* Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
*
- * Copyright (c) 2011, 2012, Intel Corporation.
+ * Copyright (c) 2012, 2014, Intel Corporation.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
*/
#define DEBUG_SUBSYSTEM S_RPC
-#ifndef __KERNEL__
-#include <liblustre.h>
-#endif
#include <obd_support.h>
#include <lustre_net.h>
#include <lustre_lib.h>
RETURN(0);
}
-EXPORT_SYMBOL(ptlrpc_start_bulk_transfer);
/**
* Server side bulk abort. Idempotent. Not thread-safe (i.e. only
*/
void ptlrpc_abort_bulk(struct ptlrpc_bulk_desc *desc)
{
- struct l_wait_info lwi;
- int rc;
+ struct l_wait_info lwi;
+ int rc;
- LASSERT(!cfs_in_interrupt()); /* might sleep */
+ LASSERT(!in_interrupt()); /* might sleep */
- if (!ptlrpc_server_bulk_active(desc)) /* completed or */
- return; /* never started */
+ if (!ptlrpc_server_bulk_active(desc)) /* completed or */
+ return; /* never started */
- /* We used to poison the pages with 0xab here because we did not want to
- * send any meaningful data over the wire for evicted clients (bug 9297)
- * However, this is no longer safe now that we use the page cache on the
- * OSS (bug 20560) */
+ /* We used to poison the pages with 0xab here because we did not want to
+ * send any meaningful data over the wire for evicted clients (bug 9297)
+ * However, this is no longer safe now that we use the page cache on the
+ * OSS (bug 20560) */
- /* The unlink ensures the callback happens ASAP and is the last
- * one. If it fails, it must be because completion just happened,
- * but we must still l_wait_event() in this case, to give liblustre
- * a chance to run server_bulk_callback()*/
+ /* The unlink ensures the callback happens ASAP and is the last
+ * one. If it fails, it must be because completion just happened,
+ * but we must still l_wait_event() in this case, to give liblustre
+ * a chance to run server_bulk_callback()*/
mdunlink_iterate_helper(desc->bd_mds, desc->bd_md_count);
- for (;;) {
- /* Network access will complete in finite time but the HUGE
- * timeout lets us CWARN for visibility of sluggish NALs */
- lwi = LWI_TIMEOUT_INTERVAL(cfs_time_seconds(LONG_UNLINK),
- cfs_time_seconds(1), NULL, NULL);
- rc = l_wait_event(desc->bd_waitq,
- !ptlrpc_server_bulk_active(desc), &lwi);
- if (rc == 0)
- return;
-
- LASSERT(rc == -ETIMEDOUT);
- CWARN("Unexpectedly long timeout: desc %p\n", desc);
- }
+ for (;;) {
+ /* Network access will complete in finite time but the HUGE
+ * timeout lets us CWARN for visibility of sluggish NALs */
+ lwi = LWI_TIMEOUT_INTERVAL(cfs_time_seconds(LONG_UNLINK),
+ cfs_time_seconds(1), NULL, NULL);
+ rc = l_wait_event(desc->bd_waitq,
+ !ptlrpc_server_bulk_active(desc), &lwi);
+ if (rc == 0)
+ return;
+
+ LASSERT(rc == -ETIMEDOUT);
+ CWARN("Unexpectedly long timeout: desc %p\n", desc);
+ }
}
-EXPORT_SYMBOL(ptlrpc_abort_bulk);
#endif /* HAVE_SERVER_SUPPORT */
/**
LASSERT(desc->bd_type == BULK_PUT_SINK ||
desc->bd_type == BULK_GET_SOURCE);
+ /* cleanup the state of the bulk for it will be reused */
+ if (req->rq_resend || req->rq_send_state == LUSTRE_IMP_REPLAY)
+ desc->bd_nob_transferred = 0;
+ else
+ LASSERT(desc->bd_nob_transferred == 0);
+
desc->bd_failure = 0;
peer = desc->bd_import->imp_connection->c_peer;
LNET_UNLINK, LNET_INS_AFTER, &me_h);
if (rc != 0) {
CERROR("%s: LNetMEAttach failed x"LPU64"/%d: rc = %d\n",
- desc->bd_export->exp_obd->obd_name, xid,
+ desc->bd_import->imp_obd->obd_name, xid,
posted_md, rc);
break;
}
&desc->bd_mds[posted_md]);
