X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Fllite%2Fnamei.c;h=7383c7c716aa08f24a3c165468d9e4a23ab62a88;hp=93d063d88e1a7c872e635f6e8bdce660ffe80318;hb=ea766bd959cc2347b223d597b3f8a31e010bed72;hpb=9b6f9d17a35188f5f4dbfae840164b999a7a78a2;ds=sidebyside diff --git a/lustre/llite/namei.c b/lustre/llite/namei.c index 93d063d..7383c7c 100644 --- a/lustre/llite/namei.c +++ b/lustre/llite/namei.c @@ -1,105 +1,80 @@ -/* -*- 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. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. * - * Copyright (c) 2002, 2003 Cluster File Systems, Inc. + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). * - * This file is part of Lustre, http://www.lustre.org. + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf * - * Lustre is free software; you can redistribute it and/or - * modify it under the terms of version 2 of the GNU General Public - * License as published by the Free Software Foundation. + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. * - * Lustre is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - * GNU General Public License for more details. + * GPL HEADER END + */ +/* + * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved. + * Use is subject to license terms. * - * You should have received a copy of the GNU General Public License - * along with Lustre; if not, write to the Free Software - * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + * Copyright (c) 2011, 2013, Intel Corporation. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. */ #include #include #include -#include #include #include #include +#include #define DEBUG_SUBSYSTEM S_LLITE #include +#include #include #include -#include +#include #include "llite_internal.h" -/* methods */ +static int ll_create_it(struct inode *, struct dentry *, + int, struct lookup_intent *); -/* called from iget{4,5_locked}->find_inode() under inode_lock spinlock */ -#if (LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0)) -static int ll_test_inode(struct inode *inode, unsigned long ino, void *opaque) -#else -static int ll_test_inode(struct inode *inode, void *opaque) -#endif +/* + * Check if we have something mounted at the named dchild. + * In such a case there would always be dentry present. + */ +static int ll_d_mountpoint(struct dentry *dparent, struct dentry *dchild, + struct qstr *name) { - static int last_ino, last_gen, last_count; - struct lustre_md *md = opaque; - - if (!(md->body->valid & (OBD_MD_FLGENER | OBD_MD_FLID))) { - CERROR("MDS body missing inum or generation\n"); - return 0; - } - - if (last_ino == md->body->ino && last_gen == md->body->generation && - last_count < 500) { - last_count++; - } else { - if (last_count > 1) - CDEBUG(D_VFSTRACE, "compared %u/%u %u times\n", - last_ino, last_gen, last_count); - last_count = 0; - last_ino = md->body->ino; - last_gen = md->body->generation; - CDEBUG(D_VFSTRACE, - "comparing inode %p ino %lu/%u to body "LPU64"/%u\n", - inode, inode->i_ino, inode->i_generation, - md->body->ino, md->body->generation); - } - -#if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0)) - if (inode->i_ino != md->body->ino) - return 0; -#endif - if (inode->i_generation != md->body->generation) { - struct ll_sb_info *sbi = ll_i2sbi(inode); - struct ll_inode_info *lli = ll_i2info(inode); - - if (inode->i_state & (I_FREEING | I_CLEAR)) - return 0; - if (inode->i_nlink == 0) - return 0; - - /* add "duplicate" inode into deathrow for destroy */ - spin_lock(&sbi->ll_deathrow_lock); - if (list_empty(&lli->lli_dead_list)) { - __iget(inode); - list_add(&lli->lli_dead_list, &sbi->ll_deathrow); + int mounted = 0; + + if (unlikely(dchild)) { + mounted = d_mountpoint(dchild); + } else if (dparent) { + dchild = d_lookup(dparent, name); + if (dchild) { + mounted = d_mountpoint(dchild); + dput(dchild); } - spin_unlock(&sbi->ll_deathrow_lock); - - return 0; } - - /* Apply the attributes in 'opaque' to this inode */ - if (!(inode->i_state & (I_FREEING | I_CLEAR))) - ll_update_inode(inode, md); - return 1; + return mounted; } -extern struct dentry_operations ll_d_ops; - int ll_unlock(__u32 mode, struct lustre_handle *lockh) { ENTRY; @@ -109,49 +84,119 @@ int ll_unlock(__u32 mode, struct lustre_handle *lockh) RETURN(0); } -/* Get an inode by inode number (already instantiated by the intent lookup). - * Returns inode or NULL - */ -#if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0)) -int ll_set_inode(struct inode *inode, void *opaque) + +/* called from iget5_locked->find_inode() under inode_lock spinlock */ +static int ll_test_inode(struct inode *inode, void *opaque) { - ll_read_inode2(inode, opaque); + struct ll_inode_info *lli = ll_i2info(inode); + struct lustre_md *md = opaque; + + if (unlikely(!(md->body->valid & OBD_MD_FLID))) { + CERROR("MDS body missing FID\n"); + return 0; + } + + if (!lu_fid_eq(&lli->lli_fid, &md->body->fid1)) + return 0; + + return 1; +} + +static int ll_set_inode(struct inode *inode, void *opaque) +{ + struct ll_inode_info *lli = ll_i2info(inode); + struct mdt_body *body = ((struct lustre_md *)opaque)->body; + + if (unlikely(!(body->valid & OBD_MD_FLID))) { + CERROR("MDS body missing FID\n"); + return -EINVAL; + } + + lli->lli_fid = body->fid1; + if (unlikely(!(body->valid & OBD_MD_FLTYPE))) { + CERROR("Can not initialize inode "DFID" without object type: " + "valid = "LPX64"\n", PFID(&lli->lli_fid), body->valid); + return -EINVAL; + } + + inode->i_mode = (inode->i_mode & ~S_IFMT) | (body->mode & S_IFMT); + if (unlikely(inode->i_mode == 0)) { + CERROR("Invalid inode "DFID" type\n", PFID(&lli->lli_fid)); + return -EINVAL; + } + + ll_lli_init(lli); + return 0; } + +/* + * Get an inode by inode number (already instantiated by the intent lookup). + * Returns inode or NULL + */ struct inode *ll_iget(struct super_block *sb, ino_t hash, struct lustre_md *md) { - struct inode *inode; + struct inode *inode; + ENTRY; LASSERT(hash != 0); inode = iget5_locked(sb, hash, ll_test_inode, ll_set_inode, md); if (inode) { - if (inode->i_state & I_NEW) - unlock_new_inode(inode); - CDEBUG(D_VFSTRACE, "inode: %lu/%u(%p)\n", inode->i_ino, - inode->i_generation, inode); + if (inode->i_state & I_NEW) { + int rc = 0; + + ll_read_inode2(inode, md); + if (S_ISREG(inode->i_mode) && + ll_i2info(inode)->lli_clob == NULL) { + CDEBUG(D_INODE, + "%s: apply lsm %p to inode "DFID".\n", + ll_get_fsname(sb, NULL, 0), md->lsm, + PFID(ll_inode2fid(inode))); + rc = cl_file_inode_init(inode, md); + } + if (rc != 0) { + make_bad_inode(inode); + unlock_new_inode(inode); + iput(inode); + inode = ERR_PTR(rc); + } else + unlock_new_inode(inode); + } else if (!