1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278
|
/*
* Copyright (c) 2013. Intel Corporation. All rights reserved.
* Copyright (c) 2006-2012. QLogic Corporation. All rights reserved.
* Copyright (c) 2003-2006, PathScale, Inc. All rights reserved.
*
* This software is available to you under a choice of one of two
* licenses. You may choose to be licensed under the terms of the GNU
* General Public License (GPL) Version 2, available from the file
* COPYING in the main directory of this source tree, or the
* OpenIB.org BSD license below:
*
* Redistribution and use in source and binary forms, with or
* without modification, are permitted provided that the following
* conditions are met:
*
* - Redistributions of source code must retain the above
* copyright notice, this list of conditions and the following
* disclaimer.
*
* - Redistributions in binary form must reproduce the above
* copyright notice, this list of conditions and the following
* disclaimer in the documentation and/or other materials
* provided with the distribution.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
#include <netdb.h> /* gethostbyname */
#include "psm_user.h"
#include "psm_mq_internal.h"
int psmi_ep_device_is_enabled(const psm_ep_t ep, int devid);
struct psmi_epid_table psmi_epid_table;
/* Iterator to access the epid table.
* 'ep' can be NULL if remote endpoints from all endpoint handles are requested
*/
void
psmi_epid_itor_init(struct psmi_eptab_iterator *itor, psm_ep_t ep)
{
itor->i = 0;
itor->ep = ep;
pthread_mutex_lock(&psmi_epid_table.tablock);
}
void *
psmi_epid_itor_next(struct psmi_eptab_iterator *itor)
{
int i;
struct psmi_epid_tabentry *e;
if (itor->i >= psmi_epid_table.tabsize)
return NULL;
for (i = itor->i; i < psmi_epid_table.tabsize; i++) {
e = &psmi_epid_table.table[i];
if (!e->entry || e->entry == EPADDR_DELETED)
continue;
if (itor->ep && e->ep != itor->ep)
continue;
itor->i = i+1;
return e->entry;
}
itor->i = psmi_epid_table.tabsize; /* put at end of table */
return NULL;
}
void
psmi_epid_itor_fini(struct psmi_eptab_iterator *itor)
{
pthread_mutex_unlock(&psmi_epid_table.tablock);
itor->i = 0;
}
#define mix64(a,b,c) \
{ \
a -= b; a -= c; a ^= (c>>43); \
b -= c; b -= a; b ^= (a<<9); \
c -= a; c -= b; c ^= (b>>8); \
a -= b; a -= c; a ^= (c>>38); \
b -= c; b -= a; b ^= (a<<23); \
c -= a; c -= b; c ^= (b>>5); \
a -= b; a -= c; a ^= (c>>35); \
b -= c; b -= a; b ^= (a<<49); \
c -= a; c -= b; c ^= (b>>11); \
a -= b; a -= c; a ^= (c>>12); \
b -= c; b -= a; b ^= (a<<18); \
c -= a; c -= b; c ^= (b>>22); \
}
psm_error_t
psmi_epid_init()
{
pthread_mutexattr_t attr;
psmi_epid_table.table = NULL,
psmi_epid_table.tabsize = 0;
psmi_epid_table.tabsize_used = 0;
pthread_mutexattr_init(&attr);
pthread_mutexattr_settype(&attr, PTHREAD_MUTEX_RECURSIVE);
pthread_mutex_init(&psmi_epid_table.tablock, &attr);
pthread_mutexattr_destroy(&attr);
return PSM_OK;
};
psm_error_t
psmi_epid_fini()
{
if (psmi_epid_table.table != NULL) {
psmi_free(psmi_epid_table.table);
psmi_epid_table.table = NULL;
}
psmi_epid_table.tabsize = 0;
psmi_epid_table.tabsize_used = 0;
return PSM_OK;
}
PSMI_ALWAYS_INLINE(
uint64_t
hash_this(const psm_ep_t ep, const psm_epid_t epid))
{
uint64_t ep_i = (uint64_t)(uintptr_t)ep;
uint64_t epid_i = (uint64_t) epid;
uint64_t hash = 0x9e3779b97f4a7c13LL;
mix64(ep_i,epid_i,hash);
return hash;
}
PSMI_ALWAYS_INLINE(
void *
