Skip to contentSkip to conversation

correct seg fault address on ia-64??

Linux-ia64 Pereira, D LailaX E 7 participants 8 messages Nov 14, 2001

lore
Share & tools
lore
#0 lore raw

Hi ,

In the following program, I was trying to obtain the address where a segmentation fault is caused.
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <signal.h>
#include <sys/mman.h>

int *x;
int page_size;

void segv_handler( int sig, siginfo_t *sip,struct sigcontext *scp){

  void *vadr=sip->si_addr;
  unsigned long *pc;
  unsigned long instr;
  int  readwrite;

  if (sig=SIGSEGV){
    printf("\nSegv handler ..\n");
  }
  printf("vaddr : %x \n",vadr);

readwrite=(((*(unsigned long*)scp->sc_ip)>>21)&1); printf("Read or write %d\n",readwrite); exit(1);
}

int main(){
int s;
struct sigaction sa;
int temp;

  page_size=getpagesize();
  printf("Pagesize : %d ,PID=%d\n",page_size,(int)getpid());
  fflush(stdout);

  /*SEGV handler setup*/
  sa.sa_handler=(void*)&segv_handler;
  sigemptyset(&sa.sa_mask);
  sigaddset(&sa.sa_mask,SIGIO);
  sigaddset(&sa.sa_mask,SIGALRM);

  sa.sa_flags=SA_SIGINFO;

  if (sigaction(SIGSEGV,&sa,NULL)){
    printf(" Error assigning signal!\n");
  }

  x=(int*)malloc(2*page_size);

/* Align to a multiple of page_size, assumed to be a power of two */
x = (int *)((long)(((int) (long)x + page_size-1) & ~(page_size-1))); printf("Address: %x\n",x );
//x[0]E6;

s=mprotect(x,page_size,PROT_NONE); //make it none access printf("Page protection : NONE : try reading the page\n "); printf("segv should arise now ...\n");
//read the page ... segv?
temp=x[0];
return 0;
}

When I run the program on a ia-64 machine (linux OS) , then I get the following output:
Pagesize : 16384 ,PID(513
Address: 4000
Page protection : NONE : try reading the page
segv should arise now ...

Segv handler ..
vaddr : 4000
Read or write 0

.............
However , when I ran the program thru gdb, I got the following:
Reading symbols from a.out...done.
(gdb) run ex3.c
......
Pagesize : 16384 ,PID(471
Address: 4000
Page protection : NONE : try reading the page
segv should arise now ...

Program received signal SIGSEGV, Segmentation fault. 0x40000000000010e0 in main ()
(gdb) where
#0 0x40000000000010e0 in main ()
#1 0x20000000000906b0 in __libc_start_main (

    main=0x40000000000013e8 <_fini+408>, argc=2, 
ubp_av=0x80000fffffffb848, 
    init=0x40000000000013b8 <_fini+360>, 
    fini=0x200000000003db30 <_dl_debug_mask>, 
    rtld_fini=0x20000000002581c8 <_dl_get_origin+632>, 
    stack_end=0x80000fffffffb840) at ../sysdeps/generic/libc-start.c:129

#2 0x40000000000010b0 in main ()
(gdb)....

So thru, gdb the seg fault address is "0x40000000000010e0" , is my interpretation right ??
Why is that when I run the program on itanium , I get the address as "4000" and thru gdb as " 0x40000000000010e0". What is wrong?? Please let me know.

d'laila

#1 lore raw

D'Laila-

Remember that longs and pointers are 64 bits. The %x' format to printf' says to print out a 32-bit integer so you wind up truncating the upper 32-bits of the address. Just change your format to `%lx' and you should print out the right address.

Quoted 115 lines · D LailaX E

Hi ,

In the following program, I was trying to obtain the address where a segmentation fault is caused.
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <signal.h>
#include <sys/mman.h>

int *x;
int page_size;

void segv_handler( int sig, siginfo_t *sip,struct sigcontext *scp){

  void *vadr=sip->si_addr;
  unsigned long *pc;
  unsigned long instr;
  int  readwrite;

  if (sig=SIGSEGV){
    printf("\nSegv handler ..\n");
  }
  printf("vaddr : %x \n",vadr);

readwrite=(((*(unsigned long*)scp->sc_ip)>>21)&1); printf("Read or write %d\n",readwrite); exit(1);
}

int main(){
int s;
struct sigaction sa;
int temp;

  page_size=getpagesize();
  printf("Pagesize : %d ,PID=%d\n",page_size,(int)getpid());
  fflush(stdout);

