1
1
openmpi/ompi/mca/pml/dr/pml_dr_recvfrag.c
Ralph Castain 9613b3176c Effectively revert the orte_output system and return to direct use of opal_output at all levels. Retain the orte_show_help subsystem to allow aggregation of show_help messages at the HNP.
After much work by Jeff and myself, and quite a lot of discussion, it has become clear that we simply cannot resolve the infinite loops caused by RML-involved subsystems calling orte_output. The original rationale for the change to orte_output has also been reduced by shifting the output of XML-formatted vs human readable messages to an alternative approach.

I have globally replaced the orte_output/ORTE_OUTPUT calls in the code base, as well as the corresponding .h file name. I have test compiled and run this on the various environments within my reach, so hopefully this will prove minimally disruptive.

This commit was SVN r18619.
2008-06-09 14:53:58 +00:00

1009 строки
41 KiB
C

/* -*- Mode: C; c-basic-offset:4 ; -*- */
/*
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2007 The University of Tennessee and The University
* of Tennessee Research Foundation. All rights
* reserved.
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
* University of Stuttgart. All rights reserved.
* Copyright (c) 2004-2006 The Regents of the University of California.
* All rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*/
/**
* @file
*/
#include "ompi_config.h"
#include "opal/class/opal_list.h"
#include "opal/threads/mutex.h"
#include "opal/util/crc.h"
#include "ompi/constants.h"
#include "ompi/communicator/communicator.h"
#include "ompi/datatype/datatype.h"
#include "ompi/mca/pml/pml.h"
#include "ompi/class/ompi_seq_tracker.h"
#include "pml_dr.h"
#include "pml_dr_comm.h"
#include "pml_dr_recvfrag.h"
#include "pml_dr_recvreq.h"
#include "pml_dr_sendreq.h"
#include "pml_dr_hdr.h"
#include "orte/mca/errmgr/errmgr.h"
#define MCA_PML_DR_HDR_VALIDATE_ACK(do_csum, hdr, type) \
do { \
mca_pml_dr_endpoint_t* ep; \
if(do_csum) { \
uint16_t csum = (uint16_t)opal_csum(hdr, sizeof(type)); \
if(hdr->hdr_common.hdr_csum != csum) { \
MCA_PML_DR_DEBUG(0, (0, "%s:%d: invalid header checksum: 0x%04x != 0x%04x\n", \
__FILE__, __LINE__, hdr->hdr_common.hdr_csum, csum)); \
return; \
} \
} \
ep = (mca_pml_dr_endpoint_t*)opal_pointer_array_get_item(&mca_pml_dr.endpoints, hdr->hdr_common.hdr_src); \
assert(ep != NULL); \
if(ompi_seq_tracker_check_duplicate(&ep->seq_sends, hdr->hdr_common.hdr_vid)) { \
MCA_PML_DR_DEBUG(0, (0, "%s:%d: dropping duplicate ack, vfrag ID %d", \
__FILE__, __LINE__, hdr->hdr_common.hdr_vid)); \
return; \
} \
MCA_PML_DR_DEBUG(1, (0, "%s:%d: couldn't find vfrag ID %d \n", \
__FILE__, __LINE__, hdr->hdr_common.hdr_vid)); \
} while (0)
OBJ_CLASS_INSTANCE(
mca_pml_dr_buffer_t,
ompi_free_list_item_t,
NULL,
NULL
);
OBJ_CLASS_INSTANCE(
mca_pml_dr_recv_frag_t,
ompi_free_list_item_t,
NULL,
NULL
);
/*
* Release resources.
*/
static void mca_pml_dr_ctl_completion(
mca_btl_base_module_t* btl,
struct mca_btl_base_endpoint_t* ep,
struct mca_btl_base_descriptor_t* des,
int status)
{
mca_bml_base_btl_t* bml_btl = (mca_bml_base_btl_t*)des->des_context;
mca_bml_base_free(bml_btl, des);
}
/**
* Callback from BTL on receive.
