[PATCH 3/4] block: add IOC_PR_READ_KEYS ioctl

[Krzysztof Kozlowski]

On 01/12/2025 16:14, Stefan Hajnoczi wrote:
> On Sat, Nov 29, 2025 at 03:32:35PM +0100, Krzysztof Kozlowski wrote:
>> On 27/11/2025 08:07, Hannes Reinecke wrote:
>>>
>>>> +	size_t keys_info_len = struct_size(keys_info, keys, inout.num_keys);
>>>> +
>>>> +	keys_info = kzalloc(keys_info_len, GFP_KERNEL);
>>>> +	if (!keys_info)
>>>> +		return -ENOMEM;
>>>> +
>>>> +	keys_info->num_keys = inout.num_keys;
>>>> +
>>>> +	ret = ops->pr_read_keys(bdev, keys_info);
>>>> +	if (ret)
>>>> +		return ret;
>>>> +
>>>> +	/* Copy out individual keys */
>>>> +	u64 __user *keys_ptr = u64_to_user_ptr(inout.keys_ptr);
>>>> +	u32 num_copy_keys = min(inout.num_keys, keys_info->num_keys);
>>>> +	size_t keys_copy_len = num_copy_keys * sizeof(keys_info->keys[0]);
>>>
>>> We just had the discussion about variable declarations on the ksummit 
>>> lists; I really would prefer to have all declarations at the start of 
>>> the scope (read: at the start of the function here).
>>
>> Then also cleanup.h should not be used here.
> 
> Hi Krzysztof,
> Christoph Hellwig replied to the v2 series, also against using __free().

That's perfectly fine to dislike cleanup.h. It's fair. What is not fine
is using it against its recommendations. Either you take entire
cleanup.h with its oddities or don't use it.

> Regardless of the reply I just sent to you about whether cleanup.h may
> or may not be used in code that forbids declarations midway through a
> scope, I will be dropping it in v3.
> 

Best regards,
Krzysztof