/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
* vim:expandtab:shiftwidth=8:tabstop=8:
*
- * Copyright (C) 2001-2003 Cluster File Systems, Inc.
+ * GPL HEADER START
*
- * This file is part of the Lustre file system, http://www.lustre.org
- * Lustre is a trademark of Cluster File Systems, Inc.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
- * You may have signed or agreed to another license before downloading
- * this software. If so, you are bound by the terms and conditions
- * of that agreement, and the following does not apply to you. See the
- * LICENSE file included with this distribution for more information.
+ * 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.
*
- * If you did not agree to a different license, then this copy of Lustre
- * is open source 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 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).
*
- * In either case, 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
- * license text for more details.
+ * 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.
*/
#ifndef EXPORT_SYMTAB
# include <liblustre.h>
#endif
-#include <linux/lustre_acl.h>
+#include <lustre_acl.h>
#include <obd_class.h>
#include <lustre_dlm.h>
/* fid_res_name_eq() */
}
EXPORT_SYMBOL(it_clear_disposition);
-static int it_to_lock_mode(struct lookup_intent *it)
-{
- ENTRY;
-
- /* CREAT needs to be tested before open (both could be set) */
- if (it->it_op & IT_CREAT)
- return LCK_PW;
- else if (it->it_op & (IT_READDIR | IT_GETATTR | IT_OPEN | IT_LOOKUP))
- return LCK_PR;
-
- LBUG();
- RETURN(-EINVAL);
-}
-
int it_open_error(int phase, struct lookup_intent *it)
{
if (it_disposition(it, DISP_OPEN_OPEN)) {
EXPORT_SYMBOL(it_open_error);
/* this must be called on a lockh that is known to have a referenced lock */
-int mdc_set_lock_data(struct obd_export *exp, __u64 *lockh, void *data)
+int mdc_set_lock_data(struct obd_export *exp, __u64 *lockh, void *data,
+ __u32 *bits)
{
struct ldlm_lock *lock;
ENTRY;
+ if(bits)
+ *bits = 0;
+
if (!*lockh) {
EXIT;
RETURN(0);
}
#endif
lock->l_ast_data = data;
+ if (bits)
+ *bits = lock->l_policy_data.l_inodebits.bits;
+
unlock_res_and_lock(lock);
LDLM_LOCK_PUT(lock);
ldlm_policy_data_t *policy, ldlm_mode_t mode,
struct lustre_handle *lockh)
{
- struct ldlm_res_id res_id =
- { .name = {fid_seq(fid),
- fid_oid(fid),
- fid_ver(fid)} };
+ struct ldlm_res_id res_id;
ldlm_mode_t rc;
ENTRY;
+ fid_build_reg_res_name(fid, &res_id);
rc = ldlm_lock_match(class_exp2obd(exp)->obd_namespace, flags,
- &res_id, type, policy, mode, lockh);
+ &res_id, type, policy, mode, lockh, 0);
RETURN(rc);
}
ldlm_policy_data_t *policy,
ldlm_mode_t mode, int flags, void *opaque)
{
- struct ldlm_res_id res_id =
- { .name = {fid_seq(fid),
- fid_oid(fid),
- fid_ver(fid)} };
+ struct ldlm_res_id res_id;
struct obd_device *obd = class_exp2obd(exp);
int rc;
ENTRY;
+ fid_build_reg_res_name(fid, &res_id);
rc = ldlm_cli_cancel_unused_resource(obd->obd_namespace, &res_id,
policy, mode, flags, opaque);
RETURN(rc);
const struct lu_fid *fid,
ldlm_iterator_t it, void *data)
{
- struct ldlm_res_id res_id = { .name = {0} };
+ struct ldlm_res_id res_id;
ENTRY;
- res_id.name[0] = fid_seq(fid);
- res_id.name[1] = fid_oid(fid);