*/
void mca_pml_dr_recv_frag_callback(
mca_btl_base_module_t* btl,
mca_btl_base_tag_t tag,
mca_btl_base_descriptor_t* des,
void* cbdata)
{
mca_btl_base_segment_t* segments = des->des_dst;
mca_pml_dr_hdr_t* hdr = (mca_pml_dr_hdr_t*)segments->seg_addr.pval;
mca_pml_dr_comm_t *comm;
mca_pml_dr_comm_proc_t *proc;
mca_pml_dr_endpoint_t *ep;
ompi_communicator_t* ompi_comm;
uint16_t csum;
bool do_csum = mca_pml_dr.enable_csum &&
(btl->btl_flags & MCA_BTL_FLAGS_NEED_CSUM);
if(segments->seg_len < sizeof(mca_pml_dr_common_hdr_t)) {
MCA_PML_DR_DEBUG(0,(0, "%s:%d: wtf? segments->seg_len:%d < sizeof(mca_pml_dr_common_hdr_t):%lu\n",
__FILE__, __LINE__, segments->seg_len, (unsigned long)sizeof(mca_pml_dr_common_hdr_t)));
return;
}
MCA_PML_DR_DEBUG(0,(0, "%s:%d: got a hdr of type %d\n",
__FILE__, __LINE__, hdr->hdr_common.hdr_type));
switch(hdr->hdr_common.hdr_type) {
case MCA_PML_DR_HDR_TYPE_MATCH:
{
if(do_csum) {
csum = (uint16_t)opal_csum(hdr, sizeof(mca_pml_dr_match_hdr_t));
if(hdr->hdr_common.hdr_csum != csum) {
MCA_PML_DR_DEBUG(0,(0, "%s:%d: invalid header checksum: 0x%04x != 0x%04x\n",
__FILE__, __LINE__, hdr->hdr_common.hdr_csum, csum));
return;
}
}
if(hdr->hdr_common.hdr_dst != mca_pml_dr.my_rank ) {
MCA_PML_DR_DEBUG(0, (0, "%s:%d: misdelivered packet [rank %d -> rank %d]\n",
__FILE__, __LINE__, hdr->hdr_common.hdr_src, hdr->hdr_common.hdr_dst));
return;
}
ep = (mca_pml_dr_endpoint_t*)opal_pointer_array_get_item(&mca_pml_dr.endpoints, hdr->hdr_common.hdr_src);
assert(ep != NULL);
if(ompi_seq_tracker_check_duplicate(&ep->seq_recvs, hdr->hdr_common.hdr_vid)) {
MCA_PML_DR_DEBUG(0,(0, "%s:%d: got a duplicate vfrag vfrag id %d\n",
__FILE__, __LINE__, hdr->hdr_common.hdr_vid));
mca_pml_dr_recv_frag_ack(btl,
ep->bml_endpoint,
&hdr->hdr_common,
hdr->hdr_match.hdr_src_ptr.pval,
1, 0);
return;
}
ompi_comm = ompi_comm_lookup(hdr->hdr_common.hdr_ctx);
if(NULL == ompi_comm ) {
OPAL_OUTPUT((0, "%s:%d: invalid communicator %d\n",
__FILE__, __LINE__, hdr->hdr_common.hdr_ctx));
return;
}
comm = (mca_pml_dr_comm_t*)ompi_comm->c_pml_comm;
assert(hdr->hdr_common.hdr_src < opal_pointer_array_get_size(&comm->sparse_procs));
proc = (mca_pml_dr_comm_proc_t*)opal_pointer_array_get_item(&comm->sparse_procs, hdr->hdr_common.hdr_src);
assert(proc != NULL);
assert(ep == proc->pml_endpoint);
mca_pml_dr_recv_frag_match(comm,proc,btl,&hdr->hdr_match,segments,des->des_dst_cnt);
break;
}
case MCA_PML_DR_HDR_TYPE_MATCH_ACK:
{
MCA_PML_DR_HDR_VALIDATE_ACK(do_csum, hdr, mca_pml_dr_ack_hdr_t);
mca_pml_dr_send_request_match_ack(btl, &hdr->hdr_ack);
break;
}
case MCA_PML_DR_HDR_TYPE_RNDV:
{
if(do_csum) {
csum = (uint16_t)opal_csum(hdr, sizeof(mca_pml_dr_rendezvous_hdr_t));
if(hdr->hdr_common.hdr_csum != csum) {
MCA_PML_DR_DEBUG(0, (0, "%s:%d: invalid header checksum: 0x%04x != 0x%04x\n",
__FILE__, __LINE__, hdr->hdr_common.hdr_csum, csum));
return;
}
}
if(hdr->hdr_common.hdr_dst != mca_pml_dr.my_rank ) {
MCA_PML_DR_DEBUG(0, (0, "%s:%d: misdelivered packet [rank %d -> rank %d]\n",
__FILE__, __LINE__, hdr->hdr_common.hdr_src, hdr->hdr_common.hdr_dst));
return;
}
ep = (mca_pml_dr_endpoint_t*)opal_pointer_array_get_item(&mca_pml_dr.endpoints, hdr->hdr_common.hdr_src);
assert(ep != NULL);
/* seq_recvs protected by matching lock */
if(ompi_seq_tracker_check_duplicate(&ep->seq_recvs, hdr->hdr_common.hdr_vid)){
mca_pml_dr_recv_request_t* recvreq;
ompi_comm = ompi_comm_lookup(hdr->hdr_common.hdr_ctx);
if(NULL == ompi_comm) {
if(ompi_seq_tracker_check_duplicate(&ep->seq_recvs_matched, hdr->hdr_common.hdr_vid)) {
MCA_PML_DR_DEBUG(0, (0, "%s:%d: acking duplicate matched rendezvous from sequence tracker\n",
__FILE__, __LINE__));
mca_pml_dr_recv_frag_ack(btl,
ep->bml_endpoint,
&hdr->hdr_common,
hdr->hdr_match.hdr_src_ptr.pval,
~(uint64_t) 0, hdr->hdr_rndv.hdr_msg_length);
return;
} else {
OPAL_OUTPUT((0, "%s:%d: the world as we know it is bad\n", __FILE__, __LINE__));
orte_errmgr.abort(-1, NULL);
}
}
comm = (mca_pml_dr_comm_t*)ompi_comm->c_pml_comm;
assert(hdr->hdr_common.hdr_src < opal_pointer_array_get_size(&comm->sparse_procs));
