linux-kernel - Re: [PATCH for-next 1/3] selftests/watchdog: add count parameter for watchdog-test

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <3ee0d14a-7f6b-4ef4-9349-d6b0f14ba9e8@fujitsu.com>
Date: Mon, 28 Oct 2024 05:45:29 +0000
From: "Zhijian Li (Fujitsu)" <lizhijian@...itsu.com>
To: Shuah Khan <skhan@...uxfoundation.org>, "linux-kselftest@...r.kernel.org"
	<linux-kselftest@...r.kernel.org>
CC: "shuah@...nel.org" <shuah@...nel.org>, "linux-kernel@...r.kernel.org"
	<linux-kernel@...r.kernel.org>
Subject: Re: [PATCH for-next 1/3] selftests/watchdog: add count parameter for
 watchdog-test



On 28/10/2024 13:31, Shuah Khan wrote:
> On 10/27/24 22:02, Zhijian Li (Fujitsu) wrote:
>>
>>
>> On 28/10/2024 11:29, Shuah Khan wrote:
>>> On 10/27/24 18:50, Zhijian Li (Fujitsu) wrote:
>>>>
>>>>
>>>> On 27/10/2024 08:28, Shuah Khan wrote:
>>>>> On 10/24/24 19:39, Li Zhijian wrote:
>>>>>> Currently, watchdog-test keep running until it gets a SIGINT. However,
>>>>>> when watchdog-test is executed from the kselftests framework, where it
>>>>>> launches test via timeout which will send SIGTERM in time up. This could
>>>>>> lead to
>>>>>> 1. watchdog haven't stop, a watchdog reset is triggered to reboot the OS
>>>>>>       in silent.
>>>>>> 2. kselftests gets an timeout exit code, and judge watchdog-test as
>>>>>>      'not ok'
>>>>>>
>>>>> This test isn't really supposed to be run from kselftest framework.
>>>>> This is the reason why it isn't included in the default run.
>>>>
>>>> May I know what's the default run, is it different from `make run_tests` ?
>>>
>>> No it isn't. "make kselftest" runs only the targets mentioned in the
>>> selftests Makefile. That is considered the kselftest default run.
>>
>> Hey, Shuah,
>>
>>
>> Thanks for your explanation.
>> If that is the case, I do not have an urgent need for the current patch, expect
>> I'd like to avoid the reboot issue after an accidentally `make run_tests`
>>
>> Some changes are make as below, please take a look. I will send it out we reach a consensus.
>>
>>
>> commit 2296f9d88fde4921758a45bf160a7f1b9d4678a0 (HEAD)
>> Author: Li Zhijian <lizhijian@...itsu.com>
>> Date:   Mon Oct 28 11:54:03 2024 +0800
>>
>>       selftests/watchdog-test: Fix system accidentally reset after watchdog-test
>>       After `make run_tests` to run watchdog-test, a system reboot would
>>       happen due to watchdog not stop.
>>       ```
> 
> The system shouldn't reboot just because watchdog test is left running.
> watchdog test keeps calling ioctl() with WDIOC_KEEPALIVE to make sure
> the watchdog card timer is reset.

Err..

How watchdog test keep calling ioctl() with WDIOC_KEEPALIVE after ./watchdog_test has finished?

In my understanding, the cause is that, ./watchdog_test didn't goto neither
A)
347 end:
348         /*
349          * Send specific magic character 'V' just in case Magic Close is
350          * enabled to ensure watchdog gets disabled on close.
351          */
352         ret = write(fd, &v, 1);
353         if (ret < 0)
354                 printf("Stopping watchdog ticks failed (%d)...\n", errno);

nor B)

  68 static void term(int sig)
  69 {
  70         int ret = write(fd, &v, 1);
  71
  72         close(fd);
  73         if (ret < 0)
  74                 printf("\nStopping watchdog ticks failed (%d)...\n", errno);
  75         else
  76                 printf("\nStopping watchdog ticks...\n");
  77         exit(0);
  78 }

to "ensure watchdog gets disabled on close"


The timeout default signal is SIGTERM, watchdog_test only registered SIGINT handler.

Thanks
Zhijian



> 
> If you are seeing reboots, that means watchdog test couldn't reset the
> timer. This usually mean system is unresponsive or something is wrong
> with the watchdog card on your system.
> 
> This is the behavior you would expect from a watchdog timer. Does your
> system have a watchdog card ot or you enabling softdog module?
> 
> Either way there is some other reason for the system reboot.
> 
> thanks,
> -- Shuah