minix/test/test42.c

1441 lines
28 KiB
C
Raw Normal View History

Merge of David's ptrace branch. Summary: o Support for ptrace T_ATTACH/T_DETACH and T_SYSCALL o PM signal handling logic should now work properly, even with debuggers being present o Asynchronous PM/VFS protocol, full IPC support for senda(), and AMF_NOREPLY senda() flag DETAILS Process stop and delay call handling of PM: o Added sys_runctl() kernel call with sys_stop() and sys_resume() aliases, for PM to stop and resume a process o Added exception for sending/syscall-traced processes to sys_runctl(), and matching SIGKREADY pseudo-signal to PM o Fixed PM signal logic to deal with requests from a process after stopping it (so-called "delay calls"), using the SIGKREADY facility o Fixed various PM panics due to race conditions with delay calls versus VFS calls o Removed special PRIO_STOP priority value o Added SYS_LOCK RTS kernel flag, to stop an individual process from running while modifying its process structure Signal and debugger handling in PM: o Fixed debugger signals being dropped if a second signal arrives when the debugger has not retrieved the first one o Fixed debugger signals being sent to the debugger more than once o Fixed debugger signals unpausing process in VFS; removed PM_UNPAUSE_TR protocol message o Detached debugger signals from general signal logic and from being blocked on VFS calls, meaning that even VFS can now be traced o Fixed debugger being unable to receive more than one pending signal in one process stop o Fixed signal delivery being delayed needlessly when multiple signals are pending o Fixed wait test for tracer, which was returning for children that were not waited for o Removed second parallel pending call from PM to VFS for any process o Fixed process becoming runnable between exec() and debugger trap o Added support for notifying the debugger before the parent when a debugged child exits o Fixed debugger death causing child to remain stopped forever o Fixed consistently incorrect use of _NSIG Extensions to ptrace(): o Added T_ATTACH and T_DETACH ptrace request, to attach and detach a debugger to and from a process o Added T_SYSCALL ptrace request, to trace system calls o Added T_SETOPT ptrace request, to set trace options o Added TO_TRACEFORK trace option, to attach automatically to children of a traced process o Added TO_ALTEXEC trace option, to send SIGSTOP instead of SIGTRAP upon a successful exec() of the tracee o Extended T_GETUSER ptrace support to allow retrieving a process's priv structure o Removed T_STOP ptrace request again, as it does not help implementing debuggers properly o Added MINIX3-specific ptrace test (test42) o Added proper manual page for ptrace(2) Asynchronous PM/VFS interface: o Fixed asynchronous messages not being checked when receive() is called with an endpoint other than ANY o Added AMF_NOREPLY senda() flag, preventing such messages from satisfying the receive part of a sendrec() o Added asynsend3() that takes optional flags; asynsend() is now a #define passing in 0 as third parameter o Made PM/VFS protocol asynchronous; reintroduced tell_fs() o Made PM_BASE request/reply number range unique o Hacked in a horrible temporary workaround into RS to deal with newly revealed RS-PM-VFS race condition triangle until VFS is asynchronous System signal handling: o Fixed shutdown logic of device drivers; removed old SIGKSTOP signal o Removed is-superuser check from PM's do_procstat() (aka getsigset()) o Added sigset macros to allow system processes to deal with the full signal set, rather than just the POSIX subset Miscellaneous PM fixes: o Split do_getset into do_get and do_set, merging common code and making structure clearer o Fixed setpriority() being able to put to sleep processes using an invalid parameter, or revive zombie processes o Made find_proc() global; removed obsolete proc_from_pid() o Cleanup here and there Also included: o Fixed false-positive boot order kernel warning o Removed last traces of old NOTIFY_FROM code THINGS OF POSSIBLE INTEREST o It should now be possible to run PM at any priority, even lower than user processes o No assumptions are made about communication speed between PM and VFS, although communication must be FIFO o A debugger will now receive incoming debuggee signals at kill time only; the process may not yet be fully stopped o A first step has been made towards making the SYSTEM task preemptible
2009-09-30 11:57:22 +02:00
/* Tests for MINIX3 ptrace(2) - by D.C. van Moolenbroek */
#define _POSIX_SOURCE 1
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <signal.h>
#include <unistd.h>
#include <errno.h>
#include <sys/wait.h>
#include <sys/select.h>
