diff --git a/lnet/include/linux/kp30.h b/lnet/include/linux/kp30.h
index 6ef28a8535e985f668dcbf675399140c6c071516..b13f161638e66d3772c585a06aa9afa5b3da76b2 100644
--- a/lnet/include/linux/kp30.h
+++ b/lnet/include/linux/kp30.h
@@ -326,6 +326,7 @@ void portals_debug_dumplog(void);
/* support decl needed both by kernel and liblustre */
char *portals_nid2str(int nal, ptl_nid_t nid, char *str);
+char *portals_id2str(int nal, ptl_process_id_t nid, char *str);
#ifndef CURRENT_TIME
# define CURRENT_TIME time(0)
@@ -644,7 +645,7 @@ enum {
NAL_ENUM_END_MARKER
};
-#define PTL_NALFMT_SIZE 26 /* %u:%u.%u.%u.%u (10+4+4+4+3+1) */
+#define PTL_NALFMT_SIZE 30 /* %u:%u.%u.%u.%u,%u (10+4+4+4+3+4+1) */
#define NAL_MAX_NR (NAL_ENUM_END_MARKER - 1)
diff --git a/lnet/klnds/gmlnd/gmlnd_api.c b/lnet/klnds/gmlnd/gmlnd_api.c
index 002587d1b5962f3b73acbcb9fc29dde9d0be25ae..bdf5cfe6cfb547f8ed8d3af0b05857456f1c9773 100644
--- a/lnet/klnds/gmlnd/gmlnd_api.c
+++ b/lnet/klnds/gmlnd/gmlnd_api.c
@@ -318,7 +318,7 @@ gmnal_api_startup(nal_t *nal, ptl_pid_t requested_pid,
/*
pid = gm_getpid();
*/
- process_id.pid = 0;
+ process_id.pid = requested_pid;
process_id.nid = global_nid;
CDEBUG(D_INFO, "portals_pid is [%u]\n", process_id.pid);
diff --git a/lnet/klnds/iblnd/ibnal.c b/lnet/klnds/iblnd/ibnal.c
index 86c2a6355206923d845a8321c343316a9ba7eb7b..c6bd7ed044c1c7f0390168566df82e97da05bf91 100644
--- a/lnet/klnds/iblnd/ibnal.c
+++ b/lnet/klnds/iblnd/ibnal.c
@@ -246,7 +246,7 @@ kibnal_init(int interface, // no use here
rc = lib_init(&kibnal_lib,
kibnal_data.kib_nid,
- 0, // process id is set as 0
+ requested_pid , // process id is set as requested_pid instead of 0
ptl_size,
ac_size);
diff --git a/lnet/klnds/qswlnd/qswlnd.c b/lnet/klnds/qswlnd/qswlnd.c
index c595450684aadb55fb7bb7baf11f7cc1dc7319ac..38d1636166d57950555ae99a3d4e977220ccee7d 100644
--- a/lnet/klnds/qswlnd/qswlnd.c
+++ b/lnet/klnds/qswlnd/qswlnd.c
@@ -696,7 +696,7 @@ kqswnal_startup (nal_t *nal, ptl_pid_t requested_pid,
/* Network interface ready to initialise */
my_process_id.nid = kqswnal_elanid2nid(kqswnal_data.kqn_elanid);
- my_process_id.pid = 0;
+ my_process_id.pid = requested_pid;
rc = lib_init(&kqswnal_lib, nal, my_process_id,
requested_limits, actual_limits);
@@ -802,7 +802,7 @@ kqswnal_initialise (void)
/* Pure gateways, and the workaround for 'EKC blocks forever until
* the service is active' want the NAL started up at module load
* time... */
- rc = PtlNIInit(QSWNAL, 0, NULL, NULL, &kqswnal_ni);
+ rc = PtlNIInit(QSWNAL, LUSTRE_SRV_PTL_PID, NULL, NULL, &kqswnal_ni);
if (rc != PTL_OK && rc != PTL_IFACE_DUP) {
ptl_unregister_nal(QSWNAL);
return (-ENODEV);
diff --git a/lnet/klnds/scimaclnd/scimacnal.c b/lnet/klnds/scimaclnd/scimacnal.c
index e77bd8ee278dc69428c1e5d1525825e6907dfab5..75188e97dcb1b35fd4cc90929b1e8e6665db763f 100644
--- a/lnet/klnds/scimaclnd/scimacnal.c
+++ b/lnet/klnds/scimaclnd/scimacnal.c
@@ -205,7 +205,7 @@ static int kscimacnal_startup(nal_t *nal, ptl_pid_t requested_pid,
}
kscimacnal_data.ksci_nid = (ptl_nid_t)(ntohl(mac_physaddr));
- process_id.pid = 0;
+ process_id.pid = requested_pid;
process_id.nid = kscimacnal_data.ksci_nid;
CDEBUG(D_NET, "calling lib_init with nid "LPX64"\n",
diff --git a/lnet/klnds/socklnd/socklnd.c b/lnet/klnds/socklnd/socklnd.c
index 9d39cb1bd522b4ec6e115e5b604adc658f45d98f..3a3629b873d45b64e163b0d8b4123cf72b6ec2b1 100644
--- a/lnet/klnds/socklnd/socklnd.c
+++ b/lnet/klnds/socklnd/socklnd.c
@@ -1589,7 +1589,7 @@ ksocknal_api_startup (nal_t *nal, ptl_pid_t requested_pid,
}
/* NB we have to wait to be told our true NID... */
- process_id.pid = 0;
+ process_id.pid = requested_pid; //LUSTRE_SRV_PTL_PID;
process_id.nid = 0;
rc = lib_init(&ksocknal_lib, nal, process_id,
@@ -1739,7 +1739,7 @@ ksocknal_module_init (void)
}
/* Pure gateways want the NAL started up at module load time... */
- rc = PtlNIInit(SOCKNAL, 0, NULL, NULL, &ksocknal_ni);
+ rc = PtlNIInit(SOCKNAL, LUSTRE_SRV_PTL_PID, NULL, NULL, &ksocknal_ni);
if (rc != PTL_OK && rc != PTL_IFACE_DUP) {
ptl_unregister_nal(SOCKNAL);
return (-ENODEV);
diff --git a/lnet/klnds/socklnd/socklnd.h b/lnet/klnds/socklnd/socklnd.h
index ff73f71ce51cacad07e14ac0e76001f340d4e4a9..e14205951eb4d76ce34272dab93c5e14e74300eb 100644
--- a/lnet/klnds/socklnd/socklnd.h
