Hi,
On 10/26/20 4:39 PM, Limonciello, Mario wrote:
>> This was present in previous versions too, but I just noticed this are you
>> sure that using
>> .string.pointer is correct here? That seems wrong since the pointer gets
>> allocated by
>> the Linux ACPI core, so it is not under influence of th
> This was present in previous versions too, but I just noticed this are you
> sure that using
> .string.pointer is correct here? That seems wrong since the pointer gets
> allocated by
> the Linux ACPI core, so it is not under influence of the AML code?
>
> I think you want / need to use ".integer
Hi,
On 10/26/20 4:25 PM, Limonciello, Mario wrote:
>>> +
>>> + print_hex_dump_bytes("set attribute data: ", DUMP_PREFIX_NONE, buffer,
>> buffer_size);
>>
>> This seems to be a debugging left-over?
>
> Yes it was for debugging, but its configurable to turn on by dynamic
> debug as I can tell. I
> > +
> > + print_hex_dump_bytes("set attribute data: ", DUMP_PREFIX_NONE, buffer,
> buffer_size);
>
> This seems to be a debugging left-over?
Yes it was for debugging, but its configurable to turn on by dynamic
debug as I can tell. Is that not correct?
Hi,
Full review below. Almost there, just a few small items remaining,
see remarks inline.
The most important 2 issues to fix are:
1. Rework the new populate_string_buffer function a bitl and add
a new calculate_string_buffer helper.
2. Drop the !capable(CAP_SYS_ADMIN) checks, instead make t
5 matches
Mail list logo