if (rc != 0) {
CERROR("%s: LNetMDAttach failed x"LPU64"/%d: rc = %d\n",
- desc->bd_export->exp_obd->obd_name, xid,
+ desc->bd_import->imp_obd->obd_name, xid,
posted_md, rc);
rc2 = LNetMEUnlink(me_h);
LASSERT(rc2 == 0);
/* Holler if peer manages to touch buffers before he knows the xid */
if (desc->bd_md_count != total_md)
CWARN("%s: Peer %s touched %d buffers while I registered\n",
- desc->bd_export->exp_obd->obd_name, libcfs_id2str(peer),
+ desc->bd_import->imp_obd->obd_name, libcfs_id2str(peer),
total_md - desc->bd_md_count);
spin_unlock(&desc->bd_lock);
RETURN(0);
}
-EXPORT_SYMBOL(ptlrpc_register_bulk);
/**
* Disconnect a bulk desc from the network. Idempotent. Not
*/
int ptlrpc_unregister_bulk(struct ptlrpc_request *req, int async)
{
- struct ptlrpc_bulk_desc *desc = req->rq_bulk;
- cfs_waitq_t *wq;
- struct l_wait_info lwi;
- int rc;
- ENTRY;
+ struct ptlrpc_bulk_desc *desc = req->rq_bulk;
+ struct l_wait_info lwi;
+ int rc;
+ ENTRY;
- LASSERT(!cfs_in_interrupt()); /* might sleep */
+ LASSERT(!in_interrupt()); /* might sleep */
- /* Let's setup deadline for reply unlink. */
- if (OBD_FAIL_CHECK(OBD_FAIL_PTLRPC_LONG_BULK_UNLINK) &&
- async && req->rq_bulk_deadline == 0)
- req->rq_bulk_deadline = cfs_time_current_sec() + LONG_UNLINK;
+ /* Let's setup deadline for reply unlink. */
+ if (OBD_FAIL_CHECK(OBD_FAIL_PTLRPC_LONG_BULK_UNLINK) &&
+ async && req->rq_bulk_deadline == 0)
+ req->rq_bulk_deadline = cfs_time_current_sec() + LONG_UNLINK;
if (ptlrpc_client_bulk_active(req) == 0) /* completed or */
RETURN(1); /* never registered */
if (async)
RETURN(0);
- if (req->rq_set != NULL)
- wq = &req->rq_set->set_waitq;
- else
- wq = &req->rq_reply_waitq;
-
for (;;) {
+ /* The wq argument is ignored by user-space wait_event macros */
+ wait_queue_head_t *wq = (req->rq_set != NULL) ?
+ &req->rq_set->set_waitq :
+ &req->rq_reply_waitq;
/* Network access will complete in finite time but the HUGE
* timeout lets us CWARN for visibility of sluggish NALs */
lwi = LWI_TIMEOUT_INTERVAL(cfs_time_seconds(LONG_UNLINK),
}
RETURN(0);
}
-EXPORT_SYMBOL(ptlrpc_unregister_bulk);
static void ptlrpc_at_set_reply(struct ptlrpc_request *req, int flags)
{
/* Report actual service time for client latency calc */
lustre_msg_set_service_time(req->rq_repmsg, service_time);
/* Report service time estimate for future client reqs, but report 0
- * (to be ignored by client) if it's a error reply during recovery.
+ * (to be ignored by client) if it's an error reply during recovery.
* (bz15815) */
if (req->rq_type == PTL_RPC_MSG_ERR &&
(req->rq_export == NULL || req->rq_export->exp_obd->obd_recovering))
lustre_msg_set_timeout(req->rq_repmsg,
at_get(&svcpt->scp_at_estimate));
- if (req->rq_reqmsg &&
- !(lustre_msghdr_get_flags(req->rq_reqmsg) & MSGHDR_AT_SUPPORT)) {
- CDEBUG(D_ADAPTTO, "No early reply support: flags=%#x "
- "req_flags=%#x magic=%d:%x/%x len=%d\n",
- flags, lustre_msg_get_flags(req->rq_reqmsg),
- lustre_msg_is_v1(req->rq_reqmsg),
- lustre_msg_get_magic(req->rq_reqmsg),
- lustre_msg_get_magic(req->rq_repmsg), req->rq_replen);
- }
+ if (req->rq_reqmsg &&
+ !(lustre_msghdr_get_flags(req->rq_reqmsg) & MSGHDR_AT_SUPPORT)) {
+ CDEBUG(D_ADAPTTO, "No early reply support: flags=%#x "
+ "req_flags=%#x magic=%x/%x len=%d\n",
+ flags, lustre_msg_get_flags(req->rq_reqmsg),
+ lustre_msg_get_magic(req->rq_reqmsg),
+ lustre_msg_get_magic(req->rq_repmsg), req->rq_replen);
+ }
}
/**
* Send request reply from request \a req reply buffer.