psmi_epid_lookup_inner(psm_ep_t ep, psm_epid_t epid, int remove))
{
uint64_t key = hash_this(ep, epid);
struct psmi_epid_tabentry *e;
void *entry = NULL;
int idx;
pthread_mutex_lock(&psmi_epid_table.tablock);
if (!psmi_epid_table.table)
goto ret;
idx = (int)(key % psmi_epid_table.tabsize);
while (psmi_epid_table.table[idx].entry != NULL) {
/* An epid can be added twice if there's more than one opened endpoint,
* but really we match on epid *and* on endpoint */
e = &psmi_epid_table.table[idx];
if (e->entry != EPADDR_DELETED && e->key == key)
{
entry = e->entry;
if (remove)
psmi_epid_table.table[idx].entry = EPADDR_DELETED;
goto ret;
}
if (++idx == psmi_epid_table.tabsize)
idx = 0;
}
ret:
pthread_mutex_unlock(&psmi_epid_table.tablock);
return entry;
}
void *
psmi_epid_lookup(psm_ep_t ep, psm_epid_t epid)
{
void *entry = psmi_epid_lookup_inner(ep, epid, 0);
if (PSMI_EP_HOSTNAME != ep)
_IPATH_VDBG("lookup of (%p,%" PRIx64 ") returns %p\n", ep, epid, entry);
return entry;
}
void *
psmi_epid_remove(psm_ep_t ep, psm_epid_t epid)
{
if (PSMI_EP_HOSTNAME != ep)
_IPATH_VDBG("remove of (%p,%" PRIx64 ")\n", ep, epid);
return psmi_epid_lookup_inner(ep, epid, 1);
}
psm_error_t
psmi_epid_add(psm_ep_t ep, psm_epid_t epid, void *entry)
{
uint64_t key;
int idx, i, newsz;
struct psmi_epid_tabentry *e;
psm_error_t err = PSM_OK;
if (PSMI_EP_HOSTNAME != ep)
_IPATH_VDBG("add of (%p,%" PRIx64 ") with entry %p\n", ep, epid, entry);
pthread_mutex_lock(&psmi_epid_table.tablock);
/* Leave this here, mostly for sanity and for the fact that the epid
* table is currently not used in the critical path */
if (++psmi_epid_table.tabsize_used >
(int)(psmi_epid_table.tabsize * PSMI_EPID_TABLOAD_FACTOR))
{
struct psmi_epid_tabentry *newtab;
newsz = psmi_epid_table.tabsize + PSMI_EPID_TABSIZE_CHUNK;
newtab = (struct psmi_epid_tabentry *)
psmi_calloc(ep, PER_PEER_ENDPOINT,
newsz, sizeof(struct psmi_epid_tabentry));
if (newtab == NULL) {
err = PSM_NO_MEMORY;
goto fail;
}
if (psmi_epid_table.table) { /* rehash the table */
for (i = 0; i < psmi_epid_table.tabsize; i++) {
e = &psmi_epid_table.table[i];
if (e->entry == NULL)
continue;
/* When rehashing, mark deleted as free again */
if (e->entry == EPADDR_DELETED) {
psmi_epid_table.tabsize_used--;
continue;
}
idx = (int)(e->key % newsz);
while (newtab[idx].entry != NULL)
if (++idx == newsz)
idx = 0;
newtab[idx].entry = e->entry;
newtab[idx].key = e->key;
newtab[idx].ep = e->ep;
newtab[idx].epid = e->epid;
}
psmi_free(psmi_epid_table.table);
}
psmi_epid_table.table = newtab;
psmi_epid_table.tabsize = newsz;
}
key = hash_this(ep, epid);
idx = (int)(key % psmi_epid_table.tabsize);
e = &psmi_epid_table.table[idx];
while (e->entry && e->entry != EPADDR_DELETED) {
if (++idx == psmi_epid_table.tabsize)
idx = 0;
e = &psmi_epid_table.table[idx];
}
e->entry = entry;
e->key = key;
e->epid = epid;
e->ep = ep;
fail:
pthread_mutex_unlock(&psmi_epid_table.tablock);
return err;
}
char *
psmi_gethostname(void)
{
/* XXX this will need a lock in a multi-threaded environment */
static char hostname[80] = {'\0'};
char *c;
if (hostname[0] == '\0') {
gethostname(hostname, sizeof(hostname));
hostname[sizeof(hostname) - 1] = '\0'; /* no guarantee of nul termination */
if ((c = strchr(hostname, '.')))
*c = '\0';
}
return hostname;
}
/*
* Hostname stuff. We really only register the network portion of the epid
* since all epids from the same nid are assumed to have the same hostname.
