On Sun, 2013-12-01 at 22:18 +0100, Helge Deller wrote:
> This is now the next try to fix the aio bug in our kmap/kunmap
> implementation. It's basically stuff from
I thought the implementation required us to move to coherency management
in the kmap API?, in which case ...
[...]
> -void kunmap_parisc(void *addr);
> -
> static inline void *kmap(struct page *page)
> {
> might_sleep();
> + /* Flush dcache page on CPUs that require coherency. */
> + if (parisc_requires_coherency())
> + flush_dcache_page(page);
> return page_address(page);
> }
We have to call flush_dcache_page() unconditionally here. It's the API
that coheres all the user space mappings. It has to be called somehow
before the kernel begins operating on the mapped address.
> static inline void kunmap(struct page *page)
> {
> - kunmap_parisc(page_address(page));
> + /* Always flush kernel dcache, even if CPU doesn't require conherency.
> + * Needed on CPUs < PA8800/PA8900 for AIO support. */
> + flush_kernel_dcache_page_addr(page_address(page));
> }
>
> static inline void *kmap_atomic(struct page *page)
> {
> pagefault_disable();
> + /* Flush dcache page on CPUs that require coherency. */
> + if (parisc_requires_coherency())
> + flush_dcache_page(page);
> return page_address(page);
> }
And unconditionally here too.
And if we do that, I don't see that we need special code for the
pa88/8900 any more.
James
--
To unsubscribe from this list: send the line "unsubscribe linux-parisc" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
