[PATCH] iommu/arm-smmu-v3: Retire disable_bypass parameter

[Jason Gunthorpe]

On Fri, Apr 05, 2024 at 05:52:07PM +0100, Robin Murphy wrote:
> The disable_bypass parameter has been mostly meaningless for a long time
> since the introduction of default domains. Its original intent is now
> fulfilled by the controls users have over the default domain type, and
> its remaining effect in the brief window between Stream Table
> initialisation and default domain creation hardly seems worth the
> complication. Furthermore, thanks to 2-level Stream Tables, disabling
> disable_bypass (there's another reason not to like it right there) has
> never guaranteed that any particular StreamID *will* bypass anyway - any
> device which might actually care about that wants RMRs - so there's not
> really much lost by taking away that option (which has already been
> non-default for nearing 6 years now).
> 
> As part of this, also remove the weird behaviour where we "successfully"
> probe and register a non-functional SMMU if the DT "#iommu-cells"
> property is wrong. I have no memory of what possessed me to think that
> was a good idea at the time, and by now I suspect it's likely to break
> things worse than simply failing probe would.
> 
> Signed-off-by: Robin Murphy <robin.murphy at arm.com>
> ---
>  drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c | 46 ++++++---------------
>  1 file changed, 13 insertions(+), 33 deletions(-)

Reviewed-by: Jason Gunthorpe <jgg at nvidia.com>

Jason