/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
* vim:expandtab:shiftwidth=8:tabstop=8:
*
- * Lustre Light name resolution
+ * 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.
+ *
+ * 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).
*
- * Copyright (c) 2002, 2003 Cluster File Systems, Inc.
+ * 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
*
- * This file is part of Lustre, http://www.lustre.org.
+ * 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 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.
+ * 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.
*
- * 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.
+ * lustre/liblustre/namei.c
*
- * 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.
+ * Lustre Light name resolution
*/
#define DEBUG_SUBSYSTEM S_LLITE
#include <time.h>
#include <sys/types.h>
#include <sys/stat.h>
-#include <sys/fcntl.h>
+#include <fcntl.h>
#include <sys/queue.h>
#include <sysio.h>
+#ifdef HAVE_XTIO_H
+#include <xtio.h>
+#endif
#include <fs.h>
#include <mount.h>
#include <inode.h>
+#ifdef HAVE_FILE_H
#include <file.h>
+#endif
#undef LIST_HEAD
#include "llite_lib.h"
-static void ll_intent_drop_lock(struct lookup_intent *it)
+void ll_intent_drop_lock(struct lookup_intent *it)
{
struct lustre_handle *handle;
}
}
-static void ll_intent_release(struct lookup_intent *it)
+void ll_intent_release(struct lookup_intent *it)
{
ENTRY;
EXIT;
}
-/*
- * remove the stale inode from pnode
- */
-void unhook_stale_inode(struct pnode *pno)
-{
- struct inode *inode = pno->p_base->pb_ino;
- ENTRY;
-
- LASSERT(inode);
- LASSERT(llu_i2info(inode)->lli_stale_flag);
-
- pno->p_base->pb_ino = NULL;
- I_RELE(inode);
-
- if (!llu_i2info(inode)->lli_open_count) {
- CDEBUG(D_INODE, "unhook inode %p (ino %lu) from pno %p\n",
- inode, llu_i2info(inode)->lli_st_ino, pno);
- if (!inode->i_ref)
- _sysio_i_gone(inode);
- }
-
- EXIT;
- return;
-}
-
void llu_lookup_finish_locks(struct lookup_intent *it, struct pnode *pnode)
{
+ struct inode *inode;
LASSERT(it);
LASSERT(pnode);
- if (it && pnode->p_base->pb_ino != NULL) {
- struct inode *inode = pnode->p_base->pb_ino;
- CDEBUG(D_DLMTRACE, "setting l_data to inode %p (%lu/%lu)\n",
- inode, llu_i2info(inode)->lli_st_ino,
+ inode = pnode->p_base->pb_ino;
+ if (it->d.lustre.it_lock_mode && inode != NULL) {
+ struct llu_sb_info *sbi;
+
+ CDEBUG(D_DLMTRACE, "setting l_data to inode %p (%llu/%lu)\n",
+ inode, (long long)llu_i2stat(inode)->st_ino,
llu_i2info(inode)->lli_st_generation);
- mdc_set_lock_data(NULL, &it->d.lustre.it_lock_handle, inode);
+
+ sbi = llu_i2sbi(inode);
+ md_set_lock_data(sbi->ll_md_exp,
+ &it->d.lustre.it_lock_handle, inode);
}
/* drop lookup/getattr locks */
- if (it->it_op == IT_LOOKUP || it->it_op == IT_GETATTR)
+ if (it->it_op & (IT_LOOKUP | IT_GETATTR))
ll_intent_release(it);
}
/* do nothing */
}
-int llu_mdc_blocking_ast(struct ldlm_lock *lock,
- struct ldlm_lock_desc *desc,
- void *data, int flag)
+int llu_md_blocking_ast(struct ldlm_lock *lock,
+ struct ldlm_lock_desc *desc,
+ void *data, int flag)
{
- int rc;
struct lustre_handle lockh;
+ int rc;
ENTRY;
case LDLM_CB_CANCELING: {
struct inode *inode = llu_inode_from_lock(lock);
struct llu_inode_info *lli;
+ struct intnl_stat *st;