+++ b/lnet/klnds/socklnd/socklnd.h
@@ -66,6 +66,7 @@
#include <portals/lib-p30.h>
#include <portals/nal.h>
#include <portals/socknal.h>
+#include <linux/lustre_idl.h>
#if CONFIG_SMP
# define SOCKNAL_N_SCHED num_online_cpus() /* # socknal schedulers */
diff --git a/lnet/libcfs/debug.c b/lnet/libcfs/debug.c
index 2b2ce3b99247aa72c1b99ad6659024177f866c22..eb75e60442e470d08c4b96dcd265d376fb500445 100644
--- a/lnet/libcfs/debug.c
+++ b/lnet/libcfs/debug.c
@@ -985,6 +985,30 @@ char *portals_nid2str(int nal, ptl_nid_t nid, char *str)
}
return str;
}
+/* bug #4615 */
+char *portals_id2str(int nal, ptl_process_id_t id, char *str)
+{
+ switch(nal){
+ case TCPNAL:
+ /* userspace NAL */
+ case SOCKNAL:
+ snprintf(str, PTL_NALFMT_SIZE - 1, "%u:%u.%u.%u.%u,%u",
+ (__u32)(id.nid >> 32), HIPQUAD((id.nid)) , id.pid);
+ break;
+ case QSWNAL:
+ case GMNAL:
+ case IBNAL:
+ snprintf(str, PTL_NALFMT_SIZE - 1, "%u:%u,%u",
+ (__u32)(id.nid >> 32), (__u32)id.nid, id.pid);
+ break;
+ default:
+ snprintf(str, PTL_NALFMT_SIZE - 1, "?%d? %llx,%lx",
+ nal, (long long)id.nid, (long)id.pid );
+ break;
+ }
+ return str;
+}
+
#ifdef __KERNEL__
char stack_backtrace[LUSTRE_TRACE_SIZE];
@@ -1077,3 +1101,4 @@ EXPORT_SYMBOL(portals_debug_set_level);
EXPORT_SYMBOL(portals_run_upcall);
EXPORT_SYMBOL(portals_run_lbug_upcall);
EXPORT_SYMBOL(portals_nid2str);
+EXPORT_SYMBOL(portals_id2str);
diff --git a/lnet/lnet/lib-move.c b/lnet/lnet/lib-move.c
index 9dcc06e343baea58fbeb05faf2f3cfa75bd9c645..854a452e55a4081d8fc9b778972ff4305e81d424 100644
--- a/lnet/lnet/lib-move.c
+++ b/lnet/lnet/lib-move.c
@@ -82,6 +82,8 @@ lib_match_md(lib_nal_t *nal, int index, int op_mask,
if (me->match_id.nid != PTL_NID_ANY &&
me->match_id.nid != src_nid)
continue;
+
+ CDEBUG(D_NET,"match_id.pid [%x], src_pid [%x]\n", me->match_id.pid, src_pid);
if (me->match_id.pid != PTL_PID_ANY &&
me->match_id.pid != src_pid)
diff --git a/lnet/lnet/module.c b/lnet/lnet/module.c
index 5615a7244329138a1242d378fefc27ebcd12f8cf..d292a501a057de31388bd608e0c385be870287f5 100644
--- a/lnet/lnet/module.c
+++ b/lnet/lnet/module.c
@@ -50,6 +50,7 @@
#include <linux/kp30.h>
#include <linux/kpr.h>
#include <linux/portals_compat25.h>
+#include <linux/lustre_idl.h>
extern void (kping_client)(struct portal_ioctl_data *);
@@ -83,7 +84,7 @@ static int kportal_ioctl(struct portal_ioctl_data *data,
CDEBUG (D_IOCTL, "Getting nid for nal [%d]\n", data->ioc_nal);
- err = PtlNIInit(data->ioc_nal, 0, NULL, NULL, &nih);
+ err = PtlNIInit(data->ioc_nal, LUSTRE_SRV_PTL_PID, NULL, NULL, &nih);
if (!(err == PTL_OK || err == PTL_IFACE_DUP))
RETURN (-EINVAL);
@@ -104,7 +105,7 @@ static int kportal_ioctl(struct portal_ioctl_data *data,
CDEBUG (D_IOCTL, "fail nid: [%d] "LPU64" count %d\n",
data->ioc_nal, data->ioc_nid, data->ioc_count);
- err = PtlNIInit(data->ioc_nal, 0, NULL, NULL, &nih);
+ err = PtlNIInit(data->ioc_nal, LUSTRE_SRV_PTL_PID, NULL, NULL, &nih);
if (!(err == PTL_OK || err == PTL_IFACE_DUP))
return (-EINVAL);
diff --git a/lnet/ulnds/socklnd/tcplnd.c b/lnet/ulnds/socklnd/tcplnd.c
index 34a9c9d5e3d43b6a57deccb423d7e116b0e2f70a..6e9cca99ca7647fb842cd3c3a592f4e1a46ee3ec 100644
--- a/lnet/ulnds/socklnd/tcplnd.c
+++ b/lnet/ulnds/socklnd/tcplnd.c
@@ -252,7 +252,7 @@ int tcpnal_init(bridge b)
return(PTL_NAL_FAILED);
}
/* XXX cfs hack */
- b->lib_nal->libnal_ni.ni_pid.pid=0;
+// b->lib_nal->libnal_ni.ni_pid.pid=0;
b->lower=m;
return(PTL_OK);
}
diff --git a/lnet/ulnds/tcplnd.c b/lnet/ulnds/tcplnd.c
index 34a9c9d5e3d43b6a57deccb423d7e116b0e2f70a..6e9cca99ca7647fb842cd3c3a592f4e1a46ee3ec 100644
--- a/lnet/ulnds/tcplnd.c
+++ b/lnet/ulnds/tcplnd.c
@@ -252,7 +252,7 @@ int tcpnal_init(bridge b)
return(PTL_NAL_FAILED);
}
/* XXX cfs hack */
- b->lib_nal->libnal_ni.ni_pid.pid=0;
+// b->lib_nal->libnal_ni.ni_pid.pid=0;
b->lower=m;
return(PTL_OK);
}
diff --git a/lustre/include/linux/lustre_idl.h b/lustre/include/linux/lustre_idl.h
index 8f5bd37d5369319f6bfdc53003cfcfa470c134e2..b69e1e195c3febd4edc69f74a4b4e6077b707d70 100644
--- a/lustre/include/linux/lustre_idl.h
+++ b/lustre/include/linux/lustre_idl.h
@@ -133,6 +133,10 @@
#define LUSTRE_OST_VERSION (0x00040000|PTLRPC_MSG_VERSION)
#define LUSTRE_DLM_VERSION (0x00040000|PTLRPC_MSG_VERSION)
+/* initial pid */
