Re: [PATCH 1/1] x86/platform/x86: Fix count of CHas on multi-pci-segment arches

[Liang, Kan]

On 3/13/2018 1:16 PM, Liang, Kan wrote:
> On 3/13/2018 11:58 AM, Andy Shevchenko wrote:
> > On Tue, Mar 13, 2018 at 3:42 AM, Liang, Kan 
> > <kan.liang@xxxxxxxxxxxxxxx> wrote:
> >
> > > +#define SKX_CAPID6ÂÂÂÂÂÂÂÂÂÂÂÂ 0x9c
> > > +#define SKX_CHA_BIT_WIDTHÂÂÂÂÂ 28
> > > +
> > > Â static int skx_count_chabox(void)
> > > Â {
> > > +ÂÂÂÂÂÂ struct pci_dev *dev = NULL;
> > > +ÂÂÂÂÂÂ u32 val = 0;
> >
> > > +ÂÂÂÂÂÂ dev = pci_get_device(PCI_VENDOR_ID_INTEL, 0x2083, dev);
> > > +ÂÂÂÂÂÂ if (!dev)
> > > +ÂÂÂÂÂÂÂÂÂÂÂÂÂÂ return 0;
> >
> > Where is pci_dev_put()?
> >
> > > +ÂÂÂÂÂÂ pci_read_config_dword(dev, SKX_CAPID6, &val);
> > > +ÂÂÂÂÂÂ return bitmap_weight((unsigned long *)&val, SKX_CHA_BIT_WIDTH);
> >
> > UB is here.
> > Fix is simple, use unsigned long and drop this ugly casting.

Just noticed that we have to do casting anyway.
pci_read_config_dword uses u32.
bitmap_weight uses unsigned long.

Thanks,
Kan