Re: [PATCH v1] memory: tegra20: Add debug statistics

[Krzysztof Kozlowski]

On 18/03/2021 17:18, Dmitry Osipenko wrote:
> 18.03.2021 18:23, Krzysztof Kozlowski пишет:
> 
>>> +
>>> +	/* collect memory controller utilization percent for each client */
>>> +	for (i = 0; i < mc->soc->num_clients; i += 2) {
>>> +		client0 = &mc->soc->clients[i];
>>> +		client1 = &mc->soc->clients[i + 1];
>>> +
>>> +		if (i + 1 == mc->soc->num_clients)
>>> +			client1 = NULL;
>>> +
>>> +		tegra20_mc_stat_events(mc, client0, client1,
>>> +				       MC_STAT_CONTROL_FILTER_PRI_DISABLE,
>>> +				       MC_STAT_CONTROL_PRI_EVENT_HP,
>>> +				       MC_STAT_CONTROL_EVENT_QUALIFIED,
>>> +				       &stats[i + 0].events,
>>> +				       &stats[i + 1].events);
>>> +	}
>>> +
>>> +	/* collect more info from active clients */
>>> +	for (i = 0; i < mc->soc->num_clients; i++) {
>>> +		clientb = mc->soc->num_clients;
>>> +
>>> +		for (client0 = NULL; i < mc->soc->num_clients; i++) {
>>> +			if (stats[i].events) {
>>> +				client0 = &mc->soc->clients[i];
>>> +				clienta = i++;
>>> +				break;
>>> +			}
>>> +		}
>>
>> Could all clients have 0 events ending with "clienta" being undefined?
>> "client0" would be non-NULL because of loop before.
> 
> As per 6.8.5.3 of C99 standard, the declaration of a for-loop "is
> reached in the order of execution before the first evaluation of the
> controlling expression".
> 
> http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1256.pdf
> 
>>> +
>>> +		for (client1 = NULL; i < mc->soc->num_clients; i++) {
>>> +			if (stats[i].events) {
>>> +				client1 = &mc->soc->clients[i];
>>> +				clientb = i;
>>> +				break;
>>> +			}
>>> +		}
>>> +
>>> +		if (!client0 && !client1)
>>> +			break;
> 
> this means that both client0 and client1 are set t0 NULL here if all
> clients have 0 events.
> 

Yes, you're right. I missed the assignment client0=NULL in the for().

Best regards,
Krzysztof