X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Fllite%2Fnamei.c;h=48303a3c4031820beb00deb027c899acf96e7fcc;hp=c70a5f1c0c02e6326ee46e91f0c1648ee9d09a34;hb=784cd144103871bd421c139c09bfbf4d5d29ca08;hpb=00097a75276ff4be42b25364d37351b54e741c90 diff --git a/lustre/llite/namei.c b/lustre/llite/namei.c index c70a5f1..48303a3 100644 --- a/lustre/llite/namei.c +++ b/lustre/llite/namei.c @@ -1,152 +1,198 @@ -/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*- - * vim:expandtab:shiftwidth=8:tabstop=8: - * - * Copyright (c) 2002, 2003 Cluster File Systems, Inc. +/* + * GPL HEADER START * - * This file is part of Lustre, http://www.lustre.org. + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * - * 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. + * 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. * - * 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. + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). * - * You should have received a copy of the GNU General Public License - * along with Lustre; if not, write to the Free Software - * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + * 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 * - * derived in small part from linux/fs/ext2/namei.c + * 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. * - * Copyright (C) 1991, 1992 Linus Torvalds + * GPL HEADER END + */ +/* + * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved. + * Use is subject to license terms. * - * Big-endian to little-endian byte-swapping/bitmaps by - * David S. Miller (davem@caip.rutgers.edu), 1995 - * Directory entry file type support and forward compatibility hooks - * for B-tree directories by Theodore Ts'o (tytso@mit.edu), 1998 + * Copyright (c) 2011, 2012, Whamcloud, Inc. + */ +/* + * 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 +#include +#include #include "llite_internal.h" -/* methods */ +static int ll_create_it(struct inode *, struct dentry *, + int, struct lookup_intent *); + +/* + * 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) +{ + 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); + } + } + return mounted; +} + +int ll_unlock(__u32 mode, struct lustre_handle *lockh) +{ + ENTRY; + + ldlm_lock_decref(lockh, mode); -#if (LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0)) -static int ll_test_inode(struct inode *inode, unsigned long ino, void *opaque) -#else + RETURN(0); +} + + +/* called from iget5_locked->find_inode() under inode_lock spinlock */ static int ll_test_inode(struct inode *inode, void *opaque) -#endif { - static int last_ino, last_gen, last_count; - struct lustre_md *md = opaque; + struct ll_inode_info *lli = ll_i2info(inode); + struct lustre_md *md = opaque; - if (!(md->body->valid & (OBD_MD_FLGENER | OBD_MD_FLID))) { - CERROR("MDS body missing inum or generation\n"); + if (unlikely(!(md->body->valid & OBD_MD_FLID))) { + CERROR("MDS body missing FID\n"); return 0; } - if (last_ino == id_ino(&md->body->id1) && - last_gen == id_gen(&md->body->id1) && - 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 = id_ino(&md->body->id1); - last_gen = id_gen(&md->body->id1); - CDEBUG(D_VFSTRACE, - "comparing inode %p ino "DLID4" to body "DLID4"\n", - inode, OLID4(&ll_i2info(inode)->lli_id), - OLID4(&md->body->id1)); - } - -#if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0)) - if (inode->i_ino != id_ino(&md->body->id1)) - return 0; -#endif - if (inode->i_generation != id_gen(&md->body->id1)) + if (!lu_fid_eq(&lli->lli_fid, &md->body->fid1)) return 0; - if (id_group(&ll_i2info(inode)->lli_id) != id_group(&md->body->id1)) - return 0; - - /* apply the attributes in 'opaque' to this inode. */ - ll_update_inode(inode, md); return 1; } -extern struct dentry_operations ll_d_ops; - -int ll_unlock(__u32 mode, struct lustre_handle *lockh) +static int ll_set_inode(struct inode *inode, void *opaque) { - ENTRY; + struct ll_inode_info *lli = ll_i2info(inode); + struct mdt_body *body = ((struct lustre_md *)opaque)->body; - ldlm_lock_decref(lockh, mode); + if (unlikely(!