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
|
/*___INFO__MARK_BEGIN__*/
/*************************************************************************
*
* The Contents of this file are made available subject to the terms of
* the Sun Industry Standards Source License Version 1.2
*
* Sun Microsystems Inc., March, 2001
*
*
* Sun Industry Standards Source License Version 1.2
* =================================================
* The contents of this file are subject to the Sun Industry Standards
* Source License Version 1.2 (the "License"); You may not use this file
* except in compliance with the License. You may obtain a copy of the
* License at http://gridengine.sunsource.net/Gridengine_SISSL_license.html
*
* Software provided under this License is provided on an "AS IS" basis,
* WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING,
* WITHOUT LIMITATION, WARRANTIES THAT THE SOFTWARE IS FREE OF DEFECTS,
* MERCHANTABLE, FIT FOR A PARTICULAR PURPOSE, OR NON-INFRINGING.
* See the License for the specific provisions governing your rights and
* obligations concerning the Software.
*
* The Initial Developer of the Original Code is: Sun Microsystems, Inc.
*
* Copyright: 2001 by Sun Microsystems, Inc.
*
* All Rights Reserved.
*
************************************************************************/
/*___INFO__MARK_END__*/
#include <stdlib.h>
#include <stdio.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <errno.h>
#include <signal.h>
#include <string.h>
#include <unistd.h>
#include <sys/wait.h>
#include <sys/time.h>
#include <sys/timeb.h>
#include <sys/resource.h>
#if defined(DARWIN)
# include <termios.h>
# include <sys/ttycom.h>
# include <sys/ioctl.h>
#elif defined(HP11) || defined(HP1164)
# include <termios.h>
#elif defined(INTERIX)
# include <termios.h>
# include <sys/ioctl.h>
#else
# include <termio.h>
#endif
#include "basis_types.h"
#include "err_trace.h"
#include "sge_stdlib.h"
#include "sge_dstring.h"
#include "sge_pty.h"
#include "sge_ijs_comm.h"
#include "sge_ijs_threads.h"
#include "sge_io.h"
#include "sge_fileio.h"
#include "sge_uidgid.h"
#include "sge_unistd.h"
#include "sge_signal.h"
#define RESPONSE_MSG_TIMEOUT 120
#define COMM_SERVER "qrsh_ijs"
#define COMM_CLIENT "shepherd_ijs"
/*
* Compile with EXTENSIVE_TRACING defined to get lots of trace messages from
* the two worker threads.
*/
#undef EXTENSIVE_TRACING
/*#define EXTENSIVE_TRACING*/
static int g_ptym = -1;
static int g_fd_pipe_in[2] = {-1,-1};
static int g_fd_pipe_out[2] = {-1, -1};
static int g_fd_pipe_err[2] = {-1, -1};
static int g_fd_pipe_to_child[2] = {-1, -1};
static COMMUNICATION_HANDLE *g_comm_handle = NULL;
static THREAD_HANDLE g_thread_main;
static int g_raised_event = 0;
static int g_job_pid = 0;
char *g_hostname = NULL;
extern bool g_csp_mode; /* defined in shepherd.c */
extern int received_signal; /* defined in shepherd.c */
/*
* Forward declarations of functions from shepherd.c
*/
int wait_my_child(
int pid, /* pid of job */
int ckpt_pid, /* pid of restarted job or same as pid */
int ckpt_type, /* type of checkpointing */
struct rusage *rusage, /* accounting information */
int timeout, /* used for prolog/epilog script, 0 for job */
int ckpt_interval, /* action depend on ckpt_type */
char *childname /* "job", "pe_start", ... */
);
void
shepherd_signal_job(pid_t pid, int sig);
/*
* static functions
*/
/****** trace_buf() **********************************************************
* NAME
* trace_buf() -- writes contents of a buffer partially to the trace file
*
* SYNOPSIS
* int trace_buf(const char *buffer, int length, const char *format, ...)
*
* FUNCTION
* Writes the contents of a buffer, preceeded by a formatted string,
* partially to the trace file. I.e. it first writes the formatted
* string to the trace file, replacing the printf placeholders with
* the variables provided in ..., and then adds the first 99 or "length"
* characters (which ever is smaller), encapsulated by hyphens, to
* the formatted string.
