X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Finclude%2Flustre_net.h;h=a7f37c21c8a2ba7ccdc35fd61648fd8ffde494ab;hb=892078e3b566c04471e7dcf2c28e66f2f3584f93;hp=dbeea10decf054553924b159aab7cd30847abde9;hpb=0f8b7f951e7f43dbf389a431afea2091388a805e;p=fs%2Flustre-release.git diff --git a/lustre/include/lustre_net.h b/lustre/include/lustre_net.h index dbeea10..a7f37c2 100644 --- a/lustre/include/lustre_net.h +++ b/lustre/include/lustre_net.h @@ -27,7 +27,7 @@ * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved. * Use is subject to license terms. * - * Copyright (c) 2010, 2012, Intel Corporation. + * Copyright (c) 2010, 2014, Intel Corporation. */ /* * This file is part of Lustre, http://www.lustre.org/ @@ -55,19 +55,10 @@ * @{ */ -#if defined(__linux__) -#include -#elif defined(__APPLE__) -#include -#elif defined(__WINNT__) -#include -#else -#error Unsupported operating system. -#endif - +#include #include -// #include -#include +#include +#include #include #include #include @@ -75,7 +66,6 @@ #include #include #include - #include #include @@ -108,30 +98,28 @@ */ #define PTLRPC_MAX_BRW_BITS (LNET_MTU_BITS + PTLRPC_BULK_OPS_BITS) #define PTLRPC_MAX_BRW_SIZE (1 << PTLRPC_MAX_BRW_BITS) -#define PTLRPC_MAX_BRW_PAGES (PTLRPC_MAX_BRW_SIZE >> CFS_PAGE_SHIFT) +#define PTLRPC_MAX_BRW_PAGES (PTLRPC_MAX_BRW_SIZE >> PAGE_CACHE_SHIFT) #define ONE_MB_BRW_SIZE (1 << LNET_MTU_BITS) #define MD_MAX_BRW_SIZE (1 << LNET_MTU_BITS) -#define MD_MAX_BRW_PAGES (MD_MAX_BRW_SIZE >> CFS_PAGE_SHIFT) +#define MD_MAX_BRW_PAGES (MD_MAX_BRW_SIZE >> PAGE_CACHE_SHIFT) #define DT_MAX_BRW_SIZE PTLRPC_MAX_BRW_SIZE -#define DT_MAX_BRW_PAGES (DT_MAX_BRW_SIZE >> CFS_PAGE_SHIFT) +#define DT_MAX_BRW_PAGES (DT_MAX_BRW_SIZE >> PAGE_CACHE_SHIFT) #define OFD_MAX_BRW_SIZE (1 << LNET_MTU_BITS) /* When PAGE_SIZE is a constant, we can check our arithmetic here with cpp! */ -#ifdef __KERNEL__ -# if ((PTLRPC_MAX_BRW_PAGES & (PTLRPC_MAX_BRW_PAGES - 1)) != 0) -# error "PTLRPC_MAX_BRW_PAGES isn't a power of two" -# endif -# if (PTLRPC_MAX_BRW_SIZE != (PTLRPC_MAX_BRW_PAGES * CFS_PAGE_SIZE)) -# error "PTLRPC_MAX_BRW_SIZE isn't PTLRPC_MAX_BRW_PAGES * CFS_PAGE_SIZE" -# endif -# if (PTLRPC_MAX_BRW_SIZE > LNET_MTU * PTLRPC_BULK_OPS_COUNT) -# error "PTLRPC_MAX_BRW_SIZE too big" -# endif -# if (PTLRPC_MAX_BRW_PAGES > LNET_MAX_IOV * PTLRPC_BULK_OPS_COUNT) -# error "PTLRPC_MAX_BRW_PAGES too big" -# endif -#endif /* __KERNEL__ */ +#if ((PTLRPC_MAX_BRW_PAGES & (PTLRPC_MAX_BRW_PAGES - 1)) != 0) +# error "PTLRPC_MAX_BRW_PAGES isn't a power of two" +#endif +#if (PTLRPC_MAX_BRW_SIZE != (PTLRPC_MAX_BRW_PAGES * PAGE_CACHE_SIZE)) +# error "PTLRPC_MAX_BRW_SIZE isn't PTLRPC_MAX_BRW_PAGES * PAGE_CACHE_SIZE" +#endif +#if (PTLRPC_MAX_BRW_SIZE > LNET_MTU * PTLRPC_BULK_OPS_COUNT) +# error "PTLRPC_MAX_BRW_SIZE too big" +#endif +#if (PTLRPC_MAX_BRW_PAGES > LNET_MAX_IOV * PTLRPC_BULK_OPS_COUNT) +# error "PTLRPC_MAX_BRW_PAGES too big" +#endif #define PTLRPC_NTHRS_INIT 2 @@ -259,19 +247,20 @@ * threads for each partition to keep service healthy, so total threads * number should be 24 * 8 = 192. * - * So with these constants, threads number wil be at the similar level + * So with these constants, threads number will be at the similar level * of old versions, unless target machine has over a hundred cores */ #define LDLM_THR_FACTOR 8 #define LDLM_NTHRS_INIT PTLRPC_NTHRS_INIT #define LDLM_NTHRS_BASE 24 -#define LDLM_NTHRS_MAX (cfs_num_online_cpus() == 1 ? 64 : 128) +#define LDLM_NTHRS_MAX (num_online_cpus() == 1 ? 64 : 128) -#define LDLM_BL_THREADS LDLM_NTHRS_AUTO_INIT -#define LDLM_NBUFS (64 * cfs_num_online_cpus()) -#define LDLM_BUFSIZE (8 * 1024) -#define LDLM_MAXREQSIZE (5 * 1024) -#define LDLM_MAXREPSIZE (1024) +#define LDLM_BL_THREADS LDLM_NTHRS_AUTO_INIT +#define LDLM_CLIENT_NBUFS 1 +#define LDLM_SERVER_NBUFS 64 +#define LDLM_BUFSIZE (8 * 1024) +#define LDLM_MAXREQSIZE (5 * 1024) +#define LDLM_MAXREPSIZE (1024) /* * MDS threads constants: @@ -313,11 +302,15 @@ #define MDS_OTHR_NTHRS_INIT PTLRPC_NTHRS_INIT #define MDS_OTHR_NTHRS_MAX MDS_MAX_OTHR_THREADS -#define MDS_NBUFS (64 * cfs_num_online_cpus()) +#define MDS_NBUFS 64 + /** * Assume file name length = FNAME_MAX = 256 (true for ext3). - * path name length = PATH_MAX = 4096 - * LOV MD size max = EA_MAX = 48000 (2000 stripes) + * path name length = PATH_MAX = 4096 + * LOV MD size max = EA_MAX = 24 * 2000 + * (NB: 24 is size of lov_ost_data) + * LOV LOGCOOKIE size max = 32 * 2000 + * (NB: 32 is size of llog_cookie) * symlink: FNAME_MAX + PATH_MAX <- largest * link: FNAME_MAX + PATH_MAX (mds_rec_link < mds_rec_create) * rename: FNAME_MAX + FNAME_MAX @@ -326,23 +319,95 @@ * MDS_MAXREQSIZE ~= 4736 bytes = * lustre_msg + ldlm_request + mdt_body + mds_rec_create + FNAME_MAX + PATH_MAX * MDS_MAXREPSIZE ~= 8300 bytes = lustre_msg + llog_header - * or, for mds_close() and mds_reint_unlink() on a many-OST filesystem: - * = 9210 bytes = lustre_msg + mdt_body + 160 * (easize + cookiesize) * * Realistic size is about 512 bytes (20 character name + 128 char symlink), * except in the open case where there are a large number of OSTs in a LOV. */ -#define MDS_MAXREPSIZE max(10 * 1024, 362 + LOV_MAX_STRIPE_COUNT * 56) -#define MDS_MAXREQSIZE MDS_MAXREPSIZE +#define MDS_MAXREQSIZE (5 * 1024) /* >= 4736 */ +#define MDS_MAXREPSIZE (9 * 1024) /* >= 8300 */ -/** MDS_BUFSIZE = max_reqsize + max sptlrpc payload size */ -#define MDS_BUFSIZE (MDS_MAXREQSIZE + 1024) +/** + * MDS incoming request with LOV EA + * 24 = sizeof(struct lov_ost_data), i.e: replay of opencreate + */ +#define MDS_LOV_MAXREQSIZE max(MDS_MAXREQSIZE, \ + 362 + LOV_MAX_STRIPE_COUNT * 24) +/** + * MDS outgoing reply with LOV EA + * + * NB: max reply size Lustre 2.4+ client can get from old MDS is: + * LOV_MAX_STRIPE_COUNT * (llog_cookie + lov_ost_data) + extra bytes + * + * but 2.4 or later MDS will never send reply with llog_cookie to any + * version client. This macro is defined for server side reply buffer size. + */ +#define MDS_LOV_MAXREPSIZE MDS_LOV_MAXREQSIZE + +/** + * This is the size of a maximum REINT_SETXATTR request: + * + * lustre_msg 56 (32 + 4 x 5 + 4) + * ptlrpc_body 184 + * mdt_rec_setxattr 136 + * lustre_capa 120 + * name 256 (XATTR_NAME_MAX) + * value 65536 (XATTR_SIZE_MAX) + */ +#define MDS_EA_MAXREQSIZE 66288 + +/** + * These are the maximum request and reply sizes (rounded up to 1 KB + * boundaries) for the "regular" MDS_REQUEST_PORTAL and MDS_REPLY_PORTAL. + */ +#define MDS_REG_MAXREQSIZE (((max(MDS_EA_MAXREQSIZE, \ + MDS_LOV_MAXREQSIZE) + 1023) >> 10) << 10) +#define MDS_REG_MAXREPSIZE MDS_REG_MAXREQSIZE + +/** + * The update request includes all of updates from the create, which might + * include linkea (4K maxim), together with other updates, we set it to 1000K: + * lustre_msg + ptlrpc_body + OUT_UPDATE_BUFFER_SIZE_MAX + */ +#define OUT_MAXREQSIZE (1000 * 1024) +#define OUT_MAXREPSIZE MDS_MAXREPSIZE + +/** MDS_BUFSIZE = max_reqsize (w/o LOV EA) + max sptlrpc payload size */ +#define MDS_BUFSIZE max(MDS_MAXREQSIZE + SPTLRPC_MAX_PAYLOAD, \ + 8 * 1024) + +/** + * MDS_REG_BUFSIZE should at least be MDS_REG_MAXREQSIZE + SPTLRPC_MAX_PAYLOAD. + * However, we need to allocate a much larger buffer for it because LNet + * requires each MD(rqbd) has at least MDS_REQ_MAXREQSIZE bytes left to avoid + * dropping of maximum-sized incoming request. So if MDS_REG_BUFSIZE is only a + * little larger than MDS_REG_MAXREQSIZE, then it can only fit in one request + * even there are about MDS_REG_MAX_REQSIZE bytes left in a rqbd, and memory + * utilization is very low. + * + * In the meanwhile, size of rqbd can't be too large, because rqbd can't be + * reused until all requests fit in it have been processed and released, + * which means one long blocked request can prevent the rqbd be reused. + * Now we set request buffer size to 160 KB, so even each rqbd is unlinked + * from LNet with unused 65 KB, buffer utilization will be about 59%. + * Please check LU-2432 for details. + */ +#define MDS_REG_BUFSIZE max(MDS_REG_MAXREQSIZE + SPTLRPC_MAX_PAYLOAD, \ + 160 * 1024) + +/** + * OUT_BUFSIZE = max_out_reqsize + max sptlrpc payload (~1K) which is + * about 10K, for the same reason as MDS_REG_BUFSIZE, we also give some + * extra bytes to each request buffer to improve buffer utilization rate. + */ +#define OUT_BUFSIZE max(OUT_MAXREQSIZE + SPTLRPC_MAX_PAYLOAD, \ + 24 * 1024) /** FLD_MAXREQSIZE == lustre_msg + __u32 padding + ptlrpc_body + opc */ #define FLD_MAXREQSIZE (160) /** FLD_MAXREPSIZE == lustre_msg + ptlrpc_body */ #define FLD_MAXREPSIZE (152) +#define FLD_BUFSIZE (1 << 12) /** * SEQ_MAXREQSIZE == lustre_msg + __u32 padding + ptlrpc_body + opc + lu_range + @@ -351,12 +416,13 @@ /** SEQ_MAXREPSIZE == lustre_msg + ptlrpc_body + lu_range */ #define SEQ_MAXREPSIZE (152) +#define SEQ_BUFSIZE (1 << 12) /** MGS threads must be >= 3, see bug 22458 comment #28 */ #define MGS_NTHRS_INIT (PTLRPC_NTHRS_INIT + 1) #define MGS_NTHRS_MAX 32 -#define MGS_NBUFS (64 * cfs_num_online_cpus()) +#define MGS_NBUFS 64 #define MGS_BUFSIZE (8 * 1024) #define MGS_MAXREQSIZE (7 * 1024) #define MGS_MAXREPSIZE (9 * 1024) @@ -387,7 +453,7 @@ */ /* depress threads factor for VM with small memory size */ #define OSS_THR_FACTOR min_t(int, 8, \ - CFS_NUM_CACHEPAGES >> (28 - CFS_PAGE_SHIFT)) + NUM_CACHEPAGES >> (28 - PAGE_CACHE_SHIFT)) #define OSS_NTHRS_INIT (PTLRPC_NTHRS_INIT + 1) #define OSS_NTHRS_BASE 64 #define OSS_NTHRS_MAX 512 @@ -399,40 +465,61 @@ #define OSS_CR_NTHRS_MAX 64 /** - * OST_MAXREQSIZE ~= - * lustre_msg + obdo + obd_ioobj + DT_MAX_BRW_PAGES * niobuf_remote + * OST_IO_MAXREQSIZE ~= + * lustre_msg + ptlrpc_body + obdo + obd_ioobj + + * DT_MAX_BRW_PAGES * niobuf_remote * * - single object with 16 pages is 512 bytes - * - OST_MAXREQSIZE must be at least 1 page of cookies plus some spillover + * - OST_IO_MAXREQSIZE must be at least 1 page of cookies plus some spillover * - Must be a multiple of 1024 + * - actual size is about 18K + */ +#define _OST_MAXREQSIZE_SUM (sizeof(struct lustre_msg) + \ + sizeof(struct ptlrpc_body) + \ + sizeof(struct obdo) + \ + sizeof(struct obd_ioobj) + \ + sizeof(struct niobuf_remote) * DT_MAX_BRW_PAGES) +/** + * FIEMAP request can be 4K+ for now */ -#define _OST_MAXREQSIZE_SUM (sizeof(struct lustre_msg) + sizeof(struct obdo) + \ - sizeof(struct obd_ioobj) + DT_MAX_BRW_PAGES * \ - sizeof(struct niobuf_remote)) -#define OST_MAXREQSIZE (((_OST_MAXREQSIZE_SUM - 1) | (1024 - 1)) + 1) +#define OST_MAXREQSIZE (16 * 1024) +#define OST_IO_MAXREQSIZE max_t(int, OST_MAXREQSIZE, \ + (((_OST_MAXREQSIZE_SUM - 1) | (1024 - 1)) + 1)) -#define OST_MAXREPSIZE (9 * 1024) +#define OST_MAXREPSIZE (9 * 1024) +#define OST_IO_MAXREPSIZE OST_MAXREPSIZE -#define OST_NBUFS (64 * cfs_num_online_cpus()) -#define OST_BUFSIZE (OST_MAXREQSIZE + 1024) +#define OST_NBUFS 64 +/** OST_BUFSIZE = max_reqsize + max sptlrpc payload size */ +#define OST_BUFSIZE max_t(int, OST_MAXREQSIZE + 1024, 16 * 1024) +/** + * OST_IO_MAXREQSIZE is 18K, giving extra 46K can increase buffer utilization + * rate of request buffer, please check comment of MDS_LOV_BUFSIZE for details. + */ +#define OST_IO_BUFSIZE max_t(int, OST_IO_MAXREQSIZE + 1024, 64 * 1024) /* Macro to hide a typecast. */ #define ptlrpc_req_async_args(req) ((void *)&req->rq_async_args) +struct ptlrpc_replay_async_args { + int praa_old_state; + int