-/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
- * vim:expandtab:shiftwidth=8:tabstop=8:
- *
+/*
* GPL HEADER START
*
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
* Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
*
- * Copyright (c) 2011, 2012, Whamcloud, Inc.
+ * Copyright (c) 2011, 2013, Intel Corporation.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
* Author: Yury Umanets <umka@clusterfs.com>
*/
-#ifndef EXPORT_SYMTAB
-# define EXPORT_SYMTAB
-#endif
#define DEBUG_SUBSYSTEM S_FID
#ifdef __KERNEL__
#include <obd.h>
#include <obd_class.h>
-#include <dt_object.h>
-#include <md_object.h>
#include <obd_support.h>
-#include <lustre_req_layout.h>
#include <lustre_fid.h>
/* mdc RPC locks */
#include <lustre_mdc.h>
struct lu_seq_range *output, __u32 opc,
const char *opcname)
{
- struct obd_export *exp = seq->lcs_exp;
- struct ptlrpc_request *req;
- struct lu_seq_range *out, *in;
- __u32 *op;
- int rc;
- ENTRY;
-
- req = ptlrpc_request_alloc_pack(class_exp2cliimp(exp), &RQF_SEQ_QUERY,
- LUSTRE_MDS_VERSION, SEQ_QUERY);
- if (req == NULL)
- RETURN(-ENOMEM);
-
- /* Init operation code */
- op = req_capsule_client_get(&req->rq_pill, &RMF_SEQ_OPC);
- *op = opc;
-
- /* Zero out input range, this is not recovery yet. */
- in = req_capsule_client_get(&req->rq_pill, &RMF_SEQ_RANGE);
- range_init(in);
-
- ptlrpc_request_set_replen(req);
-
- if (seq->lcs_type == LUSTRE_SEQ_METADATA) {
- req->rq_request_portal = SEQ_METADATA_PORTAL;
- in->lsr_flags = LU_SEQ_RANGE_MDT;
- } else {
- LASSERTF(seq->lcs_type == LUSTRE_SEQ_DATA,
- "unknown lcs_type %u\n", seq->lcs_type);
- req->rq_request_portal = SEQ_DATA_PORTAL;
- in->lsr_flags = LU_SEQ_RANGE_OST;
- }
-
- if (opc == SEQ_ALLOC_SUPER) {
- /* Update index field of *in, it is required for
- * FLD update on super sequence allocator node. */
- in->lsr_index = seq->lcs_space.lsr_index;
- req->rq_request_portal = SEQ_CONTROLLER_PORTAL;
- } else {
- LASSERTF(opc == SEQ_ALLOC_META,
- "unknown opcode %u\n, opc", opc);
- }
-
- ptlrpc_at_set_req_timeout(req);
-
- mdc_get_rpc_lock(exp->exp_obd->u.cli.cl_rpc_lock, NULL);
- rc = ptlrpc_queue_wait(req);
- mdc_put_rpc_lock(exp->exp_obd->u.cli.cl_rpc_lock, NULL);
-
- if (rc)
- GOTO(out_req, rc);
-
- out = req_capsule_server_get(&req->rq_pill, &RMF_SEQ_RANGE);
- *output = *out;
-
- if (!range_is_sane(output)) {
- CERROR("%s: Invalid range received from server: "
- DRANGE"\n", seq->lcs_name, PRANGE(output));
- GOTO(out_req, rc = -EINVAL);
- }
-
- if (range_is_exhausted(output)) {
- CERROR("%s: Range received from server is exhausted: "
- DRANGE"]\n", seq->lcs_name, PRANGE(output));
- GOTO(out_req, rc = -EINVAL);
- }
-
- CDEBUG(D_INFO, "%s: Allocated %s-sequence "DRANGE"]\n",
- seq->lcs_name, opcname, PRANGE(output));
-
- EXIT;
