Re: [Re: [RPX Classic/Lite and byte ordering]]

Re: [Re: [RPX Classic/Lite and byte ordering]]

[Eric]

Tim,

Have you verified that a kernel compiled with this io.h
change still works properly on your system?  This sounds
shakey to me if your kernel was already working as a whole
without your changes.

I would think that using __raw_writeX and __raw_readX combined
with the appropriate endian setting in cfi_endian.h would have
the same effect without such a drastic change.  I am more concerned
about the statement that other folks with the same boardset do not
experience this problem-- likely, the correct fix is an endian-map
(similar in spirit to LART) in cfi_endian.h rather than io.h in such 
a case.  If you can find a mythical RPX system that does not have this
error, perhaps you can modify the driver to read/write/read/write some
location to determine if you have an 'affected' board or not.

E

Tim Bowman <tbowman@emware.com> wrote:

> Ta-Da.... I fixed the problem in include/asm-ppc/io.h:
> ...
> #define readb(addr) in_8((volatile u8 *)(addr))
> #define writeb(b,addr) out_8((volatile u8 *)(addr), (b))
> #if defined(CONFIG_APUS) || defined(RPXCLASSIC)
> #define readw(addr) (*(volatile u16 *) (addr))
> #define readl(addr) (*(volatile u32 *) (addr))
> #define writew(b,addr) ((*(volatile u16 *) (addr)) = (b))
> #define writel(b,addr) ((*(volatile u32 *) (addr)) = (b))
> #else

[ ... ]

> This solution doesn't explain why some of you haven't had this problem. 
> Maybe the flash really is wired up two different ways on RPX boards...



____________________________________________________________________
Get free email and a permanent address at http://www.netaddress.com/?N=1


To unsubscribe, send "unsubscribe mtd" to majordomo@infradead.org

Re: [Re: [RPX Classic/Lite and byte ordering]]

[Tim Bowman]

Eric wrote:
> 
> Tim,
> 
> Have you verified that a kernel compiled with this io.h
> change still works properly on your system?  This sounds
> shakey to me if your kernel was already working as a whole
> without your changes.
> 

It seemed to, but I haven't really tested everything thoroughly.

> I would think that using __raw_writeX and __raw_readX combined
> with the appropriate endian setting in cfi_endian.h would have
> the same effect without such a drastic change.  I am more concerned
> about the statement that other folks with the same boardset do not
> experience this problem-- likely, the correct fix is an endian-map
> (similar in spirit to LART) in cfi_endian.h rather than io.h in such
> a case.  If you can find a mythical RPX system that does not have this
> error, perhaps you can modify the driver to read/write/read/write some
> location to determine if you have an 'affected' board or not.
> 

Sound advice, I think.  It looks like some net and char code uses the
readX stuff, so I'll go ahead and make the changes in rpxlite.c
instead.  I have 4 or 5 other RPX boards I can try, so I'll let you know
if I find one that is different.  Otherwise, we might want to make these
changes to the tree.

-T


To unsubscribe, send "unsubscribe mtd" to majordomo@infradead.org