praa_old_status; +}; + /** * Structure to single define portal connection. */ struct ptlrpc_connection { - /** linkage for connections hash table */ - cfs_hlist_node_t c_hash; - /** Our own lnet nid for this connection */ - lnet_nid_t c_self; - /** Remote side nid for this connection */ - lnet_process_id_t c_peer; - /** UUID of the other side */ - struct obd_uuid c_remote_uuid; - /** reference counter for this connection */ - cfs_atomic_t c_refcount; + /** linkage for connections hash table */ + struct hlist_node c_hash; + /** Our own lnet nid for this connection */ + lnet_nid_t c_self; + /** Remote side nid for this connection */ + lnet_process_id_t c_peer; + /** UUID of the other side */ + struct obd_uuid c_remote_uuid; + /** reference counter for this connection */ + atomic_t c_refcount; }; /** Client definition for PortalRPC */ @@ -480,26 +567,26 @@ typedef int (*set_producer_func)(struct ptlrpc_request_set *, void *); * returned. */ struct ptlrpc_request_set { - cfs_atomic_t set_refcount; + atomic_t set_refcount; /** number of in queue requests */ - cfs_atomic_t set_new_count; + atomic_t set_new_count; /** number of uncompleted requests */ - cfs_atomic_t set_remaining; + atomic_t set_remaining; /** wait queue to wait on for request events */ - cfs_waitq_t set_waitq; - cfs_waitq_t *set_wakeup_ptr; + wait_queue_head_t set_waitq; + wait_queue_head_t *set_wakeup_ptr; /** List of requests in the set */ - cfs_list_t set_requests; + struct list_head set_requests; /** * List of completion callbacks to be called when the set is completed * This is only used if \a set_interpret is NULL. * Links struct ptlrpc_set_cbdata. */ - cfs_list_t set_cblist; + struct list_head set_cblist; /** Completion callback, if only one. */ - set_interpreter_func set_interpret; + set_interpreter_func set_interpret; /** opaq argument passed to completion \a set_interpret callback. */ - void *set_arg; + void *set_arg; /** * Lock for \a set_new_requests manipulations * locked so that any old caller can communicate requests to @@ -507,29 +594,29 @@ struct ptlrpc_request_set { */ spinlock_t set_new_req_lock; /** List of new yet unsent requests. Only used with ptlrpcd now. */ - cfs_list_t set_new_requests; + struct list_head set_new_requests; /** rq_status of requests that have been freed already */ - int set_rc; + int set_rc; /** Additional fields used by the flow control extension */ /** Maximum number of RPCs in flight */ - int set_max_inflight; + int set_max_inflight; /** Callback function used to generate RPCs */ - set_producer_func set_producer; + set_producer_func set_producer; /** opaq argument passed to the producer callback */ - void *set_producer_arg; + void *set_producer_arg; }; /** * Description of a single ptrlrpc_set callback */ struct ptlrpc_set_cbdata { - /** List linkage item */ - cfs_list_t psc_item; - /** Pointer to interpreting function */ - set_interpreter_func psc_interpret; - /** Opaq argument to pass to the callback */ - void *psc_data; + /** List linkage item */ + struct list_head psc_item; + /** Pointer to interpreting function */ + set_interpreter_func psc_interpret; + /** Opaq argument to pass to the callback */ + void *psc_data; }; struct ptlrpc_bulk_desc; @@ -556,20 +643,20 @@ struct ptlrpc_cb_id { * added to the state for replay/failover consistency guarantees. */ struct ptlrpc_reply_state { - /** Callback description */ - struct ptlrpc_cb_id rs_cb_id; - /** Linkage for list of all reply states in a system */ - cfs_list_t rs_list; - /** Linkage for list of all reply states on same export */ - cfs_list_t rs_exp_list; - /** Linkage for list of all reply states for same obd */ - cfs_list_t rs_obd_list; + /** Callback description */ + struct ptlrpc_cb_id rs_cb_id; + /** Linkage for list of all reply states in a system */ + struct list_head rs_list; + /** Linkage for list of all reply states on same export */ + struct list_head rs_exp_list; + /** Linkage for list of all reply states for same obd */ + struct list_head rs_obd_list; #if RS_DEBUG - cfs_list_t rs_debug_list; + struct list_head rs_debug_list; #endif - /** A spinlock to protect the reply state flags */ + /** A spinlock to protect the reply state flags */ spinlock_t rs_lock; - /** Reply state flags */ + /** Reply state flags */ unsigned long rs_difficult:1; /* ACK/commit stuff */ unsigned long rs_no_ack:1; /* no ACK, even for difficult requests */ @@ -589,16 +676,16 @@ struct ptlrpc_reply_state { __u64 rs_transno; /** xid */ __u64 rs_xid; - struct obd_export *rs_export; + struct obd_export *rs_export; struct ptlrpc_service_part *rs_svcpt; - /** Lnet metadata handle for the reply */ - lnet_handle_md_t rs_md_h; - cfs_atomic_t rs_refcount; - - /** Context for the sevice thread */ - struct ptlrpc_svc_ctx *rs_svc_ctx; - /** Reply buffer (actually sent to the client), encoded if needed */ - struct lustre_msg *rs_repbuf; /* wrapper */ + /** Lnet metadata handle for the reply */ + lnet_handle_md_t rs_md_h; + atomic_t rs_refcount; + + /** Context for the sevice thread */ + struct ptlrpc_svc_ctx *rs_svc_ctx; + /** Reply buffer (actually sent to the client), encoded if needed */ + struct lustre_msg *rs_repbuf; /* wrapper */ /** Size of the reply buffer */ int rs_repbuf_len; /* wrapper buf length */ /** Size of the reply message */ @@ -610,12 +697,12 @@ struct ptlrpc_reply_state { */ struct lustre_msg *rs_msg; /* reply message */ - /** Number of locks awaiting client ACK */ - int rs_nlocks; - /** Handles of locks awaiting client reply ACK */ - struct lustre_handle rs_locks[RS_MAX_LOCKS]; - /** Lock modes of locks in \a rs_locks */ - ldlm_mode_t rs_modes[RS_MAX_LOCKS]; + /** Number of locks awaiting client ACK */ + int rs_nlocks; + /** Handles of locks awaiting client reply ACK */ + struct lustre_handle rs_locks[RS_MAX_LOCKS]; + /** Lock modes of locks in \a rs_locks */ + enum ldlm_mode rs_modes[RS_MAX_LOCKS]; }; struct ptlrpc_thread; @@ -635,6 +722,9 @@ enum rq_phase { typedef int (*ptlrpc_interpterer_t)(const struct lu_env *env, struct ptlrpc_request *req, void *arg, int rc); +/** Type of request resend call-back */ +typedef void (*ptlrpc_resend_cb_t)(struct ptlrpc_request *req, + void *arg); /** * Definition of request pool structure. @@ -644,13 +734,13 @@ typedef int (*ptlrpc_interpterer_t)(const struct lu_env *env, */ struct ptlrpc_request_pool { /** Locks the list */ - spinlock_t prp_lock; - /** list of ptlrpc_request structs */ - cfs_list_t prp_req_list; - /** Maximum message size that would fit into a rquest from this pool */ - int prp_rq_size; - /** Function to allocate more requests for this pool */ - void (*prp_populate)(struct ptlrpc_request_pool *, int); + spinlock_t prp_lock; + /** list of ptlrpc_request structs */ + struct list_head prp_req_list; + /** Maximum message size that would fit into a rquest from this pool */ + int prp_rq_size; + /** Function to allocate more requests for this pool */ + int (*prp_populate)(struct ptlrpc_request_pool *, int); }; struct lu_context; @@ -658,657 +748,7 @@ struct lu_env; struct ldlm_lock; -/** - * \defgroup nrs Network Request Scheduler - * @{ - */ -struct ptlrpc_nrs_policy; -struct ptlrpc_nrs_resource; -struct ptlrpc_nrs_request; - -/** - * NRS control operations. - * - * These are common for all policies. - */ -enum ptlrpc_nrs_ctl { - /** - * Activate the policy. - */ - PTLRPC_NRS_CTL_START, - /** - * Reserved for multiple primary policies, which may be a possibility - * in the future. - */ - PTLRPC_NRS_CTL_STOP, - /** - * Recycle resources for inactive policies. - */ - PTLRPC_NRS_CTL_SHRINK, - /** - * Not a valid opcode. - */ - PTLRPC_NRS_CTL_INVALID, - /** - * Policies can start using opcodes from this value and onwards for - * their own purposes; the assigned value itself is arbitrary. - */ - PTLRPC_NRS_CTL_1ST_POL_SPEC = 0x20, -}; - -/** - * NRS policy operations. - * - * These determine the behaviour of a policy, and are called in response to - * NRS core events. - */ -struct ptlrpc_nrs_pol_ops { - /** - * Called during policy registration; this operation is optional. - * - * \param[in] policy The policy being initialized - */ - int (*op_policy_init) (struct ptlrpc_nrs_policy *policy); - /** - * Called during policy unregistration; this operation is optional. - * - * \param[in] policy The policy being unregistered/finalized - */ - void (*op_policy_fini) (struct ptlrpc_nrs_policy *policy); - /** - * Called when activating a policy via lprocfs; policies allocate and - * initialize their resources here; this operation is optional. - * - * \param[in] policy The policy being started - * - * \see nrs_policy_start_locked() - */ - int (*op_policy_start) (struct ptlrpc_nrs_policy *policy); - /** - * Called when deactivating a policy via lprocfs; policies deallocate - * their resources here; this operation is optional - * - * \param[in] policy The policy being stopped - * - * \see nrs_policy_stop_final() - */ - void (*op_policy_stop) (struct ptlrpc_nrs_policy *policy); - /** - * Used for policy-specific operations; i.e. not generic ones like - * \e PTLRPC_NRS_CTL_START and \e PTLRPC_NRS_CTL_GET_INFO; analogous - * to an ioctl; this operation is optional. - * - * \param[in] policy The policy carrying out operation \a opc - * \param[in] opc The command operation being carried out - * \param[in,out] arg An generic buffer for communication between the - * user and the control operation - * - * \retval -ve error - * \retval 0 success - * - * \see ptlrpc_nrs_policy_control() - */ - int (*op_policy_ctl) (struct ptlrpc_nrs_policy *policy, - enum ptlrpc_nrs_ctl opc, void *arg); - - /** - * Called when obtaining references to the resources of the resource - * hierarchy for a request that has arrived for handling at the PTLRPC - * service. Policies should return -ve for requests they do not wish - * to handle. This operation is mandatory. - * - * \param[in] policy The policy we're getting resources for. - * \param[in] nrq The request we are getting resources for. - * \param[in] parent The parent resource of the resource being - * requested; set to NULL if none. - * \param[out] resp The resource is to be returned here; the - * fallback policy in an NRS head should - * \e always return a non-NULL pointer value. - * \param[in] moving_req When set, signifies that this is an attempt - * to obtain resources for a request being moved - * to the high-priority NRS head by - * ldlm_lock_reorder_req(). - * This implies two things: - * 1. We are under obd_export::exp_rpc_lock and - * so should not sleep. - * 2. We should not perform non-idempotent or can - * skip performing idempotent operations that - * were carried out when resources were first - * taken for the request when it was initialized - * in ptlrpc_nrs_req_initialize(). - * - * \retval 0, +ve The level of the returned resource in the resource - * hierarchy; currently only 0 (for a non-leaf resource) - * and 1 (for a leaf resource) are supported by the - * framework. - * \retval -ve error - * - * \see ptlrpc_nrs_req_initialize() - * \see ptlrpc_nrs_hpreq_add_nolock() - * \see ptlrpc_nrs_req_hp_move() - */ - int (*op_res_get) (struct ptlrpc_nrs_policy *policy, - struct ptlrpc_nrs_request *nrq, - struct ptlrpc_nrs_resource *parent, - struct ptlrpc_nrs_resource **resp, - bool moving_req); - /** - * Called when releasing references taken for resources in the resource - * hierarchy for the