  /*SEGV handler setup*/
  sa.sa_handler=(void*)&segv_handler;
  sigemptyset(&sa.sa_mask);
  sigaddset(&sa.sa_mask,SIGIO);
  sigaddset(&sa.sa_mask,SIGALRM);

  sa.sa_flags=SA_SIGINFO;

  if (sigaction(SIGSEGV,&sa,NULL)){
    printf(" Error assigning signal!\n");
  }

  x=(int*)malloc(2*page_size);

/* Align to a multiple of page_size, assumed to be a power of two */
x = (int *)((long)(((int) (long)x + page_size-1) & ~(page_size-1))); printf("Address: %x\n",x );
//x[0]E6;

s=mprotect(x,page_size,PROT_NONE); //make it none access printf("Page protection : NONE : try reading the page\n "); printf("segv should arise now ...\n");
//read the page ... segv?
temp=x[0];
return 0;
}

When I run the program on a ia-64 machine (linux OS) , then I get the following output:
Pagesize : 16384 ,PID(513
Address: 4000
Page protection : NONE : try reading the page
segv should arise now ...

Segv handler ..
vaddr : 4000
Read or write 0

.............
However , when I ran the program thru gdb, I got the following:
Reading symbols from a.out...done.
(gdb) run ex3.c
......
Pagesize : 16384 ,PID(471
Address: 4000
Page protection : NONE : try reading the page
segv should arise now ...

Program received signal SIGSEGV, Segmentation fault. 0x40000000000010e0 in main ()
(gdb) where
#0 0x40000000000010e0 in main ()
#1 0x20000000000906b0 in __libc_start_main (

    main=0x40000000000013e8 <_fini+408>, argc=2, 
ubp_av=0x80000fffffffb848, 
    init=0x40000000000013b8 <_fini+360>, 
    fini=0x200000000003db30 <_dl_debug_mask>, 
    rtld_fini=0x20000000002581c8 <_dl_get_origin+632>, 
    stack_end=0x80000fffffffb840) at ../sysdeps/generic/libc-start.c:129

#2 0x40000000000010b0 in main ()
(gdb)....

So thru, gdb the seg fault address is "0x40000000000010e0" , is my interpretation right ??
Why is that when I run the program on itanium , I get the address as "4000" and thru gdb as " 0x40000000000010e0". What is wrong?? Please let me know.

d'laila
_______________________________________________
Linux-IA64 mailing list
[email protected]
http://lists.linuxia64.org/lists/listinfo/linux-ia64

Signature · 4 lines
Don Dugger
"Censeo Toto nos in Kansa esse decisse." - D. Gale
[email protected]
Ph: 303/652-0870x117
#2 lore raw
In reply to Pereira, D LailaX E

"Pereira, D LailaX E" <[email protected]> writes:

|> /* Align to a multiple of page_size, assumed to be a power of two */
|> x = (int *)((long)(((int) (long)x + page_size-1) & ~(page_size-1)));

Never ever cast pointers to int! Modern C gives you all you need:
uintptr_t.

Andreas.

Signature · 4 lines
Andreas Schwab                                  "And now for something
[email protected]				completely different."
SuSE Labs, SuSE GmbH, Schanzäckerstr. 10, D-90443 Nürnberg
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
#3 lore raw
In reply to Pereira, D LailaX E
Quoted 4 lines · D LailaX E
  x=(int*)malloc(2*page_size);

/* Align to a multiple of page_size, assumed to be a power of two */
x = (int *)((long)(((int) (long)x + page_size-1) & ~(page_size-1)));

                        ^^^^^

This conversion is the WRONG thing: you are losing significant bits from the 64-bit pointer's value!

Quoted 1 line
  printf("Address:  %x\n",x );

And this is the wrong way of displaying a pointer, which is not a 32-bit entity and definitely not an unsigned int. The right way of doing it is:

    printf("Address:  %p\n", (void *)x);
Quoted 21 lines

.............
However , when I ran the program thru gdb, I got the following:
Reading symbols from a.out...done.
(gdb) run ex3.c
......
Pagesize : 16384 ,PID(471
Address: 4000
Page protection : NONE : try reading the page
segv should arise now ...

