i40e panic on link loss or after changing NIC ring buffer count

Issue

• System with Intel 700 series i40e NIC suffered error:

[3408318.640794] i40e 0000:03:00.0 ens64: NIC Link is Down
[3408319.378069] i40e 0000:03:00.0 ens64: NIC Link is Up, 10 Gbps Full Duplex, Flow Control: None
[3408332.303305] i40e 0000:03:00.0 ens64: NIC Link is Down
[3408338.343124] i40e 0000:03:00.0 ens64: Changing Tx descriptor count from 8160 to 512.
[3408338.344477] i40e 0000:03:00.0 ens64: Changing Rx descriptor count from 8160 to 512
[3408761.504733] i40e 0000:03:00.0 ens64: NIC Link is Up, 10 Gbps Full Duplex, Flow Control: None
[3408761.543703] i40e 0000:03:00.0 ens64: Changing Tx descriptor count from 512 to 8160.
[3408761.554146] i40e 0000:03:00.0 ens64: Changing Rx descriptor count from 512 to 8160

[3408761.583181] NetworkManager: page allocation failure: order:6, mode:0x40dc0(GFP_KERNEL|__GFP_COMP|__GFP_ZERO)
[3408761.583595] CPU: 7 PID: 1428 Comm: NetworkManager Kdump: loaded Not tainted 5.14.0-570.28.1.el9_6.x86_64 #1

[3408761.584360] Call Trace:
[3408761.584734]  <TASK>
[3408761.585105]  dump_stack_lvl+0x34/0x48
[3408761.585481]  warn_alloc+0x129/0x150
[3408761.585854]  ? __alloc_pages_direct_compact+0xa7/0x210
[3408761.586220]  __alloc_pages_slowpath.constprop.0+0xa73/0xb20
[3408761.586580]  ? get_page_from_freelist+0x2aa/0x590
[3408761.586936]  __alloc_pages+0x21d/0x250
[3408761.587287]  __kmalloc_large_node+0x79/0x110
[3408761.587636]  __kmalloc+0x322/0x440
[3408761.587982]  ? i40e_setup_rx_descriptors+0x91/0xb0 [i40e]
[3408761.588363]  ? i40e_setup_rx_descriptors+0x91/0xb0 [i40e]
[3408761.588717]  i40e_setup_rx_descriptors+0x91/0xb0 [i40e]
[3408761.589075]  i40e_set_ringparam.cold+0x11d/0x56e [i40e]
[3408761.589425]  __dev_ethtool+0x8fc/0x1b40
[3408761.589752]  ? __sk_destruct+0x155/0x230
[3408761.590073]  ? kmem_cache_free+0x3f1/0x420
[3408761.590390]  ? kmalloc_trace+0x176/0x330
[3408761.590704]  dev_ethtool+0xa8/0x170
[3408761.591013]  dev_ioctl+0x1b5/0x580
[3408761.591317]  ? sk_ioctl+0x4a/0x110
[3408761.591616]  sock_do_ioctl+0xab/0xf0
[3408761.591919]  sock_ioctl+0x1ce/0x2e0
[3408761.592210]  ? auditd_test_task+0x3c/0x50
[3408761.592498]  ? __audit_syscall_entry+0xef/0x140
[3408761.592782]  __x64_sys_ioctl+0x87/0xc0
[3408761.593062]  do_syscall_64+0x5c/0xe0
[3408761.593336]  ? __irq_exit_rcu+0x46/0xc0
[3408761.593608]  ? common_interrupt+0x43/0xa0
[3408761.593873]  entry_SYSCALL_64_after_hwframe+0x78/0x80

• The system panicked shortly after:

[3408762.143799] BUG: kernel NULL pointer dereference, address: 0000000000000008
[3408762.143802] #PF: supervisor write access in kernel mode
[3408762.143803] #PF: error_code(0x0002) - not-present page
[3408762.143804] PGD 0 P4D 0 
[3408762.143806] Oops: 0002 [#1] PREEMPT SMP NOPTI
[3408762.143811] RIP: 0010:i40e_xmit_frame_ring+0xff/0x500 [i40e]

[3408762.143878] Call Trace:
[3408762.143880]  <TASK>
[3408762.143881]  ? show_trace_log_lvl+0x1c4/0x2df
[3408762.143885]  ? show_trace_log_lvl+0x1c4/0x2df
[3408762.143887]  ? dev_hard_start_xmit+0x85/0x1d0
[3408762.143890]  ? __die_body.cold+0x8/0xd
[3408762.143892]  ? page_fault_oops+0x134/0x170
[3408762.143895]  ? _copy_to_iter+0x61/0x570
[3408762.143900]  ? exc_page_fault+0x62/0x150
[3408762.143903]  ? asm_exc_page_fault+0x22/0x30
[3408762.143907]  ? i40e_xmit_frame_ring+0xff/0x500 [i40e]
[3408762.143928]  dev_hard_start_xmit+0x85/0x1d0
[3408762.143930]  sch_direct_xmit+0x9b/0x360
[3408762.143934]  __dev_xmit_skb+0x22a/0x570
[3408762.143937]  __dev_queue_xmit+0x2c2/0x6b0
[3408762.143939]  ? packet_parse_headers+0x107/0x220
[3408762.143942]  ? packet_parse_headers+0x107/0x220
[3408762.143944]  packet_snd+0x382/0x760
[3408762.143946]  __sys_sendto+0x1dc/0x1f0
[3408762.143950]  ? syscall_exit_to_user_mode+0x19/0x40
[3408762.143953]  ? auditd_test_task+0x3c/0x50
[3408762.143956]  ? __audit_syscall_entry+0xef/0x140
[3408762.143959]  __x64_sys_sendto+0x20/0x30
[3408762.143961]  do_syscall_64+0x5c/0xe0
[3408762.143963]  ? audit_reset_context.part.0.constprop.0+0x273/0x2e0
[3408762.143965]  ? syscall_exit_work+0x103/0x130
[3408762.143967]  ? syscall_exit_to_user_mode+0x19/0x40
[3408762.143970]  ? do_syscall_64+0x6b/0xe0
[3408762.143971]  ? syscall_exit_work+0x103/0x130
[3408762.143972]  ? syscall_exit_to_user_mode+0x19/0x40
[3408762.143975]  ? do_syscall_64+0x6b/0xe0
[3408762.143976]  entry_SYSCALL_64_after_hwframe+0x78/0x80