+#define LUSTRE_PTL_PID 999999
+#define LUSTRE_SRV_PTL_PID LUSTRE_PTL_PID
+
struct lustre_handle {
__u64 cookie;
};
diff --git a/lustre/include/linux/lustre_net.h b/lustre/include/linux/lustre_net.h
index 03d9ab3a7794d2d78ce4cc24223c731f22735f24..aa8309d53ca6b79d69ae564417a235c89bf89c3b 100644
--- a/lustre/include/linux/lustre_net.h
+++ b/lustre/include/linux/lustre_net.h
@@ -153,7 +153,9 @@
#define PTLBD_MAXREQSIZE 1024
struct ptlrpc_peer {
- ptl_nid_t peer_nid;
+/* bugfix #4615
+ */
+ ptl_process_id_t peer_id;
struct ptlrpc_ni *peer_ni;
};
@@ -515,9 +517,17 @@ struct ptlrpc_service {
static inline char *ptlrpc_peernid2str(struct ptlrpc_peer *p, char *str)
{
LASSERT(p->peer_ni != NULL);
- return (portals_nid2str(p->peer_ni->pni_number, p->peer_nid, str));
+ return (portals_nid2str(p->peer_ni->pni_number, p->peer_id.nid, str));
+}
+
+/* For bug #4615 */
+static inline char *ptlrpc_id2str(struct ptlrpc_peer *p, char *str)
+{
+ LASSERT(p->peer_ni != NULL);
+ return (portals_id2str(p->peer_ni->pni_number, p->peer_id, str));
}
+
/* ptlrpc/events.c */
extern struct ptlrpc_ni ptlrpc_interfaces[];
extern int ptlrpc_ninterfaces;
@@ -539,6 +549,7 @@ int ptlrpc_put_connection(struct ptlrpc_connection *c);
struct ptlrpc_connection *ptlrpc_connection_addref(struct ptlrpc_connection *);
void ptlrpc_init_connection(void);
void ptlrpc_cleanup_connection(void);
+extern ptl_pid_t ptl_get_pid(void);
/* ptlrpc/niobuf.c */
int ptlrpc_start_bulk_transfer(struct ptlrpc_bulk_desc *desc);
diff --git a/lustre/ldlm/ldlm_lib.c b/lustre/ldlm/ldlm_lib.c
index fb70b653d9b16d0fb7e0bd0f169fc07e76b11eb5..8d415871329844559b3d3335a9ae2f7d2d5c7a37 100644
--- a/lustre/ldlm/ldlm_lib.c
+++ b/lustre/ldlm/ldlm_lib.c
@@ -416,11 +416,11 @@ int target_handle_connect(struct ptlrpc_request *req)
/* NB the casts only avoid compiler warnings */
case 8:
snprintf(remote_uuid.uuid, sizeof remote_uuid,
- "NET_"LPX64"_UUID", (__u64)req->rq_peer.peer_nid);
+ "NET_"LPX64"_UUID", (__u64)req->rq_peer.peer_id.nid);
break;
case 4:
snprintf(remote_uuid.uuid, sizeof remote_uuid,
- "NET_%x_UUID", (__u32)req->rq_peer.peer_nid);
+ "NET_%x_UUID", (__u32)req->rq_peer.peer_id.nid);
break;
default:
LBUG();
diff --git a/lustre/ldlm/ldlm_lockd.c b/lustre/ldlm/ldlm_lockd.c
index b4acd619e4a200d9ecade6b6d765de8c5b375be1..d1a60feb1450311fccd530209799a82ac9914f65 100644
--- a/lustre/ldlm/ldlm_lockd.c
+++ b/lustre/ldlm/ldlm_lockd.c
@@ -311,7 +311,7 @@ static void ldlm_failed_ast(struct ldlm_lock *lock, int rc,const char *ast_type)
LDLM_ERROR(lock, "%s AST failed (%d): evicting client %s@%s NID "LPX64
" (%s)", ast_type, rc, lock->l_export->exp_client_uuid.uuid,
- conn->c_remote_uuid.uuid, conn->c_peer.peer_nid,
+ conn->c_remote_uuid.uuid, conn->c_peer.peer_id.nid,
ptlrpc_peernid2str(&conn->c_peer, str));
ptlrpc_fail_export(lock->l_export);
}
diff --git a/lustre/liblustre/llite_lib.c b/lustre/liblustre/llite_lib.c
index 16437bcc99852b1149b4d4576777b56d85a11969..73f97da7e47ecc025c134555edacfbb2391d68d4 100644
--- a/lustre/liblustre/llite_lib.c
+++ b/lustre/liblustre/llite_lib.c
@@ -101,7 +101,29 @@ char *portals_nid2str(int nal, ptl_nid_t nid, char *str)
}
return str;
}
-
+/* bug #4615 */
+char *portals_id2str(int nal, ptl_process_id_t id, char *str)
+{
+ switch(nal){
+ case TCPNAL:
+ /* userspace NAL */
+ case SOCKNAL:
+ snprintf(str, PTL_NALFMT_SIZE - 1, "%u:%u.%u.%u.%u,%u",
+ (__u32)(id.nid >> 32), HIPQUAD((id.nid)) , id.pid);
+ break;
+ case QSWNAL:
+ case GMNAL:
+ case IBNAL:
+ snprintf(str, PTL_NALFMT_SIZE - 1, "%u:%u,%u",
+ (__u32)(id.nid >> 32), (__u32)id.nid, id.pid);
+ break;
+ default:
+ snprintf(str, PTL_NALFMT_SIZE - 1, "?%d? %llx,%lx",
+ nal, (long long)id.nid, (long)id.pid );
+ break;
+ }
+ return str;
+}
/*
* random number generator stuff
*/
diff --git a/lustre/liblustre/tests/echo_test.c b/lustre/liblustre/tests/echo_test.c
index 19fd83a9d1dd1923077bae91a769e502ad1de294..13157cdb3b400c489dcbd0144f4491d53142803a 100644
--- a/lustre/liblustre/tests/echo_test.c
+++ b/lustre/liblustre/tests/echo_test.c
@@ -82,6 +82,29 @@ struct pingcli_args {
int count;
int size;
};
+/* bug #4615 */
+char *portals_id2str(int nal, ptl_process_id_t id, char *str)
+{
+ switch(nal){
+ case TCPNAL:
+ /* userspace NAL */
+ case SOCKNAL:
