/* -*- 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
+ *
+ * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
+ * CA 95054 USA or visit www.sun.com if you need additional information or
+ * have any questions.
+ *
+ * GPL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. All rights reserved
+ * Use is subject to license terms.
+ */
+/*
+ * This file is part of Lustre, http://www.lustre.org/
+ * Lustre is a trademark of Sun Microsystems, Inc.
*/
#include <linux/fs.h>
#include <lustre_mdc.h>
#include "llite_internal.h"
-/* methods */
-extern struct dentry_operations ll_d_ops;
-
/*
* Check if we have something mounted at the named dchild.
* In such a case there would always be dentry present.
RETURN(0);
}
+
+/* called from iget5_locked->find_inode() under inode_lock spinlock */
+static int ll_test_inode(struct inode *inode, void *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)
+{
+ return 0;
+}
+
+
/*
* Get an inode by inode number (already instantiated by the intent lookup).
* Returns inode or NULL
struct lustre_md *md)
{
struct ll_inode_info *lli;
- struct inode *inode;
+ struct inode *inode;
+ ENTRY;
+
LASSERT(hash != 0);
+ inode = iget5_locked(sb, hash, ll_test_inode, ll_set_inode, md);
- inode = iget_locked(sb, hash);
if (inode) {
+ lli = ll_i2info(inode);
if (inode->i_state & I_NEW) {
- lli = ll_i2info(inode);
+ int rc;
+
ll_read_inode2(inode, md);
- unlock_new_inode(inode);
- } else {
- if (!(inode->i_state & (I_FREEING | I_CLEAR)))
+ rc = cl_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, "inode: %lu/%u(%p)\n",
- inode->i_ino, inode->i_generation, inode);
+ CDEBUG(D_VFSTRACE, "got inode: %p for "DFID"\n",
+ inode, PFID(&md->body->fid1));
}
-
- return inode;
+ RETURN(inode);
}
static void ll_drop_negative_dentry(struct inode *dir)
-{
+{
struct dentry *dentry, *tmp_alias, *tmp_subdir;
+ spin_lock(&ll_lookup_lock);
spin_lock(&dcache_lock);
restart:
list_for_each_entry_safe(dentry, tmp_alias,
}
}
spin_unlock(&dcache_lock);
+ spin_unlock(&ll_lookup_lock);
}
RETURN(0);
}
+__u32 ll_i2suppgid(struct inode *i)
+{
+ if (in_group_p(i->i_gid))
+ return (__u32)i->i_gid;
+ else
+ return (__u32)(-1);
+}
+
/* 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
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])
}
break;
}
+#endif
}
static void ll_d_add(struct dentry *de, struct inode *inode)
de->d_name.len, de->d_name.name, de, de->d_hash.next);
LBUG();
}
- __d_rehash(de, 0);
+ d_rehash_cond(de, 0);
}
/* Search "inode"'s alias list for a dentry that has the same name and parent
* in ll_revalidate_it. After revaliadate inode will be have hashed aliases
* and it triggers BUG_ON in d_instantiate_unique (bug #10954).
*/
-struct dentry *ll_find_alias(struct inode *inode, struct dentry *de)
+static struct dentry *ll_find_alias(struct inode *inode, struct dentry *de)
{
struct list_head *tmp;
struct dentry *dentry;
struct dentry *last_discon = NULL;
-
+
+ spin_lock(&ll_lookup_lock);
spin_lock(&dcache_lock);
list_for_each(tmp, &inode->i_dentry) {
dentry = list_entry(tmp, struct dentry, d_alias);
dget_locked(dentry);
lock_dentry(dentry);
__d_drop(dentry);
-#ifdef DCACHE_LUSTRE_INVALID
- dentry->d_flags &= ~DCACHE_LUSTRE_INVALID;
-#endif
unlock_dentry(dentry);
- __d_rehash(dentry, 0); /* avoid taking dcache_lock inside */
+ ll_dops_init(dentry, 0);
+ d_rehash_cond(dentry, 0); /* avoid taking dcache_lock inside */
spin_unlock(&dcache_lock);
+ spin_unlock(&ll_lookup_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));
- __d_rehash(de, 0);
dget_locked(last_discon);
- __d_move(last_discon, de);
+ lock_dentry(last_discon);
+ last_discon->d_flags |= DCACHE_LUSTRE_INVALID;
+ unlock_dentry(last_discon);
spin_unlock(&dcache_lock);
+ spin_unlock(&ll_lookup_lock);
+ ll_dops_init(last_discon, 1);
d_rehash(de);
+ d_move(last_discon, de);
iput(inode);
return last_discon;
}
-
+ de->d_flags |= DCACHE_LUSTRE_INVALID;
ll_d_add(de, inode);
spin_unlock(&dcache_lock);
+ spin_unlock(&ll_lookup_lock);
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 ll_sb_info *sbi = ll_i2sbi(parent);
struct inode *inode = NULL;
int rc;
+ ENTRY;
/* NB 1 request reference will be taken away by ll_intent_lock()
* when I return */
if (!it_disposition(it, DISP_LOOKUP_NEG)) {
- ENTRY;
+ struct dentry *save = *de;
+ __u32 bits;
- rc = ll_prep_inode(&inode, request, offset,
- (*de)->d_sb);
+ 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);
md_set_lock_data(sbi->ll_md_exp,
- &it->d.lustre.it_lock_handle, inode);
+ &it->d.lustre.it_lock_handle, inode, &bits);
/* We used to query real size from OSTs here, but actually
this is not needed. For stat() calls size would be updated
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.