*/
psm_error_t
psmi_epid_set_hostname(uint64_t nid, const char *hostname, int overwrite)
{
size_t hlen;
char *h;
psm_error_t err = PSM_OK;
if (hostname == NULL)
return PSM_OK;
/* First see if a hostname already exists */
if ((h = psmi_epid_lookup(PSMI_EP_HOSTNAME, nid)) != NULL) {
if (!overwrite)
return PSM_OK;
h = psmi_epid_remove(PSMI_EP_HOSTNAME, nid);
if (h != NULL) /* free the previous hostname if so exists */
psmi_free(h);
}
hlen = min(PSMI_EP_HOSTNAME_LEN, strlen(hostname)+1);
h = (char *) psmi_malloc(PSMI_EP_NONE, PER_PEER_ENDPOINT, hlen);
if (h == NULL)
return PSM_NO_MEMORY;
snprintf(h, hlen, "%s", hostname);
h[hlen-1] = '\0';
err = psmi_epid_add(PSMI_EP_HOSTNAME, nid, h);
return err;
}
/* XXX These two functions are not thread safe, we'll use a rotating buffer
* trick whenever we need to make them thread safe */
const char *
psmi_epaddr_get_hostname(psm_epid_t epid)
{
static char hostnamebufs[4][PSMI_EP_HOSTNAME_LEN];
static int bufno = 0;
uint64_t nid = psm_epid_nid(epid);
char *h, *hostname;
hostname = hostnamebufs[bufno];
bufno = (bufno + 1) % 4;
/* First, if we have registered a host for this epid, just return that, or
* else try to return something with lid and context */
h = psmi_epid_lookup(PSMI_EP_HOSTNAME, nid);
if (h != NULL)
return h;
else {
uint64_t lid, context, subcontext;
lid = PSMI_EPID_GET_LID(epid);
context = PSMI_EPID_GET_CONTEXT(epid);
subcontext = PSMI_EPID_GET_SUBCONTEXT(epid);
snprintf(hostname, PSMI_EP_HOSTNAME_LEN-1, "LID=0x%04x:%d.%d",
(unsigned int) lid, (int) context, (int) subcontext);
hostname[PSMI_EP_HOSTNAME_LEN-1] = '\0';
return hostname;
}
}
/* This one gives the hostname with a lid */
const char *
psmi_epaddr_get_name(psm_epid_t epid)
{
static char hostnamebufs[4][PSMI_EP_HOSTNAME_LEN];
static int bufno = 0;
char *h, *hostname;
uint64_t lid, context, subcontext;
lid = PSMI_EPID_GET_LID(epid);
context = PSMI_EPID_GET_CONTEXT(epid);
subcontext = PSMI_EPID_GET_SUBCONTEXT(epid);
hostname = hostnamebufs[bufno];
bufno = (bufno + 1) % 4;
h = psmi_epid_lookup(PSMI_EP_HOSTNAME, psm_epid_nid(epid));
if (h == NULL)
return psmi_epaddr_get_hostname(epid);
else {
snprintf(hostname, PSMI_EP_HOSTNAME_LEN-1,
"%s (LID=0x%04x:%d.%d)", h,
(unsigned int) lid, (int) context, (int) subcontext);
hostname[PSMI_EP_HOSTNAME_LEN-1] = '\0';
}
return hostname;
}
/* Wrapper, in case we port to OS xyz that doesn't have sysconf */
uintptr_t
psmi_getpagesize(void)
{
static uintptr_t pagesz = (uintptr_t) -1;
long sz;
if (pagesz != (uintptr_t) -1)
return pagesz;
sz = sysconf(_SC_PAGESIZE);
if (sz == -1) {
psmi_handle_error(PSMI_EP_NORETURN, PSM_INTERNAL_ERR,
"Can't query system page size");
}
pagesz = (uintptr_t) sz;
return pagesz;
}
/* If PSM_VERBOSE_ENV is set in the environment, we determine
* what its verbose level is and print the environment at "INFO"
* level if the environment's level matches the desired printlevel.
*/
static int psmi_getenv_verblevel = -1;
static int
psmi_getenv_is_verblevel(int printlevel)
{
if (psmi_getenv_verblevel == -1) {
char *env = getenv("PSM_VERBOSE_ENV");
if (env && *env) {
char *ep;
int val = (int) strtol(env, &ep, 0);
if (ep == env)
psmi_getenv_verblevel = 0;
else if (val == 2)
psmi_getenv_verblevel = 2;
else
psmi_getenv_verblevel = 1;
}
else
psmi_getenv_verblevel = 0;
}
return (printlevel <= psmi_getenv_verblevel);
}
#define GETENV_PRINTF(_level,_fmt,...) \
do { \
int nlevel = _level; \
if (psmi_getenv_is_verblevel(nlevel)) \
nlevel = 0; \
_IPATH_ENVDBG(nlevel,_fmt,##__VA_ARGS__); \
} while (0)
int
psmi_getenv(const char *name, const char *descr, int level,
int type, union psmi_envvar_val defval,
union psmi_envvar_val *newval)
{
int used_default = 0;
union psmi_envvar_val tval;
char *env = getenv(name);
int ishex = (type == PSMI_ENVVAR_TYPE_ULONG_FLAGS ||
type == PSMI_ENVVAR_TYPE_UINT_FLAGS);
/* If we're not using the default, always reset the print
* level to '1' so the changed value gets seen at low