- res_id.name[2] = fid_ver(fid);
-
+ fid_build_reg_res_name(fid, &res_id);
ldlm_resource_iterate(class_exp2obd(exp)->obd_namespace,
&res_id, it, data);
* but this is incredibly unlikely, and questionable whether the client
* could do MDS recovery under OOM anyways... */
static void mdc_realloc_openmsg(struct ptlrpc_request *req,
- struct mdt_body *body)
+ struct mdt_body *body)
{
int rc;
struct ptlrpc_request *req;
struct obd_device *obddev = class_exp2obd(exp);
struct ldlm_intent *lit;
- int joinfile = !!((it->it_flags & O_JOIN_FILE) &&
- op_data->op_data);
+ int joinfile = !!((it->it_create_mode & M_JOIN_FILE) &&
+ op_data->op_data);
CFS_LIST_HEAD(cancels);
int count = 0;
int mode;
}
spin_lock(&req->rq_lock);
- req->rq_replay = 1;
+ req->rq_replay = req->rq_import->imp_replayable;
spin_unlock(&req->rq_lock);
/* pack the intent */
static struct ptlrpc_request *mdc_intent_getattr_pack(struct obd_export *exp,
struct lookup_intent *it,
- struct md_op_data *op_data)
+ struct md_op_data *op_data)
{
struct ptlrpc_request *req;
struct obd_device *obddev = class_exp2obd(exp);
RETURN(req);
}
-/* We always reserve enough space in the reply packet for a stripe MD, because
- * we don't know in advance the file type. */
-int mdc_enqueue(struct obd_export *exp, struct ldlm_enqueue_info *einfo,
- struct lookup_intent *it, struct md_op_data *op_data,
- struct lustre_handle *lockh, void *lmm, int lmmsize,
- int extra_lock_flags)
+static int mdc_finish_enqueue(struct obd_export *exp,
+ struct ptlrpc_request *req,
+ struct ldlm_enqueue_info *einfo,
+ struct lookup_intent *it,
+ struct lustre_handle *lockh,
+ int rc)
{
- struct obd_device *obddev = class_exp2obd(exp);
- struct ptlrpc_request *req;
- struct req_capsule *pill;
- struct ldlm_request *lockreq;
- struct ldlm_reply *lockrep;
- int flags = extra_lock_flags | LDLM_FL_HAS_INTENT;
- int rc;
- struct ldlm_res_id res_id =
- { .name = {fid_seq(&op_data->op_fid1),
- fid_oid(&op_data->op_fid1),
- fid_ver(&op_data->op_fid1)} };
- ldlm_policy_data_t policy = { .l_inodebits = { MDS_INODELOCK_LOOKUP } };
+ struct req_capsule *pill = &req->rq_pill;
+ struct ldlm_request *lockreq;
+ struct ldlm_reply *lockrep;
ENTRY;
- LASSERTF(einfo->ei_type == LDLM_IBITS,"lock type %d\n", einfo->ei_type);
-
- if (it->it_op & (IT_UNLINK | IT_GETATTR | IT_READDIR))
- policy.l_inodebits.bits = MDS_INODELOCK_UPDATE;
-
- if (it->it_op & IT_OPEN) {
- int joinfile = !!((it->it_flags & O_JOIN_FILE) &&
- op_data->op_data);
-
- req = mdc_intent_open_pack(exp, it, op_data, lmm, lmmsize,
- einfo->ei_cbdata);
- if (!joinfile) {
- policy.l_inodebits.bits = MDS_INODELOCK_UPDATE;
- einfo->ei_cbdata = NULL;
- lmm = NULL;
- } else
- it->it_flags &= ~O_JOIN_FILE;
- } else if (it->it_op & IT_UNLINK)
- req = mdc_intent_unlink_pack(exp, it, op_data);
- else if (it->it_op & (IT_GETATTR | IT_LOOKUP))
- req = mdc_intent_getattr_pack(exp, it, op_data);
- else if (it->it_op == IT_READDIR)
- req = ldlm_enqueue_pack(exp);
- else {
- LBUG();
- RETURN(-EINVAL);
- }
-
- if (IS_ERR(req))
- RETURN(PTR_ERR(req));
- pill = &req->rq_pill;
-
- /* It is important to obtain rpc_lock first (if applicable), so that
- * threads that are serialised with rpc_lock are not polluting our
- * rpcs in flight counter */
