stacktrace: Add flexibility in stacktrace ouptut

- New MCA option: opal_stacktrace_output
   - Specifies where the stack trace output stream goes.
   - Accepts: none, stdout, stderr, file[:filename]
   - Default filename 'stacktrace'
     - Filename will be `stacktrace.PID`, or if VPID is available,
       then the filename will be `stacktrace.VPID.PID`
 - Update util/stacktrace to allow for different output avenues
   including files. Previously this was hardcoded to 'stderr'.
 - Since opal_backtrace_print needs to be signal safe, passing it a
   FILE object that actually represents a file stream is difficult. This
   is because we cannot open the file in the signal handler using
   `fopen` (not safe), but have to use `open` (safe). Additionally, we
   cannot use `fdopen` to convert the `int fd` to a `FILE *fh` since it
   is also not signal safe.
   - I did not want to break the backtrace.h API so I introduced a new
     rule (documented in `backtrace.c`) that if the `FILE *file`
     argument is `NULL` then look for the `opal_stacktrace_output_fileno`
     variable to tell you which file descriptor to use for output.

Signed-off-by: Joshua Hursey <jhursey@us.ibm.com>

opal/mca/backtrace/backtrace.h

@@ -12,6 +12,7 @@
  *                         All rights reserved.
  * Copyright (c) 2015      Los Alamos National Security, LLC. All rights
  *                         reserved.
+ * Copyright (c) 2017      IBM Corporation.  All rights reserved.
  * $COPYRIGHT$
  *
  * Additional copyrights may follow
@@ -26,6 +27,7 @@
 
 #include "opal/mca/mca.h"
 #include "opal/mca/base/base.h"
+#include "opal/util/stacktrace.h"
 
 BEGIN_C_DECLS
 
@@ -39,6 +41,8 @@ BEGIN_C_DECLS
 /*
  * Print back trace to FILE file with a prefix for each line.
  * First strip lines are not printed.
+ * If 'file' is NULL then the component should try to use the file descriptor
+ * saved in opal_stacktrace_output_fileno
  *
  * \note some attempts made to be signal safe.
  */

opal/mca/backtrace/execinfo/backtrace_execinfo.c

@@ -10,6 +10,7 @@
  * Copyright (c) 2004-2006 The Regents of the University of California.
  *                         All rights reserved.
  * Copyright (c) 2011 Cisco Systems, Inc.  All rights reserved.
+ * Copyright (c) 2017      IBM Corporation.  All rights reserved.
  * $COPYRIGHT$
  *
  * Additional copyrights may follow
@@ -34,12 +35,16 @@
 int
 opal_backtrace_print(FILE *file, char *prefix, int strip)
 {
-    int i, fd, len;
+    int i, len;
     int trace_size;
     void * trace[32];
     char buf[6];
+    int fd = opal_stacktrace_output_fileno;
+
+    if( NULL != file ) {
+        fd = fileno(file);
+    }
 
-    fd = fileno (file);
     if (-1 == fd) {
         return OPAL_ERR_BAD_PARAM;
     }

opal/mca/backtrace/printstack/backtrace_printstack.c

@@ -10,6 +10,7 @@
  * Copyright (c) 2004-2006 The Regents of the University of California.
  *                         All rights reserved.
  * Copyright (c) 2006      Sun Microsystems, Inc.  All rights reserved.
+ * Copyright (c) 2017      IBM Corporation.  All rights reserved.
  * $COPYRIGHT$
  *
  * Additional copyrights may follow
@@ -27,7 +28,13 @@
 int
 opal_backtrace_print(FILE *file, char *prefix, int strip)
 {
-    printstack(fileno(file));
+    int fd = opal_stacktrace_output_fileno;
+
+    if( NULL != file ) {
+        fd = fileno(file);
+    }
+
+    printstack(fd);
 
     return OPAL_SUCCESS;
 }

opal/runtime/opal_params.c

@@ -21,6 +21,7 @@
  *                         and Technology (RIST). All rights reserved.
  * Copyright (c) 2015      Mellanox Technologies, Inc.
  *                         All rights reserved.
+ * Copyright (c) 2017      IBM Corporation.  All rights reserved.
  * $COPYRIGHT$
  *
  * Additional copyrights may follow
@@ -48,6 +49,7 @@
 #include "opal/util/timings.h"
 
 char *opal_signal_string = NULL;
+char *opal_stacktrace_output_filename = NULL;
 char *opal_net_private_ipv4 = NULL;
 char *opal_set_max_sys_limits = NULL;
 