* verbosity */
#define _GETENV_PRINT(used_default,fmt,val,defval) do { \
if (used_default) \
GETENV_PRINTF(level, "%s%-25s %-40s =>%s" #fmt \
"\n", level>1?"*":" ", name, descr, ishex?" \
0x":" ", val); \
else \
GETENV_PRINTF(1, "%s%-25s %-40s =>%s" #fmt \
" (default was%s" #fmt ")\n",level>1?"*":" ", \
name, descr, ishex?" 0x":" ", val, \
ishex?" 0x":" ", defval); \
} while (0)
switch (type) {
case PSMI_ENVVAR_TYPE_YESNO:
if (!env || *env == '\0') {
tval = defval;
used_default = 1;
}
else if (env[0] == 'Y' || env[0] == 'y')
tval.e_int = 1;
else if (env[0] == 'N' || env[0] == 'n')
tval.e_int = 0;
else {
char *ep;
tval.e_ulong = strtoul(env, &ep, 0);
if (ep == env) {
used_default = 1;
tval = defval;
}
else if (tval.e_ulong != 0)
tval.e_ulong = 1;
}
_GETENV_PRINT(used_default,%s,tval.e_long?"YES":"NO",
defval.e_int?"YES":"NO");
break;
case PSMI_ENVVAR_TYPE_STR:
if (!env || *env == '\0') {
tval = defval;
used_default = 1;
}
else
tval.e_str = env;
_GETENV_PRINT(used_default,%s,tval.e_str,defval.e_str);
break;
case PSMI_ENVVAR_TYPE_INT:
if (!env || *env == '\0') {
tval = defval;
used_default = 1;
}
else {
char *ep;
tval.e_int = (int) strtol(env, &ep, 0);
if (ep == env) {
used_default = 1;
tval = defval;
}
}
_GETENV_PRINT(used_default,%d,tval.e_int,defval.e_int);
break;
case PSMI_ENVVAR_TYPE_UINT:
case PSMI_ENVVAR_TYPE_UINT_FLAGS:
if (!env || *env == '\0') {
tval = defval;
used_default = 1;
}
else {
char *ep;
tval.e_int = (unsigned int) strtoul(env, &ep, 0);
if (ep == env) {
used_default = 1;
tval = defval;
}
}
if (type == PSMI_ENVVAR_TYPE_UINT_FLAGS)
_GETENV_PRINT(used_default,%x,tval.e_uint,defval.e_uint);
else
_GETENV_PRINT(used_default,%u,tval.e_uint,defval.e_uint);
break;
case PSMI_ENVVAR_TYPE_LONG:
if (!env || *env == '\0') {
tval = defval;
used_default = 1;
}
else {
char *ep;
tval.e_long = strtol(env, &ep, 0);
if (ep == env) {
used_default = 1;
tval = defval;
}
}
_GETENV_PRINT(used_default,%ld,tval.e_long,defval.e_long);
break;
case PSMI_ENVVAR_TYPE_ULONG_ULONG:
if (!env || *env == '\0') {
tval = defval;
used_default = 1;
}
else {
char *ep;
tval.e_ulonglong = (unsigned long long) strtoull(env, &ep, 0);
if (ep == env) {
used_default = 1;
tval = defval;
}
}
_GETENV_PRINT(used_default,%llu,
tval.e_ulonglong, defval.e_ulonglong);
break;
case PSMI_ENVVAR_TYPE_ULONG:
case PSMI_ENVVAR_TYPE_ULONG_FLAGS:
default:
if (!env || *env == '\0') {
tval = defval;
used_default = 1;
}
else {
char *ep;
tval.e_ulong = (unsigned long) strtoul(env, &ep, 0);
if (ep == env) {
used_default = 1;
tval = defval;
}
}
if (type == PSMI_ENVVAR_TYPE_ULONG_FLAGS)
_GETENV_PRINT(used_default,%lx,tval.e_ulong,defval.e_ulong);
else
_GETENV_PRINT(used_default,%lu,tval.e_ulong,defval.e_ulong);
break;
}
#undef _GETENV_PRINT
*newval = tval;
return used_default;
}
/*
* Parsing int parameters set in string tuples.
* Output array int *vals should be able to store 'ntup' elements.
* Values are only overwritten if they are parsed.
* Tuples are always separated by colons ':'
*/
int psmi_parse_str_tuples(const char *string, int ntup, int *vals)
{
char *b = (char *) string;
char *e = b;
int tup_i = 0;
int n_parsed = 0;
char *buf = psmi_strdup(NULL, string);
psmi_assert_always(buf != NULL);
while (*e && tup_i < ntup) {
b = e;
while (*e && *e != ':')
e++;
if (e > b) { /* something to parse */
char *ep;
int len = e - b;
long int l;
strncpy(buf, b, len);
buf[len] = '\0';
l = strtol(buf, &ep, 0);
if (ep != buf) { /* successful conversion */
vals[tup_i] = (int) l;
n_parsed++;
}
}
if (*e == ':')
e++; /* skip delimiter */
tup_i++;
}
psmi_free(buf);
return n_parsed;
}
/*
* Memory footprint/usage mode.
*
* This can be used for debug or for separating large installations from
* small/medium ones. The default is to assume a medium installation. Large
* is not that much larger in memory footprint, but we make a conscious effort
* an consuming only the amount of memory we need.
