[PATCH 5/6] dma-mapping: Allow batched DMA sync operations if supported by the arch

[Leon Romanovsky]

On Fri, Dec 26, 2025 at 02:31:42AM +1300, Barry Song wrote:
> On Fri, Dec 26, 2025 at 1:36 AM Leon Romanovsky <leon at kernel.org> wrote:
> >
> > On Thu, Dec 25, 2025 at 06:45:09PM +1300, Barry Song wrote:
> > > > > >
> > > > >
> > > > > OK. Could you take a look at [1] and see if any further
> > > > > improvements are needed before I send v2?
> > > >
> > > > Everything looks ok, except these renames:
> > > > -                       arch_sync_dma_for_cpu(paddr, sg->length, dir);
> > > > +                       arch_sync_dma_for_cpu_batch_add(paddr, sg->length, dir);
> > >
> > > Thanks!
> > > I'm happy to drop the rename as outlined below-feedback welcome :-)
> > >
> > > diff --git a/arch/arm64/include/asm/cache.h b/arch/arm64/include/asm/cache.h
> > > index dd2c8586a725..487fb7c355ed 100644
> > > --- a/arch/arm64/include/asm/cache.h
> > > +++ b/arch/arm64/include/asm/cache.h
> > > @@ -87,6 +87,12 @@ int cache_line_size(void);
> > >
> > >  #define dma_get_cache_alignment      cache_line_size
> > >
> > > +static inline void arch_sync_dma_flush(void)
> > > +{
> > > +     dsb(sy);
> > > +}
> > > +#define arch_sync_dma_flush arch_sync_dma_flush
> > > +
> > >  /* Compress a u64 MPIDR value into 32 bits. */
> > >  static inline u64 arch_compact_of_hwid(u64 id)
> > >  {
> > > diff --git a/arch/arm64/mm/dma-mapping.c b/arch/arm64/mm/dma-mapping.c
> > > index b2b5792b2caa..ae1ae0280eef 100644
> > > --- a/arch/arm64/mm/dma-mapping.c
> > > +++ b/arch/arm64/mm/dma-mapping.c
> > > @@ -17,7 +17,7 @@ void arch_sync_dma_for_device(phys_addr_t paddr, size_t size,
> > >  {
> > >       unsigned long start = (unsigned long)phys_to_virt(paddr);
> > >
> > > -     dcache_clean_poc(start, start + size);
> > > +     dcache_clean_poc_nosync(start, start + size);
> > >  }
> > >
> > >  void arch_sync_dma_for_cpu(phys_addr_t paddr, size_t size,
> > > @@ -28,7 +28,7 @@ void arch_sync_dma_for_cpu(phys_addr_t paddr, size_t size,
> > >       if (dir == DMA_TO_DEVICE)
> > >               return;
> > >
> > > -     dcache_inval_poc(start, start + size);
> > > +     dcache_inval_poc_nosync(start, start + size);
> > >  }
> > >
> > >  void arch_dma_prep_coherent(struct page *page, size_t size)
> > > diff --git a/include/linux/dma-map-ops.h b/include/linux/dma-map-ops.h
> > > index 4809204c674c..e7dd8a63b40e 100644
> > > --- a/include/linux/dma-map-ops.h
> > > +++ b/include/linux/dma-map-ops.h
> > > @@ -361,6 +361,12 @@ static inline void arch_sync_dma_for_cpu(phys_addr_t paddr, size_t size,
> > >  }
> > >  #endif /* ARCH_HAS_SYNC_DMA_FOR_CPU */
> > >
> > > +#ifndef arch_sync_dma_flush
> >
> > You likely need to wrap this in "#ifdef CONFIG_ARCH_HAS_SYNC_DMA_FLUSH"
> > as done in the surrounding code.
> 
> I've dropped the new Kconfig option and now rely on whether
> arch_sync_dma_flush() is provided by the architecture. If an arch
> does not define arch_sync_dma_flush() in its asm/cache.h, a no-op
> implementation is used instead.

I know.

> 
> Do you still prefer keeping a config option to match the surrounding
> code style?

I don't have a strong preference here. Go ahead and try your current
version and see how people respond.

> Note that on arm64, arch_sync_dma_flush() is already a
> static inline rather than an extern, so it is not strictly aligned
> with the others.
> Having both CONFIG_ARCH_HAS_SYNC_DMA_FLUSH and
> "#ifndef arch_sync_dma_flush" seems duplicated.
> 
> Another potential optimization would be to drop these options
> entirely and handle this via ifndefs, letting each architecture
> define the macros in asm/cache.h instead.
> 
> Whether arch implements arch_sync_dma_for_xx() as static inline or
> as external functions makes no difference.
> 
> - #ifdef CONFIG_ARCH_HAS_SYNC_DMA_FOR_CPU
> - void arch_sync_dma_for_cpu(phys_addr_t paddr, size_t size,-
>                 enum dma_data_direction dir);
> - #else
> + #ifndef arch_sync_dma_for_cpu
> static inline void arch_sync_dma_for_cpu(phys_addr_t paddr, size_t size,
>                 enum dma_data_direction dir)
> {
> }
> #endif /* ARCH_HAS_SYNC_DMA_FOR_CPU */
> 
> >
> > Thanks
> >
> > > +static inline void arch_sync_dma_flush(void)
> > > +{
> > > +}
> > > +#endif
> > > +
> > >  #ifdef CONFIG_ARCH_HAS_SYNC_DMA_FOR_CPU_ALL
> > >  void arch_sync_dma_for_cpu_all(void);
> > >  #else
> > >
> 
> Thanks
> Barry
>