timeout(sc->sc_node)); + o2net_idle_timeout()) { + printk(KERN_NOTICE "o2net: " SC_NODEF_FMT " uses a network " + "idle timeout of %u ms, but we use %u ms locally. " + "Disconnecting.\n", SC_NODEF_ARGS(sc), + be32_to_cpu(hand->o2net_idle_timeout_ms), + o2net_idle_timeout()); o2net_ensure_shutdown(nn, sc, -ENOTCONN); return -1; } if (be32_to_cpu(hand->o2net_keepalive_delay_ms) != - o2net_keepalive_delay(sc->sc_node)) { - mlog(ML_NOTICE, SC_NODEF_FMT " uses a keepalive delay of " - "%u ms, but we use %u ms locally. disconnecting\n", - SC_NODEF_ARGS(sc), - be32_to_cpu(hand->o2net_keepalive_delay_ms), - o2net_keepalive_delay(sc->sc_node)); + o2net_keepalive_delay()) { + printk(KERN_NOTICE "o2net: " SC_NODEF_FMT " uses a keepalive " + "delay of %u ms, but we use %u ms locally. " + "Disconnecting.\n", SC_NODEF_ARGS(sc), + be32_to_cpu(hand->o2net_keepalive_delay_ms), + o2net_keepalive_delay()); o2net_ensure_shutdown(nn, sc, -ENOTCONN); return -1; } if (be32_to_cpu(hand->o2hb_heartbeat_timeout_ms) != O2HB_MAX_WRITE_TIMEOUT_MS) { - mlog(ML_NOTICE, SC_NODEF_FMT " uses a heartbeat timeout of " - "%u ms, but we use %u ms locally. disconnecting\n", - SC_NODEF_ARGS(sc), - be32_to_cpu(hand->o2hb_heartbeat_timeout_ms), - O2HB_MAX_WRITE_TIMEOUT_MS); + printk(KERN_NOTICE "o2net: " SC_NODEF_FMT " uses a heartbeat " + "timeout of %u ms, but we use %u ms locally. " + "Disconnecting.\n", SC_NODEF_ARGS(sc), + be32_to_cpu(hand->o2hb_heartbeat_timeout_ms), + O2HB_MAX_WRITE_TIMEOUT_MS); o2net_ensure_shutdown(nn, sc, -ENOTCONN); return -1; } @@ -1193,6 +1329,7 @@ static int o2net_check_handshake(struct o2net_sock_container *sc) * shut down already */ if (nn->nn_sc == sc) { o2net_sc_reset_idle_timer(sc); + atomic_set(&nn->nn_timeout, 0); o2net_set_nn_state(nn, sc, 1, 0); } spin_unlock(&nn->nn_lock); @@ -1216,7 +1353,7 @@ static int o2net_advance_rx(struct o2net_sock_container *sc) size_t datalen; sclog(sc, "receiving\n"); - do_gettimeofday(&sc->sc_tv_advance_start); + o2net_set_advance_start_time(sc); if (unlikely(sc->sc_handshake_ok == 0)) { if(sc->sc_page_off < sizeof(struct o2net_handshake)) { @@ -1291,7 +1428,7 @@ static int o2net_advance_rx(struct o2net_sock_container *sc) out: sclog(sc, "ret = %d\n", ret); - do_gettimeofday(&sc->sc_tv_advance_stop); + o2net_set_advance_stop_time(sc); return ret; } @@ -1347,12 +1484,11 @@ static void o2net_initialize_handshake(void) { o2net_hand->o2hb_heartbeat_timeout_ms = cpu_to_be32( O2HB_MAX_WRITE_TIMEOUT_MS); - o2net_hand->o2net_idle_timeout_ms = cpu_to_be32( - o2net_idle_timeout(NULL)); + o2net_hand->o2net_idle_timeout_ms = cpu_to_be32(o2net_idle_timeout()); o2net_hand->o2net_keepalive_delay_ms = cpu_to_be32( - o2net_keepalive_delay(NULL)); + o2net_keepalive_delay()); o2net_hand->o2net_reconnect_delay_ms = cpu_to_be32( - o2net_reconnect_delay(NULL)); + o2net_reconnect_delay()); } /* ------------------------------------------------------------ */ @@ -1391,27 +1527,23 @@ static void o2net_sc_send_keep_req(struct work_struct *work) static void o2net_idle_timer(unsigned long data) { struct o2net_sock_container *sc = (struct o2net_sock_container *)data; - struct timeval now; - - do_gettimeofday(&now); - - printk(KERN_INFO "o2net: connection to " SC_NODEF_FMT " has been idle for %u.%u " - "seconds, shutting it down.