Merge of David's ptrace branch. Summary: o Support for ptrace T_ATTACH/T_DETACH and T_SYSCALL o PM signal handling logic should now work properly, even with debuggers being present o Asynchronous PM/VFS protocol, full IPC support for senda(), and AMF_NOREPLY senda() flag DETAILS Process stop and delay call handling of PM: o Added sys_runctl() kernel call with sys_stop() and sys_resume() aliases, for PM to stop and resume a process o Added exception for sending/syscall-traced processes to sys_runctl(), and matching SIGKREADY pseudo-signal to PM o Fixed PM signal logic to deal with requests from a process after stopping it (so-called "delay calls"), using the SIGKREADY facility o Fixed various PM panics due to race conditions with delay calls versus VFS calls o Removed special PRIO_STOP priority value o Added SYS_LOCK RTS kernel flag, to stop an individual process from running while modifying its process structure Signal and debugger handling in PM: o Fixed debugger signals being dropped if a second signal arrives when the debugger has not retrieved the first one o Fixed debugger signals being sent to the debugger more than once o Fixed debugger signals unpausing process in VFS; removed PM_UNPAUSE_TR protocol message o Detached debugger signals from general signal logic and from being blocked on VFS calls, meaning that even VFS can now be traced o Fixed debugger being unable to receive more than one pending signal in one process stop o Fixed signal delivery being delayed needlessly when multiple signals are pending o Fixed wait test for tracer, which was returning for children that were not waited for o Removed second parallel pending call from PM to VFS for any process o Fixed process becoming runnable between exec() and debugger trap o Added support for notifying the debugger before the parent when a debugged child exits o Fixed debugger death causing child to remain stopped forever o Fixed consistently incorrect use of _NSIG Extensions to ptrace(): o Added T_ATTACH and T_DETACH ptrace request, to attach and detach a debugger to and from a process o Added T_SYSCALL ptrace request, to trace system calls o Added T_SETOPT ptrace request, to set trace options o Added TO_TRACEFORK trace option, to attach automatically to children of a traced process o Added TO_ALTEXEC trace option, to send SIGSTOP instead of SIGTRAP upon a successful exec() of the tracee o Extended T_GETUSER ptrace support to allow retrieving a process's priv structure o Removed T_STOP ptrace request again, as it does not help implementing debuggers properly o Added MINIX3-specific ptrace test (test42) o Added proper manual page for ptrace(2) Asynchronous PM/VFS interface: o Fixed asynchronous messages not being checked when receive() is called with an endpoint other than ANY o Added AMF_NOREPLY senda() flag, preventing such messages from satisfying the receive part of a sendrec() o Added asynsend3() that takes optional flags; asynsend() is now a #define passing in 0 as third parameter o Made PM/VFS protocol asynchronous; reintroduced tell_fs() o Made PM_BASE request/reply number range unique o Hacked in a horrible temporary workaround into RS to deal with newly revealed RS-PM-VFS race condition triangle until VFS is asynchronous System signal handling: o Fixed shutdown logic of device drivers; removed old SIGKSTOP signal o Removed is-superuser check from PM's do_procstat() (aka getsigset()) o Added sigset macros to allow system processes to deal with the full signal set, rather than just the POSIX subset Miscellaneous PM fixes: o Split do_getset into do_get and do_set, merging common code and making structure clearer o Fixed setpriority() being able to put to sleep processes using an invalid parameter, or revive zombie processes o Made find_proc() global; removed obsolete proc_from_pid() o Cleanup here and there Also included: o Fixed false-positive boot order kernel warning o Removed last traces of old NOTIFY_FROM code THINGS OF POSSIBLE INTEREST o It should now be possible to run PM at any priority, even lower than user processes o No assumptions are made about communication speed between PM and VFS, although communication must be FIFO o A debugger will now receive incoming debuggee signals at kill time only; the process may not yet be fully stopped o A first step has been made towards making the SYSTEM task preemptible
2009-09-30 11:57:22 +02:00
#include <sys/ptrace.h>
#define ITERATIONS 3
#define MAX_ERROR 4
#define _WIFSTOPPED(s) (WIFSTOPPED(s) && !WIFSIGNALED(s) && !WIFEXITED(s))
#define _WIFSIGNALED(s) (!WIFSTOPPED(s) && WIFSIGNALED(s) && !WIFEXITED(s))
#define _WIFEXITED(s) (!WIFSTOPPED(s) && !WIFSIGNALED(s) && WIFEXITED(s))
_PROTOTYPE(int main, (int argc, char **argv));
_PROTOTYPE(void test, (int m, int a));
_PROTOTYPE(pid_t traced_fork, (_PROTOTYPE(void (*c), (void))));
_PROTOTYPE(pid_t traced_pfork, (_PROTOTYPE(void (*c), (void))));
_PROTOTYPE(void WRITE, (int value));
_PROTOTYPE(int READ, (void));
_PROTOTYPE(void traced_wait, (void));
_PROTOTYPE(void detach_running, (pid_t pid));
_PROTOTYPE(void dummy_handler, (int sig));
_PROTOTYPE(void exit_handler, (int sig));
