On Thu, Sep 15, 2022 at 10:44 AM Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> wrote: > > On Wed, Sep 14, 2022 at 10:00:17AM +0200, Nam Cao wrote: > > In vt6655 driver, there is a single interrupt handler: vnt_interrupt(), > > and it does not take the spinlock. The interrupt handler only schedules > > a workqueue, and the spinlock is taken in this workqueue. Thus, there is > > no need to use spin_lock_irqsave, as the spinlock is never taken by an > > interrupt. Replace spin_lock_irqsave (and spin_unlock_irqsave) with > > spin_lock (and spin_unlock). > > What is the speed difference before and after this change? And how are > interrupts properly handled anymore in this driver if you took away the > lock that was being accessed in the irq? My understanding is that no lock is accessed in the irq in this driver. But it seems like I am wrong. Sorry for sending a buggy patch, I may follow-up with this later when I figure out what I did wrong. Best regards, Nam.
