Re: [RFC PATCH 0/2] Alloc kfence_pool after system startup

[Marco Elver]

On Fri, 4 Mar 2022 at 03:25, Tianchen Ding <dtcccc@xxxxxxxxxxxxxxxxx> wrote:
>
> On 2022/3/3 17:30, Marco Elver wrote:
>
> Thanks for your replies.
> I do see setting a large sample_interval means almost disabling KFENCE.
> In fact, my point is to provide a more “flexible” way. Since some Ops
> may be glad to use something like on/off switch than 10000ms interval. :-)

Have you already successfully caught bugs by turning KFENCE on _in
reaction_ to some suspected issues? We really do not think that
switching on KFENCE _after_ having observed a bug, especially on a
completely different machine, is at all reliable.

While your patches are appreciated, I think your usecase doesn't make
sense to us (based on our experience). I think this flexibility is
nice-to-have, so I think the justification just needs changing, to
avoid misleading other folks. Please see comments on the other
patches.

Thanks,
-- Marco