\n", SC_NODEF_ARGS(sc), - o2net_idle_timeout(sc->sc_node) / 1000, - o2net_idle_timeout(sc->sc_node) % 1000); - mlog(ML_NOTICE, "here are some times that might help debug the " - "situation: (tmr %ld.%ld now %ld.%ld dr %ld.%ld adv " - "%ld.%ld:%ld.%ld func (%08x:%u) %ld.%ld:%ld.%ld)\n", - sc->sc_tv_timer.tv_sec, (long) sc->sc_tv_timer.tv_usec, - now.tv_sec, (long) now.tv_usec, - sc->sc_tv_data_ready.tv_sec, (long) sc->sc_tv_data_ready.tv_usec, - sc->sc_tv_advance_start.tv_sec, - (long) sc->sc_tv_advance_start.tv_usec, - sc->sc_tv_advance_stop.tv_sec, - (long) sc->sc_tv_advance_stop.tv_usec, - sc->sc_msg_key, sc->sc_msg_type, - sc->sc_tv_func_start.tv_sec, (long) sc->sc_tv_func_start.tv_usec, - sc->sc_tv_func_stop.tv_sec, (long) sc->sc_tv_func_stop.tv_usec); + struct o2net_node *nn = o2net_nn_from_num(sc->sc_node->nd_num); +#ifdef CONFIG_DEBUG_FS + unsigned long msecs = ktime_to_ms(ktime_get()) - + ktime_to_ms(sc->sc_tv_timer); +#else + unsigned long msecs = o2net_idle_timeout(); +#endif + + printk(KERN_NOTICE "o2net: Connection to " SC_NODEF_FMT " has been " + "idle for %lu.%lu secs, shutting it down.\n", SC_NODEF_ARGS(sc), + msecs / 1000, msecs % 1000); + + /* + * Initialize the nn_timeout so that the next connection attempt + * will continue in o2net_start_connect. + */ + atomic_set(&nn->nn_timeout, 1); o2net_sc_queue_work(sc, &sc->sc_shutdown_work); } @@ -1420,10 +1552,10 @@ static void o2net_sc_reset_idle_timer(struct o2net_sock_container *sc) { o2net_sc_cancel_delayed_work(sc, &sc->sc_keepalive_work); o2net_sc_queue_delayed_work(sc, &sc->sc_keepalive_work, - msecs_to_jiffies(o2net_keepalive_delay(sc->sc_node))); - do_gettimeofday(&sc->sc_tv_timer); + msecs_to_jiffies(o2net_keepalive_delay())); + o2net_set_sock_timer(sc); mod_timer(&sc->sc_idle_timeout, - jiffies + msecs_to_jiffies(o2net_idle_timeout(sc->sc_node))); + jiffies + msecs_to_jiffies(o2net_idle_timeout())); } static void o2net_sc_postpone_idle(struct o2net_sock_container *sc) @@ -1447,6 +1579,7 @@ static void o2net_start_connect(struct work_struct *work) struct socket *sock = NULL; struct sockaddr_in myaddr = {0, }, remoteaddr = {0, }; int ret = 0, stop; + unsigned int timeout; /* if we're greater we initiate tx, otherwise we accept */ if (o2nm_this_node() <= o2net_num_from_nn(nn)) @@ -1466,8 +1599,17 @@ static void o2net_start_connect(struct work_struct *work) } spin_lock(&nn->nn_lock); - /* see if we already have one pending or have given up */ - stop = (nn->nn_sc || nn->nn_persistent_error); + /* + * see if we already have one pending or have given up. + * For nn_timeout, it is set when we close the connection + * because of the idle time out. So it means that we have + * at least connected to that node successfully once, + * now try to connect to it again. + */ + timeout = atomic_read(&nn->nn_timeout); + stop = (nn->nn_sc || + (nn->nn_persistent_error && + (nn->nn_persistent_error != -ENOTCONN || timeout == 0))); spin_unlock(&nn->nn_lock); if (stop) goto out; @@ -1497,8 +1639,8 @@ static void o2net_start_connect(struct work_struct *work) ret = sock->ops->bind(sock, (struct sockaddr *)&myaddr, sizeof(myaddr)); if (ret) { - mlog(ML_ERROR, "bind failed with %d at address %u.%u.%u.