(inode->i_state & (I_FREEING | I_CLEAR))) + ll_update_inode(inode, md); + CDEBUG(D_VFSTRACE, "got inode: %p for "DFID"\n", + inode, PFID(&md->body->fid1)); } - - return inode; + RETURN(inode); } -#else -struct inode *ll_iget(struct super_block *sb, ino_t hash, - struct lustre_md *md) + +static void ll_invalidate_negative_children(struct inode *dir) { - struct inode *inode; - LASSERT(hash != 0); - inode = iget4(sb, hash, ll_test_inode, md); - if (inode) - CDEBUG(D_VFSTRACE, "inode: %lu/%u(%p)\n", inode->i_ino, - inode->i_generation, inode); - return inode; + struct dentry *dentry, *tmp_subdir; + struct ll_d_hlist_node *p; + + ll_lock_dcache(dir); + ll_d_hlist_for_each_entry(dentry, p, &dir->i_dentry, d_alias) { + spin_lock(&dentry->d_lock); + if (!list_empty(&dentry->d_subdirs)) { + struct dentry *child; + + list_for_each_entry_safe(child, tmp_subdir, + &dentry->d_subdirs, + d_u.d_child) { + if (child->d_inode == NULL) + d_lustre_invalidate(child, 1); + } + } + spin_unlock(&dentry->d_lock); + } + ll_unlock_dcache(dir); } -#endif -int ll_mdc_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc, - void *data, int flag) +int ll_md_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc, + void *data, int flag) { int rc; struct lustre_handle lockh; @@ -160,24 +205,43 @@ int ll_mdc_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc, switch (flag) { case LDLM_CB_BLOCKING: ldlm_lock2handle(lock, &lockh); - rc = ldlm_cli_cancel(&lockh); + rc = ldlm_cli_cancel(&lockh, LCF_ASYNC); if (rc < 0) { CDEBUG(D_INODE, "ldlm_cli_cancel: %d\n", rc); RETURN(rc); } break; case LDLM_CB_CANCELING: { - struct inode *inode = ll_inode_from_lock(lock); + struct inode *inode = ll_inode_from_resource_lock(lock); + struct ll_inode_info *lli; __u64 bits = lock->l_policy_data.l_inodebits.bits; + struct lu_fid *fid; + ldlm_mode_t mode = lock->l_req_mode; + + /* Inode is set to lock->l_resource->lr_lvb_inode + * for mdc - bug 24555 */ + LASSERT(lock->l_ast_data == NULL); /* Invalidate all dentries associated with this inode */ if (inode == NULL) break; - if (lock->l_resource->lr_name.name[0] != inode->i_ino || - lock->l_resource->lr_name.name[1] != inode->i_generation) { - LDLM_ERROR(lock, "data mismatch with ino %lu/%u (%p)", - inode->i_ino, inode->i_generation, inode); + LASSERT(lock->l_flags & LDLM_FL_CANCELING); + /* For OPEN locks we differentiate between lock modes + * LCK_CR, LCK_CW, LCK_PR - bug 22891 */ + if (bits & (MDS_INODELOCK_LOOKUP | MDS_INODELOCK_UPDATE | + MDS_INODELOCK_LAYOUT | MDS_INODELOCK_PERM)) + ll_have_md_lock(inode, &bits, LCK_MINMODE); + + if (bits & MDS_INODELOCK_OPEN) + ll_have_md_lock(inode, &bits, mode); + + fid = ll_inode2fid(inode); + if (lock->l_resource->lr_name.name[0] != fid_seq(fid) || + lock->l_resource->lr_name.name[1] != fid_oid(fid) || + lock->l_resource->lr_name.name[2] != fid_ver(fid)) { + LDLM_ERROR(lock, "data mismatch with object " + DFID" (%p)", PFID(fid), inode); } if (bits & MDS_INODELOCK_OPEN) { @@ -188,8 +252,6 @@ int ll_mdc_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc, break; case LCK_PR: flags = FMODE_EXEC; - if (!FMODE_EXEC) - CERROR("open PR lock without FMODE_EXEC\n"); break; case LCK_CR: flags = FMODE_READ; @@ -199,63 +261,35 @@ int ll_mdc_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc, "%d, inode %ld\n", lock->l_req_mode, inode->i_ino); } - ll_mdc_real_close(inode, flags); + ll_md_real_close(inode, flags); } + lli = ll_i2info(inode); + if (bits & MDS_INODELOCK_LAYOUT) { + struct cl_object_conf conf = { { 0 } }; + + conf.coc_opc = OBJECT_CONF_INVALIDATE; + conf.coc_inode = inode; + rc = ll_layout_conf(inode, &conf); + if (rc) + CDEBUG(D_INODE, "invaliding layout %d.\n", rc); + } + if (bits & MDS_INODELOCK_UPDATE) - clear_bit(LLI_F_HAVE_MDS_SIZE_LOCK, - &(ll_i2info(inode)->lli_flags)); + lli->lli_flags &= ~LLIF_MDS_SIZE_LOCK; if (S_ISDIR(inode->i_mode) && (bits & MDS_INODELOCK_UPDATE)) { - struct dentry *dentry, *tmp, *dir; - struct list_head *list; - CDEBUG(D_INODE, "invalidating inode %lu\n", inode->i_ino); truncate_inode_pages(inode->i_mapping, 0); + ll_invalidate_negative_children(inode); + } - /* Drop possible cached negative dentries */ - list = &inode->i_dentry; - dir = NULL; - spin_lock(&dcache_lock); - - /* It is possible to have several dentries (with - racer?) */ - while ((list = list->next) != &inode->i_dentry) { - dir = list_entry(list, struct dentry, d_alias); -#ifdef LUSTRE_KERNEL_VERSION - if (!(dir->d_flags & DCACHE_LUSTRE_INVALID)) -#else - if (!d_unhashed(dir)) -#endif - break; - - dir = NULL; - } - - if (dir) { -restart: - list_for_each_entry_safe(dentry, tmp, - &dir->d_subdirs, - d_child) - { - /* XXX Print some debug here? */ - if (!dentry->d_inode) - /* Negative dentry. If we were - dropping dcache lock, go - throught the list again */ - if (ll_drop_dentry(dentry)) - goto restart; - } - } - spin_unlock(&dcache_lock); - } - - if (inode->i_sb->s_root && - inode != inode->i_sb->s_root->d_inode && - (bits & MDS_INODELOCK_LOOKUP)) - ll_unhash_aliases(inode); + if (inode->i_sb->s_root && + inode != inode->i_sb->s_root->d_inode && + (bits & (MDS_INODELOCK_LOOKUP | MDS_INODELOCK_PERM))) + ll_invalidate_aliases(inode); iput(inode); break; } @@ -266,16 +300,12 @@ restart: RETURN(0); } -int ll_mdc_cancel_unused(struct lustre_handle *conn, struct inode *inode, - int flags, void *opaque) +__u32 ll_i2suppgid(struct inode *i) { - struct ldlm_res_id res_id = - { .name = {inode->i_ino, inode->i_generation} }; - struct obd_device *obddev = class_conn2obd(conn); - ENTRY; - - RETURN(ldlm_cli_cancel_unused(obddev->obd_namespace, &res_id, flags, - opaque)); + if (cfs_curproc_is_in_groups(i->i_gid)) + return (__u32)i->i_gid; + else + return (__u32)(-1); } /* Pack the required supplementary groups into the supplied groups array. @@ -284,25 +314,21 @@ int ll_mdc_cancel_unused(struct lustre_handle *conn, struct inode *inode, * array in case it might be useful. Not needed if doing an MDS-side upcall. */ void ll_i2gids(__u32 *suppgids, struct inode *i1, struct inode *i2) { +#if 0 int i; +#endif LASSERT(i1 != NULL); LASSERT(suppgids != NULL); - if (in_group_p(i1->i_gid)) - suppgids[0] = i1->i_gid; - else - suppgids[0] = -1; + suppgids[0] = ll_i2suppgid(i1); - if (i2) { - if (in_group_p(i2->i_gid)) - suppgids[1] = i2->i_gid; + if (i2) + suppgids[1] = ll_i2suppgid(i2); else suppgids[1] = -1; - } else { - suppgids[1] = -1; - } +#if 0 for (i = 0; i < current_ngroups; i++) { if (suppgids[0] == -1) { if (current_groups[i] != suppgids[1]) @@ -316,155 +342,106 @@ void ll_i2gids(__u32 *suppgids, struct inode *i1, struct inode *i2) } break; } +#endif } -int ll_prepare_mdc_op_data(struct mdc_op_data *data, struct inode *i1, - struct inode *i2, const char *name, int namelen, - int mode) -{ - LASSERT(i1); - - if (namelen > ll_i2sbi(i1)->ll_namelen) - return -ENAMETOOLONG; - ll_i2gids(data->suppgids, i1, i2); - ll_inode2fid(&data->fid1, i1); - - if (i2) - ll_inode2fid(&data->fid2, i2); - else - memset(&data->fid2, 0, sizeof(data->fid2)); - - data->name = name; - data->namelen = namelen; - data->create_mode = mode; - data->mod_time = CURRENT_SECONDS; - - return 0; -} - -static void ll_d_add(struct dentry *de, struct inode *inode) +/* + * try to reuse three types of dentry: + * 1. unhashed alias, this one is unhashed by d_invalidate (but it may be valid + * by concurrent .revalidate). + * 2. INVALID alias (common case for no valid ldlm lock held, but this flag may + * be cleared by others calling d_lustre_revalidate). + * 3. DISCONNECTED alias. + */ +static struct dentry *ll_find_alias(struct inode *inode, struct dentry *dentry) { - CDEBUG(D_DENTRY, "adding inode %p to dentry %p\n", inode, de); - /* d_instantiate */ - if (!list_empty(&de->d_alias)) { - spin_unlock(&dcache_lock); - CERROR("dentry %.