[PATCH v11 8/9] iommu/arm-smmu-v3: Add in-kernel support for NVIDIA Tegra241 (Grace) CMDQV

[Jason Gunthorpe]

On Tue, Aug 06, 2024 at 07:11:53PM -0700, Nicolin Chen wrote:

> +static struct arm_smmu_device *
> +tegra241_cmdqv_probe(struct arm_smmu_device *smmu,
> +		     struct resource *res, int irq)
> +{
> +	struct tegra241_cmdqv *cmdqv = NULL;
> +	struct tegra241_vintf *vintf;
> +	void __iomem *base;
> +	u32 regval;
> +	int lidx;
> +	int ret;
> +
> +	base = ioremap(res->start, resource_size(res));
> +	if (IS_ERR(base)) {
> +		dev_err(smmu->dev, "failed to ioremap: %ld\n", PTR_ERR(base));
> +		goto iounmap;
> +	}
> +
> +	regval = readl(base + TEGRA241_CMDQV_CONFIG);
> +	if (disable_cmdqv) {
> +		dev_info(smmu->dev, "Detected disable_cmdqv=true\n");
> +		writel(regval & ~CMDQV_EN, base + TEGRA241_CMDQV_CONFIG);
> +		goto iounmap;
> +	}
> +
> +	cmdqv = devm_krealloc(smmu->dev, smmu, sizeof(*cmdqv), GFP_KERNEL);
> +	if (!cmdqv)
> +		goto iounmap;
> +	smmu = &cmdqv->smmu;

Should stick a:

static_assert(offsetof(struct tegra241_cmdqv,smmu) == 0);

In here. The copy inside the krealloc won't work otherwise.

But this still seems Ok to me with the new ops

Jason