_PROTOTYPE(void count_handler, (int sig));
_PROTOTYPE(void catch_handler, (int sig));
_PROTOTYPE(void test_wait_child, (void));
_PROTOTYPE(void test_wait, (void));
_PROTOTYPE(void test_exec_child, (void));
_PROTOTYPE(void test_exec, (void));
_PROTOTYPE(void test_step_child, (void));
_PROTOTYPE(void test_step, (void));
_PROTOTYPE(void test_sig_child, (void));
_PROTOTYPE(void test_sig, (void));
_PROTOTYPE(void test_exit_child, (void));
_PROTOTYPE(void test_exit, (void));
_PROTOTYPE(void test_term_child, (void));
_PROTOTYPE(void test_term, (void));
_PROTOTYPE(void test_catch_child, (void));
_PROTOTYPE(void test_catch, (void));
_PROTOTYPE(void test_kill_child, (void));
_PROTOTYPE(void test_kill, (void));
_PROTOTYPE(void test_attach_child, (void));
_PROTOTYPE(void test_attach, (void));
_PROTOTYPE(void test_detach_child, (void));
_PROTOTYPE(void test_detach, (void));
_PROTOTYPE(void test_death_child, (void));
_PROTOTYPE(void test_death, (void));
_PROTOTYPE(void test_zdeath_child, (void));
_PROTOTYPE(void test_zdeath, (void));
_PROTOTYPE(void test_syscall_child, (void));
_PROTOTYPE(void test_syscall, (void));
_PROTOTYPE(void test_tracefork_child, (void));
_PROTOTYPE(void test_tracefork, (void));
_PROTOTYPE(void test_reattach_child, (void));
_PROTOTYPE(void test_reattach, (void));
Merge of David's ptrace branch. Summary: o Support for ptrace T_ATTACH/T_DETACH and T_SYSCALL o PM signal handling logic should now work properly, even with debuggers being present o Asynchronous PM/VFS protocol, full IPC support for senda(), and AMF_NOREPLY senda() flag DETAILS Process stop and delay call handling of PM: o Added sys_runctl() kernel call with sys_stop() and sys_resume() aliases, for PM to stop and resume a process o Added exception for sending/syscall-traced processes to sys_runctl(), and matching SIGKREADY pseudo-signal to PM o Fixed PM signal logic to deal with requests from a process after stopping it (so-called "delay calls"), using the SIGKREADY facility o Fixed various PM panics due to race conditions with delay calls versus VFS calls o Removed special PRIO_STOP priority value o Added SYS_LOCK RTS kernel flag, to stop an individual process from running while modifying its process structure Signal and debugger handling in PM: o Fixed debugger signals being dropped if a second signal arrives when the debugger has not retrieved the first one o Fixed debugger signals being sent to the debugger more than once o Fixed debugger signals unpausing process in VFS; removed PM_UNPAUSE_TR protocol message o Detached debugger signals from general signal logic and from being blocked on VFS calls, meaning that even VFS can now be traced o Fixed debugger being unable to receive more than one pending signal in one process stop o Fixed signal delivery being delayed needlessly when multiple signals are pending o Fixed wait test for tracer, which was returning for children that were not waited for o Removed second parallel pending call from PM to VFS for any process o Fixed process becoming runnable between exec() and debugger trap o Added support for notifying the debugger before the parent when a debugged child exits o Fixed debugger death causing child to remain stopped forever o Fixed consistently incorrect use of _NSIG Extensions to ptrace(): o Added T_ATTACH and T_DETACH ptrace request, to attach and detach a debugger to and from a process o Added T_SYSCALL ptrace request, to trace system calls o Added T_SETOPT ptrace request, to set trace options o Added TO_TRACEFORK trace option, to attach automatically to children of a traced process o Added TO_ALTEXEC trace option, to send SIGSTOP instead of SIGTRAP upon a successful exec() of the tracee o Extended T_GETUSER ptrace support to allow retrieving a process's priv structure o Removed T_STOP ptrace request again, as it does not help implementing debuggers properly o Added MINIX3-specific ptrace test (test42) o Added proper manual page for ptrace(2) Asynchronous PM/VFS interface: o Fixed asynchronous messages not being checked when receive() is called with an endpoint other than ANY o Added AMF_NOREPLY senda() flag, preventing such messages from satisfying the receive part of a sendrec() o Added asynsend3() that takes optional flags; asynsend() is now a #define passing in 0 as third parameter o Made PM/VFS protocol asynchronous; reintroduced tell_fs() o Made PM_BASE request/reply number range unique o Hacked in a horrible temporary workaround into RS to deal with newly revealed RS-PM-VFS race condition triangle until VFS is asynchronous System signal handling: o Fixed shutdown logic of device drivers; removed old SIGKSTOP signal o Removed is-superuser check from PM's do_procstat() (aka getsigset()) o Added sigset macros to allow system processes to deal with the full signal set, rather than just the POSIX subset Miscellaneous PM fixes: o Split do_getset into do_get and do_set, merging common