Re: [PATCH] iommu: parse pci device scope even only intr remap is defined

[Andrew Vasquez <andrew.vasquez@qlogic.com>]

On Mon, 25 Jul 2011, Ingo Molnar wrote:

> * Yinghai Lu <yinghai@kernel.org> wrote:
> 
> > > This looks very messy.
> > >
> > > CONFIG_DMAR has no clear meaning. The DMAR table parsing 
> > > functionality is intermixed with the DMAR feature itself. The 
> > > kernel code is littered with a couple of dozen CONFIG_DMAR 
> > > #ifdefs with no clear structure to the initialization and to the 
> > > separation of functionality.
> > 
> > CONFIG_DMAR is actually DMA_REMAP instead DMAR table.
> > 
> > or Do you prefer to clean them up further with following depency?
> > 
> > CONFIG_DMAR_TBL for DMAR table
> > CONFIG_DMA_REMAP for DMA remapping
> > CONFIG_INTR_REMAP for Interrupt remapping
> > and XXX_REMAP will select DMAR_TBL
> 
> 'DMAR', 'TBL' and 'INTR' are all misnomers!
> 
> 	CONFIG_DMA_REMAP_TABLE
> 	CONFIG_DMA_REMAP
> 	CONFIG_IRQ_REMAP
> 
> That way we'd get the 'DMAR tables' via CONFIG_DMA_REMAP_TABLE - on 
> top of which enabling CONFIG_DMA_REMAP and CONFIG_IRQ_REMAP would 
> enable and handle various hw remapping features.
> 
> (Does anyone else have better/other code structure suggestions?)
> 
> But yes, we should first do this rename/cleanup to clarify what it 
> all means, then fix whatever config-combos don't work perfectly yet.

All,

Is this re-work effort going to make it into 3.1?  With 3.1.0-rc1, we
are seeing the same 'no interupts being routed' issue.  Applying
Yinghai's patch to 3.1-rc1:

	http://www.spinics.net/lists/linux-scsi/msg53416.html

get's interrupts routing again....

Thanks, AV