Re: Interrupt issue on m68k platform and some fix

[Lanttor <lanttor.guo@freescale.com>]

>
> No, it isn't.  It is equivalent to
> *(long *)((char *)&current_thread_info()->preempt_count + 1) += 1
> which is of course completely bogus.
Sorry, I think I paste the redundant character for the codes (no '*'
char at front or end of addql and subql instruction)
I paste it again:

ENTRY(inthandler)
SAVE_ALL_INT
GET_CURRENT(%d0)
addql #1,%curptr@(TASK_INFO+TINFO_PREEMPT+1)
/* put exception # in d0 */
movel %sp@(PT_VECTOR),%d0
swap %d0 /* extract bits 25:18 */
lsrl #2,%d0
andl #0x0ff,%d0

movel %sp,%sp@-
movel %d0,%sp@- /* put vector # on stack */
auto_irqhandler_fixup = . + 2
jbsr process_int /* process the IRQ */
addql #8,%sp /* pop parameters off stack */
ret_from_interrupt:

subql #1,%curptr@(TASK_INFO+TINFO_PREEMPT+1)
jeq ret_from_last_interrupt
2: RESTORE_ALL

ALIGN
ret_from_last_interrupt:
moveb %sp@(PT_SR),%d0
andl #(~ALLOWINT>>8)&0xff,%d0
jne 2b

/* check if we need to do software interrupts */
tstl irq_stat+CPUSTAT_SOFTIRQ_PENDING
jeq .Lret_from_exception
pea ret_from_exception
jra do_softirq

For code "addql #1,%curptr@(TASK_INFO+TINFO_PREEMPT+1)",
My understanding is that
%curptr@(TASK_INFO+TINFO_PREEMPT) is the address of
current_thread_info()->preempt_count, and
%curptr@(TASK_INFO+TINFO_PREEMPT+1) point to the second byte of
preempt_count field (preempt_count field is 4 bytes).

So this code sets value 1 to the second byte of preempt_count field.

HARDIRQ_OFFSET equals (1 << HARDIRQ_SHIT) , and HARDIRQ_SHIFT value is
16, so I think this code equals
current_thread_info()->preempt_count += HARDIRQ_OFFSET

Am I wrong?

Regards,
Lanttor




------------------------------------------------------------------------
*From:* Andreas Schwab <schwab@linux-m68k.org>
*Sent:* 07/02/2009 4:24:35 PM +0800
*To:* Lanttor <lanttor.guo@freescale.com>
*CC:* linux-m68k@vger.kernel.org, uClinux development list
<uclinux-dev@uclinux.org>
*Subject:* Interrupt issue on m68k platform and some fix

> Lanttor <lanttor.guo@freescale.com> writes:
>
>   
>> My understanding of "*addql   #1,%curptr@(TASK_INFO+TINFO_PREEMPT+1)*"
>> is that its meaning is equal to
>> "current_thread_info()->preempt_count += HARDIRQ_OFFSET",
>>     
>
> No, it isn't.  It is equivalent to
> *(long *)((char *)&current_thread_info()->preempt_count + 1) += 1
> which is of course completely bogus.
>
> Andreas.
>
>