+ snprintf(str, PTL_NALFMT_SIZE - 1, "%u:%u.%u.%u.%u,%u",
+ (__u32)(id.nid >> 32), HIPQUAD((id.nid)) , id.pid);
+ break;
+ case QSWNAL:
+ case GMNAL:
+ case IBNAL:
+ snprintf(str, PTL_NALFMT_SIZE - 1, "%u:%u,%u",
+ (__u32)(id.nid >> 32), (__u32)id.nid, id.pid);
+ break;
+ default:
+ snprintf(str, PTL_NALFMT_SIZE - 1, "?%d? %llx,%lx",
+ nal, (long long)id.nid, (long)id.pid );
+ break;
+ }
+ return str;
+}
struct task_struct *current;
diff --git a/lustre/obdfilter/filter.c b/lustre/obdfilter/filter.c
index 649e6fe53bdb0dbf6713284740d596552539effe..4faa1eff860872ad14a224d668b18fd71dd46a3e 100644
--- a/lustre/obdfilter/filter.c
+++ b/lustre/obdfilter/filter.c
@@ -1626,9 +1626,9 @@ static int filter_connect_post(struct obd_export *exp)
rc = llog_receptor_accept(ctxt, exp->exp_imp_reverse);
portals_nid2str(exp->exp_connection->c_peer.peer_ni->pni_number,
- exp->exp_connection->c_peer.peer_nid, str);
+ exp->exp_connection->c_peer.peer_id.nid, str);
CDEBUG(D_OTHER, "%s: init llog ctxt for export "LPX64"/%s, group %d\n",
- obd->obd_name, exp->exp_connection->c_peer.peer_nid,
+ obd->obd_name, exp->exp_connection->c_peer.peer_id.nid,
str, fed->fed_group);
RETURN(rc);
@@ -1685,11 +1685,11 @@ static int filter_connect(struct lustre_handle *conn, struct obd_device *obd,
if (fed->fed_group != 0 && fed->fed_group != group) {
char str[PTL_NALFMT_SIZE];
portals_nid2str(exp->exp_connection->c_peer.peer_ni->pni_number,
- exp->exp_connection->c_peer.peer_nid, str);
+ exp->exp_connection->c_peer.peer_id.nid, str);
CERROR("!!! This export (nid "LPX64"/%s) used object group %d "
"earlier; now it's trying to use group %d! This could "
"be a bug in the MDS. Tell CFS.\n",
- exp->exp_connection->c_peer.peer_nid, str,
+ exp->exp_connection->c_peer.peer_id.nid, str,
fed->fed_group, group);
GOTO(cleanup, rc = -EPROTO);
}
@@ -2392,9 +2392,9 @@ static int filter_create(struct obd_export *exp, struct obdo *oa,
if (!(oa->o_valid & OBD_MD_FLGROUP) || group == 0) {
portals_nid2str(exp->exp_connection->c_peer.peer_ni->pni_number,
- exp->exp_connection->c_peer.peer_nid, str);
+ exp->exp_connection->c_peer.peer_id.nid, str);
CERROR("!!! nid "LPX64"/%s sent invalid object group %d\n",
- exp->exp_connection->c_peer.peer_nid, str, group);
+ exp->exp_connection->c_peer.peer_id.nid, str, group);
RETURN(-EINVAL);
}
@@ -2410,11 +2410,11 @@ static int filter_create(struct obd_export *exp, struct obdo *oa,
if (fed->fed_group != group && !recreate_objs &&
!(oa->o_valid & OBD_MD_REINT)) {
portals_nid2str(exp->exp_connection->c_peer.peer_ni->pni_number,
- exp->exp_connection->c_peer.peer_nid, str);
+ exp->exp_connection->c_peer.peer_id.nid, str);
CERROR("!!! This export (nid "LPX64"/%s) used object group %d "
"earlier; now it's trying to use group %d! This could "
"be a bug in the MDS. Tell CFS.\n",
- exp->exp_connection->c_peer.peer_nid, str,
+ exp->exp_connection->c_peer.peer_id.nid, str,
fed->fed_group, group);
RETURN(-ENOTUNIQ);
}
diff --git a/lustre/osc/osc_request.c b/lustre/osc/osc_request.c
index ed75cfd7dace2b2460ea40ff916d8a06e6d2b3cf..1f84b4b52c26c4544cc3d769f7267c5940c03bfe 100644
--- a/lustre/osc/osc_request.c
+++ b/lustre/osc/osc_request.c
@@ -870,12 +870,12 @@ static int osc_brw_fini_request(struct ptlrpc_request *req, struct obdo *oa,
if (server_cksum != cksum) {
CERROR("Bad checksum: server %x, client %x, server NID "
LPX64" (%s)\n", server_cksum, cksum,
- peer->peer_nid, str);
+ peer->peer_id.nid, str);
cksum_counter = 0;
oa->o_cksum = cksum;
} else if ((cksum_counter & (-cksum_counter)) == cksum_counter){
CWARN("Checksum %u from "LPX64" (%s) OK: %x\n",
- cksum_counter, peer->peer_nid, str, cksum);
+ cksum_counter, peer->peer_id.nid, str, cksum);
}
} else {
static int cksum_missed;
@@ -884,7 +884,7 @@ static int osc_brw_fini_request(struct ptlrpc_request *req, struct obdo *oa,
if ((cksum_missed & (-cksum_missed)) == cksum_missed)
CERROR("Request checksum %u from "LPX64", no reply\n",
cksum_missed,
- req->rq_import->imp_connection->c_peer.peer_nid);
+ req->rq_import->imp_connection->c_peer.peer_id.nid);
}
#endif
RETURN(0);
diff --git a/lustre/ost/ost_handler.c b/lustre/ost/ost_handler.c
index 2271c6c4081dab41debf12978c5fcdd437d2971a..d96936e397a98bb16014ff647e5e3fb7142fdd91 100644
--- a/lustre/ost/ost_handler.c