*s %p alias next %p, prev %p\n", - de->d_name.len, de->d_name.name, de, - de->d_alias.next, de->d_alias.prev); - LBUG(); - } - if (inode) - list_add(&de->d_alias, &inode->i_dentry); - de->d_inode = inode; - - /* d_rehash */ - if (!d_unhashed(de)) { - spin_unlock(&dcache_lock); - CERROR("dentry %.*s %p hash next %p\n", - de->d_name.len, de->d_name.name, de, de->d_hash.next); - LBUG(); - } - __d_rehash(de, 0); + struct dentry *alias, *discon_alias, *invalid_alias; + struct ll_d_hlist_node *p; + + if (ll_d_hlist_empty(&inode->i_dentry)) + return NULL; + + discon_alias = invalid_alias = NULL; + + ll_lock_dcache(inode); + ll_d_hlist_for_each_entry(alias, p, &inode->i_dentry, d_alias) { + LASSERT(alias != dentry); + + spin_lock(&alias->d_lock); + if (alias->d_flags & DCACHE_DISCONNECTED) + /* LASSERT(last_discon == NULL); LU-405, bz 20055 */ + discon_alias = alias; + else if (alias->d_parent == dentry->d_parent && + alias->d_name.hash == dentry->d_name.hash && + alias->d_name.len == dentry->d_name.len && + memcmp(alias->d_name.name, dentry->d_name.name, + dentry->d_name.len) == 0) + invalid_alias = alias; + spin_unlock(&alias->d_lock); + + if (invalid_alias) + break; + } + alias = invalid_alias ?: discon_alias ?: NULL; + if (alias) { + spin_lock(&alias->d_lock); + dget_dlock(alias); + spin_unlock(&alias->d_lock); + } + ll_unlock_dcache(inode); + + return alias; } -/* Search "inode"'s alias list for a dentry that has the same name and parent - * as de. If found, return it. If not found, return de. - * Lustre can't use d_add_unique because don't unhash aliases for directory - * in ll_revalidate_it. After revaliadate inode will be have hashed aliases - * and it triggers BUG_ON in d_instantiate_unique (bug #10954). +/* + * Similar to d_splice_alias(), but lustre treats invalid alias + * similar to DCACHE_DISCONNECTED, and tries to use it anyway. */ -struct dentry *ll_find_alias(struct inode *inode, struct dentry *de) +struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de) { - struct list_head *tmp; - struct dentry *dentry; - struct dentry *last_discon = NULL; - - spin_lock(&dcache_lock); - list_for_each(tmp, &inode->i_dentry) { - dentry = list_entry(tmp, struct dentry, d_alias); - - /* We are called here with 'de' already on the aliases list. */ - if (unlikely(dentry == de)) { - CERROR("whoops\n"); - continue; - } - - if (dentry->d_flags & DCACHE_DISCONNECTED) { - LASSERT(last_discon == NULL); - last_discon = dentry; - continue; - } - - if (dentry->d_parent != de->d_parent) - continue; - - if (dentry->d_name.hash != de->d_name.hash) - continue; - - if (dentry->d_name.len != de->d_name.len) - continue; - - if (memcmp(dentry->d_name.name, de->d_name.name, - de->d_name.len) != 0) - continue; - - dget_locked(dentry); - lock_dentry(dentry); - __d_drop(dentry); -#ifdef LUSTRE_KERNEL_VERSION - dentry->d_flags &= ~DCACHE_LUSTRE_INVALID; -#endif - unlock_dentry(dentry); - __d_rehash(dentry, 0); /* avoid taking dcache_lock inside */ - spin_unlock(&dcache_lock); - iput(inode); - CDEBUG(D_DENTRY, "alias dentry %.*s (%p) parent %p inode %p " - "refc %d\n", de->d_name.len, de->d_name.name, de, - de->d_parent, de->d_inode, atomic_read(&de->d_count)); - return dentry; - } - if (last_discon) { - CDEBUG(D_DENTRY, "Reuse disconnected dentry %p inode %p " - "refc %d\n", last_discon, last_discon->d_inode, - atomic_read(&last_discon->d_count)); - dget_locked(last_discon); - spin_unlock(&dcache_lock); - d_rehash(de); - d_move(last_discon, de); - iput(inode); - return last_discon; - } - - ll_d_add(de, inode); - - spin_unlock(&dcache_lock); - + struct dentry *new; + + if (inode) { + new = ll_find_alias(inode, de); + if (new) { + ll_dops_init(new, 1, 1); + d_move(new, de); + iput(inode); + CDEBUG(D_DENTRY, + "Reuse dentry %p inode %p refc %d flags %#x\n", + new, new->d_inode, d_refcount(new), new->d_flags); + return new; + } + } + ll_dops_init(de, 1, 1); + __d_lustre_invalidate(de); + d_add(de, inode); + CDEBUG(D_DENTRY, "Add dentry %p inode %p refc %d flags %#x\n", + de, de->d_inode, d_refcount(de), de->d_flags); return de; } -static int lookup_it_finish(struct ptlrpc_request *request, int offset, - struct lookup_intent *it, void *data) +int ll_lookup_it_finish(struct ptlrpc_request *request, + struct lookup_intent *it, void *data) { struct it_cb_data *icbd = data; struct dentry **de = icbd->icbd_childp; struct inode *parent = icbd->icbd_parent; - struct ll_sb_info *sbi = ll_i2sbi(parent); struct inode *inode = NULL; - int rc; - - /* NB 1 request reference will be taken away by ll_intent_lock() - * when I return */ - if (!it_disposition(it, DISP_LOOKUP_NEG)) { - ENTRY; - - rc = ll_prep_inode(sbi->ll_osc_exp, &inode, request, offset, - (*de)->d_sb); + __u64 bits = 0; + int rc; + ENTRY; + + /* NB 1 request reference will be taken away by ll_intent_lock() + * when I return */ + CDEBUG(D_DENTRY, "it %p it_disposition %x\n", it, + it->d.lustre.it_disposition); + if (!it_disposition(it, DISP_LOOKUP_NEG)) { + rc = ll_prep_inode(&inode, request, (*de)->d_sb, it); if (rc) RETURN(rc); - CDEBUG(D_DLMTRACE, "setting l_data to inode %p (%lu/%u)\n", - inode, inode->i_ino, inode->i_generation); - mdc_set_lock_data(&it->d.lustre.it_lock_handle, inode); + ll_set_lock_data(ll_i2sbi(parent)->ll_md_exp, inode, it, &bits); /* We used to query real size from OSTs here, but actually this is not needed. For stat() calls size would be updated @@ -474,44 +451,53 @@ static int lookup_it_finish(struct ptlrpc_request *request, int offset, Everybody else who needs correct file size would call ll_glimpse_size or some equivalent themselves anyway. Also see bug 7198. */ - *de = ll_find_alias(inode, *de); - } else { - ENTRY; - /* Check that parent has UPDATE lock. If there is none, we - cannot afford to hash this dentry (done by ll_d_add) as it - might get picked up later when UPDATE lock will appear */ - if (ll_have_md_lock(parent, MDS_INODELOCK_UPDATE)) { - spin_lock(&dcache_lock); - ll_d_add(*de, inode); - spin_unlock(&dcache_lock); - } else { - /* We do not want to hash the dentry if don`t have a - * lock, but if this dentry is later used in d_move, - * we'd hit uninitialised list head d_hash, so we just - * do this to init d_hash field but leave dentry - * unhashed. (bug 10796). */ - d_rehash(*de); - d_drop(*de); - } + } + + /* Only hash *de if it is unhashed (new dentry). + * Atoimc_open may passin hashed dentries for open. + */ + if (d_unhashed(*de)) + *de = ll_splice_alias(inode, *de); + + if (!it_disposition(it, DISP_LOOKUP_NEG)) { + /* we have lookup look - unhide dentry */ + if (bits & MDS_INODELOCK_LOOKUP) + d_lustre_revalidate(*de); + } else if (!it_disposition(it, DISP_OPEN_CREATE)) { + /* If file created on server, don't depend on parent UPDATE + * lock to unhide it. It is left hidden and next lookup can + * find it in ll_splice_alias. + */ + /* Check that parent has UPDATE lock. */ + struct lookup_intent parent_it = { + .it_op = IT_GETATTR, + .d.lustre.it_lock_handle = 0 }; + + if (md_revalidate_lock(ll_i2mdexp(parent), &parent_it, + &ll_i2info(parent)->lli_fid, NULL)) { + d_lustre_revalidate(*de); + ll_intent_release(&parent_it); + } } - ll_set_dd(*de); - (*de)->d_op = &ll_d_ops; - RETURN(0); } static struct dentry *ll_lookup_it(struct inode *parent, struct dentry *dentry, struct lookup_intent *it, int lookup_flags) { + struct lookup_intent lookup_it = { .it_op = IT_LOOKUP }; struct dentry *save = dentry, *retval; - struct mdc_op_data op_data; - struct it_cb_data icbd; struct ptlrpc_request *req = NULL; - struct lookup_intent lookup_it = { .it_op = IT_LOOKUP }; + struct md_op_data *op_data; + struct it_cb_data icbd; + __u32 opc; int rc; ENTRY; + if (dentry->d_name.len > ll_i2sbi(parent)->ll_namelen) + RETURN(ERR_PTR(-ENAMETOOLONG)); + CDEBUG(D_VFSTRACE, "VFS Op:name=%.