Skip to contentSkip to conversation

"pidns17" problem in "2010-02-11" cvs

LTP Mitani 3 participants 6 messages Feb 12, 2010

lore
Share & tools
lore

Mitani

author
#0 lore raw

Hi,

I tried to test with "2010-02-11" cvs in RHEL5.4 system. But the test brings termination of connection.

I examined the phenomenon and discovered that "pidns17" test made accident and sshd stopped after the test.

I had some problems.

1. I think that "tst_exit()" must be added last of "cleanup()" function.
============
--- ./testcases/kernel/containers/pidns/pidns17.c 2009-12-07
05:55:16.000000000 +0900
+++ ./testcases/kernel/containers/pidns/pidns17.c.new 2010-02-12
16:26:31.000000000 +0900
@@ -104,7 +104,6 @@

        /* cleanup and exit */
        CLEANUP();

- tst_exit();
}

/***********************************************************************
@@ -136,7 +135,6 @@

        /* cleanup and exit */
        CLEANUP();

- tst_exit();
} /* End main */

/*
@@ -147,4 +145,5 @@

 {
        /* Clean the test testcase as LTP wants*/
        TEST_CLEANUP;

+ tst_exit();
}
============

After revision, connection termination didn't occur.

2. I cannot understand the purpose that "kill()" function whose first parameter is "-1" is called.
------------

        if (kill(-1, SIGUSR1) == -1) {
                tst_resm(TBROK | TERRNO, "cinit: kill(-1, SIGUSR1) failed");
                CLEANUP();
        }

------------

If kill()'s first option is "-1", "man kill" says that "All processes with pid larger than 1 will be signaled."

Therefore, not only the "sshd" but also the other processes were affected, I think.

I'm glad if I could get some opinions.

Thank you--

-Tomonori Mitani

------------------------------------------------------------------------------
SOLARIS 10 is the OS for Data Centers - provides features such as DTrace, Predictive Self Healing and Award Winning ZFS. Get Solaris 10 NOW
http://p.sf.net/sfu/solaris-dev2dev
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

#1 lore raw
Quoted 35 lines · Mitani

Hi,

I tried to test with "2010-02-11" cvs in RHEL5.4 system. But the test brings termination of connection.

I examined the phenomenon and discovered that "pidns17" test made accident and sshd stopped after the test.

I had some problems.

1. I think that "tst_exit()" must be added last of "cleanup()" function.
============
--- ./testcases/kernel/containers/pidns/pidns17.c 2009-12-07
05:55:16.000000000 +0900
+++ ./testcases/kernel/containers/pidns/pidns17.c.new 2010-02-12
16:26:31.000000000 +0900

(quoted diff) +0 -2
@@ -104,7 +104,6 @@
 
        /* cleanup and exit */
        CLEANUP();
-       tst_exit();
 }
 
 /***********************************************************************
@@ -136,7 +135,6 @@
 
        /* cleanup and exit */
        CLEANUP();
-       tst_exit();
 }      /* End main */
 
 /*

Review line 149

Quoted 5 lines
(quoted diff) +1 -0
@@ -147,4 +145,5 @@
 {
        /* Clean the test testcase as LTP wants*/
        TEST_CLEANUP;
+       tst_exit();
 }

Yeah I'm afraid I don't understand what CLEANUP and tst_exit exactly do. Hopefully Garrett can give an educated answer.

Quoted 8 lines

============

After revision, connection termination didn't occur.

2. I cannot understand the purpose that "kill()" function whose first parameter is "-1" is called.
------------

Quoted 8 lines
        if (kill(-1, SIGUSR1) == -1) {
                tst_resm(TBROK | TERRNO, "cinit: kill(-1, SIGUSR1) failed");
                CLEANUP();
        }

------------

If kill()'s first option is "-1", "man kill" says that "All processes with pid larger than 1 will be signaled."

Right, the test is checking whether kill -1 inside a private pidns kills all processes besides init in the pid namespace.

Quoted 2 lines

Therefore, not only the "sshd" but also the other processes were affected, I think.

sshd is not in the private pid namespace and should not be killed. If it is being killed by the pid -1 inside the container, then there is a kernel bug.

Quoted 7 lines

I'm glad if I could get some opinions.

Thank you--

-Tomonori Mitani

Thanks for testing.

-serge

------------------------------------------------------------------------------
SOLARIS 10 is the OS for Data Centers - provides features such as DTrace, Predictive Self Healing and Award Winning ZFS. Get Solaris 10 NOW
http://p.sf.net/sfu/solaris-dev2dev
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

#2 lore raw
Quoted 34 lines · Serge E. Hallyn
Quoted 33 lines · Mitani

Hi,

I tried to test with "2010-02-11" cvs in RHEL5.4 system. But the test brings termination of connection.

I examined the phenomenon and discovered that "pidns17" test made accident and sshd stopped after the test.

I had some problems.

