Re: [WARNING: ATTACHMENT UNSCANNED]Re: [linux-next:master 13230/13643] arch/riscv/errata/andes/errata.c:29:23: error: storage size of 'ret' isn't known

["Lad, Prabhakar" <prabhakar.csengg@gmail.com>]

Hi Conor,

On Thu, Aug 31, 2023 at 5:36 PM Conor Dooley <conor@kernel.org> wrote:
>
> On Thu, Aug 31, 2023 at 03:59:37PM +0100, Lad, Prabhakar wrote:
>
> > > > Currently ARCH_R9A07G043 selects DMA_GLOBAL_POOL and we have
> > > > RISCV_DMA_NONCOHERENT selecting DMA_DIRECT_REMAP which causes below
> > > > build issue when MMU is disabled:
> > > >
> > > > kernel/dma/pool.c: In function 'atomic_pool_expand':
> > > > kernel/dma/pool.c:105:44: error: implicit declaration of function
> > > > 'pgprot_dmacoherent' [-Werror=implicit-function-declaration]
> > > >   105 |
> > > > pgprot_dmacoherent(PAGE_KERNEL),
> > > >       |                                            ^~~~~~~~~~~~~~~~~~
> > > > kernel/dma/pool.c:105:44: error: incompatible type for argument 3 of
> > > > 'dma_common_contiguous_remap'
> > > >   105 |
> > > > pgprot_dmacoherent(PAGE_KERNEL),
> > > >       |
> > > > ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > > >       |                                            |
> > > >       |                                            int
> > > > In file included from kernel/dma/pool.c:8:
> > >
> > > Hmm, kernel/dma/mapping.c has its use of pgprot_dmacoherent() inside
> > > an #ifdef CONFIG_MMU block.
> > > kernel/dma/pool.c has it inside an #ifdef CONFIG_DMA_DIRECT_REMAP block.
> > >
> > > I guess that select should get a dependency on MMU:
> > >
> > >     config RISCV_DMA_NONCOHERENT
> > >             select DMA_DIRECT_REMAP if MMU
> > >
> > > For comparison, m68k does take that into account:
> > >
> > >     select DMA_DIRECT_REMAP if HAS_DMA && MMU && !COLDFIRE
> > >
> > Thanks for the pointer. With this changes we dont get an unmet
> > dependencies warning.
> >
> > arch/riscv/errata/andes/errata.c: In function 'ax45mp_iocp_sw_workaround':
> > arch/riscv/errata/andes/errata.c:29:23: error: storage size of 'ret' isn't known
> >    29 |         struct sbiret ret;
> >       |                       ^~~
> > arch/riscv/errata/andes/errata.c:35:15: error: implicit declaration of
> > function 'sbi_ecall' [-Werror=implicit-function-declaration]
> >    35 |         ret = sbi_ecall(ANDESTECH_SBI_EXT_ANDES,
> > ANDES_SBI_EXT_IOCP_SW_WORKAROUND,
> >       |               ^~~~~~~~~
> > arch/riscv/errata/andes/errata.c:29:23: warning: unused variable 'ret'
> > [-Wunused-variable]
> >    29 |         struct sbiret ret;
> >       |                       ^~~
> > arch/riscv/errata/andes/errata.c:39:1: error: control reaches end of
> > non-void function [-Werror=return-type]
> >    39 | }
> >       | ^
> > cc1: some warnings being treated as errors
> >
> > To fix the above build issue reported by the bot, adding a dependency
> > for RISCV_SBI in ERRATA_ANDES won't help as ARCH_R9A07G043 is
> > explicitly selecting it. To avoid imply should explicitly select based
> > on dependency something like below?
> >
> > diff --git a/arch/riscv/Kconfig.errata b/arch/riscv/Kconfig.errata
> > index 92c779764b27..566bcefeab50 100644
> > --- a/arch/riscv/Kconfig.errata
> > +++ b/arch/riscv/Kconfig.errata
> > @@ -2,7 +2,7 @@ menu "CPU errata selection"
> >
> >  config ERRATA_ANDES
> >      bool "Andes AX45MP errata"
> > -    depends on RISCV_ALTERNATIVE
> > +    depends on RISCV_ALTERNATIVE && RISCV_SBI
> >      help
> >        All Andes errata Kconfig depend on this Kconfig. Disabling
> >        this Kconfig will disable all Andes errata. Please say "Y"
> > @@ -12,7 +12,7 @@ config ERRATA_ANDES
> >
> >  config ERRATA_ANDES_CMO
> >      bool "Apply Andes cache management errata"
> > -    depends on ERRATA_ANDES && MMU && ARCH_R9A07G043
> > +    depends on ERRATA_ANDES && ARCH_R9A07G043
> >      select RISCV_DMA_NONCOHERENT
> >      default y
> >      help
> > diff --git a/drivers/soc/renesas/Kconfig b/drivers/soc/renesas/Kconfig
> > index 67604f24973e..543300fe9862 100644
> > --- a/drivers/soc/renesas/Kconfig
> > +++ b/drivers/soc/renesas/Kconfig
> > @@ -334,10 +334,10 @@ if RISCV
> >  config ARCH_R9A07G043
> >      bool "RISC-V Platform support for RZ/Five"
> >      select ARCH_RZG2L
> > -    select AX45MP_L2_CACHE
> > +    select AX45MP_L2_CACHE if RISCV_DMA_NONCOHERENT
> >      select DMA_GLOBAL_POOL
> > -    select ERRATA_ANDES
> > -    select ERRATA_ANDES_CMO
> > +    select ERRATA_ANDES if RISCV_SBI
> > +    select ERRATA_ANDES_CMO if ERRATA_ANDES
> >      help
> >        This enables support for the Renesas RZ/Five SoC.
> >
> > Or am I missing a simpler way to fix this.
>
> This seems reasonable to me at least /shrug
>
Ok I will go this approach then.

Cheers,
Prabhakar