__u64 bits = lock->l_policy_data.l_inodebits.bits;
+ struct lu_fid *fid;
/* Invalidate all dentries associated with this inode */
if (inode == NULL)
break;
lli = llu_i2info(inode);
+ st = llu_i2stat(inode);
if (bits & MDS_INODELOCK_UPDATE)
- clear_bit(LLI_F_HAVE_MDS_SIZE_LOCK, &lli->lli_flags);
-
- if (lock->l_resource->lr_name.name[0] != lli->lli_st_ino ||
- lock->l_resource->lr_name.name[1] != lli->lli_st_generation) {
- LDLM_ERROR(lock, "data mismatch with ino %lu/%lu",
- lli->lli_st_ino, lli->lli_st_generation);
+ lli->lli_flags &= ~LLIF_MDS_SIZE_LOCK;
+
+ fid = &lli->lli_fid;
+ 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 ino %llu/%llu/%llu",
+ (long long)fid_seq(fid),
+ (long long)fid_oid(fid),
+ (long long)fid_ver(fid));
}
- if (S_ISDIR(lli->lli_st_mode) &&
+ if (S_ISDIR(st->st_mode) &&
(bits & MDS_INODELOCK_UPDATE)) {
- CDEBUG(D_INODE, "invalidating inode %lu\n",
- lli->lli_st_ino);
+ CDEBUG(D_INODE, "invalidating inode %llu\n",
+ (long long)st->st_ino);
llu_invalidate_inode_pages(inode);
}
}
static int pnode_revalidate_finish(struct ptlrpc_request *req,
- int offset,
struct lookup_intent *it,
struct pnode *pnode)
{
if (it_disposition(it, DISP_LOOKUP_NEG))
RETURN(-ENOENT);
- rc = mdc_req2lustre_md(llu_i2sbi(inode)->ll_mdc_exp, req, offset,
- llu_i2sbi(inode)->ll_osc_exp, &md);
+ rc = md_get_lustre_md(llu_i2sbi(inode)->ll_md_exp, req,
+ llu_i2sbi(inode)->ll_dt_exp,
+ llu_i2sbi(inode)->ll_md_exp, &md);
if (rc)
RETURN(rc);
- llu_update_inode(inode, md.body, md.lsm);
+ llu_update_inode(inode, &md);
RETURN(rc);
}
-int llu_pb_revalidate(struct pnode *pnode, int flags, struct lookup_intent *it)
+static int llu_pb_revalidate(struct pnode *pnode, int flags,
+ struct lookup_intent *it)
{
struct pnode_base *pb = pnode->p_base;
- struct ll_fid pfid, cfid;
struct it_cb_data icbd;
- struct ll_uctxt ctxt;
+ struct md_op_data op_data = {{ 0 }};
struct ptlrpc_request *req = NULL;
struct lookup_intent lookup_it = { .it_op = IT_LOOKUP };
struct obd_export *exp;
int rc;
ENTRY;
- CDEBUG(D_VFSTRACE, "VFS Op:name=%s,intent=%x\n",
- pb->pb_name.name, it ? it->it_op : 0);
+ CDEBUG(D_VFSTRACE, "VFS Op:name=%.*s,intent=%x\n",
+ (int)pb->pb_name.len, pb->pb_name.name, it ? it->it_op : 0);
/* We don't want to cache negative dentries, so return 0 immediately.
* We believe that this is safe, that negative dentries cannot be
RETURN(0);
}
- /* check stale inode */
- if (llu_i2info(pb->pb_ino)->lli_stale_flag)
- unhook_stale_inode(pnode);
-
- /* check again because unhook_stale_inode() might generate
- * negative pnode */
- if (pb->pb_ino == NULL) {
- CDEBUG(D_INODE, "negative pb\n");
- RETURN(0);
- }
-
/* This is due to bad interaction with libsysio. remove this when we
* switched to libbsdio XXX
*/
{
struct llu_inode_info *lli = llu_i2info(pb->pb_ino);
+ struct intnl_stat *st = llu_i2stat(pb->pb_ino);
if (lli->lli_it) {
- CDEBUG(D_INODE, "inode %lu still have intent "
+ CDEBUG(D_INODE, "inode %llu still have intent "
"%p(opc 0x%x), release it\n",
- lli->lli_st_ino, lli->lli_it,
+ (long long) st->st_ino, lli->lli_it,
lli->lli_it->it_op);
ll_intent_release(lli->lli_it);
OBD_FREE(lli->lli_it, sizeof(*lli->lli_it));
}
}
- exp = llu_i2mdcexp(pb->pb_ino);
- ll_inode2fid(&pfid, pnode->p_parent->p_base->pb_ino);
- ll_inode2fid(&cfid, pb->pb_ino);
+ exp = llu_i2mdexp(pb->pb_ino);
