X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Finclude%2Flustre_dlm.h;h=9edf487a5d04ca62004f3f3a2f85681d5b27e8d1;hb=5296239f81a4a568d1f41b3ddd43b47cf6b3aa30;hp=873a6a11c4012c3916521bd4ab271558f0ff8277;hpb=d4a31821ccd63fcc5197680300bdc37a9453a332;p=fs%2Flustre-release.git diff --git a/lustre/include/lustre_dlm.h b/lustre/include/lustre_dlm.h index 873a6a1..9edf487 100644 --- a/lustre/include/lustre_dlm.h +++ b/lustre/include/lustre_dlm.h @@ -1,6 +1,37 @@ /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*- - * (visit-tags-table FILE) * vim:expandtab:shiftwidth=8:tabstop=8: + * + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * 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 version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf + * + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. + * + * GPL HEADER END + */ +/* + * Copyright 2008 Sun Microsystems, Inc. All rights reserved + * Use is subject to license terms. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. */ #ifndef _LUSTRE_DLM_H__ @@ -21,14 +52,22 @@ #include #include #include /* for obd_export, for LDLM_DEBUG */ +#include /* for interval_node{}, ldlm_extent */ +#include struct obd_ops; struct obd_device; #define OBD_LDLM_DEVICENAME "ldlm" +#ifdef HAVE_BGL_SUPPORT +/* 1.5 times the maximum 128 tasks available in VN mode */ +#define LDLM_DEFAULT_LRU_SIZE 196 +#else #define LDLM_DEFAULT_LRU_SIZE (100 * num_online_cpus()) +#endif #define LDLM_DEFAULT_MAX_ALIVE (cfs_time_seconds(36000)) +#define LDLM_CTIME_AGE_LIMIT (10) typedef enum { ELDLM_OK = 0, @@ -43,8 +82,8 @@ typedef enum { } ldlm_error_t; typedef enum { - LDLM_NAMESPACE_SERVER = 0, - LDLM_NAMESPACE_CLIENT = 1 + LDLM_NAMESPACE_SERVER = 1 << 0, + LDLM_NAMESPACE_CLIENT = 1 << 1 } ldlm_side_t; #define LDLM_FL_LOCK_CHANGED 0x000001 /* extent, mode, or resource changed */ @@ -57,7 +96,8 @@ typedef enum { #define LDLM_FL_BLOCK_WAIT 0x000008 #define LDLM_FL_CBPENDING 0x000010 /* this lock is being destroyed */ -#define LDLM_FL_AST_SENT 0x000020 /* blocking or cancel packet was sent */ +#define LDLM_FL_AST_SENT 0x000020 /* blocking or cancel packet was + * queued for sending. */ #define LDLM_FL_WAIT_NOREPROC 0x000040 /* not a real flag, not saved in lock */ #define LDLM_FL_CANCEL 0x000080 /* cancellation callback already run */ @@ -105,21 +145,15 @@ typedef enum { * list. */ #define LDLM_FL_KMS_IGNORE 0x200000 -/* Don't drop lock covering mmapped file in LRU */ -#define LDLM_FL_NO_LRU 0x400000 - /* Immediatelly cancel such locks when they block some other locks. Send - cancel notification to original lock holder, but expect no reply. */ + * cancel notification to original lock holder, but expect no reply. This is + * for clients (like liblustre) that cannot be expected to reliably response + * to blocking ast. */ #define LDLM_FL_CANCEL_ON_BLOCK 0x800000 /* Flags flags inherited from parent lock when doing intents. */ #define LDLM_INHERIT_FLAGS (LDLM_FL_CANCEL_ON_BLOCK) -/* These are flags that are mapped into the flags and ASTs of blocking locks */ -#define LDLM_AST_DISCARD_DATA 0x80000000 /* Add FL_DISCARD to blocking ASTs */ -/* Flags sent in AST lock_flags to be mapped into the receiving lock. */ -#define LDLM_AST_FLAGS (LDLM_FL_DISCARD_DATA) - /* completion ast to be executed */ #define LDLM_FL_CP_REQD 0x1000000 @@ -130,12 +164,15 @@ typedef enum { * w/o involving separate thread. in order to decrease cs rate */ #define LDLM_FL_ATOMIC_CB 0x4000000 +/* Cancel lock asynchronously. See ldlm_cli_cancel_unused_resource. */ +#define LDLM_FL_ASYNC 0x8000000 + /* It may happen that a client initiate 2 operations, e.g. unlink and mkdir, * such that server send blocking ast for conflict locks to this client for * the 1st operation, whereas the 2nd operation has canceled this lock and * is waiting for rpc_lock which is taken by the 1st operation. * LDLM_FL_BL_AST is to be set by ldlm_callback_handler() to the lock not allow - * ELC code to cancel it. + * ELC code to cancel it. * LDLM_FL_BL_DONE is to be set by ldlm_cancel_callback() when lock cache is * droped to let ldlm_callback_handler() return EINVAL to the server. It is * used when ELC rpc is already prepared and is waiting for