Hi Cyril and Wanlong,

any comments for this patch?

----- Original Message -----
> From: "Zhouping Liu" <[email protected]>
> To: "LTP List" <[email protected]>
> Sent: Monday, April 29, 2013 2:07:17 PM
> Subject: [LTP] [PATCH v3] mem/oom: fixed a cpuset error
> 
> For the below special NUMA system, oom0[4|5] failed:
>  # numactl -H
>  available: 2 nodes (0-1)
>  node 0 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
>  24 25 26 27
>  node 0 size: 0 MB
>  node 0 free: 0 MB
>  node 1 cpus:
>  node 1 size: 16384 MB
>  node 1 free: 14173 MB
>  node distances:
>  node   0   1
>    0:  10  40
>    1:  40  10
> 
> failed log:
>  oom04       1  TBROK  :  write /dev/cpuset/1/cpuset.mems: errno=EINVAL(22):
>  Invalid argument
> 
> The reason is that node0 only contains all CPUs, no any memory,
> and node1 contains all memory, but no CPUs. In the previous codes,
> we only do cpuset testing on an independent node, which caused the
> sub-cpuset cgroup to only contain CPUs or memory in one node, but
> that's not permitted in the scenario described above.
> 
> Signed-off-by: Zhouping Liu <[email protected]>
> Reviewed-by: Caspar Zhang <[email protected]>
> ---
> Change log:
>  v1 - v2:
>     fixed some grammar errors, and updated some comments.
>  v2 - v3:
>     amended some comment messages.
> 
>  testcases/kernel/mem/lib/mem.c   | 13 ++++++++++++-
>  testcases/kernel/mem/oom/oom04.c | 20 ++++++++++++++------
>  testcases/kernel/mem/oom/oom05.c | 26 ++++++++++++++++----------
>  3 files changed, 42 insertions(+), 17 deletions(-)
> 
> diff --git a/testcases/kernel/mem/lib/mem.c b/testcases/kernel/mem/lib/mem.c
> index 5397177..1232b4f 100644
> --- a/testcases/kernel/mem/lib/mem.c
> +++ b/testcases/kernel/mem/lib/mem.c
> @@ -792,7 +792,18 @@ void write_cpusets(long nd)
>       write_cpuset_files(CPATH_NEW, "mems", buf);
>  
>       gather_node_cpus(cpus, nd);
> -     write_cpuset_files(CPATH_NEW, "cpus", cpus);
> +     /*
> +      * If the 'nd' node doesn't contain any CPUs,
> +      * the first ID of CPU '0' will be used as
> +      * the value of cpuset.cpus.
> +      */
> +     if (strlen(cpus) != 0) {
> +             write_cpuset_files(CPATH_NEW, "cpus", cpus);
> +     } else {
> +             tst_resm(TINFO, "No CPUs in the node%ld; "
> +                             "using only CPU0", nd);
> +             write_cpuset_files(CPATH_NEW, "cpus", "0");
> +     }
>  
>       SAFE_FILE_PRINTF(NULL, CPATH_NEW "/tasks", "%d", getpid());
>  }
> diff --git a/testcases/kernel/mem/oom/oom04.c
> b/testcases/kernel/mem/oom/oom04.c
> index 4d3f2f4..ac48c36 100644
> --- a/testcases/kernel/mem/oom/oom04.c
> +++ b/testcases/kernel/mem/oom/oom04.c
> @@ -85,6 +85,8 @@ int main(int argc, char *argv[])
>  
>  void setup(void)
>  {
> +     int memnode, ret;
> +
>       tst_require_root(NULL);
>       tst_sig(FORK, DEF_HANDLER, cleanup);
>       TEST_PAUSE;
> @@ -93,12 +95,18 @@ void setup(void)
>       set_sys_tune("overcommit_memory", 1, 1);
>  
>       mount_mem("cpuset", "cpuset", NULL, CPATH, CPATH_NEW);
> -     if (is_numa(cleanup) > 0)
> -             /* For NUMA system, using the first node for cpuset.mems */
> -             write_cpusets(get_a_numa_node(cleanup));
> -     else
> -             /* For nonNUMA system, using node0 for cpuset.mems */
> -             write_cpusets(0);
> +
> +     /*
> +      * Some nodes do not contain memory, so use
> +      * get_allowed_nodes(NH_MEMS) to get a memory
> +      * node. This operation also applies to Non-NUMA
> +      * systems.
> +      */
> +     ret = get_allowed_nodes(NH_MEMS, 1, &memnode);
> +     if (ret < 0)
> +             tst_brkm(TBROK, NULL, "Failed to get a memory node "
> +                                   "using get_allowed_nodes()");
> +     write_cpusets(memnode);
>  }
>  
>  void cleanup(void)
> diff --git a/testcases/kernel/mem/oom/oom05.c
> b/testcases/kernel/mem/oom/oom05.c
> index 15feba5..545b5f0 100644
> --- a/testcases/kernel/mem/oom/oom05.c
> +++ b/testcases/kernel/mem/oom/oom05.c
> @@ -108,24 +108,30 @@ int main(int argc, char *argv[])
>  
>  void setup(void)
>  {
> +     int ret, memnode;
> +
>       tst_require_root(NULL);
>       tst_sig(FORK, DEF_HANDLER, cleanup);
>       TEST_PAUSE;
>  
> +     overcommit = get_sys_tune("overcommit_memory");
> +     set_sys_tune("overcommit_memory", 1, 1);
> +
>       mount_mem("memcg", "cgroup", "memory", MEMCG_PATH, MEMCG_PATH_NEW);
>       mount_mem("cpuset", "cpuset", NULL, CPATH, CPATH_NEW);
>       write_memcg();
>  
> -     set_sys_tune("overcommit_memory", 1, 1);
> -
> -     if (is_numa(cleanup))
> -             /* For NUMA system, using the first node for cpuset.mems */
> -             write_cpusets(get_a_numa_node(cleanup));
> -     else
> -             /* For nonNUMA system, using node0 for cpuset.mems */
> -             write_cpusets(0);
> -
> -     overcommit = get_sys_tune("overcommit_memory");
> +     /*
> +      * Some nodes do not contain memory, so use
> +      * get_allowed_nodes(NH_MEMS) to get a memory
> +      * node. This operation also applies to Non-NUMA
> +      * systems.
> +      */
> +     ret = get_allowed_nodes(NH_MEMS, 1, &memnode);
> +     if (ret < 0)
> +             tst_brkm(TBROK, NULL, "Failed to get a memory node "
> +                                   "using get_allowed_nodes()");
> +     write_cpusets(memnode);
>  }
>  
>  void cleanup(void)
> --
> 1.7.11.7
> 
> 
> ------------------------------------------------------------------------------
> Try New Relic Now & We'll Send You this Cool Shirt
> New Relic is the only SaaS-based application performance monitoring service
> that delivers powerful full stack analytics. Optimize and monitor your
> browser, app, & servers with just a few lines of code. Try New Relic
> and get this awesome Nerd Life shirt! http://p.sf.net/sfu/newrelic_d2d_apr
> _______________________________________________
> Ltp-list mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/ltp-list
> 

-- 
Thanks,
Zhouping

------------------------------------------------------------------------------
Introducing AppDynamics Lite, a free troubleshooting tool for Java/.NET
Get 100% visibility into your production application - at no cost.
Code-level diagnostics for performance bottlenecks with <2% overhead
Download for free and get started troubleshooting in minutes.
http://p.sf.net/sfu/appdyn_d2d_ap1
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

Reply via email to