Changelog v2:
* Use clock_gettime(2) for getting timestamp since a system boot.
* Change parent_info's memory to dynamic allocation.
This patch is for changing the cgroup of a forked process while parent
changing.
This patch adds the following sequence:
1. Store both the timestamp and the process-id when changing the cgroup.
2. If receiving a PROC_EVENT_FORK packet, check its parent-pid and its
timestamp.
3. If its parent-pid and the stored process-id are same and its timestamp
is older than the stored timestamp, change the cgroup of forked process.
I have been testing a cgrulesengd daemon and I noticed it fails to
change the cgroup of child occasionally. I tested it by following
configulation file:
A cpuset subsystem and a memory subsystem are mounted on different
mount points, and a cgrulesengd daemon manages each subsystem.
I login this environment as a user "user01", and each susbystem's
tasks file is the following:
They should be the same, but they are different. I investigated this
problem, and I found the cause. The reason is that the process(31806)
was forked just after writing the process(31805) to a cpuset subsystem's
tasks file:
<1> The UID/GID CHANGE event of the process 31805 happens.
<2> The daemon writes "31805" to a cpuset subsystem's tasks file.
<3> The process 31806 is forked, and it appears on a cpuset subsystem's
tasks file.
<4> The daemon writes "31805" to a memory subsystem's tasks file.
<5> The process 31806 does not appears on a memory subsystem's tasks file.
For solving this problem, I propose the following sequence.
1. Store both the timestamp and the process-id when the step <4>.
2. If receiving a PROC_EVENT_FORK packet, check its parent-pid and its
timestamp.
3. If its parent-pid and the stored process-id are same and its timestamp
is older than the stored timestamp, change the cgroup of forked process.
Changelog of v2:
* Change only [PATCH 2/2] and there is not any changes in [PATCH 1/2].
This patch adds the method for getting euid/egid from /proc/<pid>/status
file.
For changing the cgroup of a forked process, the method is usefull because
a PROC_EVENT_FORK packet does not inform of its euid and its egid.
This is the test case for the new API. This test takes one argument, the
group name.
Sample run on my system with this test case returns,
[dhaval@gondor tests]$ ../libtool --mode=execute ./walk_task a
Printing the details of groups a
Pid is 6092
Pid is 11315
Pid is 11318
Pid is 11319
Pid is 11324
Pid is 13234
[dhaval@gondor tests]$
gcc -DHAVE_CONFIG_H -I. -I.. -I../include -g -O2 -Wall -MT api.lo -MD -MP -MF .deps/api.Tpo -c api.c -fPIC -DPIC -o .libs/api.o
api.c:52:1: warning: "VERSION" redefined
In file included from ./libcgroup-internal.h:21,
from api.c:31:
../config.h:129:1: warning: this is the location of the previous definition
api.c: In function 'cgroup_parse_rules':
api.c:353: warning: implicit declaration of function 'isblank'
api.c: In function 'cgroup_modify_cgroup':
api.c:1073: warning: implicit declaration of function 'asprintf'
This patch fixes the warning arising due to isblank, asprintf and getline.
Signed-off-by: Bharata B Rao <bharata@linux.vnet.ibm.com> Signed-off-by: Dhaval Giani <dhaval@linux.vnet.ibm.com>
Remove the dependency of libcgroup on libcgroup build.
libcgroup build checks for the availability of cgroup_creat_cgroup
from libcgroup during its build. This causes a build failure if
libcgroup is compiled in a system which already has libcgroup library
installed.
Remove this check from configure.in.
Signed-off-by: Bharata B Rao <bharata@linux.vnet.ibm.com> Signed-off-by: Dhaval Giani <dhaval@linux.vnet.ibm.com>
Two pthread_rwlock_unlock() calls are necessary if *cgroup is null
and fprintf()/fflush() fails in cgroup_attach_task_pid():
src/api.c:785
785 pthread_rwlock_rdlock(&cg_mount_table_lock);
786 for(i = 0; i < CG_CONTROLLER_MAX &&
787 cg_mount_table[i].name[0]!='\0'; i++) {
[snip]
805 ret = fprintf(tasks, "%d", tid);
806 if (ret < 0) {
<<pthread_rwlock_unlock() call is necessary>>
807 cgroup_dbg("Error writing tid %d to
%s:%s\n",
808 tid, path,
strerror(errno));
809 fclose(tasks);
810 last_errno = errno;
811 return ECGOTHER;
812 }
813
814 ret = fflush(tasks);
815 if (ret) {
<<pthread_rwlock_unlock() call is necessary>>
816 last_errno = errno;
817 cgroup_dbg("Error writing tid %d to
%s:%s\n",
818 tid, path,
strerror(errno));
819 fclose(tasks);
820 return ECGOTHER;
821 }
For the readability, this patch merges almost the same lines into one
function(__cgroup_attach_task_pid()) and adds pthread_rwlock_unlock()
call for the case the function fails.
Fix infinite loop if receiving a NLMSG_NOOP packet.
Hi,
I tested 'cgred' service and I saw the problem that some processes
are not moved to a right cgroup. This problem did not occur always,
and it did sometimes. I reviewed cgrulesengd.c and found the bug
cgrulesengd stays in an infinite loop if receiving a NLMSG_NOOP
packet.
Add some success/error messages to 'cgred' service.
