Thanks, Avadh. I don't believe the process ever wakes up again. I left it
running overnight and the 'time' field didn't increase by even a fraction
of a second, so I think it never recovers.

I was trying to wait for some simulations to finish up, but it looks like
they're taking way longer than is useful. I think I'll kill them and try
out the patch and let you know what happens.

On Wed, Aug 15, 2012 at 1:21 PM, avadh patel <[email protected]> wrote:

>
> On Wed, Aug 15, 2012 at 8:16 AM, Paul Rosenfeld <[email protected]>wrote:
>
>> Hello all,
>>
>> I'm getting some new and strange behavior and I was wondering if anyone
>> else has experienced this. Typically, I run a bunch of simulations in
>> parallel using run_bench.py and running from a checkpoint with the
>> -snapshot option (i.e., no writebacks to the disk image).
>>
>> Up until recently, I haven't had any problems. But now that I'm trying to
>> run a bunch of simulations, I'm getting into a state where some of the qemu
>> processes will go into sleep mode (i.e., they show up as status 'S' in
>> htop) and they never get back into the running state. They're using no CPU
>> time and just kind of sitting around.
>>
>> Do they ever wake up even for fraction of a second?  If they do then it
> might have issue related to all CPUs in 'halted' mode. Try following patch
> to see if this happens or not.
>
> - Avadh
>
> diff --git a/qemu/cpu-exec.c b/qemu/cpu-exec.c
> index f10b207..70d4ec3 100644
> --- a/qemu/cpu-exec.c
> +++ b/qemu/cpu-exec.c
> @@ -238,16 +238,6 @@ int sim_cpu_exec(void)
>  /* #define DECLARE_HOST_REGS 1 */
>  /* #include "hostregs_helper.h" */
>      int ret, interrupt_request;
> -       CPUState *env1;
> -
> -    uint8_t all_halted = 1;
> -    for(env1 = first_cpu; env1 != NULL; env1 = env1->next_cpu)
> -        if (cpu_halted(env1) != EXCP_HALTED) {
> -            all_halted = 0;
> -            break;
> -        }
> -    if(all_halted)
> -        return EXCP_HALTED;
>
>
>      /* first we save global registers */
>
> Has anyone else had something like this happen? Is there a way to even
>> debug such a problem? Is there a chance this behavior could somehow be
>> caused by the -snapshot option?
>>
>> Thanks,
>>  Paul
>>
>> _______________________________________________
>> http://www.marss86.org
>> Marss86-Devel mailing list
>> [email protected]
>> https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel
>>
>>
>
_______________________________________________
http://www.marss86.org
Marss86-Devel mailing list
[email protected]
https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel

Reply via email to