* \a flags defines reply types
- * Returns 0 on sucess or error code
+ * Returns 0 on success or error code
*/
int ptlrpc_send_reply(struct ptlrpc_request *req, int flags)
{
req->rq_type = PTL_RPC_MSG_REPLY;
lustre_msg_set_type(req->rq_repmsg, req->rq_type);
- lustre_msg_set_status(req->rq_repmsg, req->rq_status);
+ lustre_msg_set_status(req->rq_repmsg,
+ ptlrpc_status_hton(req->rq_status));
lustre_msg_set_opc(req->rq_repmsg,
req->rq_reqmsg ? lustre_msg_get_opc(req->rq_reqmsg) : 0);
ptlrpc_connection_put(conn);
return rc;
}
-EXPORT_SYMBOL(ptlrpc_send_reply);
int ptlrpc_reply (struct ptlrpc_request *req)
{
else
return (ptlrpc_send_reply(req, 0));
}
-EXPORT_SYMBOL(ptlrpc_reply);
/**
* For request \a req send an error reply back. Create empty
rc = ptlrpc_send_reply(req, may_be_difficult);
RETURN(rc);
}
-EXPORT_SYMBOL(ptlrpc_send_error);
int ptlrpc_error(struct ptlrpc_request *req)
{
return ptlrpc_send_error(req, 0);
}
-EXPORT_SYMBOL(ptlrpc_error);
/**
* Send request \a request.
struct ptlrpc_connection *connection;
lnet_handle_me_t reply_me_h;
lnet_md_t reply_md;
- struct obd_device *obd = request->rq_import->imp_obd;
+ struct obd_import *imp = request->rq_import;
+ struct obd_device *obd = imp->imp_obd;
ENTRY;
if (OBD_FAIL_CHECK(OBD_FAIL_PTLRPC_DROP_RPC))
/* If this is a re-transmit, we're required to have disengaged
* cleanly from the previous attempt */
LASSERT(!request->rq_receiving_reply);
-
- if (request->rq_import->imp_obd &&
- request->rq_import->imp_obd->obd_fail) {
- CDEBUG(D_HA, "muting rpc for failed imp obd %s\n",
- request->rq_import->imp_obd->obd_name);
- /* this prevents us from waiting in ptlrpc_queue_wait */
- request->rq_err = 1;
+ LASSERT(!((lustre_msg_get_flags(request->rq_reqmsg) & MSG_REPLAY) &&
+ (imp->imp_state == LUSTRE_IMP_FULL)));
+
+ if (unlikely(obd != NULL && obd->obd_fail)) {
+ CDEBUG(D_HA, "muting rpc for failed imp obd %s\n",
+ obd->obd_name);
+ /* this prevents us from waiting in ptlrpc_queue_wait */
+ spin_lock(&request->rq_lock);
+ request->rq_err = 1;
+ spin_unlock(&request->rq_lock);
request->rq_status = -ENODEV;
RETURN(-ENODEV);
}
- connection = request->rq_import->imp_connection;
-
- lustre_msg_set_handle(request->rq_reqmsg,
- &request->rq_import->imp_remote_handle);
- lustre_msg_set_type(request->rq_reqmsg, PTL_RPC_MSG_REQUEST);
- lustre_msg_set_conn_cnt(request->rq_reqmsg,
- request->rq_import->imp_conn_cnt);
- lustre_msghdr_set_flags(request->rq_reqmsg,
- request->rq_import->imp_msghdr_flags);
-
- if (request->rq_resend)
- lustre_msg_add_flags(request->rq_reqmsg, MSG_RESENT);
-
+ connection = imp->imp_connection;
+
+ lustre_msg_set_handle(request->rq_reqmsg,
+ &imp->imp_remote_handle);
+ lustre_msg_set_type(request->rq_reqmsg, PTL_RPC_MSG_REQUEST);
+ lustre_msg_set_conn_cnt(request->rq_reqmsg,
+ imp->imp_conn_cnt);
+ lustre_msghdr_set_flags(request->rq_reqmsg,
+ imp->imp_msghdr_flags);
+
+ /** For enabled AT all request should have AT_SUPPORT in the
+ * FULL import state when OBD_CONNECT_AT is set */
+ LASSERT(AT_OFF || imp->imp_state != LUSTRE_IMP_FULL ||
+ (imp->imp_msghdr_flags & MSGHDR_AT_SUPPORT) ||
+ !(imp->imp_connect_data.ocd_connect_flags &