proc = (mca_pml_dr_comm_proc_t*)opal_pointer_array_get_item(&comm->sparse_procs, hdr->hdr_common.hdr_src);
assert(proc != NULL);
assert(ep == proc->pml_endpoint);
/* ack only if the vfrag has been matched */
recvreq =
mca_pml_dr_comm_proc_check_matched(proc, hdr->hdr_common.hdr_vid);
if(NULL != recvreq) {
MCA_PML_DR_DEBUG(0,(0, "%s:%d: acking duplicate matched rendezvous from pending matched vfrag id %d\n",
__FILE__, __LINE__, hdr->hdr_common.hdr_vid));
mca_pml_dr_recv_request_ack(btl, recvreq, &hdr->hdr_common,
hdr->hdr_match.hdr_src_ptr, recvreq->req_bytes_received, 1);
} else {
if(ompi_seq_tracker_check_duplicate(&ep->seq_recvs_matched, hdr->hdr_common.hdr_vid)) {
MCA_PML_DR_DEBUG(0,(0, "%s:%d: acking duplicate matched rendezvous from sequence tracker\n", __FILE__, __LINE__));
mca_pml_dr_recv_frag_ack(btl,
ep->bml_endpoint,
&hdr->hdr_common,
hdr->hdr_match.hdr_src_ptr.pval,
~(uint64_t) 0, hdr->hdr_rndv.hdr_msg_length);
} else {
MCA_PML_DR_DEBUG(0,(0, "%s:%d: droping duplicate unmatched rendezvous\n", __FILE__, __LINE__));
}
}
} else {
ompi_comm = ompi_comm_lookup(hdr->hdr_common.hdr_ctx);
if(NULL == ompi_comm) {
OPAL_OUTPUT((0, "%s:%d: the world as we know it is bad\n", __FILE__, __LINE__));
orte_errmgr.abort(-1, NULL);
}
comm = (mca_pml_dr_comm_t*)ompi_comm->c_pml_comm;
assert(hdr->hdr_common.hdr_src < opal_pointer_array_get_size(&comm->sparse_procs));
proc = (mca_pml_dr_comm_proc_t*)opal_pointer_array_get_item(&comm->sparse_procs, hdr->hdr_common.hdr_src);
assert(proc != NULL);
assert(ep == proc->pml_endpoint);
mca_pml_dr_recv_frag_match(comm,proc,btl,&hdr->hdr_match,segments,des->des_dst_cnt);
}
break;
}
case MCA_PML_DR_HDR_TYPE_RNDV_ACK:
{
MCA_PML_DR_HDR_VALIDATE_ACK(do_csum, hdr, mca_pml_dr_ack_hdr_t);
mca_pml_dr_send_request_rndv_ack(btl, &hdr->hdr_ack);
break;
}
case MCA_PML_DR_HDR_TYPE_FRAG:
{
mca_pml_dr_recv_request_t* recvreq;
if(do_csum) {
csum = (uint16_t)opal_csum(hdr, sizeof(mca_pml_dr_frag_hdr_t));
if(hdr->hdr_common.hdr_csum != csum) {
MCA_PML_DR_DEBUG(0,(0, "%s:%d: invalid header checksum: 0x%04x != 0x%04x\n",
__FILE__, __LINE__, hdr->hdr_common.hdr_csum, csum));
return;
}
}
if(hdr->hdr_common.hdr_dst != mca_pml_dr.my_rank ) {
MCA_PML_DR_DEBUG(0,(0, "%s:%d: misdelivered packet [rank %d -> rank %d]\n",
__FILE__, __LINE__, hdr->hdr_common.hdr_src, hdr->hdr_common.hdr_dst));
return;
}
ep = (mca_pml_dr_endpoint_t*)opal_pointer_array_get_item(&mca_pml_dr.endpoints, hdr->hdr_common.hdr_src);
assert(ep != NULL);
/* seq_recvs protected by matching lock */
if(ompi_seq_tracker_check_duplicate(&ep->seq_recvs, hdr->hdr_common.hdr_vid)) {
MCA_PML_DR_DEBUG(0,(0, "%s:%d: acking duplicate fragment\n", __FILE__, __LINE__));
mca_pml_dr_recv_frag_ack(btl,
ep->bml_endpoint,
&hdr->hdr_common,
hdr->hdr_frag.hdr_src_ptr.pval,
~(uint64_t) 0, 0);
} else {
ompi_comm = ompi_comm_lookup(hdr->hdr_common.hdr_ctx);
if(NULL == ompi_comm) {
MCA_PML_DR_DEBUG(0,(0, "%s:%d: the world as we know it is bad\n", __FILE__, __LINE__));
orte_errmgr.abort(-1, NULL);
}
comm = (mca_pml_dr_comm_t*)ompi_comm->c_pml_comm;
assert(hdr->hdr_common.hdr_src < opal_pointer_array_get_size(&comm->sparse_procs));
proc = (mca_pml_dr_comm_proc_t*)opal_pointer_array_get_item(&comm->sparse_procs, hdr->hdr_common.hdr_src);
assert(proc != NULL);
assert(ep == proc->pml_endpoint);
recvreq = (mca_pml_dr_recv_request_t*)hdr->hdr_frag.hdr_dst_ptr.pval;
mca_pml_dr_recv_request_progress(recvreq,btl,segments,des->des_dst_cnt);
}
break;
}
case MCA_PML_DR_HDR_TYPE_FRAG_ACK:
{
MCA_PML_DR_HDR_VALIDATE_ACK(do_csum, hdr, mca_pml_dr_ack_hdr_t);
mca_pml_dr_send_request_frag_ack(btl, &hdr->hdr_ack);
break;
}
default:
MCA_PML_DR_DEBUG(0,(0, "%s:%d: dropping unknown header type\n", __FILE__,__LINE__));
break;
}
}
/**
* Try and match the incoming message fragment to the list of
* "wild" receives
*
* @param hdr Matching data from recived fragment (IN)
*
* @param pml_comm Pointer to the communicator structure used for
* matching purposes. (IN)
*
* @return Matched receive
*
* This routine assumes that the appropriate matching locks are
* set by the upper level routine.
