On Wed, Aug 20, 2025 at 3:19 AM Daniel Baluta <daniel.baluta@xxxxxxxxx> wrote:
>
> Hi Rob,
>
> This patch will break IMX RPROC support.
> <snip>
>
> > --- a/drivers/remoteproc/imx_rproc.c
> > +++ b/drivers/remoteproc/imx_rproc.c
>
> <snip>
> > static int imx_rproc_parse_fw(struct rproc *rproc, const struct firmware *fw)
> <snip>
>
> > - priv->mem[b].cpu_addr = devm_ioremap_wc(&pdev->dev, res.start, resource_size(&res));
> > + priv->mem[b].cpu_addr = devm_ioremap_resource_wc(&pdev->dev, &res);
>
> devm_ioremap_resource_wc works only for IORESOURCE_MEM resources:
>
> lib/devres.c:124
> static void __iomem *
> __devm_ioremap_resource(struct device *dev, const struct resource *res,
>
> » if (!res || resource_type(res) != IORESOURCE_MEM) {
> » » ret = dev_err_probe(dev, -EINVAL, "invalid resource
> %pR\n", res);
> » » return IOMEM_ERR_PTR(ret);
> » }
>
> while the devm_ioremap_wc doesn't care about this.
>
> So we cannot use devm_ioremap_resource_wc here unless you add
> IORESOURCE_MEM flags
> to of_reserved_mem_region_to_resource as discussed here:
>
> https://lkml.org/lkml/2025/4/28/759
>
> The same issue we are already experiencing with Sound Open Firmware
> where the change was already merged
> and we have a bug already reported.
>
> How should we fix this:
>
> 1) Add res->flags = IORESOURCE_MEM; in of_reserved_mem_region_to_resource
Sigh. I thought I rolled that change into it. I just sent a fix[1].
Rob
[1] https://lore.kernel.org/all/20250820192805.565568-1-robh@xxxxxxxxxx/