code and making structure clearer o Fixed setpriority() being able to put to sleep processes using an invalid parameter, or revive zombie processes o Made find_proc() global; removed obsolete proc_from_pid() o Cleanup here and there Also included: o Fixed false-positive boot order kernel warning o Removed last traces of old NOTIFY_FROM code THINGS OF POSSIBLE INTEREST o It should now be possible to run PM at any priority, even lower than user processes o No assumptions are made about communication speed between PM and VFS, although communication must be FIFO o A debugger will now receive incoming debuggee signals at kill time only; the process may not yet be fully stopped o A first step has been made towards making the SYSTEM task preemptible
2009-09-30 11:57:22 +02:00
_PROTOTYPE(void altexec, (int setflag, int *traps, int *stop));
_PROTOTYPE(void test_altexec, (void));
_PROTOTYPE(void test_noaltexec, (void));
_PROTOTYPE(void e, (int n));
_PROTOTYPE(void quit, (void));
static char *executable;
static int errct = 0, subtest;
static int child = 0, attach;
static pid_t ppid;
static int pfd[4];
static int sigs, caught;
int main(argc, argv)
int argc;
char **argv;
{
2009-11-09 10:26:09 +01:00
int i, m = 0xFFFFFF, n = 0xF;
Merge of David's ptrace branch. Summary: o Support for ptrace T_ATTACH/T_DETACH and T_SYSCALL o PM signal handling logic should now work properly, even with debuggers being present o Asynchronous PM/VFS protocol, full IPC support for senda(), and AMF_NOREPLY senda() flag DETAILS Process stop and delay call handling of PM: o Added sys_runctl() kernel call with sys_stop() and sys_resume() aliases, for PM to stop and resume a process o Added exception for sending/syscall-traced processes to sys_runctl(), and matching SIGKREADY pseudo-signal to PM o Fixed PM signal logic to deal with requests from a process after stopping it (so-called "delay calls"), using the SIGKREADY facility o Fixed various PM panics due to race conditions with delay calls versus VFS calls o Removed special PRIO_STOP priority value o Added SYS_LOCK RTS kernel flag, to stop an individual process from running while modifying its process structure Signal and debugger handling in PM: o Fixed debugger signals being dropped if a second signal arrives when the debugger has not retrieved the first one o Fixed debugger signals being sent to the debugger more than once o Fixed debugger signals unpausing process in VFS; removed PM_UNPAUSE_TR protocol message o Detached debugger signals from general signal logic and from being blocked on VFS calls, meaning that even VFS can now be traced o Fixed debugger being unable to receive more than one pending signal in one process stop o Fixed signal delivery being delayed needlessly when multiple signals are pending o Fixed wait test for tracer, which was returning for children that were not waited for o Removed second parallel pending call from PM to VFS for any process o Fixed process becoming runnable between exec() and debugger trap o Added support for notifying the debugger before the parent when a debugged child exits o Fixed debugger death causing child to remain stopped forever o Fixed consistently incorrect use of _NSIG Extensions to ptrace(): o Added T_ATTACH and T_DETACH ptrace request, to attach and detach a debugger to and from a process o Added T_SYSCALL ptrace request, to trace system calls o Added T_SETOPT ptrace request, to set trace options o Added TO_TRACEFORK trace option, to attach automatically to children of a traced process o Added TO_ALTEXEC trace option, to send SIGSTOP instead of SIGTRAP upon a successful exec() of the tracee o Extended T_GETUSER ptrace support to allow retrieving a process's priv structure o Removed T_STOP ptrace request again, as it does not help implementing debuggers properly o Added MINIX3-specific ptrace test (test42) o Added proper manual page for ptrace(2) Asynchronous PM/VFS interface: o Fixed asynchronous messages not being checked when receive() is called with an endpoint other than ANY o Added AMF_NOREPLY senda() flag, preventing such messages from satisfying the receive part of a sendrec() o Added asynsend3() that takes optional flags; asynsend() is now a #define passing in 0 as third parameter o Made PM/VFS protocol asynchronous; reintroduced tell_fs() o Made PM_BASE request/reply number range unique o Hacked in a horrible temporary workaround into RS to deal with newly revealed RS-PM-VFS race condition triangle until VFS is asynchronous System signal handling: o Fixed shutdown logic of device drivers; removed old SIGKSTOP signal o Removed is-superuser check from PM's do_procstat() (aka getsigset()) o Added sigset macros to allow system processes to deal with the full signal set, rather than just the POSIX subset Miscellaneous PM fixes: o Split do_getset into do_get and do_set, merging common code and making structure clearer o Fixed setpriority() being able to put to sleep processes using an invalid parameter, or revive zombie processes o Made find_proc() global; removed obsolete proc_from_pid() o Cleanup here and there Also included: o Fixed false-positive boot order kernel warning o Removed last traces of old NOTIFY_FROM code THINGS OF POSSIBLE INTEREST o It should now be possible to run PM at any priority, even lower than user processes o No assumptions are made about communication speed between PM and VFS, although communication must be FIFO o A debugger will now receive incoming debuggee signals at kill time only; the process may not yet be fully stopped o A first step has been made towards making the SYSTEM task preemptible
