| Message ID | 1443447044-24985-1-git-send-email-julien.grall@citrix.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Mon, 28 Sep 2015, Julien Grall wrote: > The PCI support for Xen doesn't compile on ARM/ARM64 when > CONFIG_PCI_MMCONFIG=y: > > drivers/xen/pci.c:31:25: fatal error: asm/pci_x86.h: No such file or > directory > #include <asm/pci_x86.h> > > Although, Xen is not currently involved in PCI management for ARM/ARM64. > There is plan to support it, but it would require some changes in Linux > side. > > For now, introduce a new config options XEN_PCI which will be turned off > for ARM platform. > > Reported-by: Robert Richter <robert.richter@caviumnetworks.com> > Signed-off-by: Julien Grall <julien.grall@citrix.com> > > --- > Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> > Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com> > Cc: David Vrabel <david.vrabel@citrix.com> > --- > drivers/xen/Kconfig | 6 +++++- > drivers/xen/Makefile | 2 +- > 2 files changed, 6 insertions(+), 2 deletions(-) > > diff --git a/drivers/xen/Kconfig b/drivers/xen/Kconfig > index 73708ac..6bcc8b0 100644 > --- a/drivers/xen/Kconfig > +++ b/drivers/xen/Kconfig > @@ -168,7 +168,7 @@ config XEN_TMEM > > config XEN_PCIDEV_BACKEND > tristate "Xen PCI-device backend driver" > - depends on PCI && X86 && XEN > + depends on XEN_PCI > depends on XEN_BACKEND > default m > help > @@ -189,6 +189,10 @@ config XEN_PCIDEV_BACKEND > > If in doubt, say m. > > +config XEN_PCI > + def_bool y > + depends on PCI && !ARM && !ARM64 This is confusing because x86 already has PCI_XEN, see arch/x86/Kconfig. Just move PCI_XEN from arch/x86/Kconfig to here and disable it for ARM? > config XEN_SCSI_BACKEND > tristate "XEN SCSI backend driver" > depends on XEN && XEN_BACKEND && TARGET_CORE > diff --git a/drivers/xen/Makefile b/drivers/xen/Makefile > index e293bc5..c9f23b7 100644 > --- a/drivers/xen/Makefile > +++ b/drivers/xen/Makefile > @@ -11,7 +11,7 @@ CFLAGS_features.o := $(nostackp) > > CFLAGS_efi.o += -fshort-wchar > > -dom0-$(CONFIG_PCI) += pci.o > +dom0-$(CONFIG_XEN_PCI) += pci.o > dom0-$(CONFIG_USB_SUPPORT) += dbgp.o > dom0-$(CONFIG_XEN_ACPI) += acpi.o $(xen-pad-y) > xen-pad-$(CONFIG_X86) += xen-acpi-pad.o > -- > 2.1.4 >
On 28/09/15 14:30, Julien Grall wrote: > The PCI support for Xen doesn't compile on ARM/ARM64 when > CONFIG_PCI_MMCONFIG=y: > > drivers/xen/pci.c:31:25: fatal error: asm/pci_x86.h: No such file or > directory > #include <asm/pci_x86.h> > > Although, Xen is not currently involved in PCI management for ARM/ARM64. > There is plan to support it, but it would require some changes in Linux > side. > > For now, introduce a new config options XEN_PCI which will be turned off > for ARM platform. Since PCI_MMCONFIG isn't x86 specific and its use in drivers/xen/pci.c is, can you fix this by changing to #ifdef CONFIG_X86? David
Hi David, On 28/09/15 14:52, David Vrabel wrote: > On 28/09/15 14:30, Julien Grall wrote: >> The PCI support for Xen doesn't compile on ARM/ARM64 when >> CONFIG_PCI_MMCONFIG=y: >> >> drivers/xen/pci.c:31:25: fatal error: asm/pci_x86.h: No such file or >> directory >> #include <asm/pci_x86.h> >> >> Although, Xen is not currently involved in PCI management for ARM/ARM64. >> There is plan to support it, but it would require some changes in Linux >> side. >> >> For now, introduce a new config options XEN_PCI which will be turned off >> for ARM platform. > > Since PCI_MMCONFIG isn't x86 specific and its use in drivers/xen/pci.c > is, can you fix this by changing to #ifdef CONFIG_X86? Well, it may be possible that we need this code for ARM/ARM64. From the origin thread [1], it wasn't clear what is the correct solution. So I think that disabling the Xen PCI code for now is the right solution. Note, that it's still possible to use PCI in DOM0 on ARM64 platform without this code. Regards, [1] https://lkml.org/lkml/2015/8/6/510
