On Sat, Feb 27, 2010 at 3:53 AM, David Miller [off-list ref] wrote:

From: Robert Hancock <redacted>
Date: Fri, 26 Feb 2010 21:08:04 -0600

quoted

That seems like a reasonable approach to me. Only question is how to
implement the check for DMA_64BIT. Can we just check page_to_phys on
each of the pages in the skb to see if it's > 0xffffffff ? Are there
any architectures where it's more complicated than that?

On almost every platform it's "more complicated than that".

This is the whole issue.  What matters is the final DMA address and
since we have IOMMUs and the like, it is absolutely not tenable to
solve this by checking physical address attributes.

Yeah, physical address isn't quite right. There is a precedent for
such a check in the block layer though - look at
blk_queue_bounce_limit in block/blk-settings.c:

void blk_queue_bounce_limit(struct request_queue *q, u64 dma_mask)
{
        unsigned long b_pfn = dma_mask >> PAGE_SHIFT;
        int dma = 0;

        q->bounce_gfp = GFP_NOIO;
#if BITS_PER_LONG == 64
        /*
         * Assume anything <= 4GB can be handled by IOMMU.  Actually
         * some IOMMUs can handle everything, but I don't know of a
         * way to test this here.
         */
        if (b_pfn < (min_t(u64, 0xffffffffUL, BLK_BOUNCE_HIGH) >> PAGE_SHIFT))
                dma = 1;
        q->limits.bounce_pfn = max_low_pfn;
#else
        if (b_pfn < blk_max_low_pfn)
                dma = 1;
        q->limits.bounce_pfn = b_pfn;
#endif
        if (dma) {
                init_emergency_isa_pool();
                q->bounce_gfp = GFP_NOIO | GFP_DMA;
                q->limits.bounce_pfn = b_pfn;
        }
}

and then in mm/bounce.c:

static void __blk_queue_bounce(struct request_queue *q, struct bio **bio_orig,
                               mempool_t *pool)
{
        struct page *page;
        struct bio *bio = NULL;
        int i, rw = bio_data_dir(*bio_orig);
        struct bio_vec *to, *from;

        bio_for_each_segment(from, *bio_orig, i) {
                page = from->bv_page;

                /*
                 * is destination page below bounce pfn?
                 */
                if (page_to_pfn(page) <= queue_bounce_pfn(q))
                        continue;

Following that logic then, it appears that page_to_pfn(page) >
(0xffffffff >> PAGE_SHIFT) should tell us what we want to know for the
DMA_64BIT flag.. or am I missing something?