Re: [PATCH RESEND] ptrace: make ptrace() fail if the tracee changed its pid unexpectedly

[Mathieu Desnoyers]

----- On Apr 27, 2021, at 2:26 AM, Oleg Nesterov oleg@xxxxxxxxxx wrote:
[...]
>> Is this something that should also target stable kernels ? AFAIU this change
>> won't break debuggers more that they are already in this scenario. Or maybe
>> it makes them fail in more obvious ways ?
> 
> Well, I am not sure this is stable material...
> 
> To me the problem is minor, and the patch adds the user-visible change.
> I think it would be safer to not add stable tag.

I'm fine either way. So given the relatively small impact of this problem
(not critical), this ptrace fix may not be worthy of a stable tag.

I just find it odd that a patch fixing an ABI design flaw ends up not being
CC'd to stable, but also does not expose any way for user-space to discover
this altered ABI behavior. It's a rather weird middle-ground between a fix
and a new feature.

That being said, there was no prior way for user-space to achieve a correct
behavior before this patch, so making it discoverable is kind of pointless.

Thanks,

Mathieu

-- 
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com