On Tue, Feb 08, 2011 at 01:23:06PM +0000, Russell King wrote: > On Tue, Feb 08, 2011 at 02:16:11PM +0100, Sebastian Andrzej Siewior wrote: > >>> @@ -850,6 +851,11 @@ static void enqueue_an_ATL_packet(struct usb_hcd > >>> *hcd, struct isp1760_qh *qh, > >>> skip_map&= ~queue_entry; > >>> isp1760_writel(skip_map, hcd->regs + HC_ATL_PTD_SKIPMAP_REG); > >>> > >>> + priv->atl_queued++; > >>> + if (priv->atl_queued == 2) > >>> + isp1760_writel(INTERRUPT_ENABLE_MASK, > >>> + hcd->regs + INTERRUPT_ENABLE_SOT_MASK); > >> > >> Perhaps it's: > >> > >> isp1760_writel(INTERRUPT_ENABLE_[SOT_]MASK, > >> hcd->regs + HC_INTERRUPT_ENABLE); > > > > No, it's not perhaps it is what it should be. Very good catch, thank you > > Sergei. > > > >> Othrwise it doesn't make much sense... > > > > Right. But now I wonder if this patch is still required since it worked > > without the fix. However Russell complained about a short delay. > > Maybe the additional writes - even though they don't change any bits in > the register - mean that the race condition is less likely to occur due > to timing differences. Note that it would be a _significant_ timing disturbance as INTERRUPT_ENABLE_SOT_MASK corresponds to 0x8a, which will cause a misaligned access to an IO address. That may result in the alignment handler being invoked, which will be much longer than just a single write: root@vexpress:~# cat /proc/cpu/alignment ... System: 217 ... Word: 217 ... That's after a few seconds of running minicom on the ftdi_sio port. -- Russell King Linux kernel 2.6 ARM Linux - http://www.arm.linux.org.uk/ maintainer of: -- 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