*/
int
psmi_parse_memmode(void)
{
union psmi_envvar_val env_mmode;
int used_default =
psmi_getenv("PSM_MEMORY", "Memory usage mode (normal or large)",
PSMI_ENVVAR_LEVEL_USER, PSMI_ENVVAR_TYPE_STR,
(union psmi_envvar_val) "normal", &env_mmode);
if (used_default || !strcasecmp(env_mmode.e_str, "normal"))
return PSMI_MEMMODE_NORMAL;
else if (!strcasecmp(env_mmode.e_str, "min"))
return PSMI_MEMMODE_MINIMAL;
else if (!strcasecmp(env_mmode.e_str, "large") ||
!strcasecmp(env_mmode.e_str, "big"))
return PSMI_MEMMODE_LARGE;
else {
_IPATH_PRDBG("PSM_MEMORY env value %s unrecognized, "
"using 'normal' memory mode instead\n",
env_mmode.e_str);
return PSMI_MEMMODE_NORMAL;
}
}
static
const char *
psmi_memmode_string(int mode)
{
psmi_assert(mode >= PSMI_MEMMODE_NORMAL && mode < PSMI_MEMMODE_NUM);
switch (mode) {
case PSMI_MEMMODE_NORMAL:
return "normal";
case PSMI_MEMMODE_MINIMAL:
return "minimal";
case PSMI_MEMMODE_LARGE:
return "large";
default:
return "unknown";
}
}
psm_error_t
psmi_parse_mpool_env(const psm_mq_t mq, int level,
const struct psmi_rlimit_mpool *rlim,
uint32_t *valo, uint32_t *chunkszo)
{
uint32_t val;
const char *env = rlim->env;
int mode = mq->memmode;
psm_error_t err = PSM_OK;
union psmi_envvar_val env_val;
psmi_assert_always(mode >= PSMI_MEMMODE_NORMAL && mode < PSMI_MEMMODE_NUM);
psmi_getenv(rlim->env, rlim->descr, rlim->env_level,
PSMI_ENVVAR_TYPE_UINT,
(union psmi_envvar_val) rlim->mode[mode].obj_max,
&env_val);
val = env_val.e_uint;
if (val < rlim->minval || val > rlim->maxval)
{
err = psmi_handle_error(NULL, PSM_PARAM_ERR,
"Env. var %s=%u is invalid (valid settings in mode PSM_MEMORY=%s"
" are inclusively between %u and %u)", env, val,
psmi_memmode_string(mode), rlim->minval, rlim->maxval);
goto fail;
}
_IPATH_VDBG("%s max=%u,chunk=%u (mode=%s(%u),min=%u,max=%u)\n",
env, val, rlim->mode[mode].obj_chunk, psmi_memmode_string(mode),
mode, rlim->minval, rlim->maxval);
*valo = val;
*chunkszo = rlim->mode[mode].obj_chunk;
fail:
return err;
}
uint64_t
psmi_cycles_left(uint64_t start_cycles, int64_t timeout_ns)
{
if (timeout_ns < 0)
return 0ULL;
else if (timeout_ns == 0ULL || timeout_ns == ~0ULL)
return ~0ULL;
else {
uint64_t t_end = nanosecs_to_cycles(timeout_ns);
uint64_t t_now = get_cycles() - start_cycles;
if (t_now >= t_end)
return 0ULL;
else
return (t_end - t_now);
}
}
uint32_t
psmi_get_ipv4addr()
{
struct hostent *he;
uint32_t addr = 0;
he = gethostbyname(psmi_gethostname());
if (he != NULL && he->h_addrtype == AF_INET && he->h_addr != NULL) {
memcpy(&addr, he->h_addr, sizeof(uint32_t));
return addr;
}
else
return 0;
}
#define PSMI_EP_IS_PTR(ptr) ((ptr) != NULL && (ptr) < PSMI_EP_LOGEVENT)
void
psmi_syslog(psm_ep_t ep, int to_console, int level, const char *format, ...)
{
va_list ap;
/* If we've never syslogged anything from this ep at the PSM level, make
* sure we log context information */
if (PSMI_EP_IS_PTR(ep) && !ep->did_syslog) {
char uuid_str[64];
ep->did_syslog = 1;
memset(&uuid_str, 0, sizeof uuid_str);
psmi_uuid_unparse(ep->key, uuid_str);
ipath_syslog("PSM", 0, LOG_WARNING,
"uuid_key=%s,unit=%d,context=%d,subcontext=%d",
uuid_str,
ep->context.base_info.spi_unit,
ep->context.base_info.spi_context,
ep->context.base_info.spi_subcontext);
}
va_start(ap, format);
ipath_vsyslog("PSM", to_console, level, format, ap);
va_end(ap);
}
/* Table of CRCs of all 8-bit messages. */
static uint32_t crc_table[256];
/* Flag: has the table been computed? Initially false. */
static int crc_table_computed = 0;
/* Make the table for a fast CRC. */
static void make_crc_table(void)
{
uint32_t c;
int n, k;
for (n = 0; n < 256; n++) {
c = (uint32_t) n;
for (k = 0; k < 8; k++) {
if (c & 1)
c = 0xedb88320 ^ (c >> 1);
else
c = c >> 1;
}
crc_table[n] = c;
}
crc_table_computed = 1;
}
/* Update a running CRC with the bytes buf[0..len-1]--the CRC
* should be initialized to all 1's, and the transmitted value
* is the 1's complement of the final running CRC (see the
* crc() routine below)).