Hi,
The existing 'cgred' service displays a success message ([ OK ]) when
the service is starting, but it does not display any messages when it
is stopping/reloading. So this patch adds some success/error messages
to 'cgred' service like the following:
# service cgred start
Starting CGroup Rules Engine Daemon...
[ OK ]
#
# service cgred stop
Stopping CGroup Rules Engine Daemon...
[ OK ]
#
# service cgred restart
Stopping CGroup Rules Engine Daemon...
[ OK ]
Starting CGroup Rules Engine Daemon...
[ OK ]
#
# service cgred reload
Reloading rules configuration...
[ OK ]
#
This patch's purpose is almost same as the one of 'cgconfig' service:
http://sourceforge.net/mailarchive/forum.php?thread_name=49D9B778.3000900%40mxs.nes.nec.co.jp&forum_name=libcg-devel
Add some success/error messages to 'cgconfig' service.
Hi,
The existing 'cgconfig' service does not display any message even if it
succeeds. So this patch adds some success/error messages to 'cgconfig'
service like the following:
# service cgconfig start
Starting cgconfig service: [ OK ]
#
# service cgconfig stop
Stopping cgconfig service: [ OK ]
#
# service cgconfig restart
Stopping cgconfig service: [ OK ]
Starting cgconfig service: [ OK ]
#
It makes test of 'cgconfig' service a little easy.
Here is the v2 of the patch which introduces APIs to read controller
specific stats. Added Makefile.am updates which I had missed in the
previous post.
New APIs to read controller specific statistics.
This patch introduces 3 new APIs which can be used to read controller
statistics iteratively. (Eg. stats from memory.stat etc)
Reading of stats is initiated by cgroup_read_stats_begin() API, which
returns the first stat of the requested controller in addition to returing
a handle that should be used in subsequent reads.
cgroup_read_stats_next() API can be used to read the remaining stats
one by one. This needs the handle returned by cgroup_read_stats_begin().
cgroup_read_stats_end() API will terminate the stats reading iteration
initiated by cgroup_read_stats_begin().
Changelog:
v2
- Update tests/Makefile.am so that it generates appropriate rules
for tests/read_stats.c in the Makefile. This is in addition to
the manual updates done to the generated file tests/Makefile.in.
v1
- cgroup_read_stats apis now work with relative cgroup path names instead
of absolute path names.
v0
- Initial post.
Signed-off-by: Bharata B Rao <bharata@linux.vnet.ibm.com> Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Ken'ichi Ohmichi [Tue, 31 Mar 2009 02:18:27 +0000 (11:18 +0900)]
Add fclose() calls for error handling
I reviewed the libcgroup code in the viewpoint of memory-leak and
found some lacks of fclose(). This patch adds some fclose() calls
for error handling.
Jan Safranek [Mon, 30 Mar 2009 10:24:42 +0000 (15:54 +0530)]
No matter how often I test, some last-minute change screws everything up...
This time I added duplicate -version-number argument in src/Makefile.am, which
results in following compilation error:
libtool: link: CURRENT `' must be a nonnegative integer
libtool: link: `-version-number' is not valid version information
Jan Safranek [Thu, 26 Mar 2009 14:10:09 +0000 (15:10 +0100)]
Generate .tar.bz2 on 'make dist'
Automatically generate libcgroup*.tar.bz2 on 'make dist'. Some restructure
of AC_INIT and AM_INIT was needed to archieve this and correct output
filename.
Namely AC_INIT( [package], ... ) is used as resulting tarball file name now.
Also run automake and autoconf to refresh the generated stuff... Only
configure.in contains manual changes, everything else is autoconf/automake
bloat :(.
Jan Safranek [Thu, 26 Mar 2009 14:04:36 +0000 (15:04 +0100)]
Add warrning when test script is started by normal user
The runlibcgrouptest.sh will now whine when script is started by non-root
user. Root permissions are needed to create directories and mount the cgroups
there.
Jan Safranek [Wed, 25 Mar 2009 09:28:34 +0000 (10:28 +0100)]
Disable compilation of static libraries
Disable compilation of static libraries. Static libraries (.a files) should not
be compiled by default, shared library is better for LGPL licensed code.
Still, one can enable static libraries with ./configure --enable-static
Jan Safranek [Fri, 13 Mar 2009 14:16:20 +0000 (15:16 +0100)]
Add automake makefiles
Add automake makefiles and prepare everything to run, but don't actually
run the automake - it would generate lot of noise, where my manual changes
would get lots.
Jan Safranek [Fri, 13 Mar 2009 14:16:20 +0000 (15:16 +0100)]
Clean up the scripts directory
The directory contains samples and howto, which is not relevant anymore.
Sample configuration files should be in ../samples and (relevant)
documentation in ../doc
Ivana Varekova [Wed, 25 Mar 2009 07:55:57 +0000 (08:55 +0100)]
Minor fix to cgconfig init script
cgconfig script does not test the return value of cgconfigparser command
so it continues to simulate start of the service, even if it was not pass
successfuly.
This is the first part of the patch from:
http://sourceforge.net/mailarchive/forum.php?thread_name=20090120121851.GC4422%40linux.vnet.ibm.com&forum_name=libcg-devel
which was not incorporated yet.