request; this operation is optional. - * - * \param[in] policy The policy the resource belongs to - * \param[in] res The resource to be freed - * - * \see ptlrpc_nrs_req_finalize() - * \see ptlrpc_nrs_hpreq_add_nolock() - * \see ptlrpc_nrs_req_hp_move() - */ - void (*op_res_put) (struct ptlrpc_nrs_policy *policy, - struct ptlrpc_nrs_resource *res); - - /** - * Obtain a request for handling from the policy via polling; this - * operation is mandatory. - * - * \param[in] policy The policy to poll - * - * \retval NULL No erquest available for handling - * \retval valid-pointer The request polled for handling - * - * \see ptlrpc_nrs_req_poll_nolock() - */ - struct ptlrpc_nrs_request * - (*op_req_poll) (struct ptlrpc_nrs_policy *policy); - /** - * Called when attempting to add a request to a policy for later - * handling; this operation is mandatory. - * - * \param[in] policy The policy on which to enqueue \a nrq - * \param[in] nrq The request to enqueue - * - * \retval 0 success - * \retval != 0 error - * - * \see ptlrpc_nrs_req_add_nolock() - */ - int (*op_req_enqueue) (struct ptlrpc_nrs_policy *policy, - struct ptlrpc_nrs_request *nrq); - /** - * Removes a request from the policy's set of pending requests. Normally - * called after a request has been polled successfully from the policy - * for handling; this operation is mandatory. - * - * \param[in] policy The policy the request \a nrq belongs to - * \param[in] nrq The request to dequeue - * - * \see ptlrpc_nrs_req_del_nolock() - */ - void (*op_req_dequeue) (struct ptlrpc_nrs_policy *policy, - struct ptlrpc_nrs_request *nrq); - /** - * Called before carrying out the request; should not block. Could be - * used for job/resource control; this operation is optional. - * - * \param[in] policy The policy which is starting to handle request - * \a nrq - * \param[in] nrq The request - * - * \pre spin_is_locked(&svcpt->scp_req_lock) - * - * \see ptlrpc_nrs_req_start_nolock() - */ - void (*op_req_start) (struct ptlrpc_nrs_policy *policy, - struct ptlrpc_nrs_request *nrq); - /** - * Called after the request being carried out. Could be used for - * job/resource control; this operation is optional. - * - * \param[in] policy The policy which is stopping to handle request - * \a nrq - * \param[in] nrq The request - * - * \pre spin_is_locked(&svcpt->scp_req_lock) - * - * \see ptlrpc_nrs_req_stop_nolock() - */ - void (*op_req_stop) (struct ptlrpc_nrs_policy *policy, - struct ptlrpc_nrs_request *nrq); - /** - * Registers the policy's lprocfs interface with a PTLRPC service. - * - * \param[in] svc The service - * - * \retval 0 success - * \retval != 0 error - */ - int (*op_lprocfs_init) (struct ptlrpc_service *svc); - /** - * Unegisters the policy's lprocfs interface with a PTLRPC service. - * - * \param[in] svc The service - */ - void (*op_lprocfs_fini) (struct ptlrpc_service *svc); -}; - -/** - * Policy flags - */ -enum nrs_policy_flags { - /** - * Fallback policy, use this flag only on a single supported policy per - * service. Do not use this flag for policies registering using - * ptlrpc_nrs_policy_register() (i.e. ones that are not in - * \e nrs_pols_builtin). - */ - PTLRPC_NRS_FL_FALLBACK = (1 << 0), - /** - * Start policy immediately after registering. - */ - PTLRPC_NRS_FL_REG_START = (1 << 1), - /** - * This is a polciy registering externally with NRS core, via - * ptlrpc_nrs_policy_register(), (i.e. one that is not in - * \e nrs_pols_builtin. Used to avoid ptlrpc_nrs_policy_register() - * racing with a policy start operation issued by the user via lprocfs. - */ - PTLRPC_NRS_FL_REG_EXTERN = (1 << 2), -}; - -/** - * NRS queue type. - * - * Denotes whether an NRS instance is for handling normal or high-priority - * RPCs, or whether an operation pertains to one or both of the NRS instances - * in a service. - */ -enum ptlrpc_nrs_queue_type { - PTLRPC_NRS_QUEUE_REG, - PTLRPC_NRS_QUEUE_HP, - PTLRPC_NRS_QUEUE_BOTH, -}; - -/** - * NRS head - * - * A PTLRPC service has at least one NRS head instance for handling normal - * priority RPCs, and may optionally have a second NRS head instance for - * handling high-priority RPCs. Each NRS head maintains a list of available - * policies, of which one and only one policy is acting as the fallback policy, - * and optionally a different policy may be acting as the primary policy. For - * all RPCs handled by this NRS head instance, NRS core will first attempt to - * enqueue the RPC using the primary policy (if any). The fallback policy is - * used in the following cases: - * - when there was no primary policy in the - * ptlrpc_nrs_pol_state::NRS_POL_STATE_STARTED state at the time the request - * was initialized. - * - when the primary policy that was at the - * ptlrpc_nrs_pol_state::PTLRPC_NRS_POL_STATE_STARTED state at the time the - * RPC was initialized, denoted it did not wish, or for some other reason was - * not able to handle the request, by returning a non-valid NRS resource - * reference. - * - when the primary policy that was at the - * ptlrpc_nrs_pol_state::PTLRPC_NRS_POL_STATE_STARTED state at the time the - * RPC was initialized, fails later during the request enqueueing stage. - * - * \see nrs_resource_get_safe() - * \see nrs_request_enqueue() - */ -struct ptlrpc_nrs { - spinlock_t nrs_lock; - /** XXX Possibly replace svcpt->scp_req_lock with another lock here. */ - /** - * Linkage into nrs_core_heads_list - */ - cfs_list_t nrs_heads; - /** - * List of registered policies - */ - cfs_list_t nrs_policy_list; - /** - * List of policies with queued requests. Policies that have any - * outstanding requests are queued here, and this list is queried - * in a round-robin manner from NRS core when obtaining a request - * for handling. This ensures that requests from policies that at some - * point transition away from the - * ptlrpc_nrs_pol_state::NRS_POL_STATE_STARTED state are drained. - */ - cfs_list_t nrs_policy_queued; - /** - * Service partition for this NRS head - */ - struct ptlrpc_service_part *nrs_svcpt; - /** - * Primary policy, which is the preferred policy for handling RPCs - */ - struct ptlrpc_nrs_policy *nrs_policy_primary; - /** - * Fallback policy, which is the backup policy for handling RPCs - */ - struct ptlrpc_nrs_policy *nrs_policy_fallback; - /** - * This NRS head handles either HP or regular requests - */ - enum ptlrpc_nrs_queue_type nrs_queue_type; - /** - * # queued requests from all policies in this NRS head - */ - unsigned long nrs_req_queued; - /** - * # scheduled requests from all policies in this NRS head - */ - unsigned long nrs_req_started; - /** - * # policies on this NRS - * TODO: Can we avoid having this? - */ - unsigned nrs_num_pols; - /** - * This NRS head is in progress of starting a policy - */ - unsigned nrs_policy_starting:1; - /** - * In progress of shutting down the whole NRS head; used during - * unregistration - */ - unsigned nrs_stopping:1; -}; - -#define NRS_POL_NAME_MAX 16 - -/** - * NRS policy registering descriptor - * - * Is used to hold a description of a policy that can be passed to NRS core in - * order to register the policy with NRS heads in different PTLRPC services. - */ -struct ptlrpc_nrs_pol_desc { - /** - * Human-readable policy name - */ - char pd_name[NRS_POL_NAME_MAX]; - /** - * NRS operations for this policy - */ - struct ptlrpc_nrs_pol_ops *pd_ops; - /** - * Service Compatibility function; this determines whether a policy is - * adequate for handling RPCs of a particular PTLRPC service. - * - * XXX:This should give the same result during policy - * registration and unregistration, and for all partitions of a - * service; so the result should not depend on temporal service - * or other properties, that may influence the result. - */ - bool (*pd_compat) (struct ptlrpc_service *svc, - const struct ptlrpc_nrs_pol_desc *desc); - /** - * Optionally set for policies that support a single ptlrpc service, - * i.e. ones that have \a pd_compat set to nrs_policy_compat_one() - */ - char *pd_compat_svc_name; - /** - * Bitmask of nrs_policy_flags - */ - unsigned pd_flags; - /** - * Link into nrs_core::nrs_policies - */ - cfs_list_t pd_list; -}; - -/** - * NRS policy state - * - * Policies transition from one state to the other during their lifetime - */ -enum ptlrpc_nrs_pol_state { - /** - * Not a valid policy state. - */ - NRS_POL_STATE_INVALID, - /** - * For now, this state is used exclusively for policies that register - * externally to NRS core, i.e. ones that do so via - * ptlrpc_nrs_policy_register() and are not part of nrs_pols_builtin; - * it is used to prevent a race condition between the policy registering - * with more than one service partition while service is operational, - * and the user starting the policy via lprocfs. - * - * \see nrs_pol_make_avail() - */ - NRS_POL_STATE_UNAVAIL, - /** - * Policies are at this state either at the start of their life, or - * transition here when the user selects a different policy to act - * as the primary one. - */ - NRS_POL_STATE_STOPPED, - /** - * Policy is progress of stopping - */ - NRS_POL_STATE_STOPPING, - /** - * Policy is in progress of starting - */ - NRS_POL_STATE_STARTING, - /** - * A policy is in this state in two cases: - * - it is the fallback policy, which is always in this state. - * - it has been activated by the user; i.e. it is the primary policy, - */ - NRS_POL_STATE_STARTED, -}; - -/** - * NRS policy information - * - * Used for obtaining information for the status of a policy via lprocfs - */ -struct ptlrpc_nrs_pol_info { - /** - * Policy name - */ - char pi_name[NRS_POL_NAME_MAX]; - /** - * Current policy state - */ - enum ptlrpc_nrs_pol_state pi_state; - /** - * # RPCs enqueued for later dispatching by the policy - */ - long pi_req_queued; - /** - * # RPCs started for dispatch by the policy - */ - long pi_req_started; - /** - * Is this a fallback policy? - */ - unsigned pi_fallback:1; -}; - -/** - * NRS policy - * - * There is one instance of this for each policy in each NRS head of each - * PTLRPC service partition. - */ -struct ptlrpc_nrs_policy { - /** - * Linkage into the NRS head's list of policies, - * ptlrpc_nrs:nrs_policy_list - */ - cfs_list_t pol_list; - /** - * Linkage into the NRS head's list of policies with enqueued - * requests ptlrpc_nrs:nrs_policy_queued - */ - cfs_list_t pol_list_queued; - /** - * Current state of this policy - */ - enum ptlrpc_nrs_pol_state pol_state; - /** - * Bitmask of nrs_policy_flags - */ - unsigned pol_flags; - /** - * # RPCs enqueued for later dispatching by the policy - */ - long pol_req_queued; - /** - * # RPCs started for dispatch by the policy - */ - long pol_req_started; - /** - * Usage Reference count taken on the policy instance - */ - long pol_ref; - /** - * The NRS head this policy has been created at - */ - struct ptlrpc_nrs *pol_nrs; - /** - * NRS operations for this policy; points to ptlrpc_nrs_pol_desc::pd_ops - */ - struct ptlrpc_nrs_pol_ops *pol_ops; - /** - * Private policy data; varies by policy type - */ - void *pol_private; - /** - * Human-readable policy name; point to ptlrpc_nrs_pol_desc::pd_name - */ - char *pol_name; -}; - -/** - * NRS resource - * - * Resources are embedded into two types of NRS entities: - * - Inside NRS policies, in the policy's private data in - * ptlrpc_nrs_policy::pol_private - * - In objects that act as prime-level