Subject: Re: [RFC PATCH 3/3] x86-64: Calgary IOMMU - hook it in - msg#00011

Previous Message by Date: click to view message preview

Re: [RFC PATCH 3/3] x86-64: Calgary IOMMU - hook it in

On Út 14-03-06 10:26:34, Muli Ben-Yehuda wrote: > This patch hooks Calgary into the build and the x86-64 IOMMU > initialization paths. > > Signed-Off-By: Muli Ben-Yehuda <mulix@xxxxxxxxx> > Signed-Off-By: Jon Mason <jdmason@xxxxxxxxxx> > > diff -Naurp --exclude-from /home/muli/w/dontdiff > iommu_detected/arch/x86_64/Kconfig linux/arch/x86_64/Kconfig > --- iommu_detected/arch/x86_64/Kconfig 2006-03-14 08:58:23.000000000 > +0200 > +++ linux/arch/x86_64/Kconfig 2006-03-12 10:49:04.000000000 +0200 > @@ -372,6 +372,16 @@ config GART_IOMMU > and a software emulation used on other systems. > If unsure, say Y. > > +config CALGARY_IOMMU > + bool "IBM x366 server IOMMU" > + default y > + depends on PCI && MPSC && EXPERIMENTAL > + help > + Support for hardware IOMMUs in IBM's x366 server > + systems. The IOMMU can be turned off at runtime with the > + iommu=off parameter. Normally the kernel will make the right Runtime? I think you meant boottime. > + choice by itself. If unsure, say Y. Eh? How common are those machines? Pavel -- 32: bw.Write( sbuffer );

Next Message by Date: click to view message preview

Re: [RFC PATCH 3/3] x86-64: Calgary IOMMU - hook it in

On Út 14-03-06 17:22:48, Jon Mason wrote: > On Wed, Mar 15, 2006 at 12:03:48AM +0100, Pavel Machek wrote: > > On ?t 14-03-06 10:26:34, Muli Ben-Yehuda wrote: > > > This patch hooks Calgary into the build and the x86-64 IOMMU > > > initialization paths. > > > > > > Signed-Off-By: Muli Ben-Yehuda <mulix@xxxxxxxxx> > > > Signed-Off-By: Jon Mason <jdmason@xxxxxxxxxx> > > > > > > diff -Naurp --exclude-from /home/muli/w/dontdiff > > > iommu_detected/arch/x86_64/Kconfig linux/arch/x86_64/Kconfig > > > --- iommu_detected/arch/x86_64/Kconfig 2006-03-14 08:58:23.000000000 > > > +0200 > > > +++ linux/arch/x86_64/Kconfig 2006-03-12 10:49:04.000000000 +0200 > > > @@ -372,6 +372,16 @@ config GART_IOMMU > > > and a software emulation used on other systems. > > > If unsure, say Y. > > > > > > +config CALGARY_IOMMU > > > + bool "IBM x366 server IOMMU" > > > + default y > > > + depends on PCI && MPSC && EXPERIMENTAL > > > + help > > > + Support for hardware IOMMUs in IBM's x366 server > > > + systems. The IOMMU can be turned off at runtime with the > > > + iommu=off parameter. Normally the kernel will make the right > > > > Runtime? I think you meant boottime. > > Yes, thanks for pointing it out. > > > > > > + choice by itself. If unsure, say Y. > > > > Eh? How common are those machines? > > While this code is specific to IBM's xSeries systems, it will not hurt > to have it enabled on other systems. The code is intelligent enough to > detect the existence of Calgary chips and, if not there, will go down > the standard path of no-iommu (providing that swiotlb has not been > specified at boottime). If this isn't clear enough in the description, > I can remedy that. No, it was pretty clear. But unless these machines are pretty common, I'd suggest users to say N. ... its like most drivers, it takes space but no other harm. Still we don't want to say "say Y" on all drivers. ...it is not required for boot on IBM x366 machines, is it? Pavel -- 56: MD5.TransformBlock( adNAME, 0, strName.IndexOf( '\0' ), adNAME, 0 );