Signed-off-by: Ivana Varekova <varekova@redhat.com> Signed-off-by: Dhaval Giani <dhaval@linux.vnet.ibm.com>
Jan Safranek [Mon, 2 Mar 2009 12:09:16 +0000 (13:09 +0100)]
Add PACKAGE_VERSION to Makefile
By removing PACKAGE_VERSION in SVN rev. 344 I broke 'make install' - the macro
must be defined there, 'make install' needs it to build library symlinks.
I am sorry, I should have noticed it earlier.
Signed-off-by: Jan Safranek <jsafrane@redhat.com> Signed-off-by: Dhaval Giani <dhaval@linux.vnet.ibm.com>
Jan Safranek [Sun, 8 Mar 2009 21:43:59 +0000 (22:43 +0100)]
Move binaries and libraries to /bin and /lib
Since cgconfig starts before nfs shares are mounted, it should not use
anything on /usr (it can be on nfs too). All the libcgroup libraries and
binaries should be therefore in /.
Signed-off-by: Jan Safranek <jsafrane@redhat.com> Signed-off-by: Dhaval Giani <dhaval@linux.vnet.ibm.com>
Jan Safranek [Sun, 8 Mar 2009 21:43:51 +0000 (22:43 +0100)]
Start the config earlier
We could start cgconfig service earlier, even before network and syslog.
$local_fs is not needed, because /bin, /sbin and /etc are guaranteed by
rc.sysinit (see Fedora bug #309001,
https://bugzilla.redhat.com/show_bug.cgi?id=309001).
Signed-off-by: Jan Safranek <jsafrane@redhat.com> Signed-off-by: Dhaval Giani <dhaval@linux.vnet.ibm.com>
Dhaval Giani [Wed, 4 Mar 2009 11:38:53 +0000 (17:08 +0530)]
libcgroup: Initialize buffer in config.c
Fix the warning,
config.c: In function ‘cgroup_config_parse_controller_options’:
config.c:104: warning: ‘buffer’ may be used uninitialized in this function
Dhaval Giani [Wed, 4 Mar 2009 11:29:49 +0000 (16:59 +0530)]
libcgroup: Removed unused functions from api.c
api.c: At top level:
api.c:1780: warning: ‘cg_prepare_controller_array’ defined but not used
api.c:1805: warning: ‘cg_free_controller_array’ defined but not used
Turns out no one is using this functions. Remove them.
Dhaval Giani [Wed, 4 Mar 2009 11:25:35 +0000 (16:55 +0530)]
libcgroup: Fix unused variable warnings for the walk_tree API
api.c: In function ‘cg_walk_node’:
api.c:2252: warning: unused variable ‘base_level’
api.c: In function ‘cgroup_walk_tree_end’:
api.c:2314: warning: unused variable ‘ret’
Dhaval Giani [Tue, 3 Mar 2009 14:51:03 +0000 (20:21 +0530)]
libcgroup: Check if the data structures have been initialized for walk_tree
The test for cgroup_init() was not being done by the walk tree APIs.
With this it is possible for it to run without cgroup_init() being
called and having mysterious failures.
Dhaval Giani [Tue, 3 Mar 2009 14:51:02 +0000 (20:21 +0530)]
libcgroup: Dont check for success of moved tasks
Sometime the movement of tasks in the default cgroup can fail for
various reasons. Therefore, do not check for the success of the
movement of tasks.
Justification: These are generally for cgroups which need some special
setup, for example the RT group scheduler. This is not handled by the
library whihc is generic. If a user is intersted in these working, they
will setup a default cgroup in the config files whihc handle everything
they need.
Dhaval Giani [Sat, 28 Feb 2009 09:27:31 +0000 (09:27 +0000)]
Feature: Provide new libcgroup walk tree API
From: Balbir Singh <balbir@linux.vnet.ibm.com>
Changelog v3..v4
1. Add _end() method for the iterator
Changelog v3..v2
1. Move to iterator based design
Changelog v2..v1
1. Add base path and depth semantics for walking
This patch adds the capability to walk cgroups by providing a new API
called cgroup_walk_tree. The API accepts the controller to walk and the
order in which the directories and files must be visited. The code is
implemented as an iterator, the begin function starts the walk and
we have depth control. The next function gets the following node
and returns ECGEOF when done.
libcgroup.map has been updated to reflect the same change and the prototype
is exported in libcgroup.h.
