On Wed, Sep 6, 2023, at 10:35, Thomas Zimmermann wrote:
> Rename the fbdev mmap helper fb_pgprotect() to fb_pgprot_device().
> The helper sets VMA page-access flags for framebuffers in device I/O
> memory. The new name follows pgprot_device(), which does the same for
> arbitrary devices.
>
> Also clean up the helper's parameters and return value. Instead of
> the VMA instance, pass the individial parameters separately: existing
> page-access flags, the VMAs start and end addresses and the offset
> in the underlying device memory rsp file. Return the new page-access
> flags. These changes align fb_pgprot_device() closer with pgprot_device.
>
> Signed-off-by: Thomas Zimmermann <tzimmermann@xxxxxxx>
This makes sense as a cleanup, but I'm not sure the new naming is helpful.
The 'pgprot_device' permissions are based on Arm's memory attributes,
which have slightly different behavior for "device", "uncached" and
"writecombine" mappings. I think simply calling this one pgprot_fb()
or fb_pgprot() would be less confusing, since depending on the architecture
it appears to give either uncached or writecombine mappings but not
"device" on the architectures where this is different.
Arnd
