Re: [PATCH] perf/x86/amd/uncore: fix error codes in amd_uncore_init()

[Ingo Molnar]

* Sandipan Das <sandipan.das@xxxxxxx> wrote:

> On 10/13/2023 12:48 PM, Dan Carpenter wrote:
> > Some of the error paths in this function return don't initialize the
> > error code.  Return -ENODEV.
> > 
> > Fixes: d6389d3ccc13 ("perf/x86/amd/uncore: Refactor uncore management")
> > Signed-off-by: Dan Carpenter <dan.carpenter@xxxxxxxxxx>
> > ---
> >  arch/x86/events/amd/uncore.c | 3 ++-
> >  1 file changed, 2 insertions(+), 1 deletion(-)
> > 
> > diff --git a/arch/x86/events/amd/uncore.c b/arch/x86/events/amd/uncore.c
> > index 9b444ce24108..a389828f378c 100644
> > --- a/arch/x86/events/amd/uncore.c
> > +++ b/arch/x86/events/amd/uncore.c
> > @@ -1009,7 +1009,8 @@ static struct amd_uncore uncores[UNCORE_TYPE_MAX] = {
> >  static int __init amd_uncore_init(void)
> >  {
> >  	struct amd_uncore *uncore;
> > -	int ret, i;
> > +	int ret = -ENODEV;
> > +	int i;
> >  
> >  	if (boot_cpu_data.x86_vendor != X86_VENDOR_AMD &&
> >  	    boot_cpu_data.x86_vendor != X86_VENDOR_HYGON)
> 
> 
> Thanks for catching this. I see that 'ret' remains uninitialized for cases
> where the hotplug callback registration fails and was thinking if the
> following is a better fix for this as the reason might not be ENODEV.

Yeah, passing through the real error codes is usually better.

Here's it's probably a bit academic, as I don't think we are even using the 
init return code in the init sequence iterator, see how the return code by 
do_one_initcall() gets ignored by do_initcall_level() & do_pre_smp_initcalls() ...

Nevertheless, mind submitting this as a separate patch?

Thanks,

	Ingo