scheduling entities in different NRS - * policies; e.g. on a policy that performs round robin or similar order - * scheduling across client NIDs, there would be one NRS resource per unique - * client NID. On a policy which performs round robin scheduling across - * backend filesystem objects, there would be one resource associated with - * each of the backend filesystem objects partaking in the scheduling - * performed by the policy. - * - * NRS resources share a parent-child relationship, in which resources embedded - * in policy instances are the parent entities, with all scheduling entities - * a policy schedules across being the children, thus forming a simple resource - * hierarchy. This hierarchy may be extended with one or more levels in the - * future if the ability to have more than one primary policy is added. - * - * Upon request initialization, references to the then active NRS policies are - * taken and used to later handle the dispatching of the request with one of - * these policies. - * - * \see nrs_resource_get_safe() - * \see ptlrpc_nrs_req_add() - */ -struct ptlrpc_nrs_resource { - /** - * This NRS resource's parent; is NULL for resources embedded in NRS - * policy instances; i.e. those are top-level ones. - */ - struct ptlrpc_nrs_resource *res_parent; - /** - * The policy associated with this resource. - */ - struct ptlrpc_nrs_policy *res_policy; -}; - -enum { - NRS_RES_FALLBACK, - NRS_RES_PRIMARY, - NRS_RES_MAX -}; - -/* \name fifo - * - * FIFO policy - * - * This policy is a logical wrapper around previous, non-NRS functionality. - * It dispatches RPCs in the same order as they arrive from the network. This - * policy is currently used as the fallback policy, and the only enabled policy - * on all NRS heads of all PTLRPC service partitions. - * @{ - */ - -/** - * Private data structure for the FIFO policy - */ -struct nrs_fifo_head { - /** - * Resource object for policy instance. - */ - struct ptlrpc_nrs_resource fh_res; - /** - * List of queued requests. - */ - cfs_list_t fh_list; - /** - * For debugging purposes. - */ - __u64 fh_sequence; -}; - -struct nrs_fifo_req { - /** request header, must be the first member of structure */ - cfs_list_t fr_list; - __u64 fr_sequence; -}; - -/** @} fifo */ - -/** - * NRS request - * - * Instances of this object exist embedded within ptlrpc_request; the main - * purpose of this object is to hold references to the request's resources - * for the lifetime of the request, and to hold properties that policies use - * use for determining the request's scheduling priority. - * */ -struct ptlrpc_nrs_request { - /** - * The request's resource hierarchy. - */ - struct ptlrpc_nrs_resource *nr_res_ptrs[NRS_RES_MAX]; - /** - * Index into ptlrpc_nrs_request::nr_res_ptrs of the resource of the - * policy that was used to enqueue the request. - * - * \see nrs_request_enqueue() - */ - unsigned nr_res_idx; - unsigned nr_initialized:1; - unsigned nr_enqueued:1; - unsigned nr_dequeued:1; - unsigned nr_started:1; - unsigned nr_finalized:1; - cfs_binheap_node_t nr_node; - - /** - * Policy-specific fields, used for determining a request's scheduling - * priority, and other supporting functionality. - */ - union { - /** - * Fields for the FIFO policy - */ - struct nrs_fifo_req fifo; - } nr_u; - /** - * Externally-registering policies may want to use this to allocate - * their own request properties. - */ - void *ext; -}; - -/** @} nrs */ +#include /** * Basic request prioritization operations structure. @@ -1335,6 +775,167 @@ struct ptlrpc_hpreq_ops { void (*hpreq_fini)(struct ptlrpc_request *); }; +struct ptlrpc_cli_req { + /** For bulk requests on client only: bulk descriptor */ + struct ptlrpc_bulk_desc *cr_bulk; + /** optional time limit for send attempts */ + cfs_duration_t cr_delay_limit; + /** time request was first queued */ + cfs_time_t cr_queued_time; + /** request sent timeval */ + struct timeval cr_sent_tv; + /** time for request really sent out */ + time_t cr_sent_out; + /** when req reply unlink must finish. */ + time_t cr_reply_deadline; + /** when req bulk unlink must finish. */ + time_t cr_bulk_deadline; + /** Portal to which this request would be sent */ + short cr_req_ptl; + /** Portal where to wait for reply and where reply would be sent */ + short cr_rep_ptl; + /** request resending number */ + unsigned int cr_resend_nr; + /** What was import generation when this request was sent */ + int cr_imp_gen; + enum lustre_imp_state cr_send_state; + /** Per-request waitq introduced by bug 21938 for recovery waiting */ + wait_queue_head_t cr_set_waitq; + /** Link item for request set lists */ + struct list_head cr_set_chain; + /** link to waited ctx */ + struct list_head cr_ctx_chain; + + /** client's half ctx */ + struct ptlrpc_cli_ctx *cr_cli_ctx; + /** Link back to the request set */ + struct ptlrpc_request_set *cr_set; + /** outgoing request MD handle */ + lnet_handle_md_t cr_req_md_h; + /** request-out callback parameter */ + struct ptlrpc_cb_id cr_req_cbid; + /** incoming reply MD handle */ + lnet_handle_md_t cr_reply_md_h; + wait_queue_head_t cr_reply_waitq; + /** reply callback parameter */ + struct ptlrpc_cb_id cr_reply_cbid; + /** Async completion handler, called when reply is received */ + ptlrpc_interpterer_t cr_reply_interp; + /** Resend handler, called when request is resend to update RPC data */ + ptlrpc_resend_cb_t cr_resend_cb; + /** Async completion context */ + union ptlrpc_async_args cr_async_args; + /** Opaq data for replay and commit callbacks. */ + void *cr_cb_data; + /** + * Commit callback, called when request is committed and about to be + * freed. + */ + void (*cr_commit_cb)(struct ptlrpc_request *); + /** Replay callback, called after request is replayed at recovery */ + void (*cr_replay_cb)(struct ptlrpc_request *); +}; + +/** client request member alias */ +/* NB: these alias should NOT be used by any new code, instead they should + * be removed step by step to avoid potential abuse */ +#define rq_bulk rq_cli.cr_bulk +#define rq_delay_limit rq_cli.cr_delay_limit +#define rq_queued_time rq_cli.cr_queued_time +#define rq_sent_tv rq_cli.cr_sent_tv +#define rq_real_sent rq_cli.cr_sent_out +#define rq_reply_deadline rq_cli.cr_reply_deadline +#define rq_bulk_deadline rq_cli.cr_bulk_deadline +#define rq_nr_resend rq_cli.cr_resend_nr +#define rq_request_portal rq_cli.cr_req_ptl +#define rq_reply_portal rq_cli.cr_rep_ptl +#define rq_import_generation rq_cli.cr_imp_gen +#define rq_send_state rq_cli.cr_send_state +#define rq_set_chain rq_cli.cr_set_chain +#define rq_ctx_chain rq_cli.cr_ctx_chain +#define rq_set rq_cli.cr_set +#define rq_set_waitq rq_cli.cr_set_waitq +#define rq_cli_ctx rq_cli.cr_cli_ctx +#define rq_req_md_h rq_cli.cr_req_md_h +#define rq_req_cbid rq_cli.cr_req_cbid +#define rq_reply_md_h rq_cli.cr_reply_md_h +#define rq_reply_waitq rq_cli.cr_reply_waitq +#define rq_reply_cbid rq_cli.cr_reply_cbid +#define rq_interpret_reply rq_cli.cr_reply_interp +#define rq_resend_cb rq_cli.cr_resend_cb +#define rq_async_args rq_cli.cr_async_args +#define rq_cb_data rq_cli.cr_cb_data +#define rq_commit_cb rq_cli.cr_commit_cb +#define rq_replay_cb rq_cli.cr_replay_cb + +struct ptlrpc_srv_req { + /** initial thread servicing this request */ + struct ptlrpc_thread *sr_svc_thread; + /** + * Server side list of incoming unserved requests sorted by arrival + * time. Traversed from time to time to notice about to expire + * requests and sent back "early replies" to clients to let them + * know server is alive and well, just very busy to service their + * requests in time + */ + struct list_head sr_timed_list; + /** server-side per-export list */ + struct list_head sr_exp_list; + /** server-side history, used for debuging purposes. */ + struct list_head sr_hist_list; + /** history sequence # */ + __u64 sr_hist_seq; + /** the index of service's srv_at_array into which request is linked */ + time_t sr_at_index; + /** authed uid */ + uid_t sr_auth_uid; + /** authed uid mapped to */ + uid_t sr_auth_mapped_uid; + /** RPC is generated from what part of Lustre */ + enum lustre_sec_part sr_sp_from; + /** request session context */ + struct lu_context sr_ses; + /** \addtogroup nrs + * @{ + */ + /** stub for NRS request */ + struct ptlrpc_nrs_request sr_nrq; + /** @} nrs */ + /** request arrival time */ + struct timeval sr_arrival_time; + /** server's half ctx */ + struct ptlrpc_svc_ctx *sr_svc_ctx; + /** (server side), pointed directly into req buffer */ + struct ptlrpc_user_desc *sr_user_desc; + /** separated reply state */ + struct ptlrpc_reply_state *sr_reply_state; + /** server-side hp handlers */ + struct ptlrpc_hpreq_ops *sr_ops; + /** incoming request buffer */ + struct ptlrpc_request_buffer_desc *sr_rqbd; +}; + +/** server request member alias */ +/* NB: these alias should NOT be used by any new code, instead they should + * be removed step by step to avoid potential abuse */ +#define rq_svc_thread rq_srv.sr_svc_thread +#define rq_timed_list rq_srv.sr_timed_list +#define rq_exp_list rq_srv.sr_exp_list +#define rq_history_list rq_srv.sr_hist_list +#define rq_history_seq rq_srv.sr_hist_seq +#define rq_at_index rq_srv.sr_at_index +#define rq_auth_uid rq_srv.sr_auth_uid +#define rq_auth_mapped_uid rq_srv.sr_auth_mapped_uid +#define rq_sp_from rq_srv.sr_sp_from +#define rq_session rq_srv.sr_ses +#define rq_nrq rq_srv.sr_nrq +#define rq_arrival_time rq_srv.sr_arrival_time +#define rq_reply_state rq_srv.sr_reply_state +#define rq_svc_ctx rq_srv.sr_svc_ctx +#define rq_user_desc rq_srv.sr_user_desc +#define rq_ops rq_srv.sr_ops +#define rq_rqbd rq_srv.sr_rqbd + /** * Represents remote procedure call. * @@ -1343,47 +944,19 @@ struct ptlrpc_hpreq_ops { */ struct ptlrpc_request { /* Request type: one of PTL_RPC_MSG_* */ - int rq_type; + int rq_type; /** Result of request processing */ - int rq_status; - /** - * Linkage item through which this request is included into - * sending/delayed lists on client and into rqbd list on server - */ - cfs_list_t rq_list; - /** - * Server side list of incoming unserved requests sorted by arrival - * time. Traversed from time to time to notice about to expire - * requests and sent back "early replies" to clients to let them - * know server is alive and well, just very busy to service their - * requests in time - */ - cfs_list_t rq_timed_list; - /** server-side history, used for debuging purposes. */ - cfs_list_t rq_history_list; - /** server-side per-export list */ - cfs_list_t rq_exp_list; - /** server-side hp handlers */ - struct ptlrpc_hpreq_ops *rq_ops; - - /** initial thread servicing this request */ - struct ptlrpc_thread *rq_svc_thread; - - /** history sequence # */ - __u64 rq_history_seq; - /** \addtogroup nrs - * @{ + int rq_status; + /** + * Linkage item through which this request is included into + * sending/delayed lists on client and into rqbd list on server */ - /** stub for NRS request */ - struct ptlrpc_nrs_request rq_nrq; - /** @} nrs */ - /** the index of service's srv_at_array into which request is linked */ - time_t rq_at_index; - /** Lock to protect request flags and some other important bits, like - * rq_list - */ - spinlock_t rq_lock; - /** client-side flags are serialized by rq_lock */ + struct list_head rq_list; + /** Lock to protect request flags and some other important bits, like + * rq_list + */ + spinlock_t rq_lock; + /** client-side flags are serialized by rq_lock @{ */ unsigned int rq_intr:1, rq_replied:1, rq_err:1, rq_timedout:1, rq_resend:1, rq_restart:1, /** @@ -1398,35 +971,40 @@ struct ptlrpc_request { rq_replay:1, rq_no_resend:1, rq_waiting:1, rq_receiving_reply:1, rq_no_delay:1, rq_net_err:1, rq_wait_ctx:1, - rq_early:1, rq_must_unlink:1, - rq_memalloc:1, /* req originated from "kswapd" */ - /* server-side flags */ - rq_packed_final:1, /* packed final reply */ - rq_hp:1, /* high priority RPC */ - rq_at_linked:1, /* link into service's srv_at_array */ - rq_reply_truncate:1, - rq_committed:1, - /* whether the "rq_set" is a valid one */ - rq_invalid_rqset:1, + rq_early:1, + rq_req_unlinked:1, /* unlinked request buffer from lnet */ + rq_reply_unlinked:1, /* unlinked reply buffer from lnet */ + rq_memalloc:1, /* req originated from "kswapd" */ + rq_committed:1, + rq_reply_truncated:1, + /** whether the "rq_set" is a valid one */ + rq_invalid_rqset:1, rq_generation_set:1, - /* do not resend request on -EINPROGRESS */ + /** do not resend request on -EINPROGRESS */ rq_no_retry_einprogress:1, /* allow the req to be sent if the import is in recovery * status */ - rq_allow_replay:1; - - unsigned int rq_nr_resend; - - enum rq_phase rq_phase; /* one of RQ_PHASE_* */ - enum rq_phase rq_next_phase; /* one of RQ_PHASE_* to be used next */ - cfs_atomic_t rq_refcount;/* client-side refcount for SENT race, - server-side refcounf for multiple replies */ + rq_allow_replay:1, + /* bulk request, sent to server, but uncommitted */ + rq_unstable:1; + /** @} */ - /** Portal to which this request would be sent */ - short rq_request_portal; /* XXX FIXME bug 249 */ - /** Portal where to wait for reply and where reply would be sent */ - short rq_reply_portal; /* XXX FIXME bug 249 */ + /** server-side flags @{ */ + unsigned int + rq_hp:1, /**< high priority RPC */ + rq_at_linked:1, /**< link into service's srv_at_array */ + rq_packed_final:1; /**< packed final reply */ + /** @} */ + /** one of RQ_PHASE_* */ + enum rq_phase rq_phase; + /** one of RQ_PHASE_* to be used next */ + enum rq_phase rq_next_phase; + /** + * client-side refcount for SENT race, server-side refcounf + * for multiple replies + */ + atomic_t rq_refcount; /** * client-side: * !rq_truncate : # reply bytes actually received, @@ -1437,6 +1015,8 @@ struct ptlrpc_request { int rq_reqlen; /** Reply length */ int rq_replen; + /** Pool if request is from preallocated list */ + struct ptlrpc_request_pool *rq_pool; /** Request message - what client sent */ struct lustre_msg *rq_reqmsg; /** Reply message - server response */ @@ -1444,23 +1024,26 @@ struct ptlrpc_request { /** Transaction number */ __u64 rq_transno; /** xid */ - __u64 rq_xid; - /** - * List item to for replay list. Not yet commited requests get linked - * there. - * Also see \a rq_replay comment above. - */ - cfs_list_t rq_replay_list; - - /** - * security and encryption data - * @{ */ - struct ptlrpc_cli_ctx *rq_cli_ctx; /**< client's half ctx */ - struct ptlrpc_svc_ctx *rq_svc_ctx; /**< server's half ctx */ - cfs_list_t rq_ctx_chain; /**< link to waited ctx */ - - struct sptlrpc_flavor rq_flvr; /**< for client & server */ - enum lustre_sec_part rq_sp_from; + __u64 rq_xid; + /** bulk match bits */ + __u64 rq_mbits; + /** + * List item to for replay list. Not yet commited requests get linked + * there. + * Also see \a rq_replay comment above. + * It's also link chain on obd_export::exp_req_replay_queue + */ + struct list_head rq_replay_list; + /** non-shared members for client & server request*/ + union { + struct ptlrpc_cli_req rq_cli; + struct ptlrpc_srv_req rq_srv; + }; + /** + * security and encryption data + * @{ */ + /** description of flavors for client & server */ + struct sptlrpc_flavor rq_flvr; /* client/server security flags */ unsigned int @@ -1479,19 +1062,16 @@ struct ptlrpc_request { rq_pack_bulk:1, /* doesn't expect reply FIXME */ rq_no_reply:1, - rq_pill_init:1; /* pill initialized */ - - uid_t rq_auth_uid; /* authed uid */ - uid_t rq_auth_mapped_uid; /* authed uid mapped to */ + rq_pill_init:1, /* pill initialized */ + rq_srv_req:1; /* server request */ - /* (server side), pointed directly into req buffer */ - struct ptlrpc_user_desc *rq_user_desc; - /* various buffer pointers */ - struct lustre_msg *rq_reqbuf; /* req wrapper */ - char *rq_repbuf; /* rep buffer */ - struct lustre_msg *rq_repdata; /* rep wrapper msg */ - struct lustre_msg *rq_clrbuf; /* only in priv mode */ + /** various buffer pointers */ + struct lustre_msg *rq_reqbuf; /**< req wrapper */ + char *rq_repbuf; /**< rep buffer */ + struct lustre_msg *rq_repdata; /**< rep wrapper msg */ + /** only in priv mode */ + struct lustre_msg *rq_clrbuf; int rq_reqbuf_len; /* req wrapper buf len */ int rq_reqdata_len; /* req wrapper msg len */ int rq_repbuf_len; /* rep buffer len */ @@ -1500,106 +1080,36 @@ struct ptlrpc_request { int rq_clrdata_len; /* only in priv mode */ /** early replies go to offset 0, regular replies go after that */ - unsigned int rq_reply_off; - - /** @} */ - - /** Fields that help to see if request and reply were swabbed or not */ - __u32 rq_req_swab_mask; - __u32 rq_rep_swab_mask; - - /** What was import generation when this request was sent */ - int rq_import_generation; - enum lustre_imp_state rq_send_state; - - /** how many early replies (for stats) */ - int rq_early_count; - - /** client+server request */ - lnet_handle_md_t rq_req_md_h; - struct ptlrpc_cb_id rq_req_cbid; - /** optional time limit for send attempts */ - cfs_duration_t rq_delay_limit; - /** time request was first queued */ - cfs_time_t rq_queued_time; - - /* server-side... */ - /** request arrival time */ - struct timeval rq_arrival_time; - /** separated reply state */ - struct ptlrpc_reply_state *rq_reply_state; - /** incoming request buffer */ - struct ptlrpc_request_buffer_desc *rq_rqbd; - - /** client-only incoming reply */ - lnet_handle_md_t rq_reply_md_h; - cfs_waitq_t rq_reply_waitq; - struct ptlrpc_cb_id rq_reply_cbid; - - /** our LNet NID */ - lnet_nid_t rq_self; - /** Peer description (the other side) */ - lnet_process_id_t rq_peer; - /** Server-side, export on which request was received */ - struct obd_export *rq_export; - /** Client side, import where request is being sent */ - struct obd_import *rq_import; - - /** Replay callback, called after request is replayed at recovery */ - void (*rq_replay_cb)(struct ptlrpc_request *); - /** - * Commit callback, called when request is committed and about to be - * freed. - */ - void (*rq_commit_cb)(struct ptlrpc_request *); - /** Opaq data for replay and commit callbacks. */ - void *rq_cb_data; - - /** For bulk requests on client only: bulk descriptor */ - struct ptlrpc_bulk_desc *rq_bulk; - - /** client outgoing req */ - /** - * when request/reply sent (secs), or time when request should be sent - */ - time_t rq_sent; - /** time for request really sent out */ - time_t rq_real_sent; - - /** when request must finish. volatile - * so that servers' early reply updates to the deadline aren't - * kept in per-cpu cache */ - volatile time_t rq_deadline; - /** when req reply unlink must finish. */ - time_t rq_reply_deadline; - /** when req bulk unlink must finish. */ - time_t rq_bulk_deadline; - /** - * service time estimate (secs) - * If the requestsis not served by this time, it is marked as timed out. - */ - int rq_timeout; + unsigned int rq_reply_off; + /** @} */ - /** Multi-rpc bits */ - /** Per-request waitq introduced by bug 21938 for recovery waiting */ - cfs_waitq_t rq_set_waitq; - /** Link item for request set lists */ - cfs_list_t rq_set_chain; - /** Link back to the request set */ - struct ptlrpc_request_set *rq_set; - /** Async completion handler, called when reply is received */ - ptlrpc_interpterer_t rq_interpret_reply; - /** Async completion context */ - union ptlrpc_async_args rq_async_args; - - /** Pool if request is from preallocated list */ - struct ptlrpc_request_pool *rq_pool; - - struct lu_context rq_session; - struct lu_context rq_recov_session; - - /** request format description */ - struct req_capsule rq_pill; + /** Fields that help to see if request and reply were swabbed or not */ + __u32 rq_req_swab_mask; + __u32 rq_rep_swab_mask; + + /** how many early replies (for stats) */ + int rq_early_count; + /** Server-side, export on which request was received */ + struct obd_export *rq_export; + /** import where request is being sent */ + struct obd_import *rq_import; + /** our LNet NID */ + lnet_nid_t rq_self; + /** Peer description (the other side) */ + lnet_process_id_t rq_peer; + /** + * service time estimate (secs) + * If the request is not served by this time, it is marked as timed out. + */ + int rq_timeout; + /** + * when request/reply sent (secs), or time when request should be sent + */ + time_t rq_sent; + /** when request must finish. */ + time_t rq_deadline; + /** request format description */ + struct req_capsule rq_pill; }; /** @@ -1621,8 +1131,8 @@ static inline int ptlrpc_req_interpret(const struct lu_env *env, /** \addtogroup nrs * @{ */ -int ptlrpc_nrs_policy_register(struct ptlrpc_nrs_pol_desc *desc); -int ptlrpc_nrs_policy_unregister(struct ptlrpc_nrs_pol_desc *desc); +int ptlrpc_nrs_policy_register(struct ptlrpc_nrs_pol_conf *conf); +int ptlrpc_nrs_policy_unregister(struct ptlrpc_nrs_pol_conf *conf); void ptlrpc_nrs_req_hp_move(struct ptlrpc_request *req); void nrs_policy_get_info_locked(struct ptlrpc_nrs_policy *policy, struct ptlrpc_nrs_pol_info *info); @@ -1633,8 +1143,7 @@ void nrs_policy_get_info_locked(struct ptlrpc_nrs_policy *policy, * * For a reliable result, this should be checked under svcpt->scp_req lock. */ -static inline bool -ptlrpc_nrs_req_can_move(struct ptlrpc_request *req) +static inline bool ptlrpc_nrs_req_can_move(struct ptlrpc_request *req) { struct ptlrpc_nrs_request *nrq = &req->rq_nrq; @@ -1651,7 +1160,7 @@ ptlrpc_nrs_req_can_move(struct ptlrpc_request *req) /** * Returns 1 if request buffer at offset \a index was already swabbed */ -static inline int lustre_req_swabbed(struct ptlrpc_request *req, int index) +static inline int lustre_req_swabbed(struct ptlrpc_request *req, size_t index) { LASSERT(index < sizeof(req->rq_req_swab_mask) * 8); return req->rq_req_swab_mask & (1 << index); @@ -1660,7 +1169,7 @@ static inline int lustre_req_swabbed(struct ptlrpc_request *req, int index) /** * Returns 1 if request reply buffer at offset \a index was already swabbed */ -static inline int lustre_rep_swabbed(struct ptlrpc_request *req, int index) +static inline int lustre_rep_swabbed(struct ptlrpc_request *req, size_t index) { LASSERT(index < sizeof(req->rq_rep_swab_mask) * 8); return req->rq_rep_swab_mask & (1 << index); @@ -1685,7 +1194,8 @@ static inline int ptlrpc_rep_need_swab(struct ptlrpc_request *req) /** * Mark request buffer at offset \a index that it was already swabbed */ -static inline void lustre_set_req_swabbed(struct ptlrpc_request *req, int index) +static inline void lustre_set_req_swabbed(struct ptlrpc_request *req, + size_t index) { LASSERT(index < sizeof(req->rq_req_swab_mask) * 8); LASSERT((req->rq_req_swab_mask & (1 << index)) == 0); @@ -1695,7 +1205,8 @@ static inline void lustre_set_req_swabbed(struct ptlrpc_request *req, int index) /** * Mark request reply buffer at offset \a index that