*s,dir=%lu/%u(%p),intent=%s\n", dentry->d_name.len, dentry->d_name.name, parent->i_ino, parent->i_generation, parent, LL_IT2STR(it)); @@ -525,28 +511,47 @@ static struct dentry *ll_lookup_it(struct inode *parent, struct dentry *dentry, * not valid, revalidate it here. */ if (parent->i_sb->s_root && (parent->i_sb->s_root->d_inode == parent) && (it->it_op & (IT_OPEN | IT_CREAT))) { - rc = ll_inode_revalidate_it(parent->i_sb->s_root, it); + rc = ll_inode_revalidate_it(parent->i_sb->s_root, it, + MDS_INODELOCK_LOOKUP); if (rc) RETURN(ERR_PTR(rc)); } + if (it->it_op == IT_GETATTR) { + rc = ll_statahead_enter(parent, &dentry, 0); + if (rc == 1) { + if (dentry == save) + GOTO(out, retval = NULL); + GOTO(out, retval = dentry); + } + } + icbd.icbd_childp = &dentry; icbd.icbd_parent = parent; - rc = ll_prepare_mdc_op_data(&op_data, parent, NULL, dentry->d_name.name, - dentry->d_name.len, lookup_flags); - if (rc) - RETURN(ERR_PTR(rc)); + if (it->it_op & IT_CREAT || + (it->it_op & IT_OPEN && it->it_create_mode & O_CREAT)) + opc = LUSTRE_OPC_CREATE; + else + opc = LUSTRE_OPC_ANY; - it->it_create_mode &= ~current->fs->umask; + op_data = ll_prep_md_op_data(NULL, parent, NULL, dentry->d_name.name, + dentry->d_name.len, lookup_flags, opc, + NULL); + if (IS_ERR(op_data)) + RETURN((void *)op_data); - rc = mdc_intent_lock(ll_i2mdcexp(parent), &op_data, NULL, 0, it, - lookup_flags, &req, ll_mdc_blocking_ast, 0); + /* enforce umask if acl disabled or MDS doesn't support umask */ + if (!IS_POSIXACL(parent) || !exp_connect_umask(ll_i2mdexp(parent))) + it->it_create_mode &= ~cfs_curproc_umask(); + rc = md_intent_lock(ll_i2mdexp(parent), op_data, NULL, 0, it, + lookup_flags, &req, ll_md_blocking_ast, 0); + ll_finish_md_op_data(op_data); if (rc < 0) GOTO(out, retval = ERR_PTR(rc)); - rc = lookup_it_finish(req, DLM_REPLY_REC_OFF, it, &icbd); + rc = ll_lookup_it_finish(req, it, &icbd); if (rc != 0) { ll_intent_release(it); GOTO(out, retval = ERR_PTR(rc)); @@ -566,53 +571,146 @@ static struct dentry *ll_lookup_it(struct inode *parent, struct dentry *dentry, out: if (req) ptlrpc_req_finished(req); + if (it->it_op == IT_GETATTR && (retval == NULL || retval == dentry)) + ll_statahead_mark(parent, dentry); return retval; } -#if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0)) -#ifdef LUSTRE_KERNEL_VERSION +#ifdef HAVE_IOP_ATOMIC_OPEN static struct dentry *ll_lookup_nd(struct inode *parent, struct dentry *dentry, - struct nameidata *nd) + unsigned int flags) { - struct dentry *de; - ENTRY; - - if (nd && nd->flags & LOOKUP_LAST && !(nd->flags & LOOKUP_LINK_NOTLAST)) - de = ll_lookup_it(parent, dentry, &nd->intent, nd->flags); - else - de = ll_lookup_it(parent, dentry, NULL, 0); + struct lookup_intent *itp, it = { .it_op = IT_GETATTR }; + struct dentry *de; + + CDEBUG(D_VFSTRACE, "VFS Op:name=%.*s,dir=%lu/%u(%p),flags=%u\n", + dentry->d_name.len, dentry->d_name.name, parent->i_ino, + parent->i_generation, parent, flags); + + /* Optimize away (CREATE && !OPEN). Let .create handle the race. */ + if ((flags & LOOKUP_CREATE ) && !(flags & LOOKUP_OPEN)) { + ll_dops_init(dentry, 1, 1); + __d_lustre_invalidate(dentry); + d_add(dentry, NULL); + return NULL; + } + + if (flags & (LOOKUP_PARENT|LOOKUP_OPEN|LOOKUP_CREATE)) + itp = NULL; + else + itp = ⁢ + de = ll_lookup_it(parent, dentry, itp, 0); + + if (itp != NULL) + ll_intent_release(itp); + + return de; +} - RETURN(de); +/* + * For cached negative dentry and new dentry, handle lookup/create/open + * together. + */ +static int ll_atomic_open(struct inode *dir, struct dentry *dentry, + struct file *file, unsigned open_flags, + umode_t mode, int *opened) +{ + struct lookup_intent *it; + struct dentry *de; + long long lookup_flags = LOOKUP_OPEN; + int rc = 0; + ENTRY; + + CDEBUG(D_VFSTRACE, "VFS Op:name=%.*s,dir=%lu/%u(%p),file %p," + "open_flags %x,mode %x opened %d\n", + dentry->d_name.len, dentry->d_name.name, dir->i_ino, + dir->i_generation, dir, file, open_flags, mode, *opened); + + OBD_ALLOC(it, sizeof(*it)); + if (!it) + RETURN(-ENOMEM); + + it->it_op = IT_OPEN; + if (mode) { + it->it_op |= IT_CREAT; + lookup_flags |= LOOKUP_CREATE; + } + it->it_create_mode = (mode & S_IALLUGO) | S_IFREG; + it->it_flags = (open_flags & ~O_ACCMODE) | OPEN_FMODE(open_flags); + + /* Dentry added to dcache tree in ll_lookup_it */ + de = ll_lookup_it(dir, dentry, it, lookup_flags); + if (IS_ERR(de)) + rc = PTR_ERR(de); + else if (de != NULL) + dentry = de; + + if (!rc) { + if (it_disposition(it, DISP_OPEN_CREATE)) { + /* Dentry instantiated in ll_create_it. */ + rc = ll_create_it(dir, dentry, mode, it); + if (rc) { + /* We dget in ll_splice_alias. */ + if (de != NULL) + dput(de); + goto out_release; + } + + *opened |= FILE_CREATED; + } + if (dentry->d_inode && it_disposition(it, DISP_OPEN_OPEN)) { + /* Open dentry. */ + if (S_ISFIFO(dentry->d_inode->i_mode)) { + /* We cannot call open here as it would + * deadlock. + */ + if (it_disposition(it, DISP_ENQ_OPEN_REF)) + ptlrpc_req_finished( + (struct ptlrpc_request *) + it->d.lustre.it_data); + rc = finish_no_open(file, de); + } else { + file->private_data = it; + rc = finish_open(file, dentry, NULL, opened); + /* We dget in ll_splice_alias. finish_open takes + * care of dget for fd open. + */ + if (de != NULL) + dput(de); + } + } else { + rc = finish_no_open(file, de); + } + } + +out_release: + ll_intent_release(it); + OBD_FREE(it, sizeof(*it)); + + RETURN(rc); } -#else + +#else /* !HAVE_IOP_ATOMIC_OPEN */ struct lookup_intent *ll_convert_intent(struct open_intent *oit, int lookup_flags) { - struct lookup_intent *it; - - OBD_ALLOC(it, sizeof(*it)); - if (!it) - return ERR_PTR(-ENOMEM); - - if (lookup_flags & LOOKUP_OPEN) { - it->it_op = IT_OPEN; - if (lookup_flags & LOOKUP_CREATE) - it->it_op |= IT_CREAT; - it->it_create_mode = oit->create_mode; - it->it_flags = oit->flags; - } else { - it->it_op = IT_GETATTR; - } - -#ifndef HAVE_FILE_IN_STRUCT_INTENT - /* Since there is no way to pass our intent to ll_file_open, - * just check the file is there. Actual open will be done - * in ll_file_open */ - if (it->it_op & IT_OPEN) - it->it_op = IT_LOOKUP; -#endif - - return it; + struct lookup_intent *it; + + OBD_ALLOC(it, sizeof(*it)); + if (!it) + return ERR_PTR(-ENOMEM); + + if (lookup_flags & LOOKUP_OPEN) { + it->it_op = IT_OPEN; + if (lookup_flags & LOOKUP_CREATE) + it->it_op |= IT_CREAT; + it->it_create_mode = (oit->create_mode & S_IALLUGO) | S_IFREG; + it->it_flags = ll_namei_to_lookup_intent_flag(oit->flags); + } else { + it->it_op = IT_GETATTR; + } + + return it; } static struct dentry *ll_lookup_nd(struct inode *parent, struct dentry *dentry, @@ -624,18 +722,18 @@ static struct dentry *ll_lookup_nd(struct inode *parent, struct dentry *dentry, if (nd && !