Re: [PATCH] mm/kasan/kasan.h: Fix boolean checking issue for kasan_report_enabled()

[Chen Gang]

On 5/2/16 19:34, Alexander Potapenko wrote:
> On Mon, May 2, 2016 at 7:36 AM,  <chengang@xxxxxxxxxxxxxxxx> wrote:
>> From: Chen Gang <chengang@xxxxxxxxxxxxxxxx>
>>
>> According to kasan_[dis|en]able_current() comments and the kasan_depth'
>> s initialization, if kasan_depth is zero, it means disable.
> The comments for those functions are really poor, but there's nothing
> there that says kasan_depth==0 disables KASAN.
> Actually, kasan_report_enabled() is currently the only place that
> denotes the semantics of kasan_depth, so it couldn't be wrong.
> 
> init_task.kasan_depth is 1 during bootstrap and is then set to zero by
> kasan_init()
> For every other thread, current->kasan_depth is zero-initialized.
> 

OK, what you said sound reasonable to me. and do you also mean:

 - kasan_depth == 0 means enable KASAN, others means disable KASAN.

 - If always let kasan_[en|dis]able_current() be pair, and notice about
   the overflow, it should be OK: "kasan_enable_current() can let
   kasan_depth++, and kasan_disable_current() will let kasan_depth--".

 - If we check the related overflow, "kasan_depth == 1" mean "the KASAN
   should be always in disable state".


Thanks.
-- 
Chen Gang (éå)

Managing Natural Environments is the Duty of Human Beings.