I've also added test cases (tests/walk_test.c). Sample output is show
root is /cgroup/cpu///
path , parent , relative /, full /cgroup/cpu///
path l3, parent , relative /l3, full /cgroup/cpu///l3
path ll1, parent l3, relative /l3/ll1, full /cgroup/cpu///l3/ll1
path lll1, parent ll1, relative /l3/ll1/lll1, full /cgroup/cpu///l3/ll1/lll1
path l2, parent , relative /l2, full /cgroup/cpu///l2
path ll1, parent l2, relative /l2/ll1, full /cgroup/cpu///l2/ll1
path lll1, parent ll1, relative /l2/ll1/lll1, full /cgroup/cpu///l2/ll1/lll1
path l1, parent , relative /l1, full /cgroup/cpu///l1
path ll1, parent l1, relative /l1/ll1, full /cgroup/cpu///l1/ll1
path lll1, parent ll1, relative /l1/ll1/lll1, full /cgroup/cpu///l1/ll1/lll1
path a, parent , relative /a, full /cgroup/cpu///a
path e, parent a, relative /a/e, full /cgroup/cpu///a/e
path f, parent e, relative /a/e/f, full /cgroup/cpu///a/e/f
path f, parent a, relative /a/f, full /cgroup/cpu///a/f
path x, parent a, relative /a/x, full /cgroup/cpu///a/x
path b, parent a, relative /a/b, full /cgroup/cpu///a/b
path c, parent b, relative /a/b/c, full /cgroup/cpu///a/b/c
path d, parent c, relative /a/b/c/d, full /cgroup/cpu///a/b/c/d
path default, parent , relative /default, full /cgroup/cpu///default
root is /cgroup/cpu//a/
path , parent , relative /, full /cgroup/cpu//a/
path e, parent , relative /e, full /cgroup/cpu//a/e
path f, parent e, relative /e/f, full /cgroup/cpu//a/e/f
path f, parent , relative /f, full /cgroup/cpu//a/f
path x, parent , relative /x, full /cgroup/cpu//a/x
path b, parent , relative /b, full /cgroup/cpu//a/b
path c, parent b, relative /b/c, full /cgroup/cpu//a/b/c
Walking the first 5 nodes
root is /cgroup/cpu///
path , parent , relative /, full /cgroup/cpu///
path l3, parent , relative /l3, full /cgroup/cpu///l3
path ll1, parent l3, relative /l3/ll1, full /cgroup/cpu///l3/ll1
path lll1, parent ll1, relative /l3/ll1/lll1, full /cgroup/cpu///l3/ll1/lll1
NOTE: Parent directory is represented by an empty (not NULL) string "".
The length of the string is 0.
Dhaval Giani [Fri, 27 Feb 2009 20:38:54 +0000 (20:38 +0000)]
libcgroup: Update specfile for v0.33 and bump up the version
From: Ivana Varekova <varekova@redhat.com>
OK, here goes
This patch Originally from Ivana and with changes I've made, get the
spec file working again (rpmlint warnings need some care and
ownership).
The new version will be 0.33 and will have new man pages and newer
API and many more bug fixes
[varekova@redhat.com] Incorporate pam module instalation to the spec file -
file part change add dependency
[balbir@linux.vnet.ibm.com] Bumped up the version, getting ready for a new
release. Updated makefile to install new man pages and to move the pam
module to the correct place at the time of installation.
Dhaval Giani [Thu, 26 Feb 2009 11:53:08 +0000 (11:53 +0000)]
bcgroup: Rename 'flash' to 'reload'
LSB specifies that 'reload' is the right initscript parameter to stimulate
daemon to reread its configuration file. I let the 'flash' still working, in
case somebody is used to it, but do not advertise it anywhere.
Dhaval Giani [Wed, 25 Feb 2009 13:34:30 +0000 (13:34 +0000)]
libcgroup: Remove PACKAGE_VERSION from Makefile.in
From: Jan Safranek <jsafrane@redhat.com>
PACKAGE_VERSION is defined both in config.h and Makefile.in, which results
in 'PACKAGE_VERSION redefined' warnings. It should be defined only once,
let's say in config.h.
Dhaval Giani [Wed, 25 Feb 2009 13:31:29 +0000 (13:31 +0000)]
libcgroup: Move dbg() to libcgroup-internal.h
From: Jan Safranek <jsafrane@redhat.com>
I think dbg() should be in internal header file and not in the public one.
The tools get their own definition of dbg() so they don't need to include
library private header.
dbg() is defined on three places now, it will make more sense when the sources
are distributed to different directories.
Dhaval Giani [Wed, 25 Feb 2009 13:05:56 +0000 (13:05 +0000)]
cgclassify: Add -g option to cgclassify
From: Jan Safranek <jsafrane@redhat.com>
cgclassify tools currently moves processes to groups only as specified in
cgrules.conf. It would be nice to move processes to another group,
specified on command line, like cgexec does.
Balbir Singh [Mon, 23 Feb 2009 14:41:10 +0000 (14:41 +0000)]
From: Ivana Varekova <varekova@redhat.com>
This patch fixes spec file handling of daemon parts of libcg
(add chkconfig to post and including dependency, fix preun script).
And fix the levels in which is this daemon start by default (should not be any).
Balbir Singh [Mon, 23 Feb 2009 14:10:58 +0000 (14:10 +0000)]
From: Ivana Varekova <varekova@redhat.com>
This patch changes cgconfig.conf - comment all rows and rename
example values - to have example which should be installed to /etc/.
Move cgred to /etc/sysconfig and patch spec to install all
/etc/* configuration files (at least they shows examples and
help to user to do their own configuration).).
Signed-off-by: Ivana Varekova <varekova@redhat.com> Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
NOTE: The spec file is still broken due to lack of packaging of pam_cgroup.so.
Balbir Singh [Sat, 21 Feb 2009 15:34:24 +0000 (15:34 +0000)]
Impact: Fix the parsing issue when two or more values are specified
This patch fixes an issue where when two or more values are specified for
the controllers, only the last one is applied. This patch fixes that issue.
The parser is modified to collate all the name value pairs, seperated by
":". The reason for implementing it this way is because we need to pass
the controller name and splitting the rules would make it very hard to pass
the controller name to each rule, irrespective of where the controller name
was parsed.