%u\n", - ret, NIPQUAD(mynode->nd_ipv4_address)); + mlog(ML_ERROR, "bind failed with %d at address %pI4\n", + ret, &mynode->nd_ipv4_address); goto out; } @@ -1527,13 +1669,12 @@ static void o2net_start_connect(struct work_struct *work) ret = 0; out: - if (ret) { - mlog(ML_NOTICE, "connect attempt to " SC_NODEF_FMT " failed " - "with errno %d\n", SC_NODEF_ARGS(sc), ret); + if (ret && sc) { + printk(KERN_NOTICE "o2net: Connect attempt to " SC_NODEF_FMT + " failed with errno %d\n", SC_NODEF_ARGS(sc), ret); /* 0 err so that another will be queued and attempted * from set_nn_state */ - if (sc) - o2net_ensure_shutdown(nn, sc, 0); + o2net_ensure_shutdown(nn, sc, 0); } if (sc) sc_put(sc); @@ -1552,11 +1693,11 @@ static void o2net_connect_expired(struct work_struct *work) spin_lock(&nn->nn_lock); if (!nn->nn_sc_valid) { - mlog(ML_ERROR, "no connection established with node %u after " - "%u.%u seconds, giving up and returning errors.\n", + printk(KERN_NOTICE "o2net: No connection established with " + "node %u after %u.%u seconds, giving up.\n", o2net_num_from_nn(nn), - o2net_idle_timeout(NULL) / 1000, - o2net_idle_timeout(NULL) % 1000); + o2net_idle_timeout() / 1000, + o2net_idle_timeout() % 1000); o2net_set_nn_state(nn, NULL, 0, -ENOTCONN); } @@ -1579,6 +1720,7 @@ void o2net_disconnect_node(struct o2nm_node *node) /* don't reconnect until it's heartbeating again */ spin_lock(&nn->nn_lock); + atomic_set(&nn->nn_timeout, 0); o2net_set_nn_state(nn, NULL, 0, -ENOTCONN); spin_unlock(&nn->nn_lock); @@ -1595,6 +1737,9 @@ static void o2net_hb_node_down_cb(struct o2nm_node *node, int node_num, { o2quo_hb_down(node_num); + if (!node) + return; + if (node_num != o2nm_this_node()) o2net_disconnect_node(node); @@ -1608,22 +1753,19 @@ static void o2net_hb_node_up_cb(struct o2nm_node *node, int node_num, o2quo_hb_up(node_num); + BUG_ON(!node); + /* ensure an immediate connect attempt */ nn->nn_last_connect_attempt = jiffies - - (msecs_to_jiffies(o2net_reconnect_delay(node)) + 1); + (msecs_to_jiffies(o2net_reconnect_delay()) + 1); if (node_num != o2nm_this_node()) { - /* heartbeat doesn't work unless a local node number is - * configured and doing so brings up the o2net_wq, so we can - * use it.. */ - queue_delayed_work(o2net_wq, &nn->nn_connect_expired, - msecs_to_jiffies(o2net_idle_timeout(node))); - /* believe it or not, accept and node hearbeating testing * can succeed for this node before we got here.. so * only use set_nn_state to clear the persistent error * if that hasn't already happened */ spin_lock(&nn->nn_lock); + atomic_set(&nn->nn_timeout, 0); if (nn->nn_persistent_error) o2net_set_nn_state(nn, NULL, 0, 0); spin_unlock(&nn->nn_lock); @@ -1657,16 +1799,18 @@ int o2net_register_hb_callbacks(void) /* ------------------------------------------------------------ */ -static int o2net_accept_one(struct socket *sock) +static int o2net_accept_one(struct socket *sock, int *more) { int ret, slen; struct sockaddr_in sin; struct socket *new_sock = NULL; struct o2nm_node *node = NULL; + struct o2nm_node *local_node = NULL; struct o2net_sock_container *sc = NULL; struct o2net_node *nn; BUG_ON(sock == NULL); + *more = 0; ret = sock_create_lite(sock->sk->sk_family, sock->sk->sk_type, sock->sk->sk_protocol, &new_sock); if (ret) @@ -1678,6 +1822,7 @@ static int o2net_accept_one(struct socket *sock) if (ret < 0) goto out; + *more = 1; new_sock->sk->sk_allocation = GFP_ATOMIC; ret = o2net_set_nodelay(new_sock); @@ -1694,18 +1839,25 @@ static int o2net_accept_one(struct socket *sock) node = o2nm_get_node_by_ip(sin.sin_addr.s_addr); if (node == NULL) { - mlog(ML_NOTICE, "attempt to connect from unknown node at " - "%u.