On Wed, Oct 22, 2025 at 1:33 AM Alessandro Decina
<[email protected]> wrote:
>
> Whenever a status descriptor is received, i40e processes and skips over
> it, correctly updating next_to_process but forgetting to update
> next_to_clean. In the next iteration this accidentally causes the
> creation of an invalid multi-buffer xdp_buff where the first fragment
> is the status descriptor.
>
> If then a skb is constructed from such an invalid buffer - because the
> eBPF program returns XDP_PASS - a panic occurs:
>
> [ 5866.367317] BUG: unable to handle page fault for address: ffd31c37eab1c980
> [ 5866.375050] #PF: supervisor read access in kernel mode
> [ 5866.380825] #PF: error_code(0x0000) - not-present page
> [ 5866.386602] PGD 0
> [ 5866.388867] Oops: Oops: 0000 [#1] SMP NOPTI
> [ 5866.393575] CPU: 34 UID: 0 PID: 0 Comm: swapper/34 Not tainted
> 6.17.0-custom #1 PREEMPT(voluntary)
> [ 5866.403740] Hardware name: Supermicro AS -2115GT-HNTR/H13SST-G, BIOS 3.2
> 03/20/2025
> [ 5866.412339] RIP: 0010:memcpy+0x8/0x10
> [ 5866.416454] Code: cc cc 90 cc cc cc cc cc cc cc cc cc cc cc cc cc cc cc 90
> 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 66 90 48 89 f8 48 89 d1 <f3> a4
> e9 fc 26 c0 fe 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90
> [ 5866.437538] RSP: 0018:ff428d9ec0bb0ca8 EFLAGS: 00010286
> [ 5866.443415] RAX: ff2dd26dbd8f0000 RBX: ff2dd265ad161400 RCX:
> 00000000000004e1
> [ 5866.451435] RDX: 00000000000004e1 RSI: ffd31c37eab1c980 RDI:
> ff2dd26dbd8f0000
> [ 5866.459454] RBP: ff428d9ec0bb0d40 R08: 0000000000000000 R09:
> 0000000000000000
> [ 5866.467470] R10: 0000000000000000 R11: 0000000000000000 R12:
> ff428d9eec726ef8
> [ 5866.475490] R13: ff2dd26dbd8f0000 R14: ff2dd265ca2f9fc0 R15:
> ff2dd26548548b80
> [ 5866.483509] FS: 0000000000000000(0000) GS:ff2dd2c363592000(0000)
> knlGS:0000000000000000
> [ 5866.492600] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> [ 5866.499060] CR2: ffd31c37eab1c980 CR3: 0000000178d7b040 CR4:
> 0000000000f71ef0
> [ 5866.507079] PKRU: 55555554
> [ 5866.510125] Call Trace:
> [ 5866.512867] <IRQ>
> [ 5866.515132] ? i40e_clean_rx_irq_zc+0xc50/0xe60 [i40e]
> [ 5866.520921] i40e_napi_poll+0x2d8/0x1890 [i40e]
> [ 5866.526022] ? srso_alias_return_thunk+0x5/0xfbef5
> [ 5866.531408] ? raise_softirq+0x24/0x70
> [ 5866.535623] ? srso_alias_return_thunk+0x5/0xfbef5
> [ 5866.541011] ? srso_alias_return_thunk+0x5/0xfbef5
> [ 5866.546397] ? rcu_sched_clock_irq+0x225/0x1800
> [ 5866.551493] __napi_poll+0x30/0x230
> [ 5866.555423] net_rx_action+0x20b/0x3f0
> [ 5866.559643] handle_softirqs+0xe4/0x340
> [ 5866.563962] __irq_exit_rcu+0x10e/0x130
> [ 5866.568283] irq_exit_rcu+0xe/0x20
> [ 5866.572110] common_interrupt+0xb6/0xe0
> [ 5866.576425] </IRQ>
> [ 5866.578791] <TASK>
>
> Advance next_to_clean to ensure invalid xdp_buff(s) aren't created.
>
> Fixes: 1c9ba9c14658 ("i40e: xsk: add RX multi-buffer support")
> Signed-off-by: Alessandro Decina <[email protected]>
> ---
> drivers/net/ethernet/intel/i40e/i40e_xsk.c | 7 ++++++-
> 1 file changed, 6 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/ethernet/intel/i40e/i40e_xsk.c
> b/drivers/net/ethernet/intel/i40e/i40e_xsk.c
> index 9f47388eaba5..dbc19083bbb7 100644
> --- a/drivers/net/ethernet/intel/i40e/i40e_xsk.c
> +++ b/drivers/net/ethernet/intel/i40e/i40e_xsk.c
> @@ -441,13 +441,18 @@ int i40e_clean_rx_irq_zc(struct i40e_ring *rx_ring, int
> budget)
> dma_rmb();
>
> if (i40e_rx_is_programming_status(qword)) {
> + u16 ntp;
> +
> i40e_clean_programming_status(rx_ring,
> rx_desc->raw.qword[0],
> qword);
> bi = *i40e_rx_bi(rx_ring, next_to_process);
> xsk_buff_free(bi);
> - if (++next_to_process == count)
> + ntp = next_to_process++;
> + if (next_to_process == count)
> next_to_process = 0;
> + if (next_to_clean == ntp)
> + next_to_clean = next_to_process;
> continue;
> }
>
> --
> 2.43.0
>
>
I'm copying your reply from v1 as shown below so that we can continue
with the discussion :)
> It really depends on whether a status descriptor can be received in the
> middle of multi-buffer packet. Based on the existing code, I assumed it
> can. Therefore, consider this case:
>
> [valid_1st_packet][status_descriptor][valid_2nd_packet]
>
> In this case you want to skip status_descriptor but keep the existing
> logic that leads to:
>
> first = next_to_clean = valid_1st_packet
>
> so then you can go and add valid_2nd_packet as a fragment to the first.
Sorry, honestly, I still don't follow you.
Looking at the case you provided, I think @first always pointing to
valid_1st_packet is valid which does not bring any trouble. You mean
the case is what you're trying to handle?
You patch updates next_to_clean that is only used at the very
beginning, so it will not affect @first. Imaging the following case:
[status_descriptor][valid_1st_packet][valid_2nd_packet]
Even if the next_to_clean is updated, the @first still points to
[status_descriptor] that is invalid and that will later cause the
panic when constructing the skb.
I'm afraid that we're not on the same page. Let me confirm that it is
@first that points to the status descriptor that causes the panic,
right? Could you share with us the exact case just like you did as
above. Thank you.
Thanks,
Jason