Dhaval Giani [Sat, 21 Feb 2009 07:28:19 +0000 (07:28 +0000)]
libcgorup: Fix a chown security issue
From: Balbir Singh <balbir@linux.vnet.ibm.com>
Impact: Bug fix causes incorrect chown
This patch fixes a potential security issue, we free path and add
reallocate it using asprintf, but that breaks chown, since that relies on
fts_path[0] and path to point to the same address location.
Dhaval Giani [Sat, 21 Feb 2009 07:22:19 +0000 (07:22 +0000)]
libcgroup: Fix return values
From: Balbir Singh <balbir@linux.vnet.ibm.com>
Impact: Bug fix, fix incorrect return values
This patch fixes incorrect return value of -1 being returned from
some call sites. Please review, comment. printf's are ugly, so they've
been converted to dbg.
Dhaval Giani [Mon, 16 Feb 2009 08:41:19 +0000 (08:41 +0000)]
libcgroup: Compile cgrulesengd without -DDEBUG
From: Jan Safranek <jsafrane@redhat.com>
Cgrulesengd writes some debugging info to stdout by default. IMHO
this feature should be enabled only when user explicitly asks so
during compilation, like the other tools.
Dhaval Giani [Thu, 12 Feb 2009 16:03:39 +0000 (16:03 +0000)]
libcgroup: Add command line options to set log level and log destinations.
From: Jan Safranek <jsafrane@redhat.com>
I rewrote the parsing to getopt - it's IMHO less error-prone and
new options can be added quite easily. Each new option has short
and long variant, both shown in --help.
Dhaval Giani [Thu, 12 Feb 2009 16:02:22 +0000 (16:02 +0000)]
libcgroup: Make the group change logs more readable
From: Jan Safranek <jsafrane@redhat.com>
Function cgre_process_event() logs success/failure of group change two
times, each with different log severity. It should be logged just once.
[dhaval@linux.vnet.ibm.com: Minor changes to the changelog] Signed-off-by: Jan Safranek <jsafrane@redhat.com> Acked-by: Dhaval Giani <dhaval@linux.vnet.ibm.com>
git-svn-id: https://libcg.svn.sourceforge.net/svnroot/libcg/trunk@320 4f4bb910-9a46-0410-90c8-c897d4f1cd53
Dhaval Giani [Thu, 12 Feb 2009 16:01:07 +0000 (16:01 +0000)]
libcgroup: Send log message to a log file, to syslog or to both.
From: Jan Safranek <jsafrane@redhat.com>
Rework whole cgrulesengd logging to be able to send log messages to syslog.
Also introduce more log levels:
LOG_NOTICE is the default log level, it logs startup/shutdown/configuration
reload and errors.
LOG_INFO shows in addition content of configuration file (only
if logging to file is enabled) and changing cgroup of a PID.
LOG_DEBUG show all details as it was before the patch.
TODO: if something gets wrong with mounted hierarchies (e.g. root unmounts
them), the log gets full of "Cgroup change for PID: ... FAILED". Some
suppression should be implemented.
Balbir Singh [Fri, 6 Feb 2009 16:02:24 +0000 (16:02 +0000)]
I've noticed it's possible to set permissions and other parameters of root
control group, i.e. the group which is implicitly created on mount and contains
all processes.
Admin can use '.' as the group name and he/she can set group's parameters as
usual. I am not sure it's a bug or feature, I've just documented current
behavior on appropriate man page.
Balbir Singh [Tue, 3 Feb 2009 09:43:57 +0000 (09:43 +0000)]
Here is my proposal for a man page describing content of /etc/cgconfig.conf.
I did not find any realistic example of configuration file, I just thought
of something mounting groups to /mnt/cgroups. Is it suitable location of
such mount points? Or is /container mentioned on libcg web the right place
for such mounts?
Feel free to rephrase and reformat anything, especially RECOMMENDATIONS
section would appreciate some 'official' content - I wrote there just my
experiences. I'm not native speaker, please look for grammar errors too.
Signed-off-by: Jan Safranek <jsafrane@redhat.com> Reviewed-by: Ivana Varekova <varekova@redhat.com>
git-svn-id: https://libcg.svn.sourceforge.net/svnroot/libcg/trunk@315 4f4bb910-9a46-0410-90c8-c897d4f1cd53
Dhaval Giani [Wed, 14 Jan 2009 17:33:05 +0000 (17:33 +0000)]
libcgrouptest: rename get_controller() to is_subsystem_enabled()
From: Sudhir Kumar <skumar@linux.vnet.ibm.com>
This patch renames the function get_controllers() to a more
meaningful name is_subsystem_enabled().
history:
The code in the patch is part of a patch sent in an earlier patchset,
which somehow got dropped (looks an older version of that patch was
merged to the tree). Adding this patch makes everything upstream upto
the discussion held on the list.
Dhaval Giani [Fri, 9 Jan 2009 16:27:48 +0000 (16:27 +0000)]
libcgrouptest: let check_task() receive a pid as argument
From: Sudhir Kumar <skumar@linux.vnet.ibm.com>
The patch would get a reject because of modifications in an earlier
patch. Hence updating the patch by fixing the reject.
This patch makes the function check_task() capable of checking if a pid
other than the calling process is attached to a particlular group. Earlier
it was checking for the calling process only.