(body->valid & OBD_MD_FLID))) { + CERROR("MDS body missing FID\n"); + return -EINVAL; + } - RETURN(0); -} + 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); -/* - * 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) -{ - ll_read_inode2(inode, opaque); 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) + rc = cl_file_inode_init(inode, md); + if (rc != 0) { + md->lsm = NULL; + 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); + } + } + 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; @@ -163,22 +209,31 @@ int ll_mdc_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc, break; case LDLM_CB_CANCELING: { struct inode *inode = ll_inode_from_lock(lock); - struct ll_inode_info *li = ll_i2info(inode); + 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; - /* For lookup locks: Invalidate all dentries associated with - this inode, for UPDATE locks - invalidate directory pages */ - if (inode == NULL) { - if (bits & MDS_INODELOCK_OPEN) - LDLM_ERROR(lock, "null inode"); + /* Invalidate all dentries associated with this inode */ + if (inode == NULL) break; - } - if (lock->l_resource->lr_name.name[0] != id_fid(&li->lli_id) || - lock->l_resource->lr_name.name[1] != id_group(&li->lli_id)) { - LDLM_ERROR(lock, "data mismatch with object %lu/%lu", - (unsigned long)id_fid(&li->lli_id), - (unsigned long)id_group(&li->lli_id)); + 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)) + 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) { @@ -198,32 +253,33 @@ 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_md_real_close(ll_i2mdexp(inode), inode, flags); + ll_md_real_close(inode, flags); } - if (bits & MDS_INODELOCK_UPDATE) - clear_bit(LLI_F_HAVE_MDS_SIZE_LOCK, - &(ll_i2info(inode)->lli_flags)); + lli = ll_i2info(inode); + if (bits & MDS_INODELOCK_LAYOUT) { + struct cl_object_conf conf = { .coc_inode = inode, + .coc_invalidate = true }; + rc = ll_layout_conf(inode, &conf); + if (rc) + CDEBUG(D_INODE, "invaliding layout %d.\n", rc); + } + if (bits & MDS_INODELOCK_UPDATE) + lli->lli_flags &= ~LLIF_MDS_SIZE_LOCK; - /* If lookup lock is cancelled, we just drop the dentry and - this will cause us to reget data from MDS when we'd want to - access this dentry/inode again. If this is lock on - other parts of inode that is cancelled, we do not need to do - much (but need to discard data from readdir, if any), since - abscence of lock will cause ll_revalidate_it (called from - stat() and similar functions) to renew the data anyway */ if (S_ISDIR(inode->i_mode) && - (bits & MDS_INODELOCK_UPDATE)) { - CDEBUG(D_INODE, "invalidating inode %lu/%u(%p)\n", - inode->i_ino, inode->i_generation, inode); + (bits & MDS_INODELOCK_UPDATE)) { + CDEBUG(D_INODE, "invalidating inode %lu\n", + inode->i_ino); truncate_inode_pages(inode->i_mapping, 0); - } + ll_invalidate_negative_children(inode); + } - 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)) + ll_invalidate_aliases(inode); iput(inode); break; } @@ -234,123 +290,202 @@ int ll_mdc_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc, RETURN(0); } -/* 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. */ -struct dentry *ll_find_alias(struct inode *inode, struct dentry *de) +__u32 ll_i2suppgid(struct inode *i) { - struct list_head *tmp; + if (cfs_curproc_is_in_groups(i->i_gid)) + return (__u32)i->i_gid; + else + return (__u32)(-1); +} - spin_lock(&dcache_lock); - list_for_each(tmp, &inode->i_dentry) { - struct dentry *dentry = list_entry(tmp, struct dentry, d_alias); +/* Pack the required supplementary groups into the supplied groups array. + * If we don't need to use the groups from the target inode(s) then we + * instead pack one or more groups from the user's supplementary group + * 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 - /* We are called here with 'de' already on the aliases list. */ - if (dentry == de) { - CERROR("whoops\n"); - continue; - } + LASSERT(i1 != NULL); + LASSERT(suppgids != NULL); - if (dentry->d_parent != de->d_parent) - continue; + suppgids[0] = ll_i2suppgid(i1); - if (dentry->d_name.len != de->d_name.len) - continue; + if (i2) + suppgids[1] = ll_i2suppgid(i2); + else + suppgids[1] = -1; - if (memcmp(dentry->d_name.name, de->d_name.name, - de->d_name.len) != 0) +#if 0 + for (i = 0; i < current_ngroups; i++) { + if (suppgids[0] == -1) { + if (current_groups[i] != suppgids[1]) + suppgids[0] = current_groups[i]; continue; - - if (!list_empty(&dentry->d_lru)) - list_del_init(&dentry->d_lru); - - hlist_del_init(&dentry->d_hash); - __d_rehash(dentry); /* avoid taking dcache_lock inside */ - spin_unlock(&dcache_lock); - atomic_inc(&dentry->d_count); - iput(inode); - dentry->d_flags &= ~DCACHE_LUSTRE_INVALID; - 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 (suppgids[1] == -1) { + if (current_groups[i] != suppgids[0]) + suppgids[1] = current_groups[i]; + continue; + } + break; } +#endif +} - spin_unlock(&dcache_lock); +/* + * 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) +{ + 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; +} +/* + * Similar to d_splice_alias(), but lustre treats invalid alias + * similar to DCACHE_DISCONNECTED, and tries to use it anyway. + */ +struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de) +{ + 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 dentry *dentry = *de, *saved = *de; 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_dt_exp, sbi->ll_md_exp, - &inode, request, offset, dentry->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); 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(NULL, &LUSTRE_IT(it)->it_lock_handle, inode); - - /* If this is a stat, get the authoritative file size */ - if (it->it_op == IT_GETATTR && S_ISREG(inode->i_mode) && - ll_i2info(inode)->lli_smd != NULL) { - struct lov_stripe_md *lsm = ll_i2info(inode)->lli_smd; - ldlm_error_t rc; - - LASSERT(lsm->lsm_object_id != 0); - - /* bug 2334: drop MDS lock before acquiring OST lock */ - ll_intent_drop_lock(it); - - rc = ll_glimpse_size(inode); - if (rc) { - iput(inode); - RETURN(rc); - } - } - - dentry = *de = ll_find_alias(inode, dentry); - } else { - ENTRY; + 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 + from subsequent do_revalidate()->ll_inode_revalidate_it() in + 2.4 and + vfs_getattr_it->ll_getattr()->ll_inode_revalidate_it() in 2.6 + Everybody else who needs correct file size would call + ll_glimpse_size or some equivalent themselves anyway. + Also see bug 7198. */ + } + + /* 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); + } } - dentry->d_op = &ll_d_ops; - ll_set_dd(dentry); - - if (dentry == saved) - d_add(dentry, inode); - RETURN(0); } static struct dentry *ll_lookup_it(struct inode *parent, struct dentry *dentry, - struct nameidata *nd, int flags) + struct lookup_intent *it, int lookup_flags) { - struct lookup_intent *it = flags ? &nd->intent.open : NULL; struct lookup_intent lookup_it = { .it_op = IT_LOOKUP }; struct dentry *save = dentry, *retval; struct ptlrpc_request *req = NULL; - int rc, gns_it, gns_flags; + struct md_op_data *op_data; struct it_cb_data icbd; - struct lustre_id pid; + __u32 opc; + int rc; ENTRY; - if (dentry->d_name.len > EXT3_NAME_LEN) + 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", @@ -360,60 +495,65 @@ static struct dentry *ll_lookup_it(struct inode *parent, struct dentry *dentry, if (d_mountpoint(dentry)) CERROR("Tell Peter, lookup on mtpt, it %s\n", LL_IT2STR(it)); - if (nd != NULL) - nd->mnt->mnt_last_used = jiffies; - - gns_it = nd ? nd->intent.open.it_op : IT_OPEN; - gns_flags = nd ? nd->flags : LOOKUP_CONTINUE; ll_frob_intent(&it, &lookup_it); + /* As do_lookup is called before follow_mount, root dentry may be left + * 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, + 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; - ll_inode2id(&pid, parent); - rc = md_intent_lock(ll_i2mdexp(parent), &pid, (char *)dentry->d_name.name, - dentry->d_name.len, NULL, 0, NULL, it, flags, &req, - ll_mdc_blocking_ast); + 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; + + 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); + + /* 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, 1, it, &icbd); + rc = ll_lookup_it_finish(req, it, &icbd); if (rc != 0) { ll_intent_release(it); GOTO(out, retval = ERR_PTR(rc)); } + if ((it->it_op & IT_OPEN) && dentry->d_inode && + !S_ISREG(dentry->d_inode->i_mode) && + !S_ISDIR(dentry->d_inode->i_mode)) { + ll_release_openhandle(dentry, it); + } ll_lookup_finish_locks(it, dentry); - if (nd && dentry->d_inode != NULL && - dentry->d_inode->i_mode & S_ISUID && S_ISDIR(dentry->d_inode->i_mode) && - ((gns_flags & LOOKUP_CONTINUE) || (gns_it & (IT_CHDIR | IT_OPEN)))) - { - CDEBUG(D_DENTRY, "possible GNS dentry %*s %p found, " - "mounting it\n", (int)dentry->d_name.len, - dentry->d_name.name, dentry); - - rc = ll_gns_mount_object(dentry, nd->mnt); - if (rc) { - /* - * just reporting about GNS failures, lookup() is - * successful, do not stop it. - * - * GNS failure may be that object is found in SUID bit - * marked dir but it is not regular file and we should - * lookup further until we find correct mount - * object. This will allow to perform GNS mount is the - * following case for instance: - * - * /mnt/lustre/gns_mount/.mntinfo/.mntinfo/..../.mntinfo - * where all ".mntinfo" are dirs and only last one is - * reg file. - */ - CDEBUG(D_DENTRY, "failed to mount %*s, err %d\n", - (int)dentry->d_name.len, dentry->d_name.name, rc); - } - } - if (dentry == save) GOTO(out, retval = NULL); else @@ -421,40 +561,221 @@ static struct dentry *ll_lookup_it(struct inode *parent, struct dentry *dentry, out: if (req) ptlrpc_req_finished(req); - if (it == &lookup_it) - ll_intent_release(it); - if (dentry->d_inode) - CDEBUG(D_INODE, "lookup 0x%p in %lu/%lu: %*s -> %lu/%lu\n", - dentry, - (unsigned long) parent->i_ino, - (unsigned long) parent->i_generation, - dentry->d_name.len, dentry->d_name.name, - (unsigned long) dentry->d_inode->i_ino, - (unsigned long) dentry->d_inode->i_generation); - else - CDEBUG(D_INODE, "lookup 0x%p in %lu/%lu: %*s -> ??\n", - dentry, - (unsigned long) parent->i_ino, - (unsigned long) parent->i_generation, - dentry->d_name.len, dentry->d_name.name); + 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 HAVE_IOP_ATOMIC_OPEN +static struct dentry *ll_lookup_nd(struct inode *parent, struct dentry *dentry, + unsigned int flags) +{ + 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; +} + +/* + * 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 /* !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 & 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, struct nameidata *nd) { struct dentry *de; ENTRY; - if (nd && nd->flags & LOOKUP_LAST && !(nd->flags & LOOKUP_LINK_NOTLAST)) - de = ll_lookup_it(parent, dentry, nd, nd->flags); - else - de = ll_lookup_it(parent, dentry, nd, 0); + if (nd && !(nd->flags & (LOOKUP_CONTINUE|LOOKUP_PARENT))) { + struct lookup_intent *it; + + 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); + } + + de = ll_lookup_it(parent, dentry, it, nd->flags); + if (de) + dentry = de; + if ((nd->flags & LOOKUP_OPEN) && !IS_ERR(dentry)) { /* Open */ + if (dentry->d_inode && + it_disposition(it, DISP_OPEN_OPEN)) { /* nocreate */ + if (S_ISFIFO(dentry->d_inode->i_mode)) { + // We cannot call open here as it would + // deadlock. + ptlrpc_req_finished( + (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 (IS_ERR(filp)) { + if (de) + dput(de); + de = (struct dentry *)filp; + } + } + } else if (it_disposition(it, DISP_OPEN_CREATE)) { + // XXX This can only reliably work on assumption + // that there are NO hashed negative dentries. + ll_d2d(dentry)->lld_it = it; + it = NULL; /* Will be freed in ll_create_nd */ + /* We absolutely depend on ll_create_nd to be + * called to not leak this intent and possible + * data attached to it */ + } + } + + if (it) { + ll_intent_release(it); + OBD_FREE(it, sizeof(*it)); + } + } else { + de = ll_lookup_it(parent, dentry, NULL, 0); + } RETURN(de); } -#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, @@ -468,23 +789,23 @@ static struct inode *ll_create_node(struct inode *dir, const char *name, int rc; ENTRY; + LASSERT(it && it->d.lustre.it_disposition); - LASSERT(it && LUSTRE_IT(it)->it_disposition); - - request = LUSTRE_IT(it)->it_data; - rc = ll_prep_inode(sbi->ll_dt_exp, sbi->ll_md_exp, - &inode, request, 1, dir->i_sb); + 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(&inode, request, dir->i_sb); 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(NULL, &LUSTRE_IT(it)->it_lock_handle, inode); + ll_set_lock_data(sbi->ll_md_exp, inode, it, NULL); EXIT; out: ptlrpc_req_finished(request); @@ -509,8 +830,6 @@ static int ll_create_it(struct inode *dir, struct dentry *dentry, int mode, struct lookup_intent *it) { struct inode *inode; - struct ptlrpc_request *request = LUSTRE_IT(it)->it_data; - struct obd_export *md_exp = ll_i2mdexp(dir); int rc = 0; ENTRY; @@ -522,9 +841,8 @@ static int ll_create_it(struct inode *dir, struct dentry *dentry, int mode, if (rc) RETURN(rc); - mdc_store_inode_generation(md_exp, request, MDS_REQ_INTENT_REC_OFF, 1); - inode = ll_create_node(dir, (char *)dentry->d_name.name, - dentry->d_name.len, NULL, 0, mode, 0, it); + 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)); @@ -532,23 +850,16 @@ static int ll_create_it(struct inode *dir, struct dentry *dentry, int mode, RETURN(0); } -#if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0)) -static int ll_create_nd(struct inode *dir, struct dentry *dentry, int mode, struct nameidata *nd) -{ - return ll_create_it(dir, dentry, mode, &nd->intent.open); -} -#endif - -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); + 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 %u\n", + 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; } @@ -557,43 +868,72 @@ static void ll_update_times(struct ptlrpc_request *request, int offset, LTIME_S(inode->i_ctime) = body->ctime; } -static int ll_mknod_raw(struct nameidata *nd, int mode, dev_t rdev) +static int ll_new_node(struct inode *dir, struct qstr *name, + const char *tgt, int mode, int rdev, + struct dentry *dchild, __u32 opc) { struct ptlrpc_request *request = NULL; - struct inode *dir = nd->dentry->d_inode; + 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 err = -EMLINK; + int tgt_len = 0; + int err; + ENTRY; + if (unlikely(tgt != NULL)) + tgt_len = strlen(tgt) + 1; - CDEBUG(D_VFSTRACE, "VFS Op:name=%.*s,dir=%lu/%u(%p)\n", - nd->last.len, nd->last.name, dir->i_ino, - dir->i_generation, dir); + 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 = 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); - mode &= ~current->fs->umask; + ll_update_times(request, dir); + + if (dchild) { + err = ll_prep_inode(&inode, request, dchild->d_sb); + if (err) + GOTO(err_exit, err); + + d_instantiate(dchild, inode); + } + 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) +{ + int err; + ENTRY; + + CDEBUG(D_VFSTRACE, "VFS Op:name=%.*s,dir=%lu/%u(%p) mode %o dev %x\n", + name->len, name->name, dir->i_ino, dir->i_generation, dir, + mode, rdev); + + if (!IS_POSIXACL(dir) || !exp_connect_umask(ll_i2mdexp(dir))) + mode &= ~cfs_curproc_umask(); switch (mode & S_IFMT) { case 0: - case S_IFREG: mode |= S_IFREG; /* for mode = 0 case, fallthrough */ + case S_IFREG: case S_IFCHR: case S_IFBLK: case S_IFIFO: case S_IFSOCK: - OBD_ALLOC(op_data, sizeof(*op_data)); - if (op_data == NULL) - RETURN(-ENOMEM); - ll_prepare_mdc_data(op_data, dir, NULL, - (char *)nd->last.name, - nd->last.len, 0); - - err = md_create(sbi->ll_md_exp, op_data, NULL, 0, mode, - current->fsuid, current->fsgid, rdev, - &request); - OBD_FREE(op_data, sizeof(*op_data)); - if (err == 0) - ll_update_times(request, 0, dir); - ptlrpc_req_finished(request); + err = ll_new_node(dir, name, NULL, mode, rdev, dchild, + LUSTRE_OPC_MKNOD); break; case S_IFDIR: err = -EPERM; @@ -601,313 +941,441 @@ static int ll_mknod_raw(struct nameidata *nd, int mode, dev_t rdev) default: err = -EINVAL; } + + if (!err) + ll_stats_ops_tally(ll_i2sbi(dir), LPROC_LL_MKNOD, 1); + RETURN(err); } -static int ll_mknod(struct inode *dir, struct dentry *dchild, - int mode, ll_dev_t rdev) +#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) { - struct ptlrpc_request *request = NULL; - struct inode *inode = NULL; - struct ll_sb_info *sbi = ll_i2sbi(dir); - struct mdc_op_data *op_data; - int err = -EMLINK; - ENTRY; + int rc; - CDEBUG(D_VFSTRACE, "VFS Op:name=%.