Re: [PATCH bpf-next v2 2/6] net: tun: enable transfer of XDP metadata to skb

From
Willem de Bruijn <willemdebruijn.kernel@gmail.com>
Date
2025-02-18 01:47:36
Also in
bpf, linux-kselftest, lkml
Reply to
Marcus Wichelmann, 2025-02-17
Marcus Wichelmann wrote:
quoted hunk ↗ jump to hunk
When the XDP metadata area was used, it is expected that the same
metadata can also be accessed from TC, as can be read in the description
of the bpf_xdp_adjust_meta helper function. In the tun driver, this was
not yet implemented.

To make this work, the skb that is being built on XDP_PASS should know
of the current size of the metadata area. This is ensured by adding
calls to skb_metadata_set. For the tun_xdp_one code path, an additional
check is necessary to handle the case where the externally initialized
xdp_buff has no metadata support (xdp->data_meta == xdp->data + 1).

More information about this feature can be found in the commit message
of commit de8f3a83b0a0 ("bpf: add meta pointer for direct access").

Signed-off-by: Marcus Wichelmann <redacted>
---
 drivers/net/tun.c | 21 ++++++++++++++++++---
 1 file changed, 18 insertions(+), 3 deletions(-)
diff --git a/drivers/net/tun.c b/drivers/net/tun.c
index c95ab9c46bd2..3dde6cd29a84 100644
--- a/drivers/net/tun.c
+++ b/drivers/net/tun.c
@@ -1604,7 +1604,8 @@ static bool tun_can_build_skb(struct tun_struct *tun, struct tun_file *tfile,
 
 static struct sk_buff *__tun_build_skb(struct tun_file *tfile,
 				       struct page_frag *alloc_frag, char *buf,
-				       int buflen, int len, int pad)
+				       int buflen, int len, int pad,
+				       int metasize)
 {
 	struct sk_buff *skb = build_skb(buf, buflen);
 
@@ -1613,6 +1614,8 @@ static struct sk_buff *__tun_build_skb(struct tun_file *tfile,
 
 	skb_reserve(skb, pad);
 	skb_put(skb, len);
+	if (metasize)
+		skb_metadata_set(skb, metasize);
 	skb_set_owner_w(skb, tfile->socket.sk);
 
 	get_page(alloc_frag->page);
@@ -1672,6 +1675,7 @@ static struct sk_buff *tun_build_skb(struct tun_struct *tun,
 	char *buf;
 	size_t copied;
 	int pad = TUN_RX_PAD;
+	int metasize = 0;
 	int err = 0;
 
 	rcu_read_lock();
@@ -1699,7 +1703,7 @@ static struct sk_buff *tun_build_skb(struct tun_struct *tun,
 	if (hdr->gso_type || !xdp_prog) {
 		*skb_xdp = 1;
 		return __tun_build_skb(tfile, alloc_frag, buf, buflen, len,
-				       pad);
+				       pad, metasize);
 	}
 
 	*skb_xdp = 0;
@@ -1734,12 +1738,18 @@ static struct sk_buff *tun_build_skb(struct tun_struct *tun,
 
 		pad = xdp.data - xdp.data_hard_start;
 		len = xdp.data_end - xdp.data;
+
+		/* It is known that the xdp_buff was prepared with metadata
+		 * support, so no additional check is necessary.
+		 */
+		metasize = xdp.data - xdp.data_meta;
 	}
 	bpf_net_ctx_clear(bpf_net_ctx);
 	rcu_read_unlock();
 	local_bh_enable();
 
-	return __tun_build_skb(tfile, alloc_frag, buf, buflen, len, pad);
+	return __tun_build_skb(tfile, alloc_frag, buf, buflen, len, pad,
+			       metasize);
 
 out:
 	bpf_net_ctx_clear(bpf_net_ctx);
@@ -2456,6 +2466,7 @@ static int tun_xdp_one(struct tun_struct *tun,
 	struct sk_buff_head *queue;
 	u32 rxhash = 0, act;
 	int buflen = hdr->buflen;
+	int metasize = 0;
 	int ret = 0;
 	bool skb_xdp = false;
 	struct page *page;
@@ -2510,6 +2521,10 @@ static int tun_xdp_one(struct tun_struct *tun,
 	skb_reserve(skb, xdp->data - xdp->data_hard_start);
 	skb_put(skb, xdp->data_end - xdp->data);
 
+	metasize = max(xdp->data - xdp->data_meta, 0);
Can xdp->data_meta ever be greater than xdp->data?

This is pointer comparison, which is tricky wrt type. It likely is
ptrdiff_t and thus signed. But may want to use max_t(long int, ..) to
make this explicit.
quoted hunk ↗ jump to hunk
+	if (metasize)
+		skb_metadata_set(skb, metasize);
+
Not strictly needed. As skb_metadata_clear is just
skb_metadata_set(skb, 0). But also not wrong, so fine to keep.
 	if (virtio_net_hdr_to_skb(skb, gso, tun_is_little_endian(tun))) {
 		atomic_long_inc(&tun->rx_frame_errors);
 		kfree_skb(skb);
-- 
2.43.0
  
Thread (16 messages) 16 messages, 3 authors, last reply 2025-02-19
Keyboard shortcuts
hback out one level
jnext message in thread / next series-diff position
kprevious message in thread / previous series-diff position
ldrill in
Escclose help / fold thread tree
?toggle this help