[bug report] perf: Add driver for Arm NI-700 interconnect PMU

[bug report] perf: Add driver for Arm NI-700 interconnect PMU

[Dan Carpenter]

Hello Robin Murphy,

Commit 4d5a7680f2b4 ("perf: Add driver for Arm NI-700 interconnect
PMU") from Sep 4, 2024 (linux-next), leads to the following Smatch
static checker warning:

drivers/perf/arm-ni.c:726 arm_ni_pmu_online_cpu() warn: 'cd' was set to NULL
drivers/perf/arm-ni.c:738 arm_ni_pmu_offline_cpu() warn: 'cd' was set to NULL

drivers/perf/arm-ni.c
    719 static int arm_ni_pmu_online_cpu(unsigned int cpu, struct hlist_node *cpuhp_node)
    720 {
    721         struct arm_ni_cd *cd;
    722         int node;
    723 
    724         cd = hlist_entry_safe(cpuhp_node, struct arm_ni_cd, cpuhp_node);
    725         node = dev_to_node(cd_to_ni(cd)->dev);

There isn't any point in using the _safe() version of hlist_entry() unless we're
going to check for NULL.

--> 726         if (cpu_to_node(cd->cpu) != node && cpu_to_node(cpu) == node)
    727                 arm_ni_pmu_migrate(cd, cpu);
    728         return 0;
    729 }

regards,
dan carpenter

Re: [bug report] perf: Add driver for Arm NI-700 interconnect PMU

[Robin Murphy]

On 2024-09-11 8:35 am, Dan Carpenter wrote:
> Hello Robin Murphy,
> 
> Commit 4d5a7680f2b4 ("perf: Add driver for Arm NI-700 interconnect
> PMU") from Sep 4, 2024 (linux-next), leads to the following Smatch
> static checker warning:
> 
> drivers/perf/arm-ni.c:726 arm_ni_pmu_online_cpu() warn: 'cd' was set to NULL
> drivers/perf/arm-ni.c:738 arm_ni_pmu_offline_cpu() warn: 'cd' was set to NULL
> 
> drivers/perf/arm-ni.c
>      719 static int arm_ni_pmu_online_cpu(unsigned int cpu, struct hlist_node *cpuhp_node)
>      720 {
>      721         struct arm_ni_cd *cd;
>      722         int node;
>      723
>      724         cd = hlist_entry_safe(cpuhp_node, struct arm_ni_cd, cpuhp_node);
>      725         node = dev_to_node(cd_to_ni(cd)->dev);
> 
> There isn't any point in using the _safe() version of hlist_entry() unless we're
> going to check for NULL.

Hmm, this was just following the same pattern as every other PMU driver, 
which dates all the way back to the original multi-instance 
implementation[1]. Seems like one for a tree-wide cleanup if anyone's 
sufficiently bothered.

Thanks,
Robin.

[1] 
https://lore.kernel.org/all/1471024183-12666-1-git-send-email-bigeasy@linutronix.de/

> 
> --> 726         if (cpu_to_node(cd->cpu) != node && cpu_to_node(cpu) == node)
>      727                 arm_ni_pmu_migrate(cd, cpu);
>      728         return 0;
>      729 }
> 
> regards,
> dan carpenter