- mdc_get_rpc_lock(obddev->u.cli.cl_rpc_lock, it);
- mdc_enter_request(&obddev->u.cli);
- rc = ldlm_cli_enqueue(exp, &req, einfo, &res_id, &policy, &flags, NULL,
- 0, NULL, lockh, 0);
- mdc_exit_request(&obddev->u.cli);
- mdc_put_rpc_lock(obddev->u.cli.cl_rpc_lock, it);
-
+ LASSERT(rc >= 0);
/* Similarly, if we're going to replay this request, we don't want to
* actually get a lock, just perform the intent. */
if (req->rq_transno || req->rq_replay) {
einfo->ei_mode = 0;
memset(lockh, 0, sizeof(*lockh));
rc = 0;
- } else if (rc != 0) {
- CERROR("ldlm_cli_enqueue: %d\n", rc);
- LASSERTF(rc < 0, "rc %d\n", rc);
- mdc_clear_replay_flag(req, rc);
- ptlrpc_req_finished(req);
- RETURN(rc);
} else { /* rc = 0 */
struct ldlm_lock *lock = ldlm_handle2lock(lockh);
LASSERT(lock);
it->d.lustre.it_disposition = (int)lockrep->lock_policy_res1;
it->d.lustre.it_status = (int)lockrep->lock_policy_res2;
it->d.lustre.it_lock_mode = einfo->ei_mode;
+ it->d.lustre.it_lock_handle = lockh->cookie;
it->d.lustre.it_data = req;
if (it->d.lustre.it_status < 0 && req->rq_replay)
if (eadata == NULL)
RETURN(-EPROTO);
- if (body->valid & OBD_MD_FLMODEASIZE) {
- if (obddev->u.cli.cl_max_mds_easize <
- body->max_mdsize) {
- obddev->u.cli.cl_max_mds_easize =
- body->max_mdsize;
- CDEBUG(D_INFO, "maxeasize become %d\n",
- body->max_mdsize);
- }
- if (obddev->u.cli.cl_max_mds_cookiesize <
- body->max_cookiesize) {
- obddev->u.cli.cl_max_mds_cookiesize =
- body->max_cookiesize;
- CDEBUG(D_INFO, "cookiesize become %d\n",
- body->max_cookiesize);
- }
- }
-
/*
* We save the reply LOV EA in case we have to replay a
* create for recovery. If we didn't allocate a large
* (for example error one).
*/
if ((it->it_op & IT_OPEN) && req->rq_replay) {
+ void *lmm;
if (req_capsule_get_size(pill, &RMF_EADATA,
RCL_CLIENT) <
- body->eadatasize) {
+ body->eadatasize)
mdc_realloc_openmsg(req, body);
- req_capsule_set_size(pill, &RMF_EADATA,
- RCL_CLIENT,
- body->eadatasize);
- }
+ else
+ req_capsule_shrink(pill, &RMF_EADATA,
+ body->eadatasize,
+ RCL_CLIENT);
+
+ req_capsule_set_size(pill, &RMF_EADATA,
+ RCL_CLIENT,
+ body->eadatasize);
+
lmm = req_capsule_client_get(pill, &RMF_EADATA);
if (lmm)
memcpy(lmm, eadata, body->eadatasize);
RETURN(rc);
}
+
+/* We always reserve enough space in the reply packet for a stripe MD, because
+ * we don't know in advance the file type. */
+int mdc_enqueue(struct obd_export *exp, struct ldlm_enqueue_info *einfo,
+ struct lookup_intent *it, struct md_op_data *op_data,
+ struct lustre_handle *lockh, void *lmm, int lmmsize,
+ struct ptlrpc_request **reqp, int extra_lock_flags)
+{
+ struct obd_device *obddev = class_exp2obd(exp);
+ struct ptlrpc_request *req = NULL;
+ struct req_capsule *pill;
+ int flags = extra_lock_flags;
+ int rc;
+ struct ldlm_res_id res_id;
+ ldlm_policy_data_t policy = { .l_inodebits = { MDS_INODELOCK_LOOKUP } };
+ ENTRY;
+
+ LASSERTF(!it || einfo->ei_type == LDLM_IBITS, "lock type %d\n",
+ einfo->ei_type);
+
+ fid_build_reg_res_name(&op_data->op_fid1, &res_id);
+
+ if (it)
+ flags |= LDLM_FL_HAS_INTENT;
+ if (it && it->it_op & (IT_UNLINK | IT_GETATTR | IT_READDIR))
+ policy.l_inodebits.bits = MDS_INODELOCK_UPDATE;
+
+ if (reqp)
+ req = *reqp;
+
+ if (!it) {
