Re: [Qemu-devel] [PATCH] xlnx-zynqmp: Don't create rpu-cluster if there are no RPUs

[Philippe Mathieu-Daudé]

On 1/22/19 10:28 AM, Peter Maydell wrote:
> On Mon, 21 Jan 2019 at 20:12, Philippe Mathieu-Daudé <address@hidden> wrote:
>>
>> Hi Peter,
>>
>> On 1/21/19 7:43 PM, Peter Maydell wrote:
>>> If we aren't going to create any RPUs, then don't create the
>>> rpu-cluster unit. This allows us to add an assertion to the
>>> cluster object that it contains at least one CPU, which helps
>>> to avoid bugs in creating clusters and putting CPUs in them.
>>>
>>> Signed-off-by: Peter Maydell <address@hidden>
>>> ---
>>> This is a preparatory patch that is necessary for the series
>>> "[PATCH v3 0/4] tcg: support heterogenous CPU clusters"
>>> (address@hidden)
>>> in order to avoid the xlnx-zcu102 board asserting if started with
>>> fewer than 5 CPUs.
>>>
>>>  hw/arm/xlnx-zynqmp.c | 5 +++++
>>>  1 file changed, 5 insertions(+)
>>>
>>> diff --git a/hw/arm/xlnx-zynqmp.c b/hw/arm/xlnx-zynqmp.c
>>> index 370b0e44a38..16cba433cb7 100644
>>> --- a/hw/arm/xlnx-zynqmp.c
>>> +++ b/hw/arm/xlnx-zynqmp.c
>>> @@ -178,6 +178,11 @@ static void xlnx_zynqmp_create_rpu(XlnxZynqMPState *s, 
>>> const char *boot_cpu,
>>>      int i;
>>>      int num_rpus = MIN(smp_cpus - XLNX_ZYNQMP_NUM_APU_CPUS, 
>>> XLNX_ZYNQMP_NUM_RPU_CPUS);
>>
>> Not related to this patch, but this check seems dangerous, i.e. using
>> "-smp 2" we get num_rpus=-2 which luckyly doesn't enter the for() loop.
>>
>>>
>>> +    if (num_rpus == 0) {
>>
>> With the current codebase, you'd have to check for "num_rpus <= 0", ...
> 
> Oops, nice catch.
> 
>> What about this instead?
>>
>> -- >8 --
>> @@ -451,10 +451,12 @@ static void xlnx_zynqmp_realize(DeviceState *dev,
>> Error **errp)
>>                      "RPUs just use -smp 6.");
>>      }
>>
>> -    xlnx_zynqmp_create_rpu(s, boot_cpu, &err);
>> -    if (err) {
>> -        error_propagate(errp, err);
>> -        return;
>> +    if (smp_cpus > XLNX_ZYNQMP_NUM_APU_CPUS) {
>> +        xlnx_zynqmp_create_rpu(s, boot_cpu, &err);
>> +        if (err) {
>> +            error_propagate(errp, err);
>> +            return;
>> +        }
>>      }
> 
> Yeah, that would work too. I think I would just go for
> using "if (num_rpus <= 0)" in the function, though.

OK, whichever patch you prefer, you can add:
Reviewed-by: Philippe Mathieu-Daudé <address@hidden>
Tested-by: Philippe Mathieu-Daudé <address@hidden>

Regards,

Phil.

> 
> thanks
> -- PMM
>