*s,dir=%lu/%u(%p)\n", - dchild->d_name.len, dchild->d_name.name, - dir->i_ino, dir->i_generation, dir); + 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); - mode &= ~current->fs->umask; + rc = ll_mknod_generic(dir, &dentry->d_name, mode, 0, dentry); - switch (mode & S_IFMT) { - case 0: - case S_IFREG: - mode |= S_IFREG; /* for mode = 0 case, fallthrough */ - case S_IFCHR: - case S_IFBLK: - case S_IFIFO: - case S_IFSOCK: - OBD_ALLOC(op_data, sizeof(*op_data)); - if (op_data == NULL) - RETURN(-ENOMEM); - - ll_prepare_mdc_data(op_data, dir, NULL, - (char *)dchild->d_name.name, - dchild->d_name.len, 0); - - err = md_create(sbi->ll_md_exp, op_data, NULL, 0, mode, - current->fsuid, current->fsgid, rdev, - &request); - OBD_FREE(op_data, sizeof(*op_data)); - if (err) - GOTO(out_err, err); + ll_stats_ops_tally(ll_i2sbi(dir), LPROC_LL_CREATE, 1); - ll_update_times(request, 0, dir); - err = ll_prep_inode(sbi->ll_dt_exp, sbi->ll_md_exp, - &inode, request, 0, dchild->d_sb); - if (err) - GOTO(out_err, err); - break; - case S_IFDIR: - RETURN(-EPERM); - break; - default: - RETURN(-EINVAL); + 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; + + if (!it) + return ll_mknod_generic(dir, &dentry->d_name, mode, 0, dentry); + + ll_d2d(dentry)->lld_it = NULL; + + /* Was there an error? Propagate it! */ + if (it->d.lustre.it_status) { + rc = it->d.lustre.it_status; + goto out; } - d_instantiate(dchild, inode); - EXIT; - out_err: - ptlrpc_req_finished(request); - return err; + rc = ll_create_it(dir, dentry, mode, it); + if (nd && (nd->flags & LOOKUP_OPEN) && dentry->d_inode) { /* Open */ + 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); + } + +out: + ll_intent_release(it); + OBD_FREE(it, sizeof(*it)); + + if (!rc) + ll_stats_ops_tally(ll_i2sbi(dir), LPROC_LL_CREATE, 1); + + return rc; } +#endif /* HAVE_IOP_ATOMIC_OPEN */ -static int ll_symlink_raw(struct nameidata *nd, const char *tgt) +static int ll_symlink_generic(struct inode *dir, struct qstr *name, + const char *tgt, struct dentry *dchild) { - const char *name = (char *)nd->last.name; - struct inode *dir = nd->dentry->d_inode; - struct ptlrpc_request *request = NULL; - struct ll_sb_info *sbi = ll_i2sbi(dir); - struct mdc_op_data *op_data; - int len = nd->last.len; - int err = -EMLINK; + int err; ENTRY; - CDEBUG(D_VFSTRACE, "VFS Op:name=%*s,dir=%lu/%u(%p),target=%s\n", - nd->last.len, nd->last.name, dir->i_ino, dir->i_generation, - dir, tgt); - - if (dir->i_nlink >= EXT3_LINK_MAX) - RETURN(err); - - OBD_ALLOC(op_data, sizeof(*op_data)); - if (op_data == NULL) - RETURN(-ENOMEM); - ll_prepare_mdc_data(op_data, dir, NULL, name, len, 0); - LASSERT(tgt); - err = md_create(sbi->ll_md_exp, op_data, - tgt, strlen(tgt) + 1, S_IFLNK | S_IRWXUGO, - current->fsuid, current->fsgid, 0, &request); - OBD_FREE(op_data, sizeof(*op_data)); - if (err == 0) - ll_update_times(request, 0, dir); - - ptlrpc_req_finished(request); + CDEBUG(D_VFSTRACE, "VFS Op:name=%.*s,dir=%lu/%u(%p),target=%.*s\n", + name->len, name->name, dir->i_ino, dir->i_generation, + dir, 3000, tgt); + + err = ll_new_node(dir, name, (char *)tgt, S_IFLNK | S_IRWXUGO, + 0, dchild, LUSTRE_OPC_SYMLINK); + + if (!err) + ll_stats_ops_tally(ll_i2sbi(dir), LPROC_LL_SYMLINK, 1); + RETURN(err); } -static int ll_link_raw(struct nameidata *srcnd, struct nameidata *tgtnd) +static int ll_link_generic(struct inode *src, struct inode *dir, + struct qstr *name, struct dentry *dchild) { - struct inode *src = srcnd->dentry->d_inode; - struct inode *dir = tgtnd->dentry->d_inode; - struct ptlrpc_request *request = NULL; struct ll_sb_info *sbi = ll_i2sbi(dir); - struct mdc_op_data *op_data; + struct ptlrpc_request *request = NULL; + struct md_op_data *op_data; int err; - ENTRY; + ENTRY; CDEBUG(D_VFSTRACE, "VFS Op: inode=%lu/%u(%p), dir=%lu/%u(%p), target=%.