On Wed, Sep 24, 2025 at 5:09 PM Jakub Kicinski [off-list ref] wrote:

On Wed, 24 Sep 2025 06:08:42 +0000 Octavian Purdila wrote:

quoted

When a BPF program that supports BPF_F_XDP_HAS_FRAGS is issuing
bpf_xdp_adjust_tail and a large packet is injected via /dev/net/tun a
crash occurs due to detecting a bad page state (page_pool leak).

This is because xdp_buff does not record the type of memory and
instead relies on the netdev receive queue xdp info. Since the TUN/TAP
driver is using a MEM_TYPE_PAGE_SHARED memory model buffer, shrinking
will eventually call page_frag_free. But with current multi-buff
support for BPF_F_XDP_HAS_FRAGS programs buffers are allocated via the
page pool.

To fix this issue check that the receive queue memory mode is of
MEM_TYPE_PAGE_POOL before using multi-buffs.

This can also happen on veth, right? And veth re-stamps the Rx queues.

I am not sure if re-stamps will have ill effects.

The allocation and deallocation for this issue happens while
processing the same packet (receive skb -> skb_pp_cow_data ->
page_pool alloc ... __bpf_prog_run ->  bpf_xdp_adjust_tail).

IIUC, if the veth re-stamps the RX queue to MEM_TYPE_PAGE_POOL
skb_pp_cow_data will proceed to allocate from page_pool and
bpf_xdp_adjust_tail will correctly free from page_pool.