For optimization purpose 0 is used to indicate that the task to be attached
is the current task and hence call cgroup_gettid() only once.
Dhaval Giani [Fri, 9 Jan 2009 16:26:15 +0000 (16:26 +0000)]
libcgrouptest:attach exclusive pid support in test_cgroup_attach_task() function
From: Sudhir Kumar <skumar@linux.vnet.ibm.com>
The function test_attach_task_pid() was attaching the process calling the
function to a group and did not have support to take any pid for attaching
it to a group. This patch does that.
Dhaval Giani [Fri, 9 Jan 2009 16:23:01 +0000 (16:23 +0000)]
libcgrouptest: Move functions into a new file to reuse
From: Sudhir Kumar <skumar@linux.vnet.ibm.com>
There were some coding style warnings that have been fixed in this
patch. Basicaly the ones reported by checkpatch.pl
This patch puts the test functions in a separate file so that they
can be reused for further testing. Some of the functions were static
and hence changed to be non static. It also edits the Makefile so
that the patch compiles.
Dhaval Giani [Fri, 9 Jan 2009 10:19:02 +0000 (10:19 +0000)]
libcgrouptest: Avoid external variables usage
From: Sudhir Kumar <skumar@linux.vnet.ibm.com>
A lack of proper initial thought let me use some variables globaly, which
is not a good idea at all. However fixing this issue at this stage is not
very good from the point that it changes variables for a lot of functions.
But still it has to be done for a better and reusable code. This patch does
that while putting related variables in meaningful structs.
Thanks to Balbir for a good suggestion.
TODO:
Fill these data structures from a configuration file after analyzing
requirements from different test functions.
Dhaval Giani [Tue, 6 Jan 2009 08:47:57 +0000 (08:47 +0000)]
libcgroup: Fix low hanging cleanups
Some of the cleanups possible are obvious.
1. Change usage of strcat to strncat
2. Change usage of tge following type
char *s = malloc();
strcpy(s, "somestring");
strcat(s, "someotherstring");
to something more easily understandble such as
asprintf(&s, "%s%s", somestring, someotherstring);
Changes for v2:
1. Fix some memory leaks discovered using valgrind
2. Fix Balbir's comments regarding codingstyle.
3. Move the controllers array memory leak fix into another patch
Changes from v1:
1. Correct the error handling of asprintf as pointed out by Dan Smith
TODO:
1. Figure out what is the correct value of n for cg_build_path_locked
valgrind output,
[root@gondor tests]# sh runlibcgrouptest.sh
Running first set of testcases
==============================
==5067== Memcheck, a memory error detector.
==5067== Copyright (C) 2002-2007, and GNU GPL'd, by Julian Seward et al.
==5067== Using LibVEX rev 1804, a library for dynamic binary translation.
==5067== Copyright (C) 2004-2007, and GNU GPL'd, by OpenWorks LLP.
==5067== Using valgrind-3.3.0, a dynamic binary instrumentation framework.
==5067== Copyright (C) 2000-2007, and GNU GPL'd, by Julian Seward et al.
==5067== For more details, rerun with: -v
==5067==
C:DBG: fs_mounted as recieved from script=0
TEST 1:PASS : cgroup_init() Ret Value = 50001
TEST 2:PASS : cgroup_attach_task() Ret Value = 50014 Parameter nullcgroup
TEST 3:PASS : cgroup_new_cgroup() Ret Value = 0
TEST 4:PASS : cgroup_create_cgroup() Ret Value = 50014
TEST 5:PASS : cgroup_delete_cgroup() Ret Value = 50014
TEST 6:PASS : cgroup_create_cgroup() Ret Value = 50014
TEST 7:PASS : cgroup_delete_cgroup() Ret Value = 50014
TEST 8:PASS : cgroup_add_controller() Ret Value = 0
TEST 9:PASS : cgroup_add_controller() Ret Value = 0
TEST10:PASS : cgroup_add_controller() Ret Value = 0
==5067==
==5067== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 18 from 1)
==5067== malloc/free: in use at exit: 0 bytes in 0 blocks.
==5067== malloc/free: 18 allocs, 18 frees, 32,293 bytes allocated.
==5067== For counts of detected errors, rerun with: -v
==5067== All heap blocks were freed -- no leaks are possible.
Running second set of testcases
==============================
==5083== Memcheck, a memory error detector.
==5083== Copyright (C) 2002-2007, and GNU GPL'd, by Julian Seward et al.
==5083== Using LibVEX rev 1804, a library for dynamic binary translation.
==5083== Copyright (C) 2004-2007, and GNU GPL'd, by OpenWorks LLP.
==5083== Using valgrind-3.3.0, a dynamic binary instrumentation framework.
==5083== Copyright (C) 2000-2007, and GNU GPL'd, by Julian Seward et al.