*
* INPUTS
* const char *buffer - the buffer that is to be written to the trace file
* int length - length of the content of the buffer
* const char *format -
* int fd - file descriptor to read from
* char *pbuf - working buffer, must be of size BUFSIZE
* int *buf_bytes - number of bytes already in the buffer
*
* OUTPUTS
* int *buf_bytes - number of bytes in the buffer
*
* RESULT
* int - >0: OK, number of bytes read from the fd
* =0: EINTR or EAGAIN occured, just call select() and read again
* -1: error occcured, connection was closed
*
* NOTES
* MT-NOTE:
*
* SEE ALSO
*******************************************************************************/
#ifdef EXTENSIVE_TRACING
static int trace_buf(const char *buffer, int length, const char *format, ...)
{
int ret;
va_list ap;
dstring message = DSTRING_INIT;
char tmpbuf[100];
if (length > 0) {
snprintf(tmpbuf, MIN(99,length), "%s", buffer);
} else {
strcpy(tmpbuf, "");
}
va_start(ap, format);
sge_dstring_vsprintf(&message, format, ap);
sge_dstring_append(&message, "\"");
sge_dstring_append(&message, tmpbuf);
sge_dstring_append(&message, "\"");
ret = shepherd_trace("%s", sge_dstring_get_string(&message));
sge_dstring_free(&message);
va_end(ap);
return ret;
}
#endif
/****** append_to_buf() ******************************************************
* NAME
* append_to_buf() -- copies data from pty (or pipe) to a buffer
*
* SYNOPSIS
* int append_to_buf(int fd, char *pbuf, int *buf_bytes)
*
* FUNCTION
* Reads data from the pty or pipe and appends it to the given buffer.
*
* INPUTS
* int fd - file descriptor to read from
* char *pbuf - working buffer, must be of size BUFSIZE
* int *buf_bytes - number of bytes already in the buffer
*
* OUTPUTS
* int *buf_bytes - number of bytes in the buffer
*
* RESULT
* int - >0: OK, number of bytes read from the fd
* =0: EINTR or EAGAIN occured, just call select() and read again
* -1: error occcured, connection was closed
*
* NOTES
* MT-NOTE:
*
* SEE ALSO
*******************************************************************************/
static int append_to_buf(int fd, char *pbuf, int *buf_bytes)
{
int nread = 0;
if (fd >= 0) {
nread = read(fd, &pbuf[*buf_bytes], BUFSIZE-1-(*buf_bytes));
if (nread < 0 && (errno == EINTR || errno == EAGAIN)) {
nread = 0;
} else if (nread <= 0) {
nread = -1;
} else {
*buf_bytes += nread;
}
}
return nread;
}
/****** send_buf() ***********************************************************
* NAME
* send_buf() -- sends the content of the buffer over the commlib
*
* SYNOPSIS
* int send_buf(char *pbuf, int buf_bytes, int message_type)
*
* FUNCTION
* Sends the content of the buffer over to the commlib to the receiver.
*
* INPUTS
* char *pbuf - buffer to send
* int buf_bytes - number of bytes in the buffer to send
* int message_tye - type of the message that is to be sent over
* commlib. Can be STDOUT_DATA_MSG or
* STDERR_DATA_MSG, depending on where the data
* came from (pty and stdout = STDOUT_DATA_MSG,
* stderr = STDERR_DATA_MSG)
*
* RESULT
* int - 0: OK
* 1: could'nt write all data
*
* NOTES
* MT-NOTE:
*
* SEE ALSO
*******************************************************************************/
static int send_buf(char *pbuf, int buf_bytes, int message_type)
{
int ret = 0;
dstring err_msg = DSTRING_INIT;
if (comm_write_message(g_comm_handle, g_hostname,
COMM_SERVER, 1, (unsigned char*)pbuf,
(unsigned long)buf_bytes, message_type, &err_msg) != buf_bytes) {
shepherd_trace("couldn't write all data: %s\n",
sge_dstring_get_string(&err_msg));
ret = 1;
} else {
#ifdef EXTENSIVE_TRACING
shepherd_trace("successfully wrote all data: %s\n",
sge_dstring_get_string(&err_msg));
#endif
}
sge_dstring_free(&err_msg);
return ret;
}
/****** pty_to_commlib() *******************************************************
* NAME
* pty_to_commlib() -- pty_to_commlib thread entry point and main loop
*
* SYNOPSIS
* void* pty_to_commlib(void *t_conf)
*
* FUNCTION
* Entry point and main loop of the pty_to_commlib thread.
* Reads data from the pty and writes it to the commlib.
