Re: [PATCH v4 2/6] mm/slab: Plumb kmem_buckets into __do_kmalloc_node()

[Vlastimil Babka]

On 6/4/24 12:44 AM, Kees Cook wrote:
> On Mon, Jun 03, 2024 at 07:06:15PM +0200, Vlastimil Babka wrote:
>> On 5/31/24 9:14 PM, Kees Cook wrote:
>> > Introduce CONFIG_SLAB_BUCKETS which provides the infrastructure to
>> > support separated kmalloc buckets (in the follow kmem_buckets_create()
>> > patches and future codetag-based separation). Since this will provide
>> > a mitigation for a very common case of exploits, enable it by default.
>> 
>> Are you sure? I thought there was a policy that nobody is special enough
>> to have stuff enabled by default. Is it worth risking Linus shouting? :)
> 
> I think it's important to have this enabled given how common the
> exploitation methodology is and how cheap this solution is. Regardless,
> if you want it "default n", I can change it.

Yeah, I'd just recommend it in the help, noting it has a bit of memory
overhead. Defaults are not that important anyway IMHO, either it's distro
doing the config, and individually security conscious people should know
what they are doing.

> 
> This looks really nice, thank you! This is well aligned with the codetag
> followup, which also needs to have "size" be very easy to find (to the
> macros can check for compile-time-constant or not).
> 
> I will go work from your branch...

Great!

> Thanks!
> 
> -Kees
>