@@ -76,6 +78,7 @@ static bool opal_register_done = false;
 int opal_register_params(void)
 {
     int ret;
+    char *string = NULL;
 
     if (opal_register_done) {
         return OPAL_SUCCESS;
@@ -87,7 +90,6 @@ int opal_register_params(void)
      * This string is going to be used in opal/util/stacktrace.c
      */
     {
-        char *string = NULL;
         int j;
         int signals[] = {
 #ifdef SIGABRT
@@ -127,6 +129,28 @@ int opal_register_params(void)
         }
     }
 
+    /*
+     * Where should the stack trace output be directed
+     * This string is going to be used in opal/util/stacktrace.c
+     */
+    string = strdup("stderr");
+    opal_stacktrace_output_filename = string;
+    ret = mca_base_var_register ("opal", "opal", NULL, "stacktrace_output",
+                                 "Specifies where the stack trace output stream goes.  "
+                                 "Accepts one of the following: none (disabled), stderr (default), stdout, file[:filename].   "
+                                 "If 'filename' is not specified, a default filename of 'stacktrace' is used.  "
+                                 "The 'filename' is appended with either '.PID' or '.RANK.PID', if RANK is available.  "
+                                 "The 'filename' can be an absolute path or a relative path to the current working directory.",
+                                 MCA_BASE_VAR_TYPE_STRING, NULL, 0, MCA_BASE_VAR_FLAG_SETTABLE,
+                                 OPAL_INFO_LVL_3,
+                                 MCA_BASE_VAR_SCOPE_LOCAL,
+                                 &opal_stacktrace_output_filename);
+    free (string);
+    if (0 > ret) {
+        return ret;
+    }
+
+
 #if defined(HAVE_SCHED_YIELD)
     opal_progress_yield_when_idle = false;
     ret = mca_base_var_register ("opal", "opal", "progress", "yield_when_idle",

opal/runtime/opal_params.h

@@ -18,6 +18,7 @@
  * Copyright (c) 2014      Hochschule Esslingen.  All rights reserved.
  * Copyright (c) 2015      Mellanox Technologies, Inc.
  *                         All rights reserved.
+ * Copyright (c) 2017      IBM Corporation.  All rights reserved.
  * $COPYRIGHT$
  *
  * Additional copyrights may follow
@@ -29,6 +30,7 @@
 #define OPAL_PARAMS_H
 
 extern char *opal_signal_string;
+extern char *opal_stacktrace_output_filename;
 extern char *opal_net_private_ipv4;
 extern char *opal_set_max_sys_limits;
 

opal/util/stacktrace.c

@@ -25,6 +25,15 @@
 #ifdef HAVE_UNISTD_H
 #include <unistd.h>
 #endif
+#ifdef HAVE_SYS_TYPES_H
+#include <sys/types.h>
+#endif
+#ifdef HAVE_SYS_STAT_H
+#include <sys/stat.h>
+#endif
+#ifdef HAVE_SYS_FCNTL_H
+#include <fcntl.h>
+#endif
 
 #include <string.h>
 #include <signal.h>
@@ -35,6 +44,7 @@
 #include "opal/util/output.h"
 #include "opal/util/show_help.h"
 #include "opal/util/argv.h"
+#include "opal/util/proc.h"
 #include "opal/runtime/opal_params.h"
 
 #ifndef _NSIG
@@ -43,9 +53,35 @@
 
 #define HOSTFORMAT "[%s:%05d] "
 
+int    opal_stacktrace_output_fileno = -1;
+static char  *opal_stacktrace_output_filename_base = NULL;
+static size_t opal_stacktrace_output_filename_max_len = 0;
 static char stacktrace_hostname[OPAL_MAXHOSTNAMELEN];
 static char *unable_to_print_msg = "Unable to print stack trace!\n";
 
