utils/ps: add --children-of-pid

Show a PID and its hierarchy of children.

Refs #606

Signed-off-by: Brice Goglin <Brice.Goglin@inria.fr>

Author: bgoglin

contrib/completion/bash/hwloc

@@ -1,5 +1,5 @@
 #
-# Copyright © 2018-2022 Inria.  All rights reserved.
+# Copyright © 2018-2023 Inria.  All rights reserved.
 # See COPYING in top-level directory.
 #
 
@@ -480,6 +480,7 @@ complete -F _hwloc_distrib hwloc-distrib
 _hwloc_ps(){
     local OPTIONS=(-a
 		   --pid
+		   --children-of-pid
 		   --name
 		   --uid
 		   -l --logical
@@ -510,7 +511,7 @@ _hwloc_ps(){
 	--uid)
 	    COMPREPLY=( "<uid>" "all" "" )
 	    ;;
-	--pid)
+	--pid|--children-of-pid)
 	    COMPREPLY=( "<pid>" "" )
 	    ;;
 	--pid-cmd)

utils/hwloc/common-ps.c

@@ -393,3 +393,63 @@ int hwloc_ps_foreach_process(hwloc_topology_t topology, hwloc_const_bitmap_t top
   return -1;
 #endif /* HAVE_DIRENT_H */
 }
+
+int hwloc_ps_foreach_child(hwloc_topology_t topology, hwloc_const_bitmap_t topocpuset,
+                           long pid,
+                           void (*callback)(hwloc_topology_t topology, struct hwloc_ps_process *proc, void *cbdata),
+                           void *cbdata,
+                           unsigned long flags, const char *only_name, long uid)
+{
+#ifdef HAVE_DIRENT_H
+  struct hwloc_ps_process proc;
+  DIR *taskdir;
+  char path[512];
+
+  proc.pid = pid;
+  proc.cpuset = NULL;
+  proc.nthreads = 0;
+  proc.nboundthreads = 0;
+  proc.threads = NULL;
+  if (hwloc_ps_read_process(topology, topocpuset, &proc, flags) < 0)
+    goto next;
+  if (only_name && !strstr(proc.name, only_name))
+    goto next;
+  if (uid != HWLOC_PS_ALL_UIDS && proc.uid != HWLOC_PS_ALL_UIDS && proc.uid != uid)
+    goto next;
+  callback(topology, &proc, cbdata);
+ next:
+  hwloc_ps_free_process(&proc);
+
+  snprintf(path, sizeof(path), "/proc/%ld/task", proc.pid);
+  taskdir = opendir(path); /* should be enough for the vast majority of cases */
+  if (taskdir) {
+    struct dirent *taskdirent;
+    while ((taskdirent = readdir(taskdir))) {
+      char pidline[4096];
+      FILE *file;
+      char *begin;
+      size_t len;
+      snprintf(path, sizeof(path), "/proc/%ld/task/%s/children", proc.pid, taskdirent->d_name);
+      file = fopen(path, "r");
+      if (!file)
+        continue;
+      len = fread(pidline, 1, sizeof(pidline)-1, file);
+      fclose(file);
+      pidline[len] = '\0';
+      begin = pidline;
+      while (1) {
+        char *end;
+        long childpid = strtoul(begin, &end, 10);
+        if (end == begin)
+          break;
+        hwloc_ps_foreach_child(topology, topocpuset, childpid, callback, cbdata, flags, only_name, uid);
+        begin = end;
+      }
+    }
+    closedir(taskdir);
+  }
+  return 0;
+#else /* HAVE_DIRENT_H */
+  return -1;
+#endif /* HAVE_DIRENT_H */
+}

utils/hwloc/common-ps.h

@@ -1,5 +1,5 @@
 /*
- * Copyright © 2009-2021 Inria.  All rights reserved.
+ * Copyright © 2009-2023 Inria.  All rights reserved.
  * Copyright © 2009-2012 Université Bordeaux
  * Copyright © 2009-2011 Cisco Systems, Inc.  All rights reserved.
  * See COPYING in top-level directory.
@@ -45,6 +45,12 @@ int hwloc_ps_foreach_process(hwloc_topology_t topology, hwloc_const_bitmap_t top
 			     void *cbdata,
 			     unsigned long flags, const char *only_name, long only_uid);
 