*/
#define MCA_PML_DR_CHECK_WILD_RECEIVES_FOR_MATCH(hdr,comm,proc,return_match) \
do { \
/* local parameters */ \
opal_list_t* wild_receives = &comm->wild_receives; \
mca_pml_dr_recv_request_t *wild_recv; \
int frag_tag,recv_tag; \
\
/* initialization */ \
frag_tag=hdr->hdr_tag; \
\
/* \
* Loop over the wild irecvs - no need to lock, the upper level \
* locking is protecting from having other threads trying to \
* change this list. \
*/ \
for(wild_recv = (mca_pml_dr_recv_request_t *) \
opal_list_get_first(wild_receives); \
wild_recv != (mca_pml_dr_recv_request_t *) \
opal_list_get_end(wild_receives); \
wild_recv = (mca_pml_dr_recv_request_t *) \
((opal_list_item_t *)wild_recv)->opal_list_next) { \
\
recv_tag = wild_recv->req_recv.req_base.req_tag; \
if ( \
/* exact tag match */ \
(frag_tag == recv_tag) || \
/* wild tag match - negative tags (except for \
* OMPI_ANY_TAG) are reserved for internal use, and will \
* not be matched with OMPI_ANY_TAG */ \
( (recv_tag == OMPI_ANY_TAG) && (0 <= frag_tag) ) ) \
\
{ \
/* \
* Mark that this is the matching irecv, and go to process it. \
*/ \
return_match = wild_recv; \
\
/* remove this irecv from the postd wild ireceive list */ \
opal_list_remove_item(wild_receives, \
(opal_list_item_t *)wild_recv); \
\
/* found match - no need to continue */ \
break; \
} \
} \
} while(0)
/**
* Try and match the incoming message fragment to the list of
* "specific" receives
*
* @param hdr Matching data from recived fragment (IN)
*
* @param comm Pointer to the communicator structure used for
* matching purposes. (IN)
*
* @return Matched receive
*
* This routine assumes that the appropriate matching locks are
* set by the upper level routine.
*/
#define MCA_PML_DR_CHECK_SPECIFIC_RECEIVES_FOR_MATCH(hdr,comm,proc,return_match) \
do { \
/* local variables */ \
opal_list_t* specific_receives = &proc->specific_receives; \
mca_pml_dr_recv_request_t *specific_recv; \
int recv_tag,frag_tag; \
\
/* initialization */ \
frag_tag=hdr->hdr_tag; \
\
/* \
* Loop over the specific irecvs. \
*/ \
for(specific_recv = (mca_pml_dr_recv_request_t *) \
opal_list_get_first(specific_receives); \
specific_recv != (mca_pml_dr_recv_request_t *) \
opal_list_get_end(specific_receives); \
specific_recv = (mca_pml_dr_recv_request_t *) \
((opal_list_item_t *)specific_recv)->opal_list_next) { \
/* \
* Check for a match \
*/ \
recv_tag = specific_recv->req_recv.req_base.req_tag; \
if ( (frag_tag == recv_tag) || \
( (recv_tag == OMPI_ANY_TAG) && (0 <= frag_tag) ) ) { \
\
/* \
* Match made \
*/ \
return_match = specific_recv; \
\
/* remove descriptor from posted specific ireceive list */ \
opal_list_remove_item(specific_receives, \
(opal_list_item_t *)specific_recv); \
\
break; \
} \
} \
} while(0)
/**
* Try and match the incoming message fragment to the list of
* "wild" receives and "specific" receives. Used when both types
* of receives have been posted, i.e. when we need to coordinate
* between multiple lists to make sure ordered delivery occurs.
