X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Fmdc%2Fmdc_lib.c;h=1e861538de75cc582febcc9e25468b0de97dcd86;hp=7094b5ac254aea4d3a5983e2382621aebe550a2e;hb=98060d83459ba10409f295898f0ec917f938b4d3;hpb=80ea65839c878f18b1471d89f4db0b6ba0450813 diff --git a/lustre/mdc/mdc_lib.c b/lustre/mdc/mdc_lib.c index 7094b5a..1e86153 100644 --- a/lustre/mdc/mdc_lib.c +++ b/lustre/mdc/mdc_lib.c @@ -1,6 +1,4 @@ -/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*- - * vim:expandtab:shiftwidth=8:tabstop=8: - * +/* * GPL HEADER START * * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. @@ -26,8 +24,10 @@ * GPL HEADER END */ /* - * Copyright 2008 Sun Microsystems, Inc. All rights reserved + * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved. * Use is subject to license terms. + * + * Copyright (c) 2011, 2013, Intel Corporation. */ /* * This file is part of Lustre, http://www.lustre.org/ @@ -39,8 +39,8 @@ # include # include #endif -#include #include +#include #include "mdc_internal.h" #ifndef __KERNEL__ @@ -55,10 +55,10 @@ static void __mdc_pack_body(struct mdt_body *b, __u32 suppgid) LASSERT (b != NULL); b->suppgid = suppgid; - b->uid = current->uid; - b->gid = current->gid; - b->fsuid = current->fsuid; - b->fsgid = current->fsgid; + b->uid = cfs_curproc_uid(); + b->gid = cfs_curproc_gid(); + b->fsuid = cfs_curproc_fsuid(); + b->fsgid = cfs_curproc_fsgid(); b->capability = cfs_curproc_cap_pack(); } @@ -94,6 +94,21 @@ void mdc_is_subdir_pack(struct ptlrpc_request *req, const struct lu_fid *pfid, b->flags = flags; } +void mdc_swap_layouts_pack(struct ptlrpc_request *req, + struct md_op_data *op_data) +{ + struct mdt_body *b = req_capsule_client_get(&req->rq_pill, + &RMF_MDT_BODY); + + __mdc_pack_body(b, op_data->op_suppgids[0]); + b->fid1 = op_data->op_fid1; + b->fid2 = op_data->op_fid2; + b->valid |= OBD_MD_FLID; + + mdc_pack_capa(req, &RMF_CAPA1, op_data->op_capa1); + mdc_pack_capa(req, &RMF_CAPA2, op_data->op_capa2); +} + void mdc_pack_body(struct ptlrpc_request *req, const struct lu_fid *fid, struct obd_capa *oc, __u64 valid, int ea_size, __u32 suppgid, int flags) @@ -132,43 +147,49 @@ void mdc_create_pack(struct ptlrpc_request *req, struct md_op_data *op_data, const void *data, int datalen, __u32 mode, __u32 uid, __u32 gid, cfs_cap_t cap_effective, __u64 rdev) { - struct mdt_rec_create *rec; - char *tmp; - - CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_create)); - rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT); - - - rec->cr_opcode = REINT_CREATE; - rec->cr_fsuid = uid; - rec->cr_fsgid = gid; - rec->cr_cap = cap_effective; - rec->cr_fid1 = op_data->op_fid1; - rec->cr_fid2 = op_data->op_fid2; - rec->cr_mode = mode; - rec->cr_rdev = rdev; - rec->cr_time = op_data->op_mod_time; - rec->cr_suppgid1 = op_data->op_suppgids[0]; - rec->cr_suppgid2 = op_data->op_suppgids[1]; - rec->cr_flags = op_data->op_flags & ~MF_SOM_LOCAL_FLAGS; - rec->cr_bias = op_data->op_bias; - - mdc_pack_capa(req, &RMF_CAPA1, op_data->op_capa1); - - tmp = req_capsule_client_get(&req->rq_pill, &RMF_NAME); - LOGL0(op_data->op_name, op_data->op_namelen, tmp); - - if (data) { - tmp = req_capsule_client_get(&req->rq_pill, &RMF_EADATA); - memcpy(tmp, data, datalen); - } + struct mdt_rec_create *rec; + char *tmp; + __u64 flags; + + CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_create)); + rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT); + + + rec->cr_opcode = REINT_CREATE; + rec->cr_fsuid = uid; + rec->cr_fsgid = gid; + rec->cr_cap = cap_effective; + rec->cr_fid1 = op_data->op_fid1; + rec->cr_fid2 = op_data->op_fid2; + rec->cr_mode = mode; + rec->cr_rdev = rdev; + rec->cr_time = op_data->op_mod_time; + rec->cr_suppgid1 = op_data->op_suppgids[0]; + rec->cr_suppgid2 = op_data->op_suppgids[1]; + flags = op_data->op_flags & MF_SOM_LOCAL_FLAGS; + if (op_data->op_bias & MDS_CREATE_VOLATILE) + flags |= MDS_OPEN_VOLATILE; + set_mrc_cr_flags(rec, flags); + rec->cr_bias = op_data->op_bias; + rec->cr_umask = cfs_curproc_umask(); + + mdc_pack_capa(req, &RMF_CAPA1, op_data->op_capa1); + + tmp = req_capsule_client_get(&req->rq_pill, &RMF_NAME); + LOGL0(op_data->op_name, op_data->op_namelen, tmp); + + if (data) { + tmp = req_capsule_client_get(&req->rq_pill, &RMF_EADATA); + memcpy(tmp, data, datalen); + } } -static __u32 mds_pack_open_flags(__u32 flags) +static __u64 mds_pack_open_flags(__u32 flags, __u32 mode) { - __u32 cr_flags = (flags & (FMODE_READ | FMODE_WRITE | - MDS_OPEN_HAS_EA | MDS_OPEN_HAS_OBJS | - MDS_OPEN_OWNEROVERRIDE | MDS_OPEN_LOCK)); + __u64 cr_flags = (flags & (FMODE_READ | FMODE_WRITE | + MDS_OPEN_HAS_EA | MDS_OPEN_HAS_OBJS | + MDS_OPEN_OWNEROVERRIDE | MDS_OPEN_LOCK | + MDS_OPEN_BY_FID)); if (flags & O_CREAT) cr_flags |= MDS_OPEN_CREAT; if (flags & O_EXCL) @@ -181,8 +202,6 @@ static __u32 mds_pack_open_flags(__u32 flags) cr_flags |= MDS_OPEN_SYNC; if (flags & O_DIRECTORY) cr_flags |= MDS_OPEN_DIRECTORY; - if (flags & O_JOIN_FILE) - cr_flags |= MDS_OPEN_JOIN_FILE; #ifdef FMODE_EXEC if (flags & FMODE_EXEC) cr_flags |= MDS_FMODE_EXEC; @@ -190,67 +209,63 @@ static __u32 mds_pack_open_flags(__u32 flags) if (flags & O_LOV_DELAY_CREATE) cr_flags |= MDS_OPEN_DELAY_CREATE; + if ((flags & O_NOACCESS) || (flags & O_NONBLOCK)) + cr_flags |= MDS_OPEN_NORESTORE; + return cr_flags; } /* packing of MDS records */ -void mdc_join_pack(struct ptlrpc_request *req, - struct md_op_data *op_data, - __u64 head_size) -{ - struct mdt_rec_join *rec; - - rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_JOINFILE); - LASSERT(rec != NULL); - rec->jr_fid = op_data->op_fid2; - rec->jr_headsize = head_size; -} - void mdc_open_pack(struct ptlrpc_request *req, struct md_op_data *op_data, - __u32 mode, __u64 rdev, __u32 flags, const void *lmm, - int lmmlen) + __u32 mode, __u64 rdev, __u32 flags, const void *lmm, + int lmmlen) { - struct mdt_rec_create *rec; - char *tmp; - - CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_create)); - rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT); - - /* XXX do something about time, uid, gid */ - rec->cr_opcode = REINT_OPEN; - rec->cr_fsuid = current->fsuid; - rec->cr_fsgid = current->fsgid; - rec->cr_cap = cfs_curproc_cap_pack(); - if (op_data != NULL) { - rec->cr_fid1 = op_data->op_fid1; - rec->cr_fid2 = op_data->op_fid2; - } - rec->cr_mode = mode; - rec->cr_flags = mds_pack_open_flags(flags); - rec->cr_rdev = rdev; - rec->cr_time = op_data->op_mod_time; - rec->cr_suppgid1 = op_data->op_suppgids[0]; - rec->cr_suppgid2 = op_data->op_suppgids[1]; - rec->cr_bias = op_data->op_bias; - - mdc_pack_capa(req, &RMF_CAPA1, op_data->op_capa1); - /* the next buffer is child capa, which is used for replay, - * will be packed from the data in reply message. */ - - if (op_data->op_name) { - tmp = req_capsule_client_get(&req->rq_pill, &RMF_NAME); - LOGL0(op_data->op_name, op_data->op_namelen, tmp); - } - - if (lmm) { - rec->cr_flags |= MDS_OPEN_HAS_EA; + struct mdt_rec_create *rec; + char *tmp; + __u64 cr_flags; + + CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_create)); + rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT); + + /* XXX do something about time, uid, gid */ + rec->cr_opcode = REINT_OPEN; + rec->cr_fsuid = cfs_curproc_fsuid(); + rec->cr_fsgid = cfs_curproc_fsgid(); + rec->cr_cap = cfs_curproc_cap_pack(); + if (op_data != NULL) { + rec->cr_fid1 = op_data->op_fid1; + rec->cr_fid2 = op_data->op_fid2; + } + rec->cr_mode = mode; + cr_flags = mds_pack_open_flags(flags, mode); + rec->cr_rdev = rdev; + rec->cr_time = op_data->op_mod_time; + rec->cr_suppgid1 = op_data->op_suppgids[0]; + rec->cr_suppgid2 = op_data->op_suppgids[1]; + rec->cr_bias = op_data->op_bias; + rec->cr_umask = cfs_curproc_umask(); + + mdc_pack_capa(req, &RMF_CAPA1, op_data->op_capa1); + /* the next buffer is child capa, which is used for replay, + * will be packed from the data in reply message. */ + + if (op_data->op_name) { + tmp = req_capsule_client_get(&req->rq_pill, &RMF_NAME); + LOGL0(op_data->op_name, op_data->op_namelen, tmp); + if (op_data->op_bias & MDS_CREATE_VOLATILE) + cr_flags |= MDS_OPEN_VOLATILE; + } + + if (lmm) { + cr_flags |= MDS_OPEN_HAS_EA; #ifndef __KERNEL__ - /*XXX a hack for liblustre to set EA (LL_IOC_LOV_SETSTRIPE) */ - rec->cr_fid2 = op_data->op_fid2; + /*XXX a hack for liblustre to set EA (LL_IOC_LOV_SETSTRIPE) */ + rec->cr_fid2 = op_data->op_fid2; #endif - tmp = req_capsule_client_get(&req->rq_pill, &RMF_EADATA); - memcpy (tmp, lmm, lmmlen); - } + tmp = req_capsule_client_get(&req->rq_pill, &RMF_EADATA); + memcpy(tmp, lmm, lmmlen); + } + set_mrc_cr_flags(rec, cr_flags); } static inline __u64 attr_pack(unsigned int ia_valid) { @@ -298,8 +313,8 @@ static void mdc_setattr_pack_rec(struct mdt_rec_setattr *rec, struct md_op_data *op_data) { rec->sa_opcode = REINT_SETATTR; - rec->sa_fsuid = current->fsuid; - rec->sa_fsgid = current->fsgid; + rec->sa_fsuid = cfs_curproc_fsuid(); + rec->sa_fsgid = cfs_curproc_fsgid(); rec->sa_cap = cfs_curproc_cap_pack(); rec->sa_suppgid = -1; @@ -315,25 +330,29 @@ static void mdc_setattr_pack_rec(struct mdt_rec_setattr *rec, rec->sa_ctime = LTIME_S(op_data->op_attr.ia_ctime); rec->sa_attr_flags = ((struct ll_iattr *)&op_data->op_attr)->ia_attr_flags; if ((op_data->op_attr.ia_valid & ATTR_GID) && - in_group_p(op_data->op_attr.ia_gid)) + cfs_curproc_is_in_groups(op_data->op_attr.ia_gid)) rec->sa_suppgid = op_data->op_attr.ia_gid; else rec->sa_suppgid = op_data->op_suppgids[0]; + + rec->sa_bias = op_data->op_bias; } -static void mdc_epoch_pack(struct mdt_epoch *epoch, struct md_op_data *op_data) +static void mdc_ioepoch_pack(struct mdt_ioepoch *epoch, + struct md_op_data *op_data) { memcpy(&epoch->handle, &op_data->op_handle, sizeof(epoch->handle)); epoch->ioepoch = op_data->op_ioepoch; - epoch->flags = op_data->op_flags & ~MF_SOM_LOCAL_FLAGS; + epoch->flags = op_data->op_flags & MF_SOM_LOCAL_FLAGS; } void mdc_setattr_pack(struct ptlrpc_request *req, struct md_op_data *op_data, void *ea, int ealen, void *ea2, int ea2len) { struct mdt_rec_setattr *rec; - struct mdt_epoch *epoch; - + struct mdt_ioepoch *epoch; + struct lov_user_md *lum = NULL; + CLASSERT(sizeof(struct mdt_rec_reint) ==sizeof(struct mdt_rec_setattr)); rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT); mdc_setattr_pack_rec(rec, op_data); @@ -342,13 +361,21 @@ void mdc_setattr_pack(struct ptlrpc_request *req, struct md_op_data *op_data, if (op_data->op_flags & (MF_SOM_CHANGE | MF_EPOCH_OPEN)) { epoch = req_capsule_client_get(&req->rq_pill, &RMF_MDT_EPOCH); - mdc_epoch_pack(epoch, op_data); + mdc_ioepoch_pack(epoch, op_data); } if (ealen == 0) return; - memcpy(req_capsule_client_get(&req->rq_pill, &RMF_EADATA), ea, ealen); + lum = req_capsule_client_get(&req->rq_pill, &RMF_EADATA); + if (ea == NULL) { /* Remove LOV EA */ + lum->lmm_magic = LOV_USER_MAGIC_V1; + lum->lmm_stripe_size = 0; + lum->lmm_stripe_count = 0; + lum->lmm_stripe_offset = (typeof(lum->lmm_stripe_offset))(-1); + } else { + memcpy(lum, ea, ealen); + } if (ea2len == 0) return; @@ -359,14 +386,15 @@ void mdc_setattr_pack(struct ptlrpc_request *req, struct md_op_data *op_data, void mdc_unlink_pack(struct ptlrpc_request *req, struct md_op_data *op_data) { - struct mdt_rec_unlink *rec; - char *tmp; - - CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_unlink)); - rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT); - LASSERT (rec != NULL); + struct mdt_rec_unlink *rec; + char *tmp; - rec->ul_opcode = REINT_UNLINK; + CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_unlink)); + rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT); + LASSERT(rec != NULL); + + rec->ul_opcode = op_data->op_cli_flags & CLI_RM_ENTRY ? + REINT_RMENTRY : REINT_UNLINK; rec->ul_fsuid = op_data->op_fsuid; rec->ul_fsgid = op_data->op_fsgid; rec->ul_cap = op_data->op_cap; @@ -447,21 +475,19 @@ void mdc_rename_pack(struct ptlrpc_request *req, struct md_op_data *op_data, } void mdc_getattr_pack(struct ptlrpc_request *req, __u64 valid, int flags, - struct md_op_data *op_data) + struct md_op_data *op_data, int ea_size) { struct mdt_body *b = req_capsule_client_get(&req->rq_pill, &RMF_MDT_BODY); - b->fsuid = current->fsuid; - b->fsgid = current->fsgid; - b->capability = cfs_curproc_cap_pack(); b->valid = valid; if (op_data->op_bias & MDS_CHECK_SPLIT) b->valid |= OBD_MD_FLCKSPLIT; if (op_data->op_bias & MDS_CROSS_REF) b->valid |= OBD_MD_FLCROSSREF; - b->flags = flags | MDS_BFLAG_EXT_FLAGS; - b->suppgid = op_data->op_suppgids[0]; + b->eadatasize = ea_size; + b->flags = flags; + __mdc_pack_body(b, op_data->op_suppgids[0]); b->fid1 = op_data->op_fid1; b->fid2 = op_data->op_fid2; @@ -478,7 +504,7 @@ void mdc_getattr_pack(struct ptlrpc_request *req, __u64 valid, int flags, void mdc_close_pack(struct ptlrpc_request *req, struct md_op_data *op_data) { - struct mdt_epoch *epoch; + struct mdt_ioepoch *epoch; struct mdt_rec_setattr *rec; epoch = req_capsule_client_get(&req->rq_pill, &RMF_MDT_EPOCH); @@ -486,59 +512,67 @@ void mdc_close_pack(struct ptlrpc_request *req, struct md_op_data *op_data) mdc_setattr_pack_rec(rec, op_data); mdc_pack_capa(req, &RMF_CAPA1, op_data->op_capa1); - mdc_epoch_pack(epoch, op_data); + mdc_ioepoch_pack(epoch, op_data); } static int mdc_req_avail(struct client_obd *cli, struct mdc_cache_waiter *mcw) { int rc; ENTRY; - spin_lock(&cli->cl_loi_list_lock); - rc = list_empty(&mcw->mcw_entry); - spin_unlock(&cli->cl_loi_list_lock); + client_obd_list_lock(&cli->cl_loi_list_lock); + rc = cfs_list_empty(&mcw->mcw_entry); + client_obd_list_unlock(&cli->cl_loi_list_lock); RETURN(rc); }; /* We record requests in flight in cli->cl_r_in_flight here. * There is only one write rpc possible in mdc anyway. If this to change * in the future - the code may need to be revisited. */ -void mdc_enter_request(struct client_obd *cli) +int mdc_enter_request(struct client_obd *cli) { + int rc = 0; struct mdc_cache_waiter mcw; - struct l_wait_info lwi = { 0 }; + struct l_wait_info lwi = LWI_INTR(LWI_ON_SIGNAL_NOOP, NULL); - spin_lock(&cli->cl_loi_list_lock); + client_obd_list_lock(&cli->cl_loi_list_lock); if (cli->cl_r_in_flight >= cli->cl_max_rpcs_in_flight) { - list_add_tail(&mcw.mcw_entry, &cli->cl_cache_waiters); + cfs_list_add_tail(&mcw.mcw_entry, &cli->cl_cache_waiters); cfs_waitq_init(&mcw.mcw_waitq); - spin_unlock(&cli->cl_loi_list_lock); - l_wait_event(mcw.mcw_waitq, mdc_req_avail(cli, &mcw), &lwi); + client_obd_list_unlock(&cli->cl_loi_list_lock); + rc = l_wait_event(mcw.mcw_waitq, mdc_req_avail(cli, &mcw), &lwi); + if (rc) { + client_obd_list_lock(&cli->cl_loi_list_lock); + if (cfs_list_empty(&mcw.mcw_entry)) + cli->cl_r_in_flight--; + cfs_list_del_init(&mcw.mcw_entry); + client_obd_list_unlock(&cli->cl_loi_list_lock); + } } else { cli->cl_r_in_flight++; - spin_unlock(&cli->cl_loi_list_lock); + client_obd_list_unlock(&cli->cl_loi_list_lock); } + return rc; } void mdc_exit_request(struct client_obd *cli) { - struct list_head *l, *tmp; + cfs_list_t *l, *tmp; struct mdc_cache_waiter *mcw; - spin_lock(&cli->cl_loi_list_lock); + client_obd_list_lock(&cli->cl_loi_list_lock); cli->cl_r_in_flight--; - list_for_each_safe(l, tmp, &cli->cl_cache_waiters) { - + cfs_list_for_each_safe(l, tmp, &cli->cl_cache_waiters) { if (cli->cl_r_in_flight >= cli->cl_max_rpcs_in_flight) { /* No free request slots anymore */ break; } - mcw = list_entry(l, struct mdc_cache_waiter, mcw_entry); - list_del_init(&mcw->mcw_entry); + mcw = cfs_list_entry(l, struct mdc_cache_waiter, mcw_entry); + cfs_list_del_init(&mcw->mcw_entry); cli->cl_r_in_flight++; cfs_waitq_signal(&mcw->mcw_waitq); } /* Empty waiting list? Decrease reqs in-flight number */ - - spin_unlock(&cli->cl_loi_list_lock); + + client_obd_list_unlock(&cli->cl_loi_list_lock); }