This appears to be a regression in 6.8.0-50-generic.
The following warning from the Intel ice driver is reliably triggered on reboot or shutdown on DGXH100:
[ 97.538724] ------------[ cut here ]------------
[ 97.543943] RTNL: assertion failed at net/core/dev.c (6434)
[ 97.550255] WARNING: CPU: 45 PID: 1 at net/core/dev.c:6434 netif_queue_set_napi+0x11d/0x130
[ 97.559676] Modules linked in: qrtr intel_rapl_msr intel_rapl_common intel_uncore_frequency intel_uncore_frequency_common intel_ifs i10nm_edac skx_edac_common nfit x86_pkg_temp_thermal intel_powerclamp coretemp nouveau mxm_wmi drm_gpuvm cfg80211 drm_exec gpu_sched drm_ttm_helper ttm drm_display_helper cmdlinepart cec spi_nor iaa_crypto kvm_intel rc_core pmt_telemetry mtd intel_sdsi pmt_class binfmt_misc video kvm irqbypass rapl qat_4xxx dax_hmem cxl_acpi intel_cstate irdma intel_qat cxl_core ast idxd crc8 i40e isst_if_mbox_pci isst_if_mmio mei_me i2c_algo_bit intel_vsec idxd_bus authenc isst_if_common switchtec intel_th_gth ipmi_ssif i2c_i801 intel_th_pci spi_intel_pci intel_th i2c_smbus mei spi_intel i2c_ismt nls_iso8859_1 acpi_ipmi ipmi_si ipmi_devintf ipmi_msghandler mac_hid sch_fq_codel dm_multipath msr efi_pstore nfnetlink dmi_sysfs ip_tables x_tables autofs4 btrfs blake2b_generic raid10 raid456 async_raid6_recov async_memcpy async_pq async_xor async_tx xor raid6_pq libcrc32c raid1 raid0 mlx5_ib ib_uverbs macsec
[ 97.559783] ib_core mlx5_core crct10dif_pclmul crc32_pclmul polyval_clmulni ixgbe mlxfw polyval_generic psample ghash_clmulni_intel nvme xfrm_algo ice sha256_ssse3 tls xhci_pci sha1_ssse3 dca gnss nvme_core pci_hyperv_intf xhci_pci_renesas mdio nvme_auth wmi pinctrl_emmitsburg aesni_intel crypto_simd cryptd
[ 97.691627] CPU: 45 PID: 1 Comm: shutdown Not tainted 6.8.0-50-generic #51-Ubuntu
[ 97.700056] Hardware name: NVIDIA DGXH100/DGXH100, BIOS 1.1.3 10/30/2023
[ 97.707606] RIP: 0010:netif_queue_set_napi+0x11d/0x130
[ 97.713399] Code: 00 41 83 e7 01 0f 85 39 ff ff ff ba 22 19 00 00 48 c7 c6 86 10 24 86 48 c7 c7 98 0c 28 86 c6 05 7d 13 90 01 01 e8 83 93 20 ff <0f> 0b e9 13 ff ff ff 66 66 2e 0f 1f 84 00 00 00 00 00 90 90 90 90
[ 97.734504] RSP: 0018:ff4fe9c7c0073c10 EFLAGS: 00010246
[ 97.740392] RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000000000
[ 97.748429] RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000000
[ 97.756464] RBP: ff4fe9c7c0073c38 R08: 0000000000000000 R09: 0000000000000000
[ 97.764494] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000000
[ 97.772529] R13: 0000000000000001 R14: ff39b1cf1a6f9000 R15: 0000000000000000
[ 97.780561] FS: 00007c4b1229f440(0000) GS:ff39b1cafde80000(0000) knlGS:0000000000000000
[ 97.789671] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 97.796144] CR2: 00007c4b12ef94c0 CR3: 000000013094a006 CR4: 0000000000f71ef0
[ 97.804177] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 97.812210] DR3: 0000000000000000 DR6: 00000000fffe07f0 DR7: 0000000000000400
[ 97.820245] PKRU: 55555554
[ 97.823304] Call Trace:
[ 97.826073] <TASK>
[ 97.828455] ? show_regs+0x6d/0x80
[ 97.832305] ? __warn+0x89/0x160
[ 97.835955] ? netif_queue_set_napi+0x11d/0x130
[ 97.841062] ? report_bug+0x17e/0x1b0
