[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH v9 1/3] xen/libxc: Allow changing max number of hypervisor cpuid leaves

On 04/23/2014 08:36 PM, Zhang, Yang Z wrote:
Jan Beulich wrote on 2014-04-23:
I'll definitely want a VMX maintainer's ack (or at least Yang's consent) on patch 3 before applying the whole series (the first two patches alone don't make that much sense).
I have discussed with Boris before (offlist) and I don't think it is necessary 
to expose two bits. Per my understanding, only one bit to indicate whether 
APICv is enabled should be enough. The reason is that all hardwires that 
support APICv must also support virtualized x2apic(Intel SDM doesn't say it 
explicitly, but I get confirmed from hardware guys). And in current Xen's 
implementation, it will enable APICv and virtualize_x2apic unconditionally if 
hardware support them and not disabled by user. So based on current xen's 
implementation, APICv is enabled means virtulized x2apic is enabled, vice 
versa. That's the reason why I think one bit is enough.
But if you guys think two bits also is acceptable, I am ok.

Jan, how do you want to proceed? Although I didn't realize that x2apic virtualization is guaranteed for APICv-enabled silicon I still prefer to keep separate bits for x2apic vs plain APIC for guests that can only use the latter.

(I know that I need to resend this in any case with stray printk removed)


Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.