+ cl_glimpse_size or some equivalent themselves anyway.
Also see bug 7198. */
+ ll_dops_init(*de, 1);
*de = ll_find_alias(inode, *de);
+ if (*de != save) {
+ struct ll_dentry_data *lld = ll_d2d(*de);
+
+ /* just make sure the ll_dentry_data is ready */
+ if (unlikely(lld == NULL)) {
+ ll_set_dd(*de);
+ lld = ll_d2d(*de);
+ if (likely(lld != NULL))
+ lld->lld_sa_generation = 0;
+ }
+ }
+ /* we have lookup look - unhide dentry */
+ if (bits & MDS_INODELOCK_LOOKUP) {
+ lock_dentry(*de);
+ (*de)->d_flags &= ~(DCACHE_LUSTRE_INVALID);
+ unlock_dentry(*de);
+ }
} else {
- ENTRY;
+ ll_dops_init(*de, 1);
/* 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);
+ ll_d_add(*de, NULL);
spin_unlock(&dcache_lock);
} else {
+ /* negative lookup - and don't have update lock to
+ * parent */
+ lock_dentry(*de);
+ (*de)->d_flags |= DCACHE_LUSTRE_INVALID;
+ unlock_dentry(*de);
+
(*de)->d_inode = NULL;
/* We do not want to hash the dentry if don`t have a
* lock, but if this dentry is later used in d_move,
}
}
- ll_set_dd(*de);
- (*de)->d_op = &ll_d_ops;
-
RETURN(0);
}
struct md_op_data *op_data;
struct it_cb_data icbd;
__u32 opc;
- int rc;
+ int rc, first = 0;
ENTRY;
if (dentry->d_name.len > ll_i2sbi(parent)->ll_namelen)
RETURN(ERR_PTR(rc));
}
+ if (it->it_op == IT_GETATTR) {
+ first = ll_statahead_enter(parent, &dentry, 1);
+ if (first >= 0) {
+ ll_statahead_exit(dentry, first);
+ if (first == 1)
+ RETURN(retval = dentry);
+ }
+ }
+
icbd.icbd_childp = &dentry;
icbd.icbd_parent = parent;
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));
}
+ if (first == -EEXIST)
+ ll_statahead_mark(dentry);
+
if ((it->it_op & IT_OPEN) && dentry->d_inode &&
!S_ISREG(dentry->d_inode->i_mode) &&
!S_ISDIR(dentry->d_inode->i_mode)) {
(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 */
+ 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
+ nd->intent.open.file->private_data = it;
+ (void)lookup_instantiate_filp(nd,dentry,
+ NULL);
#endif
}
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, DLM_REPLY_REC_OFF, dir->i_sb);
+ rc = ll_prep_inode(&inode, request, dir->i_sb);
if (rc)
GOTO(out, inode = ERR_PTR(rc));
CDEBUG(D_DLMTRACE, "setting l_ast_data to inode %p (%lu/%u)\n",
inode, inode->i_ino, inode->i_generation);
md_set_lock_data(sbi->ll_md_exp,
- &it->d.lustre.it_lock_handle, inode);
+ &it->d.lustre.it_lock_handle, inode, NULL);
EXIT;
out:
ptlrpc_req_finished(request);
* successful create. Hash it here. */
spin_lock(&dcache_lock);
if (d_unhashed(dentry))
- __d_rehash(dentry, 0);
+ d_rehash_cond(dentry, 0);
spin_unlock(&dcache_lock);
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 mdt_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);
/* 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
err = md_create(sbi->ll_md_exp, op_data, tgt, tgt_len, mode,
current->fsuid, current->fsgid,
- current->cap_effective, rdev, &request);
+ 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(&inode, request, REPLY_REC_OFF,
- dchild->d_sb);
+ err = ll_prep_inode(&inode, request, dchild->d_sb);
if (err)
GOTO(err_exit, err);
if (dchild)
d_drop(dchild);
- ll_update_times(request, REPLY_REC_OFF, dir);
+ ll_update_times(request, dir);
EXIT;
out:
ptlrpc_req_finished(request);
struct lu_fid *fid)
{
struct dentry *parent, *child;
-
+
parent = list_entry(dir->i_dentry.next, struct dentry, d_alias);
child = d_lookup(parent, name);
if (child) {
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 = 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, REPLY_REC_OFF, dir);
+ ll_update_times(request, dir);
ptlrpc_req_finished(request);
RETURN(rc);
}
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);
* 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_i2dtexp(dir), &lsm, eadata, body->eadatasize);
if (rc < 0) {
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_i2dtexp(dir), oa, lsm, &oti, ll_i2mdexp(dir));
+ 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(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",
if (rc)
GOTO(out, rc);
- ll_update_times(request, REPLY_REC_OFF, dir);
+ ll_update_times(request, dir);
rc = ll_objects_destroy(request, dir);
out:
tgt_name->name, tgt_name->len, &request);
ll_finish_md_op_data(op_data);
if (!err) {
- ll_update_times(request, REPLY_REC_OFF, src);
- ll_update_times(request, REPLY_REC_OFF, tgt);
+ ll_update_times(request, src);
+ ll_update_times(request, tgt);
err = ll_objects_destroy(request, src);
}
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, NULL,
+ 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 = {
git://git.whamcloud.com / fs / lustre-release.git / blobdiff