Re: [PATCH v2] kasan: remove kasan_find_vm_area() to prevent possible deadlock

[Andrey Ryabinin <ryabinin.a.a@gmail.com>]

On 7/3/25 9:05 PM, Andrey Konovalov wrote:
> On Thu, Jul 3, 2025 at 8:55 PM Yeoreum Yun <yeoreum.yun@arm.com> wrote:
>>
>> Hi Andrey,
>>
>>>>
>>>> find_vm_area() couldn't be called in atomic_context.
>>>> If find_vm_area() is called to reports vm area information,
>>>> kasan can trigger deadlock like:
>>>>
>>>> CPU0                                CPU1
>>>> vmalloc();
>>>>  alloc_vmap_area();
>>>>   spin_lock(&vn->busy.lock)
>>>>                                     spin_lock_bh(&some_lock);
>>>>    <interrupt occurs>
>>>>    <in softirq>
>>>>    spin_lock(&some_lock);
>>>>                                     <access invalid address>
>>>>                                     kasan_report();
>>>>                                      print_report();
>>>>                                       print_address_description();
>>>>                                        kasan_find_vm_area();
>>>>                                         find_vm_area();
>>>>                                          spin_lock(&vn->busy.lock) // deadlock!
>>>>
>>>> To prevent possible deadlock while kasan reports, remove kasan_find_vm_area().
>>>
>>> Can we keep it for when we are in_task()?
>>
>> We couldn't do. since when kasan_find_vm_area() is called,
>> the report_lock is grabbed with irq disabled.
>>
>> Please check discuss with Andrey Ryabinin:
>>   https://lore.kernel.org/all/4599f645-f79c-4cce-b686-494428bb9e2a@gmail.com/
> 
> That was about checking for !in_interrupt(), but I believe checking
> for in_task() is different? But I'm not an expert on these checks.

The problem is that CPU1 grabs '&vn->busy.lock' after the '&some_lock'. This could
happen both in task and in irq contexts, so the in_task() guard just won't change anything.