*/
static uint32_t update_crc(uint32_t crc, unsigned char *buf, int len)
{
uint32_t c = crc;
int n;
if_pf (!crc_table_computed)
make_crc_table();
for (n = 0; n < len; n++) {
c = crc_table[(c ^ buf[n]) & 0xff] ^ (c >> 8);
}
return c;
}
/* Return the CRC of the bytes buf[0..len-1]. */
uint32_t psmi_crc(unsigned char *buf, int len)
{
return update_crc(0xffffffff, buf, len) ^ 0xffffffff;
}
/* Return the HCA type being used for a context */
uint32_t psmi_get_hca_type(psmi_context_t *context)
{
uint32_t hca_type;
/* Determine HCA type. Use heuristics based on runtime flags
*
* Header suppression available: QLE73XX
* NODMA_RTAIL: QLE72XX
* <AnythingElse>: QLE71XX
*/
if (context->runtime_flags & IPATH_RUNTIME_HDRSUPP)
hca_type = PSMI_HCA_TYPE_QLE73XX;
else if (context->runtime_flags & IPATH_RUNTIME_NODMA_RTAIL)
hca_type = PSMI_HCA_TYPE_QLE72XX;
else
hca_type = PSMI_HCA_TYPE_QLE71XX;
return hca_type;
}
#define PSMI_FAULTINJ_SPEC_NAMELEN 32
struct psmi_faultinj_spec {
STAILQ_ENTRY(psmi_faultinj_spec) next;
char spec_name[PSMI_FAULTINJ_SPEC_NAMELEN];
unsigned long long num_faults;
unsigned long long num_calls;
unsigned int seedp;
int num;
int denom;
};
int psmi_faultinj_enabled = 0;
int psmi_faultinj_verbose = 0;
char *psmi_faultinj_outfile = NULL;
static struct psmi_faultinj_spec psmi_faultinj_dummy;
static STAILQ_HEAD(, psmi_faultinj_spec) psmi_faultinj_head =
STAILQ_HEAD_INITIALIZER(psmi_faultinj_head);
void
psmi_faultinj_init()
{
union psmi_envvar_val env_fi;
psmi_getenv("PSM_FI", "PSM Fault Injection (yes/no)",
PSMI_ENVVAR_LEVEL_HIDDEN, PSMI_ENVVAR_TYPE_YESNO,
PSMI_ENVVAR_VAL_NO, &env_fi);
psmi_faultinj_enabled = !!env_fi.e_uint;
if (psmi_faultinj_enabled) {
char *def = NULL;
if (!psmi_getenv("PSM_FI_TRACEFILE", "PSM Fault Injection output file",
PSMI_ENVVAR_LEVEL_HIDDEN, PSMI_ENVVAR_TYPE_STR,
(union psmi_envvar_val) def, &env_fi))
{
psmi_faultinj_outfile = psmi_strdup(NULL, env_fi.e_str);
}
}
return;
}
void
psmi_faultinj_fini()
{
struct psmi_faultinj_spec *fi;
FILE *fp;
int do_fclose = 0;
if (!psmi_faultinj_enabled || psmi_faultinj_outfile == NULL)
return;
if (strncmp(psmi_faultinj_outfile, "stdout", 7) == 0)
fp = stdout;
else if (strncmp(psmi_faultinj_outfile, "stderr", 7) == 0)
fp = stderr;
else {
char *c = psmi_faultinj_outfile;
char buf[192];
int append = 0;
if (*c == '+') {
append = 1;
++c;
}
do_fclose = 1;
snprintf(buf, sizeof buf - 1, "%s.%s", c, __ipath_mylabel);
buf[sizeof buf - 1] = '\0';
fp = fopen(buf, append ? "a" : "w");
}
if (fp != NULL) {
STAILQ_FOREACH(fi, &psmi_faultinj_head, next) {
fprintf(fp, "%s:%s PSM_FI_%-12s %2.3f%% => "
"%2.3f%% %10lld faults/%10lld events\n", __progname,
__ipath_mylabel, fi->spec_name,
(double) fi->num * 100.0 / fi->denom,
(double) fi->num_faults * 100.0 / fi->num_calls,
fi->num_faults, fi->num_calls);
}
fflush(fp);
if (do_fclose)
fclose(fp);
}
psmi_free(psmi_faultinj_outfile);
return;
}
/*
* Intended to be used only once, not in the critical path
*/
struct psmi_faultinj_spec *
psmi_faultinj_getspec(char *spec_name, int num, int denom)
{
struct psmi_faultinj_spec *fi;
if (!psmi_faultinj_enabled)
return &psmi_faultinj_dummy;
STAILQ_FOREACH(fi, &psmi_faultinj_head, next) {
if (strcmp(fi->spec_name, spec_name) == 0)
return fi;
}
/* We got here, so no spec -- allocate one */
fi = psmi_malloc(PSMI_EP_NONE, UNDEFINED, sizeof(struct psmi_faultinj_spec));
strncpy(fi->spec_name, spec_name, PSMI_FAULTINJ_SPEC_NAMELEN-1);
fi->spec_name[PSMI_FAULTINJ_SPEC_NAMELEN-1] = '\0';
fi->num = num;
fi->denom = denom;
fi->num_faults = 0;
fi->num_calls = 0;
/*
* See if we get a hint from the environment.
