On Wed, Nov 8, 2017 at 6:55 AM, Greentime Hu [off-list ref] wrote:

+
+#define ioremap(cookie,size)           __ioremap(cookie,size,0,1)
+#define ioremap_nocache(cookie,size)   __ioremap(cookie,size,0,1)
+#define iounmap(cookie)                        __iounmap(cookie)

+#include <asm-generic/io.h>

asm-generic/io.h now provides an ioremap_nocache() helper along with
ioremap_uc/ioremap_wc/ioremap_wt, so I think you can remove the
ioremap_nocache definition here. You might also be able to remove
__ioremap and __iounmap, and only provide ioremap/iounmap, plus
the identity macro 'define ioremap ioremap'

+void __iomem *__ioremap(unsigned long phys_addr, size_t size,
+                       unsigned long flags, unsigned long align)

The 'align' argument is unused here, and not used on other architectures
either.

+{
+       struct vm_struct *area;
+       unsigned long addr, offset, last_addr;
+       pgprot_t prot;
+
+       /* Don't allow wraparound or zero size */
+       last_addr = phys_addr + size - 1;
+       if (!size || last_addr < phys_addr)
+               return NULL;
+
+       /*
+        * Mappings have to be page-aligned
+        */
+       offset = phys_addr & ~PAGE_MASK;
+       phys_addr &= PAGE_MASK;
+       size = PAGE_ALIGN(last_addr + 1) - phys_addr;
+
+       /*
+        * Ok, go for it..
+        */
+       area = get_vm_area(size, VM_IOREMAP);

Better use get_vm_area_caller here to have the ioremap areas show up
in a more useful form in /proc/vmallocinfo

Please also have a look at what you can do for memremap().

Since you have no cacheable version of ioremap_wb/wt, it will
return an uncached mapping all the time, which is not ideal.

     Arnd