Re: [RFC][PATCH 1/3] radix priority search tree - objrmap complexity fix

[Andrea Arcangeli]

On Sat, Apr 03, 2004 at 12:02:27PM -0800, Andrew Morton wrote:
> It might be better to switch over to address masking in get_user_pages()
> and just dump all the compound page logic.  I don't immediately see how the

I'm all for it, this is how the 2.4 get_user_pages deals with bigpages
too, I've never enjoyed the compound thing.

> get_user_pages() caller can subsequently do put_page() against the correct
> pageframe, but I assume you worked that out?

see this patch:

	http://www.us.kernel.org/pub/linux/kernel/people/andrea/kernels/v2.4/2.4.23aa2/9910_shm-largepage-18.gz

it's a two liner fix in follow_page:

@@ -439,6 +457,8 @@ static struct page * follow_page(struct
        pmd = pmd_offset(pgd, address);
        if (pmd_none(*pmd))
                goto out;
+       if (pmd_bigpage(*pmd))
+               return __pmd_page(*pmd) + (address & BIGPAGE_MASK) / PAGE_SIZE;


the BIGPAGE_MASK will never expose anything but the page->private to the
get_user_pages code, and handle_mm_fault takes care of doing the page
faults properly using larepages and pmds if the vma is marked
VM_BIGPAGE.

rawio on largepages has been a must-have feature (especially on >=32G)
for more than one year, definitely no need of compound slowdown for that.

Still I would like to understand what's wrong in Christoph's ppc machine
before dumping the whole compound thing.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/