rpc_lock, too late @@ -143,20 +180,31 @@ typedef enum { #define LDLM_FL_BL_AST 0x10000000 #define LDLM_FL_BL_DONE 0x20000000 -/* Cancel lock asynchronously. See ldlm_cli_cancel_unused_resource. */ -#define LDLM_FL_ASYNC 0x40000000 +/* measure lock contention and return -EUSERS if locking contention is high */ +#define LDLM_FL_DENY_ON_CONTENTION 0x40000000 + +/* These are flags that are mapped into the flags and ASTs of blocking locks */ +#define LDLM_AST_DISCARD_DATA 0x80000000 /* Add FL_DISCARD to blocking ASTs */ + +/* Flags sent in AST lock_flags to be mapped into the receiving lock. */ +#define LDLM_AST_FLAGS (LDLM_FL_DISCARD_DATA) + +/* + * -------------------------------------------------------------------------- + * NOTE! Starting from this point, that is, LDLM_FL_* flags with values above + * 0x80000000 will not be sent over the wire. + * -------------------------------------------------------------------------- + */ + +/* Used for marking lock as an target for -EINTR while cp_ast sleep + * emulation + race with upcoming bl_ast. */ +#define LDLM_FL_FAIL_LOC 0x100000000ULL /* The blocking callback is overloaded to perform two functions. These flags * indicate which operation should be performed. */ #define LDLM_CB_BLOCKING 1 #define LDLM_CB_CANCELING 2 -/* position flag of skip list pointers */ -#define LDLM_SL_HEAD(skip_list) ((skip_list)->next != NULL) -#define LDLM_SL_TAIL(skip_list) ((skip_list)->prev != NULL) -#define LDLM_SL_EMPTY(skip_list) ((skip_list)->next == NULL && \ - (skip_list)->prev == NULL) - /* compatibility matrix */ #define LCK_COMPAT_EX LCK_NL #define LCK_COMPAT_PW (LCK_COMPAT_EX | LCK_CR) @@ -165,6 +213,7 @@ typedef enum { #define LCK_COMPAT_CR (LCK_COMPAT_CW | LCK_PR | LCK_PW) #define LCK_COMPAT_NL (LCK_COMPAT_CR | LCK_EX | LCK_GROUP) #define LCK_COMPAT_GROUP (LCK_GROUP | LCK_NL) +#define LCK_COMPAT_COS (LCK_COS) extern ldlm_mode_t lck_compat_array[]; @@ -173,9 +222,9 @@ static inline void lockmode_verify(ldlm_mode_t mode) LASSERT(mode > LCK_MINMODE && mode < LCK_MAXMODE); } -static inline int lockmode_compat(ldlm_mode_t exist, ldlm_mode_t new) +static inline int lockmode_compat(ldlm_mode_t exist_mode, ldlm_mode_t new_mode) { - return (lck_compat_array[exist] & new); + return (lck_compat_array[exist_mode] & new_mode); } /* @@ -217,69 +266,101 @@ struct ldlm_lock; struct ldlm_resource; struct ldlm_namespace; -typedef int (*ldlm_pool_recalc_t)(struct ldlm_pool *pl); - -typedef int (*ldlm_pool_shrink_t)(struct ldlm_pool *pl, - int nr, unsigned int gfp_mask); - -enum { - LDLM_POOL_CTL_RECALC = 1 << 0, /* Pool recalc is enabled */ - LDLM_POOL_CTL_SHRINK = 1 << 1, /* Pool shrink is enabled */ - LDLM_POOL_CTL_FULL = (LDLM_POOL_CTL_RECALC | LDLM_POOL_CTL_SHRINK) +struct ldlm_pool_ops { + int (*po_recalc)(struct ldlm_pool *pl); + int (*po_shrink)(struct ldlm_pool *pl, int nr, + unsigned int gfp_mask); + int (*po_setup)(struct ldlm_pool *pl, int limit); }; -/* One second for pools thread check interval. */ +/** + * One second for pools thread check interval. Each pool has own period. + */ #define LDLM_POOLS_THREAD_PERIOD (1) -/* 5% margin for modest pools. See ldlm_pool.c for details. */ +/** + * 5% margin for modest pools. See ldlm_pool.c for details. + */ #define LDLM_POOLS_MODEST_MARGIN (5) -/* A change to SLV in % after which we want to wake up pools thread asap. */ -#define LDLM_POOLS_FAST_SLV_CHANGE (5) +/** + * Default recalc period for server side pools in sec. + */ +#define LDLM_POOL_SRV_DEF_RECALC_PERIOD (1) + +/** + * Default recalc period for client side pools in sec. + */ +#define LDLM_POOL_CLI_DEF_RECALC_PERIOD (10) struct ldlm_pool { - /* Common pool fields */ - cfs_proc_dir_entry_t *pl_proc_dir; /* Pool proc directory. */ - char pl_name[100]; /* Pool name, should be long - * enough to contain complex - * proc entry name. */ - spinlock_t pl_lock; /* Lock for protecting slv/clv - * updates. */ - atomic_t pl_limit; /* Number of allowed locks in - * in pool, both, client and - * server side. */ - atomic_t pl_granted; /* Number of granted locks. */ - atomic_t pl_grant_rate; /* Grant rate per T. */ - atomic_t pl_cancel_rate; /* Cancel rate per T. */ - atomic_t pl_grant_speed; /* Grant speed (GR - CR) per T. */ - __u64 pl_server_lock_volume; /* Server lock volume. Protected - * by pl_lock. */ - cfs_time_t pl_update_time; /* Time when last slv from server - * was obtained. */ - ldlm_pool_recalc_t pl_recalc; /* Recalc callback func pointer. */ - ldlm_pool_shrink_t pl_shrink; /* Shrink callback func pointer. */ - int pl_control; /* Pool features mask */ - - /* Server side pool fields */ - atomic_t pl_grant_plan; /* Planned number of granted - * locks for next T. */ - atomic_t pl_grant_step; /* Grant plan step for next T. */ - - /* Client side pool related fields */ - atomic_t pl_lock_volume_factor; /* Lock volume factor. */ - struct lprocfs_stats *pl_stats; /* Pool statistics. */ + /** + * Pool proc directory. + */ + cfs_proc_dir_entry_t *pl_proc_dir; + /** + * Pool name, should be long enough to contain compound proc entry name. + */ + char pl_name[100]; + /** + * Lock for protecting slv/clv updates. + */ + spinlock_t pl_lock; + /** + * Number of allowed locks in in pool, both, client and server side. + */ + atomic_t pl_limit; + /** + * Number of granted locks in + */ + atomic_t pl_granted; + /** + * Grant rate per T. + */ + atomic_t pl_grant_rate; + /** + * Cancel rate per T. + */ + atomic_t pl_cancel_rate; + /** + * Grant speed (GR-CR) per T. + */ + atomic_t pl_grant_speed; + /** + * Server lock volume. Protected by pl_lock. + */ + __u64 pl_server_lock_volume; + /** + * Current biggest client lock volume. Protected by pl_lock. + */ + __u64 pl_client_lock_volume; + /** + * Lock volume factor. SLV on client is calculated as following: + * server_slv * lock_volume_factor. + */ + atomic_t pl_lock_volume_factor; + /** + * Time when last slv from server was obtained. + */ + time_t pl_recalc_time; + /** + * Recalc period for pool. + */ + time_t pl_recalc_period; + /** + * Recalc and shrink ops. + */ + struct ldlm_pool_ops *pl_ops; + /** + * Number of planned locks for next period. + */ + int pl_grant_plan; + /** + * Pool statistics. + */ + struct lprocfs_stats *pl_stats; }; -static inline int pool_recalc_enabled(struct ldlm_pool *pl) -{ - return pl->pl_control & LDLM_POOL_CTL_RECALC; -} - -static inline int pool_shrink_enabled(struct ldlm_pool *pl) -{ - return pl->pl_control & LDLM_POOL_CTL_SHRINK; -} - typedef int (*ldlm_res_policy)(struct ldlm_namespace *, struct ldlm_lock **, void *req_cookie, ldlm_mode_t mode, int flags, void *data); @@ -295,24 +376,76 @@ typedef enum { LDLM_NAMESPACE_MODEST = 1 << 1 } ldlm_appetite_t; +/* + * Default values for the "max_nolock_size", "contention_time" and + * "contended_locks" namespace tunables. + */ +#define NS_DEFAULT_MAX_NOLOCK_BYTES 0 +#define NS_DEFAULT_CONTENTION_SECONDS 2 +#define NS_DEFAULT_CONTENDED_LOCKS 32 + struct ldlm_namespace { + /** + * Namespace name. Used for logging, etc. + */ char *ns_name; - ldlm_side_t ns_client; /* is this a client-side lock tree? */ - __u64 ns_connect_flags; /* client side connect flags - * supported by server */ - struct list_head *ns_hash; /* hash table for ns */ + + /** + * Is this a client-side lock tree? + */ + ldlm_side_t ns_client; + + /** + * Namespce connect flags supported by server (may be changed via proc, + * lru resize may be disabled/enabled). + */ + __u64 ns_connect_flags; + + /** + * Client side orig connect flags supported by server. + */ + __u64 ns_orig_connect_flags; + + /** + * Hash table for namespace. + */ + struct list_head *ns_hash; spinlock_t ns_hash_lock; - __u32 ns_refcount; /* count of resources in the hash */ - struct list_head ns_root_list; /* all root resources in ns */ - struct list_head ns_list_chain; /* position in global NS list */ - struct list_head ns_unused_list; /* all root resources in ns */ + /** + * Count of resources in the hash. + */ + __u32 ns_refcount; + + /** + * All root resources in namespace. + */ + struct list_head ns_root_list; + + /** + * Position in global namespace list. + */ + struct list_head ns_list_chain; + + /** + * All root resources in namespace. + */ + struct list_head ns_unused_list; int ns_nr_unused; spinlock_t ns_unused_lock; unsigned int ns_max_unused; unsigned int ns_max_age; - cfs_time_t ns_next_dump; /* next debug dump, jiffies */ + unsigned int ns_timeouts; + /** + * Seconds. + */ + unsigned int ns_ctime_age_limit; + + /** + * Next debug dump, jiffies. + */ + cfs_time_t ns_next_dump; atomic_t ns_locks; __u64 ns_resources; @@ -322,12 +455,56 @@ struct ldlm_namespace { cfs_waitq_t ns_waitq; struct ldlm_pool ns_pool; ldlm_appetite_t ns_appetite; + + /** + * If more than @ns_contented_locks found, the resource considered + * as contended. + */ + unsigned ns_contended_locks; + + /** + * The resource remembers contended state during @ns_contention_time, + * in seconds. + */ + unsigned ns_contention_time; + + /** + * Limit size of nolock requests, in bytes. + */ + unsigned ns_max_nolock_size; + + /** + * Backward link to obd, required for ldlm pool to store new SLV. + */ + struct obd_device *ns_obd; + + struct adaptive_timeout ns_at_estimate;/* estimated lock callback time*/ }; +static inline int ns_is_client(struct ldlm_namespace *ns) +{ + LASSERT(ns != NULL); + LASSERT(!(ns->ns_client & ~(LDLM_NAMESPACE_CLIENT | + LDLM_NAMESPACE_SERVER))); + LASSERT(ns->ns_client == LDLM_NAMESPACE_CLIENT || + ns->ns_client == LDLM_NAMESPACE_SERVER); + return ns->ns_client == LDLM_NAMESPACE_CLIENT; +} + +static inline int ns_is_server(struct ldlm_namespace *ns) +{ + LASSERT(ns != NULL); + LASSERT(!(ns->ns_client & ~(LDLM_NAMESPACE_CLIENT | + LDLM_NAMESPACE_SERVER))); + LASSERT(ns->ns_client == LDLM_NAMESPACE_CLIENT || + ns->ns_client == LDLM_NAMESPACE_SERVER); + return ns->ns_client == LDLM_NAMESPACE_SERVER; +} + static inline int ns_connect_lru_resize(struct ldlm_namespace *ns) { LASSERT(ns != NULL); - return ns->ns_connect_flags & OBD_CONNECT_LRU_RESIZE; + return !!(ns->ns_connect_flags & OBD_CONNECT_LRU_RESIZE); } /* @@ -336,7 +513,7 @@ static inline int ns_connect_lru_resize(struct ldlm_namespace *ns) * */ -#define RES_HASH_BITS 10 +#define RES_HASH_BITS 12 #define RES_HASH_SIZE (1UL << RES_HASH_BITS) #define RES_HASH_MASK (RES_HASH_SIZE - 1) @@ -348,78 +525,191 @@ typedef int (*ldlm_blocking_callback)(struct ldlm_lock *lock, typedef int (*ldlm_completion_callback)(struct ldlm_lock *lock, int flags, void *data); typedef int (*ldlm_glimpse_callback)(struct ldlm_lock *lock, void *data); +typedef unsigned long (*ldlm_weigh_callback)(struct ldlm_lock *lock); -struct ldlm_lock { - struct portals_handle l_handle; // must be first in the structure - atomic_t l_refc; - - /* internal spinlock protects l_resource. we should hold this lock - * first before grabbing res_lock.*/ - spinlock_t l_lock; - - /* ldlm_lock_change_resource() can change this */ - struct ldlm_resource *l_resource; - - /* protected by ns_hash_lock. FIXME */ - struct list_head l_lru; - - /* protected by lr_lock */ - struct list_head l_res_link; // position in one of three res lists - - struct list_head l_sl_mode; // skip pointer for request mode - struct list_head l_sl_policy; // skip pointer for inodebits - - /* protected by led_lock */ - struct list_head l_export_chain; // per-export chain of locks - - /* protected by lr_lock */ - ldlm_mode_t l_req_mode; - ldlm_mode_t l_granted_mode; +/* Interval node data for each LDLM_EXTENT lock */ +struct ldlm_interval { + struct interval_node li_node; /* node for tree mgmt */ + struct list_head li_group; /* the locks which have the same + * policy - group of the policy */ +}; +#define to_ldlm_interval(n) container_of(n, struct ldlm_interval, li_node) + +/* the interval tree must be accessed inside the resource lock. */ +struct ldlm_interval_tree { + /* tree size, this variable is used to count + * granted PW locks in ldlm_extent_policy()*/ + int lit_size; + ldlm_mode_t lit_mode; /* lock mode */ + struct interval_node *lit_root; /* actually ldlm_interval */ +}; +struct ldlm_lock { + /** + * Must be first in the structure. + */ + struct portals_handle l_handle; + /** + * Lock reference count. + */ + atomic_t l_refc; + /** + * Internal spinlock protects l_resource. we should hold this lock + * first before grabbing res_lock. + */ + spinlock_t l_lock; + /** + * ldlm_lock_change_resource() can change this. + */ + struct ldlm_resource *l_resource; + /** + * Protected by ns_hash_lock. List item for client side lru list. + */ + struct list_head l_lru; + /** + * Protected by lr_lock, linkage to resource's lock