icbd.icbd_parent = pnode->p_parent->p_base->pb_ino;
icbd.icbd_child = pnode;
it->it_op_release = ll_intent_release;
}
- ll_i2uctxt(&ctxt, pnode->p_parent->p_base->pb_ino, pb->pb_ino);
+ llu_prep_md_op_data(&op_data, pnode->p_parent->p_base->pb_ino,
+ pb->pb_ino, pb->pb_name.name, pb->pb_name.len,
+ 0, LUSTRE_OPC_ANY);
- rc = mdc_intent_lock(exp, &ctxt, &pfid,
- pb->pb_name.name, pb->pb_name.len,
- NULL, 0, &cfid, it, flags, &req,
- llu_mdc_blocking_ast);
- /* If req is NULL, then mdc_intent_lock only tried to do a lock match;
+ rc = md_intent_lock(exp, &op_data, NULL, 0, it, flags,
+ &req, llu_md_blocking_ast,
+ LDLM_FL_CANCEL_ON_BLOCK);
+ /* If req is NULL, then md_intent_lock only tried to do a lock match;
* if all was well, it will return 1 if it found locks, 0 otherwise. */
if (req == NULL && rc >= 0)
GOTO(out, rc);
if (rc < 0)
GOTO(out, rc = 0);
- rc = pnode_revalidate_finish(req, 1, it, pnode);
+ rc = pnode_revalidate_finish(req, it, pnode);
if (rc != 0) {
ll_intent_release(it);
GOTO(out, rc = 0);
pb->pb_ino = NULL;
} else {
llu_lookup_finish_locks(it, pnode);
- llu_i2info(pb->pb_ino)->lli_stale_flag = 0;
}
RETURN(rc);
}
struct inode *inode = NULL;
int rc;
+ /* libsysio require us generate inode right away if success.
+ * so if mds created new inode for us we need make sure it
+ * succeeded. thus for any error we can't delay to the
+ * llu_file_open() time. */
+ if (it_disposition(it, DISP_OPEN_CREATE) &&
+ it_open_error(DISP_OPEN_CREATE, it)) {
+ CDEBUG(D_INODE, "detect mds create error\n");
+ return it_open_error(DISP_OPEN_CREATE, it);
+ }
+ if (it_disposition(it, DISP_OPEN_OPEN) &&
+ it_open_error(DISP_OPEN_OPEN, it)) {
+ CDEBUG(D_INODE, "detect mds open error\n");
+ /* undo which did by md_intent_lock */
+ if (it_disposition(it, DISP_OPEN_CREATE) &&
+ !it_open_error(DISP_OPEN_CREATE, it)) {
+ LASSERT(request);
+ LASSERT(atomic_read(&request->rq_refcount) > 1);
+ CDEBUG(D_INODE, "dec a ref of req %p\n", request);
+ ptlrpc_req_finished(request);
+ }
+ return it_open_error(DISP_OPEN_OPEN, it);
+ }
+
/* NB 1 request reference will be taken away by ll_intent_lock()
* when I return
*/
- /* FIXME: for CREAT, libsysio require the inode must be generated here
- * currently here we don't know the whether the create is successful
- * or failed on mds. thus blinded return -EPERM in llu_iget(). need
- * a fix later.
- */
- if ((it->it_op & IT_CREAT) || !it_disposition(it, DISP_LOOKUP_NEG)) {
+ if (!it_disposition(it, DISP_LOOKUP_NEG) || (it->it_op & IT_CREAT)) {
struct lustre_md md;
struct llu_inode_info *lli;
+ struct intnl_stat *st;
ENTRY;
- rc = mdc_req2lustre_md(sbi->ll_mdc_exp, request, offset,
- sbi->ll_osc_exp, &md);
+ if (it_disposition(it, DISP_OPEN_CREATE))
+ ptlrpc_req_finished(request);
+
+ rc = md_get_lustre_md(sbi->ll_md_exp, request,
+ sbi->ll_dt_exp, sbi->ll_md_exp, &md);
if (rc)
RETURN(rc);
if (!inode || IS_ERR(inode)) {
/* free the lsm if we allocated one above */
if (md.lsm != NULL)
- obd_free_memmd(sbi->ll_osc_exp, &md.lsm);
+ obd_free_memmd(sbi->ll_dt_exp, &md.lsm);
RETURN(inode ? PTR_ERR(inode) : -ENOMEM);
} else if (md.lsm != NULL &&
llu_i2info(inode)->lli_smd != md.lsm) {
- obd_free_memmd(sbi->ll_osc_exp, &md.lsm);
+ obd_free_memmd(sbi->ll_dt_exp, &md.lsm);
}
lli = llu_i2info(inode);