*s\n", src->i_ino, src->i_generation, src, dir->i_ino, - dir->i_generation, dir, tgtnd->last.len, tgtnd->last.name); + dir->i_generation, dir, name->len, name->name); - OBD_ALLOC(op_data, sizeof(*op_data)); - if (op_data == NULL) - RETURN(-ENOMEM); + 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)); - ll_prepare_mdc_data(op_data, src, dir, - (char *)tgtnd->last.name, - tgtnd->last.len, 0); - err = md_link(sbi->ll_md_exp, op_data, &request); - OBD_FREE(op_data, sizeof(*op_data)); - if (err == 0) - ll_update_times(request, 0, dir); + ll_finish_md_op_data(op_data); + if (err) + GOTO(out, err); + + 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_raw(struct nameidata *nd, int mode) { - struct inode *dir = nd->dentry->d_inode; - struct ptlrpc_request *request = NULL; - struct ll_sb_info *sbi = ll_i2sbi(dir); - struct mdc_op_data *op_data; - int err = -EMLINK; + int err; ENTRY; + CDEBUG(D_VFSTRACE, "VFS Op:name=%.*s,dir=%lu/%u(%p)\n", - nd->last.len, nd->last.name, dir->i_ino, dir->i_generation, dir); - - mode = (mode & (S_IRWXUGO|S_ISVTX) & ~current->fs->umask) | S_IFDIR; - OBD_ALLOC(op_data, sizeof(*op_data)); - if (op_data == NULL) - RETURN(-ENOMEM); - - ll_prepare_mdc_data(op_data, dir, NULL, - (char *)nd->last.name, - nd->last.len, 0); - - err = md_create(sbi->ll_md_exp, op_data, NULL, 0, mode, - current->fsuid, current->fsgid, 0, &request); - OBD_FREE(op_data, sizeof(*op_data)); - if (err == 0) - ll_update_times(request, 0, dir); - ptlrpc_req_finished(request); + name->len, name->name, dir->i_ino, dir->i_generation, dir); + + 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); } -static int ll_rmdir_raw(struct nameidata *nd) +/* 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 dentry *dchild, struct qstr *name) { - struct inode *dir = nd->dentry->d_inode; 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", - nd->last.len, nd->last.name, dir->i_ino, - dir->i_generation, dir); + name->len, name->name, dir->i_ino, dir->i_generation, dir); + + if (unlikely(ll_d_mountpoint(dparent, dchild, name))) + RETURN(-EBUSY); - OBD_ALLOC(op_data, sizeof(*op_data)); - if (op_data == NULL) - RETURN(-ENOMEM); + 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_prepare_mdc_data(op_data, dir, NULL, - (char *)nd->last.name, - nd->last.len, S_IFDIR); - + ll_get_child_fid(dir, name, &op_data->op_fid3); rc = md_unlink(ll_i2sbi(dir)->ll_md_exp, op_data, &request); - OBD_FREE(op_data, sizeof(*op_data)); - if (rc == 0) - ll_update_times(request, 0, dir); + 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); } -static int ll_unlink_raw(struct nameidata *nd) +int ll_objects_destroy(struct ptlrpc_request *request, struct inode *dir) +{ + 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 = req_capsule_server_get(&request->rq_pill, &RMF_MDT_BODY); + if (!(body->valid & OBD_MD_FLEASIZE)) + RETURN(0); + + if (body->eadatasize == 0) { + CERROR("OBD_MD_FLEASIZE set but eadatasize zero\n"); + GOTO(out, rc = -EPROTO); + } + + /* The MDS sent back the EA because we unlinked the last reference + * 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 = req_capsule_server_sized_get(&request->rq_pill, &RMF_MDT_MD, + body->eadatasize); + LASSERT(eadata != NULL); + + 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)); + + OBDO_ALLOC(oa); + if (oa == NULL) + GOTO(out_free_memmd, rc = -ENOMEM); + + oa->o_id = lsm->lsm_object_id; + oa->o_seq = lsm->lsm_object_seq; + oa->o_mode = body->mode & S_IFMT; + 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 = + 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; + } + } + + 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); + 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_i2dtexp(dir), &lsm); + out: + return rc; +} + +/* 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 inode *dir = nd->dentry->d_inode; 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", - nd->last.len, nd->last.name, dir->i_ino, - dir->i_generation, dir); - - OBD_ALLOC(op_data, sizeof(*op_data)); - if (op_data == NULL) - RETURN(-ENOMEM); - ll_prepare_mdc_data(op_data, dir, NULL, - (char *)nd->last.name, nd->last.len, 0); + name->len, name->name, dir->i_ino, dir->i_generation, dir); + + /* + * 