1. I think that "tst_exit()" must be added last of "cleanup()" function.
============
--- ./testcases/kernel/containers/pidns/pidns17.c       2009-12-07 05:55:16.000000000 +0900
+++ ./testcases/kernel/containers/pidns/pidns17.c.new   2010-02-12 16:26:31.000000000 +0900

(quoted diff) +0 -2
@@ -104,7 +104,6 @@
 
         /* cleanup and exit */
         CLEANUP();
-       tst_exit();
  }
 
  /***********************************************************************
@@ -136,7 +135,6 @@
 
         /* cleanup and exit */
         CLEANUP();
-       tst_exit();
  }      /* End main */
 
Quoted 10 lines

Review line 149

Quoted 7 lines
(quoted diff) +1 -0
  /*
@@ -147,4 +145,5 @@
  {
         /* Clean the test testcase as LTP wants*/
         TEST_CLEANUP;
+       tst_exit();
  }

Yeah I'm afraid I don't understand what CLEANUP and tst_exit exactly do.  Hopefully Garrett can give an educated answer.

Mitani's correct -- this is what should be done... CLEANUP is a constant that maps to cleanup in the event that tst_brkm is called, because linux_syscall_numbers.h's copy of syscall calls tst_brkm internally if ENOSYS is returned...

Quoted 14 lines
Quoted 14 lines

============

After revision, connection termination didn't occur.

2. I cannot understand the purpose that "kill()" function whose first parameter is "-1" is called.
------------
        if (kill(-1, SIGUSR1) == -1) {
                tst_resm(TBROK | TERRNO, "cinit: kill(-1, SIGUSR1) failed"); CLEANUP();
        }
------------

Quoted 5 lines
Quoted 2 lines

If kill()'s first option is "-1", "man kill" says that "All processes with pid larger than 1 will be signaled."

Right, the test is checking whether kill -1 inside a private pidns kills all processes besides init in the pid namespace.

Yeah, that's just not smart...

Quoted 6 lines
Quoted 2 lines

Therefore, not only the "sshd" but also the other processes were affected, I think.

sshd is not in the private pid namespace and should not be killed. If it is being killed by the pid -1 inside the container, then there is a kernel bug.

No, it isn't. If the test is being run as root it'll force a reboot on the box:

If pid is -1:
If the user has super-user privileges, the signal is sent to all processes excluding system processes (with P_SYSTEM flag set), process with ID 1 (usually init(8)), and the process sending the signal. If the user is not the super user, the signal is sent to all processes with the same uid as the user excluding the process sending the signal. No error is returned if any process could be signaled.

Quoted 1 line
Quoted 1 line

I'm glad if I could get some opinions.

I'll look at fixing this testcase once the cvs tree has been fully
converted to git by Rishikesh, as the points you've made are important and valid...
Thanks :),
-Garrett

------------------------------------------------------------------------------
SOLARIS 10 is the OS for Data Centers - provides features such as DTrace, Predictive Self Healing and Award Winning ZFS. Get Solaris 10 NOW
http://p.sf.net/sfu/solaris-dev2dev
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

#3 lore raw
Quoted 82 lines · Garrett Cooper
Quoted 34 lines · Serge E. Hallyn
Quoted 33 lines · Mitani

Hi,

I tried to test with "2010-02-11" cvs in RHEL5.4 system. But the test brings termination of connection.

I examined the phenomenon and discovered that "pidns17" test made accident and sshd stopped after the test.

I had some problems.

1. I think that "tst_exit()" must be added last of "cleanup()" function.
============
--- ./testcases/kernel/containers/pidns/pidns17.c       2009-12-07 05:55:16.000000000 +0900
+++ ./testcases/kernel/containers/pidns/pidns17.c.new   2010-02-12 16:26:31.000000000 +0900

(quoted diff) +0 -2
@@ -104,7 +104,6 @@
 
         /* cleanup and exit */
         CLEANUP();
-       tst_exit();
  }
 
  /***********************************************************************
@@ -136,7 +135,6 @@
 
         /* cleanup and exit */
         CLEANUP();
-       tst_exit();
  }      /* End main */
 
Quoted 10 lines

Review line 149

Quoted 7 lines
(quoted diff) +1 -0
  /*
@@ -147,4 +145,5 @@
  {
         /* Clean the test testcase as LTP wants*/
         TEST_CLEANUP;
+       tst_exit();
  }

Yeah I'm afraid I don't understand what CLEANUP and tst_exit exactly do.  Hopefully Garrett can give an educated answer.

Mitani's correct -- this is what should be done... CLEANUP is a constant that maps to cleanup in the event that tst_brkm is called, because linux_syscall_numbers.h's copy of syscall calls tst_brkm internally if ENOSYS is returned...

Quoted 14 lines
Quoted 14 lines

============

After revision, connection termination didn't occur.

2. I cannot understand the purpose that "kill()" function whose first parameter is "-1" is called.
------------
        if (kill(-1, SIGUSR1) == -1) {
                tst_resm(TBROK | TERRNO, "cinit: kill(-1, SIGUSR1) failed"); CLEANUP();
        }
------------

Quoted 5 lines
Quoted 2 lines

If kill()'s first option is "-1", "man kill" says that "All processes with pid larger than 1 will be signaled."

Right, the test is checking whether kill -1 inside a private pidns kills all processes besides init in the pid namespace.

Yeah, that's just not smart...