+ st = llu_i2stat(inode);
/* If this is a stat, get the authoritative file size */
- if (it->it_op == IT_GETATTR && S_ISREG(lli->lli_st_mode) &&
+ if (it->it_op == IT_GETATTR && S_ISREG(st->st_mode) &&
lli->lli_smd != NULL) {
struct lov_stripe_md *lsm = lli->lli_smd;
- struct ost_lvb lvb;
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 = llu_glimpse_size(inode, &lvb);
+ rc = cl_glimpse_size(inode);
if (rc) {
I_RELE(inode);
RETURN(rc);
}
- lli->lli_st_size = lvb.lvb_size;
}
} else {
ENTRY;
struct inode *llu_inode_from_lock(struct ldlm_lock *lock)
{
struct inode *inode;
- l_lock(&lock->l_resource->lr_namespace->ns_lock);
+ lock_res_and_lock(lock);
if (lock->l_ast_data) {
inode = (struct inode *)lock->l_ast_data;
} else
inode = NULL;
- l_unlock(&lock->l_resource->lr_namespace->ns_lock);
+ unlock_res_and_lock(lock);
return inode;
}
static int llu_lookup_it(struct inode *parent, struct pnode *pnode,
struct lookup_intent *it, int flags)
{
- struct ll_fid pfid;
- struct ll_uctxt ctxt;
+ struct md_op_data op_data = {{ 0 }};
struct it_cb_data icbd;
struct ptlrpc_request *req = NULL;
struct lookup_intent lookup_it = { .it_op = IT_LOOKUP };
+ __u32 opc;
int rc;
ENTRY;
icbd.icbd_child = pnode;
icbd.icbd_parent = parent;
- icbd.icbd_child = pnode;
- ll_inode2fid(&pfid, parent);
- ll_i2uctxt(&ctxt, parent, NULL);
-
- rc = mdc_intent_lock(llu_i2mdcexp(parent), &ctxt, &pfid,
- pnode->p_base->pb_name.name,
- pnode->p_base->pb_name.len,
- NULL, 0, NULL, it, flags, &req,
- llu_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;
+ }
+
+ llu_prep_md_op_data(&op_data, parent, NULL,
+ pnode->p_base->pb_name.name,
+ pnode->p_base->pb_name.len, flags, opc);
+
+ rc = md_intent_lock(llu_i2mdexp(parent), &op_data, NULL, 0, it,
+ flags, &req, llu_md_blocking_ast,
+ LDLM_FL_CANCEL_ON_BLOCK);
if (rc < 0)
GOTO(out, rc);
-
- rc = lookup_it_finish(req, 1, it, &icbd);
+
+ rc = lookup_it_finish(req, DLM_REPLY_REC_OFF, it, &icbd);
if (rc != 0) {
ll_intent_release(it);
GOTO(out, rc);
if (it->it_flags & O_CREAT) {
it->it_op |= IT_CREAT;
it->it_create_mode = *((int*)intent->int_arg1);
+ /* bug 7278: libsysio hack. For O_EXCL, libsysio depends on
+ this lookup to return negative result, but then there is no
+ way to find out original intent in ll_iop_open(). So we just
+ clear O_EXCL from libsysio flags here to avoid checking
+ for negative result. O_EXCL will be enforced by MDS. */
+ *((int*)intent->int_arg2) &= ~O_EXCL;
}
if (intent->int_opmask & INT_GETATTR)
/* conform to kernel code, if only IT_LOOKUP was set, don't
* pass down it */
- if (!it->it_op || it->it_op == IT_LOOKUP) {
+ if (!it->it_op || it->it_op & IT_LOOKUP) {
OBD_FREE(it, sizeof(*it));
it = NULL;
}
int rc;
ENTRY;
+ liblustre_wait_event(0);
+
*inop = NULL;
/* the mount root inode have no name, so don't call
if (pnode->p_mount->mnt_root == pnode) {
struct inode *i = pnode->p_base->pb_ino;
*inop = i;
- return 0;
+ RETURN(0);
}
if (!pnode->p_base->pb_name.len)
if (llu_pb_revalidate(pnode, 0, it)) {
LASSERT(pnode->p_base->pb_ino);
*inop = pnode->p_base->pb_ino;
- RETURN(0);
+ GOTO(out, rc = 0);
}
rc = llu_lookup_it(pnode->p_parent->p_base->pb_ino, pnode, it, 0);
*inop = pnode->p_base->pb_ino;
}
+out:
+ if (it)
+ OBD_FREE(it, sizeof(*it));
+ liblustre_wait_event(0);
RETURN(rc);
}
-