Program received signal SIGSEGV, Segmentation fault. 0x40000000000010e0 in main ()
(gdb) where
#0 0x40000000000010e0 in main ()
#1 0x20000000000906b0 in __libc_start_main (

    main=0x40000000000013e8 <_fini+408>, argc=2, 
ubp_av=0x80000fffffffb848, 
    init=0x40000000000013b8 <_fini+360>, 
    fini=0x200000000003db30 <_dl_debug_mask>, 
    rtld_fini=0x20000000002581c8 <_dl_get_origin+632>, 
    stack_end=0x80000fffffffb840) at ../sysdeps/generic/libc-start.c:129
Quoted 8 lines

#2 0x40000000000010b0 in main ()
(gdb)....

So thru, gdb the seg fault address is "0x40000000000010e0" , is my interpretation right ??
Why is that when I run the program on itanium , I get the address as "4000" and thru gdb as " 0x40000000000010e0". What is wrong?? Please let me know.

You're confusing data addresses and code addresses. You're displaying the value of a pointer, which is a (badly formed) data address, while gdb is displaying the address of the instruction that generated the fault, i.e. a code address.

Dan

#4 lore raw
In reply to Pereira, D LailaX E

Such horrible things are being done in source code in the 64bit
environment. Is there any kind of "porting guide" available?
A long time ago we did a porting guide for MIPSPro, see

http://techpubs.sgi.com/library/tpl/cgi-bin/init.cgi
in search enter: mipspro 64-bit
and take a look at the MIPSpro 64-bit Porting and Transition Guide, select Chapter 3, second portion (first portion is Fortran, second is C).

What is out there for gcc?
Thanks, Rich

#5 lore raw
In reply to Pereira, D LailaX E
Quoted 1 line · Rich Altmaier

What is out there for gcc?

gcc -W -Werror..

Seriously, gcc supports 64bit targets for a _very_ long time now (more than 5 years) - people should have gotten their code cleaned up..

Christoph

Signature · 1 line
Of course it doesn't work. We've performed a software upgrade.
#6 lore raw
In reply to Pereira, D LailaX E

Rich-

A while back someone from SCO made a presentation at one of the early Intel Developer Forums about just this topic. I can't find it in a quick search of either the SCO (now Caldera) or Intel sites. Maybe someone from Caldera or Intel can see if that presentation is still available.

Quoted 18 lines · Rich Altmaier

Such horrible things are being done in source code in the 64bit
environment. Is there any kind of "porting guide" available?
A long time ago we did a porting guide for MIPSPro, see

http://techpubs.sgi.com/library/tpl/cgi-bin/init.cgi
in search enter: mipspro 64-bit
and take a look at the MIPSpro 64-bit Porting and Transition Guide, select Chapter 3, second portion (first portion is Fortran, second is C).

What is out there for gcc?
Thanks, Rich

_______________________________________________
Linux-IA64 mailing list
[email protected]
http://lists.linuxia64.org/lists/listinfo/linux-ia64

Signature · 4 lines
Don Dugger
"Censeo Toto nos in Kansa esse decisse." - D. Gale
[email protected]
Ph: 303/652-0870x117
#7 lore raw
In reply to Pereira, D LailaX E

Hi,

perhaps you meant our presentation on Turbolinux for IA64 at the August 2000 Forum which included a list of issues when porting from IA32 to IA64. The presentation is included in the attachment.

Cheers,
Rok

Quoted 39 lines · On Wed, 14 Nov 2001 [email protected] wrote:

Rich-

A while back someone from SCO made a presentation at one of the early Intel Developer Forums about just this topic. I can't find it in a quick search of either the SCO (now Caldera) or Intel sites. Maybe someone from Caldera or Intel can see if that presentation is still available.

Quoted 18 lines · Rich Altmaier

Such horrible things are being done in source code in the 64bit
environment. Is there any kind of "porting guide" available?
A long time ago we did a porting guide for MIPSPro, see

http://techpubs.sgi.com/library/tpl/cgi-bin/init.cgi
in search enter: mipspro 64-bit
and take a look at the MIPSpro 64-bit Porting and Transition Guide, select Chapter 3, second portion (first portion is Fortran, second is C).

What is out there for gcc?
Thanks, Rich

_______________________________________________
Linux-IA64 mailing list
[email protected]
http://lists.linuxia64.org/lists/listinfo/linux-ia64

Signature · 9 lines
Don Dugger
"Censeo Toto nos in Kansa esse decisse." - D. Gale
[email protected]
Ph: 303/652-0870x117

_______________________________________________
Linux-IA64 mailing list
[email protected]
http://lists.linuxia64.org/lists/listinfo/linux-ia64