Re: [PATCH v14 6/6] iommu: Remove mode argument from iommu_set_dma_strict()

[Lu Baolu]

On 2021/6/21 16:12, John Garry wrote:
> On 21/06/2021 06:17, Lu Baolu wrote:
> > On 2021/6/18 19:34, John Garry wrote:
> > > diff --git a/drivers/iommu/iommu.c b/drivers/iommu/iommu.c
> > > index 60b1ec42e73b..ff221d3ddcbc 100644
> > > --- a/drivers/iommu/iommu.c
> > > +++ b/drivers/iommu/iommu.c
> > > @@ -349,10 +349,9 @@ static int __init iommu_dma_setup(char *str)
> > >   }
> > >   early_param("iommu.strict", iommu_dma_setup);
> > > -void iommu_set_dma_strict(bool strict)
> > > +void iommu_set_dma_strict(void)
> > >   {
> > > -    if (strict || !(iommu_cmd_line & IOMMU_CMD_LINE_STRICT))
> > > -        iommu_dma_strict = strict;
> > > +    iommu_dma_strict = true;
> > >   }
>
> Hi baolu,

Hi John,

> > Sorry for this late comment.
> >  > Normally the cache invalidation policy should come from the user. We
> > have pre-build kernel option and also a kernel boot command iommu.strict
> > to override it. These seem reasonable.
> >
> > We also have a helper (iommu_set_dma_strict()) so that the vendor iommu
> > driver could squeeze in and change the previous settings mostly due to:
> >
> > a) vendor iommu driver specific kernel boot command. (We are about to
> >     deprecate those.)
> >
> > b) quirky hardware.
> >
> > c) kernel optimization (e.x. strict mode in VM environment).
> >
> > a) and b) are mandatory, while c) is optional. In any instance should c)
> > override the flush mode specified by the user. Hence, probably we should
> > also have another helper like:
> >
> > void iommu_set_dma_strict_optional()
> > {
> >      if (!(iommu_cmd_line & IOMMU_CMD_LINE_STRICT))
> >          iommu_dma_strict = true;
> > }
> >
> > Any thoughts?
>
> What you are suggesting is a change in policy from mainline code. 
> Currently for c) we always set strict enabled, regardless of any user 
> cmdline input. But now you are saying that you want iommu.strict to 
> override in particular scenario, right?
>
> In that case I would think it's better to rework the current API, like 
> adding an option to "force" strict mode:
>
> void iommu_set_dma_strict(bool force)
> {
>           if (force == true)
>          iommu_dma_strict = true;
>      else if (!(iommu_cmd_line & IOMMU_CMD_LINE_STRICT))
>          iommu_dma_strict = true;
> }
>
> So we would use iommu_set_dma_strict(true) for a) and b), but 
> iommu_set_dma_strict(false) for c).

Yes. We need to distinguish the "must" and "nice-to-have" cases of
setting strict mode.

> Then I am not sure what you want to do with the accompanying print for 
> c). It was:
> "IOMMU batching is disabled due to virtualization"
>
> And now is from this series:
> "IOMMU batching disallowed due to virtualization"
>
> Using iommu_get_dma_strict(domain) is not appropriate here to know the 
> current mode (so we know whether to print).
>
> Note that this change would mean that the current series would require 
> non-trivial rework, which would be unfortunate so late in the cycle.

This patch series looks good to me and I have added by reviewed-by.
Probably we could make another patch series to improve it so that the
kernel optimization should not override the user setting.

Best regards,
baolu