it was already swabbed */ -static inline void lustre_set_rep_swabbed(struct ptlrpc_request *req, int index) +static inline void lustre_set_rep_swabbed(struct ptlrpc_request *req, + size_t index) { LASSERT(index < sizeof(req->rq_rep_swab_mask) * 8); LASSERT((req->rq_rep_swab_mask & (1 << index)) == 0); @@ -1739,7 +1250,7 @@ ptlrpc_rqphase2str(struct ptlrpc_request *req) /** * Debugging functions and helpers to print request structure into debug log * @{ - */ + */ /* Spare the preprocessor, spoil the bugs. */ #define FLAG(field, str) (field ? str : "") @@ -1796,24 +1307,110 @@ do { \ * Structure that defines a single page of a bulk transfer */ struct ptlrpc_bulk_page { - /** Linkage to list of pages in a bulk */ - cfs_list_t bp_link; - /** - * Number of bytes in a page to transfer starting from \a bp_pageoffset - */ - int bp_buflen; - /** offset within a page */ - int bp_pageoffset; - /** The page itself */ - struct page *bp_page; + /** Linkage to list of pages in a bulk */ + struct list_head bp_link; + /** + * Number of bytes in a page to transfer starting from \a bp_pageoffset + */ + int bp_buflen; + /** offset within a page */ + int bp_pageoffset; + /** The page itself */ + struct page *bp_page; }; -#define BULK_GET_SOURCE 0 -#define BULK_PUT_SINK 1 -#define BULK_GET_SINK 2 -#define BULK_PUT_SOURCE 3 +enum ptlrpc_bulk_op_type { + PTLRPC_BULK_OP_ACTIVE = 0x00000001, + PTLRPC_BULK_OP_PASSIVE = 0x00000002, + PTLRPC_BULK_OP_PUT = 0x00000004, + PTLRPC_BULK_OP_GET = 0x00000008, + PTLRPC_BULK_BUF_KVEC = 0x00000010, + PTLRPC_BULK_BUF_KIOV = 0x00000020, + PTLRPC_BULK_GET_SOURCE = PTLRPC_BULK_OP_PASSIVE | PTLRPC_BULK_OP_GET, + PTLRPC_BULK_PUT_SINK = PTLRPC_BULK_OP_PASSIVE | PTLRPC_BULK_OP_PUT, + PTLRPC_BULK_GET_SINK = PTLRPC_BULK_OP_ACTIVE | PTLRPC_BULK_OP_GET, + PTLRPC_BULK_PUT_SOURCE = PTLRPC_BULK_OP_ACTIVE | PTLRPC_BULK_OP_PUT, +}; -/** +static inline bool ptlrpc_is_bulk_op_get(enum ptlrpc_bulk_op_type type) +{ + return (type & PTLRPC_BULK_OP_GET) == PTLRPC_BULK_OP_GET; +} + +static inline bool ptlrpc_is_bulk_get_source(enum ptlrpc_bulk_op_type type) +{ + return (type & PTLRPC_BULK_GET_SOURCE) == PTLRPC_BULK_GET_SOURCE; +} + +static inline bool ptlrpc_is_bulk_put_sink(enum ptlrpc_bulk_op_type type) +{ + return (type & PTLRPC_BULK_PUT_SINK) == PTLRPC_BULK_PUT_SINK; +} + +static inline bool ptlrpc_is_bulk_get_sink(enum ptlrpc_bulk_op_type type) +{ + return (type & PTLRPC_BULK_GET_SINK) == PTLRPC_BULK_GET_SINK; +} + +static inline bool ptlrpc_is_bulk_put_source(enum ptlrpc_bulk_op_type type) +{ + return (type & PTLRPC_BULK_PUT_SOURCE) == PTLRPC_BULK_PUT_SOURCE; +} + +static inline bool ptlrpc_is_bulk_desc_kvec(enum ptlrpc_bulk_op_type type) +{ + return ((type & PTLRPC_BULK_BUF_KVEC) | (type & PTLRPC_BULK_BUF_KIOV)) + == PTLRPC_BULK_BUF_KVEC; +} + +static inline bool ptlrpc_is_bulk_desc_kiov(enum ptlrpc_bulk_op_type type) +{ + return ((type & PTLRPC_BULK_BUF_KVEC) | (type & PTLRPC_BULK_BUF_KIOV)) + == PTLRPC_BULK_BUF_KIOV; +} + +static inline bool ptlrpc_is_bulk_op_active(enum ptlrpc_bulk_op_type type) +{ + return ((type & PTLRPC_BULK_OP_ACTIVE) | + (type & PTLRPC_BULK_OP_PASSIVE)) + == PTLRPC_BULK_OP_ACTIVE; +} + +static inline bool ptlrpc_is_bulk_op_passive(enum ptlrpc_bulk_op_type type) +{ + return ((type & PTLRPC_BULK_OP_ACTIVE) | + (type & PTLRPC_BULK_OP_PASSIVE)) + == PTLRPC_BULK_OP_PASSIVE; +} + +struct ptlrpc_bulk_frag_ops { + /** + * Add a page \a page to the bulk descriptor \a desc + * Data to transfer in the page starts at offset \a pageoffset and + * amount of data to transfer from the page is \a len + */ + void (*add_kiov_frag)(struct ptlrpc_bulk_desc *desc, + struct page *page, int pageoffset, int len); + + /* + * Add a \a fragment to the bulk descriptor \a desc. + * Data to transfer in the fragment is pointed to by \a frag + * The size of the fragment is \a len + */ + int (*add_iov_frag)(struct ptlrpc_bulk_desc *desc, void *frag, int len); + + /** + * Uninitialize and free bulk descriptor \a desc. + * Works on bulk descriptors both from server and client side. + */ + void (*release_frags)(struct ptlrpc_bulk_desc *desc); +}; + +extern const struct ptlrpc_bulk_frag_ops ptlrpc_bulk_kiov_pin_ops; +extern const struct ptlrpc_bulk_frag_ops ptlrpc_bulk_kiov_nopin_ops; +extern const struct ptlrpc_bulk_frag_ops ptlrpc_bulk_kvec_ops; + +/* * Definition of bulk descriptor. * Bulks are special "Two phase" RPCs where initial request message * is sent first and it is followed bt a transfer (o receiving) of a large @@ -1826,14 +1423,14 @@ struct ptlrpc_bulk_page { struct ptlrpc_bulk_desc { /** completed with failure */ unsigned long bd_failure:1; - /** {put,get}{source,sink} */ - unsigned long bd_type:2; /** client side */ unsigned long bd_registered:1; /** For serialization with callback */ spinlock_t bd_lock; /** Import generation when request for this bulk was sent */ int bd_import_generation; + /** {put,get}{source,sink}{kvec,kiov} */ + enum ptlrpc_bulk_op_type bd_type; /** LNet portal for this bulk */ __u32 bd_portal; /** Server side - export this bulk created for */ @@ -1842,13 +1439,14 @@ struct ptlrpc_bulk_desc { struct obd_import *bd_import; /** Back pointer to the request */ struct ptlrpc_request *bd_req; - cfs_waitq_t bd_waitq; /* server side only WQ */ + struct ptlrpc_bulk_frag_ops *bd_frag_ops; + wait_queue_head_t bd_waitq; /* server side only WQ */ int bd_iov_count; /* # entries in bd_iov */ int bd_max_iov; /* allocated size of bd_iov */ int bd_nob; /* # bytes covered */ int bd_nob_transferred; /* # bytes GOT/PUT */ - __u64 bd_last_xid; + __u64 bd_last_mbits; struct ptlrpc_cb_id bd_cbid; /* network callback info */ lnet_nid_t bd_sender; /* stash event::sender */ @@ -1857,18 +1455,32 @@ struct ptlrpc_bulk_desc { /** array of associated MDs */ lnet_handle_md_t bd_mds[PTLRPC_BULK_OPS_COUNT]; -#if defined(__KERNEL__) - /* - * encrypt iov, size is either 0 or bd_iov_count. - */ - lnet_kiov_t *bd_enc_iov; + union { + struct { + /* + * encrypt iov, size is either 0 or bd_iov_count. + */ + lnet_kiov_t *bd_enc_vec; + lnet_kiov_t bd_vec[0]; + } bd_kiov; + + struct { + struct kvec *bd_enc_kvec; + struct kvec bd_kvec[0]; + } bd_kvec; + } bd_u; - lnet_kiov_t bd_iov[0]; -#else - lnet_md_iovec_t bd_iov[0]; -#endif }; +#define GET_KIOV(desc) ((desc)->bd_u.bd_kiov.bd_vec) +#define BD_GET_KIOV(desc, i) ((desc)->bd_u.bd_kiov.bd_vec[i]) +#define GET_ENC_KIOV(desc) ((desc)->bd_u.bd_kiov.bd_enc_vec) +#define BD_GET_ENC_KIOV(desc, i) ((desc)->bd_u.bd_kiov.bd_enc_vec[i]) +#define GET_KVEC(desc) ((desc)->bd_u.bd_kvec.bd_kvec) +#define BD_GET_KVEC(desc, i) ((desc)->bd_u.bd_kvec.bd_kvec[i]) +#define GET_ENC_KVEC(desc) ((desc)->bd_u.bd_kvec.bd_enc_kvec) +#define BD_GET_ENC_KVEC(desc, i) ((desc)->bd_u.bd_kvec.bd_enc_kvec[i]) + enum { SVC_STOPPED = 1 << 0, SVC_STOPPING = 1 << 1, @@ -1886,7 +1498,7 @@ struct ptlrpc_thread { /** * List of active threads in svc->srv_threads */ - cfs_list_t t_link; + struct list_head t_link; /** * thread-private data (preallocated memory) */ @@ -1899,7 +1511,7 @@ struct ptlrpc_thread { /** * service thread pid */ - pid_t t_pid; + pid_t t_pid; /** * put watchdog in the structure per thread b=14840 */ @@ -1908,7 +1520,7 @@ struct ptlrpc_thread { * the svc this thread belonged to b=18582 */ struct ptlrpc_service_part *t_svcpt; - cfs_waitq_t t_ctl_waitq; + wait_queue_head_t t_ctl_waitq; struct lu_env *t_env; char t_name[PTLRPC_THR_NAME_LEN]; }; @@ -1981,23 +1593,23 @@ static inline int thread_test_and_clear_flags(struct ptlrpc_thread *thread, * More than one request can fit into the buffer. */ struct ptlrpc_request_buffer_desc { - /** Link item for rqbds on a service */ - cfs_list_t rqbd_list; - /** History of requests for this buffer */ - cfs_list_t rqbd_reqs; - /** Back pointer to service for which this buffer is registered */ - struct ptlrpc_service_part *rqbd_svcpt; - /** LNet descriptor */ - lnet_handle_md_t rqbd_md_h; - int rqbd_refcount; - /** The buffer itself */ - char *rqbd_buffer; - struct ptlrpc_cb_id rqbd_cbid; - /** - * This "embedded" request structure is only used for the - * last request to fit into the buffer - */ - struct ptlrpc_request rqbd_req; + /** Link item for rqbds on a service */ + struct list_head rqbd_list; + /** History of requests for this buffer */ + struct list_head rqbd_reqs; + /** Back pointer to service for which this buffer is registered */ + struct ptlrpc_service_part *rqbd_svcpt; + /** LNet descriptor */ + lnet_handle_md_t rqbd_md_h; + int rqbd_refcount; + /** The buffer itself */ + char *rqbd_buffer; + struct ptlrpc_cb_id rqbd_cbid; + /** + * This "embedded" request structure is only used for the + * last request to fit into the buffer + */ + struct ptlrpc_request rqbd_req; }; typedef int (*svc_handler_t)(struct ptlrpc_request *req); @@ -2048,9 +1660,9 @@ struct ptlrpc_service_ops { struct ptlrpc_service { /** serialize /proc operations */ spinlock_t srv_lock; - /** most often accessed fields */ - /** chain thru all services */ - cfs_list_t srv_list; + /** most often accessed fields */ + /** chain thru all services */ + struct list_head srv_list; /** service operations table */ struct ptlrpc_service_ops srv_ops; /** only statically allocated strings here; we don't clean them */ @@ -2058,13 +1670,13 @@ struct ptlrpc_service { /** only statically allocated strings here; we don't clean them */ char *srv_thread_name; /** service thread list */ - cfs_list_t srv_threads; + struct list_head srv_threads; /** threads # should be created for each partition on initializing */ int srv_nthrs_cpt_init; /** limit of threads number for each partition */ int srv_nthrs_cpt_limit; /** Root of /proc dir tree for this service */ - cfs_proc_dir_entry_t *srv_procroot; + struct proc_dir_entry *srv_procroot; /** Pointer to statistic data for this service */ struct lprocfs_stats *srv_stats; /** # hp per lp reqs to handle */ @@ -2139,7 +1751,7 @@ struct ptlrpc_service_part { /** # running threads */ int scp_nthrs_running; /** service threads list */ - cfs_list_t scp_threads; + struct list_head scp_threads; /** * serialize the following fields, used for protecting @@ -2156,23 +1768,23 @@ struct ptlrpc_service_part { /** # incoming reqs */ int scp_nreqs_incoming; /** request buffers to be reposted */ - cfs_list_t scp_rqbd_idle; + struct list_head scp_rqbd_idle; /** req buffers receiving */ - cfs_list_t scp_rqbd_posted; + struct list_head scp_rqbd_posted; /** incoming reqs */ - cfs_list_t scp_req_incoming; + struct list_head scp_req_incoming; /** timeout before re-posting reqs, in tick */ cfs_duration_t scp_rqbd_timeout; /** * all threads sleep on this. This wait-queue is signalled when new * incoming request arrives and when difficult reply has to be handled. */ - cfs_waitq_t scp_waitq; + wait_queue_head_t scp_waitq; /** request history */ - cfs_list_t scp_hist_reqs; + struct list_head scp_hist_reqs; /** request buffer history */ - cfs_list_t scp_hist_rqbds; + struct list_head scp_hist_rqbds; /** # request buffers in history */ int scp_hist_nrqbds; /** sequence number for request */ @@ -2211,7 +1823,7 @@ struct ptlrpc_service_part { /** reqs waiting for replies */ struct ptlrpc_at_array scp_at_array; /** early reply timer */ - cfs_timer_t scp_at_timer; + struct timer_list scp_at_timer; /** debug */ cfs_time_t scp_at_checktime; /** check early replies */ @@ -2224,17 +1836,13 @@ struct