+++ b/lustre/ost/ost_handler.c
@@ -669,7 +669,7 @@ int ost_brw_write(struct ptlrpc_request *req, struct obd_trans_info *oti)
if (client_cksum != cksum) {
CERROR("Bad checksum: client %x, server %x, client NID "
LPX64" (%s)\n", client_cksum, cksum,
- req->rq_connection->c_peer.peer_nid, str);
+ req->rq_connection->c_peer.peer_id.nid, str);
cksum_counter = 1;
repbody->oa.o_cksum = cksum;
} else {
@@ -677,7 +677,7 @@ int ost_brw_write(struct ptlrpc_request *req, struct obd_trans_info *oti)
if ((cksum_counter & (-cksum_counter)) == cksum_counter)
CWARN("Checksum %u from "LPX64": %x OK\n",
cksum_counter,
- req->rq_connection->c_peer.peer_nid,
+ req->rq_connection->c_peer.peer_id.nid,
cksum);
}
}
diff --git a/lustre/portals/include/linux/kp30.h b/lustre/portals/include/linux/kp30.h
index 6ef28a8535e985f668dcbf675399140c6c071516..b13f161638e66d3772c585a06aa9afa5b3da76b2 100644
--- a/lustre/portals/include/linux/kp30.h
+++ b/lustre/portals/include/linux/kp30.h
@@ -326,6 +326,7 @@ void portals_debug_dumplog(void);
/* support decl needed both by kernel and liblustre */
char *portals_nid2str(int nal, ptl_nid_t nid, char *str);
+char *portals_id2str(int nal, ptl_process_id_t nid, char *str);
#ifndef CURRENT_TIME
# define CURRENT_TIME time(0)
@@ -644,7 +645,7 @@ enum {
NAL_ENUM_END_MARKER
};
-#define PTL_NALFMT_SIZE 26 /* %u:%u.%u.%u.%u (10+4+4+4+3+1) */
+#define PTL_NALFMT_SIZE 30 /* %u:%u.%u.%u.%u,%u (10+4+4+4+3+4+1) */
#define NAL_MAX_NR (NAL_ENUM_END_MARKER - 1)
diff --git a/lustre/portals/knals/gmnal/gmnal_api.c b/lustre/portals/knals/gmnal/gmnal_api.c
index 002587d1b5962f3b73acbcb9fc29dde9d0be25ae..bdf5cfe6cfb547f8ed8d3af0b05857456f1c9773 100644
--- a/lustre/portals/knals/gmnal/gmnal_api.c
+++ b/lustre/portals/knals/gmnal/gmnal_api.c
@@ -318,7 +318,7 @@ gmnal_api_startup(nal_t *nal, ptl_pid_t requested_pid,
/*
pid = gm_getpid();
*/
- process_id.pid = 0;
+ process_id.pid = requested_pid;
process_id.nid = global_nid;
CDEBUG(D_INFO, "portals_pid is [%u]\n", process_id.pid);
diff --git a/lustre/portals/knals/ibnal/ibnal.c b/lustre/portals/knals/ibnal/ibnal.c
index 86c2a6355206923d845a8321c343316a9ba7eb7b..c6bd7ed044c1c7f0390168566df82e97da05bf91 100644
--- a/lustre/portals/knals/ibnal/ibnal.c
+++ b/lustre/portals/knals/ibnal/ibnal.c
@@ -246,7 +246,7 @@ kibnal_init(int interface, // no use here
rc = lib_init(&kibnal_lib,
kibnal_data.kib_nid,
- 0, // process id is set as 0
+ requested_pid , // process id is set as requested_pid instead of 0
ptl_size,
ac_size);
diff --git a/lustre/portals/knals/qswnal/qswnal.c b/lustre/portals/knals/qswnal/qswnal.c
index c595450684aadb55fb7bb7baf11f7cc1dc7319ac..38d1636166d57950555ae99a3d4e977220ccee7d 100644
--- a/lustre/portals/knals/qswnal/qswnal.c
+++ b/lustre/portals/knals/qswnal/qswnal.c
@@ -696,7 +696,7 @@ kqswnal_startup (nal_t *nal, ptl_pid_t requested_pid,
/* Network interface ready to initialise */
my_process_id.nid = kqswnal_elanid2nid(kqswnal_data.kqn_elanid);
- my_process_id.pid = 0;
+ my_process_id.pid = requested_pid;
rc = lib_init(&kqswnal_lib, nal, my_process_id,
requested_limits, actual_limits);
@@ -802,7 +802,7 @@ kqswnal_initialise (void)
/* Pure gateways, and the workaround for 'EKC blocks forever until
* the service is active' want the NAL started up at module load
* time... */
- rc = PtlNIInit(QSWNAL, 0, NULL, NULL, &kqswnal_ni);
+ rc = PtlNIInit(QSWNAL, LUSTRE_SRV_PTL_PID, NULL, NULL, &kqswnal_ni);
if (rc != PTL_OK && rc != PTL_IFACE_DUP) {
ptl_unregister_nal(QSWNAL);
return (-ENODEV);
diff --git a/lustre/portals/knals/scimacnal/scimacnal.c b/lustre/portals/knals/scimacnal/scimacnal.c
index e77bd8ee278dc69428c1e5d1525825e6907dfab5..75188e97dcb1b35fd4cc90929b1e8e6665db763f 100644
--- a/lustre/portals/knals/scimacnal/scimacnal.c
+++ b/lustre/portals/knals/scimacnal/scimacnal.c
@@ -205,7 +205,7 @@ static int kscimacnal_startup(nal_t *nal, ptl_pid_t requested_pid,
}
kscimacnal_data.ksci_nid = (ptl_nid_t)(ntohl(mac_physaddr));
- process_id.pid = 0;
+ process_id.pid = requested_pid;
process_id.nid = kscimacnal_data.ksci_nid;
CDEBUG(D_NET, "calling lib_init with nid "LPX64"\n",
diff --git a/lustre/portals/knals/socknal/socknal.c b/lustre/portals/knals/socknal/socknal.c
index 9d39cb1bd522b4ec6e115e5b604adc658f45d98f..3a3629b873d45b64e163b0d8b4123cf72b6ec2b1 100644