+ struct obd_export *exp = seq->lcs_exp;
+ struct ptlrpc_request *req;
+ struct lu_seq_range *out, *in;
+ __u32 *op;
+ unsigned int debug_mask;
+ int rc;
+ ENTRY;
+
+ req = ptlrpc_request_alloc_pack(class_exp2cliimp(exp), &RQF_SEQ_QUERY,
+ LUSTRE_MDS_VERSION, SEQ_QUERY);
+ if (req == NULL)
+ RETURN(-ENOMEM);
+
+ /* Init operation code */
+ op = req_capsule_client_get(&req->rq_pill, &RMF_SEQ_OPC);
+ *op = opc;
+
+ /* Zero out input range, this is not recovery yet. */
+ in = req_capsule_client_get(&req->rq_pill, &RMF_SEQ_RANGE);
+ range_init(in);
+
+ ptlrpc_request_set_replen(req);
+
+ in->lsr_index = seq->lcs_space.lsr_index;
+ if (seq->lcs_type == LUSTRE_SEQ_METADATA)
+ fld_range_set_mdt(in);
+ else
+ fld_range_set_ost(in);
+
+ if (opc == SEQ_ALLOC_SUPER) {
+ req->rq_request_portal = SEQ_CONTROLLER_PORTAL;
+ req->rq_reply_portal = MDC_REPLY_PORTAL;
+ /* During allocating super sequence for data object,
+ * the current thread might hold the export of MDT0(MDT0
+ * precreating objects on this OST), and it will send the
+ * request to MDT0 here, so we can not keep resending the
+ * request here, otherwise if MDT0 is failed(umounted),
+ * it can not release the export of MDT0 */
+ if (seq->lcs_type == LUSTRE_SEQ_DATA)
+ req->rq_no_delay = req->rq_no_resend = 1;
+ debug_mask = D_CONSOLE;
+ } else {
+ if (seq->lcs_type == LUSTRE_SEQ_METADATA)
+ req->rq_request_portal = SEQ_METADATA_PORTAL;
+ else
+ req->rq_request_portal = SEQ_DATA_PORTAL;
+ debug_mask = D_INFO;
+ }
+
+ ptlrpc_at_set_req_timeout(req);
+
+ if (seq->lcs_type == LUSTRE_SEQ_METADATA)
+ mdc_get_rpc_lock(exp->exp_obd->u.cli.cl_rpc_lock, NULL);
+ rc = ptlrpc_queue_wait(req);
+ if (seq->lcs_type == LUSTRE_SEQ_METADATA)
+ mdc_put_rpc_lock(exp->exp_obd->u.cli.cl_rpc_lock, NULL);
+ if (rc)
+ GOTO(out_req, rc);
+
+ out = req_capsule_server_get(&req->rq_pill, &RMF_SEQ_RANGE);
+ *output = *out;
+
+ if (!range_is_sane(output)) {
+ CERROR("%s: Invalid range received from server: "
+ DRANGE"\n", seq->lcs_name, PRANGE(output));
+ GOTO(out_req, rc = -EINVAL);
+ }
+
+ if (range_is_exhausted(output)) {
+ CERROR("%s: Range received from server is exhausted: "
+ DRANGE"]\n", seq->lcs_name, PRANGE(output));
+ GOTO(out_req, rc = -EINVAL);
+ }
+
+ CDEBUG_LIMIT(debug_mask, "%s: Allocated %s-sequence "DRANGE"]\n",
+ seq->lcs_name, opcname, PRANGE(output));
+
+ EXIT;
out_req:
- ptlrpc_req_finished(req);
- return rc;
+ ptlrpc_req_finished(req);
+ return rc;
}
/* Request sequence-controller node to allocate new super-sequence. */
int rc;
ENTRY;
- cfs_down(&seq->lcs_sem);
+ mutex_lock(&seq->lcs_mutex);
-#ifdef __KERNEL__
if (seq->lcs_srv) {
+#ifdef HAVE_SEQ_SERVER
LASSERT(env != NULL);
rc = seq_server_alloc_super(seq->lcs_srv, &seq->lcs_space,
env);
- } else {
+#else
+ rc = 0;
#endif
- rc = seq_client_rpc(seq, &seq->lcs_space,