*
* INPUTS
* void *t_conf - pointer to cl_thread_settings_t struct of the thread
*
* RESULT
* void* - always NULL
*
* NOTES
* MT-NOTE:
*
* SEE ALSO
*******************************************************************************/
static void* pty_to_commlib(void *t_conf)
{
int do_exit = 0;
int fd_max = 0;
int ret;
fd_set read_fds;
struct timeval timeout;
char *stdout_buf = NULL;
char *stderr_buf = NULL;
int stdout_bytes = 0;
int stderr_bytes = 0;
bool b_select_timeout = false;
dstring err_msg = DSTRING_INIT;
/* Report to thread lib that this thread starts up now */
thread_func_startup(t_conf);
/* Set this thread to be cancelled (=killed) at any time. */
thread_setcancelstate(1);
/* Allocate working buffers, BUFSIZE = 64k */
stdout_buf = sge_malloc(BUFSIZE);
stderr_buf = sge_malloc(BUFSIZE);
/* The main loop of this thread */
while (do_exit == 0) {
/* Fill fd_set for select */
FD_ZERO(&read_fds);
if (g_ptym != -1) {
FD_SET(g_ptym, &read_fds);
}
if (g_fd_pipe_out[0] != -1) {
FD_SET(g_fd_pipe_out[0], &read_fds);
}
if (g_fd_pipe_err[0] != -1) {
FD_SET(g_fd_pipe_err[0], &read_fds);
}
fd_max = MAX(g_ptym, g_fd_pipe_out[0]);
fd_max = MAX(fd_max, g_fd_pipe_err[0]);
#ifdef EXTENSIVE_TRACING
shepherd_trace("pty_to_commlib: g_ptym = %d, g_fd_pipe_out[0] = %d, "
"g_fd_pipe_err[0] = %d, fd_max = %d",
g_ptym, g_fd_pipe_out[0], g_fd_pipe_err[0], fd_max);
#endif
/* Fill timeout struct for select */
b_select_timeout = false;
if (do_exit == 1) {
/* If we know that we have to exit, don't wait for data,
* just peek into the fds and read all data available from the buffers. */
timeout.tv_sec = 0;
timeout.tv_usec = 0;
} else {
if ((stdout_bytes > 0 && stdout_bytes < 256)
|| cl_com_messages_in_send_queue(g_comm_handle) > 0) {
/*
* If we just received few data, wait another 10 milliseconds if new
* data arrives to avoid sending data over the network in too small junks.
* Also retry to send the messages that are still in the queue ASAP.
*/
timeout.tv_sec = 0;
timeout.tv_usec = 10000;
} else {
/* Standard timeout is one second */
timeout.tv_sec = 1;
timeout.tv_usec = 0;
}
}
#ifdef EXTENSIVE_TRACING
shepherd_trace("pty_to_commlib: doing select() with %d seconds, "
"%d usec timeout", timeout.tv_sec, timeout.tv_usec);
#endif
/* Wait blocking for data from pty or pipe */
errno = 0;
ret = select(fd_max+1, &read_fds, NULL, NULL, &timeout);
thread_testcancel(t_conf);
/* This is a workaround for Darwin and HP11, where thread_testcancel() doesn't work.
* TODO: Find the reason why it doesn't work and remove the workaround
*/
if (g_raised_event > 0) {
do_exit = 1;
}
#ifdef EXTENSIVE_TRACING
shepherd_trace("pty_to_commlib: select() returned %d", ret);
#endif
if (ret < 0) {
/* select error */
#ifdef EXTENSIVE_TRACING
shepherd_trace("pty_to_commlib: select() returned %d, reason: %d, %s",
ret, errno, strerror(errno));
#endif
if (errno == EINTR) {
/* If we have a buffer, send it now (as we don't want to care about
* how long we acutally waited), then just continue, the top of the
* loop will handle signals.
* b_select_timeout tells the bottom of the loop to send the buffer.
*/
/* ATTENTION: Don't call shepherd_trace() here, because on some
* architectures it causes another EINTR, leading to a infinte loop.