+ /* The only way right now is FLOCK, in this case we hide flock
+ policy as lmm, but lmmsize is 0 */
+ LASSERT(lmm && lmmsize == 0);
+ LASSERTF(einfo->ei_type == LDLM_FLOCK, "lock type %d\n",
+ einfo->ei_type);
+ policy = *(ldlm_policy_data_t *)lmm;
+ res_id.name[3] = LDLM_FLOCK;
+ } else if (it->it_op & IT_OPEN) {
+ int joinfile = !!((it->it_create_mode & M_JOIN_FILE) &&
+ op_data->op_data);
+
+ req = mdc_intent_open_pack(exp, it, op_data, lmm, lmmsize,
+ einfo->ei_cbdata);
+ if (!joinfile) {
+ policy.l_inodebits.bits = MDS_INODELOCK_UPDATE;
+ einfo->ei_cbdata = NULL;
+ lmm = NULL;
+ } else
+ it->it_create_mode &= ~M_JOIN_FILE;
+ } else if (it->it_op & IT_UNLINK)
+ req = mdc_intent_unlink_pack(exp, it, op_data);
+ else if (it->it_op & (IT_GETATTR | IT_LOOKUP))
+ req = mdc_intent_getattr_pack(exp, it, op_data);
+ else if (it->it_op == IT_READDIR)
+ req = ldlm_enqueue_pack(exp);
+ else {
+ LBUG();
+ RETURN(-EINVAL);
+ }
+
+ if (IS_ERR(req))
+ RETURN(PTR_ERR(req));
+ pill = &req->rq_pill;
+
+ /* It is important to obtain rpc_lock first (if applicable), so that
+ * threads that are serialised with rpc_lock are not polluting our
+ * rpcs in flight counter. We do not do flock request limiting, though*/
+ if (it) {
+ mdc_get_rpc_lock(obddev->u.cli.cl_rpc_lock, it);
+ mdc_enter_request(&obddev->u.cli);
+ }
+ rc = ldlm_cli_enqueue(exp, &req, einfo, &res_id, &policy, &flags, NULL,
+ 0, NULL, lockh, 0);
+ if (reqp)
+ *reqp = req;
+
+ if (it) {
+ mdc_exit_request(&obddev->u.cli);
+ mdc_put_rpc_lock(obddev->u.cli.cl_rpc_lock, it);
+ }
+ if (!it) {
+ /* For flock requests we immediatelly return without further
+ delay and let caller deal with the rest, since rest of
+ this function metadata processing makes no sense for flock
+ requests anyway */
+ RETURN(rc);
+ }
+
+ if (rc < 0) {
+ CERROR("ldlm_cli_enqueue: %d\n", rc);
+ mdc_clear_replay_flag(req, rc);
+ ptlrpc_req_finished(req);
+ RETURN(rc);
+ }
+ rc = mdc_finish_enqueue(exp, req, einfo, it, lockh, rc);
+
+ RETURN(rc);
+}
+
+static int mdc_finish_intent_lock(struct obd_export *exp,
+ struct ptlrpc_request *request,
+ struct md_op_data *op_data,
+ struct lookup_intent *it,
+ struct lustre_handle *lockh)
+{
+ struct lustre_handle old_lock;
+ struct mdt_body *mdt_body;
+ struct ldlm_lock *lock;
+ int rc;
+
+
+ LASSERT(request != NULL);
+ LASSERT(request != LP_POISON);
+ LASSERT(request->rq_repmsg != LP_POISON);
+
+ if (!it_disposition(it, DISP_IT_EXECD)) {
+ /* The server failed before it even started executing the
+ * intent, i.e. because it couldn't unpack the request. */
+ LASSERT(it->d.lustre.it_status != 0);
+ RETURN(it->d.lustre.it_status);
+ }
+ rc = it_open_error(DISP_IT_EXECD, it);
+ if (rc)
+ RETURN(rc);
+
+ mdt_body = req_capsule_server_get(&request->rq_pill, &RMF_MDT_BODY);
+ LASSERT(mdt_body != NULL); /* mdc_enqueue checked */
+
+ /* If we were revalidating a fid/name pair, mark the intent in
+ * case we fail and get called again from lookup */
+ if (fid_is_sane(&op_data->op_fid2) &&
+ it->it_create_mode & M_CHECK_STALE &&
+ it->it_op != IT_GETATTR) {
+ it_set_disposition(it, DISP_ENQ_COMPLETE);
+
+ /* Also: did we find the same inode? */
+ /* sever can return one of two fids:
+ * op_fid2 - new allocated fid - if file is created.
