Re: [PATCH v1] iommu: Skip mapping at address 0x0 if it already exists

[Vasant Hegde <vasant.hegde@amd.com>]

Hi Antheas,

On 2/23/2026 1:16 PM, Antheas Kapenekakis wrote:
> [You don't often get email from lkml@antheas.dev. Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ]
> 
> On Mon, 23 Feb 2026 at 07:02, Vasant Hegde <vasant.hegde@amd.com> wrote:
>>
>> Antheas,
>>
>> On 2/22/2026 5:20 AM, Antheas Kapenekakis wrote:
>>> Commit 789a5913b29c ("iommu/amd: Use the generic iommu page table")
>>> introduces the shared iommu page table for AMD IOMMU. Some bioses
>>> contain an identity mapping for address 0x0, which is not parsed
>>> properly (e.g., certain Strix Halo devices). This causes the DMA
>>> components of the device to fail to initialize (e.g., the NVMe SSD
>>> controller), leading to a failed post.
>>>
>>> The failure is caused by iommu_create_device_direct_mappings(), which
>>> is the new mapping implementation. In it, address aliasing is handled
>>> via the following check:
>>>
>>> ```
>>> phys_addr = iommu_iova_to_phys(domain, addr);
>>> if (!phys_addr) {
>>>         map_size += pg_size;
>>>         continue;
>>> }
>>> ````
>>
>> Thanks for debugging and fixing it . Just wondering why can't we replace replace
>> above check with pfn_valid() ?
>>
> 
> Hi Vasant,
> I can check later today
> 
> From a cursory glance though, it does not seem very suitable... Can you expand?

Looking into the details, you are right. It won't work for direct_mapping().


May be we can have this patch for now, but I think on long run we should update
iova_to_phys() ops.

-Vasant