Here is a patch to enable Xen to run on a Unisys ES7000 x86_64 system. Signed off by Aravindh Puthiyaparambil <aravindh.puthiyaparambil@unisys.com> _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
On 19 Jul 2005, at 00:02, Puthiyaparambil, Aravindh wrote:> Here is a patch to enable Xen to run on a Unisys ES7000 x86_64 system. > > Signed off by Aravindh Puthiyaparambil > <aravindh.puthiyaparambil@unisys.com>We want to stick close to Linux 2.6 code. Comparing your code with 2.6.12: 1. MAX_IO_APICS is never set to 65 or 129 in 2.6.12. Do you really have systems with 129 IO APICs? x86_64 sets it to 128 always; i386 to 32. 2. setup_ioapic_ids_from_mpc: 2.6.12 gates this function on a genapic hook, not an explicit CPU version check. Why the difference? 3. mp_register_ioapic/mp_register_gsi: I see no equivalent of your changes in the 2.6.12 tree. Are they awaiting inclusion, or are they in a 2.6.13-rc tree that I ought to be looking at? I would like the patch closer to 2.6.12, or to know what mainline Linux tree you based it on. -- Keir _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
> 1. MAX_IO_APICS is never set to 65 or 129 in 2.6.12. Do you really > have systems with 129 IO APICs? x86_64 sets it to 128 always; i386 to > 32.We currently have systems in our lab with 65 IOAPICs and expect to ship them in the near future. As this architecture is further extended, systems with 129 IOAPICs are feasible. For the Linux kernel this issue was tracked by OSDL bugzilla #3754 at http://bugme.osdl.org/show_bug.cgi?id=3754 Len Brown has applied this patch to the acpi-test tree. He had not promoted it into 2.6.12 because it was very late in the development period for 2.6.12. We expect that Len will promote this for 2.6.13> 2. setup_ioapic_ids_from_mpc: 2.6.12 gates this function on agenapic> hook, not an explicit CPU version check. Why the difference?The check on unique IO-APIC ID is changed for setup-ioapic_ids_from_mpc() in 2.6.13-rc3 to be heuristic also: http://www.kernel.org/diff/diffview.cgi?file=%2Fpub%2Flinux%2Fkernel%2Fv 2.6%2Ftesting%2Fpatch-2.6.13-rc3.bz2;z=425 , see discussion at http://www.ussg.iu.edu/hypermail/linux/kernel/0505.1/index.html (search for "Do not enforce unique APIC ID"). Andi Kleen argued that heuristic way is better, since it will cover 32 bit platform variety more accurately. Among x86_64, there is no sub-platform that needs this check.> 3. mp_register_ioapic/mp_register_gsi: I see no equivalent of your > changes in the 2.6.12 tree. Are they awaiting inclusion, or are theyin> a 2.6.13-rc tree that I ought to be looking at?On mp_register_gsi(), the final patch is in 12.6.13-rc3-mm1 (http://www.kernel.org/diff/diffview.cgi?file=%2Fpub%2Flinux%2Fkernel%2F people%2Fakpm%2Fpatches%2F2.6%2F2.6.13-rc3%2F2.6.13-rc3-mm1%2F2.6.13-rc3 -mm1.bz2;z=138 ; http://www.kernel.org/diff/diffview.cgi?file=%2Fpub%2Flinux%2Fkernel%2Fp eople%2Fakpm%2Fpatches%2F2.6%2F2.6.13-rc3%2F2.6.13-rc3-mm1%2F2.6.13-rc3- mm1.bz2;z=341 Those are updated IRQ patches that contain fixes for VIA chipset. The original version of this patch is already in 2.6.13-rc3 http://www.kernel.org/diff/diffview.cgi?file=%2Fpub%2Flinux%2Fkernel%2Fv 2.6%2Ftesting%2Fpatch-2.6.13-rc3.bz2;z=429 ; http://www.kernel.org/diff/diffview.cgi?file=%2Fpub%2Flinux%2Fkernel%2Fv 2.6%2Ftesting%2Fpatch-2.6.13-rc3.bz2;z=940 Please let me know if you have any more questions about these patches. I hope you will apply them without waiting for 2.6.13 to be released :-) Aravindh _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Outlook messed up the URLs in my previous reply. Hope it''s better now.> 1. MAX_IO_APICS is never set to 65 or 129 in 2.6.12. Do you really > have systems with 129 IO APICs? x86_64 sets it to 128 always; i386 to > 32.We currently have systems in our lab with 65 IOAPICs and expect to ship them in the near future. As this architecture is further extended, systems with 129 IOAPICs are feasible. For the Linux kernel this issue was tracked by OSDL bugzilla #3754 at http://bugme.osdl.org/show_bug.cgi?id=3754 Len Brown has applied this patch to the acpi-test tree. He had not promoted it into 2.6.12 because it was very late in the development period for 2.6.12. We expect that Len will promote this for 2.6.13> 2. setup_ioapic_ids_from_mpc: 2.6.12 gates this function on a > genapic hook, not an explicit CPU version check. Why the difference?The check on unique IO-APIC ID is changed for setup-ioapic_ids_from_mpc() in 2.6.13-rc3 to be heuristic also: http://www.kernel.org/diff/diffview.cgi?file=%2Fpub%2Flinux%2Fkernel%2Fv 2.6%2Ftesting%2Fpatch-2.6.13-rc3.bz2;z=425 See discussion at http://www.ussg.iu.edu/hypermail/linux/kernel/0505.1/index.html search for "Do not enforce unique APIC ID"). Andi Kleen argued that heuristic way is better, since it will cover 32 bit platform variety more accurately. Among x86_64, there is no sub-platform that needs this check.> 3. mp_register_ioapic/mp_register_gsi: I see no equivalent of your > changes in the 2.6.12 tree. Are they awaiting inclusion, or are they > in a 2.6.13-rc tree that I ought to be looking at?On mp_register_gsi(), the final patch is in 12.6.13-rc3-mm1 http://www.kernel.org/diff/diffview.cgi?file=%2Fpub%2Flinux%2Fkernel%2Fp eople%2Fakpm%2Fpatches%2F2.6%2F2.6.13-rc3%2F2.6.13-rc3-mm1%2F2.6.13-rc3- mm1.bz2;z=138 http://www.kernel.org/diff/diffview.cgi?file=%2Fpub%2Flinux%2Fkernel%2Fp eople%2Fakpm%2Fpatches%2F2.6%2F2.6.13-rc3%2F2.6.13-rc3-mm1%2F2.6.13-rc3- mm1.bz2;z=341 Those are updated IRQ patches that contain fixes for VIA chipset. The original version of this patch is already in 2.6.13-rc3 http://www.kernel.org/diff/diffview.cgi?file=%2Fpub%2Flinux%2Fkernel%2Fv 2.6%2Ftesting%2Fpatch-2.6.13-rc3.bz2;z=429 http://www.kernel.org/diff/diffview.cgi?file=%2Fpub%2Flinux%2Fkernel%2Fv 2.6%2Ftesting%2Fpatch-2.6.13-rc3.bz2;z=940 Please let me know if you have any more questions about these patches. I hope you will apply them without waiting for 2.6.13 to be released :-) Aravindh _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel