Re: [LTP] "cgroup_fj" and "cgroup" have problems?

["Mitani" <mitani@ryobi.co.jp>]

Hi,

Thank you for fixing "cgroup" and "cgroup_fj".
However, I found a new problem in revised "cgroup" testcases.

It's in "cgroup_regression_test.sh".

In my original patch, I didn't revise about "tst_kvercmp 2 6 29" 
judgment.
But in fixed version, the error route of above judgment revised 
from "TBROK" to "TCONF".
If "TCONF" is used, the shell should end by exit 0, I think.
But "cgroup_regression_test.sh" does not end by exit 0.

Please consider about the following patch:
============
--- cgroup_regression_test.sh	2009-12-23 09:52:18.000000000 +0900
+++ cgroup_regression_test.sh.new	2009-12-24 15:47:06.000000000 +0900
@@ -31,7 +31,7 @@
 tst_kvercmp 2 6 29
 if [ $? -eq 0 ]; then
 	tst_brkm TCONF ignored "test must be run with kernel 2.6.29 or
newer"
-	exit 1
+	exit 0
 elif [ ! -f /proc/cgroups ]; then
 	tst_brkm TCONF ignored "Kernel does not support for control groups;
skipping testcases";
 	exit 0
============


Regards--


-Tomonori Miatani

-----Original Message-----
From: Garrett Cooper [mailto:yanegomi@gmail.com] 
Sent: Wednesday, December 23, 2009 9:53 AM
To: Mitani
Cc: ltp-list@lists.sourceforge.net
Subject: Re: [LTP] "cgroup_fj" and "cgroup" have problems?

On Mon, Dec 14, 2009 at 12:08 AM, Mitani <mitani@ryobi.co.jp> wrote:
> I took your indication and changed the patch as follows.
>
> Case 1:
> I revised "[ -f /proc/cgroups ]" case to use "tst_resm TCONF"
> instead of "echo".
> But this revision is unbalanced with the other exit messages,
> so I revised other cases too.
> I don't know how to use exit code, so I use 0 or 1.
> I might mistake those code.
>
> ============
> --- run_cgroup_test_fj.sh       2009-10-26 23:49:23.000000000 +0900
> +++ run_cgroup_test_fj.sh.new   2009-12-14 14:21:53.000000000 +0900
> @@ -34,19 +34,19 @@
>  export CASENO2=0
>
>  if [ "$(id -ru)" -ne 0 ]; then
> -       echo "***error***:you must use root to test"
> -       exit -1
> +       tst_brkm TBROK ignored "***error***:you must use root to test"
> +       exit 1
>  fi
>
>  if ! [ -f /proc/cgroups ]; then
> -       echo "***error***:you must enable cgroup config in kernel";
> -       exit -3;
> +       tst_resm TCONF "***error***:you must enable cgroup config in
> kernel";
> +       exit 0;
>  fi
>
>  if [ ! -x $TESTROOT/cgroup_fj_proc ]; then
> -       echo "***warning***:file \"cgroup_fj_proc\" is not exist or not
> executable";
> -       echo "please do make and check execution permission"
> -       exit -4;
> +       tst_brkm TBROK ignored "***warning***:file \"cgroup_fj_proc\" is
not
> exist or not executable";
> +       tst_brkm TBROK ignored "please do make and check execution
> permission"
> +       exit 1;
>  fi
>
>  CPUSET=`grep -w cpuset /proc/cgroups | cut -f1`;
> @@ -83,10 +83,10 @@
>         SYBSYSCOMPILED="$SYBSYSCOMPILED devices,"
>  fi
>  if [ "$SYBSYSCOMPILED" = "" ];then     # Warning and exit if all cgroup
> subsystem are not compiled
> -       echo "CONTROLLERS TESTCASES: WARNING";
> -       echo "Kernel does not support for any cgroup subsystem";
> -       echo "Skipping all controllers testcases....";
> -       exit -2;
> +       tst_resm TCONF "CONTROLLERS TESTCASES: WARNING";
> +       tst_resm TCONF "Kernel does not support for any cgroup subsystem";
> +       tst_resm TCONF "Skipping all controllers testcases....";
> +       exit 0;
>  fi
>
>  echo "Now, we start the test for cgroup...";
> ============
>
>
> Case 2:
> I revised "[ "$USER" != root ]" case, and revised
> "[ ! -f /proc/cgroups ]" case to use "tst_resm" instead of "echocase"
>  for balance.
>
> ============
> --- cgroup_regression_test.sh   2009-11-30 17:41:37.000000000 +0900
> +++ cgroup_regression_test.sh.new       2009-12-14 11:42:58.000000000
+0900
> @@ -29,15 +29,15 @@
>  export TST_COUNT=1
>
>  if [ ! -f /proc/cgroups ]; then
> -        echo "CONTROLLERS TESTCASES: WARNING"
> -        echo "Kernel does not support for control groups";
> -        echo "Skipping all controllers testcases....";
> +       tst_resm TCONF "CONTROLLERS TESTCASES: WARNING"
> +       tst_resm TCONF "Kernel does not support for control groups"
> +       tst_resm TCONF "Skipping all controllers testcases...."
>        exit 0
>  fi
>
>  if [ "$USER" != root ]; then
>        tst_brkm TBROK ignored "Test must be run as root"
> -       exit 0
> +       exit 1
>  fi
>
>  tst_kvercmp 2 6 29
> ============
>
>
> Regards--
>
>
> -Tomonori Mitani
>
>
> -----Original Message-----
> From: Garrett Cooper [mailto:yanegomi@gmail.com]
> Sent: Friday, December 11, 2009 6:45 PM
> To: Mitani
> Cc: ltp-list@lists.sourceforge.net
> Subject: Re: [LTP] "cgroup_fj" and "cgroup" have problems?
>
> On Dec 11, 2009, at 1:21 AM, Mitani wrote:
>
>> Hi,
>>
>> Sorry for my many many mails today.
>>
>>
>> I ran tests with "ltp-2009-12-10" cvs.
>>
>> In my system (kernel 2.6.18-164.el5), kernel does not support
>> control groups.
>> So, when I ran test of "cgroup_fj", I got following message and the
>> test faild by "FAIL 253".
>>
>> ------------
>> ***error***:you must enable cgroup config in kernel
>> ------------
>>
>> In the investigation for above error message, I found that there may be
>> two problems,.
>>
>>  (a) "cgroup_fj" test failed in the system not support "control groups".
>>  (b) "cgroup" test may passed by user-ID which isn't "root".
>>
>>
>> (a) "cgroup_fj" test
>>  Above message is correct message, but exiting by "FAIL" is not
>>  correct, I think.
>>  Referring to the example of other tests, I think that the tests
>>  for non-supported functions must be finished in "PASS 0" basically.
>>  (For example, "move_pages{01...11}" tests exit in "PASS 0" and with
>>  "NUMA support not provided" messages.)
>>
>>  I suggest following patch to resolve this problem.
>>
>> ============
>> --- run_cgroup_test_fj.sh     2009-10-26 23:49:23.000000000 +0900
>> +++ run_cgroup_test_fj.sh.new 2009-12-11 16:10:39.000000000 +0900
>> @@ -40,7 +40,7 @@
>>
>> if ! [ -f /proc/cgroups ]; then
>>       echo "***error***:you must enable cgroup config in kernel";
>> -     exit -3;
>> +     exit 0;
>> fi
>>
>> if [ ! -x $TESTROOT/cgroup_fj_proc ]; then
>>
>> ============
>>
>>
>> (b) "cgroup" test
>>  I watched "cgroup" test source for hints for (a), and found another
>>  problem.
>>  If this test run by user-ID non-"root", it will exit in "PASS", I think.
>>  But I think that it must exit in "FAIL".
>>  Because this test cannot be tested only by "root" user.
>>
>> =========
>> --- cgroup_regression_test.sh 2009-11-30 17:41:37.000000000 +0900
>> +++ cgroup_regression_test.sh.new     2009-12-11 16:22:14.000000000 +0900
>> @@ -37,7 +37,7 @@
>>
>> if [ "$USER" != root ]; then
>>       tst_brkm TBROK ignored "Test must be run as root"
>> -     exit 0
>> +     exit -1
>> fi
>>
>> tst_kvercmp 2 6 29
>
> 1. Case 1 should be tst_brkm TCONF and should have exit 0 (like you
> suggested).
> 2. Case 2 should have the equivalent exit code for TBROK (it's in
> include/test.h). $? -> -1 is undefined via POSIX and while it passes on
> bash, it doesn't pass on ash, and thus would fail with busybox.
> Thanks,
> -Garrett

Committed -- thanks!
-Garrett



------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev 
_______________________________________________
Ltp-list mailing list
Ltp-list@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/ltp-list