+ * op_fid3 - existent fid - if file only open.
+ * op_fid3 is saved in lmv_intent_open */
+ if ((!lu_fid_eq(&op_data->op_fid2, &mdt_body->fid1)) &&
+ (!lu_fid_eq(&op_data->op_fid3, &mdt_body->fid1))) {
+ CDEBUG(D_DENTRY, "Found stale data "DFID"("DFID")/"DFID
+ "\n", PFID(&op_data->op_fid2),
+ PFID(&op_data->op_fid2), PFID(&mdt_body->fid1));
+ RETURN(-ESTALE);
+ }
+ }
+
+ rc = it_open_error(DISP_LOOKUP_EXECD, it);
+ if (rc)
+ RETURN(rc);
+
+ /* keep requests around for the multiple phases of the call
+ * this shows the DISP_XX must guarantee we make it into the call
+ */
+ if (!it_disposition(it, DISP_ENQ_CREATE_REF) &&
+ it_disposition(it, DISP_OPEN_CREATE) &&
+ !it_open_error(DISP_OPEN_CREATE, it)) {
+ it_set_disposition(it, DISP_ENQ_CREATE_REF);
+ ptlrpc_request_addref(request); /* balanced in ll_create_node */
+ }
+ if (!it_disposition(it, DISP_ENQ_OPEN_REF) &&
+ it_disposition(it, DISP_OPEN_OPEN) &&
+ !it_open_error(DISP_OPEN_OPEN, it)) {
+ it_set_disposition(it, DISP_ENQ_OPEN_REF);
+ ptlrpc_request_addref(request); /* balanced in ll_file_open */
+ /* BUG 11546 - eviction in the middle of open rpc processing */
+ OBD_FAIL_TIMEOUT(OBD_FAIL_MDC_ENQUEUE_PAUSE, obd_timeout);
+ }
+
+ if (it->it_op & IT_CREAT) {
+ /* XXX this belongs in ll_create_it */
+ } else if (it->it_op == IT_OPEN) {
+ LASSERT(!it_disposition(it, DISP_OPEN_CREATE));
+ } else {
+ LASSERT(it->it_op & (IT_GETATTR | IT_LOOKUP));
+ }
+
+ /* If we already have a matching lock, then cancel the new
+ * one. We have to set the data here instead of in
+ * mdc_enqueue, because we need to use the child's inode as
+ * the l_ast_data to match, and that's not available until
+ * intent_finish has performed the iget().) */
+ lock = ldlm_handle2lock(lockh);
+ if (lock) {
+ ldlm_policy_data_t policy = lock->l_policy_data;
+ LDLM_DEBUG(lock, "matching against this");
+
+ LASSERTF(fid_res_name_eq(&mdt_body->fid1,
+ &lock->l_resource->lr_name),
+ "Lock res_id: %lu/%lu/%lu, fid: %lu/%lu/%lu.\n",
+ (unsigned long)lock->l_resource->lr_name.name[0],
+ (unsigned long)lock->l_resource->lr_name.name[1],
+ (unsigned long)lock->l_resource->lr_name.name[2],
+ (unsigned long)fid_seq(&mdt_body->fid1),
+ (unsigned long)fid_oid(&mdt_body->fid1),
+ (unsigned long)fid_ver(&mdt_body->fid1));
+ LDLM_LOCK_PUT(lock);
+
+ memcpy(&old_lock, lockh, sizeof(*lockh));
+ if (ldlm_lock_match(NULL, LDLM_FL_BLOCK_GRANTED, NULL,
+ LDLM_IBITS, &policy, LCK_NL, &old_lock, 0)) {
+ ldlm_lock_decref_and_cancel(lockh,
+ it->d.lustre.it_lock_mode);
+ memcpy(lockh, &old_lock, sizeof(old_lock));
+ it->d.lustre.it_lock_handle = lockh->cookie;
+ }
+ }
+ CDEBUG(D_DENTRY,"D_IT dentry %.*s intent: %s status %d disp %x rc %d\n",
+ op_data->op_namelen, op_data->op_name, ldlm_it2str(it->it_op),
+ it->d.lustre.it_status, it->d.lustre.it_disposition, rc);
+ RETURN(rc);
+}
+
/*
* This long block is all about fixing up the lock and request state
* so that it is correct as of the moment _before_ the operation was
ldlm_blocking_callback cb_blocking,
int extra_lock_flags)
{
- struct ptlrpc_request *request;
- struct lustre_handle old_lock;
struct lustre_handle lockh;