(nd->flags & (LOOKUP_CONTINUE|LOOKUP_PARENT))) { struct lookup_intent *it; -#if defined(HAVE_FILE_IN_STRUCT_INTENT) && (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,17)) - /* Did we came here from failed revalidate just to propagate - * its error? */ - if (nd->flags & LOOKUP_OPEN) - if (IS_ERR(nd->intent.open.file)) - RETURN((struct dentry *)nd->intent.open.file); -#endif - if (ll_d2d(dentry) && ll_d2d(dentry)->lld_it) { it = ll_d2d(dentry)->lld_it; ll_d2d(dentry)->lld_it = NULL; } else { + if ((nd->flags & LOOKUP_CREATE ) && !(nd->flags & LOOKUP_OPEN)) { + /* We are sure this is new dentry, so we need to create + our private data and set the dentry ops */ + ll_dops_init(dentry, 1, 1); + __d_lustre_invalidate(dentry); + d_add(dentry, NULL); + RETURN(NULL); + } it = ll_convert_intent(&nd->intent.open, nd->flags); if (IS_ERR(it)) RETURN((struct dentry *)it); @@ -647,7 +745,6 @@ static struct dentry *ll_lookup_nd(struct inode *parent, struct dentry *dentry, if ((nd->flags & LOOKUP_OPEN) && !IS_ERR(dentry)) { /* Open */ if (dentry->d_inode && it_disposition(it, DISP_OPEN_OPEN)) { /* nocreate */ -#ifdef HAVE_FILE_IN_STRUCT_INTENT if (S_ISFIFO(dentry->d_inode->i_mode)) { // We cannot call open here as it would // deadlock. @@ -655,27 +752,18 @@ static struct dentry *ll_lookup_nd(struct inode *parent, struct dentry *dentry, (struct ptlrpc_request *) it->d.lustre.it_data); } else { - struct file *filp; - nd->intent.open.file->private_data = it; - filp =lookup_instantiate_filp(nd,dentry, - NULL); -#if (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,17)) -/* 2.6.1[456] have a bug in open_namei() that forgets to check - * nd->intent.open.file for error, so we need to return it as lookup's result - * instead */ - if (IS_ERR(filp)) { - if (de) - dput(de); - de = (struct dentry *) filp; - } -#endif - + struct file *filp; + + nd->intent.open.file->private_data = it; + filp = lookup_instantiate_filp(nd, + dentry, + NULL); + if (IS_ERR(filp)) { + if (de) + dput(de); + de = (struct dentry *)filp; + } } -#else /* HAVE_FILE_IN_STRUCT_INTENT */ - /* Release open handle as we have no way to - * pass it to ll_file_open */ - ll_release_openhandle(dentry, it); -#endif /* HAVE_FILE_IN_STRUCT_INTENT */ } else if (it_disposition(it, DISP_OPEN_CREATE)) { // XXX This can only reliably work on assumption // that there are NO hashed negative dentries. @@ -697,8 +785,7 @@ static struct dentry *ll_lookup_nd(struct inode *parent, struct dentry *dentry, RETURN(de); } -#endif -#endif +#endif /* HAVE_IOP_ATOMIC_OPEN */ /* We depend on "mode" being set with the proper file type/umask by now */ static struct inode *ll_create_node(struct inode *dir, const char *name, @@ -717,19 +804,18 @@ static struct inode *ll_create_node(struct inode *dir, const char *name, LASSERT(it_disposition(it, DISP_ENQ_CREATE_REF)); request = it->d.lustre.it_data; it_clear_disposition(it, DISP_ENQ_CREATE_REF); - rc = ll_prep_inode(sbi->ll_osc_exp, &inode, request, DLM_REPLY_REC_OFF, - dir->i_sb); + rc = ll_prep_inode(&inode, request, dir->i_sb, it); if (rc) GOTO(out, inode = ERR_PTR(rc)); - LASSERT(list_empty(&inode->i_dentry)); + LASSERT(ll_d_hlist_empty(&inode->i_dentry)); /* We asked for a lock on the directory, but were granted a * lock on the inode. Since we finally have an inode pointer, * stuff it in the lock. */ CDEBUG(D_DLMTRACE, "setting l_ast_data to inode %p (%lu/%u)\n", inode, inode->i_ino, inode->i_generation); - mdc_set_lock_data(&it->d.lustre.it_lock_handle, inode); + ll_set_lock_data(sbi->ll_md_exp, inode, it, NULL); EXIT; out: ptlrpc_req_finished(request); @@ -751,109 +837,93 @@ static struct inode *ll_create_node(struct inode *dir, const char *name, * with d_instantiate(). */ static int ll_create_it(struct inode *dir, struct dentry *dentry, int mode, - struct lookup_intent *it) + struct lookup_intent *it) { - struct inode *inode; - struct ptlrpc_request *request = it->d.lustre.it_data; - int rc = 0; - ENTRY; + struct inode *inode; + int rc = 0; + ENTRY; - CDEBUG(D_VFSTRACE, "VFS Op:name=%.*s,dir=%lu/%u(%p),intent=%s\n", - dentry->d_name.len, dentry->d_name.name, dir->i_ino, - dir->i_generation, dir, LL_IT2STR(it)); + CDEBUG(D_VFSTRACE, "VFS Op:name=%.*s,dir=%lu/%u(%p),intent=%s\n", + dentry->d_name.len, dentry->d_name.name, dir->i_ino, + dir->i_generation, dir, LL_IT2STR(it)); - rc = it_open_error(DISP_OPEN_CREATE, it); - if (rc) - RETURN(rc); - - mdc_store_inode_generation(request, DLM_INTENT_REC_OFF, - DLM_REPLY_REC_OFF); - inode = ll_create_node(dir, dentry->d_name.name, dentry->d_name.len, - NULL, 0, mode, 0, it); - if (IS_ERR(inode)) { - RETURN(PTR_ERR(inode)); - } + rc = it_open_error(DISP_OPEN_CREATE, it); + if (rc) + RETURN(rc); - d_instantiate(dentry, inode); - /* Negative dentry may be unhashed if parent does not have UPDATE lock, - * but some callers, e.g. do_coredump, expect dentry to be hashed after - * successful create. Hash it here. */ - spin_lock(&dcache_lock); - if (d_unhashed(dentry)) - __d_rehash(dentry, 0); - spin_unlock(&dcache_lock); - RETURN(0); + inode = ll_create_node(dir, dentry->d_name.name, dentry->d_name.len, + NULL, 0, mode, 0, it); + if (IS_ERR(inode)) + RETURN(PTR_ERR(inode)); + + if (filename_is_volatile(dentry->d_name.name, dentry->d_name.len, NULL)) + ll_i2info(inode)->lli_volatile = true; + + d_instantiate(dentry, inode); + RETURN(0); } -static void ll_update_times(struct ptlrpc_request *request, int offset, +static void ll_update_times(struct ptlrpc_request *request, struct inode *inode) { - struct mds_body *body = lustre_msg_buf(request->rq_repmsg, offset, - sizeof(*body)); - LASSERT(body); + struct mdt_body *body = req_capsule_server_get(&request->rq_pill, + &RMF_MDT_BODY); - /* mtime is always updated with ctime, but can be set in past. - As write and utime(2) may happen within 1 second, and utime's - mtime has a priority over write's one, so take mtime from mds - for the same ctimes. */ + LASSERT(body); + if (body->valid & OBD_MD_FLMTIME && + body->mtime > LTIME_S(inode->i_mtime)) { + CDEBUG(D_INODE, "setting ino %lu mtime from %lu to "LPU64"\n", + inode->i_ino, LTIME_S(inode->i_mtime), body->mtime); + LTIME_S(inode->i_mtime) = body->mtime; + } if (body->valid & OBD_MD_FLCTIME && - body->ctime >= LTIME_S(inode->i_ctime)) { + body->ctime > LTIME_S(inode->i_ctime)) LTIME_S(inode->i_ctime) = body->ctime; - - if (body->valid & OBD_MD_FLMTIME) { - CDEBUG(D_INODE, "setting ino %lu mtime from %lu " - "to "LPU64"\n", inode->i_ino, - LTIME_S(inode->i_mtime), body->mtime); - LTIME_S(inode->i_mtime) = body->mtime; - } - } } static int ll_new_node(struct inode *dir, struct qstr *name, - const char *tgt, int mode, - int