Hi The background: Interrupts are disabled just before stopping other cores during handling restart: See example from: /arch/arm/kernel/reboot.c: /* * Restart requires that the secondary CPUs stop performing any activity * while the primary CPU resets the system. */ void machine_restart(char *cmd) { local_irq_disable(); smp_send_stop(); do_kernel_restart(cmd); /* Give a grace period for failure to restart of 1s */ mdelay(1000); /* Whoops - the platform was unable to reboot. Tell the user! */ printk("Reboot failed -- System halted\n"); while (1); } where function do_kernel_restart() invoke restart handlers: void do_kernel_restart(char *cmd) { atomic_notifier_call_chain(&restart_handler_list, reboot_mode, cmd); } One and the only one! restart handler is mine where I do i2c transaction in order to reset the board. i2c transaction requires interrupts to be enabled. Can I safely do: local_irq_enable on my handler entry and local_irq_disable just before exit? In the meantime I will do i2c_smbus transaction and add some 500ms delay before disable back irq just to make sure transaction is succeed. I know that it is strange question, but in company in which I work I received subtask to get to know. The other alternative is to get response by execute test with for example loop of 10 thousands of restarts.