Problem with using PT_SYSCALL.
Hi!
I've been experimenting with ptrace() and found out that simply following
man (2) ptrace is not enough.
The attached problem is supposed to run /bin/ls / and stop at every system
call it makes. The problem is that ptrace(PT_SYSCALL,...) at the end of the
main loop seems simply to resume the child and make it run until it exits, but
not until it makes a system call.
Attached program works as expected under FreeBSD and Linux, so I believe
there is some detail that needs to be taken into account for NetBSD and that I
have missed.
Can you please give a hint how to correct this program?
---
#include stdio.h
#include stdlib.h
#include unistd.h
#include signal.h
#include string.h
#include sys/types.h
#include sys/wait.h
#include sys/ptrace.h
#include errno.h
#if defined(BSD)
#define REQ_TRACE_MEPT_TRACE_ME
#define REQ_SYSCALL PT_SYSCALL
#define REQ_KILLPT_KILL
#define CONT_ADDR (caddr_t)1
#elif defined(LINUX)
#define REQ_TRACE_MEPTRACE_TRACEME
#define REQ_SYSCALL PTRACE_SYSCALL
#define REQ_KILLPTRACE_KILL
#define CONT_ADDR NULL
#else
#error Define BSD or LINUX.
#endif
static void die(const char *msg)
{
printf(%s, errno=%d\n,msg,errno);
_exit(1);
}
static void child()
{
const char *args[] = {/bin/ls,/,NULL};
const char *env[] = {NULL};
if(0ptrace(REQ_TRACE_ME,0,NULL,0))
die(failed to trace myself);
execve(args[0],(char * const *)args,(char * const *)env);
die(failed to start the child);
}
int main()
{
pid_t p;
int status;
int sig;
p=fork();
if(p0)
die(failed to fork);
if(!p)
child();
for(;;)
{
if(0waitpid(p,status,0))
die(failed to wait for child);
if(WIFSTOPPED(status))
{
sig=WSTOPSIG(status);
printf(child stopped by signal %d
(%s)\n,sig,strsignal(sig));
if(sig!=SIGTRAP)
{
printf(signal is not SIGTRAP; killing
child\n);
if(0ptrace(REQ_KILL,p,NULL,0))
die(failed to kill the child);
break;
}
}
else if(WIFSIGNALED(status))
{
sig=WTERMSIG(status);
printf(child was signaled by signal %d (%d)\n,
sig,strsignal(sig));
if(0ptrace(REQ_KILL,p,NULL,0))
die(failed to kill the child);
break;
}
else if(WIFEXITED(status))
{
printf(child exited with status
%d\n,WEXITSTATUS(status));
break;
}
else
{
printf(neither stopped, nor signaled, nor killed?\n);
if(0ptrace(PT_KILL,p,NULL,0))
die(failed to kill the child);
break;
}
printf(resuming the child\n);
if(0ptrace(REQ_SYSCALL,p,CONT_ADDR,0))
die(failed to resume the child);
}
return 0;
}
---
Re: Problem with using PT_SYSCALL.
In article 201007161241.30007.aanisi...@inbox.ru, Ð Ñ Ñ ÐµÐ¼ Ð Ð½Ð¸Ñ Ð¸Ð¼Ð¾Ð² aanisi...@inbox.ru wrote: Hi! I've been experimenting with ptrace() and found out that simply following man (2) ptrace is not enough. The attached problem is supposed to run /bin/ls / and stop at every system call it makes. The problem is that ptrace(PT_SYSCALL,...) at the end of the main loop seems simply to resume the child and make it run until it exits, but not until it makes a system call. Attached program works as expected under FreeBSD and Linux, so I believe there is some detail that needs to be taken into account for NetBSD and that I have missed. Can you please give a hint how to correct this program? I think that the problem is not with your program, someone broke PT_SYSCALL. Please file a PR with the code so that it does not get lost. christos
Add NULL power handler to x86/ipmi(4)
Is there any reason anyone can think of to not add a NULL power handler
to the ipmi(4) driver? I can't see any reason for anything special to
happen either at suspend or resume, and the lack of a power handler
prevents the system from going to sleep at all.
Proposed patch is attached.
-
| Paul Goyette | PGP Key fingerprint: | E-mail addresses: |
| Customer Service | FA29 0E3B 35AF E8AE 6651 | paul at whooppee.com|
| Network Engineer | 0786 F758 55DE 53BA 7731 | pgoyette at juniper.net |
| Kernel Developer | | pgoyette at netbsd.org |
-Index: ipmi.c
===
RCS file: /cvsroot/src/sys/arch/x86/x86/ipmi.c,v
retrieving revision 1.46
diff -u -p -r1.46 ipmi.c
--- ipmi.c 10 Apr 2010 19:02:39 - 1.46
+++ ipmi.c 16 Jul 2010 12:19:09 -
@@ -1934,6 +1934,10 @@ ipmi_thread(void *cookie)
sc-sc_wdog.smw_tickle = ipmi_watchdog_tickle;
sysmon_wdog_register(sc-sc_wdog);
+ /* Set up a NULL power handler so we can possibly sleep */
+ if (!pmf_device_register(self, NULL, NULL))
+aprint_error_dev(self, couldn't establish a power handler\n);
+
mutex_enter(sc-sc_poll_mtx);
while (sc-sc_thread_running) {
ipmi_refresh_sensors(sc);
Re: Add NULL power handler to x86/ipmi(4)
On Fri, Jul 16, 2010 at 04:43:02PM -0700, Paul Goyette wrote: Is there any reason anyone can think of to not add a NULL power handler to the ipmi(4) driver? I can't see any reason for anything special to happen either at suspend or resume, and the lack of a power handler prevents the system from going to sleep at all. ipmi(4) should probably not suspend if its watchdog timer is active. Dave -- David Young OJC Technologies dyo...@ojctech.com Urbana, IL * (217) 278-3933