rdev, struct dentry *dchild) + const char *tgt, int mode, int rdev, + struct dentry *dchild, __u32 opc) { struct ptlrpc_request *request = NULL; + struct md_op_data *op_data; struct inode *inode = NULL; struct ll_sb_info *sbi = ll_i2sbi(dir); - struct mdc_op_data op_data; int tgt_len = 0; int err; ENTRY; if (unlikely(tgt != NULL)) - tgt_len = strlen(tgt)+1; + tgt_len = strlen(tgt) + 1; - err = ll_prepare_mdc_op_data(&op_data, dir, NULL, name->name, - name->len, 0); - if (err) - GOTO(err_exit, err); + op_data = ll_prep_md_op_data(NULL, dir, NULL, name->name, + name->len, 0, opc, NULL); + if (IS_ERR(op_data)) + GOTO(err_exit, err = PTR_ERR(op_data)); - err = mdc_create(sbi->ll_mdc_exp, &op_data, tgt, tgt_len, - mode, current->fsuid, current->fsgid, - current->cap_effective, rdev, &request); + err = md_create(sbi->ll_md_exp, op_data, tgt, tgt_len, mode, + cfs_curproc_fsuid(), cfs_curproc_fsgid(), + cfs_curproc_cap_pack(), rdev, &request); + ll_finish_md_op_data(op_data); if (err) GOTO(err_exit, err); - ll_update_times(request, REPLY_REC_OFF, dir); + ll_update_times(request, dir); if (dchild) { - err = ll_prep_inode(sbi->ll_osc_exp, &inode, request, - REPLY_REC_OFF, dchild->d_sb); + err = ll_prep_inode(&inode, request, dchild->d_sb, NULL); if (err) GOTO(err_exit, err); - d_drop(dchild); d_instantiate(dchild, inode); - EXIT; } + EXIT; err_exit: ptlrpc_req_finished(request); return err; } - static int ll_mknod_generic(struct inode *dir, struct qstr *name, int mode, unsigned rdev, struct dentry *dchild) { @@ -864,7 +934,8 @@ static int ll_mknod_generic(struct inode *dir, struct qstr *name, int mode, name->len, name->name, dir->i_ino, dir->i_generation, dir, mode, rdev); - mode &= ~current->fs->umask; + if (!IS_POSIXACL(dir) || !exp_connect_umask(ll_i2mdexp(dir))) + mode &= ~cfs_curproc_umask(); switch (mode & S_IFMT) { case 0: @@ -874,7 +945,8 @@ static int ll_mknod_generic(struct inode *dir, struct qstr *name, int mode, case S_IFBLK: case S_IFIFO: case S_IFSOCK: - err = ll_new_node(dir, name, NULL, mode, rdev, dchild); + err = ll_new_node(dir, name, NULL, mode, rdev, dchild, + LUSTRE_OPC_MKNOD); break; case S_IFDIR: err = -EPERM; @@ -882,12 +954,39 @@ static int ll_mknod_generic(struct inode *dir, struct qstr *name, int mode, default: err = -EINVAL; } + + if (!err) + ll_stats_ops_tally(ll_i2sbi(dir), LPROC_LL_MKNOD, 1); + RETURN(err); } -#if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0)) -#ifndef LUSTRE_KERNEL_VERSION -static int ll_create_nd(struct inode *dir, struct dentry *dentry, int mode, struct nameidata *nd) +#ifdef HAVE_IOP_ATOMIC_OPEN +/* + * Plain create. Intent create is handled in atomic_open. + */ +static int ll_create_nd(struct inode *dir, struct dentry *dentry, + umode_t mode, bool want_excl) +{ + int rc; + + CDEBUG(D_VFSTRACE, "VFS Op:name=%.*s,dir=%lu/%u(%p)," + "flags=%u, excl=%d\n", + dentry->d_name.len, dentry->d_name.name, dir->i_ino, + dir->i_generation, dir, mode, want_excl); + + rc = ll_mknod_generic(dir, &dentry->d_name, mode, 0, dentry); + + ll_stats_ops_tally(ll_i2sbi(dir), LPROC_LL_CREATE, 1); + + CDEBUG(D_VFSTRACE, "VFS Op:name=%.*s, unhashed %d\n", + dentry->d_name.len, dentry->d_name.name, d_unhashed(dentry)); + + return rc; +} +#else /* !HAVE_IOP_ATOMIC_OPEN */ +static int ll_create_nd(struct inode *dir, struct dentry *dentry, + ll_umode_t mode, struct nameidata *nd) { struct lookup_intent *it = ll_d2d(dentry)->lld_it; int rc; @@ -904,33 +1003,25 @@ static int ll_create_nd(struct inode *dir, struct dentry *dentry, int mode, stru } rc = ll_create_it(dir, dentry, mode, it); -#ifdef HAVE_FILE_IN_STRUCT_INTENT if (nd && (nd->flags & LOOKUP_OPEN) && dentry->d_inode) { /* Open */ - nd->intent.open.file->private_data = it; - lookup_instantiate_filp(nd, dentry, NULL); + struct file *filp; + + nd->intent.open.file->private_data = it; + filp = lookup_instantiate_filp(nd, dentry, NULL); + if (IS_ERR(filp)) + rc = PTR_ERR(filp); } -#else - ll_release_openhandle(dentry,it); -#endif out: ll_intent_release(it); OBD_FREE(it, sizeof(*it)); - return rc; -} -#else -static int ll_create_nd(struct inode *dir, struct dentry *dentry, int mode, struct nameidata *nd) -{ - - if (!nd || !nd->intent.d.lustre.it_disposition) - /* No saved request? Just mknod the file */ - return ll_mknod_generic(dir, &dentry->d_name, mode, 0, dentry); + if (!rc) + ll_stats_ops_tally(ll_i2sbi(dir), LPROC_LL_CREATE, 1); - return ll_create_it(dir, dentry, mode, &nd->intent); + return rc; } -#endif -#endif +#endif /* HAVE_IOP_ATOMIC_OPEN */ static int ll_symlink_generic(struct inode *dir, struct qstr *name, const char *tgt, struct dentry *dchild) @@ -943,17 +1034,21 @@ static int ll_symlink_generic(struct inode *dir, struct qstr *name, dir, 3000, tgt); err = ll_new_node(dir, name, (char *)tgt, S_IFLNK | S_IRWXUGO, - 0, dchild); + 0, dchild, LUSTRE_OPC_SYMLINK); + + if (!err) + ll_stats_ops_tally(ll_i2sbi(dir), LPROC_LL_SYMLINK, 1); + RETURN(err); } static int ll_link_generic(struct inode *src, struct inode *dir, struct qstr *name, struct dentry *dchild) { + struct ll_sb_info *sbi = ll_i2sbi(dir); struct ptlrpc_request *request = NULL; - struct mdc_op_data op_data; + struct md_op_data *op_data; int err; - struct ll_sb_info *sbi = ll_i2sbi(dir); ENTRY; CDEBUG(D_VFSTRACE, @@ -961,92 +1056,135 @@ static int ll_link_generic(struct inode *src, struct inode *dir, src->i_ino, src->i_generation, src, dir->i_ino, dir->i_generation, dir, name->len, name->name); - err = ll_prepare_mdc_op_data(&op_data, src, dir, name->name, - name->len, 0); + op_data = ll_prep_md_op_data(NULL, src, dir, name->name, name->len, + 0, LUSTRE_OPC_ANY, NULL); + if (IS_ERR(op_data)) + RETURN(PTR_ERR(op_data)); + + err = md_link(sbi->ll_md_exp, op_data, &request); + ll_finish_md_op_data(op_data); if (err) GOTO(out, err); - err = mdc_link(sbi->ll_mdc_exp, &op_data, &request); - if (err) - GOTO(out, err); - - if (dchild) { - d_drop(dchild); - } - ll_update_times(request, REPLY_REC_OFF, dir); + ll_update_times(request, dir); + ll_stats_ops_tally(sbi, LPROC_LL_LINK, 1); EXIT; out: ptlrpc_req_finished(request); RETURN(err); } -static int ll_mkdir_generic(struct inode *dir, struct qstr *name, int mode, - struct dentry *dchild) +static int ll_mkdir_generic(struct inode *dir, struct qstr *name, + int mode, struct dentry *dchild) { int err; - ENTRY; + CDEBUG(D_VFSTRACE, "VFS Op:name=%.*s,dir=%lu/%u(%p)\n", name->len, name->name, dir->i_ino, dir->i_generation, dir); - mode = (mode & (S_IRWXUGO|S_ISVTX) & ~current->fs->umask) | S_IFDIR; - err = ll_new_node(dir, name, NULL, mode, 0, dchild); + if (!IS_POSIXACL(dir) || !exp_connect_umask(ll_i2mdexp(dir))) + mode &= ~cfs_curproc_umask(); + mode = (mode & (S_IRWXUGO|S_ISVTX)) | S_IFDIR; + err = ll_new_node(dir, name, NULL, mode, 0, dchild, LUSTRE_OPC_MKDIR); + + if (!err) + ll_stats_ops_tally(ll_i2sbi(dir), LPROC_LL_MKDIR, 1); RETURN(err); } +/* Try to find the child dentry by its name. + If found, put the result fid into @fid. */ +static void ll_get_child_fid(struct inode * dir, struct qstr *name, + struct lu_fid *fid) +{ + struct dentry *parent, *child; + + parent = ll_d_hlist_entry(dir->i_dentry, struct dentry, d_alias); + child = d_lookup(parent, name); + if (child) { + if (child->d_inode) + *fid = *ll_inode2fid(child->d_inode); + dput(child); + } +} + static int ll_rmdir_generic(struct inode *dir, struct dentry *dparent, - struct qstr *name) + struct dentry *dchild, struct qstr *name) { struct ptlrpc_request *request = NULL; - struct mdc_op_data op_data; - struct dentry *dentry; + struct md_op_data *op_data; int rc; ENTRY; + CDEBUG(D_VFSTRACE, "VFS Op:name=%.