*/
b_select_timeout = true;
} else {
shepherd_trace("pty_to_commlib: select() error -> exiting");
do_exit = 1;
}
} else if (ret == 0) {
/* timeout, if we have a buffer, send it now */
#ifdef EXTENSIVE_TRACING
shepherd_trace("pty_to_commlib: select() timeout");
#endif
b_select_timeout = true;
} else {
/* at least one fd is ready to read from */
ret = 1;
/* now we can be sure that our child has started the job,
* we can close the pipe_to_child now
*/
if (g_fd_pipe_to_child[1] != -1) {
shepherd_trace("pty_to_commlib: closing pipe to child");
close(g_fd_pipe_to_child[1]);
g_fd_pipe_to_child[1] = -1;
}
if (g_ptym != -1 && FD_ISSET(g_ptym, &read_fds)) {
#ifdef EXTENSIVE_TRACING
shepherd_trace("pty_to_commlib: reading from ptym");
#endif
ret = append_to_buf(g_ptym, stdout_buf, &stdout_bytes);
#ifdef EXTENSIVE_TRACING
trace_buf(stdout_buf, ret, "pty_to_commlib: appended %d bytes, stdout_buf = ",
ret, stdout_buf);
#endif
}
if (ret >= 0 && g_fd_pipe_out[0] != -1
&& FD_ISSET(g_fd_pipe_out[0], &read_fds)) {
#ifdef EXTENSIVE_TRACING
shepherd_trace("pty_to_commlib: reading from pipe_out");
#endif
ret = append_to_buf(g_fd_pipe_out[0], stdout_buf, &stdout_bytes);
#ifdef EXTENSIVE_TRACING
trace_buf(stdout_buf, ret, "pty_to_commlib: appended %d bytes, stdout_buf = ",
ret, stdout_buf);
#endif
}
if (ret >= 0 && g_fd_pipe_err[0] != -1
&& FD_ISSET(g_fd_pipe_err[0], &read_fds)) {
#ifdef EXTENSIVE_TRACING
shepherd_trace("pty_to_commlib: reading from pipe_err");
#endif
ret = append_to_buf(g_fd_pipe_err[0], stderr_buf, &stderr_bytes);
#ifdef EXTENSIVE_TRACING
trace_buf(stderr_buf, ret, "pty_to_commlib: appended %d bytes, stderr_buf = ",
ret, stderr_buf);
#endif
}
if (ret < 0) {
/* A fd was closed, likely our child has exited, we can exit, too. */
shepherd_trace("pty_to_commlib: append_to_buf() returned %d, "
"errno %d, %s -> exiting", ret, errno, strerror(errno));
do_exit = 1;
}
}
/* Always send stderr buffer immediately */
if (stderr_bytes != 0) {
ret = send_buf(stderr_buf, stderr_bytes, STDERR_DATA_MSG);
if (ret == 0) {
stderr_bytes = 0;
} else {
shepherd_trace("pty_to_commlib: send_buf() returned %d "
"-> exiting", ret);
do_exit = 1;
}
}
/*
* Send stdout_buf if there is enough data in it
* OR if there was a select timeout (don't wait too long to send data)
* OR if we will exit the loop now and there is data in the stdout_buf
*/
if (stdout_bytes >= 256 ||
(b_select_timeout == true && stdout_bytes > 0) ||
(do_exit == 1 && stdout_bytes > 0)) {
#ifdef EXTENSIVE_TRACING
shepherd_trace("pty_to_commlib: sending stdout buffer");
#endif
ret = send_buf(stdout_buf, stdout_bytes, STDOUT_DATA_MSG);
stdout_bytes = 0;
if (ret != 0) {
shepherd_trace("pty_to_commlib: send_buf() failed -> exiting");
do_exit = 1;
}
comm_flush_write_messages(g_comm_handle, &err_msg);
}
}
#ifdef EXTENSIVE_TRACING
shepherd_trace("pty_to_commlib: shutting down thread");
#endif
FREE(stdout_buf);
FREE(stderr_buf);
thread_func_cleanup(t_conf);
/* TODO: This could cause race conditions in the main thread, replace with pthread_condition */
#ifdef EXTENSIVE_TRACING
shepherd_trace("pty_to_commlib: raising event for main thread");
#endif
g_raised_event = 2;
thread_trigger_event(&g_thread_main);
sge_dstring_free(&err_msg);
#ifdef EXTENSIVE_TRACING
shepherd_trace("pty_to_commlib: leaving pty_to_commlib thread");
#endif
return NULL;
}
/****** commlib_to_pty() *******************************************************
* NAME
* commlib_to_pty() -- commlib_to_pty thread entry point and main loop
*
* SYNOPSIS
* void* commlib_to_pty(void *t_conf)
*
* FUNCTION
* Entry point and main loop of the commlib_to_pty thread.
* Reads data from the commlib and writes it to the pty.