queues. + */ + struct list_head l_res_link; + /** + * Tree node for ldlm_extent. + */ + struct ldlm_interval *l_tree_node; + /** + * Protected by per-bucket exp->exp_lock_hash locks. Per export hash + * of locks. + */ + struct hlist_node l_exp_hash; + /** + * Protected by lr_lock. Requested mode. + */ + ldlm_mode_t l_req_mode; + /** + * Granted mode, also protected by lr_lock. + */ + ldlm_mode_t l_granted_mode; + /** + * Lock enqueue completion handler. + */ ldlm_completion_callback l_completion_ast; + /** + * Lock blocking ast handler. + */ ldlm_blocking_callback l_blocking_ast; + /** + * Lock glimpse handler. + */ ldlm_glimpse_callback l_glimpse_ast; - - struct obd_export *l_export; - struct obd_export *l_conn_export; - - struct lustre_handle l_remote_handle; - ldlm_policy_data_t l_policy_data; - - /* protected by lr_lock */ - __u32 l_flags; + ldlm_weigh_callback l_weigh_ast; + + /** + * Lock export. + */ + struct obd_export *l_export; + /** + * Lock connection export. + */ + struct obd_export *l_conn_export; + + /** + * Remote lock handle. + */ + struct lustre_handle l_remote_handle; + + ldlm_policy_data_t l_policy_data; + + /* + * Protected by lr_lock. Various counters: readers, writers, etc. + */ + __u64 l_flags; __u32 l_readers; __u32 l_writers; + /* + * Set for locks that were removed from class hash table and will be + * destroyed when last reference to them is released. Set by + * ldlm_lock_destroy_internal(). + * + * Protected by lock and resource locks. + */ __u8 l_destroyed; - /* If the lock is granted, a process sleeps on this waitq to learn when + /** + * If the lock is granted, a process sleeps on this waitq to learn when * it's no longer in use. If the lock is not granted, a process sleeps - * on this waitq to learn when it becomes granted. */ + * on this waitq to learn when it becomes granted. + */ cfs_waitq_t l_waitq; - struct timeval l_enqueued_time; - cfs_time_t l_last_used; /* jiffies */ + /** + * Seconds. it will be updated if there is any activity related to + * the lock, e.g. enqueue the lock or send block AST. + */ + cfs_time_t l_last_activity; + + /** + * Jiffies. Should be converted to time if needed. + */ + cfs_time_t l_last_used; + struct ldlm_extent l_req_extent; - /* Client-side-only members */ - __u32 l_lvb_len; /* temporary storage for */ - void *l_lvb_data; /* an LVB received during */ - void *l_lvb_swabber; /* an enqueue */ + /* + * Client-side-only members. + */ + + /** + * Temporary storage for an LVB received during an enqueue operation. + */ + __u32 l_lvb_len; + void *l_lvb_data; + void *l_lvb_swabber; + void *l_ast_data; + spinlock_t l_extents_list_lock; + struct list_head l_extents_list; - /* Server-side-only members */ + struct list_head l_cache_locks_list; - /* protected by elt_lock */ - struct list_head l_pending_chain; /* callbacks pending */ - cfs_time_t l_callback_timeout; /* jiffies */ + /* + * Server-side-only members. + */ - __u32 l_pid; /* pid which created this lock */ + /** connection cookie for the client originated the operation. */ + __u64 l_client_cookie; - /* for ldlm_add_ast_work_item() */ + /** + * Protected by elt_lock. Callbacks pending. + */ + struct list_head l_pending_chain; + + cfs_time_t l_callback_timeout; + + /** + * Pid which created this lock. + */ + __u32 l_pid; + + /** + * For ldlm_add_ast_work_item(). + */ struct list_head l_bl_ast; + /** + * For ldlm_add_ast_work_item(). + */ struct list_head l_cp_ast; + /** + * For ldlm_add_ast_work_item(). + */ + struct list_head l_rk_ast; + struct ldlm_lock *l_blocking_lock; int l_bl_ast_run; + + /** + * Protected by lr_lock, linkages to "skip lists". + */ + struct list_head l_sl_mode; + struct list_head l_sl_policy; + struct lu_ref l_reference; }; struct ldlm_resource { @@ -442,10 +732,19 @@ struct ldlm_resource { struct ldlm_res_id lr_name; atomic_t lr_refcount; + struct ldlm_interval_tree lr_itree[LCK_MODE_NUM]; /* interval trees*/ + /* Server-side-only lock value block elements */ struct semaphore lr_lvb_sem; __u32 lr_lvb_len; void *lr_lvb_data; + + /* when the resource was considered as contended */ + cfs_time_t lr_contention_time; + /** + * List of references to this resource. For debugging. + */ + struct