2009-09-30 11:57:22 +02:00
if (strcmp(argv[0], "DO CHECK") == 0) {
exit(42);
}
printf("Test 42 ");
fflush(stdout);
executable = argv[0];
if (argc >= 2) m = atoi(argv[1]);
if (argc >= 3) n = atoi(argv[2]);
for (i = 0; i < ITERATIONS; i++) {
if (n & 001) test(m, 0);
if (n & 002) test(m, 1);
if (n & 004) test(m, 2);
if (n & 010) test(m, 3);
}
quit();
return(-1); /* impossible */
}
void test(m, a)
int m;
int a;
{
attach = a;
if (m & 0000001) test_wait();
if (m & 0000002) test_exec();
if (m & 0000004) test_step();
if (m & 0000010) test_sig();
if (m & 0000020) test_exit();
if (m & 0000040) test_term();
if (m & 0000100) test_catch();
if (m & 0000200) test_kill();
if (m & 0000400) test_attach();
if (m & 0001000) test_detach();
if (m & 0002000) test_death();
if (m & 0004000) test_zdeath();
if (m & 0010000) test_syscall();
if (m & 0020000) test_tracefork();
if (m & 0040000) test_altexec();
if (m & 0100000) test_noaltexec();
if (m & 0200000) test_reattach();
Merge of David's ptrace branch. Summary: o Support for ptrace T_ATTACH/T_DETACH and T_SYSCALL o PM signal handling logic should now work properly, even with debuggers being present o Asynchronous PM/VFS protocol, full IPC support for senda(), and AMF_NOREPLY senda() flag DETAILS Process stop and delay call handling of PM: o Added sys_runctl() kernel call with sys_stop() and sys_resume() aliases, for PM to stop and resume a process o Added exception for sending/syscall-traced processes to sys_runctl(), and matching SIGKREADY pseudo-signal to PM o Fixed PM signal logic to deal with requests from a process after stopping it (so-called "delay calls"), using the SIGKREADY facility o Fixed various PM panics due to race conditions with delay calls versus VFS calls o Removed special PRIO_STOP priority value o Added SYS_LOCK RTS kernel flag, to stop an individual process from running while modifying its process structure Signal and debugger handling in PM: o Fixed debugger signals being dropped if a second signal arrives when the debugger has not retrieved the first one o Fixed debugger signals being sent to the debugger more than once o Fixed debugger signals unpausing process in VFS; removed PM_UNPAUSE_TR protocol message o Detached debugger signals from general signal logic and from being blocked on VFS calls, meaning that even VFS can now be traced o Fixed debugger being unable to receive more than one pending signal in one process stop o Fixed signal delivery being delayed needlessly when multiple signals are pending o Fixed wait test for tracer, which was returning for children that were not waited for o Removed second parallel pending call from PM to VFS for any process o Fixed process becoming runnable between exec() and debugger trap o Added support for notifying the debugger before the parent when a debugged child exits o Fixed debugger death causing child to remain stopped forever o Fixed consistently incorrect use of _NSIG Extensions to ptrace(): o Added T_ATTACH and T_DETACH ptrace request, to attach and detach a debugger to and from a process o Added T_SYSCALL ptrace request, to trace system calls o Added T_SETOPT ptrace request, to set trace options o Added TO_TRACEFORK trace option, to attach automatically to children of a traced process o Added TO_ALTEXEC trace option, to send SIGSTOP instead of SIGTRAP upon a successful exec() of the tracee o Extended T_GETUSER ptrace support to allow retrieving a process's priv structure o Removed T_STOP ptrace request again, as it does not help implementing debuggers properly o Added MINIX3-specific ptrace test (test42) o Added proper manual page for ptrace(2) Asynchronous PM/VFS interface: o Fixed asynchronous messages not being checked when receive() is called with an endpoint other than ANY o Added AMF_NOREPLY senda() flag, preventing such messages from satisfying the receive part of a sendrec() o Added asynsend3() that takes optional flags; asynsend() is now a #define passing in 0 as third parameter o Made PM/VFS protocol asynchronous; reintroduced tell_fs() o Made PM_BASE request/reply number range unique o Hacked in a horrible temporary workaround into RS to deal with newly revealed RS-PM-VFS race condition triangle until VFS is asynchronous System