- struct mdt_body *mdt_body;
- struct ldlm_lock *lock;
int rc = 0;
ENTRY;
LASSERT(it);
PFID(&op_data->op_fid1), ldlm_it2str(it->it_op),
it->it_flags);
+ lockh.cookie = 0;
if (fid_is_sane(&op_data->op_fid2) &&
(it->it_op & (IT_LOOKUP | IT_GETATTR))) {
/* We could just return 1 immediately, but since we should only
&op_data->op_fid2, LDLM_IBITS, &policy,
LCK_CR|LCK_CW|LCK_PR|LCK_PW, &lockh);
if (mode) {
- memcpy(&it->d.lustre.it_lock_handle, &lockh,
- sizeof(lockh));
+ it->d.lustre.it_lock_handle = lockh.cookie;
it->d.lustre.it_lock_mode = mode;
}
if (!it_disposition(it, DISP_ENQ_COMPLETE)) {
struct ldlm_enqueue_info einfo =
{ LDLM_IBITS, it_to_lock_mode(it), cb_blocking,
- ldlm_completion_ast, NULL, NULL };
+ ldlm_completion_ast, NULL, NULL, NULL };
/* For case if upper layer did not alloc fid, do it now. */
if (!fid_is_sane(&op_data->op_fid2) && it->it_op & IT_CREAT) {
}
}
rc = mdc_enqueue(exp, &einfo, it, op_data, &lockh,
- lmm, lmmsize, extra_lock_flags);
+ lmm, lmmsize, NULL, extra_lock_flags);
if (rc < 0)
RETURN(rc);
- memcpy(&it->d.lustre.it_lock_handle, &lockh, sizeof(lockh));
} else if (!fid_is_sane(&op_data->op_fid2) ||
- !(it->it_flags & O_CHECK_STALE)) {
+ !(it->it_create_mode & M_CHECK_STALE)) {
/* DISP_ENQ_COMPLETE set means there is extra reference on
* request referenced from this intent, saved for subsequent
* lookup. This path is executed when we proceed to this
* lookup, so we clear DISP_ENQ_COMPLETE */
it_clear_disposition(it, DISP_ENQ_COMPLETE);
}
- request = *reqp = it->d.lustre.it_data;
- LASSERT(request != NULL);
- LASSERT(request != LP_POISON);
- LASSERT(request->rq_repmsg != LP_POISON);
+ *reqp = it->d.lustre.it_data;
+ rc = mdc_finish_intent_lock(exp, *reqp, op_data, it, &lockh);
+ RETURN(rc);
+}
- if (!it_disposition(it, DISP_IT_EXECD)) {
- /* The server failed before it even started executing the
- * intent, i.e. because it couldn't unpack the request. */
- LASSERT(it->d.lustre.it_status != 0);
- RETURN(it->d.lustre.it_status);
- }
- rc = it_open_error(DISP_IT_EXECD, it);
- if (rc)
- RETURN(rc);
+static int mdc_intent_getattr_async_interpret(const struct lu_env *env,
+ struct ptlrpc_request *req,
+ void *unused, int rc)
+{
+ struct obd_export *exp = req->rq_async_args.pointer_arg[0];
+ struct md_enqueue_info *minfo = req->rq_async_args.pointer_arg[1];
+ struct ldlm_enqueue_info *einfo = req->rq_async_args.pointer_arg[2];
+ struct lookup_intent *it;
+ struct lustre_handle *lockh;
+ struct obd_device *obddev;
+ int flags = LDLM_FL_HAS_INTENT;
+ ENTRY;
- mdt_body = req_capsule_server_get(&request->rq_pill, &RMF_MDT_BODY);
- LASSERT(mdt_body != NULL); /* mdc_enqueue checked */
+ it = &minfo->mi_it;
+ lockh = &minfo->mi_lockh;
- /* If we were revalidating a fid/name pair, mark the intent in
- * case we fail and get called again from lookup */
- if (fid_is_sane(&op_data->op_fid2) &&
- (it->it_flags & O_CHECK_STALE) &&
- it->it_op != IT_GETATTR) {
- it_set_disposition(it, DISP_ENQ_COMPLETE);
+ obddev = class_exp2obd(exp);
- /* Also: did we find the same inode? */
- /* sever can return one of two fids:
- * op_fid2 - new allocated fid - if file is created.
