On Mon, Apr 15, 2024 at 12:59:37PM +0200, Kurt Kanzenbach wrote: > From: Lukas Wunner <[email protected]> > > The removal of the igc module leads to a deadlock: > > |[Mon Apr 8 17:38:55 2024] __mutex_lock.constprop.0+0x3e5/0x7a0 > |[Mon Apr 8 17:38:55 2024] ? preempt_count_add+0x85/0xd0 > |[Mon Apr 8 17:38:55 2024] __mutex_lock_slowpath+0x13/0x20 > |[Mon Apr 8 17:38:55 2024] mutex_lock+0x3b/0x50 > |[Mon Apr 8 17:38:55 2024] rtnl_lock+0x19/0x20 > |[Mon Apr 8 17:38:55 2024] unregister_netdevice_notifier+0x2a/0xc0 > |[Mon Apr 8 17:38:55 2024] netdev_trig_deactivate+0x25/0x70 > |[Mon Apr 8 17:38:55 2024] led_trigger_set+0xe2/0x2d0 > |[Mon Apr 8 17:38:55 2024] led_classdev_unregister+0x4f/0x100 > |[Mon Apr 8 17:38:55 2024] devm_led_classdev_release+0x15/0x20 > |[Mon Apr 8 17:38:55 2024] release_nodes+0x47/0xc0 > |[Mon Apr 8 17:38:55 2024] devres_release_all+0x9f/0xe0 > |[Mon Apr 8 17:38:55 2024] device_del+0x272/0x3c0 > |[Mon Apr 8 17:38:55 2024] netdev_unregister_kobject+0x8c/0xa0 > |[Mon Apr 8 17:38:55 2024] unregister_netdevice_many_notify+0x530/0x7c0 > |[Mon Apr 8 17:38:55 2024] unregister_netdevice_queue+0xad/0xf0 > |[Mon Apr 8 17:38:55 2024] unregister_netdev+0x21/0x30 > |[Mon Apr 8 17:38:55 2024] igc_remove+0xfb/0x1f0 [igc] > |[Mon Apr 8 17:38:55 2024] pci_device_remove+0x42/0xb0 > |[Mon Apr 8 17:38:55 2024] device_remove+0x43/0x70 > > unregister_netdev() acquires the RNTL lock and releases the LEDs bound > to that netdevice. However, netdev_trig_deactivate() and later > unregister_netdevice_notifier() try to acquire the RTNL lock again. > > Avoid this situation by not using the device-managed LED class > functions. > > Link: > https://lore.kernel.org/r/CAEhC_B=ksywxcg_+aqqxurgegkq+4mqnsv8ebhokbc3-obj...@mail.gmail.com/ > Link: https://lore.kernel.org/r/ZhRD3cOtz5i-61PB@mail-itl/ > Reported-by: Roman Lozko <[email protected]> > Reported-by: "Marek Marczykowski-Górecki" <[email protected]> > Fixes: ea578703b03d ("igc: Add support for LEDs on i225/i226") > Signed-off-by: Lukas Wunner <[email protected]> > [Kurt: Wrote commit message and tested on i225] > Signed-off-by: Kurt Kanzenbach <[email protected]>
I am aware that this patch seems to have also been submitted by Lucas himself. I'd like to suggest that we focus on review of that submission. https://lore.kernel.org/netdev/2f1be6b1cf2b3346929b0049f2ac7d7d79acb5c9.1713188539.git.lu...@wunner.de/