+/*
+ * Set the stacktrace filename:
+ * stacktrace.PID
+ * -or, if VPID is available-
+ * stacktrace.VPID.PID
+ */
+static void set_stacktrace_filename(void) {
+    opal_proc_t *my_proc = opal_proc_local_get();
+
+    if( NULL == my_proc ) {
+        snprintf(opal_stacktrace_output_filename, opal_stacktrace_output_filename_max_len,
+                 "%s.%lu",
+                 opal_stacktrace_output_filename_base, (unsigned long)getpid());
+    }
+    else {
+        snprintf(opal_stacktrace_output_filename, opal_stacktrace_output_filename_max_len,
+                 "%s.%lu.%lu",
+                 opal_stacktrace_output_filename_base, (unsigned long)my_proc->proc_name.vpid, (unsigned long)getpid());
+    }
+
+    return;
+}
+
 /**
  * This function is being called as a signal-handler in response
  * to a user-specified signal (e.g. SIGFPE or SIGSEGV).
@@ -69,12 +105,37 @@ static void show_stackframe (int signo, siginfo_t * info, void * p)
     int ret;
     char *si_code_str = "";
 
+    /* Do not print the stack trace */
+    if( 0 > opal_stacktrace_output_fileno && 0 == opal_stacktrace_output_filename_max_len ) {
+        /* Raise the signal again, so we don't accidentally mask critical signals.
+         * For critical signals, it is preferred that we call 'raise' instead of
+         * 'exit' or 'abort' so that the return status is set properly for this
+         * process.
+         */
+        signal(signo, SIG_DFL);
+        raise(signo);
+
+        return;
+    }
+
+    /* Update the file name with the RANK, if available */
+    if( 0 < opal_stacktrace_output_filename_max_len ) {
+        set_stacktrace_filename();
+        opal_stacktrace_output_fileno = open(opal_stacktrace_output_filename,
+                                             O_CREAT|O_WRONLY|O_TRUNC, S_IRUSR|S_IWUSR);
+        if( 0 > opal_stacktrace_output_fileno ) {
+            opal_output(0, "Error: Failed to open the stacktrace output file. Default: stderr\n\tFilename: %s\n\tErrno: %s",
+                        opal_stacktrace_output_filename, strerror(errno));
+            opal_stacktrace_output_fileno = fileno(stderr);
+        }
+    }
+
     /* write out the footer information */
     memset (print_buffer, 0, sizeof (print_buffer));
     ret = snprintf(print_buffer, sizeof(print_buffer),
                    HOSTFORMAT "*** Process received signal ***\n",
                    stacktrace_hostname, getpid());
-    write(fileno(stderr), print_buffer, ret);
+    write(opal_stacktrace_output_fileno, print_buffer, ret);
 
 
     memset (print_buffer, 0, sizeof (print_buffer));
@@ -324,14 +385,14 @@ static void show_stackframe (int signo, siginfo_t * info, void * p)
     }
 
     /* write out the signal information generated above */
-    write(fileno(stderr), print_buffer, sizeof(print_buffer)-size);
+    write(opal_stacktrace_output_fileno, print_buffer, sizeof(print_buffer)-size);
 
     /* print out the stack trace */
     snprintf(print_buffer, sizeof(print_buffer), HOSTFORMAT,
              stacktrace_hostname, getpid());
-    ret = opal_backtrace_print(stderr, print_buffer, 2);
+    ret = opal_backtrace_print(NULL, print_buffer, 2);
     if (OPAL_SUCCESS != ret) {
-        write(fileno(stderr), unable_to_print_msg, strlen(unable_to_print_msg));
+        write(opal_stacktrace_output_fileno, unable_to_print_msg, strlen(unable_to_print_msg));
     }
 
     /* write out the footer information */
@@ -340,9 +401,15 @@ static void show_stackframe (int signo, siginfo_t * info, void * p)
                    HOSTFORMAT "*** End of error message ***\n",
                    stacktrace_hostname, getpid());
     if (ret > 0) {
-        write(fileno(stderr), print_buffer, ret);
+        write(opal_stacktrace_output_fileno, print_buffer, ret);
     } else {
-        write(fileno(stderr), unable_to_print_msg, strlen(unable_to_print_msg));
+        write(opal_stacktrace_output_fileno, unable_to_print_msg, strlen(unable_to_print_msg));
+    }
+
+    if( fileno(stdout) != opal_stacktrace_output_fileno &&
+        fileno(stderr) != opal_stacktrace_output_fileno ) {
+        close(opal_stacktrace_output_fileno);
+        opal_stacktrace_output_fileno = -1;
     }
 