%u.%u.%u:%d\n", NIPQUAD(sin.sin_addr.s_addr), - ntohs(sin.sin_port)); + printk(KERN_NOTICE "o2net: Attempt to connect from unknown " + "node at %pI4:%d\n", &sin.sin_addr.s_addr, + ntohs(sin.sin_port)); ret = -EINVAL; goto out; } - if (o2nm_this_node() > node->nd_num) { - mlog(ML_NOTICE, "unexpected connect attempted from a lower " - "numbered node '%s' at " "%u.%u.%u.%u:%d with num %u\n", - node->nd_name, NIPQUAD(sin.sin_addr.s_addr), - ntohs(sin.sin_port), node->nd_num); + if (o2nm_this_node() >= node->nd_num) { + local_node = o2nm_get_node_by_num(o2nm_this_node()); + if (local_node) + printk(KERN_NOTICE "o2net: Unexpected connect attempt " + "seen at node '%s' (%u, %pI4:%d) from " + "node '%s' (%u, %pI4:%d)\n", + local_node->nd_name, local_node->nd_num, + &(local_node->nd_ipv4_address), + ntohs(local_node->nd_ipv4_port), + node->nd_name, + node->nd_num, &sin.sin_addr.s_addr, + ntohs(sin.sin_port)); ret = -EINVAL; goto out; } @@ -1714,8 +1866,8 @@ static int o2net_accept_one(struct socket *sock) * and tries to connect before we see their heartbeat */ if (!o2hb_check_node_heartbeating_from_callback(node->nd_num)) { mlog(ML_CONN, "attempt to connect from node '%s' at " - "%u.%u.%u.%u:%d but it isn't heartbeating\n", - node->nd_name, NIPQUAD(sin.sin_addr.s_addr), + "%pI4:%d but it isn't heartbeating\n", + node->nd_name, &sin.sin_addr.s_addr, ntohs(sin.sin_port)); ret = -EINVAL; goto out; @@ -1730,10 +1882,10 @@ static int o2net_accept_one(struct socket *sock) ret = 0; spin_unlock(&nn->nn_lock); if (ret) { - mlog(ML_NOTICE, "attempt to connect from node '%s' at " - "%u.%u.%u.%u:%d but it already has an open connection\n", - node->nd_name, NIPQUAD(sin.sin_addr.s_addr), - ntohs(sin.sin_port)); + printk(KERN_NOTICE "o2net: Attempt to connect from node '%s' " + "at %pI4:%d but it already has an open connection\n", + node->nd_name, &sin.sin_addr.s_addr, + ntohs(sin.sin_port)); goto out; } @@ -1747,6 +1899,7 @@ static int o2net_accept_one(struct socket *sock) new_sock = NULL; spin_lock(&nn->nn_lock); + atomic_set(&nn->nn_timeout, 0); o2net_set_nn_state(nn, sc, 0, 0); spin_unlock(&nn->nn_lock); @@ -1761,21 +1914,48 @@ out: sock_release(new_sock); if (node) o2nm_node_put(node); + if (local_node) + o2nm_node_put(local_node); if (sc) sc_put(sc); return ret; } +/* + * This function is invoked in response to one or more + * pending accepts at softIRQ level. We must drain the + * entire que before returning. + */ + static void o2net_accept_many(struct work_struct *work) { struct socket *sock = o2net_listen_sock; - while (o2net_accept_one(sock) == 0) + int more; + int err; + + /* + * It is critical to note that due to interrupt moderation + * at the network driver level, we can't assume to get a + * softIRQ for every single conn since tcp SYN packets + * can arrive back-to-back, and therefore many pending + * accepts may result in just 1 softIRQ. If we terminate + * the o2net_accept_one() loop upon seeing an err, what happens + * to the rest of the conns in the queue? If no new SYN + * arrives for hours, no softIRQ will be delivered, + * and the connections will just sit in the queue. + */ + + for (;;) { + err = o2net_accept_one(sock, &more); + if (!more) + break; cond_resched(); + } } -static void o2net_listen_data_ready(struct sock *sk, int bytes) +static void o2net_listen_data_ready(struct sock *sk) { - void (*ready)(struct sock *sk, int bytes); + void (*ready)(struct sock *sk); read_lock(&sk->sk_callback_lock); ready = sk->sk_user_data; @@ -1784,18 +1964,29 @@ static void o2net_listen_data_ready(struct sock *sk, int bytes) goto out; } - /* ->sk_data_ready is also called for a newly established child socket - * before it has been accepted and the acceptor has set up their - * data_ready.. we only want to queue listen work for our listening - * socket */ + /* This callback may called twice when a new connection + * is being established as a child socket inherits everything + * from a parent LISTEN socket, including the data_ready cb of + * the parent. This leads to a hazard. In o2net_accept_one() + * we are still initializing the child socket but have not + * changed the inherited data_ready callback yet when + * data starts arriving. + * We avoid this hazard by checking the state. + * For the listening socket, the state will be TCP_LISTEN; for the new + * socket, will be TCP_ESTABLISHED. Also, in this case, + * sk->sk_user_data is not a valid function pointer. + */ + if (sk->sk_state == TCP_LISTEN) { - mlog(ML_TCP, "bytes: %d\n", bytes); queue_work(o2net_wq, &o2net_listen_work); + } else { + ready = NULL; } out: read_unlock(&sk->sk_callback_lock); - ready(sk, bytes); + if (ready != NULL) + ready(sk); } static int o2net_open_listening_sock(__be32 addr, __be16 port) @@ -1810,7 +2001,7 @@ static int o2net_open_listening_sock(__be32 addr, __be16 port) ret = sock_create(PF_INET, SOCK_STREAM, IPPROTO_TCP, &sock); if (ret < 0) { - mlog(ML_ERROR, "unable to create socket, ret=%d\n", ret); + printk(KERN_ERR "o2net: Error %d while creating socket\n", ret); goto out; } @@ -1824,19 +2015,18 @@ static int o2net_open_listening_sock(__be32 addr, __be16 port) o2net_listen_sock = sock; INIT_WORK(&o2net_listen_work, o2net_accept_many); - sock->sk->sk_reuse = 1; + sock->sk->sk_reuse = SK_CAN_REUSE; ret = sock->ops->bind(sock, (struct sockaddr *)&sin, sizeof(sin)); if (ret < 0) { - mlog(ML_ERROR, "unable to bind socket at %u.%u.%u.%u:%u, " - "ret=%d\n", NIPQUAD(addr), ntohs(port), ret); + printk(KERN_ERR "o2net: Error %d while binding socket at " + "%pI4:%u\n", ret, &addr, ntohs(port)); goto out; } ret = sock->ops->listen(sock, 64); - if (ret < 0) { - mlog(ML_ERROR, "unable to listen on %u.%u.%u.%u:%u, ret=%d\n", - NIPQUAD(addr), ntohs(port), ret); - } + if (ret < 0) + printk(KERN_ERR "o2net: Error %d while listening on %pI4:%u\n", + ret, &addr, ntohs(port)); out: if (ret) { @@ -1922,6 +2112,9 @@ int o2net_init(void) o2quo_init(); + if (o2net_debugfs_init()) + return -ENOMEM; + o2net_hand = kzalloc(sizeof(struct o2net_handshake), GFP_KERNEL); o2net_keep_req = kzalloc(sizeof(struct o2net_msg), GFP_KERNEL); o2net_keep_resp = kzalloc(sizeof(struct o2net_msg), GFP_KERNEL); @@ -1941,6 +2134,7 @@ int o2net_init(void) for (i = 0; i < ARRAY_SIZE(o2net_nodes); i++) { struct o2net_node *nn = o2net_nn_from_num(i); + atomic_set(&nn->nn_timeout, 0); spin_lock_init(&nn->nn_lock); INIT_DELAYED_WORK(&nn->nn_connect_work, o2net_start_connect); INIT_DELAYED_WORK(&nn->nn_connect_expired, @@ -1962,4 +2156,5 @@ void o2net_exit(void) kfree(o2net_hand); kfree(o2net_keep_req); kfree(o2net_keep_resp); + o2net_debugfs_exit(); } diff --git a/fs/ocfs2/cluster/tcp.h b/fs/ocfs2/cluster/tcp.h index f36f66aab3d..5bada2a69b5 100644 --- a/fs/ocfs2/cluster/tcp.h +++ b/fs/ocfs2/cluster/tcp.h @@ -106,6 +106,8 @@ int o2net_register_handler(u32 msg_type, u32 key, u32 max_len, struct list_head *unreg_list); void o2net_unregister_handler_list(struct list_head *list); +void o2net_fill_node_map(unsigned long *map, unsigned bytes); + struct o2nm_node; int o2net_register_hb_callbacks(void); void o2net_unregister_hb_callbacks(void); @@ -117,4 +119,36 @@ int o2net_num_connected_peers(void); int o2net_init(void); void o2net_exit(void); +struct o2net_send_tracking; +struct o2net_sock_container; + +#ifdef CONFIG_DEBUG_FS +int o2net_debugfs_init(void); +void o2net_debugfs_exit(void); +void o2net_debug_add_nst(struct o2net_send_tracking *nst); +void o2net_debug_del_nst(struct o2net_send_tracking *nst); +void o2net_debug_add_sc(struct o2net_sock_container *sc); +void o2net_debug_del_sc(struct o2net_sock_container *sc); +#else +static inline int o2net_debugfs_init(void) +{ + return 0; +} +static inline void o2net_debugfs_exit(void) +{ +} +static inline void o2net_debug_add_nst(struct o2net_send_tracking *nst) +{ +} +static inline void o2net_debug_del_nst(struct o2net_send_tracking *nst) +{ +} +static inline void o2net_debug_add_sc(struct o2net_sock_container *sc) +{ +} +static inline void o2net_debug_del_sc(struct o2net_sock_container *sc) +{ +} +#endif /* CONFIG_DEBUG_FS */ + #endif /* O2CLUSTER_TCP_H */ diff --git a/fs/ocfs2/cluster/tcp_internal.h b/fs/ocfs2/cluster/tcp_internal.h index d25b9af2850..dc024367110 100644 --- a/fs/ocfs2/cluster/tcp_internal.h +++ b/fs/ocfs2/cluster/tcp_internal.h @@ -32,10 +32,10 @@ * on their number */ #define O2NET_QUORUM_DELAY_MS ((o2hb_dead_threshold + 2) * O2HB_REGION_TIMEOUT_MS) -/* +/* * This version number represents quite a lot, unfortunately. It not * only represents the raw network message protocol on the wire but also - * locking semantics of the file system using the protocol. It should + * locking semantics of the file system using the protocol. It should * be somewhere else, I'm sure, but right now it isn't. * * With version 11, we separate out the filesystem locking portion. The @@ -95,6 +95,8 @@ struct o2net_node { unsigned nn_sc_valid:1; /* if this is set tx just returns it */ int nn_persistent_error; + /* It is only set to 1 after the idle time out. */ + atomic_t nn_timeout; /* threads waiting for an sc to arrive wait on the wq for generation * to increase. it is increased when a connecting socket succeeds @@ -127,7 +129,7 @@ struct o2net_node { struct o2net_sock_container { struct kref sc_kref; - /* the next two are vaild for the life time of the sc */ + /* the next two are valid for the life time of the sc */ struct socket *sc_sock; struct o2nm_node *sc_node; @@ -163,17 +165,28 @@ struct o2net_sock_container { /* original handlers for the sockets */ void (*sc_state_change)(struct sock *sk); - void (*sc_data_ready)(struct sock *sk, int bytes); - - struct timeval sc_tv_timer; - struct timeval sc_tv_data_ready; - struct timeval sc_tv_advance_start; - struct timeval sc_tv_advance_stop; - struct timeval sc_tv_func_start; - struct timeval sc_tv_func_stop; + void (*sc_data_ready)(struct sock *sk); + u32 sc_msg_key; u16 sc_msg_type; +#ifdef