Re: [PATCH v16 3/3] of: Respect #{iommu,msi}-cells in maps

[Vijayanand Jitta <vijayanand.jitta@xxxxxxxxxxxxxxxx>]

On 7/28/2026 3:34 PM, Robin Murphy wrote:
> On 28/07/2026 5:05 am, Vijayanand Jitta wrote:
>>
>>
>> On 7/23/2026 6:47 PM, Neil Armstrong wrote:
>>> Hi,
>>>
>>> On 6/3/26 09:13, Vijayanand Jitta wrote:
>>>> From: Robin Murphy <robin.murphy@xxxxxxx>
>>>>
>>>> So far our parsing of {iommu,msi}-map properties has always blindly
>>>> assumed that the output specifiers will always have exactly 1 cell.
>>>> This typically does happen to be the case, but is not actually enforced
>>>> (and the PCI msi-map binding even explicitly states support for 0 or 1
>>>> cells) - as a result we've now ended up with dodgy DTs out in the field
>>>> which depend on this behaviour to map a 1-cell specifier for a 2-cell
>>>> provider, despite that being bogus per the bindings themselves.
>>>>
>>>> Since there is some potential use in being able to map at least single
>>>> input IDs to multi-cell output specifiers (and properly support 0-cell
>>>> outputs as well), add support for properly parsing and using the target
>>>> nodes' #cells values, albeit with the unfortunate complication of still
>>>> having to work around expectations of the old behaviour too.
>>>>
>>>> Since there are multi-cell output specifiers, the callers of of_map_id()
>>>> may need to get the exact cell output value for further processing.
>>>> Update of_map_id() to set args_count in the output to reflect the actual
>>>> number of output specifier cells.
>>>>
>>>> Signed-off-by: Robin Murphy <robin.murphy@xxxxxxx>
>>>> Signed-off-by: Charan Teja Kalla <charan.kalla@xxxxxxxxxxxxxxxx>
>>>> Signed-off-by: Vijayanand Jitta <vijayanand.jitta@xxxxxxxxxxxxxxxx>
>>>> ---
>>>>    drivers/of/base.c  | 168 
>>>> +++++++++++++++++++++++++++++++++++++++++------------
>>>>    include/linux/of.h |   6 +-
>>>>    2 files changed, 135 insertions(+), 39 deletions(-)
>>>>
>>>> diff --git a/drivers/of/base.c b/drivers/of/base.c
>>>> index d658c2620135..ac7961cbab94 100644
>>>> --- a/drivers/of/base.c
>>>> +++ b/drivers/of/base.c
>>>> @@ -2116,19 +2116,49 @@ int of_find_last_cache_level(unsigned int cpu)
>>>>        return cache_level;
>>>>    }
>>>>    +/*
>>>> + * Some DTs have an iommu-map targeting a 2-cell IOMMU node while
>>>> + * specifying only 1 cell. Fortunately they all consist of value '1'
>>>> + * as the 2nd cell entry with the same target, so check for that pattern.
>>>> + *
>>>> + * Example:
>>>> + *    IOMMU node:
>>>> + *        #iommu-cells = <2>;
>>>> + *
>>>> + *    Device node:
>>>> + *        iommu-map = <0x0000 &smmu 0x0000 0x1>,
>>>> + *                <0x0100 &smmu 0x0100 0x1>;
>>>
>>> So the sm8650 PCIe controllers has:
>>>
>>> pcie@1c08000:
>>>              iommu-map = <0     &apps_smmu 0x1480 0x1>,
>>>                      <0x100 &apps_smmu 0x1481 0x1>;
>>>
>>> and
>>>
>>> pcie@1c00000:
>>>
>>>              iommu-map = <0     &apps_smmu 0x1400 0x1>,
>>>                      <0x100 &apps_smmu 0x1401 0x1>;
>>>
>>> and apps_smmu has #iommu-cells = <2>, but gets flagged at wrong:
>>>
>>> [    7.538800] OF: /soc@0/pcie@1c08000: iommu-map has 1-cell entries 
>>> targeting 2-cell #iommu-cells, treating as 1-cell output
>>>
>>> Returning false in of_check_bad_map() triggers:
>>>
>>> [    7.642680] OF: /soc@0/pcie@1c08000: Unsupported iommu-map - cannot 
>>> handle 256-ID range with 2-cell output specifier
>>>
>>> I don't understand the issue here, we use 2 cells as expected by
>>> the iommu-cells, so why is it wrong ? can somebody explain in
>>> comprehensive words ? I'm super confused, it worked like a charm until now.
>>>
>>> Neil
>>>
>>
>> Hi Neil,
>>
>> iommu-map = <0     &apps_smmu 0x1480 0x1>,
>>              <0x100 &apps_smmu 0x1481 0x1>;
>>
>>
>> Entries here are not 2-cell format, Even though apps_smmu declares 
>> #iommu-cells = <2>,
>> this DT only supplies one output cell (0x1480/0x1481) — the trailing 0x1 is 
>> the length field,
>> not a second output cell. (<id-base phandle out-base length>)
>>
>> The new code detects exactly this pattern (same target phandle across all 
>> entries, length always 1)
>> and falls back to treating the map as 1-cell output for backward 
>> compatibility — hence the pr_warn_once.
>> It's harmless and expected, your RIDs still resolve to the correct SIDs (0 → 
>> 0x1480, 0x100 → 0x1481).
>>
>> The second message is a different case and shouldn't be coming from this 
>> same map — once the 1-cell
>> fallback triggers on the first entry, it applies to the whole map, so you 
>> shouldn't hit both warnings
>> together on the same node. That error only fires for a genuine 2-cell output 
>> specifier combined with
>> an id_len > 1, e.g.:
>>
>> iommu-map = <0x0 &apps_smmu 0x1480 0x1 0x100>;
>>
>> (<id-base, phandle, out0, out1, length=256>) — which isn't supported, since 
>> there's no way to
>> linearly scale a multi-cell output specifier across a range of IDs.
>>
>> Are you seeing that second error on the same pcie node, or a different one?
>> If it's the same node, can you share the exact iommu-map entry that triggers 
>> it?
> 
> I think Neil is saying he bypassed the fallback check so that it *did* try to 
> parse the given map with the real #iommu-cells=2 in precisely the way you've 
> shown - so even if it could have got past that point, it would have then 
> blown trying to parse the second "entry" of just <&apps_smmu 0x1481 0x1>, 
> since 0x1481 almost certainly isn't a valid phandle to read an #iommu-cells 
> value from.
> 
> Cheers,
> Robin.

Right, I get it now, so when it tried to parse with iommu-cells as 2,

iommu-map = <0     &apps_smmu 0x1480 0x1>,
            <0x100 &apps_smmu 0x1481 0x1>;

Above tuples would look something like <0  &apps_smmu 0x1480 0x1 0x100>, where 
0x100 from next tuple would be seen as length. Hence, the above error log.
And the next tuple <&apps_smmu 0x1481 0x1> won't be able to get parsed as you 
mentioned.


Thanks,
Vijay