[LTP] [PATCH] lib/tst_cgroup: Reset cgroup.clone_children value

[Li Wang]

On Fri, Aug 7, 2020 at 7:16 PM Yang Xu <xuyang2018.jy@cn.fujitsu.com> wrote:

> Hi Li
>
>
> >
> >
> > On Fri, Aug 7, 2020 at 5:42 PM Yang Xu <xuyang2018.jy@cn.fujitsu.com
> > <mailto:xuyang2018.jy@cn.fujitsu.com>> wrote:
> >
> >     When running cgroup test cases(like cpuset01 or oom05) about cpuset
> >     subsystem
> >     firstly, then cpuset_inherit case will fail because the value of
> >     cgroup.clone_children has been changed into 1 on cgroup-v1. Reset
> >     this value
> >     when calling tst_cgroupN_umount.
> >
> >     Signed-off-by: Yang Xu <xuyang2018.jy@cn.fujitsu.com
> >     <mailto:xuyang2018.jy@cn.fujitsu.com>>
> >     ---
> >       lib/tst_cgroup.c | 8 ++++++++
> >       1 file changed, 8 insertions(+)
> >
> >     diff --git a/lib/tst_cgroup.c b/lib/tst_cgroup.c
> >     index 9459f7ea0..764951afa 100644
> >     --- a/lib/tst_cgroup.c
> >     +++ b/lib/tst_cgroup.c
> >     @@ -9,6 +9,8 @@
> >       #include <stdio.h>
> >       #include <stdlib.h>
> >       #include <sys/mount.h>
> >     +#include <fcntl.h>
> >     +#include <unistd.h>
> >
> >       #include "tst_test.h"
> >       #include "tst_safe_macros.h"
> >     @@ -123,6 +125,7 @@ static void tst_cgroupN_umount(const char
> >     *mnt_path, const char *new_path)
> >              FILE *fp;
> >              int fd;
> >              char s_new[BUFSIZ], s[BUFSIZ], value[BUFSIZ];
> >     +       char knob_path[PATH_MAX];
> >
> >              if (!tst_is_mounted(mnt_path))
> >                      return;
> >     @@ -151,6 +154,11 @@ static void tst_cgroupN_umount(const char
> >     *mnt_path, const char *new_path)
> >                                  != (ssize_t)strlen(value) - 1)
> >                                      tst_res(TWARN | TERRNO, "write %s",
> s);
> >              }
> >     +       if (tst_cg_ver & TST_CGROUP_V1) {
> >
> >
> > To recognize cgroup_v1 is not enough here, because it will
> > be failed "with no such cgroup.clone_children file" on MEMCG umount if
> > the system no CPUSET mounting.
> I has umounted cpuset, but I still see this file in memory as below:
>

It's not the same file, here set to 1 is '../cpuset/cgroup.clone_children'
which belong to the CPUSET but not the MEMCG one.

i.e.
# echo 1 >/sys/fs/cgroup/memory/cgroup.clone_children
# cat /sys/fs/cgroup/memory/cgroup.clone_children
1
# cat /sys/fs/cgroup/cpuset/cgroup.clone_children
0



>
> #mount |grep cpuset
> nothing
> # ls -l /sys/fs/cgroup/memory/cgroup.clone_children
> -rw-r--r--. 1 root root 0 Aug  7 08:16
> /sys/fs/cgroup/memory/cgroup.clone_children
>
> >
> > Maybe a smart way is to save the cgroup.clone_children value, restore it
> > if it has been changed in the setup phase. What do u think?
> We can just use print and scanf api to do this.
> But I don't know this whether takes bad effects on complexd nesting
> situation(has sub cgroup).
>

We could just do scanf that for CPUSET subsystem, and do nothing if no
CPUSET mounting.



>
> >
> >     +               sprintf(knob_path, "%s/cgroup.clone_children",
> >     mnt_path);
> >     +               if (!access(knob_path, F_OK))
> Here has a check for cgroup.clone_children so it should not trigger un
> such file error.
>

No, the tst_cgroupN_umount just unmount the DIR but do nothing to
distinguish what kind of group it is, so it will also do same thing for all
cgroup types.



>
> ps: I will add clone_children swith into cpu_inherit case.
>
> Best Regards
> Yang Xu
> >     +                       SAFE_FILE_PRINTF(knob_path, "%d", 0);
> >     +       }
> >              if (fd != -1)
> >                      close(fd);
> >              if (fp != NULL)
> >     --
> >     2.23.0
> >
> >
> >
> >
> >
> > --
> > Regards,
> > Li Wang
>
>
>

-- 
Regards,
Li Wang
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linux.it/pipermail/ltp/attachments/20200807/ad1286a3/attachment-0001.htm>