On 7/10/2020 10:32 pm, Jerome Brunet wrote:
With arm64 defconfig on Khadas vim3, no obvious regression. Looks good. Tested-by: Jerome Brunet <jbrunet@xxxxxxxxxxxx> I did not test with RT. Brad, Could you let us know is Thomas's patch works for you ? Thx
There was a merge conflict in applying against v5.9-rc8-rt12 with particular this patch: https://git.kernel.org/pub/scm/linux/kernel/git/rt/linux-rt-devel.git/commit/kernel/irq/manage.c?h=linux-5.9.y-rt&id=18df00ef0b2b1513dc8f1a9ed26b11fff2261c30 I did manage to add the patch after attempting to resolve the conflict which solves the deadlock issue I am seeing with mmc and works fine during testing (a kernel compilation on preempt_rt configured kernel). irq_thread in /kernel/irq/manage.c Looks like this (not 100% sure I should have placed the irq_finalize_oneshot before add_interrupt_randomness). Based on this I can provide Tested-by: Brad Harper <bjharper@xxxxxxxxx> --- static int irq_thread(void *data) { struct callback_head on_exit_work; struct irqaction *action = data; struct irq_desc *desc = irq_to_desc(action->irq); irqreturn_t (*handler_fn)(struct irq_desc *desc, struct irqaction *action); if (force_irqthreads && test_bit(IRQTF_FORCED_THREAD, &action->thread_flags)) handler_fn = irq_forced_thread_fn; else handler_fn = irq_thread_fn; init_task_work(&on_exit_work, irq_thread_dtor); task_work_add(current, &on_exit_work, false); irq_thread_check_affinity(desc, action); while (!irq_wait_for_interrupt(action)) { irqreturn_t action_ret; irq_thread_check_affinity(desc, action); action_ret = handler_fn(desc, action); if (action_ret == IRQ_HANDLED) atomic_inc(&desc->threads_handled); if (action_ret == IRQ_WAKE_THREAD) irq_wake_secondary(desc, action); irq_finalize_oneshot(desc, action); if (IS_ENABLED(CONFIG_PREEMPT_RT)) { migrate_disable(); add_interrupt_randomness(action->irq, 0, desc->random_ip ^ (unsigned long) action); migrate_enable(); } wake_threads_waitq(desc); } /* * This is the regular exit path. __free_irq() is stopping the * thread via kthread_stop() after calling * synchronize_hardirq(). So neither IRQTF_RUNTHREAD nor the * oneshot mask bit can be set. */ task_work_cancel(current, irq_thread_dtor); return 0; } ---
