On Mon, Jun 06, 2022 at 12:16:22PM +0100, Marc Zyngier wrote:
> > dc97fd6fec0099 Hector Martin 2022-03-10   926  	switch (irqc->info.version) {
> > dc97fd6fec0099 Hector Martin 2022-03-10   927  	case 1: {
> > dc97fd6fec0099 Hector Martin 2022-03-10   928  		u32 info;
> > dc97fd6fec0099 Hector Martin 2022-03-10   929  
> > 76cde26394114f Hector Martin 2021-01-21   930  		info = aic_ic_read(irqc, AIC_INFO);
> > 7c841f5f6fa3f9 Hector Martin 2022-03-10   931  		irqc->nr_irq = FIELD_GET(AIC_INFO_NR_IRQ, info);
> > dc97fd6fec0099 Hector Martin 2022-03-10   932  		irqc->max_irq = AIC_MAX_IRQ;
> > dc97fd6fec0099 Hector Martin 2022-03-10   933  
> > dc97fd6fec0099 Hector Martin 2022-03-10   934  		off = irqc->info.target_cpu;
> > dc97fd6fec0099 Hector Martin 2022-03-10   935  		off += sizeof(u32) * irqc->max_irq; /* TARGET_CPU */
> > dc97fd6fec0099 Hector Martin 2022-03-10   936  
> > dc97fd6fec0099 Hector Martin 2022-03-10   937  		break;
> > 
> > "off" uninitialized if irqc->info.version != 1.
> 
> Which is pretty much impossible by construction, as as of this
> particular commit, all the possible implementations have version == 1.
> 
> Subsequent patches do add cases where version == 2, and that gets
> covered too.
> 
> So unless I'm missing something obvious, there is no bug here.

I assumed it was something like that.  The kbuild bot doesn't do cross
function analysis.  Neither does Clang really.

regards,
dan carpenter