Transcript ‘struct sigcontext’ On using Linux’s signaling mechanism for debugqing application programs

‘struct sigcontext’
On using Linux’s signaling
mechanism for debugqing
application programs
Installing a signal-handler
#include <signal.h>
void handler( int signo, siginfo_t *si, void *sc );
int main( void )
{
struct sigaction oa, sa = {0};
sa.sa_flags = SA_SIGINFO;
sa.sa_sigaction = handler;
sigaction( SIGSEGV, &sa, &oa );
}
Layout of kernel-stack (by cpu)
Upon entry to kernel’s exception-handler:
SS
ESP
saved pointer
to user-stack
EFLAGS
CS
EIP
SS:ESP
saved pointer
to user-opcode
error
code
‘error code’ = segment-selector for the illegal memory-segment access
Layout of kernel-stack (by Linux)
ss
esp
eflags
cs
eip
error
gs
fs
es
ds
SS:ESP
eax
ebp
edi
esi
edx
ecx
ebx
Layout of user-stack
Upon entering ‘main()’
tos
Upon entering ‘handler()’
environment strings
environment strings
command-line arguments
command-line arguments
envp
argv
argc
return-address
envp
argv
argc
return-address
sigstub
sigcontext
siginfo
tos
sc
si
signum
return-address
‘segvtrap.cpp’
• We have constructed this demo-program
to show how you could utilize the context
information that Linux can provide to your
signal-handler (if you use SA_SIGINFO)
• You can use our ‘run.cpp’ tool to examine
an application-program’s exit-status
In-class exercises
• Try commenting out the ‘exit(1)’ statement
• Then add statement: sc ->eip += 1;
• Try replacing ‘asm(“ hlt “)’ statement with
this privileged instruction: inb( 0x1F7 );
• Try replacing ‘asm(“ hlt “)’ statement with
this illegal assignment: *(char*)0 = 0;
• Try replacing ‘asm(“ hlt” )’ statement with
an instruction that reads a kernel-address