Re: [PATCH 1/3] crypto: hisilicon/qm - modify the process of regs dfx

[yekai (A)]

On 2022/11/4 18:07, yekai (A) wrote:
>
> On 2022/11/4 17:05, Herbert Xu wrote:
>> On Sat, Oct 29, 2022 at 09:47:59AM +0000, Kai Ye wrote:
>>> diff --git a/drivers/crypto/hisilicon/qm.c b/drivers/crypto/hisilicon/qm.c
>>> index 363a02810a16..832cfd9a7728 100644
>>> --- a/drivers/crypto/hisilicon/qm.c
>>> +++ b/drivers/crypto/hisilicon/qm.c
>>> @@ -1722,8 +1722,22 @@ static int qm_regs_show(struct seq_file *s, void *unused)
>>>  
>>>  DEFINE_SHOW_ATTRIBUTE(qm_regs);
>>>  
>>> +static void dfx_regs_uninit(struct hisi_qm *qm,
>>> +		struct dfx_diff_registers *dregs, int reg_len)
>>> +{
>>> +	int i;
>>> +
>>> +	/* Setting the pointer is NULL to prevent double free */
>>> +	for (i = 0; i < reg_len; i++) {
>>> +		kfree(dregs[i].regs);
>>> +		dregs[i].regs = NULL;
>>> +	}
>>> +	kfree(dregs);
>>> +	dregs = NULL;
>> What's the point of this last NULL assignment?
>>
>> Cheers,
> To prevent pointer reuse and reduce register data security risks. So the original code setting the pointer is NULL.
>  
> Thanks
> Kai
I misunderstood the pointer. There is no security data in the pointer. It is only a value transferred as a structure.
This patch-set does not modify the logic. Can i add an extra patch to delete this redundant line next version?

Thanks
Kai