Re: [PATCH] iommu/arm-smmu-v3: Fix soft lockup triggered by arm_smmu_mm_invalidate_range

[Nicolin Chen <nicolinc@nvidia.com>]

On Mon, Sep 18, 2023 at 10:22:36AM +0100, Will Deacon wrote:
> > @@ -201,9 +201,14 @@ static void arm_smmu_mm_invalidate_range(struct mmu_notifier *mn,
> >        */
> >       size = end - start;
> >
> > -     if (!(smmu_domain->smmu->features & ARM_SMMU_FEAT_BTM))
> > -             arm_smmu_tlb_inv_range_asid(start, size, smmu_mn->cd->asid,
> > -                                         PAGE_SIZE, false, smmu_domain);
> > +     if (!(smmu_domain->smmu->features & ARM_SMMU_FEAT_BTM)) {
> > +             if (!(smmu_domain->smmu->features & ARM_SMMU_FEAT_RANGE_INV) &&
> > +                 size >= CMDQ_MAX_TLBI_OPS * PAGE_SIZE)
> > +                     arm_smmu_tlb_inv_asid(smmu_domain->smmu, smmu_mn->cd->asid);
> > +             else
> > +                     arm_smmu_tlb_inv_range_asid(start, size, smmu_mn->cd->asid,
> > +                                                 PAGE_SIZE, false, smmu_domain);
> 
> cosmetic nit: Please use braces for the multi-line conditionals.

Ack.

> > +     }
> >       arm_smmu_atc_inv_domain(smmu_domain, mm->pasid, start, size);
> >  }
> >
> > diff --git a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h
> > index dcab85698a4e..79a81eed1dcc 100644
> > --- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h
> > +++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h
> > @@ -326,6 +326,15 @@
> >   */
> >  #define CMDQ_BATCH_ENTRIES           BITS_PER_LONG
> >
> > +/*
> > + * Cloned from the MAX_TLBI_OPS in arch/arm64/include/asm/tlbflush.h, this
> > + * is used as a threshold to replace per-page TLBI commands to issue in the
> > + * command queue with an address-space TLBI command, when SMMU w/o a range
> > + * invalidation feature handles too many per-page TLBI commands, which will
> > + * otherwise result in a soft lockup.
> > + */
> > +#define CMDQ_MAX_TLBI_OPS            (1 << (PAGE_SHIFT - 3))
> 
> Maybe stick "SVA" in the name of this somewhere, since that's the reason why
> looking at PAGE_SHIFT is relevant?

Hmm, that does make sense, yet it wouldn't apply to the non-SVA
pathway, which makes putting it in the common header meaningless.
Perhaps I should have just left it in arm-smmu-v3-sva.c file...

Meanwhile, we'd need to figure out another definition for non-SVA
pathway.

Thanks
Nicolin