On 01/04/2022 16.52, Bart Van Assche wrote: > On 3/31/22 23:20, Adrian Hunter wrote: >> On 01/04/2022 1.34, Bart Van Assche wrote: >>> Quiescing LUNs falls outside the scope of a shutdown callback. The shutdown >>> callback is called from inside the reboot() system call and the reboot() >>> system call is called after user space has stopped accessing block devices. >>> Hence this patch that removes the quiescing calls from >>> ufshcd_wl_shutdown(). This patch makes shutdown faster since multiple >>> synchronize_rcu() calls are removed. >> >> AFAIK there is nothing stopping shutdown being called during intense UFS I/O. >> What happens then? > > Hmm ... how could this happen? Am I perhaps misunderstanding something about the Linux shutdown sequence? > > The UFS driver is the only driver I know that tries to stop I/O from inside its shutdown callback. I'm not aware of any other Linux kernel driver that tries to pause I/O from inside its shutdown callback. We are putting the UFS device into powerdown mode. I am not sure what will happen if the device is processing requests at the same time.