signal handling: o Fixed shutdown logic of device drivers; removed old SIGKSTOP signal o Removed is-superuser check from PM's do_procstat() (aka getsigset()) o Added sigset macros to allow system processes to deal with the full signal set, rather than just the POSIX subset Miscellaneous PM fixes: o Split do_getset into do_get and do_set, merging common code and making structure clearer o Fixed setpriority() being able to put to sleep processes using an invalid parameter, or revive zombie processes o Made find_proc() global; removed obsolete proc_from_pid() o Cleanup here and there Also included: o Fixed false-positive boot order kernel warning o Removed last traces of old NOTIFY_FROM code THINGS OF POSSIBLE INTEREST o It should now be possible to run PM at any priority, even lower than user processes o No assumptions are made about communication speed between PM and VFS, although communication must be FIFO o A debugger will now receive incoming debuggee signals at kill time only; the process may not yet be fully stopped o A first step has been made towards making the SYSTEM task preemptible
2009-09-30 11:57:22 +02:00
}
pid_t traced_fork(c)
_PROTOTYPE(void (*c), (void));
{
pid_t pid;
int r, status;
if (pipe(pfd) != 0) e(200);
if (pipe(&pfd[2]) != 0) e(201);
switch (attach) {
case 0: /* let child volunteer to be traced */
pid = fork();
if (pid < 0) e(202);
if (pid == 0) {
child = 1;
if (ptrace(T_OK, 0, 0, 0) != 0) e(203);
WRITE(0);
c();
e(204);
}
if (READ() != 0) e(205);
break;
case 1: /* attach to child process */
pid = fork();
if (pid < 0) e(206);
if (pid == 0) {
child = 1;
if (READ() != 0) e(207);
c();
e(208);
}
if (ptrace(T_ATTACH, pid, 0, 0) != 0) e(209);
if (waitpid(pid, &status, 0) != pid) e(210);
if (!_WIFSTOPPED(status)) e(211);
if (WSTOPSIG(status) != SIGSTOP) e(212);
if (ptrace(T_RESUME, pid, 0, 0) != 0) e(213);
WRITE(0);
break;
case 2: /* attach to non-child process */
ppid = fork();
if (ppid < 0) e(214);
if (ppid == 0) {
pid = fork();
if (pid < 0) exit(215);
if (pid == 0) {
child = 1;
if (READ() != 0) e(216);
c();
e(217);
}
child = 1;
WRITE(pid);
if (waitpid(pid, &status, 0) != pid) e(218);
if (_WIFSTOPPED(status)) e(219);
if (_WIFEXITED(status) && (r = WEXITSTATUS(status)) != 42) e(r);
exit(0);
}
pid = READ();
if (ptrace(T_ATTACH, pid, 0, 0) != 0) e(220);
if (waitpid(pid, &status, 0) != pid) e(221);
if (!_WIFSTOPPED(status)) e(222);
if (WSTOPSIG(status) != SIGSTOP) e(223);
if (ptrace(T_RESUME, pid, 0, 0) != 0) e(224);
WRITE(0);
break;
case 3: /* attach by forking from child */
ppid = fork();
if (ppid < 0) e(225);
if (ppid == 0) {
child = 1;
if (ptrace(T_OK, 0, 0, 0) != 0) e(226);
WRITE(0);
if (READ() != 0) e(227);
pid = fork();
if (pid < 0) e(228);
if (pid == 0) {
c();
e(229);
}
WRITE(pid);
if (waitpid(pid, &status, 0) != pid) e(230);
if (_WIFSTOPPED(status)) e(231);
if (_WIFEXITED(status) && (r = WEXITSTATUS(status)) != 42) e(r);
exit(0);
}
if (READ() != 0) e(232);
if (kill(ppid, SIGSTOP) != 0) e(233);
if (waitpid(ppid, &status, 0) != ppid) e(234);
if (!_WIFSTOPPED(status)) e(235);
if (WSTOPSIG(status) != SIGSTOP) e(236);
if (ptrace(T_SETOPT, ppid, 0, TO_TRACEFORK) != 0) e(237);
if (ptrace(T_RESUME, ppid, 0, 0) != 0) e(238);
WRITE(0);
pid = READ();
if (waitpid(pid, &status, 0) != pid) e(239);
if (!_WIFSTOPPED(status)) e(240);
if (WSTOPSIG(status) != SIGSTOP) e(241);
if (ptrace(T_SETOPT, pid, 0, 0) != 0) e(242);
if (ptrace(T_RESUME, pid, 0, 0) != 0) e(243);
detach_running(ppid);
break;
}
return pid;
}
pid_t traced_pfork(c)
_PROTOTYPE(void (*c), (void));
{
pid_t pid;
if (pipe(pfd) != 0) e(300);
if (pipe(&pfd[2]) != 0) e(301);
pid = fork();
if (pid < 0) e(302);
if (pid == 0) {
child = 1;
c();
e(303);
}
return pid;
}
void WRITE(value)
int value;
{
if (write(pfd[child*2+1], &value, sizeof(value)) != sizeof(value)) e(400);
}
int READ()
{
int value;
if (read(pfd[2-child*2], &value, sizeof(value)) != sizeof(value)) e(401);
return value;
}
void traced_wait()
{
int r, status;
if (attach == 2) {
if (waitpid(ppid, &status, 0) != ppid) e(500);
if (!_WIFEXITED(status)) e(501);
if ((r = WEXITSTATUS(status)) != 0) e(r);
}
else {
/* Quick hack to clean up detached children */
waitpid(-1, NULL, WNOHANG);
}
close(pfd[0]);
close(pfd[1]);
close(pfd[2]);
close(pfd[3]);
}
void detach_running(pid)
pid_t pid;
{
/* Detach from a process that is not already stopped. This is the way to do it.
* We have to stop the child in order to detach from it, but as the child may
* have other signals pending for the tracer, we cannot assume we get our own
* signal back immediately. However, because we know that the kill is instant
* and resuming with pending signals will only stop the process immediately
* again, we can use T_RESUME for all the signals until we get our own signal,
* and then detach. A complicating factor is that anywhere during this
* procedure, the child may die (e.g. by getting a SIGKILL). In our tests, this
* will not happen.
