[PATCH] riscv: Fix efi header for RV32

[Leo Liang <ycliang@andestech.com>]

On Wed, Nov 11, 2020 at 04:25:53PM +0800, Leo Liang wrote:
> Hi Atish and Heinrich,
> 
> On Tue, Oct 13, 2020 at 12:23:31PM -0700, Atish Patra wrote:
> > RV32 should use PE32 format instead of PE32+ as the efi header format.
> > This requires following changes
> > 1. A different header magic value
> > 2. An additional parameter known as BaseOfData. Currently, it is set to
> >    zero in absence of any usage.
> > 
> > Signed-off-by: Atish Patra <atish.patra@wdc.com>
> > Reviewed-by: Bin Meng <bin.meng@windriver.com>
> > Reviewed-by: Rick Chen <rick@andestech.com>
> > ---
> >  arch/riscv/lib/crt0_riscv_efi.S | 7 ++++++-
> >  1 file changed, 6 insertions(+), 1 deletion(-)
> > 
> > diff --git a/arch/riscv/lib/crt0_riscv_efi.S b/arch/riscv/lib/crt0_riscv_efi.S
> > index 87fe1e56f906..4aaa49ad0777 100644
> > --- a/arch/riscv/lib/crt0_riscv_efi.S
> > +++ b/arch/riscv/lib/crt0_riscv_efi.S
> > @@ -15,11 +15,13 @@
> >  #define SAVE_LONG(reg, idx)	sd	reg, (idx*SIZE_LONG)(sp)
> >  #define LOAD_LONG(reg, idx)	ld	reg, (idx*SIZE_LONG)(sp)
> >  #define PE_MACHINE	IMAGE_FILE_MACHINE_RISCV64
> > +#define PE_MAGIC        IMAGE_NT_OPTIONAL_HDR64_MAGIC
> >  #else
> >  #define SIZE_LONG	4
> >  #define SAVE_LONG(reg, idx)	sw	reg, (idx*SIZE_LONG)(sp)
> >  #define LOAD_LONG(reg, idx)	lw	reg, (idx*SIZE_LONG)(sp)
> >  #define PE_MACHINE	IMAGE_FILE_MACHINE_RISCV32
> > +#define PE_MAGIC        IMAGE_NT_OPTIONAL_HDR32_MAGIC
> >  #endif
> >  
> >  
> > @@ -48,7 +50,7 @@ coff_header:
> >  		 IMAGE_FILE_LOCAL_SYMS_STRIPPED | \
> >  		 IMAGE_FILE_DEBUG_STRIPPED)
> >  optional_header:
> > -	.short	IMAGE_NT_OPTIONAL_HDR64_MAGIC	/* PE32+ format */
> > +	.short	PE_MAGIC			/* PE32+ format */
> >  	.byte	0x02				/* MajorLinkerVersion */
> >  	.byte	0x14				/* MinorLinkerVersion */
> >  	.long	_edata - _start			/* SizeOfCode */
> > @@ -56,6 +58,9 @@ optional_header:
> >  	.long	0				/* SizeOfUninitializedData */
> >  	.long	_start - ImageBase		/* AddressOfEntryPoint */
> >  	.long	_start - ImageBase		/* BaseOfCode */
> > +#if __riscv_xlen == 32
> > +	.long	0				/* BaseOfData */
> > +#endif
> >  
> >  extra_header_fields:
> >  	.quad	0				/* ImageBase */
> 
> diff --git a/arch/riscv/lib/crt0_riscv_efi.S b/arch/riscv/lib/crt0_riscv_efi.S
> index 4aaa49ad07..f47c4a6d82 100644
> --- a/arch/riscv/lib/crt0_riscv_efi.S
> +++ b/arch/riscv/lib/crt0_riscv_efi.S
> @@ -63,7 +63,11 @@ optional_header:
>  #endif
> 
>  extra_header_fields:
> +#if __riscv_xlen == 32
> +       .long   0                               /* ImageBase */
> +#else if
>         .quad   0                               /* ImageBase */
> +#endif
>         .long   0x20                            /* SectionAlignment */
>         .long   0x8                             /* FileAlignment */
>         .short  0                               /* MajorOperatingSystemVersion */
> @@ -83,10 +87,17 @@ extra_header_fields:
>         .long   0                               /* CheckSum */
>         .short  IMAGE_SUBSYSTEM_EFI_APPLICATION /* Subsystem */
>         .short  0                               /* DllCharacteristics */
> +#if __riscv_xlen == 32
> +       .long   0                               /* SizeOfStackReserve */
> +       .long   0                               /* SizeOfStackCommit */
> +       .long   0                               /* SizeOfHeapReserve */
> +       .long   0                               /* SizeOfHeapCommit */
> +#else if
>         .quad   0                               /* SizeOfStackReserve */
>         .quad   0                               /* SizeOfStackCommit */
>         .quad   0                               /* SizeOfHeapReserve */
>         .quad   0                               /* SizeOfHeapCommit */
> +#endif
>         .long   0                               /* LoaderFlags */
>         .long   0x6                             /* NumberOfRvaAndSizes */
> 
> Based on Atish's patch, 
> would the above modification fit more closely to the specification that Heinrich provided ?
> 
> Best regards,
> Leo

Sending a new patch seems more reasonable.
I will send a new patch later based on this patch.

Best regards,
Leo