On Thu, 7 Jun 2012, megha dey wrote: > SCENARIO: > In host mode, I am trying to copy some files from the pendrive to the > sdcard or vice-versa using the astro file manager.During the file > transfer, if I disconnect the pendrive,and then reconnect it back,it > is observed that re-enumeration is not taking place. What version of the kernel are you using? > ANALYSIS: > From the stack trace,we see that there are 2 main threads involved: > 1) usb-storage thread: > When file transfer is in progress,the pendrive is removed.In this > context,usb_sg_wait() function is waiting infinitely for the IO to > complete (wait_for_completion(&io->complete)). Since the drive is > removed,this wait never completes. No, the wait should complete even after the drive is removed. Disconnect processing cancels all outstanding I/O operations. > 2) khubd thread > On disconnecting the pendrive, during the device delete operation, > usb_stor_disconnect() function is called which is trying to release > the storage resources and then tries to stop the usb-storage thread > via kthread_stop(). > > The usb_storage thread is not listening on kthread_should_stop as it > is waiting infinitely on usb_sg_wait(). Hence the hub thread is also > blocked as it tries to stop the storage thread. > > QUESTIONS: > 1)Why is the wait_for_completion(&io->complete) not a timed wait? (* > After adding a timeout of 5 seconds,this issue is no longer seen , the > pendrive enumerates each time) Because requests already are timed out by the block layer, and because disconnects should cause the request to be cancelled immediately. > 2)If we do put a timed wait,what should be the ideal timeout value? It should not be a timed wait. 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
