On Fri, 16 Jan 2015, Oliver Neukum wrote: > Hi, > > I was looking at the CDC-WDM driver. It uses cancel_work_sync() > in the pre_reset() method. I looked at the work and saw a GFP_KERNEL > there. > Now suppose a device has a wdm interface and a storage interface. > > Is the following sequence possible? > > kmalloc with GFP_KERNEL in the work > VM layer decides to page to storage interface > task blocks waiting for completion of IO > > SCSI layer sees an error > SCSI error handling starts > SCSI layer requests a bus reset > > USB layer calls pre_reset() of cdc-wdm > wdm_pre_reset() calls cancel_work_sync() > cancel_work_sync() waits for work to finish > DEADLOCK > > What do you think? Yes, that could happen. It seems unlikely but it is possible. Alan Stern -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
