[dm-devel] [patch 4/4] dm-writecache: use new API for flushing

[Mikulas Patocka]

On Tue, 22 May 2018, Jeff Moyer wrote:

> Hi, Mike,
> 
> Mike Snitzer <snitzer at redhat.com> writes:
> 
> > Looking at Mikulas' wrapper API that you and hch are calling into
> > question:
> >
> > For ARM it is using arch/arm64/mm/flush.c:arch_wb_cache_pmem().
> > (And ARM does seem to be providing CONFIG_ARCH_HAS_PMEM_API.)
> >
> > Whereas x86_64 is using memcpy_flushcache() as provided by
> > CONFIG_ARCH_HAS_UACCESS_FLUSHCACHE.
> > (Yet ARM does provide arch/arm64/lib/uaccess_flushcache.c:memcpy_flushcache)
> >
> > Just seems this isn't purely about ARM lacking on an API level (given on
> > x86_64 Mikulas isn't only using CONFIG_ARCH_HAS_PMEM_API).
> >
> > Seems this is more to do with x86_64 having efficient Non-temporal
> > stores?
> 
> Yeah, I think you've got that all right.
> 
> > Anyway, I'm still trying to appreciate the details here before I can
> > make any forward progress.
> 
> Making data persistent on x64 requires 3 steps:
> 1) copy the data into pmem   (store instructions)
> 2) flush the cache lines associated with the data (clflush, clflush_opt, clwb)
> 3) wait on the flush to complete (sfence)

In theory it works this way. In practice, this sequence is useless because 
the cache flusing instructions are horribly slow.

So, the dm-writecache driver uses non-temporal stores instead of cache 
flushing.

Now, the problem with arm64 is that it doesn't have non-temporal stores. 
So, memcpy_flushcache on arm64 does cached stores and flushes the cache 
afterwards. And this eager flushing is slower than late flushing. On arm4, 
you want to do cached stores, then do something else, and flush the cache 
as late as possible.

> I'm not sure if other architectures require step 3.  Mikulas'
> implementation seems to imply that arm64 doesn't require the fence.

I suppose that arch_wb_cache_pmem() does whatever it needs to do to flush 
the cache. If not, add something like arch_wb_cache_pmem_commit().

> The current pmem api provides:
> 
> memcpy*           -- step 1
> memcpy_flushcache -- this combines steps 1 and 2
> dax_flush         -- step 2
> wmb*              -- step 3
> 
> * not strictly part of the pmem api
> 
> So, if you didn't care about performance, you could write generic code
> that only used memcpy, dax_flush, and wmb (assuming other arches
> actually need the wmb).  What Mikulas did was to abstract out an API
> that could be called by generic code that would work optimally on all
> architectures.
> 
> This looks like a worth-while addition to the PMEM API, to me.  Mikulas,
> what do you think about refactoring the code as Christoph suggested?

I sent this patch 
https://www.redhat.com/archives/dm-devel/2018-May/msg00054.html so that 
you can take the functions pmem_memcpy, pmem_assign, pmem_flush and 
pmem_commit and move them to the generic linux headers. If you want to do 
it, do it.

> Cheers,
> Jeff

Mikulas