==5083== For more details, rerun with: -v
==5083==
C:DBG: fs_mounted as recieved from script=1
C:DBG: mountpoint1 as recieved from script=/dev/cgroup_controllers-1
sanity check pass. cgroup
TEST 1:PASS : cgroup_attach_task() Ret Value = 50014 Parameter nullcgroup
TEST 2:PASS : cgroup_init() Ret Value = 0
TEST 3:PASS : cgroup_attach_task() Ret Value = 0 Task found in group/s
TEST 4:PASS : cgroup_attach_task_pid() Ret Value = 50016
TEST 5:PASS : cgroup_new_cgroup() Ret Value = 0
TEST 6:PASS : cgroup_create_cgroup() Ret Value = 0 group found in filesystem
TEST 7:PASS : cgroup_attach_task() Ret Value = 0 Task found in group/s
==5083== Conditional jump or move depends on uninitialised value(s)
==5083== at 0x40070D8: strncat (mc_replace_strmem.c:214)
==5083== by 0x804BC92: main (libcgrouptest01.c:1271)
==5083== at 0x40070D8: strncat (mc_replace_strmem.c:214)
==5083== by 0x804BC92: main (libcgrouptest01.c:1271)
==5083==
==5083== Conditional jump or move depends on uninitialised value(s)
==5083== at 0x8049471: _ZL14group_modifiedPci (libcgrouptest01.c:1076)
==5083== by 0x804C583: main (libcgrouptest01.c:254)
==5083==
==5083== Conditional jump or move depends on uninitialised value(s)
==5083== at 0x8049479: _ZL14group_modifiedPci (libcgrouptest01.c:1076)
==5083== by 0x804C583: main (libcgrouptest01.c:254)
==5083==
==5083== Conditional jump or move depends on uninitialised value(s)
==5083== at 0x4007470: strncmp (mc_replace_strmem.c:314)
==5083== by 0x804949B: _ZL14group_modifiedPci (libcgrouptest01.c:1076)
==5083== by 0x804C583: main (libcgrouptest01.c:254)
==5083==
==5083== Conditional jump or move depends on uninitialised value(s)
==5083== at 0x4007475: strncmp (mc_replace_strmem.c:314)
==5083== by 0x804949B: _ZL14group_modifiedPci (libcgrouptest01.c:1076)
==5083== by 0x804C583: main (libcgrouptest01.c:254)
==5083==
==5083== Conditional jump or move depends on uninitialised value(s)
==5083== at 0x4007497: strncmp (mc_replace_strmem.c:314)
==5083== by 0x804949B: _ZL14group_modifiedPci (libcgrouptest01.c:1076)
==5083== by 0x804C583: main (libcgrouptest01.c:254)
==5083==
==5083== Conditional jump or move depends on uninitialised value(s)
==5083== at 0x40074A0: strncmp (mc_replace_strmem.c:314)
==5083== by 0x804949B: _ZL14group_modifiedPci (libcgrouptest01.c:1076)
==5083== by 0x804C583: main (libcgrouptest01.c:254)
TEST 8:FAIL : cgroup_modify_cgroup() Ret Value = 0 Parameter same cgroup
TEST 9:PASS : cgroup_new_cgroup() Ret Value = 0
TEST10:PASS : cgroup_modify_cgroup() Ret Value = 0
TEST11:PASS : cgroup_modify_cgroup() Ret Value = 50007
TEST12:PASS : cgroup_new_cgroup() Ret Value = 0
TEST13:PASS : cgroup_modify_cgroup() Ret Value = 0
TEST14:PASS : cgroup_get_cgroup() Ret Value = 50007 Parameter nullcgroup
TEST15:PASS : cgroup_get_cgroup() Ret Value = 50002 Parameter not created group
TEST16:PASS : cgroup_get_cgroup() Ret Value = 0
TEST 0:PASS : cgroup_new_cgroup() Ret Value = 0
TEST 0:PASS : cgroup_create_cgroup() Ret Value = 0 group found in filesystem
TEST17:FAIL : cgroup_get_cgroup() Ret Value = 50017
TEST16:PASS : cgroup_delete_cgroup() Ret Value = 0 Group deleted from filesystem
TEST17:PASS : cgroup_create_cgroup() Ret Value = 50007
TEST18:PASS : cgroup_delete_cgroup() Ret Value = 50007
TEST19:PASS : cgroup_compare_cgroup() Ret Value = 50011 Parameter nullcgroup
TEST20:PASS : cgroup_compare_cgroup() Ret Value = 0
TEST21:PASS : cgroup_compare_cgroup() Ret Value = 50017
==5083==
==5083== ERROR SUMMARY: 9 errors from 7 contexts (suppressed: 18 from 1)
==5083== malloc/free: in use at exit: 0 bytes in 0 blocks.
==5083== malloc/free: 279 allocs, 279 frees, 237,968 bytes allocated.
==5083== For counts of detected errors, rerun with: -v
==5083== All heap blocks were freed -- no leaks are possible.
Cleanup done
Running third set of testcases
==============================
==5134== Memcheck, a memory error detector.
==5134== Copyright (C) 2002-2007, and GNU GPL'd, by Julian Seward et al.
==5134== Using LibVEX rev 1804, a library for dynamic binary translation.
==5134== Copyright (C) 2004-2007, and GNU GPL'd, by OpenWorks LLP.
==5134== Using valgrind-3.3.0, a dynamic binary instrumentation framework.
==5134== Copyright (C) 2000-2007, and GNU GPL'd, by Julian Seward et al.
