Re: [Xen-devel] Re: [Xen-ia64-devel] [PATCH 0/3] IA64: add cpufreq support
On Thu, Oct 09, 2008 at 11:37:23AM +0800, Yu, Ke wrote: Oh, looks like the BIOS is broken. From the _PSS object , only the P0 state is valid, other two P states P1 and P2 are all zeros. This is not correct. Can you try another machine? Not only DSDT, but also MADE was broken, I confirmed that xenpm worked by patching those tables. That's okay. Is it assumed that the dom0 acpi layer sees real tables so that it understands all physical cpus? The assumption isn't true on xen/ia64. Xen/ia64 modifies MADT to tell dom0 how many vcpu are assigned to dom0. Please see acpi_update_lsapic() in xen/arch/ia64/xen/dom_fw_dom0.c Something needs to be done. thanks, -- yamahata ___ Xen-ia64-devel mailing list Xen-ia64-devel@lists.xensource.com http://lists.xensource.com/xen-ia64-devel
RE: [Xen-devel] Re: [Xen-ia64-devel] [PATCH 0/3] IA64: add cpufreq support
Isaku Yamahata wrote: On Thu, Oct 09, 2008 at 11:37:23AM +0800, Yu, Ke wrote: Oh, looks like the BIOS is broken. From the _PSS object , only the P0 state is valid, other two P states P1 and P2 are all zeros. This is not correct. Can you try another machine? Not only DSDT, but also MADE was broken, I confirmed that xenpm worked by patching those tables. That's okay. Is it assumed that the dom0 acpi layer sees real tables so that it understands all physical cpus? The assumption isn't true on xen/ia64. Xen/ia64 modifies MADT to tell dom0 how many vcpu are assigned to dom0. Please see acpi_update_lsapic() in xen/arch/ia64/xen/dom_fw_dom0.c Something needs to be done. thanks, Right, we already take this into consideration at the design phase. Fortunately, Only the DSDT is needed for Px state, which is unmodifed, so that is Ok for Px support. And for the acpi id to apic id maping, we dedicatedly parse the MADT before the acpi_update_lsapic, so the maping is also correct. Best Regards Ke ___ Xen-ia64-devel mailing list Xen-ia64-devel@lists.xensource.com http://lists.xensource.com/xen-ia64-devel
RE: [Xen-ia64-devel] add pci configuration code, which is needed by VTD
Isaku Yamahata wrote: On Sat, Sep 27, 2008 at 01:40:49PM +0800, Xu, Anthony wrote: Add pci configuration code, which is needed by VTD Signed-off-by; Anthony Xu [EMAIL PROTECTED] The new one, I put pci.c under arch/ia64/xen, since it only uses two functions of linux pci.c, I don't want to pull the whole pci.c in xen. I see. I understood why you ended in the patch. Hmm, looking at xen-unstable.hg/xen/arch/x86/pci.c, Only pci_conf_read()/pci_conf_write() are different, other functions seems same to x86's. So please split out same functions into arch common file, probably xen/drivers/pci/pci.c or something acceptable for Keir. Then add pci_conf_read()/pci_conf_write() in xen/arch/ia64/xen/pci.c. Good suggestion! I'll make the patches. BTW, how about the other two patches I sent out? Thanks, Anthony ___ Xen-ia64-devel mailing list Xen-ia64-devel@lists.xensource.com http://lists.xensource.com/xen-ia64-devel
Re: [Xen-devel] Re: [Xen-ia64-devel] [PATCH 0/3] IA64: add cpufreq support
On Thu, Oct 09, 2008 at 05:26:04PM +0800, Yu, Ke wrote: Isaku Yamahata wrote: On Thu, Oct 09, 2008 at 11:37:23AM +0800, Yu, Ke wrote: Oh, looks like the BIOS is broken. From the _PSS object , only the P0 state is valid, other two P states P1 and P2 are all zeros. This is not correct. Can you try another machine? Not only DSDT, but also MADE was broken, I confirmed that xenpm worked by patching those tables. That's okay. Is it assumed that the dom0 acpi layer sees real tables so that it understands all physical cpus? The assumption isn't true on xen/ia64. Xen/ia64 modifies MADT to tell dom0 how many vcpu are assigned to dom0. Please see acpi_update_lsapic() in xen/arch/ia64/xen/dom_fw_dom0.c Something needs to be done. thanks, Right, we already take this into consideration at the design phase. Fortunately, Only the DSDT is needed for Px state, which is unmodifed, so that is Ok for Px support. And for the acpi id to apic id maping, we dedicatedly parse the MADT before the acpi_update_lsapic, so the maping is also correct. Great. I confirmed it myself. I applied all your 3 patches. thanks, -- yamahata ___ Xen-ia64-devel mailing list Xen-ia64-devel@lists.xensource.com http://lists.xensource.com/xen-ia64-devel
Re: [Xen-ia64-devel] add pci configuration code, which is needed by VTD
On Thu, Oct 09, 2008 at 05:36:57PM +0800, Xu, Anthony wrote: Isaku Yamahata wrote: On Sat, Sep 27, 2008 at 01:40:49PM +0800, Xu, Anthony wrote: Add pci configuration code, which is needed by VTD Signed-off-by; Anthony Xu [EMAIL PROTECTED] The new one, I put pci.c under arch/ia64/xen, since it only uses two functions of linux pci.c, I don't want to pull the whole pci.c in xen. I see. I understood why you ended in the patch. Hmm, looking at xen-unstable.hg/xen/arch/x86/pci.c, Only pci_conf_read()/pci_conf_write() are different, other functions seems same to x86's. So please split out same functions into arch common file, probably xen/drivers/pci/pci.c or something acceptable for Keir. Then add pci_conf_read()/pci_conf_write() in xen/arch/ia64/xen/pci.c. Good suggestion! I'll make the patches. I applied the patch without two functions in order to fix compilation error. BTW, how about the other two patches I sent out? I applied both. thanks, -- yamahata ___ Xen-ia64-devel mailing list Xen-ia64-devel@lists.xensource.com http://lists.xensource.com/xen-ia64-devel
[Xen-ia64-devel] Please pull ia64 trees
Hi Keir, Please pull the ia64 trees: http://xenbits.xensource.com/ext/ia64/xen-unstable.hg http://xenbits.xensource.com/ext/ia64/linux-2.6.18-xen.hg This includes cpufreq support by Yu Ke and Liu Jinsong, some preliminary patches for VT-d support by Anthony Xu and severla bug fixes. Note: This touches the common file, linux/drivers/acpi/Kconfig, to compile necessary files. Thanks, -- yamahata ___ Xen-ia64-devel mailing list Xen-ia64-devel@lists.xensource.com http://lists.xensource.com/xen-ia64-devel