*
* INPUTS
* void *t_conf - pointer to cl_thread_settings_t struct of the thread
*
* RESULT
* void* - always NULL
*
* NOTES
* MT-NOTE:
*
* SEE ALSO
*******************************************************************************/
static void* commlib_to_pty(void *t_conf)
{
recv_message_t recv_mess;
int b_was_connected = 0;
bool b_sent_to_child = false;
int ret;
int do_exit = 0;
int fd_write = -1;
dstring err_msg = DSTRING_INIT;
/* report to thread lib that thread starts up now */
thread_func_startup(t_conf);
thread_setcancelstate(1);
if (g_ptym != -1) {
fd_write = g_ptym;
} else if (g_fd_pipe_in[1] != -1) {
fd_write = g_fd_pipe_in[1];
} else {
do_exit = 1;
shepherd_trace("commlib_to_pty: no valid handle for stdin available. Exiting!");
}
cl_com_set_synchron_receive_timeout(g_comm_handle, 1);
while (do_exit == 0) {
/* wait blocking for a message from commlib */
recv_mess.cl_message = NULL;
recv_mess.data = NULL;
sge_dstring_free(&err_msg);
sge_dstring_sprintf(&err_msg, "");
ret = comm_recv_message(g_comm_handle, CL_TRUE, &recv_mess, &err_msg);
/*
* Check if the thread was cancelled. Exit thread if it was.
* It shouldn't be neccessary to do the check here, as the cancel state
* of the thread is 1, i.e. the thread may be cancelled at any time,
* but this doesn't work on some architectures (Darwin, older Solaris).
*/
thread_testcancel(t_conf);
if (g_raised_event > 0) {
do_exit = 1;
continue;
}
#ifdef EXTENSIVE_TRACING
shepherd_trace("commlib_to_pty: comm_recv_message() returned %d, err_msg: %s",
ret, sge_dstring_get_string(&err_msg));
#endif
if (ret != COMM_RETVAL_OK) {
/* handle error cases */
switch (ret) {
case COMM_NO_SELECT_DESCRIPTORS:
/*
* As long as we're not connected, this return value is expected.
* If we were already connected, it means the connection was closed.
*/
if (b_was_connected == 1) {
shepherd_trace("commlib_to_pty: was connected, "
"but lost connection -> exiting");
do_exit = 1;
}
break;
case COMM_CONNECTION_NOT_FOUND:
if (b_was_connected == 0) {
shepherd_trace("commlib_to_pty: our server is not running -> exiting");
/*
* On some architectures (e.g. Darwin), the child doesn't recognize
* when the pipe gets closed, so we have to send the information
* that the parent will exit soon to the child.
* "noshell" may be 0 or 1, everything else indicates an error.
*/
if (b_sent_to_child == false) {
if (write(g_fd_pipe_to_child[1], "noshell = 9", 11) != 11) {
shepherd_trace("commlib_to_pty: error in communicating "
"with child -> exiting");
} else {
b_sent_to_child = true;
}
}
} else {
shepherd_trace("commlib_to_pty: was connected and still have "
"selectors, but lost connection -> exiting");
}
do_exit = 1;
break;
case COMM_INVALID_PARAMETER:
shepherd_trace("commlib_to_pty: communication handle or "
"message buffer is invalid -> exiting");
do_exit = 1;
break;
case COMM_CANT_TRIGGER:
shepherd_trace("commlib_to_pty: can't trigger communication, likely the "
"communication was shut down by other thread -> exiting");
shepherd_trace("commlib_to_pty: err_msg: %s",
sge_dstring_get_string(&err_msg));
do_exit = 1;
break;
case COMM_CANT_RECEIVE_MESSAGE:
if (check_client_alive(g_comm_handle,
COMM_SERVER, &err_msg) != COMM_RETVAL_OK) {
shepherd_trace("commlib_to_pty: not connected any more -> exiting.");
do_exit = 1;
} else {
#ifdef EXTENSIVE_TRACING
shepherd_trace("commlib_to_pty: can't receive message, reason: %s "
"-> trying again",
sge_dstring_get_string(&err_msg));
#endif
}
b_was_connected = 1;
break;
case COMM_GOT_TIMEOUT:
#ifdef EXTENSIVE_TRACING
shepherd_trace("commlib_to_pty: got timeout -> trying again");
#endif
b_was_connected = 1;
break;
case COMM_SELECT_INTERRUPT:
/* Don't do tracing here */
/* shepherd_trace("commlib_to_pty: interrupted select"); */
b_was_connected = 1;
break;
default:
/* Unknown error, just try again */
#ifdef EXTENSIVE_TRACING
shepherd_trace("commlib_to_pty: comm_recv_message() returned %d -> "
"trying again\n", ret);
#endif
b_was_connected = 1;
break;
}
} else { /* if (ret == COMM_RETVAL_OK) */
/* We received a message, 'parse' it */
switch (recv_mess.type) {
case STDIN_DATA_MSG:
/* data message, write data to stdin of child */
#ifdef EXTENSIVE_TRACING
shepherd_trace("commlib_to_pty: received data message");
shepherd_trace("commlib_to_pty: writing data to stdin of child, "
"length = %d", recv_mess.cl_message->message_length-1);
#endif
if (sge_writenbytes(fd_write,
recv_mess.data,
(int)(recv_mess.cl_message->message_length-1))
!= (int)(recv_mess.cl_message->message_length-1)) {
shepherd_trace("commlib_to_pty: error writing to stdin of "
"child: %d, %s", errno, strerror(errno));
}
b_was_connected = 1;
break;
case WINDOW_SIZE_CTRL_MSG:
/* control message, set size of pty */
shepherd_trace("commlib_to_pty: received window size message, "
"changing window size");
ioctl(g_ptym, TIOCSWINSZ, &(recv_mess.ws));
b_was_connected = 1;
break;
case SETTINGS_CTRL_MSG:
/* control message */
shepherd_trace("commlib_to_pty: recevied settings message\n");
/* Forward the settings to the child process.
