On Mon, Jul 3, 2023 at 4:45 PM David Ahern [off-list ref] wrote:

On 7/3/23 12:22 AM, Mina Almasry wrote:

quoted

tcpdump is able to access the header of these skbs which is in host
memory, but not the payload in device memory. Here is an example
session with my netcat-like test for device memory TCP:
https://pastebin.com/raw/FRjKf0kv

tcpdump seems to work, and the length of the packets above is correct.
tcpdump -A however isn't able to print the payload of the packets:
https://pastebin.com/raw/2PcNxaZV

That is my expectation. The tcpdump is just an easy example of accessing
the skb page frags. skb_copy_and_csum_bits used by icmp is another
example that can walk frags wanting access to device memory. You did not
cause a panic or trip a WARN_ON for example with the tcpdump?

Change for af_packet was not too hard :)

diff --git a/net/packet/af_packet.c b/net/packet/af_packet.c
index a2dbeb264f260e5b8923ece9aac99fe19ddfeb62..aa4133d1b1e0676e408499ea4534b51262394432

100644

--- a/net/packet/af_packet.c
+++ b/net/packet/af_packet.c

@@ -2152,7 +2152,7 @@ static int packet_rcv(struct sk_buff *skb,

struct net_device *dev,
                }
        }

-       snaplen = skb->len;
+       snaplen = skb->devmem ? skb_headlen(skb) : skb->len;

        res = run_filter(skb, sk, snaplen);
        if (!res)

@@ -2275,7 +2275,7 @@ static int tpacket_rcv(struct sk_buff *skb,

struct net_device *dev,
                }
        }

-       snaplen = skb->len;
+       snaplen = skb->devmem ? skb_headlen(skb) : skb->len;

        res = run_filter(skb, sk, snaplen);
        if (!res)


And a generic change in pskb_may_pull() ( __pskb_pull_tail() more
exactly) was enough to cover most other cases.