Quoted 6 lines
Quoted 2 lines

Therefore, not only the "sshd" but also the other processes were affected, I think.

sshd is not in the private pid namespace and should not be killed. If it is being killed by the pid -1 inside the container, then there is a kernel bug.

No, it isn't. If the test is being run as root it'll force a reboot on the box:

Quoted 8 lines

    If pid is -1:
            If the user has super-user privileges, the signal is sent to all processes excluding system processes (with P_SYSTEM flag set), process with ID 1 (usually init(8)), and the process sending the signal.  If the user is not the super user, the signal is sent to all processes with the same uid as the user excluding the process sending the signal.  No error is returned if any process could be signaled.

Oh wait.. containers isolate PIDs and resources, correct (a weak form of BSD jails or Solaris zones)? If so, then I'd watch the console // /var/log/messages, etc and see whether or not things stay alive after the signal is tossed...

Quoted 5 lines
Quoted 1 line
Quoted 1 line

I'm glad if I could get some opinions.

   I'll look at fixing this testcase once the cvs tree has been fully converted to git by Rishikesh, as the points you've made are important and valid...

Thanks,
-Garrett

------------------------------------------------------------------------------
SOLARIS 10 is the OS for Data Centers - provides features such as DTrace, Predictive Self Healing and Award Winning ZFS. Get Solaris 10 NOW
http://p.sf.net/sfu/solaris-dev2dev
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

#4 lore raw
Quoted 25 lines · Garrett Cooper
Quoted 15 lines · Garrett Cooper
Quoted 2 lines · Serge E. Hallyn

Right, the test is checking whether kill -1 inside a private pidns kills all processes besides init in the pid namespace.

Yeah, that's just not smart...

Quoted 6 lines
Quoted 2 lines

Therefore, not only the "sshd" but also the other processes were affected, I think.

sshd is not in the private pid namespace and should not be killed. If it is being killed by the pid -1 inside the container, then there is a kernel bug.

No, it isn't. If the test is being run as root it'll force a reboot on the box:

Quoted 8 lines

    If pid is -1:
            If the user has super-user privileges, the signal is sent to all processes excluding system processes (with P_SYSTEM flag set), process with ID 1 (usually init(8)), and the process sending the signal.  If the user is not the super user, the signal is sent to all processes with the same uid as the user excluding the process sending the signal.  No error is returned if any process could be signaled.

Quoted 1 line

Oh wait.. containers isolate PIDs and resources, correct (a weak form

weak? :)

Quoted 3 lines

of BSD jails or Solaris zones)? If so, then I'd watch the console // /var/log/messages, etc and see whether or not things stay alive after the signal is tossed...

Right. To approximate this testcase by hand I just tried on today's fresh kernel git pull

ns_exec -cp /bin/bash
for i in seq 1 100; do sleep 30 & done
kill -USR1 -1

and only the sleeps were killed.

thanks,
-serge

------------------------------------------------------------------------------
SOLARIS 10 is the OS for Data Centers - provides features such as DTrace, Predictive Self Healing and Award Winning ZFS. Get Solaris 10 NOW
http://p.sf.net/sfu/solaris-dev2dev
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

#5 lore raw
In reply to Garrett Cooper
Quoted 46 lines · Garrett Cooper
Quoted 34 lines · Serge E. Hallyn
Quoted 33 lines · Mitani

Hi,

I tried to test with "2010-02-11" cvs in RHEL5.4 system. But the test brings termination of connection.

I examined the phenomenon and discovered that "pidns17" test made accident and sshd stopped after the test.

I had some problems.

1. I think that "tst_exit()" must be added last of "cleanup()" function.
============
--- ./testcases/kernel/containers/pidns/pidns17.c       2009-12-07 05:55:16.000000000 +0900
+++ ./testcases/kernel/containers/pidns/pidns17.c.new   2010-02-12 16:26:31.000000000 +0900

(quoted diff) +0 -2
@@ -104,7 +104,6 @@
 
         /* cleanup and exit */
         CLEANUP();
-       tst_exit();
  }
 
  /***********************************************************************
@@ -136,7 +135,6 @@
 
         /* cleanup and exit */
         CLEANUP();
-       tst_exit();
  }      /* End main */
 
Quoted 10 lines

Review line 149

Quoted 7 lines
(quoted diff) +1 -0
  /*
@@ -147,4 +145,5 @@
  {
         /* Clean the test testcase as LTP wants*/
         TEST_CLEANUP;
+       tst_exit();
  }

Yeah I'm afraid I don't understand what CLEANUP and tst_exit exactly do.  Hopefully Garrett can give an educated answer.

Quoted 4 lines

Mitani's correct -- this is what should be done... CLEANUP is a constant that maps to cleanup in the event that tst_brkm is called, because linux_syscall_numbers.h's copy of syscall calls tst_brkm internally if ENOSYS is returned...

Cool, thanks

-serge

------------------------------------------------------------------------------
SOLARIS 10 is the OS for Data Centers - provides features such as DTrace, Predictive Self Healing and Award Winning ZFS. Get Solaris 10 NOW
http://p.sf.net/sfu/solaris-dev2dev
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list