* This is also tells the child process that it can start
* the job 'in' the pty now.
*/
shepherd_trace("commlib_to_pty: writing to child %d bytes: %s",
strlen(recv_mess.data), recv_mess.data);
if (write(g_fd_pipe_to_child[1], recv_mess.data,
strlen(recv_mess.data)) != strlen(recv_mess.data)) {
shepherd_trace("commlib_to_pty: error in communicating "
"with child -> exiting");
do_exit = 1;
} else {
b_sent_to_child = true;
}
b_was_connected = 1;
break;
default:
shepherd_trace("commlib_to_pty: received unknown message");
break;
}
}
comm_free_message(&recv_mess, &err_msg);
}
/*
* When we get here, likely the commlib connection was shut down
* from the other side. We have to kill the job here to wake up
* the main thread.
*/
/*
* TODO: Use SIGINT if qrsh client was quit with Ctrl-C
*/
shepherd_signal_job(g_job_pid, SIGKILL);
sge_dstring_free(&err_msg);
#ifdef EXTENSIVE_TRACING
shepherd_trace("commlib_to_pty: leaving commlib_to_pty function\n");
#endif
thread_func_cleanup(t_conf);
/* TODO: pthread_condition, see other thread*/
#ifdef EXTENSIVE_TRACING
shepherd_trace("commlib_to_pty: raising event for main thread");
#endif
g_raised_event = 3;
thread_trigger_event(&g_thread_main);
#ifdef EXTENSIVE_TRACING
shepherd_trace("commlib_to_pty: leaving commlib_to_pty thread");
#endif
return NULL;
}
int parent_loop(char *hostname, int port, int ptym,
int *fd_pipe_in, int *fd_pipe_out, int *fd_pipe_err,
int *fd_pipe_to_child,
int ckpt_pid, int ckpt_type, int timeout,
int ckpt_interval, char *childname,
char *user_name, int *exit_status, struct rusage *rusage,
int job_pid, dstring *err_msg)
{
int ret;
THREAD_LIB_HANDLE *thread_lib_handle = NULL;
THREAD_HANDLE *thread_pty_to_commlib = NULL;
THREAD_HANDLE *thread_commlib_to_pty = NULL;
cl_raw_list_t *cl_com_log_list = NULL;
shepherd_trace("main: starting parent_loop");
g_hostname = strdup(hostname);
g_ptym = ptym;
g_fd_pipe_in[0] = fd_pipe_in[0];
g_fd_pipe_in[1] = fd_pipe_in[1];
g_fd_pipe_out[0] = fd_pipe_out[0];
g_fd_pipe_out[1] = fd_pipe_out[1];
g_fd_pipe_err[0] = fd_pipe_err[0];
g_fd_pipe_err[1] = fd_pipe_err[1];
g_fd_pipe_to_child[0] = fd_pipe_to_child[0];
g_fd_pipe_to_child[1] = fd_pipe_to_child[1];
g_job_pid = job_pid;
/*
* Initialize err_msg, so it's never NULL.
*/
sge_dstring_sprintf(err_msg, "");
ret = comm_init_lib(err_msg);
if (ret != 0) {
shepherd_trace("main: can't open communication library");
return 1;
}
/*
* Get log list of communication before a connection is opened.
