Re: [PATCH 20/22] x86,word-at-a-time: Remove .fixup usage

[Joe Lawrence]

On 11/15/21 7:36 AM, Miroslav Benes wrote:
> On Fri, 12 Nov 2021, Josh Poimboeuf wrote:
> 
>> On Fri, Nov 12, 2021 at 10:33:36AM +0100, Peter Zijlstra wrote:
>>> On Thu, Nov 11, 2021 at 05:50:03PM -0800, Josh Poimboeuf wrote:
>>>
>>>> Hm, I think there is actually a livepatch problem here.
>>>
>>> I suspected as much, because I couldn't find any code dealing with it
>>> when I looked in a hurry.. :/
>>>
>>>> Some ideas to fix:
>>>
>>>> c) Update the reliable stacktrace code to mark the stack unreliable if
>>>>    it has a function with ".cold" in the name?
>>>
>>> Why not simply match func.cold as func in the transition thing? Then
>>> func won't get patched as long as it (or it's .cold part) is in use.
>>> This seems like the natural thing to do.
>>
>> Well yes, you're basically hinting at my first two options a and b:
>>
>> a) Add a field to 'klp_func' which allows the patch module to specify a
>>    function's .cold counterpart?
>>
>> b) Detect such cold counterparts in klp_enable_patch()?  Presumably it
>>    would require searching kallsyms for "<func>.cold", which is somewhat
>>    problematic as there might be duplicates.
>>
>> It's basically a two-step process:  1) match func to .cold if it exists;
>> 2) check for both in klp_check_stack_func().  The above two options are
>> proposals for the 1st step.  The 2nd step was implied.
> 
> This reminded me... one of the things I have on my todo list for a long 
> time is to add an option for a live patch creator to specify functions 
> which are not contained in the live patch but their presence on stacks 
> should be checked for. It could save some space in the final live patch 
> when one would add functions (callers) just because the consistency 
> requires it.
> 

Yea, I've used this technique once (adding a nop to a function so
kpatch-build would detect and include in klp_funcs[]) to make a set of
changes safer with respect to the consistency model.  Making it simpler
to for the livepatch author to say, "I'm not changing foo(), but I don't
want it doing anything while patching a task" sounds reasonable.

> I took as a convenience feature with a low priority and forgot about it. 
> The problem above changes it. So should we take the opportunity and 
> implement both in one step? I wanted to include a list of functions in 
> on a patch level (klp_patch structure) and klp_check_stack() would just 
> have more to check.
> 

As far as I read the original problem, I think it would solve for that,
too, so I would say go for it.

-- 
Joe