Re: [PATCH] clocksource/drivers/arm_arch_timer: Update sched_clock when non-boot CPUs need counter workaround

[王法杰 <wangfajie@longcheer.com>]

Hi Marc Zyngier

We found the APPS Crash issue on MTBF test.
Brief crash information, APPS Crash - Kernel BUG at /sched/walt/sched_avg.c:281! in sched_update_nr_prod flow

[Test equipment]
1. Number of phone:  200 pcs
2. CPU info of phone:  CPU architecture with Quad Cortex-A73 and Quad Cortex-A53

[Preset conditions]
1. Android 13 with kernel 5.15
2. Install application of top 20
3. Connected to Wi-Fi
4. Connect the adapter to charge the phone
5. Test duration 48 hours

[Expected results]
0 crash happened.

[Test results without [PATCH] clocksource/drivers/arm_arch_timer: Update sched_clock when non-boot CPUs need counter workaround ]
1. First round of test, found 3 phones with APPS Crash issue on /sched/walt/sched_avg.c:281! in sched_update_nr_prod flow
2. Second round of test, found 7 phones with APP Crash issue on /sched/walt/sched_avg.c:281! in sched_update_nr_prod flow

[Test results with [PATCH] clocksource/drivers/arm_arch_timer: Update sched_clock when non-boot CPUs need counter workaround ] 
1. First round of test, 0 crash happened.
2. Second round of test, 0 crash happened.

Tested-by: wangfajie <wangfajie@longcheer.com>
Tested-by: liurenwang <liurenwang@longcheer.com>
Tested-by: zhanghui <zhanghui5@longcheer.com>
Tested-by: liangke <liangke1@xiaomi.com>

So we think the PATCH is working and it can fix APPS crash issue. Many thanks your time.

Best regards!
Wangfajie


-----邮件原件-----
发件人: Marc Zyngier [mailto:maz@kernel.org] 
发送时间: 2023年1月19日 17:52
收件人: 刘琦 <liuqi405@icloud.com>
抄送: daniel.lezcano@kernel.org; linux-arm-kernel@lists.infradead.org; linux-kernel@vger.kernel.org; mark.rutland@arm.com; quic_ylal@quicinc.com; stable@vger.kernel.org; will@kernel.org; 王法杰 <wangfajie@longcheer.com>; 刘仁旺 <liurenwang@longcheer.com>; 张辉 <zhanghui5@longcheer.com>; liangke1@xiaomi.com
主题: Re: [PATCH] clocksource/drivers/arm_arch_timer: Update sched_clock when non-boot CPUs need counter workaround

On 2023-01-19 07:59, 刘琦 wrote:
> [Test Report]
> Result: Test Pass
> 
> A total of two rounds of pending testing
>      a. The first round of hanging test
>           Number of machines: 200
>           Hanging test duration: 48h
>           Hanging test results: no walt crash problem
>      b. The second round of hanging test
>           Number of machines: 200
>           Hanging test duration: 72h
>           Hanging test results: no walt crash problem
> 
> Tested-by: wangfajie <wangfajie@longcheer.com>
> Tested-by: liurenwang <liurenwang@longcheer.com>
> Tested-by: zhanghui <zhanghui5@longcheer.com>
> Tested-by: liangke <liangke1@xiaomi.com>

Thanks for this.

The only issue here is that that you don't explain what you tested, nor how you tested it.

It is also a patch that has known defects (you just have to read the thread for the details)... This makes this testing, no matter how thorough it is, rather ineffective.

         M.
--
Jazz is not dead. It just smells funny...