     /* Raise the signal again, so we don't accidentally mask critical signals.
@@ -373,7 +440,30 @@ void opal_stackframe_output(int stream)
             opal_output(stream, "%s", traces[i]);
         }
     } else {
-        opal_backtrace_print(stderr, NULL, 2);
+        /* Do not print the stack trace */
+        if( 0 > opal_stacktrace_output_fileno && 0 == opal_stacktrace_output_filename_max_len ) {
+            return;
+        }
+
+        /* Update the file name with the RANK, if available */
+        if( 0 < opal_stacktrace_output_filename_max_len ) {
+            set_stacktrace_filename();
+            opal_stacktrace_output_fileno = open(opal_stacktrace_output_filename,
+                                                 O_CREAT|O_WRONLY|O_TRUNC, S_IRUSR|S_IWUSR);
+            if( 0 > opal_stacktrace_output_fileno ) {
+                opal_output(0, "Error: Failed to open the stacktrace output file. Default: stderr\n\tFilename: %s\n\tErrno: %s",
+                            opal_stacktrace_output_filename, strerror(errno));
+                opal_stacktrace_output_fileno = fileno(stderr);
+            }
+        }
+
+        opal_backtrace_print(NULL, NULL, 2);
+
+        if( fileno(stdout) != opal_stacktrace_output_fileno &&
+            fileno(stderr) != opal_stacktrace_output_fileno ) {
+            close(opal_stacktrace_output_fileno);
+            opal_stacktrace_output_fileno = -1;
+        }
     }
 }
 
@@ -444,6 +534,50 @@ int opal_util_register_stackhandlers (void)
         }
     }
 
+    /* Setup the output stream to use */
+    if( NULL == opal_stacktrace_output_filename ||
+        0 == strcasecmp(opal_stacktrace_output_filename, "none") ) {
+        opal_stacktrace_output_fileno = -1;
+    }
+    else if( 0 == strcasecmp(opal_stacktrace_output_filename, "stdout") ) {
+        opal_stacktrace_output_fileno = fileno(stdout);
+    }
+    else if( 0 == strcasecmp(opal_stacktrace_output_filename, "stderr") ) {
+        opal_stacktrace_output_fileno = fileno(stdout);
+    }
+    else if( 0 == strcasecmp(opal_stacktrace_output_filename, "file" ) ||
+             0 == strcasecmp(opal_stacktrace_output_filename, "file:") ) {
+        opal_stacktrace_output_filename_base = strdup("stacktrace");
+
+        free(opal_stacktrace_output_filename);
+        // Magic number: 8 = space for .PID and .RANK (allow 7 digits each)
+        opal_stacktrace_output_filename_max_len = strlen("stacktrace") + 8 + 8;
+        opal_stacktrace_output_filename = (char*)malloc(sizeof(char) * opal_stacktrace_output_filename_max_len);
+        set_stacktrace_filename();
+        opal_stacktrace_output_fileno = -1;
+    }
+    else if( 0 == strncasecmp(opal_stacktrace_output_filename, "file:", 5) ) {
+        char *filename_cpy = NULL;
+        next = strchr(opal_stacktrace_output_filename, ':');
+        next++; // move past the ':' to the filename specified
+
+        opal_stacktrace_output_filename_base = strdup(next);
+
+        free(opal_stacktrace_output_filename);
+        // Magic number: 8 = space for .PID and .RANK (allow 7 digits each)
+        opal_stacktrace_output_filename_max_len = strlen(opal_stacktrace_output_filename_base) + 8 + 8;
+        opal_stacktrace_output_filename = (char*)malloc(sizeof(char) * opal_stacktrace_output_filename_max_len);
+        set_stacktrace_filename();
+        opal_stacktrace_output_fileno = -1;
+
+        free(filename_cpy);
+    }
+    else {
+        opal_stacktrace_output_fileno = fileno(stderr);
+    }
+
+
+    /* Setup the signals to catch */
     memset(&act, 0, sizeof(act));
     act.sa_sigaction = show_stackframe;
     act.sa_flags = SA_SIGINFO;

opal/util/stacktrace.h

@@ -10,6 +10,7 @@
  * Copyright (c) 2004-2005 The Regents of the University of California.
  *                         All rights reserved.
  * Copyright (c) 2008      Cisco Systems, Inc.  All rights reserved.
+ * Copyright (c) 2017      IBM Corporation.  All rights reserved.
  * $COPYRIGHT$
  *
  * Additional copyrights may follow
@@ -24,6 +25,12 @@
 
 #include "opal_config.h"
 
+/*
+ * File descriptor to be used by the backtrace framework if opal_backtrace_print
+ * is passed NULL for it's FILE file pointer.
+ */
+extern int opal_stacktrace_output_fileno;
+
 /**
  * Output the current stack trace (not including the call to this
  * function) to the stream indicated.