percpu_irq

[prasadae31]

Hi,

This is my first post here and hence I am not sure if this is the right place. So if I am totally off coast, please show me the right way!

Ok, here is the deal,
I am writing a kernel module read the hardware performance counters on the core 2 duo. And I want to enable the register overflow interrupt.
Got all the hardware stuff done, like setting the interrupt enable bit and things like that, but now I do not know how to associate a handler to it. I googled and found out that the vector is 0xee.

And from another place that I can use register_percpu_irq.
But unfortunately later realized that is only for ia64 (As far as I know this is the itanium architecture)

If someone can help me on this one, I can get out of this little hole and continue.

Note: I am a newbie with kernel programming, and never dealt with interrupts.... Till NOW.

[cbogdan]

Hi,

First of all, interrupt control methods :

local_irq_disable() - disable local interrupt delivery

local_irq_enable() - enable local interrupt delivery

local_irq_save() - save the current state of local interrupt delivery and then disable it

local_irq_restore() - restore local interrupt delivery to the given state

disable_irq() - disable the given interrupt line and ensure no handler on the line is executing before returning

disable_irq_nosync() - disable the given interrupt line

enable_irq() - enable the given interrupt line

irqs_disabled() - returns nonzero if local interrupt delivery is disabled; otherwise returns zero

in_interrupt() - returns nonzero if in interrupt context and zero if in process context

in_irq() - returns nonzero if currently executing an interrupt handler and zero otherwise

So, declaration of an interrupt handler:

static irqreturn_t intr_handler(int irq, void *dev_id, struct pt_regs *regs)

The reason the interrupt handler is static is that it's never called directly from another file.

Drivers can register an interrupt handler and enable a given interrupt line for handling via the function:

/* request_irq: allocate a given interrupt line */
int request_irq(unsigned int irq,
irqreturn_t (*handler)(int, void *, struct pt_regs *),
unsigned long irqflags,
const char *devname,
void *dev_id)

Unregister a given interrupt handler - if no handlers now exist on the line, the given interrupt line is disabled:

void free_irq(unsigned int irq, void *dev_id)

I hope this will help you.

[prasadae31]

Thanks a lot. I will look into this.

The problem I was facing was that I do not know the irq no for the interrupt. Only that the local APIC ctrl for the particular interrupt is APIC_LVTPC -> APIC_BASE + 0x340

I just realized that I can set any vector (not only 0xee) as long as it does not interfere with others and I register it with the local APIC.

I was looking at patches provided by popular performance counter monitoring software liker perfctr and perfmon and they do everything from within the kernel. (But not by a pluggable module.) using apic_write(APIC_LVTPC, LOCAL_PERF*_VECTOR)

where LOCAL_PERF*_VECTOR is LOCAL_PERFMON_VECTOR or LOCAL_PERFCTR_VECTOR and both are defined to 0xee and the patch also changes the FIRST_SYSTEM_VECTOR from 0xef to 0xee.
And then they use BUILD_INTERRUPT(LOCAL_PERF*_INTERRUPT, apic_perf_interrupt) -> in include/asm-x86/mach-default/arch_entry.h
And then they define a fucntion:
fastcall smp_apic_perf_interrupt(<forgot> *regs){
IRQ_Ack();
IRQ_enter();
/* Do whatever */
IRQ_exit()
}

Now from your post and what I have observed from these patches, I am extremely confused. I am now aware of the complexity and power of the APIC architecture on x86 and the beauty of the linux kernel in handling these things!

I have read the interrupts section from the following books:
Linux device drivers
Understanding the linux kernel
Linux kernel Development
Linux kernel primer

But I am missing 1 basic understanding. What is the IRQ parameter? Is it related to 1 line to the IO APIC or the Local APIC? And the NR_IRQS is arount 238 (I do not remember the exact number from a test module which just printed that value from the init function) and not 256. Why is this?

Please do not judge me based on my naivete!

[prasadae31]

Hi, Thanks, I figured it out. What I had done, worked. I realized that when I put in printk's Thanks a lot, I will definitely return the favor by helping those I can help. The true GNU way!