From f49b231d18ed4b54645570ed3cfeed004fd9bb7f Mon Sep 17 00:00:00 2001 From: Eduardo Habkost Date: Thu, 15 Jul 2010 19:32:14 -0300 Subject: [PATCH 7/9] Change default CPU model (qemu64) to model=6 RH-Author: Eduardo Habkost Message-id: <1279222335-10031-2-git-send-email-ehabkost@redhat.com> Patchwork-id: 10742 O-Subject: [PATCH 1/2] Change default CPU model (qemu64) to model=6 Bugzilla: 614537 RH-Acked-by: Gleb Natapov RH-Acked-by: Andrea Arcangeli RH-Acked-by: john cooper This is a port of a patch we included on RHEL5. Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=614537 Upstream status: I am still trying to understand what's the status of the RHEL5 patch, and why it was not accepted in the first place. Andrea has more details Original patch description below: -------------- enable sysenter on 32bit guests (skype fix) #2 Update, see comment on model=3... such subtle details there seem to be in windows, and then somebody on qemu-devel thought they could use a brand not-existing qemu new vendor_id != intel/amd with the potential for bigger screwup of all the guest OS... (not just windows) to fix this ;). NOTE: you should check the patch applied with no offset-warnings because this risks to go to modify some other cpu definition and not qemu64 if there are offsets (if applied on current source it changes athlon because qemu64 is 6,3,3 in rhel5/master). This is not against rhel5/master but against rhel5/master minus the previous version of this patch and it obsoletes the previous patch. I guess it also invalidates all my testing (however intel/windows32 + skype has already been tested, and linux32/64 is sure fine as well on intel). I think I only need to re-test windows32+skype on svm and linux32 on svm, and linux64 with compatibility mode again on svm. ------------ From: Andrea Arcangeli model=2 is not existent when vendor is intel and an errata of P6 says that any model <= 2 when family is 6 lack sap feature, so windows and linux 32bit guests disable sap in software and slowdown for no good reason when running inside kvm on intel CPU. model=3 would fix sysenter troubles but it prevents windows XP from using all cpus, so model=6 matches "Athlon (PM core)" on AMD and "P2 with on-die L2 cache" on Intel and it allows windows to use all CPUs as well as fixing sysenter. Patch will follow, workaround without this fix is -cpu qemu64,model=6. There is a bug in skype that it checks if the sep feature is set to run sysenter on intel chip inside its binary without passing through ntdll, without verifying the model was <= 2 like windows does, so windows forbids sysenter but skype calls it anyway. This is skype bug and it would trigger on real P6 hardware too (but nobody tests skype on P6). qemu64,-sep fixed skype. But because every time we have an intel cpu running kvm, we also have sysenter functioning, we boots the model to 6 so windows 32bit (and linux32bit) gets a boost with sysenter too. Signed-off-by: Eduardo Habkost --- target-i386/helper.c | 3 ++- 1 files changed, 2 insertions(+), 1 deletions(-) Signed-off-by: Eduardo Habkost --- target-i386/helper.c | 3 ++- 1 files changed, 2 insertions(+), 1 deletions(-) diff --git a/target-i386/helper.c b/target-i386/helper.c index e271b31..0b3b239 100644 --- a/target-i386/helper.c +++ b/target-i386/helper.c @@ -219,7 +219,8 @@ static x86_def_t builtin_x86_defs[] = { .vendor2 = CPUID_VENDOR_AMD_2, .vendor3 = CPUID_VENDOR_AMD_3, .family = 6, - .model = 2, + /* Athlon (PM core) || P2 with on-die L2 cache - P6 has no sep */ + .model = 6, .stepping = 3, .features = PPRO_FEATURES | /* these features are needed for Win64 and aren't fully implemented */ -- 1.7.0.3