*
* @param hdr Matching data from recived fragment (IN)
*
* @param comm Pointer to the communicator structure used for
* matching purposes. (IN)
*
* @return Matched receive
*
* This routine assumes that the appropriate matching locks are
* set by the upper level routine.
*/
#define MCA_PML_DR_CHECK_SPECIFIC_AND_WILD_RECEIVES_FOR_MATCH( \
hdr,comm,proc,return_match) \
do { \
/* local variables */ \
mca_pml_dr_recv_request_t *specific_recv, *wild_recv; \
mca_pml_sequence_t wild_recv_seq, specific_recv_seq; \
int frag_tag, wild_recv_tag, specific_recv_tag; \
\
/* initialization */ \
frag_tag=hdr->hdr_tag; \
\
/* \
* We know that when this is called, both specific and wild irecvs \
* have been posted. \
*/ \
specific_recv = (mca_pml_dr_recv_request_t *) \
opal_list_get_first(&(proc)->specific_receives); \
wild_recv = (mca_pml_dr_recv_request_t *) \
opal_list_get_first(&comm->wild_receives); \
\
specific_recv_seq = specific_recv->req_recv.req_base.req_sequence; \
wild_recv_seq = wild_recv->req_recv.req_base.req_sequence; \
\
while (true) { \
if (wild_recv_seq < specific_recv_seq) { \
/* \
* wild recv is earlier than the specific one. \
*/ \
/* \
* try and match \
*/ \
wild_recv_tag = wild_recv->req_recv.req_base.req_tag; \
if ( (frag_tag == wild_recv_tag) || \
( (wild_recv_tag == OMPI_ANY_TAG) && (0 <= frag_tag) ) ) { \
/* \
* Match made \
*/ \
return_match=wild_recv; \
\
/* remove this recv from the wild receive queue */ \
opal_list_remove_item(&comm->wild_receives, \
(opal_list_item_t *)wild_recv); \
break; \
} \
\
/* \
* No match, go to the next. \
*/ \
wild_recv=(mca_pml_dr_recv_request_t *) \
((opal_list_item_t *)wild_recv)->opal_list_next; \
\
/* \
* If that was the last wild one, just look at the \
* rest of the specific ones. \
*/ \
if (wild_recv == (mca_pml_dr_recv_request_t *) \
opal_list_get_end(&comm->wild_receives) ) \
{ \
MCA_PML_DR_CHECK_SPECIFIC_RECEIVES_FOR_MATCH(hdr, comm, proc, return_match); \
break; \
} \
\
/* \
* Get the sequence number for this recv, and go \
* back to the top of the loop. \
*/ \
wild_recv_seq = wild_recv->req_recv.req_base.req_sequence; \
\
} else { \
/* \
* specific recv is earlier than the wild one. \
*/ \
specific_recv_tag=specific_recv->req_recv.req_base.req_tag; \
if ( (frag_tag == specific_recv_tag) || \
( (specific_recv_tag == OMPI_ANY_TAG) && (0<=frag_tag)) ) \
{ \
/* \
* Match made \
*/ \
return_match = specific_recv; \
/* remove descriptor from specific receive list */ \
opal_list_remove_item(&(proc)->specific_receives, \
(opal_list_item_t *)specific_recv); \
break; \
} \
\
/* \
* No match, go on to the next specific irecv. \
*/ \
specific_recv = (mca_pml_dr_recv_request_t *) \
((opal_list_item_t *)specific_recv)->opal_list_next; \
\
/* \
* If that was the last specific irecv, process the \
* rest of the wild ones. \
*/ \
if (specific_recv == (mca_pml_dr_recv_request_t *) \
opal_list_get_end(&(proc)->specific_receives)) \
{ \
MCA_PML_DR_CHECK_WILD_RECEIVES_FOR_MATCH(hdr, comm, proc, return_match); \
break; \
} \
/* \
* Get the sequence number for this recv, and go \
* back to the top of the loop. \
*/ \
specific_recv_seq = specific_recv->req_recv.req_base.req_sequence; \
} \
} \
} while(0)
/*
* Specialized matching routines for internal use only.
*/
static bool mca_pml_dr_check_cantmatch_for_match(
opal_list_t *additional_matches,
mca_pml_dr_comm_t* comm,
mca_pml_dr_comm_proc_t *proc);
/**
* RCS/CTS receive side matching
*
* @param hdr list of parameters needed for matching
* This list is also embeded in frag,
* but this allows to save a memory copy when
* a match is made in this routine. (IN)
* @param frag pointer to receive fragment which we want
* to match (IN/OUT). If a match is not made,
* hdr is copied to frag.
* @param match_made parameter indicating if we matched frag/
* hdr (OUT)
* @param additional_matches if a match is made with frag, we
* may be able to match fragments that previously
* have arrived out-of-order. If this is the
* case, the associated fragment descriptors are
* put on this list for further processing. (OUT)
*
* @return OMPI error code
*
* This routine is used to try and match a newly arrived message fragment
* to pre-posted receives. The following assumptions are made
* - fragments are received out of order
* - for long messages, e.g. more than one fragment, a RTS/CTS algorithm
* is used.
