Re: [RFC PATCH v3 2/2] soc: renesas: Add L2 cache management for RZ/Five SoC

From: Conor Dooley
Date: Fri Oct 21 2022 - 18:32:14 EST


On Fri, Oct 21, 2022 at 11:05:40PM +0100, Lad, Prabhakar wrote:
> Hi Rob,
>
> Thank you for the review.
>
> On Fri, Oct 21, 2022 at 3:05 AM Rob Herring <robh@xxxxxxxxxx> wrote:
> >
> > On Wed, Oct 19, 2022 at 11:02:42PM +0100, Prabhakar wrote:
> > > From: Lad Prabhakar <prabhakar.mahadev-lad.rj@xxxxxxxxxxxxxx>
> > >
> > > On the AX45MP core, cache coherency is a specification option so it may
> > > not be supported. In this case DMA will fail. As a workaround, firstly we
> > > allocate a global dma coherent pool from which DMA allocations are taken
> > > and marked as non-cacheable + bufferable using the PMA region as specified
> > > in the device tree. Synchronization callbacks are implemented to
> > > synchronize when doing DMA transactions.
> > >
> > > The Andes AX45MP core has a Programmable Physical Memory Attributes (PMA)
> > > block that allows dynamic adjustment of memory attributes in the runtime.
> > > It contains a configurable amount of PMA entries implemented as CSR
> > > registers to control the attributes of memory locations in interest.
> > >
> > > Below are the memory attributes supported:
> > > * Device, Non-bufferable
> > > * Device, bufferable
> > > * Memory, Non-cacheable, Non-bufferable
> > > * Memory, Non-cacheable, Bufferable
> > > * Memory, Write-back, No-allocate
> > > * Memory, Write-back, Read-allocate
> > > * Memory, Write-back, Write-allocate
> > > * Memory, Write-back, Read and Write-allocate
> > >
> > > This patch adds support to configure the memory attributes of the memory
> > > regions as passed from the l2 cache node and exposes the cache management
> > > ops.
> > >
> > > More info about PMA (section 10.3):
> > > http://www.andestech.com/wp-content/uploads/AX45MP-1C-Rev.-5.0.0-Datasheet.pdf
> > >
> > > This feature is based on the work posted [0] by Vincent Chen
> > > <vincentc@xxxxxxxxxxxxx> for the Andes AndeStart RISC-V CPU.
> > >
> > > [0] https://lore.kernel.org/lkml/1540982130-28248-1-git-send-email-vincentc@xxxxxxxxxxxxx/
> > >
> > > Signed-off-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@xxxxxxxxxxxxxx>
> > > ---
> > > arch/riscv/include/asm/cacheflush.h | 8 +
> > > arch/riscv/include/asm/errata_list.h | 2 +
> > > arch/riscv/mm/dma-noncoherent.c | 20 ++
> > > drivers/soc/renesas/Kconfig | 5 +
> > > drivers/soc/renesas/Makefile | 4 +
> > > drivers/soc/renesas/rzf/Kconfig | 6 +
> > > drivers/soc/renesas/rzf/Makefile | 3 +
> > > drivers/soc/renesas/rzf/ax45mp_cache.c | 431 +++++++++++++++++++++++++
> >
> > How many cache drivers do we have around now? I've seen a few bindings
> > go by. I'm guessing it is time to stop putting the drivers in the
> > drivers/soc/ dumping ground.
> >
> The main reason this driver is not in arch/riscv is that it has vendor
> specific extensions. Due to this reason it was agreed during the LPC
> that vendor specific extension should be maintained by SoC vendors and
> was agreed that this can go into drivers/soc/renesas folder instead.

Does not in drivers/soc mean they need to go into arch/riscv?
The outcome of the chat at the LPC BoF was more that the cache drivers
themselves should not be be routed via the arch maintainers, no?