+ } else {
+ /* Check whether the connection to seq controller has been
+ * setup (lcs_exp != NULL) */
+ if (seq->lcs_exp == NULL) {
+ mutex_unlock(&seq->lcs_mutex);
+ RETURN(-EINPROGRESS);
+ }
+
+ rc = seq_client_rpc(seq, &seq->lcs_space,
SEQ_ALLOC_SUPER, "super");
-#ifdef __KERNEL__
}
-#endif
- cfs_up(&seq->lcs_sem);
+ mutex_unlock(&seq->lcs_mutex);
RETURN(rc);
}
int rc;
ENTRY;
-#ifdef __KERNEL__
if (seq->lcs_srv) {
+#ifdef HAVE_SEQ_SERVER
LASSERT(env != NULL);
rc = seq_server_alloc_meta(seq->lcs_srv, &seq->lcs_space, env);
- } else {
+#else
+ rc = 0;
#endif
- rc = seq_client_rpc(seq, &seq->lcs_space,
- SEQ_ALLOC_META, "meta");
-#ifdef __KERNEL__
+ } else {
+ do {
+ /* If meta server return -EINPROGRESS or EAGAIN,
+ * it means meta server might not be ready to
+ * allocate super sequence from sequence controller
+ * (MDT0)yet */
+ rc = seq_client_rpc(seq, &seq->lcs_space,
+ SEQ_ALLOC_META, "meta");
+ } while (rc == -EINPROGRESS || rc == -EAGAIN);
}
-#endif
+
RETURN(rc);
}
if (seq->lcs_update) {
cfs_waitq_add(&seq->lcs_waitq, link);
cfs_set_current_state(CFS_TASK_UNINT);
- cfs_up(&seq->lcs_sem);
+ mutex_unlock(&seq->lcs_mutex);
cfs_waitq_wait(link, CFS_TASK_UNINT);
- cfs_down(&seq->lcs_sem);
+ mutex_lock(&seq->lcs_mutex);
cfs_waitq_del(&seq->lcs_waitq, link);
cfs_set_current_state(CFS_TASK_RUNNING);
return -EAGAIN;
}
++seq->lcs_update;
- cfs_up(&seq->lcs_sem);
+ mutex_unlock(&seq->lcs_mutex);
return 0;
}
static void seq_fid_alloc_fini(struct lu_client_seq *seq)
{
LASSERT(seq->lcs_update == 1);
- cfs_down(&seq->lcs_sem);
+ mutex_lock(&seq->lcs_mutex);
--seq->lcs_update;
cfs_waitq_signal(&seq->lcs_waitq);
}
-/* Allocate the whole seq to the caller*/
+/**
+ * Allocate the whole seq to the caller.
+ **/
int seq_client_get_seq(const struct lu_env *env,
struct lu_client_seq *seq, seqno_t *seqnr)
{
int rc;
LASSERT(seqnr != NULL);
- cfs_down(&seq->lcs_sem);
+ mutex_lock(&seq->lcs_mutex);
cfs_waitlink_init(&link);
while (1) {
CERROR("%s: Can't allocate new sequence, "
"rc %d\n", seq->lcs_name, rc);
seq_fid_alloc_fini(seq);
- cfs_up(&seq->lcs_sem);
+ mutex_unlock(&seq->lcs_mutex);
return rc;
}
CDEBUG(D_INFO, "%s: allocate sequence "
"[0x%16.16"LPF64"x]\n", seq->lcs_name, *seqnr);
- /*Since the caller require the whole seq,
- *so marked this seq to be used*/
- seq->lcs_fid.f_oid = LUSTRE_SEQ_MAX_WIDTH;
- seq->lcs_fid.f_seq = *seqnr;
- seq->lcs_fid.f_ver = 0;
+ /* Since the caller require the whole seq,
+ * so marked this seq to be used */
+ if (seq->lcs_type == LUSTRE_SEQ_METADATA)
+ seq->lcs_fid.f_oid = LUSTRE_METADATA_SEQ_MAX_WIDTH;
+ else
+ seq->lcs_fid.f_oid = LUSTRE_DATA_SEQ_MAX_WIDTH;
+ seq->lcs_fid.f_seq = *seqnr;
+ seq->lcs_fid.f_ver = 0;
/*
* Inform caller that sequence switch is performed to allow it
* to setup FLD for it.