*s,dir=%lu/%u(%p)\n", name->len, name->name, dir->i_ino, dir->i_generation, dir); - /* Check if we have something mounted at the dir we are going to delete - * In such a case there would always be dentry present. */ - if (dparent) { - dentry = d_lookup(dparent, name); - if (dentry) { - int mounted = d_mountpoint(dentry); - dput(dentry); - if (mounted) - GOTO(out, rc = -EBUSY); - } + if (unlikely(ll_d_mountpoint(dparent, dchild, name))) + RETURN(-EBUSY); + + op_data = ll_prep_md_op_data(NULL, dir, NULL, name->name, name->len, + S_IFDIR, LUSTRE_OPC_ANY, NULL); + if (IS_ERR(op_data)) + RETURN(PTR_ERR(op_data)); + + ll_get_child_fid(dir, name, &op_data->op_fid3); + op_data->op_fid2 = op_data->op_fid3; + rc = md_unlink(ll_i2sbi(dir)->ll_md_exp, op_data, &request); + ll_finish_md_op_data(op_data); + if (rc == 0) { + ll_update_times(request, dir); + ll_stats_ops_tally(ll_i2sbi(dir), LPROC_LL_RMDIR, 1); } - rc = ll_prepare_mdc_op_data(&op_data, dir, NULL, name->name, - name->len, S_IFDIR); - if (rc) - GOTO(out, rc); - rc = mdc_unlink(ll_i2sbi(dir)->ll_mdc_exp, &op_data, &request); - if (rc) - GOTO(out, rc); - ll_update_times(request, REPLY_REC_OFF, dir); - - EXIT; -out: ptlrpc_req_finished(request); - return(rc); + RETURN(rc); +} + +/** + * Remove dir entry + **/ +int ll_rmdir_entry(struct inode *dir, char *name, int namelen) +{ + struct ptlrpc_request *request = NULL; + struct md_op_data *op_data; + int rc; + ENTRY; + + CDEBUG(D_VFSTRACE, "VFS Op:name=%.*s,dir=%lu/%u(%p)\n", + namelen, name, dir->i_ino, dir->i_generation, dir); + + op_data = ll_prep_md_op_data(NULL, dir, NULL, name, strlen(name), + S_IFDIR, LUSTRE_OPC_ANY, NULL); + if (IS_ERR(op_data)) + RETURN(PTR_ERR(op_data)); + op_data->op_cli_flags |= CLI_RM_ENTRY; + rc = md_unlink(ll_i2sbi(dir)->ll_md_exp, op_data, &request); + ll_finish_md_op_data(op_data); + if (rc == 0) { + ll_update_times(request, dir); + ll_stats_ops_tally(ll_i2sbi(dir), LPROC_LL_RMDIR, 1); + } + + ptlrpc_req_finished(request); + RETURN(rc); } int ll_objects_destroy(struct ptlrpc_request *request, struct inode *dir) { - struct mds_body *body; + struct mdt_body *body; struct lov_mds_md *eadata; struct lov_stripe_md *lsm = NULL; struct obd_trans_info oti = { 0 }; struct obdo *oa; + struct obd_capa *oc = NULL; int rc; ENTRY; /* req is swabbed so this is safe */ - body = lustre_msg_buf(request->rq_repmsg, REPLY_REC_OFF, sizeof(*body)); - + body = req_capsule_server_get(&request->rq_pill, &RMF_MDT_BODY); if (!(body->valid & OBD_MD_FLEASIZE)) RETURN(0); @@ -1059,242 +1197,228 @@ int ll_objects_destroy(struct ptlrpc_request *request, struct inode *dir) * to this file. Use this EA to unlink the objects on the OST. * It's opaque so we don't swab here; we leave it to obd_unpackmd() to * check it is complete and sensible. */ - eadata = lustre_swab_repbuf(request, REPLY_REC_OFF + 1, - body->eadatasize, NULL); + eadata = req_capsule_server_sized_get(&request->rq_pill, &RMF_MDT_MD, + body->eadatasize); LASSERT(eadata != NULL); - if (eadata == NULL) { - CERROR("Can't unpack MDS EA data\n"); - GOTO(out, rc = -EPROTO); - } - rc = obd_unpackmd(ll_i2obdexp(dir), &lsm, eadata, body->eadatasize); + rc = obd_unpackmd(ll_i2dtexp(dir), &lsm, eadata, body->eadatasize); if (rc < 0) { CERROR("obd_unpackmd: %d\n", rc); GOTO(out, rc); } LASSERT(rc >= sizeof(*lsm)); - rc = obd_checkmd(ll_i2obdexp(dir), ll_i2mdcexp(dir), lsm); - if (rc) - GOTO(out_free_memmd, rc); - - oa = obdo_alloc(); + OBDO_ALLOC(oa); if (oa == NULL) GOTO(out_free_memmd, rc = -ENOMEM); - oa->o_id = lsm->lsm_object_id; + oa->o_oi = lsm->lsm_oi; oa->o_mode = body->mode & S_IFMT; - oa->o_valid = OBD_MD_FLID | OBD_MD_FLTYPE; + oa->o_valid = OBD_MD_FLID | OBD_MD_FLTYPE | OBD_MD_FLGROUP; if (body->valid & OBD_MD_FLCOOKIE) { oa->o_valid |= OBD_MD_FLCOOKIE; oti.oti_logcookies = - lustre_msg_buf(request->rq_repmsg, REPLY_REC_OFF + 2, - sizeof(struct llog_cookie) * - lsm->lsm_stripe_count); + req_capsule_server_sized_get(&request->rq_pill, + &RMF_LOGCOOKIES, + sizeof(struct llog_cookie) * + lsm->lsm_stripe_count); if (oti.oti_logcookies == NULL) { oa->o_valid &= ~OBD_MD_FLCOOKIE; body->valid &= ~OBD_MD_FLCOOKIE; } } - rc = obd_destroy(ll_i2obdexp(dir), oa, lsm, &oti, ll_i2mdcexp(dir)); - obdo_free(oa); - if (rc) - CERROR("obd destroy objid "LPX64" error %d\n", - lsm->lsm_object_id, rc); - out_free_memmd: - obd_free_memmd(ll_i2obdexp(dir), &lsm); - out: - return rc; + if (body->valid & OBD_MD_FLOSSCAPA) { + rc = md_unpack_capa(ll_i2mdexp(dir), request, &RMF_CAPA2, &oc); + if (rc) + GOTO(out_free_memmd, rc); + } + + rc = obd_destroy(NULL, ll_i2dtexp(dir), oa, lsm, &oti, + ll_i2mdexp(dir), oc); + capa_put(oc); + if (rc) + CERROR("obd destroy objid "DOSTID" error %d\n", + POSTID(&lsm->lsm_oi), rc); +out_free_memmd: + obd_free_memmd(ll_i2dtexp(dir), &lsm); + OBDO_FREE(oa); +out: + return rc; } -static int ll_unlink_generic(struct inode * dir, struct qstr *name) +/* ll_unlink_generic() doesn't update the inode with the new link count. + * Instead, ll_ddelete() and ll_d_iput() will update it based upon if there + * is any lock existing. They will recycle dentries and inodes based upon locks + * too. b=20433 */ +static int ll_unlink_generic(struct inode *dir, struct dentry *dparent, + struct dentry *dchild, struct qstr *name) { struct ptlrpc_request *request = NULL; - struct mdc_op_data op_data; + struct md_op_data *op_data; int rc; ENTRY; - CDEBUG(D_VFSTRACE, "VFS Op:name=%.*s,dir=%lu/%u(%p)\n", name->len, name->name, dir->i_ino, dir->i_generation, dir); - rc = ll_prepare_mdc_op_data(&op_data, dir, NULL, name->name, - name->len, 0); - if (rc) - GOTO(out, rc); - rc = mdc_unlink(ll_i2sbi(dir)->ll_mdc_exp, &op_data, &request); - if (rc) - GOTO(out, rc); + /* + * XXX: unlink bind mountpoint maybe call to here, + * just check it as vfs_unlink does. + */ + if (unlikely(ll_d_mountpoint(dparent, dchild, name))) + RETURN(-EBUSY); + + op_data = ll_prep_md_op_data(NULL, dir, NULL, name->name, + name->len, 0, LUSTRE_OPC_ANY, NULL); + if (IS_ERR(op_data)) + RETURN(PTR_ERR(op_data)); - ll_update_times(request, REPLY_REC_OFF, dir); + ll_get_child_fid(dir, name, &op_data->op_fid3); + op_data->op_fid2 = op_data->op_fid3; + rc = md_unlink(ll_i2sbi(dir)->ll_md_exp, op_data, &request); + ll_finish_md_op_data(op_data); + if (rc) + GOTO(out, rc); + + ll_update_times(request, dir); + ll_stats_ops_tally(ll_i2sbi(dir), LPROC_LL_UNLINK, 1); rc = ll_objects_destroy(request, dir); - if (rc) - GOTO(out, rc); - EXIT; out: ptlrpc_req_finished(request); - return(rc); + RETURN(rc); } -static int ll_rename_generic(struct inode *src, struct qstr *src_name, - struct inode *tgt, struct qstr *tgt_name) +static int ll_rename_generic(struct inode *src, struct dentry *src_dparent, + struct dentry *src_dchild, struct qstr *src_name, + struct inode *tgt, struct dentry *tgt_dparent, + struct dentry *tgt_dchild, struct qstr *tgt_name) { struct ptlrpc_request *request = NULL; struct ll_sb_info *sbi = ll_i2sbi(src); - struct mdc_op_data op_data; + struct md_op_data *op_data; int err; - ENTRY; CDEBUG(D_VFSTRACE,"VFS Op:oldname=%.