Re: [PATCH 2/8] SFI: include/linux/sfi.h

[Feng Tang <feng.tang@intel.com>]

On Tue, 23 Jun 2009 15:28:11 +0800
Ingo Molnar <mingo@elte.hu> wrote:

> 
> * Len Brown <lenb@kernel.org> wrote:
> 
> > +/*
> > + * Table structures must be byte-packed to match the SFI
> > specification,
> > + * as they are provided by the BIOS.
> > + */
> > +#pragma pack(1)
> 
> We use __attribute__((packed)) for that generally.
> 
> > +struct sfi_table_header {
> > +	char signature[SFI_SIGNATURE_SIZE];
> > +	u32 length;
> > +	u8 revision;
> > +	u8 checksum;
> > +	char oem_id[SFI_OEM_ID_SIZE];
> > +	char oem_table_id[SFI_OEM_TABLE_ID_SIZE];
> > +};
> 
> Please use more readable structure definitions, such as:
> 
> > +struct sfi_table_header {
> > +	char			signature[SFI_SIGNATURE_SIZE];
> > +	u32			length;
> > +	u8			revision;
> > +	u8			checksum;
> > +	char			oem_id[SFI_OEM_ID_SIZE];
> > +	char
> > oem_table_id[SFI_OEM_TABLE_ID_SIZE]; +};
> 
> > +void __init __iomem *
> > +arch_early_ioremap(unsigned long phys, unsigned long size);
> > +void __init arch_early_iounmap(void __iomem *virt, unsigned long
> > size);
> 
> Why is this in sfr.h? Should be in a separate series.

Thanks for your great comments, will address them.

For these arch_early_ioremap/arch_early_iounmap API, do you mean we should
put it in a sfi.h under "asm" folder? The reason we put it here is to
give a arch independent API declaration here and let each arch implement
its own func.

Thanks,
Feng

> 
> > +static inline int sfi_init_memory_map(void) { return -1; }
> > +static inline int sfi_init(void) { return 0; }
> > +static inline void sfi_init_late(void)	{}
> 
> Seems half-aligned. Please align consistently, such as:
> 
> static inline int sfi_init_memory_map(void)	{ return -1; }
> static inline int sfi_init(void)		{ return  0; }
> static inline void sfi_init_late(void)		{ }
> 
> or dont align at all.
> 
> 	Ingo