*/
seq_fid_alloc_fini(seq);
- cfs_up(&seq->lcs_sem);
+ mutex_unlock(&seq->lcs_mutex);
return rc;
}
LASSERT(fid != NULL);
cfs_waitlink_init(&link);
- cfs_down(&seq->lcs_sem);
+ mutex_lock(&seq->lcs_mutex);
+
+ if (OBD_FAIL_CHECK(OBD_FAIL_SEQ_EXHAUST))
+ seq->lcs_fid.f_oid = seq->lcs_width;
while (1) {
seqno_t seqnr;
CERROR("%s: Can't allocate new sequence, "
"rc %d\n", seq->lcs_name, rc);
seq_fid_alloc_fini(seq);
- cfs_up(&seq->lcs_sem);
+ mutex_unlock(&seq->lcs_mutex);
RETURN(rc);
}
}
*fid = seq->lcs_fid;
- cfs_up(&seq->lcs_sem);
+ mutex_unlock(&seq->lcs_mutex);
CDEBUG(D_INFO, "%s: Allocated FID "DFID"\n", seq->lcs_name, PFID(fid));
RETURN(rc);
LASSERT(seq != NULL);
cfs_waitlink_init(&link);
- cfs_down(&seq->lcs_sem);
+ mutex_lock(&seq->lcs_mutex);
while (seq->lcs_update) {
cfs_waitq_add(&seq->lcs_waitq, &link);
cfs_set_current_state(CFS_TASK_UNINT);
- cfs_up(&seq->lcs_sem);
+ mutex_unlock(&seq->lcs_mutex);
cfs_waitq_wait(&link, CFS_TASK_UNINT);
- cfs_down(&seq->lcs_sem);
+ mutex_lock(&seq->lcs_mutex);
cfs_waitq_del(&seq->lcs_waitq, &link);
cfs_set_current_state(CFS_TASK_RUNNING);
}
seq->lcs_space.lsr_index = -1;
range_init(&seq->lcs_space);
- cfs_up(&seq->lcs_sem);
+ mutex_unlock(&seq->lcs_mutex);
}
EXPORT_SYMBOL(seq_client_flush);
-static void seq_client_proc_fini(struct lu_client_seq *seq);
-
+static void seq_client_proc_fini(struct lu_client_seq *seq)
+{
#ifdef LPROCFS
+ ENTRY;
+ if (seq->lcs_proc_dir) {
+ if (!IS_ERR(seq->lcs_proc_dir))
+ lprocfs_remove(&seq->lcs_proc_dir);
+ seq->lcs_proc_dir = NULL;
+ }
+ EXIT;
+#endif /* LPROCFS */
+}
+
static int seq_client_proc_init(struct lu_client_seq *seq)
{
+#ifdef LPROCFS
int rc;
ENTRY;
out_cleanup:
seq_client_proc_fini(seq);
return rc;
-}
-
-static void seq_client_proc_fini(struct lu_client_seq *seq)
-{
- ENTRY;
- if (seq->lcs_proc_dir) {
- if (!IS_ERR(seq->lcs_proc_dir))
- lprocfs_remove(&seq->lcs_proc_dir);
- seq->lcs_proc_dir = NULL;
- }
- EXIT;
-}
-#else
-static int seq_client_proc_init(struct lu_client_seq *seq)
-{
- return 0;
-}
-static void seq_client_proc_fini(struct lu_client_seq *seq)
-{
- return;
-}
+#else /* LPROCFS */
+ return 0;
#endif
+}
int seq_client_init(struct lu_client_seq *seq,
struct obd_export *exp,
const char *prefix,
struct lu_server_seq *srv)
{
- int rc;
- ENTRY;
+ int rc;
+ ENTRY;
- LASSERT(seq != NULL);
- LASSERT(prefix != NULL);
+ LASSERT(seq != NULL);
+ LASSERT(prefix != NULL);
- seq->lcs_exp = exp;
- seq->lcs_srv = srv;
- seq->lcs_type = type;
- cfs_sema_init(&seq->lcs_sem, 1);
- seq->lcs_width = LUSTRE_SEQ_MAX_WIDTH;
- cfs_waitq_init(&seq->lcs_waitq);
+ seq->lcs_srv = srv;
+ seq->lcs_type = type;
- /* Make sure that things are clear before work is started. */
- seq_client_flush(seq);
+ mutex_init(&seq->lcs_mutex);