+int hwloc_ps_foreach_child(hwloc_topology_t topology, hwloc_const_bitmap_t topocpuset,
+                           long pid,
+                           void (*callback)(hwloc_topology_t topology, struct hwloc_ps_process *proc, void *cbdata),
+			     void *cbdata,
+			     unsigned long flags, const char *only_name, long only_uid);
+
 void hwloc_ps_pidcmd(struct hwloc_ps_process *proc, const char *pidcmd);
 
 void hwloc_ps_free_process(struct hwloc_ps_process *proc);

utils/hwloc/hwloc-ps.1in

@@ -1,5 +1,5 @@
 .\" -*- nroff -*-
-.\" Copyright © 2010-2021 Inria.  All rights reserved.
+.\" Copyright © 2010-2023 Inria.  All rights reserved.
 .\" Copyright © 2009-2010 Cisco Systems, Inc.  All rights reserved.
 .\" See COPYING in top-level directory.
 .TH HWLOC-PS "1" "%HWLOC_DATE%" "%PACKAGE_VERSION%" "%PACKAGE_NAME%"
@@ -28,6 +28,10 @@ specific part of the machine.
 Only show process of PID \fI<pid>\fR,
 even if it is not bound to any specific part of the machine.
 .TP
+\fB\-\-children\-of\-pid <pid>\fR
+Only show process of PID \fI<pid>\fR and its hierarchy of children,
+even if they are not bound to any specific part of the machine.
+.TP
 \fB\-\-name <name>\fR
 Only show processes whose name contains \fI<name>\fR,
 even if they are not bound to any specific part of the machine.

utils/hwloc/hwloc-ps.c

@@ -1,5 +1,5 @@
 /*
- * Copyright © 2009-2022 Inria.  All rights reserved.
+ * Copyright © 2009-2023 Inria.  All rights reserved.
  * Copyright © 2009-2012 Université Bordeaux
  * Copyright © 2009-2011 Cisco Systems, Inc.  All rights reserved.
  * See COPYING in top-level directory.
@@ -35,6 +35,7 @@ static int logical = 1;
 static int single_ancestor = 0;
 #define NO_ONLY_PID -1
 static long only_pid = NO_ONLY_PID;
+static long children_of_pid = NO_ONLY_PID;
 static long only_uid;
 static int json_server = 0;
 static int json_port = JSON_PORT;
@@ -47,6 +48,7 @@ void usage(const char *name, FILE *where)
   fprintf (where, "Options:\n");
   fprintf (where, "  -a                 Show all processes, including those that are not bound\n");
   fprintf (where, "  --pid <pid>        Only show process of pid number <pid>\n");
+  fprintf (where, "  --children-of-pid <pid> Only show process of pid number <pid> and its children\n");
   fprintf (where, "  --name <name>      Only show processes whose name contains <name>\n");
 #ifdef HWLOC_LINUX_SYS
   fprintf (where, "  --uid <uid>        Only show processes of the user with the given uid\n");
@@ -197,7 +199,7 @@ static void foreach_process_cb(hwloc_topology_t topology,
   const char *pidcmd = cbdata;
 
   /* don't print anything if the process isn't bound and if no threads are bound and if not showing all */
-  if (!proc->bound && (!proc->nthreads || !proc->nboundthreads) && !show_all && !only_name)
+  if (!proc->bound && (!proc->nthreads || !proc->nboundthreads) && !show_all && !only_name && children_of_pid == NO_ONLY_PID)
     return;
 
   if (pidcmd)
@@ -212,7 +214,11 @@ static void foreach_process_cb(hwloc_topology_t topology,
 static int run(hwloc_topology_t topology, hwloc_const_bitmap_t topocpuset,
 	       unsigned long psflags, char *pidcmd)
 {
-  if (only_pid == NO_ONLY_PID) {
+  if (children_of_pid != NO_ONLY_PID) {
+    /* show children */
+    return hwloc_ps_foreach_child(topology, topocpuset, children_of_pid, foreach_process_cb, pidcmd, psflags, only_name, only_uid);
+
+  } else if (only_pid == NO_ONLY_PID) {
     /* show all */
     return hwloc_ps_foreach_process(topology, topocpuset, foreach_process_cb, pidcmd, psflags, only_name, only_uid);
 
@@ -317,6 +323,7 @@ run_json_server(hwloc_topology_t topology, hwloc_const_bitmap_t topocpuset)
 
       only_name = NULL;
       only_pid = NO_ONLY_PID;
+      children_of_pid = NO_ONLY_PID;
       current = req;
       while (*current) {
 	if (!strncmp(current, "lastcpulocation ", 16)) {
@@ -338,6 +345,10 @@ run_json_server(hwloc_topology_t topology, hwloc_const_bitmap_t topocpuset)
 	  psflags |= HWLOC_PS_FLAG_THREADS;
 	  show_all = 1;
 	  break;
+	} else if (!strncmp(current, "childrenofpid=", 14)) {
+	  children_of_pid = atoi(current+14);
+	  show_all = 1;
+	  break;
 	} else if (!strncmp(current, "name=", 5)) {
 	  only_name = current+5;
 	  show_all = 1;
@@ -418,6 +429,13 @@ int main(int argc, char *argv[])
       }
       only_pid = strtol(argv[1], NULL, 10);
       opt = 1;
+    } else if (!strcmp(argv[0], "--children-of-pid")) {
+      if (argc < 2) {
+	usage(callname, stderr);
+	exit(EXIT_FAILURE);
+      }
+      children_of_pid = strtol(argv[1], NULL, 10);
+      opt = 1;
     } else if (!strcmp(argv[0], "--name")) {
       if (argc < 2) {
 	usage(callname, stderr);