ptlrpc_service_part { */ spinlock_t scp_rep_lock __cfs_cacheline_aligned; /** all the active replies */ - cfs_list_t scp_rep_active; -#ifndef __KERNEL__ - /** replies waiting for service */ - cfs_list_t scp_rep_queue; -#endif + struct list_head scp_rep_active; /** List of free reply_states */ - cfs_list_t scp_rep_idle; + struct list_head scp_rep_idle; /** waitq to run, when adding stuff to srv_free_rs_list */ - cfs_waitq_t scp_rep_waitq; + wait_queue_head_t scp_rep_waitq; /** # 'difficult' replies */ - cfs_atomic_t scp_nreps_difficult; + atomic_t scp_nreps_difficult; }; #define ptlrpc_service_for_each_part(part, i, svc) \ @@ -2263,53 +1871,42 @@ struct ptlrpcd_ctl { * Stop completion. */ struct completion pc_finishing; - /** - * Thread requests set. - */ - struct ptlrpc_request_set *pc_set; - /** - * Thread name used in cfs_daemonize() - */ - char pc_name[16]; - /** - * Environment for request interpreters to run in. - */ - struct lu_env pc_env; + /** + * Thread requests set. + */ + struct ptlrpc_request_set *pc_set; + /** + * Thread name used in kthread_run() + */ + char pc_name[16]; + /** + * Environment for request interpreters to run in. + */ + struct lu_env pc_env; + /** + * CPT the thread is bound on. + */ + int pc_cpt; /** * Index of ptlrpcd thread in the array. */ - int pc_index; - /** - * Number of the ptlrpcd's partners. - */ - int pc_npartners; - /** - * Pointer to the array of partners' ptlrpcd_ctl structure. - */ - struct ptlrpcd_ctl **pc_partners; - /** - * Record the partner index to be processed next. - */ - int pc_cursor; -#ifndef __KERNEL__ - /** - * Async rpcs flag to make sure that ptlrpcd_check() is called only - * once. - */ - int pc_recurred; - /** - * Currently not used. - */ - void *pc_callback; - /** - * User-space async rpcs callback. - */ - void *pc_wait_callback; - /** - * User-space check idle rpcs callback. - */ - void *pc_idle_callback; -#endif + int pc_index; + /** + * Pointer to the array of partners' ptlrpcd_ctl structure. + */ + struct ptlrpcd_ctl **pc_partners; + /** + * Number of the ptlrpcd's partners. + */ + int pc_npartners; + /** + * Record the partner index to be processed next. + */ + int pc_cursor; + /** + * Error code if the thread failed to fully start. + */ + int pc_error; }; /* Bits for pc_flags */ @@ -2332,48 +1929,42 @@ enum ptlrpcd_ctl_flags { * This is a recovery ptlrpc thread. */ LIOD_RECOVERY = 1 << 3, - /** - * The ptlrpcd is bound to some CPU core. - */ - LIOD_BIND = 1 << 4, }; /** * \addtogroup nrs * @{ * - * Service compatibility function; policy is compatible with all services. + * Service compatibility function; the policy is compatible with all services. * * \param[in] svc The service the policy is attempting to register with. * \param[in] desc The policy descriptor * - * \retval true The policy is compatible with the NRS head + * \retval true The policy is compatible with the service * * \see ptlrpc_nrs_pol_desc::pd_compat() */ -static inline bool -nrs_policy_compat_all(struct ptlrpc_service *svc, - const struct ptlrpc_nrs_pol_desc *desc) +static inline bool nrs_policy_compat_all(const struct ptlrpc_service *svc, + const struct ptlrpc_nrs_pol_desc *desc) { return true; } /** - * Service compatibility function; policy is compatible with only a specific + * Service compatibility function; the policy is compatible with only a specific * service which is identified by its human-readable name at * ptlrpc_service::srv_name. * * \param[in] svc The service the policy is attempting to register with. * \param[in] desc The policy descriptor * - * \retval false The policy is not compatible with the NRS head - * \retval true The policy is compatible with the NRS head + * \retval false The policy is not compatible with the service + * \retval true The policy is compatible with the service * * \see ptlrpc_nrs_pol_desc::pd_compat() */ -static inline bool -nrs_policy_compat_one(struct ptlrpc_service *svc, - const struct ptlrpc_nrs_pol_desc *desc) +static inline bool nrs_policy_compat_one(const struct ptlrpc_service *svc, + const struct ptlrpc_nrs_pol_desc *desc) { LASSERT(desc->pd_compat_svc_name != NULL); return strcmp(svc->srv_name, desc->pd_compat_svc_name) == 0; @@ -2417,8 +2008,11 @@ extern lnet_pid_t ptl_get_pid(void); */ #ifdef HAVE_SERVER_SUPPORT struct ptlrpc_bulk_desc *ptlrpc_prep_bulk_exp(struct ptlrpc_request *req, - unsigned npages, unsigned max_brw, - unsigned type, unsigned portal); + unsigned nfrags, unsigned max_brw, + unsigned int type, + unsigned portal, + const struct ptlrpc_bulk_frag_ops + *ops); int ptlrpc_start_bulk_transfer(struct ptlrpc_bulk_desc *desc); void ptlrpc_abort_bulk(struct ptlrpc_bulk_desc *desc); @@ -2465,7 +2059,6 @@ int ptlrpc_send_reply(struct ptlrpc_request *req, int flags); int ptlrpc_reply(struct ptlrpc_request *req); int ptlrpc_send_error(struct ptlrpc_request *req, int difficult); int ptlrpc_error(struct ptlrpc_request *req); -void ptlrpc_resend_req(struct ptlrpc_request *request); int ptlrpc_at_get_net_latency(struct ptlrpc_request *req); int ptl_send_rpc(struct ptlrpc_request *request, int noreply); int ptlrpc_register_rqbd(struct ptlrpc_request_buffer_desc *rqbd); @@ -2477,6 +2070,8 @@ int ptlrpc_register_rqbd(struct ptlrpc_request_buffer_desc *rqbd); * request queues, request management, etc. * @{ */ +void ptlrpc_request_committed(struct ptlrpc_request *req, int force); + void ptlrpc_init_client(int req_portal, int rep_portal, char *name, struct ptlrpc_client *); void ptlrpc_cleanup_client(struct obd_import *imp); @@ -2484,7 +2079,6 @@ struct ptlrpc_connection *ptlrpc_uuid_to_connection(struct obd_uuid *uuid); int ptlrpc_queue_wait(struct ptlrpc_request *req); int ptlrpc_replay_req(struct ptlrpc_request *req); -int ptlrpc_unregister_reply(struct ptlrpc_request *req, int async); void ptlrpc_restart_req(struct ptlrpc_request *req); void ptlrpc_abort_inflight(struct obd_import *imp); void ptlrpc_cleanup_imp(struct obd_import *imp); @@ -2495,23 +2089,18 @@ struct ptlrpc_request_set *ptlrpc_prep_fcset(int max, set_producer_func func, void *arg); int ptlrpc_set_add_cb(struct ptlrpc_request_set *set, set_interpreter_func fn, void *data); -int ptlrpc_set_next_timeout(struct ptlrpc_request_set *); int ptlrpc_check_set(const struct lu_env *env, struct ptlrpc_request_set *set); int ptlrpc_set_wait(struct ptlrpc_request_set *); -int ptlrpc_expired_set(void *data); -void ptlrpc_interrupted_set(void *data); void ptlrpc_mark_interrupted(struct ptlrpc_request *req); void ptlrpc_set_destroy(struct ptlrpc_request_set *); void ptlrpc_set_add_req(struct ptlrpc_request_set *, struct ptlrpc_request *); -void ptlrpc_set_add_new_req(struct ptlrpcd_ctl *pc, - struct ptlrpc_request *req); void ptlrpc_free_rq_pool(struct ptlrpc_request_pool *pool); -void ptlrpc_add_rqs_to_pool(struct ptlrpc_request_pool *pool, int num_rq); +int ptlrpc_add_rqs_to_pool(struct ptlrpc_request_pool *pool, int num_rq); struct ptlrpc_request_pool * ptlrpc_init_rq_pool(int, int, - void (*populate_pool)(struct ptlrpc_request_pool *, int)); + int (*populate_pool)(struct ptlrpc_request_pool *, int)); void ptlrpc_at_set_req_timeout(struct ptlrpc_request *req); struct ptlrpc_request *ptlrpc_request_alloc(struct obd_import *imp, @@ -2539,33 +2128,45 @@ void ptlrpc_req_finished(struct ptlrpc_request *request); void ptlrpc_req_finished_with_imp_lock(struct ptlrpc_request *request); struct ptlrpc_request *ptlrpc_request_addref(struct ptlrpc_request *req); struct ptlrpc_bulk_desc *ptlrpc_prep_bulk_imp(struct ptlrpc_request *req, - unsigned npages, unsigned max_brw, - unsigned type, unsigned portal); -void __ptlrpc_free_bulk(struct ptlrpc_bulk_desc *bulk, int pin); -static inline void ptlrpc_free_bulk_pin(struct ptlrpc_bulk_desc *bulk) -{ - __ptlrpc_free_bulk(bulk, 1); -} -static inline void ptlrpc_free_bulk_nopin(struct ptlrpc_bulk_desc *bulk) -{ - __ptlrpc_free_bulk(bulk, 0); -} + unsigned nfrags, unsigned max_brw, + unsigned int type, + unsigned portal, + const struct ptlrpc_bulk_frag_ops + *ops); + +int ptlrpc_prep_bulk_frag(struct ptlrpc_bulk_desc *desc, + void *frag, int len); void __ptlrpc_prep_bulk_page(struct ptlrpc_bulk_desc *desc, - cfs_page_t *page, int pageoffset, int len, int); + struct page *page, int pageoffset, int len, + int pin); static inline void ptlrpc_prep_bulk_page_pin(struct ptlrpc_bulk_desc *desc, - cfs_page_t *page, int pageoffset, + struct page *page, int pageoffset, int len) { __ptlrpc_prep_bulk_page(desc, page, pageoffset, len, 1); } static inline void ptlrpc_prep_bulk_page_nopin(struct ptlrpc_bulk_desc *desc, - cfs_page_t *page, int pageoffset, + struct page *page, int pageoffset, int len) { __ptlrpc_prep_bulk_page(desc, page, pageoffset, len, 0); } +void ptlrpc_free_bulk(struct ptlrpc_bulk_desc *bulk); + +static inline void ptlrpc_release_bulk_page_pin(struct ptlrpc_bulk_desc *desc) +{ + int i; + + for (i = 0; i < desc->bd_iov_count ; i++) + page_cache_release(BD_GET_KIOV(desc, i).kiov_page); +} + +static inline void ptlrpc_release_bulk_noop(struct ptlrpc_bulk_desc *desc) +{ +} + void ptlrpc_retain_replayable_request(struct ptlrpc_request *req, struct obd_import *imp); __u64 ptlrpc_next_xid(void); @@ -2580,7 +2181,7 @@ int ptlrpcd_queue_work(void *handler); /** @} */ struct ptlrpc_service_buf_conf { - /* nbufs is how many buffers to post */ + /* nbufs is buffers # to allocate when growing the pool */ unsigned int bc_nbufs; /* buffer size to post */ unsigned int bc_buf_size; @@ -2664,14 +2265,12 @@ int liblustre_check_services(void *arg); void ptlrpc_daemonize(char *name); int ptlrpc_service_health_check(struct ptlrpc_service *); void ptlrpc_server_drop_request(struct ptlrpc_request *req); +void ptlrpc_request_change_export(struct ptlrpc_request *req, + struct obd_export *export); +void ptlrpc_update_export_timer(struct obd_export *exp, long extra_delay); -#ifdef __KERNEL__ int ptlrpc_hr_init(void); void ptlrpc_hr_fini(void); -#else -# define ptlrpc_hr_init() (0) -# define ptlrpc_hr_fini() do {} while(0) -#endif /** @} */ @@ -2685,21 +2284,21 @@ int ptlrpc_init_import(struct obd_import *imp); int ptlrpc_disconnect_import(struct obd_import *imp, int noclose); int ptlrpc_import_recovery_state_machine(struct obd_import *imp); void deuuidify(char *uuid, const char *prefix, char **uuid_start, - int *uuid_len); - + int *uuid_len); +void ptlrpc_import_enter_resend(struct obd_import *imp); /* ptlrpc/pack_generic.c */ int ptlrpc_reconnect_import(struct obd_import *imp); /** @} */ /** - * ptlrpc msg buffer and swab interface + * ptlrpc msg buffer and swab interface * * @{ */ int ptlrpc_buf_need_swab(struct ptlrpc_request *req, const int inout, - int index); + __u32 index); void ptlrpc_buf_set_swabbed(struct ptlrpc_request *req, const int inout, - int index); + __u32 index); int ptlrpc_unpack_rep_msg(struct ptlrpc_request *req, int len); int ptlrpc_unpack_req_msg(struct ptlrpc_request *req, int len); @@ -2719,32 +2318,32 @@ int lustre_shrink_msg(struct lustre_msg *msg, int segment, unsigned int newlen, int move_data); void lustre_free_reply_state(struct ptlrpc_reply_state *rs); int __lustre_unpack_msg(struct lustre_msg *m, int len); -int lustre_msg_hdr_size(__u32 magic, int count); -int lustre_msg_size(__u32 magic, int count, __u32 *lengths); -int lustre_msg_size_v2(int count, __u32 *lengths); -int lustre_packed_msg_size(struct lustre_msg *msg); -int