Re: 2.6.9-rc1-mm3

[Paulo Marques <pmarques@grupopie.com>]

William Lee Irwin III wrote:
> On Fri, Sep 03, 2004 at 06:51:08PM +0100, Paulo Marques wrote:
> 
>>Could you send me the .tmp_kallsyms2.S and System.map files from
>>this kernel build, please, please, please?
>>I really want to address this problem, but without hardware and
>>without more information I'm a little in the dark (although
>>looking at the resulting names already gives some clues).
>>Also, doing a "cat /proc/kallsyms" shows the same kind of behavior,
>>doesn't it? (just to be sure)
> 
> 
> cat /proc/kallsyms also exhibits this problem.
> 
> The data will appear shortly at:
> 
> ftp://ftp.kernel.org/pub/linux/kernel/people/wli/misc/kallsyms2.S-sparc64.gz
> ftp://ftp.kernel.org/pub/linux/kernel/people/wli/misc/System.map-2.6.9-rc1-mm3-sparc64.gz

Thank for all the information!

Looking at the data I found out that the "_up_up_up" is in fact the
token with code "0", which means that the token "0" was being used a
lot more than the others.

This pointed me in the direction of the bug.

My error was to assume the .word assembler directive meant a 16-bit
unsigned integer, when in fact it depends on the architecture and is
32 bits on sparc :(

This one liner should solve the problem.

Please verify that in fact it does solve it, and I'll send a proper
"[PATCH]" message to be included in the next version.

-- 
Paulo Marques - www.grupopie.com

To err is human, but to really foul things up requires a computer.
Farmers' Almanac, 1978

--- linux-2.6.9-rc1-mm3/scripts/kallsyms.c      2004-09-05 21:51:14.000000000 +0100
+++ linux-2.6.9-rc1-kall/scripts/kallsyms.c     2004-09-05 21:52:38.000000000 +0100
@@ -303,7 +303,7 @@ write_src(void)

         output_label("kallsyms_token_index");
         for (i = 0; i < 256; i++)
-               printf("\t.word\t%d\n", best_idx[i]);
+               printf("\t.short\t%d\n", best_idx[i]);
         printf("\n");
  }