On Mon, Oct 28 2002, David S. Miller wrote:
> On Mon, 2002-10-28 at 07:08, Jens Axboe wrote:
> > Ok what I make of this is that from bio_map_user() (which does a
> > get_user_pages() I need to do a
> >
> > if (write_to_vm)
> > flush_dcache_page(page);
> >
> > and in bio_unmap_user() I do
> >
> > if (!write_to_vm)
> > flush_dcache_page(page);
> >
> > is that correct?
>
> Ho hum, it is tricky :-)))
>
> At bio_map_user() you need to see the user's most recent write
> to the page if you are going "user --> device". So if "user
> --> device" bio_map_user() must flush_dcache_page().
Yes, that that is the
if (write_to_vm)
flush_dcache_page(page);
> I find the write_to_vm condition confusion which is probably why
> I am sitting here spelling this out :-)
Hehe, actually these are confusing, but write_to_vm is probably the
least confusing wording I think. It means we are reading from the
device, thus writing to vm pages.
> At bio_unmap_user(), if we are going "device --> user" you have
> to flush_dcache_page(). And actually, this flush could just as
> legitimately occur at bio_map_user() time.
Thus
if (!write_to_vm)
flush_dcache_page(page);
> Therefore, the easiest thing to do is always flush_dcache_page()
> at bio_map_user().
>
> All the other cases are going to be like this, so we might as
> well cut to the chase and flush_dcache_page() for all the pages
> inside of get_user_pages().
>
> Whoever made get_user_pages() and didn't carry over the
> flush_dcache_page calls from the mechanism it is meant to replace
> should be spanked :-)
I agree then. I will remove references to flush_dcache_page() in
bio_map_user() and bio_unmap_user(), and we'll just unconditionally do
flush_dcache_page() on every page mapped in get_user_pages(). That
sounds far better to me. Do you really expect users of get_user_pages()
to get this right (remember, we are often talking about device drivers
:-). I sure am not :)
-- Jens Axboe- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Thu Oct 31 2002 - 22:00:37 EST