On Fri, 22 Jul 2011 17:01:11 +0200 Alexander Graf <ag...@suse.de> wrote:
> On 22.07.2011, at 16:08, Elie Richa wrote: > > > > #if MAX_IPI > 0 > > > case 0x40: /* IDE */ > > > case 0x50: > > > idx = (addr - 0x40) >> 4; > > > retval = read_IRQreg(opp, opp->irq_ipi0 + idx, IRQ_IDE); > > > break; > > > #endif > > > > These are the IPI dispatch registers which are write only, so I suppose > > this shouldn't be here right? > > The code was there long before me :). No idea why it is there though - it > tries to read out the IDE register for 2 IPIs. Maybe it was read-write in > early versions of MPIC? Scott, any idea? I don't know about early MPICs, but I don't know how you'd even define that register for read -- it's a command, not state. I suspect it's just a bug -- especially given the initialization to 1, which suggests the author saw it as just another destination register. -Scott
