[LTP] [PATCH] cpuset_hotplug_test/cpuset_hotplug: Fix bug for obtaining task_cpus value in root_cpu_hotplug_test
Samuel Zou
zou_wei@huawei.com
Mon Jun 21 11:18:53 CEST 2021
Hi Xu,
You'are right, I miss to get task_cpus value in general_cpu_hotplug_test().
Thanks for your review and agree with your modification.
On 2021/6/21 16:56, xuyang2018.jy@fujitsu.com wrote:
> Hi Zou
>
> It looks like this patch also miss to get task_cpus value
> intgeneric_cpu_hotplug_test when expect cpu value is EMPTY.
>
> I can reproduce this easily by increasing the possible_cpus value in
> cmdline. Merge your patch, this case still fails as below:
> cpuset_hotplug 1 TINFO: CPUs are numbered continuously starting at 0 (0-3)
> cpuset_hotplug 1 TINFO: Nodes are numbered continuously starting at 0 (0-3)
> /opt/ltp/testcases/bin/cpuset_funcs.sh: line 179: 5676 Killed
> /bin/cat /dev/zero > /dev/null 2>&1
> cpuset_hotplug 1 TPASS: Cpuset vs CPU hotplug test succeeded.
> /opt/ltp/testcases/bin/cpuset_funcs.sh: line 179: 5714 Killed
> /bin/cat /dev/zero > /dev/null 2>&1
> cpuset_hotplug 3 TPASS: Cpuset vs CPU hotplug test succeeded.
> /opt/ltp/testcases/bin/cpuset_funcs.sh: line 179: 5750 Killed
> /bin/cat /dev/zero > /dev/null 2>&1
> cpuset_hotplug 5 TPASS: Cpuset vs CPU hotplug test succeeded.
> cpuset_hotplug 7 TFAIL: task's cpu present list isn't expected(Result:
> 0-7, Expect: 0-3).
> /opt/ltp/testcases/bin/cpuset_funcs.sh: line 179: 5795 Killed
> /bin/cat /dev/zero > /dev/null 2>&1
> /opt/ltp/testcases/bin/cpuset_funcs.sh: line 179: 5846 Killed
> /bin/cat /dev/zero > /dev/null 2>&1
> cpuset_hotplug 9 TPASS: Cpuset vs CPU hotplug test succeeded.
> /opt/ltp/testcases/bin/cpuset_funcs.sh: line 179: 5894 Killed
> /bin/cat /dev/zero > /dev/null 2>&1
> cpuset_hotplug 11 TPASS: Cpuset vs CPU hotplug test succeeded.
>
>
> Also since we have use present value, we shoud adjust output to match
> it. I modify this patch as below, What do you think about it?
>
> ---
> a/testcases/kernel/controllers/cpuset/cpuset_hotplug_test/cpuset_hotplug_test.sh
> +++
> b/testcases/kernel/controllers/cpuset/cpuset_hotplug_test/cpuset_hotplug_test.sh
> @@ -93,9 +93,7 @@ root_cpu_hotplug_test()
>
> root_cpus="`cat $CPUSET/cpuset.cpus`"
>
> - task_cpus="`cat /proc/$tst_pid/status | grep Cpus_allowed_list`"
> - task_cpus="`echo $task_cpus | sed -e 's/Cpus_allowed_list: //'`"
> -
> + task_cpus="`cat /sys/devices/system/cpu/present`"
> check_result "$root_cpus" "$expect_cpus"
> ret=$?
> if [ $ret -eq 0 ]
> @@ -103,7 +101,7 @@ root_cpu_hotplug_test()
> check_result "$task_cpus" "$expect_task_cpus"
> ret=$?
> if [ $ret -ne 0 ]; then
> - tst_resm TFAIL "task's allowed list isn't
> expected.(Result: $task_cpus, Expect: $expect_task_cpus)"
> + tst_resm TFAIL "task's present list isn't
> expected.(Result: $task_cpus, Expect: $expect_task_cpus)"
> fi
> else
> tst_resm TFAIL "root group's cpus isn't
> expected(Result: $root_cpus, Expect: $expect_cpus)."
> @@ -166,7 +164,6 @@ general_cpu_hotplug_test()
> sleep 1
>
> cpus="`cat $path/cpuset.cpus`"
> -
> task_cpus="`cat /proc/$tst_pid/status | grep Cpus_allowed_list`"
> task_cpus="`echo $task_cpus | sed -e 's/Cpus_allowed_list: //'`"
>
> @@ -186,6 +183,7 @@ general_cpu_hotplug_test()
> /bin/kill -s SIGKILL $tst_pid
> return 1
> fi
> + task_cpus="`cat /sys/devices/system/cpu/present`"
> fi
>
> check_result "$cpus" "$expect_cpus"
> @@ -194,7 +192,7 @@ general_cpu_hotplug_test()
> check_result $task_cpus $expect_task_cpus
> ret=$?
> if [ $ret -ne 0 ]; then
> - tst_resm TFAIL "task's cpu allowed list isn't
> expected(Result: $task_cpus, Expect: $expect_task_cpus)."
