From mboxrd@z Thu Jan 1 00:00:00 1970 From: tglx@linutronix.de (Thomas Gleixner) Date: Thu, 28 Apr 2011 21:48:52 +0200 (CEST) Subject: AT91: Convert to fasteoi IRQ handler, and remove ARM irq_finish In-Reply-To: <20110428192403.GA21061@linutronix.de> References: <1303909742.18407.12.camel@redbox> <20110428151819.GE13755@siel.b> <20110428192403.GA21061@linutronix.de> Message-ID: To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On Thu, 28 Apr 2011, Sebastian Andrzej Siewior wrote: > * Andrew Victor | 2011-04-28 19:12:23 [+0200]: > > >hi, > > > >>> - ? ? desc->irq_data.chip->irq_unmask(&desc->irq_data); > >>> - ? ? /* now it may re-trigger */ > >>> + ? ? /* acknowledge interrupt - now it may re-trigger */ > >>> + ? ? desc->irq_data.chip->irq_eoi(&desc->irq_data); > >>> ?} > >> > >> why does this handler mess with the parent irq chip ? > >> this looks pretty wrong to me. > > > >This is a chained-handler, and without that line the end-of-interrupt > >doesn't seem to be signalled. > > Most chain handler don't mess with other irq chips. Why is the parent > irq hanlder set level and not eoi? Wouldn't this make this superflous? > > btw: tglx removed direct desc access so this patch won't apply. It does, you still can access irq_data, but you should use the wrappers. Thanks, tglx