*/
int status;
if (kill(pid, SIGSTOP) != 0) e(600);
if (waitpid(pid, &status, 0) != pid) e(601);
while (_WIFSTOPPED(status)) {
if (WSTOPSIG(status) == SIGSTOP) {
if (ptrace(T_DETACH, pid, 0, 0) != 0) e(602);
return;
}
if (ptrace(T_RESUME, pid, 0, WSTOPSIG(status)) != 0) e(603);
if (waitpid(pid, &status, 0) != pid) e(604);
}
/* Apparently the process exited. */
if (!_WIFEXITED(status) && !_WIFSIGNALED(status)) e(605);
/* In our tests, that should not happen. */
e(606);
}
void dummy_handler(sig)
int sig;
{
}
void exit_handler(sig)
int sig;
{
exit(42);
}
void count_handler(sig)
int sig;
{
sigs++;
}
void catch_handler(sig)
int sig;
{
sigset_t set;
int bit;
switch (sig) {
case SIGUSR1: bit = 1; break;
case SIGUSR2: bit = 2; break;
case SIGTERM: bit = 4; break;
default: e(100);
}
sigfillset(&set);
sigprocmask(SIG_SETMASK, &set, NULL);
if (caught & bit) e(101);
caught |= bit;
}
void test_wait_child()
{
exit(42);
}
void test_wait()
{
pid_t pid;
int status;
subtest = 1;
pid = traced_fork(test_wait_child);
if (waitpid(pid, &status, 0) != pid) e(1);
if (!_WIFEXITED(status)) e(2);
if (WEXITSTATUS(status) != 42) e(3);
traced_wait();
}
void test_exec_child()
{
if (READ() != 0) e(100);
execl(executable, "DO CHECK", NULL);
e(101);
}
void test_exec()
{
pid_t pid;
int r, status;
subtest = 2;
pid = traced_fork(test_exec_child);
WRITE(0);
/* An exec() should result in a trap signal. */
if (waitpid(pid, &status, 0) != pid) e(1);
if (!_WIFSTOPPED(status)) e(2);
if (WSTOPSIG(status) != SIGTRAP) e(3);
if (ptrace(T_RESUME, pid, 0, 0) != 0) e(4);
if (waitpid(pid, &status, 0) != pid) e(5);
if (!_WIFEXITED(status)) e(6);
if ((r = WEXITSTATUS(status)) != 42) e(r);
traced_wait();
}
void test_step_child()
{
sigset_t set;
signal(SIGUSR1, SIG_IGN);
WRITE(0);
if (READ() != 0) e(100);
/* It must not be possible for the child to stop the single-step signal. */
signal(SIGTRAP, SIG_IGN);
sigfillset(&set);
sigprocmask(SIG_SETMASK, &set, NULL);
exit(42);
}
void test_step()
{
pid_t pid;
int r, status, count;
subtest = 3;
pid = traced_fork(test_step_child);
if (READ() != 0) e(1);
/* While the child is running, neither waitpid() nor ptrace() should work. */
if (waitpid(pid, &status, WNOHANG) != 0) e(2);
if (ptrace(T_RESUME, pid, 0, 0) != -1) e(3);
if (errno != EBUSY) e(4);
if (kill(pid, SIGUSR1) != 0) e(5);
WRITE(0);
/* A kill() signal (other than SIGKILL) should be delivered to the tracer. */
if (waitpid(pid, &status, 0) != pid) e(6);
if (!_WIFSTOPPED(status)) e(7);
if (WSTOPSIG(status) != SIGUSR1) e(8);
/* ptrace(T_STEP) should result in instruction-wise progress. */
for (count = 0; ; count++) {
if (ptrace(T_STEP, pid, 0, 0) != 0) e(9);
if (waitpid(pid, &status, 0) != pid) e(10);
if (_WIFEXITED(status)) break;
if (!_WIFSTOPPED(status)) e(11);
if (WSTOPSIG(status) != SIGTRAP) e(12);
}
if ((r = WEXITSTATUS(status)) != 42) e(r);
if (count < 10) e(13); /* in practice: hundreds */
traced_wait();
}
void test_sig_child()
{
signal(SIGUSR1, exit_handler);
if (READ() != 0) e(100);
pause();
e(101);
}
void test_sig()
{
pid_t pid;
int r, sig, status;
subtest = 4;
pid = traced_fork(test_sig_child);
WRITE(0);
/* allow the child to enter the pause */
sleep(1);
if (kill(pid, SIGUSR1) != 0) e(1);
if (kill(pid, SIGUSR2) != 0) e(2);
/* All signals should arrive at the tracer, although in "random" order. */
if (waitpid(pid, &status, 0) != pid) e(3);
if (!_WIFSTOPPED(status)) e(4);
if (WSTOPSIG(status) != SIGUSR1 && WSTOPSIG(status) != SIGUSR2) e(5);
/* The tracer should see kills arriving while the tracee is stopped. */
