On Thu, Nov 24, 2016 at 09:47:01AM +0100, Andrew Jones wrote:
[...]
> > +static inline uint64_t edu_reg_readq(struct pci_edu_dev *dev, int reg)
> > +{
> > + return __raw_readq(edu_reg(dev, reg));
> > +}
> > +
> > +static inline uint32_t edu_reg_readl(struct pci_edu_dev *dev, int reg)
> > +{
> > + return __raw_readl(edu_reg(dev, reg));
> > +}
> > +
> > +static inline void edu_reg_writeq(struct pci_edu_dev *dev, int reg,
> > + uint64_t val)
> > +{
> > + __raw_writeq(val, edu_reg(dev, reg));
> > +}
> > +
> > +static inline void edu_reg_writel(struct pci_edu_dev *dev, int reg,
> > + uint32_t val)
> > +{
> > + __raw_writel(val, edu_reg(dev, reg));
> > +}
>
> Can you refresh my memory on why we're using __raw accessors here instead
> of the __raw wrappers that include barriers and endian conversion? Hmm,
> actually it doesn't look like the EDU device specifies an endianness...
I see that __raw wrappers (read[bwlq]) are assuming the data is little
endian, while edu device is configured as native endianess
(DEVICE_NATIVE_ENDIAN). So looks like using __raw accessors should be
the correct way?
But, yes here I missed memory barriers. I have only volatile
protections (which __raw_*() provides).
Thanks,
-- peterx
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