* - 2nd and greater fragments include a receive descriptor pointer
* - fragments may be dropped
* - fragments may be corrupt
* - this routine may be called simultaneously by more than one thread
*/
bool mca_pml_dr_recv_frag_match(
mca_pml_dr_comm_t* comm,
mca_pml_dr_comm_proc_t *proc,
mca_btl_base_module_t *btl,
mca_pml_dr_match_hdr_t *hdr,
mca_btl_base_segment_t* segments,
size_t num_segments)
{
/* local variables */
uint16_t next_msg_seq_expected, frag_msg_seq;
mca_pml_dr_recv_request_t *match = NULL;
bool additional_match=false;
opal_list_t additional_matches;
ompi_proc_t* ompi_proc = proc->ompi_proc;
int rc;
uint32_t csum;
mca_pml_dr_endpoint_t* ep = (mca_pml_dr_endpoint_t*) proc->pml_endpoint;
bool do_csum = mca_pml_dr.enable_csum &&
(btl->btl_flags & MCA_BTL_FLAGS_NEED_CSUM);
/* source sequence number */
frag_msg_seq = hdr->hdr_seq;
/* get next expected message sequence number - if threaded
* run, lock to make sure that if another thread is processing
* a frag from the same message a match is made only once.
* Also, this prevents other posted receives (for a pair of
* end points) from being processed, and potentially "loosing"
* the fragment.
*/
OPAL_THREAD_LOCK(&comm->matching_lock);
/* get sequence number of next message that can be processed */
next_msg_seq_expected = (uint16_t)proc->expected_sequence;
if (frag_msg_seq == next_msg_seq_expected) {
/*
* This is the sequence number we were expecting,
* so we can try matching it to already posted
* receives.
*/
/* We're now expecting the next sequence number. */
(proc->expected_sequence)++;
rematch:
/*
* figure out what sort of matching logic to use, if need to
* look only at "specific" receives, or "wild" receives,
* or if we need to traverse both sets at the same time.
*/
if (opal_list_get_size(&proc->specific_receives) == 0 ){
/*
* There are only wild irecvs, so specialize the algorithm.
*/
MCA_PML_DR_CHECK_WILD_RECEIVES_FOR_MATCH(hdr, comm, proc, match);
} else if (opal_list_get_size(&comm->wild_receives) == 0 ) {
/*
* There are only specific irecvs, so specialize the algorithm.
*/
MCA_PML_DR_CHECK_SPECIFIC_RECEIVES_FOR_MATCH(hdr, comm, proc, match);
} else {
/*
* There are some of each.
*/
MCA_PML_DR_CHECK_SPECIFIC_AND_WILD_RECEIVES_FOR_MATCH(hdr, comm, proc, match);
}
/* if match found, process data */
if (match) {
/*
* update delivered sequence number information, if needed.
*/
MCA_PML_DR_RECV_REQUEST_MATCHED(match,comm,proc,hdr);
if( (match->req_recv.req_base.req_type == MCA_PML_REQUEST_PROBE) ) {
/* complete the probe */
mca_pml_dr_recv_request_matched_probe(match,btl,segments,num_segments);
/* retry the matchh */
match = NULL;
goto rematch;
}
} else {
/* if no match found, verify csum, if pass place on unexpected queue */
mca_pml_dr_recv_frag_t* frag;
MCA_PML_DR_RECV_FRAG_ALLOC(frag, rc);
if(OMPI_SUCCESS != rc) {
OPAL_THREAD_UNLOCK(&comm->matching_lock);
return false;
}
MCA_PML_DR_RECV_FRAG_INIT(frag,ompi_proc,hdr,segments,num_segments,btl,csum);
if(do_csum && csum != hdr->hdr_csum) {
mca_pml_dr_recv_frag_ack(btl,
(mca_bml_base_endpoint_t*)ompi_proc->proc_bml,
&hdr->hdr_common, hdr->hdr_src_ptr.pval, 0, 0);
MCA_PML_DR_DEBUG(0,(0, "%s:%d: received corrupted data 0x%08x != 0x%08x (segments %lu length %lu)\n",
__FILE__, __LINE__, csum, hdr->hdr_csum, (unsigned long)num_segments,
(unsigned long)(segments[0].seg_len - mca_pml_dr_hdr_size(hdr->hdr_common.hdr_type))));
MCA_PML_DR_RECV_FRAG_RETURN(frag);
OPAL_THREAD_UNLOCK(&comm->matching_lock);
return false;
}
opal_list_append( &proc->unexpected_frags, (opal_list_item_t *)frag );
}
/*
* Now that new message has arrived, check to see if
* any fragments on the c_c_frags_cant_match list
* may now be used to form new matchs
*/
if (0 < opal_list_get_size(&proc->frags_cant_match)) {
additional_match = mca_pml_dr_check_cantmatch_for_match(&additional_matches,comm,proc);
}
} else {
/*
* This message comes after the next expected, so it
* is ahead of sequence. If passes csum save it for later.