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_get_child_fid(dir, name, &op_data->op_fid3); rc = md_unlink(ll_i2sbi(dir)->ll_md_exp, op_data, &request); - OBD_FREE(op_data, sizeof(*op_data)); + ll_finish_md_op_data(op_data); if (rc) GOTO(out, rc); - ll_update_times(request, 0, dir); - EXIT; -out: + + ll_update_times(request, dir); + ll_stats_ops_tally(ll_i2sbi(dir), LPROC_LL_UNLINK, 1); + + rc = ll_objects_destroy(request, dir); + out: ptlrpc_req_finished(request); - return rc; + RETURN(rc); } -static int ll_rename_raw(struct nameidata *srcnd, struct nameidata *tgtnd) +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 inode *src = srcnd->dentry->d_inode; - struct inode *tgt = tgtnd->dentry->d_inode; 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; - - if (srcnd->mnt != tgtnd->mnt) - RETURN(-EXDEV); - CDEBUG(D_VFSTRACE,"VFS Op:oldname=%.*s,src_dir=%lu/%u(%p),newname=%.*s," - "tgt_dir=%lu/%u(%p)\n", srcnd->last.len, srcnd->last.name, - src->i_ino, src->i_generation, src, tgtnd->last.len, - tgtnd->last.name, tgt->i_ino, tgt->i_generation, tgt); - - OBD_ALLOC(op_data, sizeof(*op_data)); - if (op_data == NULL) - RETURN(-ENOMEM); - ll_prepare_mdc_data(op_data, src, tgt, NULL, 0, 0); - err = md_rename(sbi->ll_md_exp, op_data, (char *)srcnd->last.name, - srcnd->last.len, (char *)tgtnd->last.name, - tgtnd->last.len, &request); - OBD_FREE(op_data, sizeof(*op_data)); + "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); + + 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, 0, src); - ll_update_times(request, 0, tgt); + 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); } ptlrpc_req_finished(request); + RETURN(err); } -#if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0)) -#define LLITE_IT_RAWOPS (IT_MKNOD|IT_MKDIR|IT_SYMLINK|IT_LINK|IT_UNLINK|IT_RMDIR|IT_RENAME) -static int ll_rawop_from_intent(struct nameidata *nd) +static int ll_mknod(struct inode *dir, struct dentry *dchild, ll_umode_t mode, + dev_t rdev) { - int error = 0; + return ll_mknod_generic(dir, &dchild->d_name, mode, + old_encode_dev(rdev), dchild); +} - if (!nd || !(nd->intent.open.op & LLITE_IT_RAWOPS)) - return 0; +static int ll_unlink(struct inode * dir, struct dentry *dentry) +{ + return ll_unlink_generic(dir, NULL, dentry, &dentry->d_name); +} - switch (nd->intent.open.op) { - case IT_MKNOD: - error = ll_mknod_raw(nd, nd->intent.open.create_mode, - nd->intent.open.create.dev); - break; - case IT_MKDIR: - error = ll_mkdir_raw(nd, nd->intent.open.create_mode); - break; - case IT_RMDIR: - error = ll_rmdir_raw(nd); - break; - case IT_UNLINK: - error = ll_unlink_raw(nd); - break; - case IT_SYMLINK: - LASSERT(nd->intent.open.create.link); - error = ll_symlink_raw(nd, nd->intent.open.create.link); - break; - case IT_LINK: - error = ll_link_raw(nd->intent.open.create.source_nd, nd); - break; - case IT_RENAME: - LASSERT(nd->intent.open.create.source_nd); - error = ll_rename_raw(nd->intent.open.create.source_nd, nd); - break; - default: - LBUG(); - } - if (error != -EOPNOTSUPP) - nd->intent.open.flags |= IT_STATUS_RAW; +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, &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); +} - return error; +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); } + +static int ll_rename(struct inode *old_dir, struct dentry *old_dentry, + struct inode *new_dir, struct dentry *new_dentry) +{ + 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 = { - .mknod = ll_mknod, - .setattr = ll_setattr, -#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, - .getattr = ll_getattr, - .endparentlookup = ll_rawop_from_intent, + .mknod = ll_mknod, +#ifdef HAVE_IOP_ATOMIC_OPEN + .atomic_open = ll_atomic_open, +#endif + .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 +}; + +struct inode_operations ll_special_inode_operations = { + .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 - .setxattr = ll_setxattr, - .getxattr = ll_getxattr, - .listxattr = ll_listxattr, - .removexattr = ll_removexattr, - .permission = ll_inode_permission, };