>> That doesn't deal with long-lived mappings and changes that I just
>> mentioned in last e-mail - but it does mean that callers of ioremap()
>> don't need to worry ... they can call "blindly".
>
> I see. So you think it is better to enclose page_is_ram() with #ifndef
> CONFIG_IA64 in the patch, so that only ioremap() is used?
Functionally yes - though the aversion to #ifdef in the middle of code
would imply some extra hoops. Perhaps
#ifndef CONFIG_IA64
#define can_use_kmap(pfn) page_is_ram(pfn)
#else
#define can_use_kmap(pfn) (0) /* ioremap() will take care of cache attributes */
#endif
...
if (can_use_kmap(pfn)) {
-Tony
��.n��������+%������w��{.n�����{�����ܨ}���Ơz�j:+v�����w����ޙ��&�)ߡ�a����z�ޗ���ݢj��w�f
