[Crash-utility] non CONFIG_SMP support was: Re[2]: ARM supportfor CONFIG_

[takuo.koguchi.sw at hitachi.com]

Dave,

>With respect to the "runq" command, I tried this patch on my only 
>UP ARM sample dumpfile:
I tried a similar modification and got a similar result.

It read
 [no tasks queued]
for both RT and CFS
but there are some "RU" tasks shown by ps command, and there are some tasks with "on_rq = 1".
I have not figured out the reason.

Best Regards,

Takuo

>
>
>----- Original Message -----
>> Dave,
>>
>> Toshi's patch works fine for me.
>>
>> Apart from this, runq command does not work against my vmcore(non-SMP).
>> 
>> runq fails with the following message. 
>> I am afraid it might be another CONFIG_SMP related issue.
>> 
>> crash> set debug 10
>> debug: 10
>> crash> runq
>> runq: per_cpu_symbol_search(per_cpu__runqueues): NULL
>> runq: per-cpu runqueues does not exist
>> crash> 
>>
>> Takuo
>
>Hi Takuo,
>
>With respect to the "runq" command, I tried this patch on my only 
>UP ARM sample dumpfile:
>
>--- task.c.orig 2011-06-27 10:33:51.000000000 -0400
>+++ task.c      2011-06-27 10:47:13.000000000 -0400
>@@ -6970,8 +6970,13 @@
>                         "prio");
>        }
> 
>-       if (!(rq_sp = per_cpu_symbol_search("per_cpu__runqueues")))
>-               error(FATAL, "per-cpu runqueues does not exist\n");
>+       if (!(rq_sp = per_cpu_symbol_search("per_cpu__runqueues"))) {
>+               if (!(kt->flags & SMP) && !(kt->flags & PER_CPU_OFF)) {
>+                       if (!(rq_sp = symbol_search("runqueues")))
>+                               error(FATAL, "runqueues does not exist\n");
>+               } else
>+                       error(FATAL, "per-cpu runqueues does not exist\n");
>+       }
> 
>         runqbuf = GETBUF(SIZE(runqueue));
>        if ((init_sp = per_cpu_symbol_search("per_cpu__init_cfs_rq")))
>
>and it shows:
>
>  crash> runq
>  CPU 0 RUNQUEUE: c0396c78
>    CURRENT: PID: 0      TASK: c0388ee0  COMMAND: "swapper"
>    RT PRIO_ARRAY: c0396d28
>       [no tasks queued]
>    CFS RB_ROOT: c0396d00
>       [no tasks queued]
>  crash> 
>
>but a true test would have something other than an idle state with no
>other tasks queued.
>
>And the "timer" command will definitely require a bit more work.
>
>Dave
>
>--
>Crash-utility mailing list
>Crash-utility at redhat.com
>https://www.redhat.com/mailman/listinfo/crash-utility
>