*/
mca_pml_dr_recv_frag_t* frag;
MCA_PML_DR_RECV_FRAG_ALLOC(frag, rc);
if(OMPI_SUCCESS != rc) {
OPAL_THREAD_UNLOCK(&comm->matching_lock);
return false;
}
MCA_PML_DR_RECV_FRAG_INIT(frag,ompi_proc,hdr,segments,num_segments,btl,csum);
if(do_csum && csum != hdr->hdr_csum) {
mca_pml_dr_recv_frag_ack(btl,
(mca_bml_base_endpoint_t*)ompi_proc->proc_bml,
&hdr->hdr_common, hdr->hdr_src_ptr.pval, 0, 0);
MCA_PML_DR_DEBUG(0,(0, "%s:%d: received corrupted data 0x%08x != 0x%08x\n",
__FILE__, __LINE__, csum, hdr->hdr_csum));
MCA_PML_DR_RECV_FRAG_RETURN(frag);
OPAL_THREAD_UNLOCK(&comm->matching_lock);
return false;
}
opal_list_append(&proc->frags_cant_match, (opal_list_item_t *)frag);
}
ompi_seq_tracker_insert(&ep->seq_recvs, hdr->hdr_common.hdr_vid);
OPAL_THREAD_UNLOCK(&comm->matching_lock);
/* release matching lock before processing fragment */
if(match != NULL) {
mca_pml_dr_recv_request_progress(match,btl,segments,num_segments);
}
/* if buffered a short message - go ahead and ack */
else if (hdr->hdr_common.hdr_type == MCA_PML_DR_HDR_TYPE_MATCH) {
MCA_PML_DR_DEBUG(1,(0, "%s:%d: received short message, acking now vfrag id: %d\n",
__FILE__, __LINE__, hdr->hdr_common.hdr_vid));
mca_pml_dr_recv_frag_ack(btl,
(mca_bml_base_endpoint_t*)ompi_proc->proc_bml,
&hdr->hdr_common, hdr->hdr_src_ptr.pval, 1, 0);
}
if(additional_match) {
opal_list_item_t* item;
while(NULL != (item = opal_list_remove_first(&additional_matches))) {
mca_pml_dr_recv_frag_t* frag = (mca_pml_dr_recv_frag_t*)item;
mca_pml_dr_recv_request_progress(frag->request,frag->btl,frag->segments,frag->num_segments);
MCA_PML_DR_RECV_FRAG_RETURN(frag);
}
}
return (match != NULL);
}
void mca_pml_dr_recv_frag_ack(
mca_btl_base_module_t* btl,
mca_bml_base_endpoint_t* endpoint,
mca_pml_dr_common_hdr_t* hdr,
void *src_ptr,
uint64_t mask,
uint16_t len)
{
mca_btl_base_descriptor_t* des;
mca_bml_base_btl_t* bml_btl;
mca_pml_dr_recv_frag_t* frag;
mca_pml_dr_ack_hdr_t* ack;
int rc;
bool do_csum;
/* use the same BTL for ACK's makes failover SANE */
bml_btl = mca_bml_base_btl_array_find(&endpoint->btl_eager,
btl);
do_csum = mca_pml_dr.enable_csum &&
(bml_btl->btl_flags & MCA_BTL_FLAGS_NEED_CSUM);
/* allocate descriptor */
mca_bml_base_alloc(bml_btl, &des, MCA_BTL_NO_ORDER,
sizeof(mca_pml_dr_ack_hdr_t), MCA_BTL_DES_FLAGS_PRIORITY);
if(NULL == des) {
goto retry;
}
/* fill out header */
ack = (mca_pml_dr_ack_hdr_t*)des->des_src->seg_addr.pval;
ack->hdr_common.hdr_type = MCA_PML_DR_HDR_TYPE_ACK | hdr->hdr_type;
ack->hdr_common.hdr_flags = 0;
ack->hdr_common.hdr_src = hdr->hdr_dst;
ack->hdr_common.hdr_dst = hdr->hdr_src;
ack->hdr_common.hdr_vid = hdr->hdr_vid;
ack->hdr_common.hdr_ctx = hdr->hdr_ctx;
ack->hdr_vlen = len;
ack->hdr_vmask = mask;
ack->hdr_src_ptr.pval = src_ptr;
assert(ack->hdr_src_ptr.pval);
ack->hdr_dst_ptr.pval = NULL;
ack->hdr_common.hdr_csum = (uint16_t)(do_csum ?
opal_csum(ack, sizeof(mca_pml_dr_ack_hdr_t)) :
OPAL_CSUM_ZERO);
/* initialize descriptor */
des->des_flags |= MCA_BTL_DES_FLAGS_PRIORITY;
des->des_cbfunc = mca_pml_dr_ctl_completion;
MCA_PML_DR_DEBUG(1,(0, "%s:%d: sending ack, vfrag ID %d",
__FILE__, __LINE__, hdr->hdr_vid));
rc = mca_bml_base_send(bml_btl, des, MCA_BTL_TAG_PML);
if(rc != OMPI_SUCCESS) {
mca_bml_base_free(bml_btl, des);
goto retry;
}
return;
/* queue request to retry later */
retry:
MCA_PML_DR_RECV_FRAG_ALLOC(frag,rc);
/* frag->hdr.hdr_match = *hdr; */
frag->num_segments = 0;
opal_list_append(&mca_pml_dr.acks_pending, (opal_list_item_t*)frag);
}
/**
* Scan the list of frags that came in ahead of time to see if any
* can be processed at this time. If they can, try and match the
* frags.
