Re: gemini timer change

[Chris Gilbert <chris%dokein.co.uk@localhost>]

While I've not looked at bsd code for some time and just lurked on the list, generally the save = disable_interrupts(I32_bit) is paired with restore_interrupts(save) otherwise the f32_bit won't be restored (not that I think netbsd ever used FIQs)

IE enable_interrupts() tended to not be the right thing to use other than in the depths of interrupt handling.

Regards,

Chris

On 9 March 2014 10:35, Martin Husemann <martin%duskware.de@localhost> wrote:

I just commited the patch below to fix the build.

However, I am not sure about the last part (changing disable_interrupts to
enable_interrupts) - it looked like a copy & paste bug to me, but there
may be a deeper reasoning.

Please feel free to fix differently.

Martin


Index: gemini_timer.c
===================================================================
RCS file: /cvsroot/src/sys/arch/arm/gemini/gemini_timer.c,v
retrieving revision 1.5
diff -u -p -r1.5 gemini_timer.c
--- gemini_timer.c      1 Jul 2011 19:32:28 -0000       1.5
+++ gemini_timer.c      9 Mar 2014 10:31:13 -0000
@@ -257,9 +257,7 @@ statintr(void *frame)
 static void
 timer_init(geminitmr_softc_t *sc, int schz, boolean_t autoload, boolean_t intr)
 {
-       int psw;
-
-       psw = disable_interrupts(I32_bit);
+       disable_interrupts(I32_bit);
        timer_factors(sc, schz, autoload);
        _timer_stop(sc);
        _timer_intr_dis(sc);
@@ -267,7 +265,7 @@ timer_init(geminitmr_softc_t *sc, int sc
        if (intr)
                _timer_intr_enb(sc);
        _timer_start(sc);
-       psw = disable_interrupts(I32_bit);
+       enable_interrupts(I32_bit);
 }

 void