> 2025年1月22日 21:26，Mariusz Tkaczyk <mtkaczyk@xxxxxxxxxx> 写道：
>
> On Wed, 22 Jan 2025 20:43:19 +0800
> Coly Li <colyli@xxxxxxx> wrote:
>>
[snipped]
>> So the patch is just a best-effort-try, if the binary is not
>> installed in /sbin, /usr/sbin or /usr/local/sbin, my patch just gives
>> up.
>>
>
> Maybe we can print error then? It would be useful for programmers
> to understand the problem. Sometimes, to debug early stages I simply
> redirected all error messages from mdadm to /dev/kmsg.
>
Currently there is error message: sh: modprobe: command not found.
This message was caught from console and this was how the root cause was found.
This might be enough?
>
>>>
>>>> +
>>>> if (system("modprobe md_mod") == 0)
>>>> fd = open(new_array_file, O_WRONLY);
>>>
>>>> }
>>>
>>> The change will affect code executed later, probably we don't want
>>> that. Shouldn't we restore old PATH here to minimize risk?
>>
>> For my understanding if the code was called after boot up, these path
>> should be set already by shell initialization scripts. And for udev
>> context, it is called and exited, and NOT shared with other udev
>> tasks, almost no influence.
>
> I see, it is more theoretical problem, almost to possible to met.
> All fine then, we can continue with no clean up here as I cannot find
> normal scenario it can occur.
>
> Your change (after getenv fixes) LGTM. You can take a look into concept
> I proposed but I have no strong preference.
Copied. V3 posted.
Thanks.
Coly Li
