2.6.20.4 on ARM: "Unknown signal 64"????

G'day,

I am porting software from intel to ARM with linux 2.6.20.4. All goes
fine. Except, I got "Unknown signal 64" signal every time I started
fresh compiled software on target ARM machine. This is USB bus
software. And on start it calls kernel module of a corresponding
driver for the USB device. After "Unknown signal 64", software dies
but kernel module stays working. What this signal anyway????

Regards,
chelex

che_lex wrote:
> G'day,
> 
> I am porting software from intel to ARM with linux 2.6.20.4. All goes
> fine. Except, I got "Unknown signal 64" signal every time I started
> fresh compiled software on target ARM machine. This is USB bus
> software. And on start it calls kernel module of a corresponding
> driver for the USB device. After "Unknown signal 64", software dies
> but kernel module stays working. What this signal anyway????
> 
> Regards,
> chelex


It could be mis-aligned 16 or 32 bit data. The 80x86 family
is not fussy about alignment, but ARMs may do surprising
things with mis-aligned data.

I had similar problems when porting Linux code initially
tested on PC's to pure RISC architectures (Sparc, ARM).

-- 

Tauno Voipio
tauno voipio (at) iki fi

On Sep 24, 6:29 pm, Tauno Voipio <tauno.voi...@INVALIDiki.fi> wrote:
> che_lex wrote:
> > G'day,
>
> > I am porting software from intel to ARM with linux 2.6.20.4. All goes
> > fine. Except, I got "Unknown signal 64" signal every time I started
> > fresh compiled software on target ARM machine. This is USB bus
> > software. And on start it calls kernel module of a corresponding
> > driver for the USB device. After "Unknown signal 64", software dies
> > but kernel module stays working. What this signal anyway????
>
> > Regards,
> > chelex
>
> It could be mis-aligned 16 or 32 bit data. The 80x86 family
> is not fussy about alignment, but ARMs may do surprising
> things with mis-aligned data.
>
> I had similar problems when porting Linux code initially
> tested on PC's to pure RISC architectures (Sparc, ARM).
>
> --
>
> Tauno Voipio
> tauno voipio (at) iki fi

Thank you. Obliged for reply.

Is it possible to overcome these problems with compile settings or any
other way???

che_lex wrote:
> On Sep 24, 6:29 pm, Tauno Voipio <tauno.voi...@INVALIDiki.fi> wrote:
> 
>>che_lex wrote:
>>
>>>G'day,
>>
>>>I am porting software from intel to ARM with linux 2.6.20.4. All goes
>>>fine. Except, I got "Unknown signal 64" signal every time I started
>>>fresh compiled software on target ARM machine. This is USB bus
>>>software. And on start it calls kernel module of a corresponding
>>>driver for the USB device. After "Unknown signal 64", software dies
>>>but kernel module stays working. What this signal anyway????
>>
>>>Regards,
>>>chelex
>>
>>It could be mis-aligned 16 or 32 bit data. The 80x86 family
>>is not fussy about alignment, but ARMs may do surprising
>>things with mis-aligned data.
>>
>>I had similar problems when porting Linux code initially
>>tested on PC's to pure RISC architectures (Sparc, ARM).
>>
>>--
>>
>>Tauno Voipio
>>tauno voipio (at) iki fi
> 
> 
> Thank you. Obliged for reply.
> 
> Is it possible to overcome these problems with compile settings or any
> other way???


No - sorry.

Run the program in debug mode and check the place of the
trap. You may find the cause by examining the trapped
instruction and the register contents.

Of course, you need the sources to find the real culprit.

If it is a structure, you may get the desired results by
declaring with the attribute 'packed'. Check GCC manual
for details.

Please not that declaring everything 'packed' may slow
your program to a crawl.

HTH

-- 

Tauno Voipio