*/
cl_com_log_list = cl_com_get_log_list();
/*
* Open the connection port so we can connect to our server
*/
shepherd_trace("main: opening connection to qrsh/qlogin client");
ret = comm_open_connection(false, g_csp_mode, COMM_CLIENT, port, COMM_SERVER,
user_name, &g_comm_handle, err_msg);
if (ret != COMM_RETVAL_OK) {
shepherd_trace("main: can't open commlib stream, err_msg = %s",
sge_dstring_get_string(err_msg));
return 1;
}
/*
* register at qrsh/qlogin client, which is the server of the communication.
* The answer of the server (a WINDOW_SIZE_CTRL_MSG) will be handled in the
* commlib_to_pty thread.
*/
shepherd_trace("main: sending REGISTER_CTRL_MSG");
ret = (int)comm_write_message(g_comm_handle, g_hostname, COMM_SERVER, 1,
(unsigned char*)" ", 1, REGISTER_CTRL_MSG, err_msg);
if (ret == 0) {
/* No bytes written - error */
shepherd_trace("main: can't send REGISTER_CTRL_MSG, comm_write_message() "
"returned: %s", sge_dstring_get_string(err_msg));
/* Don't exit here, the error handling is done in the commlib_to_tty-thread */
/* Most likely, the qrsh client is not running, so it's not the shepherds fault,
* we shouldn't return 1 (which leads to a shepherd_exit which sets the whole
* queue in error!).
*/
/* return 1;*/
}
#ifdef EXTENSIVE_TRACING
else {
shepherd_trace("main: Sent %d bytes to qrsh client", ret);
}
#endif
/*
* Setup thread list.
*/
ret = thread_init_lib(&thread_lib_handle);
/*
* Register this main thread at the thread library, so it can
* be triggered and create two worker threads.
*/
ret = register_thread(thread_lib_handle, &g_thread_main, "main thread");
if (ret != CL_RETVAL_OK) {
shepherd_trace("main: registering main thread failed: %d", ret);
return 1;
}
{
sigset_t old_sigmask;
sge_thread_block_all_signals(&old_sigmask);
ret = create_thread(thread_lib_handle,
&thread_pty_to_commlib,
cl_com_log_list,
"pty_to_commlib thread",
2,
pty_to_commlib);
if (ret != CL_RETVAL_OK) {
shepherd_trace("main: creating pty_to_commlib thread failed: %d", ret);
}
ret = create_thread(thread_lib_handle,
&thread_commlib_to_pty,
cl_com_log_list,
"commlib_to_pty thread",
3,
commlib_to_pty);
pthread_sigmask(SIG_SETMASK, &old_sigmask, NULL);
}
if (ret != CL_RETVAL_OK) {
shepherd_trace("main: creating commlib_to_pty thread failed: %d", ret);
}
/* From here on, the two worker threads are doing all the work.
* This main thread is just waiting until one of the to worker threads
* wants to exit and sends an event to the main thread.
* A worker thread wants to exit when either the communciation to
* the server was shut down or the user application (likely the shell)
* exited.
* On some architectures, this thread awakens from the wait whenever
* a signal arrives. Therefore we have to check if it was a signal
* or a event that awoke this thread.
*/
shepherd_trace("main: created both worker threads, now waiting jobs end");
*exit_status = wait_my_child(job_pid, ckpt_pid, ckpt_type, rusage, timeout,
ckpt_interval, childname);
alarm(0);
shepherd_trace("main: wait_my_child returned exit_status = %d", *exit_status);
shepherd_trace("main: rusage.ru_stime.tv_sec = %d", rusage->ru_stime.tv_sec);
shepherd_trace("main: rusage.ru_stime.tv_usec = %d", rusage->ru_stime.tv_usec);
shepherd_trace("main: rusage.ru_utime.tv_sec = %d", rusage->ru_utime.tv_sec);
shepherd_trace("main: rusage.ru_utime.tv_usec = %d", rusage->ru_utime.tv_usec);
/*
* We are sure the job exited when we get here, but there could still be
* some output in the buffers, so wait for the communication threads
* to give them time to read, transmit and flush the buffers.
*/
while (g_raised_event == 0) {
ret = thread_wait_for_event(&g_thread_main, 0, 0);
}
shepherd_trace("main: received event %d, g_raised_event = %d\n",
ret, g_raised_event);
/*
* One of the threads sent an event, so shut down both threads now
*/
shepherd_trace("main: shutting down pty_to_commlib thread");
/*
* It seems that tracing between thread_shutdown() and thread_join()
* could cause deadlocks!