lustre_msg_early_size(void); -void *lustre_msg_buf_v2(struct lustre_msg_v2 *m, int n, int min_size); -void *lustre_msg_buf(struct lustre_msg *m, int n, int minlen); -int lustre_msg_buflen(struct lustre_msg *m, int n); -void lustre_msg_set_buflen(struct lustre_msg *m, int n, int len); -int lustre_msg_bufcount(struct lustre_msg *m); -char *lustre_msg_string(struct lustre_msg *m, int n, int max_len); +__u32 lustre_msg_hdr_size(__u32 magic, __u32 count); +__u32 lustre_msg_size(__u32 magic, int count, __u32 *lengths); +__u32 lustre_msg_size_v2(int count, __u32 *lengths); +__u32 lustre_packed_msg_size(struct lustre_msg *msg); +__u32 lustre_msg_early_size(void); +void *lustre_msg_buf_v2(struct lustre_msg_v2 *m, __u32 n, __u32 min_size); +void *lustre_msg_buf(struct lustre_msg *m, __u32 n, __u32 minlen); +__u32 lustre_msg_buflen(struct lustre_msg *m, __u32 n); +void lustre_msg_set_buflen(struct lustre_msg *m, __u32 n, __u32 len); +__u32 lustre_msg_bufcount(struct lustre_msg *m); +char *lustre_msg_string(struct lustre_msg *m, __u32 n, __u32 max_len); __u32 lustre_msghdr_get_flags(struct lustre_msg *msg); void lustre_msghdr_set_flags(struct lustre_msg *msg, __u32 flags); __u32 lustre_msg_get_flags(struct lustre_msg *msg); -void lustre_msg_add_flags(struct lustre_msg *msg, int flags); -void lustre_msg_set_flags(struct lustre_msg *msg, int flags); -void lustre_msg_clear_flags(struct lustre_msg *msg, int flags); +void lustre_msg_add_flags(struct lustre_msg *msg, __u32 flags); +void lustre_msg_set_flags(struct lustre_msg *msg, __u32 flags); +void lustre_msg_clear_flags(struct lustre_msg *msg, __u32 flags); __u32 lustre_msg_get_op_flags(struct lustre_msg *msg); -void lustre_msg_add_op_flags(struct lustre_msg *msg, int flags); -void lustre_msg_set_op_flags(struct lustre_msg *msg, int flags); +void lustre_msg_add_op_flags(struct lustre_msg *msg, __u32 flags); struct lustre_handle *lustre_msg_get_handle(struct lustre_msg *msg); __u32 lustre_msg_get_type(struct lustre_msg *msg); __u32 lustre_msg_get_version(struct lustre_msg *msg); -void lustre_msg_add_version(struct lustre_msg *msg, int version); +void lustre_msg_add_version(struct lustre_msg *msg, __u32 version); __u32 lustre_msg_get_opc(struct lustre_msg *msg); __u64 lustre_msg_get_last_xid(struct lustre_msg *msg); +__u16 lustre_msg_get_tag(struct lustre_msg *msg); __u64 lustre_msg_get_last_committed(struct lustre_msg *msg); __u64 *lustre_msg_get_versions(struct lustre_msg *msg); __u64 lustre_msg_get_transno(struct lustre_msg *msg); @@ -2754,22 +2353,18 @@ void lustre_msg_set_slv(struct lustre_msg *msg, __u64 slv); void lustre_msg_set_limit(struct lustre_msg *msg, __u64 limit); int lustre_msg_get_status(struct lustre_msg *msg); __u32 lustre_msg_get_conn_cnt(struct lustre_msg *msg); -int lustre_msg_is_v1(struct lustre_msg *msg); __u32 lustre_msg_get_magic(struct lustre_msg *msg); __u32 lustre_msg_get_timeout(struct lustre_msg *msg); __u32 lustre_msg_get_service_time(struct lustre_msg *msg); char *lustre_msg_get_jobid(struct lustre_msg *msg); __u32 lustre_msg_get_cksum(struct lustre_msg *msg); -#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0) -__u32 lustre_msg_calc_cksum(struct lustre_msg *msg, int compat18); -#else -# warning "remove checksum compatibility support for b1_8" +__u64 lustre_msg_get_mbits(struct lustre_msg *msg); __u32 lustre_msg_calc_cksum(struct lustre_msg *msg); -#endif void lustre_msg_set_handle(struct lustre_msg *msg,struct lustre_handle *handle); void lustre_msg_set_type(struct lustre_msg *msg, __u32 type); void lustre_msg_set_opc(struct lustre_msg *msg, __u32 opc); void lustre_msg_set_last_xid(struct lustre_msg *msg, __u64 last_xid); +void lustre_msg_set_tag(struct lustre_msg *msg, __u16 tag); void lustre_msg_set_last_committed(struct lustre_msg *msg,__u64 last_committed); void lustre_msg_set_versions(struct lustre_msg *msg, __u64 *versions); void lustre_msg_set_transno(struct lustre_msg *msg, __u64 transno); @@ -2781,6 +2376,7 @@ void lustre_msg_set_timeout(struct lustre_msg *msg, __u32 timeout); void lustre_msg_set_service_time(struct lustre_msg *msg, __u32 service_time); void lustre_msg_set_jobid(struct lustre_msg *msg, char *jobid); void lustre_msg_set_cksum(struct lustre_msg *msg, __u32 cksum); +void lustre_msg_set_mbits(struct lustre_msg *msg, __u64 mbits); static inline void lustre_shrink_reply(struct ptlrpc_request *req, int segment, @@ -2791,34 +2387,66 @@ lustre_shrink_reply(struct ptlrpc_request *req, int segment, req->rq_replen = lustre_shrink_msg(req->rq_repmsg, segment, newlen, move_data); } + +#ifdef LUSTRE_TRANSLATE_ERRNOS + +static inline int ptlrpc_status_hton(int h) +{ + /* + * Positive errnos must be network errnos, such as LUSTRE_EDEADLK, + * ELDLM_LOCK_ABORTED, etc. + */ + if (h < 0) + return -lustre_errno_hton(-h); + else + return h; +} + +static inline int ptlrpc_status_ntoh(int n) +{ + /* + * See the comment in ptlrpc_status_hton(). + */ + if (n < 0) + return -lustre_errno_ntoh(-n); + else + return n; +} + +#else + +#define ptlrpc_status_hton(h) (h) +#define ptlrpc_status_ntoh(n) (n) + +#endif /** @} */ /** Change request phase of \a req to \a new_phase */ static inline void ptlrpc_rqphase_move(struct ptlrpc_request *req, enum rq_phase new_phase) { - if (req->rq_phase == new_phase) - return; + if (req->rq_phase == new_phase) + return; - if (new_phase == RQ_PHASE_UNREGISTERING) { - req->rq_next_phase = req->rq_phase; - if (req->rq_import) - cfs_atomic_inc(&req->rq_import->imp_unregistering); - } + if (new_phase == RQ_PHASE_UNREGISTERING) { + req->rq_next_phase = req->rq_phase; + if (req->rq_import) + atomic_inc(&req->rq_import->imp_unregistering); + } - if (req->rq_phase == RQ_PHASE_UNREGISTERING) { - if (req->rq_import) - cfs_atomic_dec(&req->rq_import->imp_unregistering); - } + if (req->rq_phase == RQ_PHASE_UNREGISTERING) { + if (req->rq_import) + atomic_dec(&req->rq_import->imp_unregistering); + } - DEBUG_REQ(D_INFO, req, "move req \"%s\" -> \"%s\"", - ptlrpc_rqphase2str(req), ptlrpc_phase2str(new_phase)); + DEBUG_REQ(D_INFO, req, "move req \"%s\" -> \"%s\"", + ptlrpc_rqphase2str(req), ptlrpc_phase2str(new_phase)); - req->rq_phase = new_phase; + req->rq_phase = new_phase; } /** - * Returns true if request \a req got early reply and hard deadline is not met + * Returns true if request \a req got early reply and hard deadline is not met */ static inline int ptlrpc_client_early(struct ptlrpc_request *req) @@ -2862,7 +2490,8 @@ ptlrpc_client_recv_or_unlink(struct ptlrpc_request *req) spin_unlock(&req->rq_lock); return 1; } - rc = req->rq_receiving_reply || req->rq_must_unlink; + rc = !req->rq_req_unlinked || !req->rq_reply_unlinked || + req->rq_receiving_reply; spin_unlock(&req->rq_lock); return rc; } @@ -2870,25 +2499,25 @@ ptlrpc_client_recv_or_unlink(struct ptlrpc_request *req) static inline void ptlrpc_client_wake_req(struct ptlrpc_request *req) { - if (req->rq_set == NULL) - cfs_waitq_signal(&req->rq_reply_waitq); - else - cfs_waitq_signal(&req->rq_set->set_waitq); + if (req->rq_set == NULL) + wake_up(&req->rq_reply_waitq); + else + wake_up(&req->rq_set->set_waitq); } static inline void ptlrpc_rs_addref(struct ptlrpc_reply_state *rs) { - LASSERT(cfs_atomic_read(&rs->rs_refcount) > 0); - cfs_atomic_inc(&rs->rs_refcount); + LASSERT(atomic_read(&rs->rs_refcount) > 0); + atomic_inc(&rs->rs_refcount); } static inline void ptlrpc_rs_decref(struct ptlrpc_reply_state *rs) { - LASSERT(cfs_atomic_read(&rs->rs_refcount) > 0); - if (cfs_atomic_dec_and_test(&rs->rs_refcount)) - lustre_free_reply_state(rs); + LASSERT(atomic_read(&rs->rs_refcount) > 0); + if (atomic_dec_and_test(&rs->rs_refcount)) + lustre_free_reply_state(rs); } /* Should only be called once per req */ @@ -2985,7 +2614,6 @@ int server_disconnect_export(struct obd_export *exp); * Pinger API (client side only) * @{ */ -extern int suppress_pings; enum timeout_event { TIMEOUT_GRANT = 1 }; @@ -2994,59 +2622,24 @@ typedef int (*timeout_cb_t)(struct timeout_item *, void *); int ptlrpc_pinger_add_import(struct obd_import *imp); int ptlrpc_pinger_del_import(struct obd_import *imp); int ptlrpc_add_timeout_client(int time, enum timeout_event event, - timeout_cb_t cb, void *data, - cfs_list_t *obd_list); -int ptlrpc_del_timeout_client(cfs_list_t *obd_list, + timeout_cb_t cb, void *data, + struct list_head *obd_list); +int ptlrpc_del_timeout_client(struct list_head *obd_list, enum timeout_event event); struct ptlrpc_request * ptlrpc_prep_ping(struct obd_import *imp); int ptlrpc_obd_ping(struct obd_device *obd); -cfs_time_t ptlrpc_suspend_wakeup_time(void); -#ifdef __KERNEL__ void ping_evictor_start(void); void ping_evictor_stop(void); -#else -#define ping_evictor_start() do {} while (0) -#define ping_evictor_stop() do {} while (0) -#endif -int ptlrpc_check_and_wait_suspend(struct ptlrpc_request *req); +void ptlrpc_pinger_ir_up(void); +void ptlrpc_pinger_ir_down(void); /** @} */ - -/* ptlrpc daemon bind policy */ -typedef enum { - /* all ptlrpcd threads are free mode */ - PDB_POLICY_NONE = 1, - /* all ptlrpcd threads are bound mode */ - PDB_POLICY_FULL = 2, - /* ... */ - PDB_POLICY_PAIR = 3, - /* ... , - * means each ptlrpcd[X] has two partners: thread[X-1] and thread[X+1]. - * If kernel supports NUMA, pthrpcd threads are binded and - * grouped by NUMA node */ - PDB_POLICY_NEIGHBOR = 4, -} pdb_policy_t; - -/* ptlrpc daemon load policy - * It is caller's duty to specify how to push the async RPC into some ptlrpcd - * queue, but it is not enforced, affected by "ptlrpcd_bind_policy". If it is - * "PDB_POLICY_FULL", then the RPC will be processed by the selected ptlrpcd, - * Otherwise, the RPC may be processed by the selected ptlrpcd or its partner, - * depends on which is scheduled firstly, to accelerate the RPC processing. */ -typedef enum { - /* on the same CPU core as the caller */ - PDL_POLICY_SAME = 1, - /* within the same CPU partition, but not the same core as the caller */ - PDL_POLICY_LOCAL = 2, - /* round-robin on all CPU cores, but not the same core as the caller */ - PDL_POLICY_ROUND = 3, - /* the specified CPU core is preferred, but not enforced */ - PDL_POLICY_PREFERRED = 4, -} pdl_policy_t; +int ptlrpc_pinger_suppress_pings(void); /* ptlrpc/ptlrpcd.c */ void ptlrpcd_stop(struct ptlrpcd_ctl *pc, int force); +void ptlrpcd_free(struct ptlrpcd_ctl *pc); void ptlrpcd_wake(struct ptlrpc_request *req); -void ptlrpcd_add_req(struct ptlrpc_request *req, pdl_policy_t policy, int idx); +void ptlrpcd_add_req(struct ptlrpc_request *req); void ptlrpcd_add_rqset(struct ptlrpc_request_set *set); int ptlrpcd_addref(void); void ptlrpcd_decref(void); @@ -3057,7 +2650,7 @@ void ptlrpcd_decref(void); * @{ */ const char* ll_opcode2str(__u32 opcode); -#ifdef LPROCFS +#ifdef CONFIG_PROC_FS void ptlrpc_lprocfs_register_obd(struct obd_device *obd); void ptlrpc_lprocfs_unregister_obd(struct obd_device *obd); void ptlrpc_lprocfs_brw(struct ptlrpc_request *req, int bytes); @@ -3075,11 +2668,9 @@ int llog_origin_handle_prev_block(struct ptlrpc_request *req); int llog_origin_handle_next_block(struct ptlrpc_request *req); int llog_origin_handle_read_header(struct ptlrpc_request *req); int llog_origin_handle_close(struct ptlrpc_request *req); -int llog_origin_handle_cancel(struct ptlrpc_request *req); /* ptlrpc/llog_client.c */ extern struct llog_operations llog_client_ops; - /** @} net */ #endif