*s,src_dir=%lu/%u(%p),newname=%.*s," "tgt_dir=%lu/%u(%p)\n", src_name->len, src_name->name, src->i_ino, src->i_generation, src, tgt_name->len, tgt_name->name, tgt->i_ino, tgt->i_generation, tgt); - err = ll_prepare_mdc_op_data(&op_data, src, tgt, NULL, 0, 0); - if (err) - GOTO(out, err); - err = mdc_rename(sbi->ll_mdc_exp, &op_data, - src_name->name, src_name->len, - tgt_name->name, tgt_name->len, &request); - if (err) - GOTO(out, err); - ll_update_times(request, REPLY_REC_OFF, src); - ll_update_times(request, REPLY_REC_OFF, tgt); - err = ll_objects_destroy(request, src); - if (err) - GOTO(out, err); + if (unlikely(ll_d_mountpoint(src_dparent, src_dchild, src_name) || + ll_d_mountpoint(tgt_dparent, tgt_dchild, tgt_name))) + RETURN(-EBUSY); + + op_data = ll_prep_md_op_data(NULL, src, tgt, NULL, 0, 0, + LUSTRE_OPC_ANY, NULL); + if (IS_ERR(op_data)) + RETURN(PTR_ERR(op_data)); + + ll_get_child_fid(src, src_name, &op_data->op_fid3); + ll_get_child_fid(tgt, tgt_name, &op_data->op_fid4); + err = md_rename(sbi->ll_md_exp, op_data, + src_name->name, src_name->len, + tgt_name->name, tgt_name->len, &request); + ll_finish_md_op_data(op_data); + if (!err) { + ll_update_times(request, src); + ll_update_times(request, tgt); + ll_stats_ops_tally(sbi, LPROC_LL_RENAME, 1); + err = ll_objects_destroy(request, src); + } - EXIT; -out: ptlrpc_req_finished(request); - return(err); -} - -#ifdef LUSTRE_KERNEL_VERSION -static int ll_mknod_raw(struct nameidata *nd, int mode, dev_t rdev) -{ - return ll_mknod_generic(nd->dentry->d_inode, &nd->last, mode,rdev,NULL); -} -static int ll_rename_raw(struct nameidata *srcnd, struct nameidata *tgtnd) -{ - return ll_rename_generic(srcnd->dentry->d_inode, &srcnd->last, - tgtnd->dentry->d_inode, &tgtnd->last); -} -static int ll_link_raw(struct nameidata *srcnd, struct nameidata *tgtnd) -{ - return ll_link_generic(srcnd->dentry->d_inode, tgtnd->dentry->d_inode, - &tgtnd->last, NULL); -} -static int ll_symlink_raw(struct nameidata *nd, const char *tgt) -{ - return ll_symlink_generic(nd->dentry->d_inode, &nd->last, tgt, NULL); -} -static int ll_rmdir_raw(struct nameidata *nd) -{ - return ll_rmdir_generic(nd->dentry->d_inode, nd->dentry, &nd->last); -} -static int ll_mkdir_raw(struct nameidata *nd, int mode) -{ - return ll_mkdir_generic(nd->dentry->d_inode, &nd->last, mode, NULL); -} -static int ll_unlink_raw(struct nameidata *nd) -{ - return ll_unlink_generic(nd->dentry->d_inode, &nd->last); + RETURN(err); } -#endif -static int ll_mknod(struct inode *dir, struct dentry *dchild, int mode, - ll_dev_t rdev) +static int ll_mknod(struct inode *dir, struct dentry *dchild, ll_umode_t mode, + dev_t rdev) { - return ll_mknod_generic(dir, &dchild->d_name, mode, - old_encode_dev(rdev), dchild); + return ll_mknod_generic(dir, &dchild->d_name, mode, + old_encode_dev(rdev), dchild); } -#if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0)) static int ll_unlink(struct inode * dir, struct dentry *dentry) { - return ll_unlink_generic(dir, &dentry->d_name); + return ll_unlink_generic(dir, NULL, dentry, &dentry->d_name); } -static int ll_mkdir(struct inode *dir, struct dentry *dentry, int mode) + +static int ll_mkdir(struct inode *dir, struct dentry *dentry, ll_umode_t mode) { return ll_mkdir_generic(dir, &dentry->d_name, mode, dentry); } + static int ll_rmdir(struct inode *dir, struct dentry *dentry) { - return ll_rmdir_generic(dir, NULL, &dentry->d_name); + return ll_rmdir_generic(dir, NULL, dentry, &dentry->d_name); } + static int ll_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) { return ll_symlink_generic(dir, &dentry->d_name, oldname, dentry); } + static int ll_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry) { - return ll_link_generic(old_dentry->d_inode, dir, - &new_dentry->d_name, new_dentry); + return ll_link_generic(old_dentry->d_inode, dir, &new_dentry->d_name, + new_dentry); } + static int ll_rename(struct inode *old_dir, struct dentry *old_dentry, struct inode *new_dir, struct dentry *new_dentry) { - return ll_rename_generic(old_dir, &old_dentry->d_name, new_dir, + int err; + err = ll_rename_generic(old_dir, NULL, + old_dentry, &old_dentry->d_name, + new_dir, NULL, new_dentry, &new_dentry->d_name); -} + if (!err) { +#ifndef HAVE_FS_RENAME_DOES_D_MOVE + if (!S_ISDIR(old_dentry->d_inode->i_mode)) #endif + d_move(old_dentry, new_dentry); + } + return err; +} struct inode_operations ll_dir_inode_operations = { -#ifdef LUSTRE_KERNEL_VERSION - .link_raw = ll_link_raw, - .unlink_raw = ll_unlink_raw, - .symlink_raw = ll_symlink_raw, - .mkdir_raw = ll_mkdir_raw, - .rmdir_raw = ll_rmdir_raw, - .mknod_raw = ll_mknod_raw, - .rename_raw = ll_rename_raw, - .setattr = ll_setattr, - .setattr_raw = ll_setattr_raw, + .mknod = ll_mknod, +#ifdef HAVE_IOP_ATOMIC_OPEN + .atomic_open = ll_atomic_open, #endif - .mknod = ll_mknod, -#if (LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0)) - .create_it = ll_create_it, - .lookup_it = ll_lookup_it, - .revalidate_it = ll_inode_revalidate_it, -#else - .lookup = ll_lookup_nd, - .create = ll_create_nd, - /* We need all these non-raw things for NFSD, to not patch it. */ - .unlink = ll_unlink, - .mkdir = ll_mkdir, - .rmdir = ll_rmdir, - .symlink = ll_symlink, - .link = ll_link, - .rename = ll_rename, - .setattr = ll_setattr, - .getattr = ll_getattr, + .lookup = ll_lookup_nd, + .create = ll_create_nd, + /* We need all these non-raw things for NFSD, to not patch it. */ + .unlink = ll_unlink, + .mkdir = ll_mkdir, + .rmdir = ll_rmdir, + .symlink = ll_symlink, + .link = ll_link, + .rename = ll_rename, + .setattr = ll_setattr, + .getattr = ll_getattr, + .permission = ll_inode_permission, + .setxattr = ll_setxattr, + .getxattr = ll_getxattr, + .listxattr = ll_listxattr, + .removexattr = ll_removexattr, +#ifdef HAVE_IOP_GET_ACL + .get_acl = ll_get_acl, #endif - .permission = ll_inode_permission, - .setxattr = ll_setxattr, - .getxattr = ll_getxattr, - .listxattr = ll_listxattr, - .removexattr = ll_removexattr, }; struct inode_operations ll_special_inode_operations = { -#ifdef LUSTRE_KERNEL_VERSION - .setattr_raw = ll_setattr_raw, -#endif - .setattr = ll_setattr, -#if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0)) - .getattr = ll_getattr, -#else - .revalidate_it = ll_inode_revalidate_it, + .setattr = ll_setattr, + .getattr = ll_getattr, + .permission = ll_inode_permission, + .setxattr = ll_setxattr, + .getxattr = ll_getxattr, + .listxattr = ll_listxattr, + .removexattr = ll_removexattr, +#ifdef HAVE_IOP_GET_ACL + .get_acl = ll_get_acl, #endif - .permission = ll_inode_permission, - .setxattr = ll_setxattr, - .getxattr = ll_getxattr, - .listxattr = ll_listxattr, - .removexattr = ll_removexattr, };