* Format is
* <num:denom:initial_seed>
*
* By default, we chose the initial seed to be the 'pid'. If users need
* repeatability, they should set initial_seed to be the 'pid' when the
* error was observed or force the initial_seed to be a constant number in
* each running process. Using 'pid' is useful because core dumps store
* pids and our backtrace format does as well so if a crash is observed for
* a specific seed, programs can reuse the 'pid' to regenerate the same
* error condition.
*/
{
int fvals[3] = { num, denom, (int) getpid() };
union psmi_envvar_val env_fi;
char fvals_str[128];
char fname[128];
char fdesc[256];
snprintf(fvals_str, sizeof fvals_str - 1, "%d:%d:1", num, denom);
fvals_str[sizeof fvals_str - 1] = '\0';
snprintf(fname, sizeof fname - 1, "PSM_FI_%s", spec_name);
fname[sizeof fname - 1] = '\0';
snprintf(fdesc, sizeof fdesc - 1, "Fault Injection %s <%s>",
fname, fvals_str);
if (!psmi_getenv(fname, fdesc, PSMI_ENVVAR_LEVEL_HIDDEN,
PSMI_ENVVAR_TYPE_STR, (union psmi_envvar_val) fvals_str,
&env_fi))
{
/* not using default values */
int n_parsed = psmi_parse_str_tuples(env_fi.e_str, 3, fvals);
if (n_parsed >= 1)
fi->num = fvals[0];
if (n_parsed >= 2)
fi->denom = fvals[1];
if (n_parsed >= 3)
fi->seedp = fvals[2];
}
}
STAILQ_INSERT_TAIL(&psmi_faultinj_head, fi, next);
return fi;
}
int
psmi_faultinj_is_fault(struct psmi_faultinj_spec *fi)
{
int r;
if (!psmi_faultinj_enabled) /* never fault if disabled */
return 0;
if (fi->num == 0)
return 0;
fi->num_calls++;
r = rand_r(&fi->seedp);
if (r % fi->denom <= fi->num) {
fi->num_faults++;
return 1;
}
else
return 0;
}
/* For memory allocation, we kind of break the PSM error handling rules.
* If the caller gets NULL, it has to assume that the error has been handled
* and should always return PSM_NO_MEMORY */
/*
* Log memory increments or decrements of type memstats_t.
*/
struct psmi_memtype_hdr {
struct {
uint64_t size : 48;
uint64_t magic : 8;
uint64_t type : 8;
};
};
struct psmi_stats_malloc psmi_stats_memory;
void
psmi_log_memstats(psmi_memtype_t type, int64_t nbytes)
{
#define _add_max_total(type,nbytes) \
psmi_stats_memory.m_ ## type ## _total += (nbytes); \
psmi_stats_memory.m_ ## type ## _max = max( \
psmi_stats_memory.m_ ## type ## _total, \
psmi_stats_memory.m_ ## type ## _max);
switch (type) {
case PER_PEER_ENDPOINT:
_add_max_total(perpeer, nbytes);
break;
case NETWORK_BUFFERS:
_add_max_total(netbufs, nbytes);
break;
case DESCRIPTORS:
_add_max_total(descriptors, nbytes);
break;
case UNEXPECTED_BUFFERS:
_add_max_total(unexpbufs, nbytes);
break;
case STATS:
_add_max_total(stats, nbytes);
break;
case UNDEFINED:
_add_max_total(undefined, nbytes);
break;
default:
psmi_assert_always(type == TOTAL);
break;
}
_add_max_total(all, nbytes);
psmi_stats_memory.m_all_max++;
#undef _add_max_total
return;
}
#define psmi_stats_mask PSMI_STATSTYPE_MEMORY
#ifdef malloc
#undef malloc
#endif
void *
psmi_malloc_internal(psm_ep_t ep, psmi_memtype_t type,
size_t sz, const char *curloc)
{
size_t newsz = sz;
void *newa;
psmi_assert(sizeof(struct psmi_memtype_hdr) == 8);
if_pf (psmi_stats_mask & PSMI_STATSTYPE_MEMORY)
newsz += sizeof(struct psmi_memtype_hdr);
newa = malloc(newsz);
if (newa == NULL) {
psmi_handle_error(PSMI_EP_NORETURN, PSM_NO_MEMORY,
"Out of memory for malloc at %s", curloc);
return NULL;
}
if_pf (psmi_stats_mask & PSMI_STATSTYPE_MEMORY) {
struct psmi_memtype_hdr *hdr = (struct psmi_memtype_hdr *) newa;
hdr->size = newsz;
hdr->type = type;
hdr->magic = 0x8c;
psmi_log_memstats(type, newsz);
newa = (void *) (hdr + 1);
//_IPATH_INFO("alloc is %p\n", newa);
}
return newa;
}
#ifdef calloc
#undef calloc
#endif
void *
psmi_calloc_internal(psm_ep_t ep, psmi_memtype_t type, size_t nelem,
size_t elemsz, const char *curloc)
{
void *newa = psmi_malloc_internal(ep, type, nelem*elemsz, curloc);
if (newa == NULL) /* error handled above */
return NULL;
memset(newa, 0, nelem*elemsz);
return newa;
}
#ifdef strdup
#undef strdup
#endif
void *