[ 97.845203] ? handle_bug+0x51/0xa0
[ 97.849146] ? exc_invalid_op+0x18/0x80
[ 97.853478] ? asm_exc_invalid_op+0x1b/0x20
[ 97.858204] ? netif_queue_set_napi+0x11d/0x130
[ 97.863317] ice_vsi_clear_napi_queues+0x36/0x90 [ice]
[ 97.869192] ice_vsi_close+0x24/0xd0 [ice]
[ 97.873864] ice_deinit_eth+0x2b/0xa0 [ice]
[ 97.878619] ice_remove+0x107/0x250 [ice]
[ 97.883176] ice_shutdown+0x1b/0x70 [ice]
[ 97.887736] pci_device_shutdown+0x37/0x90
[ 97.892370] device_shutdown+0x13e/0x280
[ 97.896805] kernel_restart+0x3a/0xa0
[ 97.900945] __do_sys_reboot+0x156/0x250
[ 97.905363] __x64_sys_reboot+0x1b/0x30
[ 97.909694] x64_sys_call+0x219f/0x25a0
[ 97.914029] do_syscall_64+0x7f/0x180
[ 97.918162] ? irqentry_exit_to_user_mode+0x7b/0x260
[ 97.923759] ? irqentry_exit+0x43/0x50
[ 97.927993] ? exc_page_fault+0x94/0x1b0
[ 97.932422] entry_SYSCALL_64_after_hwframe+0x78/0x80
[ 97.938118] RIP: 0033:0x7c4b12e1ba07
[ 97.942226] Code: c7 c0 ff ff ff ff eb be 66 2e 0f 1f 84 00 00 00 00 00 90 f3 0f 1e fa 89 fa be 69 19 12 28 bf ad de e1 fe b8 a9 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 01 c3 48 8b 15 e1 c3 0d 00 f7 d8 64 89 02 b8
[ 97.963323] RSP: 002b:00007ffd36a491a8 EFLAGS: 00000246 ORIG_RAX: 00000000000000a9
[ 97.971846] RAX: ffffffffffffffda RBX: 0000000000000003 RCX: 00007c4b12e1ba07
[ 97.979879] RDX: 0000000001234567 RSI: 0000000028121969 RDI: 00000000fee1dead
[ 97.987909] RBP: 00007ffd36a493e0 R08: 0000000000000069 R09: 0000000000000000
[ 97.995941] R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
[ 98.003973] R13: 0000000000000000 R14: 0000000000000000 R15: 0000000001234567
[ 98.012004] </TASK>
[ 98.014480] ---[ end trace 0000000000000000 ]---
Patch submitted to kernel team mailing list: https:/
SRU Justification
[Impact]
An RTNL assertion failed warning is triggered from the Intel ice driver when a
PCIe device using the driver is removed or the system is shutdown/rebooted.
This was caused by commit "ice: move netif_queue_
sections" which was brought in through stable updates in 6.8.0-50-generic. The
commit adds a call to "ice_vsi_
in K6.8 "ice_vsi_close()" is called via "ice_remove()". Function "ice_remove()"
is used as the PCI remove callback, and is also called via "ice_shutdown()".
[Fix]
This is resolved by cherry picking commit "ice: Remove and readd netdev during
devlink reload" from upstream. This commit refactors "ice_remove()" to not call
"ice_vsi_close()". Upstream was not affected because in mainline "ice: Remove
and readd netdev during devlink reload" is a parent to "ice: move
netif_queue_
Only noble:linux and its derivative kernels are affected.
[Test Plan]
(1) Apply patch "ice: Remove and readd netdev during devlink reload", reboot
into patched kernel on system that utilizes the Intel ice driver.
(2) Reboot or shutdown the system. Observe in the system's console that the
warning splat from LP#2091107 no longer appears.
[Where problems could occur]
This change affects the Intel ice kernel module specifically. Issues with this
fix would manifest as misbehavior of that driver, which would be used with
select intel NICs, including "Intel Corporation Ethernet Controller E810-C for
QSFP".