On Sat, Aug 16, 2025 at 11:18:02AM +0900, Yunseong Kim wrote: > I think this part is a macro, so it appears this way. > > Link: https://github.com/torvalds/linux/blob/v6.17-rc1/include/linux/spinlock_rt.h#L96 > > #define spin_lock_irqsave(lock, flags) \ > do { \ > typecheck(unsigned long, flags); \ > flags = 0; \ > spin_lock(lock); \ > } while (0) > > My tree is indeed 6.17-rc1. I made a mistake in the diagram, > which caused the misunderstanding. I’ve redrawn the diagram: > > kworker (hub_event) > | > v > vhci_urb_enqueue() [drivers/usb/usbip/vhci_hcd.c] > | > |---> spin_unlock_irqrestore(&vhci->lock, flags); > | (Context: IRQs Enabled, Process Context) > |---> local_irq_disable(); > | > | *** STATE CHANGE: IRQs Disabled (Atomic Context) *** > | > +-----> usb_hcd_giveback_urb() [drivers/usb/core/hcd.c] > | > v > __usb_hcd_giveback_urb() > | > v > mon_complete() [drivers/usb/mon/mon_main.c] > | > |---> spin_lock_irqsave() [include/linux/spinlock_rt.h] > | > v https://github.com/torvalds/linux/blob/v6.17-rc1/include/linux/spinlock_rt.h#L96 > spin_lock() [kernel/locking/spinlock_rt.c] <--- Attempts to acquire lock > | > | [On PREEMPT_RT] > v > rt_spin_lock() [kernel/locking/spinlock_rt.c] > | > v > [May Sleep if contended] > | > X <----------- BUG: Sleeping in atomic context (IRQs are disabled!) > > | > |---> local_irq_enable(); > (Context: IRQs Enabled) So it looks like we should be using a different function instead of local_irq_disable(). We need something which in a non-RT build will disable interrupts on the local CPU, but in an RT build will merely disable preemption. (In fact, every occurrence of local_irq_disable() in the USB subsystem probably should be changed in this way.) Is there such a function? Alan Stern
