Re: [PATCH 2/6] mm: kfence: check kfence pool size at building time

[Muchun Song <muchun.song@linux.dev>]

> On Mar 28, 2023, at 18:14, Marco Elver <elver@google.com> wrote:
> 
> On Tue, 28 Mar 2023 at 11:58, 'Muchun Song' via kasan-dev
> <kasan-dev@googlegroups.com> wrote:
>> 
>> Check kfence pool size at building time to expose problem ASAP.
>> 
>> Signed-off-by: Muchun Song <songmuchun@bytedance.com>
>> ---
>> mm/kfence/core.c | 7 +++----
>> 1 file changed, 3 insertions(+), 4 deletions(-)
>> 
>> diff --git a/mm/kfence/core.c b/mm/kfence/core.c
>> index de62a84d4830..6781af1dfa66 100644
>> --- a/mm/kfence/core.c
>> +++ b/mm/kfence/core.c
>> @@ -841,10 +841,9 @@ static int kfence_init_late(void)
>>                return -ENOMEM;
>>        __kfence_pool = page_to_virt(pages);
>> #else
>> -       if (nr_pages > MAX_ORDER_NR_PAGES) {
>> -               pr_warn("KFENCE_NUM_OBJECTS too large for buddy allocator\n");
>> -               return -EINVAL;
>> -       }
>> +       BUILD_BUG_ON_MSG(get_order(KFENCE_POOL_SIZE) > MAX_ORDER,
>> +                        "CONFIG_KFENCE_NUM_OBJECTS is too large for buddy allocator");
>> +
> 
> It's perfectly valid to want to use KFENCE with a very large pool that
> is initialized on boot, and simply sacrifice the ability to initialize
> late.

You are right. I didn’t realize this.

Thanks 

> 
> Nack.