--- a/lustre/portals/knals/socknal/socknal.c
+++ b/lustre/portals/knals/socknal/socknal.c
@@ -1589,7 +1589,7 @@ ksocknal_api_startup (nal_t *nal, ptl_pid_t requested_pid,
}
/* NB we have to wait to be told our true NID... */
- process_id.pid = 0;
+ process_id.pid = requested_pid; //LUSTRE_SRV_PTL_PID;
process_id.nid = 0;
rc = lib_init(&ksocknal_lib, nal, process_id,
@@ -1739,7 +1739,7 @@ ksocknal_module_init (void)
}
/* Pure gateways want the NAL started up at module load time... */
- rc = PtlNIInit(SOCKNAL, 0, NULL, NULL, &ksocknal_ni);
+ rc = PtlNIInit(SOCKNAL, LUSTRE_SRV_PTL_PID, NULL, NULL, &ksocknal_ni);
if (rc != PTL_OK && rc != PTL_IFACE_DUP) {
ptl_unregister_nal(SOCKNAL);
return (-ENODEV);
diff --git a/lustre/portals/knals/socknal/socknal.h b/lustre/portals/knals/socknal/socknal.h
index ff73f71ce51cacad07e14ac0e76001f340d4e4a9..e14205951eb4d76ce34272dab93c5e14e74300eb 100644
--- a/lustre/portals/knals/socknal/socknal.h
+++ b/lustre/portals/knals/socknal/socknal.h
@@ -66,6 +66,7 @@
#include <portals/lib-p30.h>
#include <portals/nal.h>
#include <portals/socknal.h>
+#include <linux/lustre_idl.h>
#if CONFIG_SMP
# define SOCKNAL_N_SCHED num_online_cpus() /* # socknal schedulers */
diff --git a/lustre/portals/libcfs/debug.c b/lustre/portals/libcfs/debug.c
index 2b2ce3b99247aa72c1b99ad6659024177f866c22..eb75e60442e470d08c4b96dcd265d376fb500445 100644
--- a/lustre/portals/libcfs/debug.c
+++ b/lustre/portals/libcfs/debug.c
@@ -985,6 +985,30 @@ char *portals_nid2str(int nal, ptl_nid_t nid, char *str)
}
return str;
}
+/* bug #4615 */
+char *portals_id2str(int nal, ptl_process_id_t id, char *str)
+{
+ switch(nal){
+ case TCPNAL:
+ /* userspace NAL */
+ case SOCKNAL:
+ snprintf(str, PTL_NALFMT_SIZE - 1, "%u:%u.%u.%u.%u,%u",
+ (__u32)(id.nid >> 32), HIPQUAD((id.nid)) , id.pid);
+ break;
+ case QSWNAL:
+ case GMNAL:
+ case IBNAL:
+ snprintf(str, PTL_NALFMT_SIZE - 1, "%u:%u,%u",
+ (__u32)(id.nid >> 32), (__u32)id.nid, id.pid);
+ break;
+ default:
+ snprintf(str, PTL_NALFMT_SIZE - 1, "?%d? %llx,%lx",
+ nal, (long long)id.nid, (long)id.pid );
+ break;
+ }
+ return str;
+}
+
#ifdef __KERNEL__
char stack_backtrace[LUSTRE_TRACE_SIZE];
@@ -1077,3 +1101,4 @@ EXPORT_SYMBOL(portals_debug_set_level);
EXPORT_SYMBOL(portals_run_upcall);
EXPORT_SYMBOL(portals_run_lbug_upcall);
EXPORT_SYMBOL(portals_nid2str);
+EXPORT_SYMBOL(portals_id2str);
diff --git a/lustre/portals/portals/lib-move.c b/lustre/portals/portals/lib-move.c
index 9dcc06e343baea58fbeb05faf2f3cfa75bd9c645..854a452e55a4081d8fc9b778972ff4305e81d424 100644
--- a/lustre/portals/portals/lib-move.c
+++ b/lustre/portals/portals/lib-move.c
@@ -82,6 +82,8 @@ lib_match_md(lib_nal_t *nal, int index, int op_mask,
if (me->match_id.nid != PTL_NID_ANY &&
me->match_id.nid != src_nid)
continue;
+
+ CDEBUG(D_NET,"match_id.pid [%x], src_pid [%x]\n", me->match_id.pid, src_pid);
if (me->match_id.pid != PTL_PID_ANY &&
me->match_id.pid != src_pid)
diff --git a/lustre/portals/portals/module.c b/lustre/portals/portals/module.c
index 5615a7244329138a1242d378fefc27ebcd12f8cf..d292a501a057de31388bd608e0c385be870287f5 100644
--- a/lustre/portals/portals/module.c
+++ b/lustre/portals/portals/module.c
@@ -50,6 +50,7 @@
#include <linux/kp30.h>
#include <linux/kpr.h>
#include <linux/portals_compat25.h>
+#include <linux/lustre_idl.h>
extern void (kping_client)(struct portal_ioctl_data *);
@@ -83,7 +84,7 @@ static int kportal_ioctl(struct portal_ioctl_data *data,
CDEBUG (D_IOCTL, "Getting nid for nal [%d]\n", data->ioc_nal);
- err = PtlNIInit(data->ioc_nal, 0, NULL, NULL, &nih);
+ err = PtlNIInit(data->ioc_nal, LUSTRE_SRV_PTL_PID, NULL, NULL, &nih);
if (!(err == PTL_OK || err == PTL_IFACE_DUP))
RETURN (-EINVAL);
@@ -104,7 +105,7 @@ static int kportal_ioctl(struct portal_ioctl_data *data,
CDEBUG (D_IOCTL, "fail nid: [%d] "LPU64" count %d\n",
data->ioc_nal, data->ioc_nid, data->ioc_count);
- err = PtlNIInit(data->ioc_nal, 0, NULL, NULL, &nih);
+ err = PtlNIInit(data->ioc_nal, LUSTRE_SRV_PTL_PID, NULL, NULL, &nih);
if (!(err == PTL_OK || err == PTL_IFACE_DUP))
return (-EINVAL);
diff --git a/lustre/portals/unals/tcpnal.c b/lustre/portals/unals/tcpnal.c
index 34a9c9d5e3d43b6a57deccb423d7e116b0e2f70a..6e9cca99ca7647fb842cd3c3a592f4e1a46ee3ec 100644
