On Mon, Dec 14, 2015 at 02:23:56PM -0200, Eduardo Habkost wrote: > > -#define MCE_CAP_DEF (MCG_CTL_P|MCG_SER_P) > > +#define MCE_CAP_DEF (MCG_CTL_P|MCG_SER_P|MCG_LMCE_P) > > This makes mcg_cap change when upgrading QEMU. > > VMs with MCG_LMCE_P enabled shouldn't be migratable to hosts > running older kernels, or the guest may try to read or write > MSR_IA32_MCG_EXT_CTL after miration and get a #GP. That means: > > 1) Older machine-types (pc-2.5 and older) should keep the > old (MCG_CTL_P|MCG_SER_P) default. > 2) We can't make pc-2.6 enable LMCE by default, either, because > QEMU guarantees that just changing the machine-type shouldn't > introduce new host requirements (see: > http://article.gmane.org/gmane.comp.emulators.qemu/346651) > > It looks like we need a new -cpu option to enable the feature, > then. At least until we raise the minimum kernel version > requirements of QEMU. ... and obviously LMCE is vendor-specific so it cannot be enabled on !Intel guests with a define like that. mce_init() in qemu should check vendor too. The same mistake was done with SER_P but that's much harder to change, as we discussed previously. -- Regards/Gruss, Boris. SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nürnberg) -- -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