Previous Message by Thread: click to view message preview

Re: [RFC PATCH 3/3] x86-64: Calgary IOMMU - hook it in

On Út 14-03-06 10:26:34, Muli Ben-Yehuda wrote: > This patch hooks Calgary into the build and the x86-64 IOMMU > initialization paths. > > Signed-Off-By: Muli Ben-Yehuda <mulix@xxxxxxxxx> > Signed-Off-By: Jon Mason <jdmason@xxxxxxxxxx> > > diff -Naurp --exclude-from /home/muli/w/dontdiff > iommu_detected/arch/x86_64/Kconfig linux/arch/x86_64/Kconfig > --- iommu_detected/arch/x86_64/Kconfig 2006-03-14 08:58:23.000000000 > +0200 > +++ linux/arch/x86_64/Kconfig 2006-03-12 10:49:04.000000000 +0200 > @@ -372,6 +372,16 @@ config GART_IOMMU > and a software emulation used on other systems. > If unsure, say Y. > > +config CALGARY_IOMMU > + bool "IBM x366 server IOMMU" > + default y > + depends on PCI && MPSC && EXPERIMENTAL > + help > + Support for hardware IOMMUs in IBM's x366 server > + systems. The IOMMU can be turned off at runtime with the > + iommu=off parameter. Normally the kernel will make the right Runtime? I think you meant boottime. > + choice by itself. If unsure, say Y. Eh? How common are those machines? Pavel -- 32: bw.Write( sbuffer );

Next Message by Thread: click to view message preview

Re: [RFC PATCH 3/3] x86-64: Calgary IOMMU - hook it in

On Út 14-03-06 17:22:48, Jon Mason wrote: > On Wed, Mar 15, 2006 at 12:03:48AM +0100, Pavel Machek wrote: > > On ?t 14-03-06 10:26:34, Muli Ben-Yehuda wrote: > > > This patch hooks Calgary into the build and the x86-64 IOMMU > > > initialization paths. > > > > > > Signed-Off-By: Muli Ben-Yehuda <mulix@xxxxxxxxx> > > > Signed-Off-By: Jon Mason <jdmason@xxxxxxxxxx> > > > > > > diff -Naurp --exclude-from /home/muli/w/dontdiff > > > iommu_detected/arch/x86_64/Kconfig linux/arch/x86_64/Kconfig > > > --- iommu_detected/arch/x86_64/Kconfig 2006-03-14 08:58:23.000000000 > > > +0200 > > > +++ linux/arch/x86_64/Kconfig 2006-03-12 10:49:04.000000000 +0200 > > > @@ -372,6 +372,16 @@ config GART_IOMMU > > > and a software emulation used on other systems. > > > If unsure, say Y. > > > > > > +config CALGARY_IOMMU > > > + bool "IBM x366 server IOMMU" > > > + default y > > > + depends on PCI && MPSC && EXPERIMENTAL > > > + help > > > + Support for hardware IOMMUs in IBM's x366 server > > > + systems. The IOMMU can be turned off at runtime with the > > > + iommu=off parameter. Normally the kernel will make the right > > > > Runtime? I think you meant boottime. > > Yes, thanks for pointing it out. > > > > > > + choice by itself. If unsure, say Y. > > > > Eh? How common are those machines? > > While this code is specific to IBM's xSeries systems, it will not hurt > to have it enabled on other systems. The code is intelligent enough to > detect the existence of Calgary chips and, if not there, will go down > the standard path of no-iommu (providing that swiotlb has not been > specified at boottime). If this isn't clear enough in the description, > I can remedy that. No, it was pretty clear. But unless these machines are pretty common, I'd suggest users to say N. ... its like most drivers, it takes space but no other harm. Still we don't want to say "say Y" on all drivers. ...it is not required for boot on IBM x366 machines, is it? Pavel -- 56: MD5.TransformBlock( adNAME, 0, strName.IndexOf( '\0' ), adNAME, 0 );