>
> > > drivers/soc/renesas/rzf/ax45mp_sbi.h | 29 ++
> > > 9 files changed, 508 insertions(+)
> > > create mode 100644 drivers/soc/renesas/rzf/Kconfig
> > > create mode 100644 drivers/soc/renesas/rzf/Makefile
> > > create mode 100644 drivers/soc/renesas/rzf/ax45mp_cache.c
> > > create mode 100644 drivers/soc/renesas/rzf/ax45mp_sbi.h
> > >
> > > diff --git a/arch/riscv/include/asm/cacheflush.h b/arch/riscv/include/asm/cacheflush.h
> > > index 8a5c246b0a21..40aa790be9a3 100644
> > > --- a/arch/riscv/include/asm/cacheflush.h
> > > +++ b/arch/riscv/include/asm/cacheflush.h
> > > @@ -65,6 +65,14 @@ static inline void riscv_noncoherent_supported(void) {}
> > > #define SYS_RISCV_FLUSH_ICACHE_LOCAL 1UL
> > > #define SYS_RISCV_FLUSH_ICACHE_ALL (SYS_RISCV_FLUSH_ICACHE_LOCAL)
> > >
> > > +#ifdef CONFIG_AX45MP_L2_CACHE
> > > +void ax45mp_cpu_dma_inval_range(void *vaddr, size_t end);
> > > +void ax45mp_cpu_dma_wb_range(void *vaddr, size_t end);
> > > +
> > > +#define ALT_CMO_OP(_op, _start, _size, _cachesize) \
> > > + _op(_start, _size)
> > > +#endif
> > > +
> > > #include <asm-generic/cacheflush.h>
> > >
> > > #endif /* _ASM_RISCV_CACHEFLUSH_H */
> > > diff --git a/arch/riscv/include/asm/errata_list.h b/arch/riscv/include/asm/errata_list.h
> > > index 19a771085781..d9cbf60c3b65 100644
> > > --- a/arch/riscv/include/asm/errata_list.h
> > > +++ b/arch/riscv/include/asm/errata_list.h
> > > @@ -89,6 +89,7 @@ asm volatile(ALTERNATIVE( \
> > > #define ALT_THEAD_PMA(_val)
> > > #endif
> > >
> > > +#ifdef CONFIG_ERRATA_THEAD_CMO
> > > /*
> > > * dcache.ipa rs1 (invalidate, physical address)
> > > * | 31 - 25 | 24 - 20 | 19 - 15 | 14 - 12 | 11 - 7 | 6 - 0 |
> > > @@ -143,5 +144,6 @@ asm volatile(ALTERNATIVE_2( \
> > > : "a0")
> > >
> > > #endif /* __ASSEMBLY__ */
> > > +#endif
> > >
> > > #endif
> > > diff --git a/arch/riscv/mm/dma-noncoherent.c b/arch/riscv/mm/dma-noncoherent.c
> > > index b0add983530a..5270acca6766 100644
> > > --- a/arch/riscv/mm/dma-noncoherent.c
> > > +++ b/arch/riscv/mm/dma-noncoherent.c
> > > @@ -24,13 +24,25 @@ void arch_sync_dma_for_device(phys_addr_t paddr, size_t size,
> > >
> > > switch (dir) {
> > > case DMA_TO_DEVICE:
> > > +#ifdef CONFIG_ERRATA_THEAD_CMO
> > > ALT_CMO_OP(clean, vaddr, size, riscv_cbom_block_size);
> > > +#elif CONFIG_AX45MP_L2_CACHE
> > > + ALT_CMO_OP(ax45mp_cpu_dma_wb_range, vaddr, size, 0x0);
> > > +#endif
> >
> > How do you support more than one platform in a build?
> >
> Yes, that's one concern which I have mentioned in the cover letter too
> (At that moment it's just a single platform). Suggestions welcome!

I think I said it on one of the earlier version, but it needs to be
implemented w/ runtime patching via alternatives just like the thead
stuff patches in their functions.