On 28/09/15 14:48, Stefano Stabellini wrote: > On Mon, 28 Sep 2015, Julien Grall wrote: >> The PCI support for Xen doesn't compile on ARM/ARM64 when >> CONFIG_PCI_MMCONFIG=y: >> >> drivers/xen/pci.c:31:25: fatal error: asm/pci_x86.h: No such file or >> directory >> #include <asm/pci_x86.h> >> >> Although, Xen is not currently involved in PCI management for ARM/ARM64. >> There is plan to support it, but it would require some changes in Linux >> side. >> >> For now, introduce a new config options XEN_PCI which will be turned off >> for ARM platform. >> >> Reported-by: Robert Richter <robert.richter@caviumnetworks.com> >> Signed-off-by: Julien Grall <julien.grall@citrix.com> >> >> --- >> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> >> Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com> >> Cc: David Vrabel <david.vrabel@citrix.com> >> --- >> drivers/xen/Kconfig | 6 +++++- >> drivers/xen/Makefile | 2 +- >> 2 files changed, 6 insertions(+), 2 deletions(-) >> >> diff --git a/drivers/xen/Kconfig b/drivers/xen/Kconfig >> index 73708ac..6bcc8b0 100644 >> --- a/drivers/xen/Kconfig >> +++ b/drivers/xen/Kconfig >> @@ -168,7 +168,7 @@ config XEN_TMEM >> >> config XEN_PCIDEV_BACKEND >> tristate "Xen PCI-device backend driver" >> - depends on PCI && X86 && XEN >> + depends on XEN_PCI >> depends on XEN_BACKEND >> default m >> help >> @@ -189,6 +189,10 @@ config XEN_PCIDEV_BACKEND >> >> If in doubt, say m. >> >> +config XEN_PCI >> + def_bool y >> + depends on PCI && !ARM && !ARM64 > > This is confusing because x86 already has PCI_XEN, see arch/x86/Kconfig. > Just move PCI_XEN from arch/x86/Kconfig to here and disable it for ARM? PCI_XEN enables SWIOTLB. Do we really want the same dependencies for any architecture? Regards,
On 09/28/2015 09:59 AM, Julien Grall wrote: > Hi David, > > On 28/09/15 14:52, David Vrabel wrote: >> On 28/09/15 14:30, Julien Grall wrote: >>> The PCI support for Xen doesn't compile on ARM/ARM64 when >>> CONFIG_PCI_MMCONFIG=y: >>> >>> drivers/xen/pci.c:31:25: fatal error: asm/pci_x86.h: No such file or >>> directory >>> #include <asm/pci_x86.h> >>> >>> Although, Xen is not currently involved in PCI management for ARM/ARM64. >>> There is plan to support it, but it would require some changes in Linux >>> side. >>> >>> For now, introduce a new config options XEN_PCI which will be turned off >>> for ARM platform. >> Since PCI_MMCONFIG isn't x86 specific and its use in drivers/xen/pci.c >> is, can you fix this by changing to #ifdef CONFIG_X86? > Well, it may be possible that we need this code for ARM/ARM64. > > From the origin thread [1], it wasn't clear what is the correct > solution. So I think that disabling the Xen PCI code for now is the > right solution. > > Note, that it's still possible to use PCI in DOM0 on ARM64 platform > without this code. > > Regards, > > [1] https://lkml.org/lkml/2015/8/6/510 > Can we then implement Robert's proposals from that thread and have an arch function for pci_probe? It will be a nop on ARM. (I think that was the only problem here). -boris
On 28/09/15 15:02, Boris Ostrovsky wrote: > On 09/28/2015 09:59 AM, Julien Grall wrote: >> Hi David, >> >> On 28/09/15 14:52, David Vrabel wrote: >>> On 28/09/15 14:30, Julien Grall wrote: >>>> The PCI support for Xen doesn't compile on ARM/ARM64 when >>>> CONFIG_PCI_MMCONFIG=y: >>>> >>>> drivers/xen/pci.c:31:25: fatal error: asm/pci_x86.h: No such file or >>>> directory >>>> #include <asm/pci_x86.h> >>>> >>>> Although, Xen is not currently involved in PCI management for >>>> ARM/ARM64. >>>> There is plan to support it, but it would require some changes in Linux >>>> side. >>>> >>>> For now, introduce a new config options XEN_PCI which will be turned >>>> off >>>> for ARM platform. >>> Since PCI_MMCONFIG isn't x86 specific and its use in drivers/xen/pci.c >>> is, can you fix this by changing to #ifdef CONFIG_X86? This. >> Well, it may be possible that we need this code for ARM/ARM64. >> >> From the origin thread [1], it wasn't clear what is the correct >> solution. So I think that disabling the Xen PCI code for now is the >> right solution. >> >> Note, that it's still possible to use PCI in DOM0 on ARM64 platform >> without this code. >> >> Regards, >> >> [1] https://lkml.org/lkml/2015/8/6/510 >> > > Can we then implement Robert's proposals from that thread and have an > arch function for pci_probe? It will be a nop on ARM. (I think that was > the only problem here). Or preferably this. David