*/
ret = thread_shutdown(thread_pty_to_commlib);
ret = thread_shutdown(thread_commlib_to_pty);
close(g_ptym);
close(g_fd_pipe_in[0]);
close(g_fd_pipe_in[1]);
close(g_fd_pipe_out[0]);
close(g_fd_pipe_out[1]);
close(g_fd_pipe_err[0]);
close(g_fd_pipe_err[1]);
/*
* Wait until the pty_to_commlib threads have shut down
*/
thread_join(thread_pty_to_commlib);
shepherd_trace("main: joined pty_to_commlib thread");
#if 0
/* This causes problems on some architectures - why? */
DPRINTF(("main: cl_thread_join(thread_commlib_to_pty)\n"));
thread_join(thread_commlib_to_pty);
#endif
/*
* Wait for all messages to be sent
* TODO: This shouldn't be neccessary, the shutdown of the
* connection should do a flush.
* Test it, and if it works, remove this code.
*/
shepherd_trace("main: flushing messages");
if (g_comm_handle != NULL) {
comm_flush_write_messages(g_comm_handle, err_msg);
}
#if 0
{
struct timeb ts;
ftime(&ts);
shepherd_trace("+++++ timestamp: %d.%03d ++++", (int)ts.time, (int)ts.millitm);
}
#endif
/* From here on, only the main thread is running */
thread_cleanup_lib(&thread_lib_handle);
/* The communication will be freed in close_parent_loop() */
sge_dstring_free(err_msg);
shepherd_trace("main: leaving main loop");
return 0;
}
int close_parent_loop(int exit_status)
{
int ret = 0;
char sz_exit_status[21];
dstring err_msg = DSTRING_INIT;
/*
* Send UNREGISTER_CTRL_MSG
*/
snprintf(sz_exit_status, 20, "%d", exit_status);
shepherd_trace("sending UNREGISTER_CTRL_MSG with exit_status = \"%s\"",
sz_exit_status);
shepherd_trace("sending to host: %s",
g_hostname != NULL ? g_hostname : "<null>");
ret = (int)comm_write_message(g_comm_handle, g_hostname,
COMM_SERVER, 1, (unsigned char*)sz_exit_status, strlen(sz_exit_status),
UNREGISTER_CTRL_MSG, &err_msg);
if (ret != strlen(sz_exit_status)) {
shepherd_trace("comm_write_message returned: %s",
sge_dstring_get_string(&err_msg));
shepherd_trace("close_parent_loop: comm_write_message() returned %d "
"instead of %d!!!", ret, strlen(sz_exit_status));
}
/*
* Wait for UNREGISTER_RESPONSE_CTRL_MSG
*/
{
int count = 0;
recv_message_t recv_mess;
shepherd_trace("waiting for UNREGISTER_RESPONSE_CTRL_MSG");
while (count < RESPONSE_MSG_TIMEOUT) {
#if defined(INTERIX)
/*
* TODO: comm_recv_message() should return immediatley when the server
* is not running any more. On Interix, it waits until a timeout
* occurs (60s), so we check if the server is running before
* we wait for the message.
* This has to be fixed in comm_recv_message() or the commlib.
*/
if (check_client_alive(g_comm_handle,
COMM_SERVER, &err_msg) != COMM_RETVAL_OK) {
shepherd_trace("Server already exited");
break;
}
#endif
ret = comm_recv_message(g_comm_handle, CL_TRUE, &recv_mess, &err_msg);
if (ret == COMM_GOT_TIMEOUT) {
count++;
} else if (recv_mess.type == UNREGISTER_RESPONSE_CTRL_MSG) {
shepherd_trace("Received UNREGISTER_RESPONSE_CTRL_MSG");
comm_free_message(&recv_mess, &err_msg);
break;
} else {
shepherd_trace("No connection or timeout while waiting for message");
break;
}
comm_free_message(&recv_mess, &err_msg);
}
}
/* Now we are completely logged of from the server and can shut down */
/*
* Tell the communication to shut down immediately, don't wait for
* the next timeout
*/
shepherd_trace("main: cl_com_ignore_timeouts");
comm_ignore_timeouts(true, &err_msg);
/*
* Do cleanup
*/
shepherd_trace("main: cl_com_cleanup_commlib()");
comm_shutdown_connection(g_comm_handle, COMM_SERVER, &err_msg);
shepherd_trace("main: comm_cleanup_lib()");
ret = comm_cleanup_lib(&err_msg);
if (ret != COMM_RETVAL_OK) {
shepherd_trace("main: error in comm_cleanup_lib(): %d", ret);
}
FREE(g_hostname);
sge_dstring_free(&err_msg);
shepherd_trace("main: leaving closinge_parent_loop()");
return 0;
}
|