ADS GCP reboots when running the application!

ADS GCP reboots when running the application!

[rpm]

Hi !
   I am using  "2.4.9-ac10-rmk2-np1-ads3"  kernel on ADS Graphic Client Board 
(StrongARM). When i run the following c++  code the system reboots. 

************************************************************************
#include <iostream>
#include <fstream>
#include <string>
#include <vector>
#include <iterator>

using namespace std;

int main(int argc, char** argv)
{
    string raw_filename = "input.out";
    ifstream raw_file(raw_filename.c_str());

    vector<long> data;
   
copy(istream_iterator<long>(raw_file),istream_iterator<long>(),back_inserter(data));

    data.erase(data.begin());

    return 0;
}
**********************************************************************
Here, "input.out" is a file containing 10000 numbers. If the numbers in the 
file "input.out" are 5000 then the system works fine. 
    The kernel is not showing any OOPS or panic  , it just reboots ! what i 
think is that some double fault ( fault inside fault handler ) or  something 
similar to that might be causing the precessor to reboot. Can someone  give 
some direction  !

regards,
rpm

ps: I am not attaching the "input.out" file due to large size if some wants 
to test , i can send the file !

Re: ADS GCP reboots when running the application!

[Russell King]

On Mon, May 13, 2002 at 05:08:30PM +0530, rpm wrote:
>     The kernel is not showing any OOPS or panic  , it just reboots !

Weird.  Tried any more recent kernels?

> what i think is that some double fault ( fault inside fault handler )

No such thing on ARMs.  If you take a fault while handling one, you
re-enter the fault handler - you don't reboot.

Alternatively, you could send the ARM Linux mailing lists/me a binary
to try on various other ARM machines/kernels.  (see
http://www.arm.linux.org.uk/armlinux/mailinglists.php)

(Note, I'm unlikely to get time to try it here, between trying to get
2.5.15 working, trying to review patches people have sent over the
last two weeks, maybe actually applying some of them, sending stuff
to Linus, finding and killing bugs, tackling the odd hardware failure,
updating my web diary, and maybe even producing an ARM patch for
something later than 2.5.10-rmk1 which was released a fortnight last
Sunday ago.) ;(

-- 
Russell King (rmk@arm.linux.org.uk)                The developer of ARM Linux
             http://www.arm.linux.org.uk/personal/aboutme.html

Re: ADS GCP reboots when running the application!

[rpm]

On Monday 13 May 2002 10:07 pm, Russell King wrote:
> On Mon, May 13, 2002 at 05:08:30PM +0530, rpm wrote:
> >     The kernel is not showing any OOPS or panic  , it just reboots !
>
> Weird.  Tried any more recent kernels?
>
> > what i think is that some double fault ( fault inside fault handler )
>
> No such thing on ARMs.  If you take a fault while handling one, you
> re-enter the fault handler - you don't reboot.
>
  
What if the fault handler does a fault  ( like seg fault in seg fault handler 
)  , cause in i386, i remember such a situation causes a processor reboot as  
it becomes a infinite loop !

the same code works fine on an iPAQ  ! without any problems 
and i did a diff of the strace output on iPAQ and GCP and found that the 
following lines are extra in case of iPAQ while GCP reboots before it can 
print the lines !
***************************************
brk(0xc8000)                                         = 0xc8000     
close(4)                                                 = 0                
close(3)                                                = 0                   
munmap(0x40000000, 4096)                 = 0                   
_exit(0)                                                = ?

*****************************************
so i conclude that the system crashes in brk() sys call !

If you can point out  the cases where the kernel reboots without showing any 
message ,   then it will be easier to debug for me! 
and thanks a lot for the reply  :)  

rpm   

Re: ADS GCP reboots when running the application!

[Russell King]

On Tue, May 14, 2002 at 02:41:06PM +0530, rpm wrote:
> On Monday 13 May 2002 10:07 pm, Russell King wrote:
> > No such thing on ARMs.  If you take a fault while handling one, you
> > re-enter the fault handler - you don't reboot.
>   
> What if the fault handler does a fault  ( like seg fault in seg fault handler 
> )  , cause in i386, i remember such a situation causes a processor reboot as  
> it becomes a infinite loop !

You're right in the x86 case, but wrong in the ARM case - you just take a
fault after fault after fault (and you'll either end up overwriting the
kernel or something else of that nature).

> so i conclude that the system crashes in brk() sys call !

strace would print 'brk(' on entry to the syscall though.

> If you can point out  the cases where the kernel reboots without showing
> any message ,   then it will be easier to debug for me!

Well, if I knew of any, they'd get removed/fixed pretty damned fast.

-- 
Russell King (rmk@arm.linux.org.uk)                The developer of ARM Linux
             http://www.arm.linux.org.uk/personal/aboutme.html