Commit 4604393c authored Sep 03, 2020 by Robin Murphy Committed by Joerg Roedel Sep 04, 2020

iommu/dma: Remove broken huge page handling



The attempt to handle huge page allocations was originally added since
the comments around stripping __GFP_COMP in other implementations were
nonsensical, and we naively assumed that split_huge_page() could simply
be called equivalently to split_page(). It turns out that this doesn't
actually work correctly, so just get rid of it - there's little point
going to the effort of allocating huge pages if we're only going to
split them anyway.

Reported-by: Roman Gushchin <guro@fb.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/e287dbe69aa0933abafd97c80631940fd188ddd1.1599132844.git.robin.murphy@arm.com


Signed-off-by: Joerg Roedel <jroedel@suse.de>

parent aae4c8e2

drivers/iommu/dma-iommu.c

+5 −8

Original line number	Diff line number	Diff line
		@@ -524,6 +524,9 @@ static struct page *__iommu_dma_alloc_pages(struct device dev,
		/* IOMMU can map any pages, so himem can also be used here */
		gfp \|= __GFP_NOWARN \| __GFP_HIGHMEM;

		/* It makes no sense to muck about with huge pages */
		gfp &= ~__GFP_COMP;

		while (count) {
		struct page *page = NULL;
		unsigned int order_size;
		@@ -544,15 +547,9 @@ static struct page *__iommu_dma_alloc_pages(struct device dev,
		page = alloc_pages_node(nid, alloc_flags, order);
		if (!page)
		continue;
		if (!order)
		break;
		if (!PageCompound(page)) {
		if (order)
		split_page(page, order);
		break;
		} else if (!split_huge_page(page)) {
		break;
		}
		__free_pages(page, order);
		}
		if (!page) {
		__iommu_dma_free_pages(pages, i);

Admin message