lu_ref lr_reference; }; struct ldlm_ast_work { @@ -462,10 +761,12 @@ struct ldlm_ast_work { struct ldlm_enqueue_info { __u32 ei_type; /* Type of the lock being enqueued. */ __u32 ei_mode; /* Mode of the lock being enqueued. */ - void *ei_cb_bl; /* Different callbacks for lock handling (blocking, */ - void *ei_cb_cp; /* completion, glimpse) */ - void *ei_cb_gl; + void *ei_cb_bl; /* blocking lock callback */ + void *ei_cb_cp; /* lock completion callback */ + void *ei_cb_gl; /* lock glimpse callback */ + void *ei_cb_wg; /* lock weigh callback */ void *ei_cbdata; /* Data to be passed into callbacks. */ + short ei_async:1; /* async request */ }; extern struct obd_ops ldlm_obd_ops; @@ -493,21 +794,26 @@ void _ldlm_lock_debug(struct ldlm_lock *lock, __u32 mask, ...) __attribute__ ((format (printf, 4, 5))); -#define LDLM_ERROR(lock, fmt, a...) do { \ +#define LDLM_DEBUG_LIMIT(mask, lock, fmt, a...) do { \ static cfs_debug_limit_state_t _ldlm_cdls; \ - ldlm_lock_debug(&_ldlm_cdls, D_ERROR, lock, \ + ldlm_lock_debug(&_ldlm_cdls, mask, lock, \ __FILE__, __FUNCTION__, __LINE__, \ "### " fmt , ##a); \ } while (0) +#define LDLM_ERROR(lock, fmt, a...) LDLM_DEBUG_LIMIT(D_ERROR, lock, fmt, ## a) +#define LDLM_WARN(lock, fmt, a...) LDLM_DEBUG_LIMIT(D_WARNING, lock, fmt, ## a) + #define LDLM_DEBUG(lock, fmt, a...) do { \ ldlm_lock_debug(NULL, D_DLMTRACE, lock, \ __FILE__, __FUNCTION__, __LINE__, \ "### " fmt , ##a); \ } while (0) -#else -#define LDLM_DEBUG(lock, fmt, a...) ((void)0) -#define LDLM_ERROR(lock, fmt, a...) ((void)0) +#else /* !LIBCFS_DEBUG */ +# define LDLM_DEBUG(lock, fmt, a...) ((void)0) +# define LDLM_ERROR(lock, fmt, a...) ((void)0) +# define ldlm_lock_debuf(cdls, level, lock, file, func, line, fmt, a...) \ + ((void)0) #endif #define LDLM_DEBUG_NOLOCK(format, a...) \ @@ -557,10 +863,12 @@ int ldlm_handle_cancel(struct ptlrpc_request *req); int ldlm_request_cancel(struct ptlrpc_request *req, const struct ldlm_request *dlm_req, int first); int ldlm_del_waiting_lock(struct ldlm_lock *lock); -int ldlm_refresh_waiting_lock(struct ldlm_lock *lock); +int ldlm_refresh_waiting_lock(struct ldlm_lock *lock, int timeout); void ldlm_revoke_export_locks(struct obd_export *exp); -int ldlm_get_ref(ldlm_side_t client); -void ldlm_put_ref(ldlm_side_t client, int force); +int ldlm_get_ref(void); +void ldlm_put_ref(void); +int ldlm_init_export(struct obd_export *exp); +void ldlm_destroy_export(struct obd_export *exp); /* ldlm_lock.c */ ldlm_processing_policy ldlm_get_processing_policy(struct ldlm_resource *res); @@ -569,16 +877,27 @@ void ldlm_lock2handle(const struct ldlm_lock *lock, struct lustre_handle *lockh); struct ldlm_lock *__ldlm_handle2lock(const struct lustre_handle *, int flags); void ldlm_cancel_callback(struct ldlm_lock *); -int ldlm_lock_set_data(struct lustre_handle *, void *data); int ldlm_lock_remove_from_lru(struct ldlm_lock *); -struct ldlm_lock *ldlm_handle2lock_ns(struct ldlm_namespace *, - const struct lustre_handle *); static inline struct ldlm_lock *ldlm_handle2lock(const struct lustre_handle *h) { return __ldlm_handle2lock(h, 0); } +#define LDLM_LOCK_REF_DEL(lock) \ + lu_ref_del(&lock->l_reference, "handle", cfs_current()) + +static inline struct ldlm_lock * +ldlm_handle2lock_long(const struct lustre_handle *h, int flags) +{ + struct ldlm_lock *lock; + + lock = __ldlm_handle2lock(h, flags); + if (lock != NULL) + LDLM_LOCK_REF_DEL(lock); + return lock; +} + static inline int ldlm_res_lvbo_update(struct ldlm_resource *res, struct lustre_msg *m, int buf_idx, int increase) @@ -591,8 +910,27 @@ static inline int ldlm_res_lvbo_update(struct ldlm_resource *res, return 0; } +int ldlm_error2errno(ldlm_error_t error); +ldlm_error_t ldlm_errno2error(int err_no); /* don't call it `errno': this + * confuses user-space. */ + +/** + * Release a temporary lock reference obtained by ldlm_handle2lock() or + * __ldlm_handle2lock(). + */ #define LDLM_LOCK_PUT(lock) \ do { \ + LDLM_LOCK_REF_DEL(lock); \ + /*LDLM_DEBUG((lock), "put");*/ \ + ldlm_lock_put(lock); \ +} while (0) + +/** + * Release a lock reference obtained by some other means (see + * LDLM_LOCK_PUT()). + */ +#define LDLM_LOCK_RELEASE(lock) \ +do { \ /*LDLM_DEBUG((lock), "put");*/ \ ldlm_lock_put(lock); \ } while (0) @@ -612,7 +950,7 @@ do { \ if (c-- == 0) \ break; \ list_del_init(&_lock->member); \ - LDLM_LOCK_PUT(_lock); \ + LDLM_LOCK_RELEASE(_lock); \ } \ LASSERT(c <= 0); \ }) @@ -622,15 +960,18 @@ void ldlm_lock_put(struct ldlm_lock *lock); void ldlm_lock_destroy(struct ldlm_lock *lock); void ldlm_lock2desc(struct ldlm_lock *lock, struct ldlm_lock_desc *desc); void ldlm_lock_addref(struct lustre_handle *lockh, __u32 mode); +int ldlm_lock_addref_try(struct lustre_handle *lockh, __u32 mode); void ldlm_lock_decref(struct lustre_handle *lockh, __u32 mode); void ldlm_lock_decref_and_cancel(struct lustre_handle *lockh, __u32 mode); void ldlm_lock_allow_match(struct ldlm_lock *lock); -int ldlm_lock_match(struct ldlm_namespace *ns, int flags, - const struct ldlm_res_id *, - ldlm_type_t type, ldlm_policy_data_t *, ldlm_mode_t mode, - struct lustre_handle *); +void ldlm_lock_allow_match_locked(struct ldlm_lock *lock); +ldlm_mode_t ldlm_lock_match(struct ldlm_namespace *ns, int flags, + const struct ldlm_res_id *, ldlm_type_t type, + ldlm_policy_data_t *, ldlm_mode_t mode, + struct lustre_handle *, int unref); struct ldlm_resource *ldlm_lock_convert(struct ldlm_lock *lock, int new_mode, - int *flags); + __u32 *flags); +void ldlm_lock_downgrade(struct ldlm_lock *lock, int new_mode); void ldlm_lock_cancel(struct ldlm_lock *lock); void ldlm_cancel_locks_for_export(struct obd_export *export); void ldlm_reprocess_all(struct ldlm_resource *res); @@ -640,16 +981,20 @@ void ldlm_lock_dump_handle(int level, struct lustre_handle *); void ldlm_unlink_lock_skiplist(struct ldlm_lock *req); /* resource.c */ -struct ldlm_namespace *ldlm_namespace_new(char *name, ldlm_side_t client, - ldlm_appetite_t apt); +struct ldlm_namespace * +ldlm_namespace_new(struct obd_device *obd, char *name, + ldlm_side_t client, ldlm_appetite_t apt); int ldlm_namespace_cleanup(struct ldlm_namespace *ns, int flags); -int ldlm_namespace_free(struct ldlm_namespace *ns, int force); -void ldlm_namespace_move(struct ldlm_namespace *ns, ldlm_side_t client); -struct ldlm_namespace *ldlm_namespace_first(ldlm_side_t client); +void ldlm_namespace_free(struct ldlm_namespace *ns, + struct obd_import *imp, int force); +void ldlm_namespace_register(struct ldlm_namespace *ns, ldlm_side_t client); +void ldlm_namespace_unregister(struct ldlm_namespace *ns, ldlm_side_t client); +void ldlm_namespace_move_locked(struct ldlm_namespace *ns, ldlm_side_t client); +struct ldlm_namespace *ldlm_namespace_first_locked(ldlm_side_t client); +void ldlm_namespace_get_locked(struct ldlm_namespace *ns); +void ldlm_namespace_put_locked(struct ldlm_namespace *ns, int wakeup); void ldlm_namespace_get(struct ldlm_namespace *ns); void ldlm_namespace_put(struct ldlm_namespace *ns, int wakeup); -void ldlm_namespace_get_nolock(struct ldlm_namespace *ns); -void ldlm_namespace_put_nolock(struct ldlm_namespace *ns, int wakeup); int ldlm_proc_setup(void); #ifdef LPROCFS void ldlm_proc_cleanup(void); @@ -674,17 +1019,28 @@ void ldlm_resource_dump(int level, struct ldlm_resource *); int ldlm_lock_change_resource(struct ldlm_namespace *, struct ldlm_lock *, const struct ldlm_res_id *); +#define LDLM_RESOURCE_ADDREF(res) do { \ + lu_ref_add_atomic(&(res)->lr_reference, __FUNCTION__, cfs_current()); \ +} while (0) + +#define LDLM_RESOURCE_DELREF(res) do { \ + lu_ref_del(&(res)->lr_reference, __FUNCTION__, cfs_current()); \ +} while (0) + struct ldlm_callback_suite { ldlm_completion_callback lcs_completion; ldlm_blocking_callback lcs_blocking; ldlm_glimpse_callback lcs_glimpse; + ldlm_weigh_callback lcs_weigh; }; /* ldlm_request.c */ int ldlm_expired_completion_wait(void *data); +int ldlm_blocking_ast_nocheck(struct ldlm_lock *lock); int ldlm_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc, void *data, int flag); int ldlm_glimpse_ast(struct ldlm_lock *lock, void *reqp); +int ldlm_completion_ast_async(struct ldlm_lock *lock, int flags, void *data); int ldlm_completion_ast(struct ldlm_lock *lock, int flags, void *data); int ldlm_cli_enqueue(struct obd_export *exp, struct ptlrpc_request **reqp, struct ldlm_enqueue_info *einfo, @@ -692,9 +1048,14 @@ int ldlm_cli_enqueue(struct obd_export *exp, struct ptlrpc_request **reqp, ldlm_policy_data_t *policy, int *flags, void *lvb, __u32 lvb_len, void *lvb_swabber, struct lustre_handle *lockh, int async); -struct ptlrpc_request *ldlm_prep_enqueue_req(struct obd_export *exp, - int bufcount, int *size, - struct list_head *head, int count); +int ldlm_prep_enqueue_req(struct obd_export *exp, + struct ptlrpc_request *req, + struct list_head *cancels, + int count); +int ldlm_prep_elc_req(struct obd_export *exp, + struct ptlrpc_request *req, + int version, int opc, int canceloff, + struct list_head *cancels, int count); int ldlm_handle_enqueue0(struct ldlm_namespace *ns, struct ptlrpc_request *req, const struct ldlm_request *dlm_req, const struct ldlm_callback_suite *cbs); @@ -711,10 +1072,11 @@ int ldlm_cli_enqueue_local(struct ldlm_namespace *ns, ldlm_completion_callback completion, ldlm_glimpse_callback glimpse, void *data, __u32 lvb_len, void *lvb_swabber, + const __u64 *client_cookie, struct lustre_handle *lockh); int ldlm_server_ast(struct lustre_handle *lockh, struct ldlm_lock_desc *new, void *data, __u32 data_len); -int ldlm_cli_convert(struct lustre_handle *, int new_mode, int *flags); +int ldlm_cli_convert(struct lustre_handle *, int new_mode, __u32 *flags); int ldlm_cli_update_pool(struct ptlrpc_request *req); int ldlm_handle_convert0(struct ptlrpc_request *req, const struct ldlm_request *dlm_req); @@ -727,16 +1089,14 @@ int ldlm_cli_cancel_unused_resource(struct ldlm_namespace *ns, ldlm_mode_t mode, int flags, void *opaque); int ldlm_cli_cancel_req(struct obd_export *exp, struct list_head *head, int count, int flags); -int ldlm_cli_join_lru(struct ldlm_namespace *, - const struct ldlm_res_id *, int join); int ldlm_cancel_resource_local(struct ldlm_resource *res, struct list_head *cancels, ldlm_policy_data_t *policy, ldlm_mode_t mode, int lock_flags, - int flags, void *opaque); + int cancel_flags, void *opaque); int ldlm_cli_cancel_list(struct list_head *head, int count, - struct ptlrpc_request *req, int off, int flags); - + struct ptlrpc_request *req, int flags); + /* mds/handler.c */ /* This has to be here because recursive inclusion sucks. */ int intent_disposition(struct ldlm_reply *rep, int flag); @@ -753,11 +1113,27 @@ void intent_set_disposition(struct ldlm_reply *rep, int flag); #define IOC_LDLM_REGRESS_STOP _IOWR('f', 43, long) #define IOC_LDLM_MAX_NR 43 +/** + * "Modes" of acquiring lock_res, necessary to tell lockdep that taking more + * than one lock_res is dead-lock safe. + */ +enum lock_res_type { + LRT_NORMAL, + LRT_NEW +}; + static inline void lock_res(struct ldlm_resource *res) { spin_lock(&res->lr_lock); } +static inline void lock_res_nested(struct ldlm_resource *res, + enum lock_res_type mode) +{ + spin_lock_nested(&res->lr_lock, mode); +} + + static inline void unlock_res(struct ldlm_resource *res) { spin_unlock(&res->lr_lock); @@ -772,23 +1148,24 @@ struct ldlm_resource * lock_res_and_lock(struct ldlm_lock *lock); void unlock_res_and_lock(struct ldlm_lock *lock); /* ldlm_pool.c */ -int ldlm_pools_init(ldlm_side_t client); void ldlm_pools_recalc(ldlm_side_t client); +int ldlm_pools_init(void); void ldlm_pools_fini(void); -void ldlm_pools_wakeup(void); -int ldlm_pool_init(struct ldlm_pool *pl, struct ldlm_namespace *ns, +int ldlm_pool_init(struct ldlm_pool *pl, struct ldlm_namespace *ns, int idx, ldlm_side_t client); -int ldlm_pool_shrink(struct ldlm_pool *pl, int nr, +int ldlm_pool_shrink(struct ldlm_pool *pl, int nr, unsigned int gfp_mask); void ldlm_pool_fini(struct ldlm_pool *pl); -int ldlm_pool_setup(struct ldlm_pool *pl, __u32 limit); +int ldlm_pool_setup(struct ldlm_pool *pl, int limit); int ldlm_pool_recalc(struct ldlm_pool *pl); +__u32 ldlm_pool_get_lvf(struct ldlm_pool *pl); __u64 ldlm_pool_get_slv(struct ldlm_pool *pl); +__u64 ldlm_pool_get_clv(struct ldlm_pool *pl); __u32 ldlm_pool_get_limit(struct ldlm_pool *pl); void ldlm_pool_set_slv(struct ldlm_pool *pl, __u64 slv); +void ldlm_pool_set_clv(struct ldlm_pool *pl, __u64 clv); void ldlm_pool_set_limit(struct ldlm_pool *pl, __u32 limit); void ldlm_pool_add(struct ldlm_pool *pl, struct ldlm_lock *lock); void ldlm_pool_del(struct ldlm_pool *pl, struct ldlm_lock *lock); #endif -