Re: [PATCH v4 2/2] staging: android: Remove BUG from ion_system_heap.c

[Randy Dunlap]

On 8/22/20 2:34 AM, Tomer Samara wrote:
> On Fri, Aug 21, 2020 at 09:25:26AM -0700, Randy Dunlap wrote:
>> On 8/21/20 8:28 AM, Tomer Samara wrote:
>>> Remove BUG() from ion_sytem_heap.c
>>>
>>> this fix the following checkpatch issue:
>>> Avoid crashing the kernel - try using WARN_ON &
>>> recovery code ratherthan BUG() or BUG_ON().
>>>
>>> Signed-off-by: Tomer Samara <tomersamara98@xxxxxxxxx>
>>> ---
>>>  drivers/staging/android/ion/ion_system_heap.c | 2 +-
>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/staging/android/ion/ion_system_heap.c b/drivers/staging/android/ion/ion_system_heap.c
>>> index eac0632ab4e8..00d6154aec34 100644
>>> --- a/drivers/staging/android/ion/ion_system_heap.c
>>> +++ b/drivers/staging/android/ion/ion_system_heap.c
>>> @@ -30,7 +30,7 @@ static int order_to_index(unsigned int order)
>>>  	for (i = 0; i < NUM_ORDERS; i++)
>>>  		if (order == orders[i])
>>>  			return i;
>>> -	BUG();
>>> +	/* This is impossible. */
>>>  	return -1;
>>>  }
>>
>> Hi,
>> Please explain why this is impossible.
>>
>> If some caller calls order_to_index(5), it will return -1, yes?
>>
>> -- 
>> ~Randy
>>
> 
> As Dan Carpenter says here https://lkml.kernel.org/lkml/cover.1597865771.git.tomersamara98@xxxxxxxxx/T/#mc790b91029565b1bb0cb87997b39007d9edb6e04.
> After looking at callers we see that order_to_index called from 2 functions:
> - alloc_buffer_page called from alloc_largest_available which 
>   loop over all legit order nubmers
>   ( Flow:
>    alloc_largest_available-->alloc_buffer_page-->order_to_index
>   )
> 
> - free_buffer_page takes the order using compound_order, which return 0 or
>   the order number for the page, this function has 2 callers too,
>   ion_system_heap_allocate (which called it in case of failure at sg_alloc_table,
>   thus calling from this flow will no casue error) and ion_system_heap_free
>   (which will be called on every sg table in the buffer that allocated good,
>   meaning from this flow also error will not be created).
>   ( Flows:
>    ion_system_heap_free     --> free_buffer_page --> order_to_index
>    ion_system_heap_allocate --> free_buffer_page --> order_to_index
>   )
> 
> Of course if some user will use this function with wrong order number he will be able to get this -1.
> So should I remove this comment and resotre the error checks?

I think so, but that's just an opinion.

> Btw, this is the same reason that I dropped the error check at ion_page_pool_shrink, so should I restore here also?

IMO yes.

Getting rid of BUG()s is a good goal, but usually it's not so easy.

thanks.
-- 
~Randy