On 3/5/26 5:10 AM, David Hoppenbrouwers wrote:
> On 3/2/26 6:34 AM, Sairaj Kodilkar wrote:
>>
>>
>> On 2/25/2026 8:28 PM, David Hoppenbrouwers wrote:
>>
>> @David Could you please test above patch ?
>
> It appears to work fine with both the Linux VM and my own kernel project.
>
Thank you for reporting and testing this bug fix David. Like Sairaj I have
not been able to reproduce it but it is clear there is an issue.
>> I have rearraged some part of code. Along with it I have added a
>> requirement that caller
>> must validate DTE before calling fetch_pte. This requirement simplifies
>> the loop.
>>
>> If @David and @Alejandro are okay with this change, I can prepare and
>> send official patch.
>
> It is fine with me. I don't need to send a v2?
>
No need for now, though I would appreciate your review and testing on the
next round of patches. I am working on additional changes, essentially
adding to Sairaj's observation/approach that we are better off following
the Page Table Root pointer outside of the loop, so that we end up
iterating simply on PDEs/PTE. I've been delayed by other work, but I wanted
to ack your question.
Thank you,
Alejandro
>> Thanks
>> Sairaj
>
