From: Francois Romieu <romieu@fr.zoreil.com>
Date: Wed, 18 Mar 2009 23:36:29 +0100

quoted text
> Karsten Wiese <fzu@wemgehoertderstaat.de> :
> > Am Mittwoch 18 März 2009 schrieb Francois Romieu:
> [...]
> > > If netif_rx_schedule fails after napi is enabled, there is a
> > > racing poll thread to care for the transfer.
> > > 
> > > Stated differently the bug is noticeable because napi_enable() 
> > > should be called before request_irq() in rtl8169_open, right ?
> > 
> > Propably yes. That would mean the first interrupt is received before
> > rtl_hw_start() is called in rtl8169_open. The source of that
> > interrupt would be something else then?
> 
> All we need is thus a shared interrupt racing on a different CPU
> with rtl8169_open and some IO versus memory reordering so that
> the hardware is started and a change in the status register can
> be seen in the IRQ handler before napi_enable/NAPI_STATE_SCHED
> escapes from the CPU where rtl8169_open is running...
> 
> It does not look like a credible explanation. :o/

Agreed, looking at how this part of the driver works I can't
see how this condition can arise either.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html