Re: [PATCH v2 1/6] Fix "x86/alternatives: Lockdep-enforce text_mutex in text_poke*()"

[Nadav Amit]

at 12:53 PM, Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:

> On Thu, Sep 06, 2018 at 07:42:14PM +0000, Nadav Amit wrote:
>> at 12:40 PM, Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
>> 
>>> On Sun, Sep 02, 2018 at 10:32:19AM -0700, Nadav Amit wrote:
>>>> text_mutex is expected to be held before text_poke() is called, but we
>>>> cannot add a lockdep assertion since kgdb does not take it, and instead
>>>> *supposedly* ensures the lock is not taken and will not be acquired by
>>>> any other core while text_poke() is running.
>>>> 
>>>> The reason for the "supposedly" comment is that it is not entirely clear
>>>> that this would be the case if gdb_do_roundup is zero.
>>> 
>>> Argh, that's pretty shit code...
>>> 
>>> Not only is that text_mutex abuse ugly, so too is the fixmap usage from
>>> IRQ context. I suppose this really does require your alternative mm
>>> patches for text_poke().
>> 
>> Right, I forgot about thatâ
> 
> With that CR3 trickery, we can rid ourselves of the text_mutex
> requirement, since concurrent text_poke is 'safe'. That would clean up
> the kgdb code quite a bit.

I donât know. Iâm somewhat worried with multiple mechanisms potentially
changing the same code at the same time - and maybe ending up with some
mess.