Re: [PATCH v1 1/2] ACPI: scan: Rearrange memory allocation in acpi_device_add()

[Hans de Goede <hdegoede@redhat.com>]

Hi,

On 1/18/21 4:32 PM, Andy Shevchenko wrote:
> On Mon, Jan 18, 2021 at 04:16:16PM +0100, Rafael J. Wysocki wrote:
>> On Sat, Jan 16, 2021 at 1:37 PM Hans de Goede <hdegoede@redhat.com> wrote:
>>> On 1/14/21 7:46 PM, Rafael J. Wysocki wrote:
> 
> ...
> 
>>> When I have cases like this, where 2 mallocs are necessary I typically do it like this:
>>>
>>>         const char *bus_id;
>>>
>>>         ...
>>>
>>>         } else {
>>>                 acpi_device_bus_id = kzalloc(sizeof(*acpi_device_bus_id),
>>>                                              GFP_KERNEL);
>>>                 bus_id = kstrdup_const(acpi_device_hid(device), GFP_KERNEL);
>>>                 if (!acpi_device_bus_id || !bus_id) {
>>>                         kfree(acpi_device_bus_id);
> 
> 
>>>                         kfree(bus_id);
> 
> Just to be sure, shouldn't it be kfree_const() ?

Yes I beleive it should, my bad.

Regards,

Hans


> 
>>>                         result = -ENOMEM;
>>>                         goto err_unlock;
>>>                 }
>>>                 acpi_device_bus_id->bus_id = bus_id;
>>>                 list_add_tail(&acpi_device_bus_id->node, &acpi_bus_id_list);
>>>         }
>>>
>>>         ...
>>>
>>> So that there is only one if / 1 error-handling path for both mallocs.
>>> I personally find this a bit cleaner.
>