mlx4_core: For 64-bit systems, vmap() kernel queue buffers

Since kernel virtual memory is not a problem on 64-bit systems, there
is no reason to use our own 2-layer page mapping scheme for large
kernel queue buffers on such systems. Instead, map the page list to a
single virtually contiguous buffer with vmap(), so that can we access
buffer memory via direct indexing.

Signed-off-by: Michael S. Tsirkin <[e-mail]>
Signed-off-by: Jack Morgenstein <[e-mail]>
Signed-off-by: Roland Dreier <[e-mail]>

drivers/net/mlx4/alloc.c

···
 
 			memset(buf->u.page_list[i].buf, 0, PAGE_SIZE);
 		}
+
+		if (BITS_PER_LONG == 64) {
+			struct page **pages;
+			pages = kmalloc(sizeof *pages * buf->nbufs, GFP_KERNEL);
+			if (!pages)
+				goto err_free;
+			for (i = 0; i < buf->nbufs; ++i)
+				pages[i] = virt_to_page(buf->u.page_list[i].buf);
+			buf->u.direct.buf = vmap(pages, buf->nbufs, VM_MAP, PAGE_KERNEL);
+			kfree(pages);
+			if (!buf->u.direct.buf)
+				goto err_free;
+		}
 	}
 
 	return 0;
···
 		dma_free_coherent(&dev->pdev->dev, size, buf->u.direct.buf,
 				  buf->u.direct.map);
 	else {
+		if (BITS_PER_LONG == 64)
+			vunmap(buf->u.direct.buf);
+
 		for (i = 0; i < buf->nbufs; ++i)
 			if (buf->u.page_list[i].buf)
 				dma_free_coherent(&dev->pdev->dev, PAGE_SIZE,

include/linux/mlx4/device.h

···
 };
 
 struct mlx4_buf {
-	union {
+	struct {
 		struct mlx4_buf_list	direct;
 		struct mlx4_buf_list   *page_list;
 	} u;
···
 void mlx4_buf_free(struct mlx4_dev *dev, int size, struct mlx4_buf *buf);
 static inline void *mlx4_buf_offset(struct mlx4_buf *buf, int offset)
 {
-	if (buf->nbufs == 1)
+	if (BITS_PER_LONG == 64 || buf->nbufs == 1)
 		return buf->u.direct.buf + offset;
 	else
 		return buf->u.page_list[offset >> PAGE_SHIFT].buf +