*
* @param additional_matches List to hold new matches with fragments
* from the c_frags_cant_match list. (IN/OUT)
*
* @param pml_comm Pointer to the communicator structure used for
* matching purposes. (IN)
*
* This routine assumes that the appropriate matching locks are
* set by the upper level routine.
*/
static bool mca_pml_dr_check_cantmatch_for_match(
opal_list_t *additional_matches,
mca_pml_dr_comm_t* comm,
mca_pml_dr_comm_proc_t *proc)
{
/* local parameters */
int match_found;
uint16_t next_msg_seq_expected, frag_seq;
mca_pml_dr_recv_frag_t *frag;
bool match_made = false;
/*
* Loop over all the out of sequence messages. No ordering is assumed
* in the c_frags_cant_match list.
*/
match_found = 1;
while ((0 < opal_list_get_size(&proc->frags_cant_match)) && match_found) {
/* initialize match flag for this search */
match_found = 0;
/* get sequence number of next message that can be processed */
next_msg_seq_expected = proc->expected_sequence;
/* search the list for a fragment from the send with sequence
* number next_msg_seq_expected
*/
for(frag = (mca_pml_dr_recv_frag_t *)
opal_list_get_first(&proc->frags_cant_match);
frag != (mca_pml_dr_recv_frag_t *)
opal_list_get_end(&proc->frags_cant_match);
frag = (mca_pml_dr_recv_frag_t *)
opal_list_get_next(frag))
{
/*
* If the message has the next expected seq from that proc...
*/
frag_seq=frag->hdr.hdr_match.hdr_seq;
if (frag_seq == next_msg_seq_expected) {
mca_pml_dr_recv_request_t *match = NULL;
mca_pml_dr_match_hdr_t* hdr = &frag->hdr.hdr_match;
/* We're now expecting the next sequence number. */
(proc->expected_sequence)++;
/* signal that match was made */
match_found = 1;
/*
* remove frag from list
*/
opal_list_remove_item(&proc->frags_cant_match,
(opal_list_item_t *)frag);
rematch:
/*
* figure out what sort of matching logic to use, if need to
* look only at "specific" receives, or "wild" receives,
* or if we need to traverse both sets at the same time.
*/
proc = (mca_pml_dr_comm_proc_t*)opal_pointer_array_get_item(&comm->sparse_procs,
hdr->hdr_common.hdr_src);
if (opal_list_get_size(&proc->specific_receives) == 0 ) {
/*
* There are only wild irecvs, so specialize the algorithm.
*/
MCA_PML_DR_CHECK_WILD_RECEIVES_FOR_MATCH(hdr, comm, proc, match);
} else if (opal_list_get_size(&comm->wild_receives) == 0 ) {
/*
* There are only specific irecvs, so specialize the algorithm.
*/
MCA_PML_DR_CHECK_SPECIFIC_RECEIVES_FOR_MATCH(hdr, comm, proc, match);
} else {
/*
* There are some of each.
*/
MCA_PML_DR_CHECK_SPECIFIC_AND_WILD_RECEIVES_FOR_MATCH(hdr, comm, proc, match);
}
/* if match found, process data */
if (match) {
/*
* If this was a probe need to queue fragment on unexpected list
*/
if( (match->req_recv.req_base.req_type == MCA_PML_REQUEST_PROBE) ) {
/* complete the probe */
mca_pml_dr_recv_request_matched_probe(match,frag->btl,frag->segments,frag->num_segments);
/* retry the match */
match = NULL;
goto rematch;
} else {
/* associate the receive descriptor with the fragment
* descriptor */
frag->request=match;
match->req_proc = proc;
match->req_endpoint = (mca_pml_dr_endpoint_t*)proc->ompi_proc->proc_pml;
MCA_PML_DR_DEBUG(10, (0, "%s:%d: adding endpoint %p match %p\n",
__FILE__, __LINE__, (void*)proc->ompi_proc->proc_pml, (void*)match->req_endpoint));
/* add this fragment descriptor to the list of
* descriptors to be processed later
*/
if(match_made == false) {
match_made = true;
OBJ_CONSTRUCT(additional_matches, opal_list_t);
}
MCA_PML_DR_RECV_REQUEST_MATCHED(match,comm,proc,&frag->hdr.hdr_match);
opal_list_append(additional_matches, (opal_list_item_t *)frag);
}
} else {
/* if no match found, place on unexpected queue */
opal_list_append( &proc->unexpected_frags, (opal_list_item_t *)frag);
}
/* c_frags_cant_match is not an ordered list, so exit loop
* and re-start search for next sequence number */
break;
} /* end if (frag_seq == next_msg_seq_expected) */
} /* end for (frag) loop */
} /* end while loop */
return match_made;
}