[Question] Should retain 2M alignment if kernel image is bad aligned at entry or BSS overlaps?

[Kefeng Wang]

On 2022/3/3 19:31, Ard Biesheuvel wrote:
> On Thu, 3 Mar 2022 at 06:14, Kefeng Wang <wangkefeng.wang at huawei.com> wrote:
>>
>> On 2022/3/1 20:57, Ard Biesheuvel wrote:
>>> On Tue, 1 Mar 2022 at 11:34, Kefeng Wang <wangkefeng.wang at huawei.com> wrote:
> ...
>>>> Do you think the following adjustment make sense or it is definitely wrong?
>>>>
>>> I can only answer this if I understand which problem it solves. Why do
>>> you need the 2M alignment in this case?
>>> .
>> Sorry for the late response，my purpose is that we don't want to enable
>> KPTI if
>>
>> KASLR is disabled. For now, if there are some firmware bug, the kernel
>> image is
>>
>> relocated which lead to kaslr_requires_kpti() returen
>> ture(kaslr_offset() > 0).
>>
>> the change to 2M alignment is a workaround and according to your
>> explanation,
>>
>> I don't think the workaround is necessary.  I want to make sure that the
>> above
>>
>> scene is expected? thanks.
>>
> I don't think we need this.
Got it, thanks.
> .