[ltt-dev] [BUG] [PATCH] LTTV - interrupt handler time calculation is wrong in case of nested interrupts

Mathieu Desnoyers compudj at krystal.dyndns.org
Wed Oct 15 10:03:00 EDT 2008


Hi Gaurav,

Your fix seems appropriate, I'll merge it.

Thanks !

Mathieu

* Gaurav Singh (gausinghnsit at gmail.com) wrote:
> Hi all,
> 
> I was using the interrupt handler plugin and noticed that sometimes
> the timer interrupt  handler was taking a lot of time. Specifically
> the max time indicated by the plugin was wrong. This was because the
> time calculated by the plugin in case of nested interrupts is wrong.
> 
> Nested interrupt:
> Interrupt 1 (entry)  .... 1
> Interrupt 2(entry).....2
> Interrupt 2 (exit) ..... 3
> Interrupt 3 (exit) ..... 4
> 
> The plugin did the following calculation:
> Time for 1st interrupt = timeat (3) - timeat (1)
> Time for 2nd interrupt = timeat (4) - timeat (2)
> 
> 
> The following patch corrects this anamoly. Basically it takes the up
> the most recent interrupt and pairs the irq_exit with it. Please check
> if this works for you.
> 
> diff -uprN lttv-0.10.0-pre11-10032008/lttv/modules/gui/interrupts/interrupts.c
> ../lttv-0.10.0-pre11-10032008/lttv/modules/gui/interrupts/interrupts.c
> --- lttv-0.10.0-pre11-10032008/lttv/modules/gui/interrupts/interrupts.c
> 2007-10-16 06:44:40.000000000 +0530
> +++ ../lttv-0.10.0-pre11-10032008/lttv/modules/gui/interrupts/interrupts.c
>      2008-09-29 15:13:02.000000000 +0530
> @@ -583,7 +583,7 @@ static void CalculateData(LttTime time_e
>    LttTime duration;
>    GArray *FirstRequestIrqExit = event_data->FirstRequestIrqExit;
>    GArray *FirstRequestIrqEntry = event_data->FirstRequestIrqEntry;
> -  for(i = 0; i < FirstRequestIrqEntry->len; i++)
> +  for(i = FirstRequestIrqEntry->len-1; i >=0; i--)
>    {
>      element = &g_array_index(FirstRequestIrqEntry,irq_entry,i);
>      if(element->cpu_id == cpu_id)
> 
> 
> 
> Regards,
> Gaurav
> 

-- 
Mathieu Desnoyers
OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68




More information about the lttng-dev mailing list