Re: alpha copy_to_user (was Re: NFS fails under 2.2.6)

Keith Owens (kaos@ocs.com.au)
Thu, 22 Apr 1999 19:50:34 +1000


On 22 Apr 1999 11:16:17 +0200,
Andi Kleen <ak@muc.de> wrote:
>Keith Owens <kaos@ocs.com.au> writes:
>> Until that blissful state occurs, Alpha users need to specify "-c 4"
>> when running ksymoops. That tells ksymoops that the input is 4 byte
>> chunks that need to be byte swapped.
>
>Couldn't the endianess be guessed from the vmlinux image?

ksymoops does not read vmlinux by default. I was also thinking about
users who want to do cross system debugging. That is, those working on
embedded systems or porting to new systems. Debugging a big endianess
Oops (say i370) on a little endian machine (say i386) needs a common
interface. Fortunately we have one, it's called network order :).

The real problem is the endianess ambiguity on *some* systems when they
dump their code lines. AFAIK, all archs use network order for
/proc/ksyms, System.map, EIP/PC, registers, stack trace, objdump etc.
Some archs dump Code: in network order as well. Only a few systems
cause problems by dump the Code: line with the native endianess. IMHO,
the traps.c code for those systems should be converted to network order
instead of complicating ksymoops even further.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/