- * op_fid3 - existent fid - if file only open.
- * op_fid3 is saved in lmv_intent_open */
- if ((!lu_fid_eq(&op_data->op_fid2, &mdt_body->fid1)) &&
- (!lu_fid_eq(&op_data->op_fid3, &mdt_body->fid1))) {
- CDEBUG(D_DENTRY, "Found stale data "DFID"("DFID")/"DFID
- "\n", PFID(&op_data->op_fid2),
- PFID(&op_data->op_fid2), PFID(&mdt_body->fid1));
- RETURN(-ESTALE);
- }
+ mdc_exit_request(&obddev->u.cli);
+ if (OBD_FAIL_CHECK(OBD_FAIL_MDC_GETATTR_ENQUEUE))
+ rc = -ETIMEDOUT;
+
+ rc = ldlm_cli_enqueue_fini(exp, req, einfo->ei_type, 1, einfo->ei_mode,
+ &flags, NULL, 0, NULL, lockh, rc);
+ if (rc < 0) {
+ CERROR("ldlm_cli_enqueue_fini: %d\n", rc);
+ mdc_clear_replay_flag(req, rc);
+ GOTO(out, rc);
}
- rc = it_open_error(DISP_LOOKUP_EXECD, it);
+ rc = mdc_finish_enqueue(exp, req, einfo, it, lockh, rc);
if (rc)
- RETURN(rc);
+ GOTO(out, rc);
- /* keep requests around for the multiple phases of the call
- * this shows the DISP_XX must guarantee we make it into the call
- */
- if (!it_disposition(it, DISP_ENQ_CREATE_REF) &&
- it_disposition(it, DISP_OPEN_CREATE) &&
- !it_open_error(DISP_OPEN_CREATE, it)) {
- it_set_disposition(it, DISP_ENQ_CREATE_REF);
- ptlrpc_request_addref(request); /* balanced in ll_create_node */
- }
- if (!it_disposition(it, DISP_ENQ_OPEN_REF) &&
- it_disposition(it, DISP_OPEN_OPEN) &&
- !it_open_error(DISP_OPEN_OPEN, it)) {
- it_set_disposition(it, DISP_ENQ_OPEN_REF);
- ptlrpc_request_addref(request); /* balanced in ll_file_open */
- /* BUG 11546 - eviction in the middle of open rpc processing */
- OBD_FAIL_TIMEOUT(OBD_FAIL_MDC_ENQUEUE_PAUSE, obd_timeout);
- }
+ rc = mdc_finish_intent_lock(exp, req, &minfo->mi_data, it, lockh);
+ EXIT;
- if (it->it_op & IT_CREAT) {
- /* XXX this belongs in ll_create_it */
- } else if (it->it_op == IT_OPEN) {
- LASSERT(!it_disposition(it, DISP_OPEN_CREATE));
- } else {
- LASSERT(it->it_op & (IT_GETATTR | IT_LOOKUP));
+out:
+ OBD_FREE_PTR(einfo);
+ minfo->mi_cb(req, minfo, rc);
+ return 0;
+}
+
+int mdc_intent_getattr_async(struct obd_export *exp,
+ struct md_enqueue_info *minfo,
+ struct ldlm_enqueue_info *einfo)
+{
+ struct md_op_data *op_data = &minfo->mi_data;
+ struct lookup_intent *it = &minfo->mi_it;
+ struct ptlrpc_request *req;
+ struct obd_device *obddev = class_exp2obd(exp);
+ struct ldlm_res_id res_id;
+ ldlm_policy_data_t policy = {
+ .l_inodebits = { MDS_INODELOCK_LOOKUP }
+ };
+ int rc;
+ int flags = LDLM_FL_HAS_INTENT;
+ ENTRY;
+
+ CDEBUG(D_DLMTRACE,"name: %.*s in inode "DFID", intent: %s flags %#o\n",
+ op_data->op_namelen, op_data->op_name, PFID(&op_data->op_fid1),
+ ldlm_it2str(it->it_op), it->it_flags);
+
+ fid_build_reg_res_name(&op_data->op_fid1, &res_id);
+ req = mdc_intent_getattr_pack(exp, it, op_data);
+ if (!req)
+ RETURN(-ENOMEM);
+
+ mdc_enter_request(&obddev->u.cli);
