Re: [PATCHv4 bpf-next 1/2] xdp: add a new helper for dev map multicast support

On Fri, 12 Jun 2020 16:54:08 +0800
Hangbin Liu [off-list ref] wrote:

On Wed, Jun 10, 2020 at 12:18:59PM +0200, Jesper Dangaard Brouer wrote:

quoted

On Tue, 26 May 2020 22:05:38 +0800
Hangbin Liu [off-list ref] wrote:
  

quoted

diff --git a/net/core/xdp.c b/net/core/xdp.c
index 90f44f382115..acdc63833b1f 100644
--- a/net/core/xdp.c
+++ b/net/core/xdp.c

@@ -475,3 +475,29 @@ void xdp_warn(const char *msg, const char *func, const int line)
 	WARN(1, "XDP_WARN: %s(line:%d): %s\n", func, line, msg);
 };
 EXPORT_SYMBOL_GPL(xdp_warn);
+
+struct xdp_frame *xdpf_clone(struct xdp_frame *xdpf)
+{
+	unsigned int headroom, totalsize;
+	struct xdp_frame *nxdpf;
+	struct page *page;
+	void *addr;
+
+	headroom = xdpf->headroom + sizeof(*xdpf);
+	totalsize = headroom + xdpf->len;
+
+	if (unlikely(totalsize > PAGE_SIZE))
+		return NULL;
+	page = dev_alloc_page();
+	if (!page)
+		return NULL;
+	addr = page_to_virt(page);
+
+	memcpy(addr, xdpf, totalsize);  

I don't think this will work.  You are assuming that the memory model
(xdp_mem_info) is the same.

You happened to use i40, that have MEM_TYPE_PAGE_SHARED, and you should
have changed this to MEM_TYPE_PAGE_ORDER0, but it doesn't crash as they
are compatible.  If you were using mlx5, I suspect that this would
result in memory leaking.  

Is there anything else I should do except add the following line?
	nxdpf->mem.type = MEM_TYPE_PAGE_ORDER0;

You do realize that you also have copied over the mem.id, right?

And as I wrote below you also need to update frame_sz.

quoted

You also need to update xdpf->frame_sz, as you also cannot assume it is
the same.  

Won't the memcpy() copy xdpf->frame_sz to nxdpf? 

You obviously cannot use the frame_sz from the existing frame, as you
just allocated a new page for the new xdp_frame, that have another size
(here PAGE_SIZE).

And I didn't see xdpf->frame_sz is set in xdp_convert_zc_to_xdp_frame(),
do we need a fix?

Good catch, that sounds like a bug, that should be fixed.
Will you send a fix?

quoted

quoted

+
+	nxdpf = addr;
+	nxdpf->data = addr + headroom;
+
+	return nxdpf;
+}
+EXPORT_SYMBOL_GPL(xdpf_clone);  

struct xdp_frame {
	void *data;
	u16 len;
	u16 headroom;
	u32 metasize:8;
	u32 frame_sz:24;
	/* Lifetime of xdp_rxq_info is limited to NAPI/enqueue time,
	 * while mem info is valid on remote CPU.
	 */
	struct xdp_mem_info mem;
	struct net_device *dev_rx; /* used by cpumap */
};
  

struct xdp_mem_info {
	u32                        type;                 /*     0     4 */
	u32                        id;                   /*     4     4 */

	/* size: 8, cachelines: 1, members: 2 */
	/* last cacheline: 8 bytes */
};

-- 
Best regards,
  Jesper Dangaard Brouer
  MSc.CS, Principal Kernel Engineer at Red Hat
  LinkedIn: http://www.linkedin.com/in/brouer