psmi_strdup_internal(psm_ep_t ep, const char *string, const char *curloc)
{
size_t len = strlen(string)+1;
void *newa = psmi_malloc_internal(ep, UNDEFINED, len, curloc);
if (newa == NULL)
return NULL;
memcpy(newa, string, len); /* copy with \0 */
return newa;
}
#ifdef free
#undef free
#endif
void
psmi_free_internal(void *ptr)
{
if_pf (psmi_stats_mask & PSMI_STATSTYPE_MEMORY) {
struct psmi_memtype_hdr *hdr =
(struct psmi_memtype_hdr *) ptr - 1;
//_IPATH_INFO("hdr is %p, ptr is %p\n", hdr, ptr);
psmi_memtype_t type = hdr->type;
int64_t size = hdr->size;
int magic = (int) hdr->magic;
psmi_log_memstats(type, -size);
psmi_assert_always(magic == 0x8c);
ptr = (void *) hdr;
}
free(ptr);
}
PSMI_ALWAYS_INLINE(
psm_error_t
psmi_coreopt_ctl(const void *core_obj, int optname,
void *optval, uint64_t *optlen, int get))
{
psm_error_t err = PSM_OK;
char err_string[256];
switch(optname) {
case PSM_CORE_OPT_DEBUG:
/* Sanity check length */
if (*optlen < sizeof(unsigned)) {
snprintf(err_string, 256, "Option value length error");
*optlen = sizeof(unsigned);
goto fail;
}
if (get) {
*((unsigned *) optval) = infinipath_debug;
}
else
infinipath_debug = *(unsigned*) optval;
break;
case PSM_CORE_OPT_EP_CTXT:
{
/* core object is epaddr */
psm_epaddr_t epaddr = (psm_epaddr_t) core_obj;
/* Sanity check epaddr */
if (!epaddr) {
snprintf(err_string, 256, "Invalid endpoint address");
goto fail;
}
/* Sanity check length */
if (*optlen < sizeof(unsigned long)) {
snprintf(err_string, 256, "Option value length error");
*optlen = sizeof(void*);
goto fail;
}
if (get) {
*((unsigned long*) optval) = (unsigned long) epaddr->usr_ep_ctxt;
}
else
epaddr->usr_ep_ctxt = optval;
}
break;
default:
/* Unknown/unrecognized option */
snprintf(err_string, 256, "Unknown PSM_CORE option %u.", optname);
goto fail;
}
return err;
fail:
/* Unrecognized/unknown option */
return psmi_handle_error(NULL, PSM_PARAM_ERR, err_string, "%s");
}
psm_error_t psmi_core_setopt(const void *core_obj, int optname,
const void *optval, uint64_t optlen)
{
return psmi_coreopt_ctl(core_obj, optname, (void*) optval, &optlen, 0);
}
psm_error_t psmi_core_getopt(const void *core_obj, int optname,
void *optval, uint64_t *optlen)
{
return psmi_coreopt_ctl(core_obj, optname, optval, optlen, 1);
}
/* PSM AM component option handling */
PSMI_ALWAYS_INLINE(
psm_error_t
psmi_amopt_ctl(const void *am_obj, int optname,
void *optval, uint64_t *optlen, int get))
{
psm_error_t err = PSM_OK;
switch(optname) {
case PSM_AM_OPT_FRAG_SZ:
{
/* AM object is a psm_epaddr (or NULL for global minimum sz) */
psm_epaddr_t epaddr = (psm_epaddr_t) am_obj;
if (!get) /* Cannot set this option */
return psmi_handle_error(NULL, PSM_OPT_READONLY,
"Unable to set PSM_AM_OPT_FRAG_SZ. This is "
"a read only option.");
/* Sanity check length */
if (*optlen < sizeof(uint32_t)) {
*optlen = sizeof(uint32_t);
return err = psmi_handle_error(PSMI_EP_LOGEVENT, PSM_PARAM_ERR,
"Option value length error");
}
/* TODO: Currently all AMs occur over IPS which utilizes the PIO flows.
* These are limited to the PIO size of the chip. Once we have AM
* capability over shared memory then we can have different fragment
* sizes over both transport and the global fragment size will need to
* take the minimum of all possible transports used. For now if the
* endpoint is opened get the PIO size from it else hard code it to 2K
* which is "correct" for all supported chips.
*/
*((unsigned *) optval) =
(epaddr &&
psmi_ep_device_is_enabled(epaddr->ep, PTL_DEVID_IPS)) ?
(epaddr->ep->context.base_info.spi_piosize -
IPATH_MESSAGE_HDR_SIZE) : 2048;
}
break;
default:
err = psmi_handle_error(NULL, PSM_PARAM_ERR, "Unknown PSM_AM option %u.", optname);
}
return err;
}
psm_error_t psmi_am_setopt(const void *am_obj, int optname,
const void *optval, uint64_t optlen)
{
return psmi_amopt_ctl(am_obj, optname, (void*) optval, &optlen, 0);
}
psm_error_t psmi_am_getopt(const void *am_obj, int optname,
void *optval, uint64_t *optlen)
{
return psmi_amopt_ctl(am_obj, optname, optval, optlen, 1);
}
|