Re: [PATCH -v5 09/11] tracing: add IRQENTRY_EXIT for MIPS

[Wu Zhangjin <wuzhangjin@gmail.com>]

Hi,

On Mon, 2009-10-26 at 01:36 +0100, Frederic Weisbecker wrote:
> 2009/10/25 Wu Zhangjin <wuzhangjin@gmail.com>:
> > This patch fix the following error with FUNCTION_GRAPH_TRACER=y:
> >
> > kernel/built-in.o: In function `print_graph_irq':
> > trace_functions_graph.c:(.text+0x6dba0): undefined reference to 
> > `__irqentry_text_start'
> > trace_functions_graph.c:(.text+0x6dba8): undefined reference to 
> > `__irqentry_text_start'
> > trace_functions_graph.c:(.text+0x6dbd0): undefined reference to 
> > `__irqentry_text_end'
> > trace_functions_graph.c:(.text+0x6dbd4): undefined reference to 
> > `__irqentry_text_end'
> >
> > (This patch is need to support function graph tracer of MIPS)
> 
> 
> If you want to enjoy this section, you'd need to tag the
> mips irq entry functions with "__irq_entry"  :)
> 
> I guess there is a do_IRQ() in mips that is waiting for that (and
> probably some others).
> The effect is that interrupt areas are cut with a pair of arrows
> in the trace, so that you more easily spot interrupts in the traces
> 
> May be I missed this part in another patch in this series though...

ooh, Sorry, only this patch added(I stopped after fixing the compiling
errors, no more check! so lazy a guy!).

Just checked the source code of MIPS, the do_IRQ() is defined as a
macro, so, I must move the macro to a C file, and also, there is a
irq_enter...irq_exit block in a "big" function, I need to split it out.

[...]
/*
 * do_IRQ handles all normal device IRQ's (the special
 * SMP cross-CPU interrupts have their own specific
 * handlers).
 *
 * Ideally there should be away to get this into kernel/irq/handle.c to
 * avoid the overhead of a call for just a tiny function ...
 */
#define do_IRQ(irq)
\
do {
\
        irq_enter();
\
        __DO_IRQ_SMTC_HOOK(irq);
\
        generic_handle_irq(irq);
\
        irq_exit();
\
} while (0)
[...]

But The comment told us: do not make this tiny function be a standalone
function, so???

the same to do_IRQ_no_affinity.

and, about the following function, I need to split the
irq_enter()...irq_exit() block out.

void ipi_decode(struct smtc_ipi *pipi)
{
        [...]
        switch (type_copy) {
        case SMTC_CLOCK_TICK:
                irq_enter();
                kstat_incr_irqs_this_cpu(irq, irq_to_desc(irq));
                cd = &per_cpu(mips_clockevent_device, cpu);
                cd->event_handler(cd);
                irq_exit();
                break;

        case LINUX_SMP_IPI:
                switch ((int)arg_copy) {
                case SMP_RESCHEDULE_YOURSELF:
                        ipi_resched_interrupt();
                        break;
                case SMP_CALL_FUNCTION:
                        ipi_call_interrupt();
                        break;
        [...]

Regards,
        Wu Zhangjin