On 05/09/17 11:03, Peter Zijlstra wrote:
> On Tue, Sep 05, 2017 at 10:52:06AM +0200, Juergen Gross wrote:
>>> Hmm, that might work. Could we somehow nop that call when
>>> !X86_FEATURE_HYPERVISOR?, that saves native from having to do the call
>>> and would be a win for everyone.
>>
>> So in fact we want a "always false" shortcut for bare metal and for any
>> virtualization environment selecting bare metal behavior.
>>
>> I'll have a try.
>
> Right, so for native I think you can do this:
>
>
> diff --git a/arch/x86/kernel/paravirt_patch_64.c b/arch/x86/kernel/paravirt_patch_64.c
> index 11aaf1eaa0e4..4e9839001291 100644
> --- a/arch/x86/kernel/paravirt_patch_64.c
> +++ b/arch/x86/kernel/paravirt_patch_64.c
> @@ -21,6 +21,7 @@ DEF_NATIVE(, mov64, "mov %rdi, %rax");
> #if defined(CONFIG_PARAVIRT_SPINLOCKS)
> DEF_NATIVE(pv_lock_ops, queued_spin_unlock, "movb $0, (%rdi)");
> DEF_NATIVE(pv_lock_ops, vcpu_is_preempted, "xor %rax, %rax");
> +DEF_NATIVE(pv_lock_ops, virt_spin_lock, "xor %rax, %rax");
> #endif
>
> unsigned paravirt_patch_ident_32(void *insnbuf, unsigned len)
> @@ -77,6 +78,13 @@ unsigned native_patch(u8 type, u16 clobbers, void *ibuf,
> goto patch_site;
> }
> goto patch_default;
> + case PARAVIRT_PATCH(pv_lock_ops.virt_spin_lock):
> + if (!this_cpu_has(X86_FEATURE_HYPERVISOR)) {
> + start = start_pv_lock_ops_virt_spin_lock;
> + end = end_pv_lock_ops_virt_spin_lock;
> + goto patch_side;
> + }
> + goto patch_default;
I'd rather add a test to paravirt_patch_default() similar to the
_paravirt_ident_* cases for catching the Xen (and possibly KVM) cases,
too.
I just need to add _paravirt_zero_[32|64] functions I can use for bare
metal and the nopvspin cases.
Juergen
--
To unsubscribe from this list: send the line "unsubscribe linux-doc" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