CONFIG_DEBUG_FS + struct list_head sc_net_debug_item; + ktime_t sc_tv_timer; + ktime_t sc_tv_data_ready; + ktime_t sc_tv_advance_start; + ktime_t sc_tv_advance_stop; + ktime_t sc_tv_func_start; + ktime_t sc_tv_func_stop; +#endif +#ifdef CONFIG_OCFS2_FS_STATS + ktime_t sc_tv_acquiry_total; + ktime_t sc_tv_send_total; + ktime_t sc_tv_status_total; + u32 sc_send_count; + u32 sc_recv_count; + ktime_t sc_tv_process_total; +#endif struct mutex sc_send_lock; }; @@ -206,4 +219,24 @@ struct o2net_status_wait { struct list_head ns_node_item; }; +#ifdef CONFIG_DEBUG_FS +/* just for state dumps */ +struct o2net_send_tracking { + struct list_head st_net_debug_item; + struct task_struct *st_task; + struct o2net_sock_container *st_sc; + u32 st_id; + u32 st_msg_type; + u32 st_msg_key; + u8 st_node; + ktime_t st_sock_time; + ktime_t st_send_time; + ktime_t st_status_time; +}; +#else +struct o2net_send_tracking { + u32 dummy; +}; +#endif /* CONFIG_DEBUG_FS */ + #endif /* O2CLUSTER_TCP_INTERNAL_H */ diff --git a/fs/ocfs2/cluster/ver.c b/fs/ocfs2/cluster/ver.c deleted file mode 100644 index a56eee6abad..00000000000 --- a/fs/ocfs2/cluster/ver.c +++ /dev/null @@ -1,42 +0,0 @@ -/* -*- mode: c; c-basic-offset: 8; -*- - * vim: noexpandtab sw=8 ts=8 sts=0: - * - * ver.c - * - * version string - * - * Copyright (C) 2002, 2005 Oracle. All rights reserved. - * - * This program is free software; you can redistribute it and/or - * modify it under the terms of the GNU General Public - * License as published by the Free Software Foundation; either - * version 2 of the License, or (at your option) any later version. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - * General Public License for more details. - * - * You should have received a copy of the GNU General Public - * License along with this program; if not, write to the - * Free Software Foundation, Inc., 59 Temple Place - Suite 330, - * Boston, MA 021110-1307, USA. - */ - -#include <linux/module.h> -#include <linux/kernel.h> - -#include "ver.h" - -#define CLUSTER_BUILD_VERSION "1.5.0" - -#define VERSION_STR "OCFS2 Node Manager " CLUSTER_BUILD_VERSION - -void cluster_print_version(void) -{ - printk(KERN_INFO "%s\n", VERSION_STR); -} - -MODULE_DESCRIPTION(VERSION_STR); - -MODULE_VERSION(CLUSTER_BUILD_VERSION); diff --git a/fs/ocfs2/cluster/ver.h b/fs/ocfs2/cluster/ver.h deleted file mode 100644 index 32554c3382c..00000000000 --- a/fs/ocfs2/cluster/ver.h +++ /dev/null @@ -1,31 +0,0 @@ -/* -*- mode: c; c-basic-offset: 8; -*- - * vim: noexpandtab sw=8 ts=8 sts=0: - * - * ver.h - * - * Function prototypes - * - * Copyright (C) 2005 Oracle. All rights reserved. - * - * This program is free software; you can redistribute it and/or - * modify it under the terms of the GNU General Public - * License as published by the Free Software Foundation; either - * version 2 of the License, or (at your option) any later version. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - * General Public License for more details. - * - * You should have received a copy of the GNU General Public - * License along with this program; if not, write to the - * Free Software Foundation, Inc., 59 Temple Place - Suite 330, - * Boston, MA 021110-1307, USA. - */ - -#ifndef O2CLUSTER_VER_H -#define O2CLUSTER_VER_H - -void cluster_print_version(void); - -#endif /* O2CLUSTER_VER_H */ |