+ rc = ldlm_cli_enqueue(exp, &req, einfo, &res_id, &policy, &flags, NULL,
+ 0, NULL, &minfo->mi_lockh, 1);
+ if (rc < 0) {
+ mdc_exit_request(&obddev->u.cli);
+ RETURN(rc);
}
- /* If we already have a matching lock, then cancel the new
- * one. We have to set the data here instead of in
- * mdc_enqueue, because we need to use the child's inode as
- * the l_ast_data to match, and that's not available until
- * intent_finish has performed the iget().) */
- lock = ldlm_handle2lock(&lockh);
- if (lock) {
- ldlm_policy_data_t policy = lock->l_policy_data;
- LDLM_DEBUG(lock, "matching against this");
+ req->rq_async_args.pointer_arg[0] = exp;
+ req->rq_async_args.pointer_arg[1] = minfo;
+ req->rq_async_args.pointer_arg[2] = einfo;
+ req->rq_interpret_reply = mdc_intent_getattr_async_interpret;
+ ptlrpcd_add_req(req, PSCOPE_OTHER);
- LASSERTF(fid_res_name_eq(&mdt_body->fid1,
- &lock->l_resource->lr_name),
- "Lock res_id: %lu/%lu/%lu, fid: %lu/%lu/%lu.\n",
- (unsigned long)lock->l_resource->lr_name.name[0],
- (unsigned long)lock->l_resource->lr_name.name[1],
- (unsigned long)lock->l_resource->lr_name.name[2],
- (unsigned long)fid_seq(&mdt_body->fid1),
- (unsigned long)fid_oid(&mdt_body->fid1),
- (unsigned long)fid_ver(&mdt_body->fid1));
- LDLM_LOCK_PUT(lock);
+ RETURN(0);
+}
- memcpy(&old_lock, &lockh, sizeof(lockh));
- if (ldlm_lock_match(NULL, LDLM_FL_BLOCK_GRANTED, NULL,
- LDLM_IBITS, &policy, LCK_NL, &old_lock)) {
- ldlm_lock_decref_and_cancel(&lockh,
- it->d.lustre.it_lock_mode);
- memcpy(&lockh, &old_lock, sizeof(old_lock));
- memcpy(&it->d.lustre.it_lock_handle, &lockh,
- sizeof(lockh));
- }
+int mdc_revalidate_lock(struct obd_export *exp,
+ struct lookup_intent *it,
+ struct lu_fid *fid)
+{
+ /* We could just return 1 immediately, but since we should only
+ * be called in revalidate_it if we already have a lock, let's
+ * verify that. */
+ struct ldlm_res_id res_id;
+ struct lustre_handle lockh;
+ ldlm_policy_data_t policy;
+ ldlm_mode_t mode;
+ ENTRY;
+
+ fid_build_reg_res_name(fid, &res_id);
+ /* As not all attributes are kept under update lock, e.g.
+ owner/group/acls are under lookup lock, we need both
+ ibits for GETATTR. */
+ policy.l_inodebits.bits = (it->it_op == IT_GETATTR) ?
+ MDS_INODELOCK_UPDATE | MDS_INODELOCK_LOOKUP :
+ MDS_INODELOCK_LOOKUP;
+
+ mode = ldlm_lock_match(exp->exp_obd->obd_namespace,
+ LDLM_FL_BLOCK_GRANTED, &res_id, LDLM_IBITS,
+ &policy, LCK_CR|LCK_CW|LCK_PR|LCK_PW, &lockh, 0);
+ if (mode) {
+ it->d.lustre.it_lock_handle = lockh.cookie;
+ it->d.lustre.it_lock_mode = mode;
}
- CDEBUG(D_DENTRY,"D_IT dentry %.*s intent: %s status %d disp %x rc %d\n",
- op_data->op_namelen, op_data->op_name, ldlm_it2str(it->it_op),
- it->d.lustre.it_status, it->d.lustre.it_disposition, rc);
- RETURN(rc);
+ RETURN(!!mode);
}
git://git.whamcloud.com / fs / lustre-release.git / blobdiff