On Mon, Apr 17, 2023, at 11:03, Thomas Zimmermann wrote:

Am 11.04.23 um 10:08 schrieb Geert Uytterhoeven:

quoted

On Thu, Apr 6, 2023 at 4:30 PM Thomas Zimmermann [off-list ref] wrote:

quoted

quoted

+#ifndef fb_pgprotect
+#define fb_pgprotect fb_pgprotect
+static inline void fb_pgprotect(struct file *file, struct vm_area_struct *vma,
+                               unsigned long off)

Does this affect any noMMU platforms that relied on fb_pgprotect()
doing nothing before?
Perhaps the body below should be protected by "#ifdef CONFIG_MMU"?

I cannot conclusively answer this question, but I did some grep'ing 
('git grep ndef | grep CONFIG_MMU'):

Only the architectures in this patchset provide <asm/fb.h> but nothing 
anywhere uses <asm-generic/fb.h> yet. And of those architectures, only 
arm and m68k have !CONFIG_MMU cases. Those are handled in the rsp 
patches. I think we're good.

Agreed. The generic version is just a more elaborate way to do
nothing here, as the 

     vma->vm_page_prot = pgprot_writecombine(vma->vm_page_prot);

line on nommu just turns into a self-assignment of the same member
that was set the line before.

     Arnd