On 28/09/15 15:02, Boris Ostrovsky wrote: > On 09/28/2015 09:59 AM, Julien Grall wrote: >> Hi David, >> >> On 28/09/15 14:52, David Vrabel wrote: >>> On 28/09/15 14:30, Julien Grall wrote: >>>> The PCI support for Xen doesn't compile on ARM/ARM64 when >>>> CONFIG_PCI_MMCONFIG=y: >>>> >>>> drivers/xen/pci.c:31:25: fatal error: asm/pci_x86.h: No such file or >>>> directory >>>> #include <asm/pci_x86.h> >>>> >>>> Although, Xen is not currently involved in PCI management for >>>> ARM/ARM64. >>>> There is plan to support it, but it would require some changes in Linux >>>> side. >>>> >>>> For now, introduce a new config options XEN_PCI which will be turned >>>> off >>>> for ARM platform. >>> Since PCI_MMCONFIG isn't x86 specific and its use in drivers/xen/pci.c >>> is, can you fix this by changing to #ifdef CONFIG_X86? >> Well, it may be possible that we need this code for ARM/ARM64. >> >> From the origin thread [1], it wasn't clear what is the correct >> solution. So I think that disabling the Xen PCI code for now is the >> right solution. >> >> Note, that it's still possible to use PCI in DOM0 on ARM64 platform >> without this code. >> >> Regards, >> >> [1] https://lkml.org/lkml/2015/8/6/510 >> > > Can we then implement Robert's proposals from that thread and have an > arch function for pci_probe? It will be a nop on ARM. (I think that was > the only problem here). Hmmmm. It looks like that the mmconfig is not yet supported upstream for ARM64. So even with arch function for pci_probe so it won't work. I hit the compilation issue on a 4.2 with patched applied for some hardware enablement. Some of them if for mmconfig. I should have check it before sending it. Sorry for the noise. Regards,
diff --git a/drivers/xen/Kconfig b/drivers/xen/Kconfig index 73708ac..6bcc8b0 100644 --- a/drivers/xen/Kconfig +++ b/drivers/xen/Kconfig @@ -168,7 +168,7 @@ config XEN_TMEM config XEN_PCIDEV_BACKEND tristate "Xen PCI-device backend driver" - depends on PCI && X86 && XEN + depends on XEN_PCI depends on XEN_BACKEND default m help @@ -189,6 +189,10 @@ config XEN_PCIDEV_BACKEND If in doubt, say m. +config XEN_PCI + def_bool y + depends on PCI && !ARM && !ARM64 + config XEN_SCSI_BACKEND tristate "XEN SCSI backend driver" depends on XEN && XEN_BACKEND && TARGET_CORE diff --git a/drivers/xen/Makefile b/drivers/xen/Makefile index e293bc5..c9f23b7 100644 --- a/drivers/xen/Makefile +++ b/drivers/xen/Makefile @@ -11,7 +11,7 @@ CFLAGS_features.o := $(nostackp) CFLAGS_efi.o += -fshort-wchar -dom0-$(CONFIG_PCI) += pci.o +dom0-$(CONFIG_XEN_PCI) += pci.o dom0-$(CONFIG_USB_SUPPORT) += dbgp.o dom0-$(CONFIG_XEN_ACPI) += acpi.o $(xen-pad-y) xen-pad-$(CONFIG_X86) += xen-acpi-pad.o
The PCI support for Xen doesn't compile on ARM/ARM64 when CONFIG_PCI_MMCONFIG=y: drivers/xen/pci.c:31:25: fatal error: asm/pci_x86.h: No such file or directory #include <asm/pci_x86.h> Although, Xen is not currently involved in PCI management for ARM/ARM64. There is plan to support it, but it would require some changes in Linux side. For now, introduce a new config options XEN_PCI which will be turned off for ARM platform. Reported-by: Robert Richter <robert.richter@caviumnetworks.com> Signed-off-by: Julien Grall <julien.grall@citrix.com> --- Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com> Cc: David Vrabel <david.vrabel@citrix.com> --- drivers/xen/Kconfig | 6 +++++- drivers/xen/Makefile | 2 +- 2 files changed, 6 insertions(+), 2 deletions(-)