--- a/lustre/portals/unals/tcpnal.c
+++ b/lustre/portals/unals/tcpnal.c
@@ -252,7 +252,7 @@ int tcpnal_init(bridge b)
return(PTL_NAL_FAILED);
}
/* XXX cfs hack */
- b->lib_nal->libnal_ni.ni_pid.pid=0;
+// b->lib_nal->libnal_ni.ni_pid.pid=0;
b->lower=m;
return(PTL_OK);
}
diff --git a/lustre/ptlrpc/connection.c b/lustre/ptlrpc/connection.c
index 2e0d889da6fd6386ac9b2a8e072a3c061ffa39af..646cb0721badd79a80f87a3e10fc9028157d478f 100644
--- a/lustre/ptlrpc/connection.c
+++ b/lustre/ptlrpc/connection.c
@@ -62,12 +62,12 @@ struct ptlrpc_connection *ptlrpc_get_connection(struct ptlrpc_peer *peer,
CDEBUG(D_INFO, "peer is %s on %s\n",
- ptlrpc_peernid2str(peer, str), peer->peer_ni->pni_name);
+ ptlrpc_id2str(peer, str), peer->peer_ni->pni_name);
spin_lock(&conn_lock);
list_for_each(tmp, &conn_list) {
c = list_entry(tmp, struct ptlrpc_connection, c_link);
- if (peer->peer_nid == c->c_peer.peer_nid &&
+ if (!memcmp(peer, &c->c_peer, sizeof(struct ptlrpc_peer)) &&
peer->peer_ni == c->c_peer.peer_ni) {
ptlrpc_connection_addref(c);
GOTO(out, c);
@@ -76,7 +76,7 @@ struct ptlrpc_connection *ptlrpc_get_connection(struct ptlrpc_peer *peer,
list_for_each_safe(tmp, pos, &conn_unused_list) {
c = list_entry(tmp, struct ptlrpc_connection, c_link);
- if (peer->peer_nid == c->c_peer.peer_nid &&
+ if (!memcmp(peer, &c->c_peer, sizeof(struct ptlrpc_peer)) &&
peer->peer_ni == c->c_peer.peer_ni) {
ptlrpc_connection_addref(c);
list_del(&c->c_link);
diff --git a/lustre/ptlrpc/events.c b/lustre/ptlrpc/events.c
index 254ae3093b16e722b770167a25f2f54d7e246367..8ea1813f70f34ea6faa07d583ca224c016f51b1a 100644
--- a/lustre/ptlrpc/events.c
+++ b/lustre/ptlrpc/events.c
@@ -197,8 +197,8 @@ void request_in_callback(ptl_event_t *ev)
CERROR("Can't allocate incoming request descriptor: "
"Dropping %s RPC from %s\n",
service->srv_name,
- portals_nid2str(srv_ni->sni_ni->pni_number,
- ev->initiator.nid, str));
+ portals_id2str(srv_ni->sni_ni->pni_number,
+ ev->initiator, str));
return;
}
}
@@ -212,7 +212,7 @@ void request_in_callback(ptl_event_t *ev)
ev->ni_fail_type == PTL_NI_OK)
req->rq_reqlen = ev->mlength;
do_gettimeofday(&req->rq_arrival_time);
- req->rq_peer.peer_nid = ev->initiator.nid;
+ req->rq_peer.peer_id = ev->initiator;
req->rq_peer.peer_ni = rqbd->rqbd_srv_ni->sni_ni;
req->rq_rqbd = rqbd;
@@ -361,7 +361,8 @@ int ptlrpc_uuid_to_peer (struct obd_uuid *uuid, struct ptlrpc_peer *peer)
pni = &ptlrpc_interfaces[i];
if (pni->pni_number == peer_nal) {
- peer->peer_nid = peer_nid;
+ peer->peer_id.nid = peer_nid;
+ peer->peer_id.pid = LUSTRE_SRV_PTL_PID; //#4165:only client will call this func.
peer->peer_ni = pni;
return (0);
}
@@ -409,20 +410,37 @@ void ptlrpc_ni_fini(struct ptlrpc_ni *pni)
/* notreached */
}
+ptl_pid_t ptl_get_pid(void)
+{
+ ptl_pid_t pid;
+
+#ifndef __KERNEL__
+ pid = getpid();
+#else
+ pid = LUSTRE_SRV_PTL_PID;
+#endif
+ return pid;
+}
+
int ptlrpc_ni_init(int number, char *name, struct ptlrpc_ni *pni)
{
int rc;
char str[20];
ptl_handle_ni_t nih;
-
+ ptl_pid_t pid;
+
+ pid = ptl_get_pid();
+
/* We're not passing any limits yet... */
- rc = PtlNIInit(number, 0, NULL, NULL, &nih);
+ rc = PtlNIInit(number, pid, NULL, NULL, &nih);
if (rc != PTL_OK && rc != PTL_IFACE_DUP) {
CDEBUG (D_NET, "Can't init network interface %s: %d\n",
name, rc);
return (-ENOENT);
}
+ CDEBUG(D_NET, "My pid is: %x\n", ptl_get_pid());
+
PtlSnprintHandle(str, sizeof(str), nih);
CDEBUG (D_NET, "init %d %s: %s\n", number, name, str);
diff --git a/lustre/ptlrpc/niobuf.c b/lustre/ptlrpc/niobuf.c
index b6282164d298fce6572259ebb8020827db5c0906..e8e46607ae54a269d01dadc2638dc3e87e369927 100644
--- a/lustre/ptlrpc/niobuf.c
+++ b/lustre/ptlrpc/niobuf.c
@@ -34,7 +34,6 @@ static int ptl_send_buf (ptl_handle_md_t *mdh, void *base, int len,
ptl_ack_req_t ack, struct ptlrpc_cb_id *cbid,
struct ptlrpc_connection *conn, int portal, __u64 xid)
{
- ptl_process_id_t remote_id;
int rc;
int rc2;
ptl_md_t md;
@@ -43,14 +42,10 @@ static int ptl_send_buf (ptl_handle_md_t *mdh, void *base, int len,
LASSERT (portal != 0);
LASSERT (conn != NULL);
- CDEBUG (D_INFO, "conn=%p ni %s nid %s on %s\n",
+ CDEBUG (D_INFO, "conn=%p ni %s id %s on %s\n",
conn, conn->c_peer.peer_ni->pni_name,
- ptlrpc_peernid2str(&conn->c_peer, str),
