Hi Julian, Thanks for looking at this. In short - I agree with your review and will do most of them. As a well as a few minor changes suggested by the kbuild test robot. Expect a new version shortly. On Mon, Nov 16, 2015 at 10:25 PM, Julian Calaby <julian.calaby@xxxxxxxxx> wrote: >> >> static void __exit ath6kl_sdio_exit(void) >> { >> sdio_unregister_driver(&ath6kl_sdio_driver); >> + >> +#ifdef CONFIG_GPIOLIB >> + /* Delay to avoid pulling the plug on the chip when an irq is pending >> + * and then getting a spurious message: >> + * "ath6kl: failed to get pending recv messages: -125" >> + */ >> + msleep(300); > > Is there some actual synchronisation you can do here against the IRQ? > 300msec isn't a long time to wait, but it seems wrong here. > Considering this is only called on exit, I wasn't too worried about this, but then again, on reboot as well as a few types of reconfiguring the interface speed is an important consideration for our system. I'm open to suggestions. I had looked at this and didn't see an obvious way to try to sync with the IRQ as it goes through several subsystems. But looking at it again in a different light... 125 is ECANCELED. Which is exactly right and appropriate, and ath6kl_htc_rxmsg_pending_handler() where the message comes from does a proper cleanup, or so it looks like. I wonder if maybe we can consider ECANCELED not an error that needs an error message as it might be a reasonable case? So instead of synchronization, just consider that particular status an expected and properly handled exception condition and not print an error? Thoughts? - Steve _______________________________________________ ath6kl mailing list ath6kl@xxxxxxxxxxxxxxxxxxx http://lists.infradead.org/mailman/listinfo/ath6kl