Re: [PATCH 0/8] tip related: radix tree for spareseirq and logical flat clean up

[ebiederm@xmission.com (Eric W. Biederman)]

Yinghai Lu <yinghai@kernel.org> writes:

> On 02/13/2010 04:04 AM, Eric W. Biederman wrote:
>>>  
>>>  #ifdef CONFIG_X86_LOCAL_APIC
>>> Index: linux-2.6/arch/x86/lguest/boot.c
>>> ===================================================================
>>> --- linux-2.6.orig/arch/x86/lguest/boot.c
>>> +++ linux-2.6/arch/x86/lguest/boot.c
>>> @@ -819,7 +819,7 @@ static void __init lguest_init_IRQ(void)
>>>  
>>>  	for (i = FIRST_EXTERNAL_VECTOR; i < NR_VECTORS; i++) {
>>>  		/* Some systems map "vectors" to interrupts weirdly.  Not us! */
>>> -		__get_cpu_var(vector_irq)[i] = i - FIRST_EXTERNAL_VECTOR;
>>> +		__get_cpu_var(vector_desc)[i] = irq_to_desc(i - FIRST_EXTERNAL_VECTOR);
>>>  		if (i != SYSCALL_VECTOR)
>>>  			set_intr_gate(i, interrupt[i - FIRST_EXTERNAL_VECTOR]);
>>>  	}
>> 
>> YH It appears that irq_to_desc_alloc_node has not been called yet
>> so the setting of vector_desc needs to move to lguest_setup_irq.
>
> lguest is using sparseirq?

Apparently.  The last commit in that region of code moved some of the
work (irq_to_desc_alloc_node) into lguest_setup_irq because kmalloc
would not work that early.

Certainly the long term direction is to figure out how to enable
sparseirq unconditionally on x86.

Eric