+ ptlrpc_id2str(&conn->c_peer, str),
conn->c_peer.peer_ni->pni_name);
-
- remote_id.nid = conn->c_peer.peer_nid,
- remote_id.pid = 0;
-
md.start = base;
md.length = len;
md.threshold = (ack == PTL_ACK_REQ) ? 2 : 1;
@@ -76,13 +71,13 @@ static int ptl_send_buf (ptl_handle_md_t *mdh, void *base, int len,
CDEBUG(D_NET, "Sending %d bytes to portal %d, xid "LPD64"\n",
len, portal, xid);
- rc2 = PtlPut (*mdh, ack, remote_id, portal, 0, xid, 0, 0);
+ rc = PtlPut (*mdh, ack, conn->c_peer.peer_id, portal, 0, xid, 0, 0);
if (rc != PTL_OK) {
/* We're going to get an UNLINK event when I unlink below,
* which will complete just like any other failed send, so
* I fall through and return success here! */
CERROR("PtlPut(%s, %d, "LPD64") failed: %d\n",
- ptlrpc_peernid2str(&conn->c_peer, str),
+ ptlrpc_id2str(&conn->c_peer, str),
portal, xid, rc);
rc2 = PtlMDUnlink(*mdh);
LASSERT (rc2 == PTL_OK);
@@ -96,7 +91,6 @@ int ptlrpc_start_bulk_transfer (struct ptlrpc_bulk_desc *desc)
int rc;
int rc2;
struct ptlrpc_peer *peer;
- ptl_process_id_t remote_id;
ptl_md_t md;
__u64 xid;
char str[PTL_NALFMT_SIZE];
@@ -134,22 +128,19 @@ int ptlrpc_start_bulk_transfer (struct ptlrpc_bulk_desc *desc)
/* Client's bulk and reply matchbits are the same */
xid = desc->bd_req->rq_xid;
- remote_id.nid = peer->peer_nid;
- remote_id.pid = 0;
-
CDEBUG(D_NET, "Transferring %u pages %u bytes via portal %d on %s "
"nid %s pid %d xid "LPX64"\n", desc->bd_iov_count,
desc->bd_nob, desc->bd_portal, peer->peer_ni->pni_name,
- ptlrpc_peernid2str(peer, str), remote_id.pid, xid);
+ ptlrpc_id2str(peer, str), peer->peer_id.pid, xid);
/* Network is about to get at the memory */
desc->bd_network_rw = 1;
if (desc->bd_type == BULK_PUT_SOURCE)
- rc = PtlPut (desc->bd_md_h, PTL_ACK_REQ, remote_id,
+ rc = PtlPut (desc->bd_md_h, PTL_ACK_REQ, peer->peer_id,
desc->bd_portal, 0, xid, 0, 0);
else
- rc = PtlGet (desc->bd_md_h, remote_id,
+ rc = PtlGet (desc->bd_md_h, peer->peer_id,
desc->bd_portal, 0, xid, 0);
if (rc != PTL_OK) {
@@ -157,7 +148,7 @@ int ptlrpc_start_bulk_transfer (struct ptlrpc_bulk_desc *desc)
* event this creates will signal completion with failure,
* so we return SUCCESS here! */
CERROR("Transfer(%s, %d, "LPX64") failed: %d\n",
- ptlrpc_peernid2str(peer, str),
+ ptlrpc_id2str(peer, str),
desc->bd_portal, xid, rc);
rc2 = PtlMDUnlink(desc->bd_md_h);
LASSERT (rc2 == PTL_OK);
@@ -205,7 +196,6 @@ int ptlrpc_register_bulk (struct ptlrpc_request *req)
struct ptlrpc_peer *peer;
int rc;
int rc2;
- ptl_process_id_t source_id;
ptl_handle_me_t me_h;
ptl_md_t md;
ENTRY;
@@ -242,13 +232,10 @@ int ptlrpc_register_bulk (struct ptlrpc_request *req)
LASSERT (!desc->bd_registered || req->rq_xid != desc->bd_last_xid);
desc->bd_registered = 1;
desc->bd_last_xid = req->rq_xid;
-
- source_id.nid = desc->bd_import->imp_connection->c_peer.peer_nid;
- source_id.pid = PTL_PID_ANY;
-
+
rc = PtlMEAttach(peer->peer_ni->pni_ni_h,
- desc->bd_portal, source_id, req->rq_xid, 0,
- PTL_UNLINK, PTL_INS_AFTER, &me_h);
+ desc->bd_portal, desc->bd_import->imp_connection->c_peer.peer_id,
+ req->rq_xid, 0, PTL_UNLINK, PTL_INS_AFTER, &me_h);
if (rc != PTL_OK) {
CERROR("PtlMEAttach failed: %d\n", rc);
LASSERT (rc == PTL_NO_SPACE);
@@ -397,7 +384,6 @@ int ptl_send_rpc(struct ptlrpc_request *request)
int rc2;
struct ptlrpc_connection *connection;
unsigned long flags;
- ptl_process_id_t source_id;
ptl_handle_me_t reply_me_h;
ptl_md_t reply_md;
ENTRY;
@@ -419,10 +405,7 @@ int ptl_send_rpc(struct ptlrpc_request *request)
request->rq_reqmsg->handle = request->rq_import->imp_remote_handle;
request->rq_reqmsg->type = PTL_RPC_MSG_REQUEST;
request->rq_reqmsg->conn_cnt = request->rq_import->imp_conn_cnt;
-
- source_id.nid = connection->c_peer.peer_nid;
- source_id.pid = PTL_PID_ANY;
-
+
LASSERT (request->rq_replen != 0);
if (request->rq_repmsg == NULL)
OBD_ALLOC(request->rq_repmsg, request->rq_replen);
@@ -431,7 +414,7 @@ int ptl_send_rpc(struct ptlrpc_request *request)
rc = PtlMEAttach(connection->c_peer.peer_ni->pni_ni_h,
request->rq_reply_portal, /* XXX FIXME bug 249 */
- source_id, request->rq_xid, 0, PTL_UNLINK,
+ connection->c_peer.peer_id, request->rq_xid, 0, PTL_UNLINK,
PTL_INS_AFTER, &reply_me_h);
if (rc != PTL_OK) {
CERROR("PtlMEAttach failed: %d\n", rc);