==5134== For more details, rerun with: -v
==5134==
C:DBG: fs_mounted as recieved from script=2
C:DBG: mountpoint1 as recieved from script=/dev/cgroup_controllers-1
C:DBG: mountpoint2 as recieved from script=/dev/cgroup_controllers-2
sanity check pass. cgroup
TEST 1:PASS : cgroup_init() Ret Value = 0
TEST 2:PASS : cgroup_attach_task() Ret Value = 0 Task found in group/s
TEST 3:PASS : cgroup_new_cgroup() Ret Value = 0
TEST 4:PASS : cgroup_create_cgroup() Ret Value = 0 group found in filesystem
TEST 5:PASS : cgroup_new_cgroup() Ret Value = 0
TEST 6:PASS : cgroup_create_cgroup() Ret Value = 0 group found in filesystem
TEST 5:PASS : cgroup_new_cgroup() Ret Value = 0
TEST 6:PASS : cgroup_create_cgroup() Ret Value = 0 group found in filesystem
TEST 7:PASS : cgroup_create_cgroup() Ret Value = 0 group found in filesystem
TEST 8:PASS : cgroup_attach_task() Ret Value = 0 Task found in group/s
TEST 9:PASS : cgroup_attach_task() Ret Value = 0 Task found in group/s
TEST10:PASS : cgroup_new_cgroup() Ret Value = 0
TEST11:PASS : cgroup_attach_task() Ret Value = 50002 Parameter not created group
TEST12:PASS : cgroup_new_cgroup() Ret Value = 0
TEST13:PASS : cgroup_modify_cgroup() Ret Value = 0
TEST14:PASS : cgroup_new_cgroup() Ret Value = 0
TEST15:PASS : cgroup_modify_cgroup() Ret Value = 0
TEST16:PASS : cgroup_delete_cgroup() Ret Value = 0 Group deleted from filesystem
TEST17:PASS : cgroup_delete_cgroup() Ret Value = 0 Group deleted from filesystem
TEST18:PASS : cgroup_new_cgroup() Ret Value = 0
TEST19:PASS : cgroup_create_cgroup() Ret Value = 0 group found under both controllers
TEST20:PASS : cgroup_attach_task() Ret Value = 0 Task found in group/s
TEST21:PASS : cgroup_new_cgroup() Ret Value = 0
TEST22:PASS : cgroup_modify_cgroup() Ret Value = 0 group modified under both controllers
TEST23:PASS : cgroup_delete_cgroup() Ret Value = 0 Group deleted from filesystem
TEST 0:PASS : cgroup_new_cgroup() Ret Value = 0
TEST 0:PASS : cgroup_create_cgroup() Ret Value = 0 group found in filesystem
TEST24:FAIL : cgroup_get_cgroup() Ret Value = 50018
==5134==
==5134== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 18 from 1)
==5134== malloc/free: in use at exit: 66,044 bytes in 15 blocks.
==5134== malloc/free: 279 allocs, 264 frees, 232,094 bytes allocated.
==5134== For counts of detected errors, rerun with: -v
==5134== searching for pointers to 15 not-freed blocks.
==5134== checked 6,381,388 bytes.
==5134==
==5134== 66,044 (18,064 direct, 47,980 indirect) bytes in 4 blocks are definitely lost in loss record 1 of 3
==5134== at 0x4004BA2: calloc (vg_replace_malloc.c:397)
==5134== by 0x4030DAA: cgroup_new_cgroup (wrapper.c:28)
==5134== by 0x8049656: new_cgroup(char*, char*, char*, int, int) (libcgrouptest01.c:1132)
==5134== by 0x8049917: create_new_cgroup_ds(int, char const*, int, int) (libcgrouptest01.c:744)
==5134== by 0x804C190: main (libcgrouptest01.c:485)
==5134==
==5134==
==5134== 20,980 bytes in 5 blocks are indirectly lost in loss record 2 of 3
==5134== at 0x4004BA2: calloc (vg_replace_malloc.c:397)
==5134== by 0x4030C3C: cgroup_add_value_string (wrapper.c:122)
==5134== by 0x8048FDB: _ZL17add_control_valueP17cgroup_controllerPcS1_i (libcgrouptest01.c:1113)
==5134== by 0x80496C9: new_cgroup(char*, char*, char*, int, int) (libcgrouptest01.c:1147)
==5134== by 0x8049917: create_new_cgroup_ds(int, char const*, int, int) (libcgrouptest01.c:744)
==5134== by 0x804C190: main (libcgrouptest01.c:485)
==5134==
==5134==
==5134== 27,000 bytes in 6 blocks are indirectly lost in loss record 3 of 3
==5134== at 0x4004BA2: calloc (vg_replace_malloc.c:397)
==5134== by 0x4030F5B: cgroup_add_controller (wrapper.c:62)
==5134== by 0x80496AE: new_cgroup(char*, char*, char*, int, int) (libcgrouptest01.c:1144)
==5134== by 0x8049917: create_new_cgroup_ds(int, char const*, int, int) (libcgrouptest01.c:744)
==5134== by 0x804C190: main (libcgrouptest01.c:485)
==5134==
==5134== LEAK SUMMARY:
==5134== definitely lost: 18,064 bytes in 4 blocks.
==5134== indirectly lost: 47,980 bytes in 11 blocks.
==5134== possibly lost: 0 bytes in 0 blocks.
==5134== still reachable: 0 bytes in 0 blocks.
==5134== suppressed: 0 bytes in 0 blocks.
Cleanup done
[root@gondor tests]#
projects / thirdparty / libcgroup.git / log
