Re: [PATCH][tip] x86/paravirt: Make the virt_spin_lock_key setup after jump_label_init()

[Juergen Gross]

On 27/10/17 19:09, Vitaly Kuznetsov wrote:
> Boris Petkov <bp@xxxxxxx> writes:
> 
>> On October 27, 2017 6:02:00 PM GMT+02:00, Dou Liyang <douly.fnst@xxxxxxxxxxxxxx> wrote:
>>> Commit:
>>>
>>>  9043442b43b1 ("locking/paravirt: Use new static key for controlling
>>>  call of virt_spin_lock()")
>>>
>>> set the static virt_spin_lock_key to a value before jump_label_init()
>>> has been called, which will result in a WARN().
>>>
>>> Move the native_pv_lock_init() into xx_smp_prepare_cpus(). Make the
>>> setup later to avoid the WARN().
>>>
>>> Reported-by: Juergen Gross <jgross@xxxxxxxx>
>>> Suggested-by: Juergen Gross <jgross@xxxxxxxx>
>>> Signed-off-by: Dou Liyang <douly.fnst@xxxxxxxxxxxxxx>
>>> ---
>>> arch/x86/kernel/smpboot.c | 3 ++-
>>> arch/x86/xen/smp_pv.c     | 2 ++
>>> arch/x86/xen/spinlock.c   | 6 ++++--
>>> 3 files changed, 8 insertions(+), 3 deletions(-)
>>>
>>> diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
>>> index aed1460..6b1335a 100644
>>> --- a/arch/x86/kernel/smpboot.c
>>> +++ b/arch/x86/kernel/smpboot.c
>>> @@ -1323,6 +1323,8 @@ void __init native_smp_prepare_cpus(unsigned int
>>> max_cpus)
>>> 	pr_info("CPU0: ");
>>> 	print_cpu_info(&cpu_data(0));
>>>
>>> +	native_pv_lock_init();
>>> +
>>> 	uv_system_init();
>>>
>>> 	set_mtrr_aps_delayed_init();
>>> @@ -1350,7 +1352,6 @@ void __init native_smp_prepare_boot_cpu(void)
>>> 	/* already set me in cpu_online_mask in boot_cpu_init() */
>>> 	cpumask_set_cpu(me, cpu_callout_mask);
>>> 	cpu_set_state_online(me);
>>> -	native_pv_lock_init();
>>> }
>>>
>>> void __init native_smp_cpus_done(unsigned int max_cpus)
>>> diff --git a/arch/x86/xen/smp_pv.c b/arch/x86/xen/smp_pv.c
>>> index 5147140..570b2bc 100644
>>> --- a/arch/x86/xen/smp_pv.c
>>> +++ b/arch/x86/xen/smp_pv.c
>>> @@ -236,6 +236,8 @@ static void __init xen_pv_smp_prepare_cpus(unsigned
>>> int max_cpus)
>>> 		xen_raw_printk(m);
>>> 		panic(m);
>>> 	}
>>> +	native_pv_lock_init();
>>> +
>>> 	xen_init_lock_cpu(0);
>>>
>>> 	smp_store_boot_cpu_info();
>>> diff --git a/arch/x86/xen/spinlock.c b/arch/x86/xen/spinlock.c
>>> index e8ab80a..1e1462d 100644
>>> --- a/arch/x86/xen/spinlock.c
>>> +++ b/arch/x86/xen/spinlock.c
>>> @@ -81,8 +81,11 @@ void xen_init_lock_cpu(int cpu)
>>> 	int irq;
>>> 	char *name;
>>>
>>> -	if (!xen_pvspin)
>>> +	if (!xen_pvspin) {
>>> +		if (cpu == 0)
>>> +			static_branch_disable(&virt_spin_lock_key);
>>
>> This is assuming CPU 0 is the boot cpu. I think you want boot_cpu_data.cpu_index here or whatever is used on xen to identify the BSP reliably. 
> 
> It seems both PV and PVHVM call xen_init_lock_cpu(0) so 0 here is
> Linux's idea of CPU id, not Xen's.
> 
> In case Xen's idea is needed xen_vcpu_id mapping should be used. But I
> don't think it's the case here.
> 

Correct.


Juergen