RE: [PATCH 4/6] arm64/io: Provide a WC friendly __iowriteXX_copy()

[David Laight <David.Laight@ACULAB.COM>]

From: Jason Gunthorpe
> Sent: 22 February 2024 22:36
> To: David Laight <David.Laight@ACULAB.COM>
> 
> On Thu, Feb 22, 2024 at 10:05:04PM +0000, David Laight wrote:
> > From: Jason Gunthorpe
> > > Sent: 21 February 2024 01:17
> > >
> > > The kernel provides driver support for using write combining IO memory
> > > through the __iowriteXX_copy() API which is commonly used as an optional
> > > optimization to generate 16/32/64 byte MemWr TLPs in a PCIe environment.
> > >
> > ...
> > > Implement __iowrite32/64_copy() specifically for ARM64 and use inline
> > > assembly to build consecutive blocks of STR instructions. Provide direct
> > > support for 64/32/16 large TLP generation in this manner. Optimize for
> > > common constant lengths so that the compiler can directly inline the store
> > > blocks.
> > ...
> > > +/*
> > > + * This generates a memcpy that works on a from/to address which is aligned to
> > > + * bits. Count is in terms of the number of bits sized quantities to copy. It
> > > + * optimizes to use the STR groupings when possible so that it is WC friendly.
> > > + */
> > > +#define memcpy_toio_aligned(to, from, count, bits)                        \
> > > +	({                                                                \
> > > +		volatile u##bits __iomem *_to = to;                       \
> > > +		const u##bits *_from = from;                              \
> > > +		size_t _count = count;                                    \
> > > +		const u##bits *_end_from = _from + ALIGN_DOWN(_count, 8); \
> > > +                                                                          \
> > > +		for (; _from < _end_from; _from += 8, _to += 8)           \
> > > +			__const_memcpy_toio_aligned##bits(_to, _from, 8); \
> > > +		if ((_count % 8) >= 4) {
> >
> > If (_count & 4) {
> 
> That would be obfuscating, IMHO. The compiler doesn't need such things
> to generate optimal code.

Try it: https://godbolt.org/z/EvvGrTxv3 
And it isn't that obfuscated - no more so than your version.

> > > +			__const_memcpy_toio_aligned##bits(_to, _from, 1); \
> > > +	})
> >
> > But that looks bit a bit large to be inlined.
> 
> You trimmed alot, this #define is in a C file and it is a template to
> generate the 32 and 64 bit out of line functions. Things are done like
> this because the 32/64 version are exactly the same logic except just
> with different types and sizes.

I missed that in a quick read at 11pm :-(

Although I doubt that generating long TLP from byte writes is
really necessary.
IIRC you were merging at most 4 writes.
So better to do a single 32bit write instead.
(Unless you have misaligned source data - unlikely.)

While write-combining to generate long TLP is probably mostly
safe for PCIe targets, there are some that will only handle
TLP for single 32bit data items.
Which might be why the code is explicitly requesting 4 byte copies.
So it may be entirely wrong to write-combine anything except
the generic memcpy_toio().

	David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)