+ if (type == LUSTRE_SEQ_METADATA)
+ seq->lcs_width = LUSTRE_METADATA_SEQ_MAX_WIDTH;
+ else
+ seq->lcs_width = LUSTRE_DATA_SEQ_MAX_WIDTH;
- if (exp == NULL) {
- LASSERT(seq->lcs_srv != NULL);
- } else {
- LASSERT(seq->lcs_exp != NULL);
- seq->lcs_exp = class_export_get(seq->lcs_exp);
- }
+ cfs_waitq_init(&seq->lcs_waitq);
+ /* Make sure that things are clear before work is started. */
+ seq_client_flush(seq);
- snprintf(seq->lcs_name, sizeof(seq->lcs_name),
- "cli-%s", prefix);
+ if (exp != NULL)
+ seq->lcs_exp = class_export_get(exp);
+ else if (type == LUSTRE_SEQ_METADATA)
+ LASSERT(seq->lcs_srv != NULL);
- rc = seq_client_proc_init(seq);
- if (rc)
- seq_client_fini(seq);
- RETURN(rc);
+ snprintf(seq->lcs_name, sizeof(seq->lcs_name),
+ "cli-%s", prefix);
+
+ rc = seq_client_proc_init(seq);
+ if (rc)
+ seq_client_fini(seq);
+ RETURN(rc);
}
EXPORT_SYMBOL(seq_client_init);
EXIT;
}
EXPORT_SYMBOL(seq_client_fini);
+
+int client_fid_init(struct obd_device *obd,
+ struct obd_export *exp, enum lu_cli_type type)
+{
+ struct client_obd *cli = &obd->u.cli;
+ char *prefix;
+ int rc;
+ ENTRY;
+
+ OBD_ALLOC_PTR(cli->cl_seq);
+ if (cli->cl_seq == NULL)
+ RETURN(-ENOMEM);
+
+ OBD_ALLOC(prefix, MAX_OBD_NAME + 5);
+ if (prefix == NULL)
+ GOTO(out_free_seq, rc = -ENOMEM);
+
+ snprintf(prefix, MAX_OBD_NAME + 5, "cli-%s", obd->obd_name);
+
+ /* Init client side sequence-manager */
+ rc = seq_client_init(cli->cl_seq, exp, type, prefix, NULL);
+ OBD_FREE(prefix, MAX_OBD_NAME + 5);
+ if (rc)
+ GOTO(out_free_seq, rc);
+
+ RETURN(rc);
+out_free_seq:
+ OBD_FREE_PTR(cli->cl_seq);
+ cli->cl_seq = NULL;
+ return rc;
+}
+EXPORT_SYMBOL(client_fid_init);
+
+int client_fid_fini(struct obd_device *obd)
+{
+ struct client_obd *cli = &obd->u.cli;
+ ENTRY;
+
+ if (cli->cl_seq != NULL) {
+ seq_client_fini(cli->cl_seq);
+ OBD_FREE_PTR(cli->cl_seq);
+ cli->cl_seq = NULL;
+ }
+
+ RETURN(0);
+}
+EXPORT_SYMBOL(client_fid_fini);
+
+#ifdef __KERNEL__
+struct proc_dir_entry *seq_type_proc_dir;
+
+static int __init fid_mod_init(void)
+{
+ seq_type_proc_dir = lprocfs_register(LUSTRE_SEQ_NAME,
+ proc_lustre_root,
+ NULL, NULL);
+ if (IS_ERR(seq_type_proc_dir))
+ return PTR_ERR(seq_type_proc_dir);
+
+# ifdef HAVE_SERVER_SUPPORT
+ fid_server_mod_init();
+# endif
+
+ return 0;
+}
+
+static void __exit fid_mod_exit(void)
+{
+# ifdef HAVE_SERVER_SUPPORT
+ fid_server_mod_exit();
+# endif
+
+ if (seq_type_proc_dir != NULL && !IS_ERR(seq_type_proc_dir)) {
+ lprocfs_remove(&seq_type_proc_dir);
+ seq_type_proc_dir = NULL;
+ }
+}
+
+MODULE_AUTHOR("Sun Microsystems, Inc. <http://www.lustre.org/>");
+MODULE_DESCRIPTION("Lustre FID Module");
+MODULE_LICENSE("GPL");
+
+cfs_module(fid, "0.1.0", fid_mod_init, fid_mod_exit);
+#endif /* __KERNEL__ */