Re: [PATCH net v2] xsk: avoid data corruption on cq descriptor number

[Maciej Fijalkowski <maciej.fijalkowski@intel.com>]

On Sat, Oct 25, 2025 at 08:18:17PM +0200, Fernando Fernandez Mancera wrote:
> 
> 
> On 10/24/25 7:16 PM, Maciej Fijalkowski wrote:
> > On Fri, Oct 24, 2025 at 12:40:49PM +0200, Fernando Fernandez Mancera wrote:
> > > Since commit 30f241fcf52a ("xsk: Fix immature cq descriptor
> > > production"), the descriptor number is stored in skb control block and
> > > xsk_cq_submit_addr_locked() relies on it to put the umem addrs onto
> > > pool's completion queue.
> > > 
> > > skb control block shouldn't be used for this purpose as after transmit
> > > xsk doesn't have control over it and other subsystems could use it. This
> > > leads to the following kernel panic due to a NULL pointer dereference.
> > > 
> > >   BUG: kernel NULL pointer dereference, address: 0000000000000000
> > >   #PF: supervisor read access in kernel mode
> > >   #PF: error_code(0x0000) - not-present page
> > >   PGD 0 P4D 0
> > >   Oops: Oops: 0000 [#1] SMP NOPTI
> > >   CPU: 2 UID: 1 PID: 927 Comm: p4xsk.bin Not tainted 6.16.12+deb14-cloud-amd64 #1 PREEMPT(lazy)  Debian 6.16.12-1
> > >   Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.17.0-debian-1.17.0-1 04/01/2014
> > >   RIP: 0010:xsk_destruct_skb+0xd0/0x180
> > >   [...]
> > >   Call Trace:
> > >    <IRQ>
> > >    ? napi_complete_done+0x7a/0x1a0
> > >    ip_rcv_core+0x1bb/0x340
> > >    ip_rcv+0x30/0x1f0
> > >    __netif_receive_skb_one_core+0x85/0xa0
> > >    process_backlog+0x87/0x130
> > >    __napi_poll+0x28/0x180
> > >    net_rx_action+0x339/0x420
> > >    handle_softirqs+0xdc/0x320
> > >    ? handle_edge_irq+0x90/0x1e0
> > >    do_softirq.part.0+0x3b/0x60
> > >    </IRQ>
> > >    <TASK>
> > >    __local_bh_enable_ip+0x60/0x70
> > >    __dev_direct_xmit+0x14e/0x1f0
> > >    __xsk_generic_xmit+0x482/0xb70
> > >    ? __remove_hrtimer+0x41/0xa0
> > >    ? __xsk_generic_xmit+0x51/0xb70
> > >    ? _raw_spin_unlock_irqrestore+0xe/0x40
> > >    xsk_sendmsg+0xda/0x1c0
> > >    __sys_sendto+0x1ee/0x200
> > >    __x64_sys_sendto+0x24/0x30
> > >    do_syscall_64+0x84/0x2f0
> > >    ? __pfx_pollwake+0x10/0x10
> > >    ? __rseq_handle_notify_resume+0xad/0x4c0
> > >    ? restore_fpregs_from_fpstate+0x3c/0x90
> > >    ? switch_fpu_return+0x5b/0xe0
> > >    ? do_syscall_64+0x204/0x2f0
> > >    ? do_syscall_64+0x204/0x2f0
> > >    ? do_syscall_64+0x204/0x2f0
> > >    entry_SYSCALL_64_after_hwframe+0x76/0x7e
> > >    </TASK>
> > >   [...]
> > >   Kernel panic - not syncing: Fatal exception in interrupt
> > >   Kernel Offset: 0x1c000000 from 0xffffffff81000000 (relocation range: 0xffffffff80000000-0xffffffffbfffffff)
> > > 
> > > The approach proposed stores the first address also in the xsk_addr_node
> > > along with the number of descriptors. The head xsk_addr_node is
> > > referenced in skb_shinfo(skb)->destructor_arg. The rest of the fragments
> > > store the address on the list.
> > > 
> > > This is less efficient as 4 bytes are wasted when storing each address.
> > 
> > Hi Fernando,
> > it's not about 4 bytes being wasted but rather memory allocation that you
> > introduce here. I tried hard to avoid hurting non-fragmented traffic,
> > below you can find impact reported by Jason from similar approach as
> > yours:
> > https://lore.kernel.org/bpf/CAL+tcoD=Gn6ZmJ+_Y48vPRyHVHmP-7irsx=fRVRnyhDrpTrEtQ@mail.gmail.com/
> > 
> > I assume this patch will yield a similar performance degradation...
> > 
> 
> It does, thank you for explaining Maciej. I have been thinking about
> possible solutions and remembered skb extensions. If I am not wrong, it
> shouldn't yield a performance degratation or at least it should be a much
> less severe one. Although, XDP_SOCKETS Kconfig would require "select
> SKB_EXTENSIONS".

SGTM. However I have not used this feature in the past, so I'm looking
forward for your implementation.

> 
> What do you think about this approach? I could draft a series for net-next..

That should be targetted to 'bpf' as a fix, still.

> I am just looking for different options other than using skb control block
> because I believe similar issues will arise in the future even if we fix the
> current one on ip_rcv..

Yes I agree. Heart says 'just fix the IP layer' but as you said, we never
know if other layer would wipe out the cb.

> 
> Thanks,
> Fernando.