On Thu, Aug 29, 2019 at 4:38 PM Bartosz Golaszewski [off-list ref] wrote:
quoted hunk ↗ jump to hunk
@@ -710,6 +710,10 @@ extern void devm_free_pages(struct device *dev, unsigned long addr);
void __iomem *devm_ioremap_resource(struct device *dev,
const struct resource *res);
+void __iomem *devm_ioremap_resource_nocache(struct device *dev,
+ const struct resource *res);
+void __iomem *devm_ioremap_resource_wc(struct device *dev,
+ const struct resource *res);
void __iomem *devm_of_iomap(struct device *dev,
struct device_node *node, int index,diff --git a/lib/devres.c b/lib/devres.c
I think adding devm_ioremap_resource_wc() and
devm_platform_ioremap_resource_wc() makes sense, but I think we're
better off without devm_ioremap_resource_nocache() and
devm_ioremap_resource_cache().
The only architecture that actually has a difference between
ioremap() and ioremap_nocache() seems to be ia64. I would
generally assume that any driver using ioremap_nocache()
that is not ia64 specific should just use ioremap().
The ia64 version of ioremap() tries to guess whether it needs
a cached or uncached mapping, everyone else always
gets uncached these days.
Arnd