I am unable to convince myself that this is correct. It's playing with pmd
and pte pages which can be freed, reallocated and filled with random stuff.
I really don't see how that can work, but am willing to be taught.
Because we hold i_shared_sem we know that the pgd layer is stable and that
the mm's aren't going away.
Is it not possible to take each mm's page_table_lock? There's a ranking
problem with pte_chain_lock(), but that can presumably be resolved by doing a
trylock on the page_table_lock and if that fails, restart the whole operation.
But then again, why is it not possible to just do:
list_for_each_entry(vma, &mapping->i_mmap, shared) {
if (!pte_chain)
pte_chain = pte_chain_alloc(GFP_KERNEL);
spin_lock(&mm->page_table_lock);
pte = find_pte(vma, page, NULL);
if (pte)
pte_chain = page_add_rmap(page, pte, pte_chain);
spin_unlock(&mm->page_table_lock);
}
pte_chain_free(pte_chain);
up(&mapping->i_shared_sem);
?
-
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/