+ OBD_CONNECT_AT));
+
+ if (request->rq_resend) {
+ lustre_msg_add_flags(request->rq_reqmsg, MSG_RESENT);
+ if (request->rq_resend_cb != NULL)
+ request->rq_resend_cb(request, &request->rq_async_args);
+ }
if (request->rq_memalloc)
mpflag = cfs_memory_pressure_get_and_set();
if (rc) {
/* this prevents us from looping in
* ptlrpc_queue_wait */
- request->rq_err = 1;
+ spin_lock(&request->rq_lock);
+ request->rq_err = 1;
+ spin_unlock(&request->rq_lock);
request->rq_status = rc;
GOTO(cleanup_bulk, rc);
}
}
spin_lock(&request->rq_lock);
- /* If the MD attach succeeds, there _will_ be a reply_in callback */
- request->rq_receiving_reply = !noreply;
- /* We are responsible for unlinking the reply buffer */
- request->rq_must_unlink = !noreply;
- /* Clear any flags that may be present from previous sends. */
+ /* We are responsible for unlinking the reply buffer */
+ request->rq_reply_unlinked = noreply;
+ request->rq_receiving_reply = !noreply;
+ /* Clear any flags that may be present from previous sends. */
+ request->rq_req_unlinked = 0;
request->rq_replied = 0;
request->rq_err = 0;
request->rq_timedout = 0;
request->rq_net_err = 0;
request->rq_resend = 0;
request->rq_restart = 0;
- request->rq_reply_truncate = 0;
+ request->rq_reply_truncated = 0;
spin_unlock(&request->rq_lock);
if (!noreply) {
reply_md.user_ptr = &request->rq_reply_cbid;
reply_md.eq_handle = ptlrpc_eq_h;
- /* We must see the unlink callback to unset rq_must_unlink,
- so we can't auto-unlink */
+ /* We must see the unlink callback to set rq_reply_unlinked,
+ * so we can't auto-unlink */
rc = LNetMDAttach(reply_me_h, reply_md, LNET_RETAIN,
&request->rq_reply_md_h);
if (rc != 0) {
/* add references on request for request_out_callback */
ptlrpc_request_addref(request);
- if (obd->obd_svc_stats != NULL)
- lprocfs_counter_add(obd->obd_svc_stats, PTLRPC_REQACTIVE_CNTR,
- cfs_atomic_read(&request->rq_import->imp_inflight));
+ if (obd != NULL && obd->obd_svc_stats != NULL)
+ lprocfs_counter_add(obd->obd_svc_stats, PTLRPC_REQACTIVE_CNTR,
+ atomic_read(&imp->imp_inflight));
- OBD_FAIL_TIMEOUT(OBD_FAIL_PTLRPC_DELAY_SEND, request->rq_timeout + 5);
+ OBD_FAIL_TIMEOUT(OBD_FAIL_PTLRPC_DELAY_SEND, request->rq_timeout + 5);
- cfs_gettimeofday(&request->rq_arrival_time);
- request->rq_sent = cfs_time_current_sec();
- /* We give the server rq_timeout secs to process the req, and
- add the network latency for our local timeout. */
+ do_gettimeofday(&request->rq_sent_tv);
+ request->rq_sent = cfs_time_current_sec();
+ /* We give the server rq_timeout secs to process the req, and
+ add the network latency for our local timeout. */
request->rq_deadline = request->rq_sent + request->rq_timeout +
ptlrpc_at_get_net_latency(request);
- ptlrpc_pinger_sending_on_import(request->rq_import);
+ ptlrpc_pinger_sending_on_import(imp);
DEBUG_REQ(D_INFO, request, "send flg=%x",
lustre_msg_get_flags(request->rq_reqmsg));
connection,
request->rq_request_portal,
request->rq_xid, 0);
- if (rc == 0)
- GOTO(out, rc);
+ if (likely(rc == 0))
+ GOTO(out, rc);
+ request->rq_req_unlinked = 1;
ptlrpc_req_finished(request);
if (noreply)
GOTO(out, rc);