From fe10caee8dcb850bdb97db404947337e6918e2ed Mon Sep 17 00:00:00 2001
From: Jason Lowe <jlowe@apache.org>
Date: Wed, 25 May 2016 21:37:31 +0000
Subject: [PATCH] YARN-4459. container-executor should only kill process
 groups. Contributed by Jun Gong (cherry picked from commit
 1ba31fe9e906dbd093afd4b254216601967a4a7b)

Conflicts:

	hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/native/container-executor/test/test-container-executor.c
---
 .../impl/container-executor.c                 | 19 +++------
 .../test/test-container-executor.c            | 42 -------------------
 2 files changed, 5 insertions(+), 56 deletions(-)

diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/native/container-executor/impl/container-executor.c b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/native/container-executor/impl/container-executor.c
index 4e547705f65..d583b7efaa5 100644
--- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/native/container-executor/impl/container-executor.c
+++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/native/container-executor/impl/container-executor.c
@@ -1524,21 +1524,13 @@ int signal_container_as_user(const char *user, int pid, int sig) {
   }
 
   //Don't continue if the process-group is not alive anymore.
-  int has_group = 1;
   if (kill(-pid,0) < 0) {
-    if (kill(pid, 0) < 0) {
-      if (errno == ESRCH) {
-        return INVALID_CONTAINER_PID;
-      }
-      fprintf(LOGFILE, "Error signalling container %d with %d - %s\n",
-	      pid, sig, strerror(errno));
-      return -1;
-    } else {
-      has_group = 0;
-    }
+    fprintf(LOGFILE, "Error signalling not exist process group %d "
+            "with signal %d\n", pid, sig);
+    return INVALID_CONTAINER_PID;
   }
 
-  if (kill((has_group ? -1 : 1) * pid, sig) < 0) {
+  if (kill(-pid, sig) < 0) {
     if(errno != ESRCH) {
       fprintf(LOGFILE, 
               "Error signalling process group %d with signal %d - %s\n", 
@@ -1552,8 +1544,7 @@ int signal_container_as_user(const char *user, int pid, int sig) {
       return INVALID_CONTAINER_PID;
     }
   }
-  fprintf(LOGFILE, "Killing process %s%d with %d\n",
-	  (has_group ? "group " :""), pid, sig);
+  fprintf(LOGFILE, "Killing process group %d with %d\n", pid, sig);
   return 0;
 }
 
diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/native/container-executor/test/test-container-executor.c b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/native/container-executor/test/test-container-executor.c
index 3db75eab072..0306cb6fc6a 100644
--- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/native/container-executor/test/test-container-executor.c
+++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/native/container-executor/test/test-container-executor.c
@@ -469,47 +469,6 @@ void run_test_in_child(const char* test_name, void (*func)()) {
   }
 }
 
-void test_signal_container() {
-  printf("\nTesting signal_container\n");
-  fflush(stdout);
-  fflush(stderr);
-  pid_t child = fork();
-  if (child == -1) {
-    printf("FAIL: fork failed\n");
-    exit(1);
-  } else if (child == 0) {
-    printf("\nSwitching to user %d\n", user_detail->pw_uid);
-    if (change_user(user_detail->pw_uid, user_detail->pw_gid) != 0) {
-      exit(1);
-    }
-    sleep(3600);
-    exit(0);
-  } else {
-    printf("Child container launched as %" PRId64 "\n", (int64_t)child);
-    printf("Signaling container as user %s\n", yarn_username);
-    // there's a race condition for child calling change_user and us
-    // calling signal_container_as_user, hence sleeping
-    sleep(3);
-    if (signal_container_as_user(yarn_username, child, SIGQUIT) != 0) {
-      exit(1);
-    }
-    int status = 0;
-    if (waitpid(child, &status, 0) == -1) {
-      printf("FAIL: waitpid failed - %s\n", strerror(errno));
-      exit(1);
-    }
-    if (!WIFSIGNALED(status)) {
-      printf("FAIL: child wasn't signalled - %d\n", status);
-      exit(1);
-    }
-    if (WTERMSIG(status) != SIGQUIT) {
-      printf("FAIL: child was killed with %d instead of %d\n", 
-	     WTERMSIG(status), SIGQUIT);
-      exit(1);
-    }
-  }
-}
-
 void test_signal_container_group() {
   printf("\nTesting group signal_container\n");
   fflush(stdout);
@@ -825,7 +784,6 @@ int main(int argc, char **argv) {
 
   // the tests that change user need to be run in a subshell, so that
   // when they change user they don't give up our privs
-  run_test_in_child("test_signal_container", test_signal_container);
   run_test_in_child("test_signal_container_group", test_signal_container_group);
 
   // init app and run container can't be run if you aren't testing as root