Re: [PATCH] linux: sysconf: limit _SC_MAX_ARG to 6 MiB [BZ #25305]

[Michal Nazarewicz <mina86@mina86.com>]

>> * Adhemerval Zanella:
>>> IMHO being conservative and use the lower bound of all supported kernels.
>>> I really don't think trying to be smart here with dynamic probing
>>> will add much, specially since this upper limit is what kernel will
>>> support from now on.

> On 07/04/2021 16:04, Florian Weimer wrote:
>> Ah, if the conservative choice does not penalize newer kernels, then I
>> guess that's okay as well.

Indeed; the 6 MiB bound penalises old unsupported kernels.  Barring
Linux changing things more, the calculation in this patch is what kernel
is using going forward.

On Wed, Apr 07 2021, Adhemerval Zanella wrote:
> My understanding is newer kernel are more restrictive since they limit
> maximum argument plus environment size to up to 6MB, different than older 
> kernels that have a higher bar to up 1/4 of maximum stack size.

Specifically, Linux’s limit is:

    clamp(128 KiB, max stack size / 4, 6 MiB)

so stack size still plays a role but only if it’s in 0.5–24 MiB range.
Outside of that range one of the static bounds is used.

> So I take you don't oppose to the patch.

>> Then the argument goes like this: If you want us to make the limit
>> dynamic, add something to the auxiliary vector. 8-)

> I am not sure if kernel will be willing to make this dynamic, at least
> it seems not be an issue.

I can’t read Linus’ head but I think you’re right; it seems the attitude
in LKML is to use 128 KiB and be done with it.

-- 
Best regards
ミハウ “𝓶𝓲𝓷𝓪86” ナザレヴイツ
«If at first you don’t succeed, give up skydiving»