Re: [patch take 2][Intel-IOMMU] Fix for IOMMU early crash

2007-10-04 Thread Benjamin Herrenschmidt
> Subject: [Intel-IOMMU] Fix for IOMMU early crash > > pci_dev's->sysdata is highly overloaded and currently > IOMMU is broken due to IOMMU code depending on this field. > > This patch introduces new field in pci_dev's dev.archdata struct to > hold IOMMU specific per device IOMMU private data.

Re: [patch take 2][Intel-IOMMU] Fix for IOMMU early crash

2007-10-04 Thread Keshavamurthy, Anil S
On Thu, Oct 04, 2007 at 01:39:39PM +1000, Benjamin Herrenschmidt wrote: > > > > Why don't you use the new struct dev_archdata mechanism ? That's what I > > > use on powerpc to provide optional iommu linkage to any device in the > > > system. > > Good one. I will certainly try out your idea and

Re: [patch take 2][Intel-IOMMU] Fix for IOMMU early crash

2007-10-04 Thread Keshavamurthy, Anil S
On Thu, Oct 04, 2007 at 01:39:39PM +1000, Benjamin Herrenschmidt wrote: Why don't you use the new struct dev_archdata mechanism ? That's what I use on powerpc to provide optional iommu linkage to any device in the system. Good one. I will certainly try out your idea and will update

Re: [patch take 2][Intel-IOMMU] Fix for IOMMU early crash

2007-10-04 Thread Benjamin Herrenschmidt
Subject: [Intel-IOMMU] Fix for IOMMU early crash pci_dev's-sysdata is highly overloaded and currently IOMMU is broken due to IOMMU code depending on this field. This patch introduces new field in pci_dev's dev.archdata struct to hold IOMMU specific per device IOMMU private data.

Re: [patch take 2][Intel-IOMMU] Fix for IOMMU early crash

2007-10-03 Thread Benjamin Herrenschmidt
> > Why don't you use the new struct dev_archdata mechanism ? That's what I > > use on powerpc to provide optional iommu linkage to any device in the > > system. > Good one. I will certainly try out your idea and will update the list > tomorrow. The advantage is that it allows to completely

Re: [patch take 2][Intel-IOMMU] Fix for IOMMU early crash

2007-10-03 Thread Keshavamurthy, Anil S
On Thu, Oct 04, 2007 at 11:19:33AM +1000, Benjamin Herrenschmidt wrote: > > Index: 2.6-mm/include/linux/pci.h > > === > > --- 2.6-mm.orig/include/linux/pci.h 2007-10-03 13:48:20.0 -0700 > > +++ 2.6-mm/include/linux/pci.h

Re: [patch take 2][Intel-IOMMU] Fix for IOMMU early crash

2007-10-03 Thread Benjamin Herrenschmidt
> Index: 2.6-mm/include/linux/pci.h > === > --- 2.6-mm.orig/include/linux/pci.h 2007-10-03 13:48:20.0 -0700 > +++ 2.6-mm/include/linux/pci.h2007-10-03 13:49:08.0 -0700 > @@ -195,6 +195,7 @@ > #ifdef

[patch take 2][Intel-IOMMU] Fix for IOMMU early crash

2007-10-03 Thread Keshavamurthy, Anil S
Subject: [patch][Intel-IOMMU] Fix for IOMMU early crash pci_dev's->sysdata is highly overloaded and currently IOMMU is broken due to IOMMU code depending on this field. This patch introduces new field in pci_dev's struct to hold IOMMU specific per device IOMMU private data. Signed-off-by: Anil

[patch take 2][Intel-IOMMU] Fix for IOMMU early crash

2007-10-03 Thread Keshavamurthy, Anil S
Subject: [patch][Intel-IOMMU] Fix for IOMMU early crash pci_dev's-sysdata is highly overloaded and currently IOMMU is broken due to IOMMU code depending on this field. This patch introduces new field in pci_dev's struct to hold IOMMU specific per device IOMMU private data. Signed-off-by: Anil S

Re: [patch take 2][Intel-IOMMU] Fix for IOMMU early crash

2007-10-03 Thread Benjamin Herrenschmidt
Index: 2.6-mm/include/linux/pci.h === --- 2.6-mm.orig/include/linux/pci.h 2007-10-03 13:48:20.0 -0700 +++ 2.6-mm/include/linux/pci.h2007-10-03 13:49:08.0 -0700 @@ -195,6 +195,7 @@ #ifdef CONFIG_PCI_MSI

Re: [patch take 2][Intel-IOMMU] Fix for IOMMU early crash

2007-10-03 Thread Keshavamurthy, Anil S
On Thu, Oct 04, 2007 at 11:19:33AM +1000, Benjamin Herrenschmidt wrote: Index: 2.6-mm/include/linux/pci.h === --- 2.6-mm.orig/include/linux/pci.h 2007-10-03 13:48:20.0 -0700 +++ 2.6-mm/include/linux/pci.h

Re: [patch take 2][Intel-IOMMU] Fix for IOMMU early crash

2007-10-03 Thread Benjamin Herrenschmidt
Why don't you use the new struct dev_archdata mechanism ? That's what I use on powerpc to provide optional iommu linkage to any device in the system. Good one. I will certainly try out your idea and will update the list tomorrow. The advantage is that it allows to completely isolate the