On Fri, 4 May 2018, Sebastian Andrzej Siewior wrote: > Hi Alan, > > I posted a RFC [0] to let the HCD complete the URB in threaded-IRQ > context to avoid the softirq/tasklet. Mauro reported that this does not > help him and never came back after that. > You did not oppose the approach as long as there is no performance issue > which I did not see. Would you mind if I revisit the approach and > convert all HCDs to threaded IRQs while at it? I don't mind revisiting the approach, although it certainly would be good to find out why it doesn't help Mauro's video problem. However, converting _all_ the HCDs to threaded IRQs is a different story. It should be okay to convert the ones that currently use tasklets, but I can't approve changes to all the others. Only the drivers that I maintain (ohci-hcd and uhci-hcd). And even for those two, the conversion will not be easy. Simply changing the giveback routines would violate the documented guarantees for isochronous transfers. 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