> + tst_resm TFAIL "task's cpu present list isn't
> expected(Result: $task_cpus, Expect: $expect_task_cpus)."
> fi
> else
> if [ "$cpus" = "" ]; then
>
>
> Best Regards
> Yang XU
>> Hi Samuel, Zou
>>> Hi,
>>>
>>> A friendly reminder for the two patchs in:
>>>
>>> https://patchwork.ozlabs.org/project/ltp/patch/1617707717-63693-1-git-send-email-zou_wei@huawei.com/
>>>
>>>
>>>
>>> https://patchwork.ozlabs.org/project/ltp/patch/1617701249-62196-1-git-send-email-zou_wei@huawei.com/
>>>
>>>
>>>
>>> Can someone please review the patchs, comment, and if appropriate commit
>>> them?
>>>
>>> Regards,
>>> Zou Wei
>>>
>>> On 2021/4/6 17:27, Zou Wei wrote:
>>>> --------------------------
>>>>
>>>> 1. The method of obtaining $task_cpus from root_cpu_hotplug_test is
>>>> incorrect, because the value of Cpus_allowed_list is related
>>>> to /sys/devices/system/cpu/possible and
>>>> /sys/devices/system/cpu/isolated.
>>>>
>>>> 2. If isolcpus is configured in cmdline, the value of
>>>> Cpus_allowed_list is equal to /sys/devices/system/cpu/possible minus
>>>> /sys/devices/system/cpu/isolated. In this case, $task_cpus is
>>>> inconsistent
>>>> with $expect_task_cpus, test will be failed.
>>>>
>>>> 3. Need to change the method of obtaining $task_cpu
>>>> from /sys/devices/system/cpu/present.
>> Thanks for your patch, I also met the same problem when
>> /sys/devices/system/cpu/possible is not equal
>> to/sys/devices/system/cpu/possible/present .
>> Cpus_allowed_list: 0-511
>> but my present cpu is 0-103
>>
>> Then I use possible_cpus==104 in /etc/defaut/grub, cpu_hotplug case then
>> pass.
>>
>> According to kernel documentation[1] and kernel fs/proc/array code[2](It
>> just print cpu mask instead of present cpu), So using
>> /sys/devices/system/cpu/present is right instead of getting cpu_allow_list.
>>
>> possible: cpus that have been allocated resources and can be
>> brought online if they are present.
>>
>> present: cpus that have been identified as being present in
>> the system.
>>
>>
>> Reviewed-by: Yang Xu <xuyang2018.jy@fujitsu.com>
>>
>> [1]https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/ABI/testing/sysfs-devices-system-cpu
>>
>> [2]https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/fs/proc/array.c#n418
>>
>>
>> ps:
>> Also find a kernel documentation problem and have sent a patch to remove
>> non-existed cpu-hotplug.txt in admin-guide/cputopology.rst.
>>
>>
>> Best Regards
>> Yang Xu
>>>>
>>>> 4. The description of cpu_present_mask:
>>>> Bitmap of CPUs currently present in the system.
>>>> Not all of them may be online. When physical hotplug is processed
>>>> by the relevant subsystem (e.g ACPI) can change and new bit either be
>>>> added or removed from the map depending on the event is
>>>> hot-add/hot-remove. There are currently no locking rules as of now.
>>>> Typical usage is to init topology during boot,
>>>> at which time hotplug is disabled.
>>>>
>>>> Signed-off-by: Zou Wei <zou_wei@huawei.com>
>>>> ---
>>>> .../controllers/cpuset/cpuset_hotplug_test/cpuset_hotplug_test.sh | 3
>>>> +--
>>>> 1 file changed, 1 insertion(+), 2 deletions(-)
>>>>
>>>> diff --git
>>>> a/testcases/kernel/controllers/cpuset/cpuset_hotplug_test/cpuset_hotplug_test.sh
>>>>
>>>> b/testcases/kernel/controllers/cpuset/cpuset_hotplug_test/cpuset_hotplug_test.sh
>>>>
>>>>
>>>> index 155e536..2c6993a 100755
>>>> ---
>>>> a/testcases/kernel/controllers/cpuset/cpuset_hotplug_test/cpuset_hotplug_test.sh
>>>>
>>>>
>>>> +++
>>>> b/testcases/kernel/controllers/cpuset/cpuset_hotplug_test/cpuset_hotplug_test.sh
>>>>
>>>>
>>>> @@ -93,8 +93,7 @@ root_cpu_hotplug_test()
>>>> root_cpus="`cat $CPUSET/cpuset.cpus`"
>>>> - task_cpus="`cat /proc/$tst_pid/status | grep Cpus_allowed_list`"
>>>> - task_cpus="`echo $task_cpus | sed -e 's/Cpus_allowed_list: //'`"
>>>> + task_cpus="`cat /sys/devices/system/cpu/present`"
>>>> check_result "$root_cpus" "$expect_cpus"
>>>> ret=$?
>>>>
>>>
More information about the ltp
mailing list