if (kill(pid, WSTOPSIG(status)) != 0) e(6);
if (waitpid(pid, &status, WNOHANG) != pid) e(7);
if (!_WIFSTOPPED(status)) e(8);
if (WSTOPSIG(status) != SIGUSR1 && WSTOPSIG(status) != SIGUSR2) e(9);
sig = (WSTOPSIG(status) == SIGUSR1) ? SIGUSR2 : SIGUSR1;
if (ptrace(T_RESUME, pid, 0, 0) != 0) e(10);
if (waitpid(pid, &status, 0) != pid) e(11);
if (!_WIFSTOPPED(status)) e(12);
if (WSTOPSIG(status) != sig) e(13);
if (waitpid(pid, &status, WNOHANG) != 0) e(14);
if (ptrace(T_RESUME, pid, 0, 0) != 0) e(15);
/* Ignored signals passed via ptrace() should be ignored. */
if (kill(pid, SIGUSR1) != 0) e(16);
if (waitpid(pid, &status, 0) != pid) e(17);
if (!_WIFSTOPPED(status)) e(18);
if (WSTOPSIG(status) != SIGUSR1) e(19);
if (ptrace(T_RESUME, pid, 0, SIGCHLD) != 0) e(20);
/* if the pause has been aborted (shouldn't happen!), let the child exit */
sleep(1);
if (waitpid(pid, &status, WNOHANG) != 0) e(21);
/* Caught signals passed via ptrace() should invoke their signal handlers. */
if (kill(pid, SIGUSR1) != 0) e(22);
if (waitpid(pid, &status, 0) != pid) e(23);
if (!_WIFSTOPPED(status)) e(24);
if (WSTOPSIG(status) != SIGUSR1) e(25);
if (ptrace(T_RESUME, pid, 0, SIGUSR1) != 0) e(26);
if (waitpid(pid, &status, 0) != pid) e(27);
if (!_WIFEXITED(status)) e(28);
if ((r = WEXITSTATUS(status)) != 42) e(29);
traced_wait();
}
void test_exit_child()
{
WRITE(0);
for(;;);
}
void test_exit()
{
pid_t pid;
int r, status;
subtest = 5;
pid = traced_fork(test_exit_child);
if (READ() != 0) e(1);
sleep(1);
if (kill(pid, SIGSTOP) != 0) e(2);
if (waitpid(pid, &status, 0) != pid) e(3);
if (!_WIFSTOPPED(status)) e(4);
if (WSTOPSIG(status) != SIGSTOP) e(5);
/* There should be no more signals pending for the tracer now. */
if (waitpid(pid, &status, WNOHANG) != 0) e(6);
/* ptrace(T_EXIT) should terminate the process with the given exit value. */
if (ptrace(T_EXIT, pid, 0, 42) != 0) e(7);
if (waitpid(pid, &status, 0) != pid) e(8);
if (!_WIFEXITED(status)) e(9);
if ((r = WEXITSTATUS(status)) != 42) e(r);
traced_wait();
}
void test_term_child()
{
signal(SIGUSR2, dummy_handler);
WRITE(0);
pause();
e(100);
}
void test_term()
{
pid_t pid;
int status;
subtest = 6;
pid = traced_fork(test_term_child);
if (READ() != 0) e(1);
/* If the first of two signals terminates the traced child, the second signal
* may or may not be delivered to the tracer - this is merely a policy issue.
* However, nothing unexpected should happen.
*/
if (kill(pid, SIGUSR1) != 0) e(2);
if (kill(pid, SIGUSR2) != 0) e(3);
if (waitpid(pid, &status, 0) != pid) e(4);
if (!_WIFSTOPPED(status)) e(5);
if (ptrace(T_RESUME, pid, 0, SIGUSR1) != 0) e(6);
if (waitpid(pid, &status, 0) != pid) e(7);
if (_WIFSTOPPED(status)) {
if (ptrace(T_RESUME, pid, 0, SIGUSR1) != 0) e(8);
if (waitpid(pid, &status, 0) != pid) e(9);
}
if (!_WIFSIGNALED(status)) e(10);
if (WTERMSIG(status) != SIGUSR1) e(11);
traced_wait();
}
void test_catch_child()
{
struct sigaction sa;
sigset_t set, oset;
sa.sa_handler = catch_handler;
sigemptyset(&sa.sa_mask);
sa.sa_flags = SA_NODEFER;
sigaction(SIGUSR1, &sa, NULL);
sigaction(SIGUSR2, &sa, NULL);
sigaction(SIGTERM, &sa, NULL);
sigfillset(&set);
sigprocmask(SIG_SETMASK, &set, &oset);
caught = 0;
WRITE(0);
while (caught != 7) sigsuspend(&oset);
exit(42);
}
void test_catch()
{
pid_t pid;
int r, sig, status;
subtest = 7;
pid = traced_fork(test_catch_child);
if (READ() != 0) e(1);
if (kill(pid, SIGUSR1) != 0) e(2);
if (kill(pid, SIGUSR2) != 0) e(3);
if (waitpid(pid, &status, 0) != pid) e(4);
if (!_WIFSTOPPED(status)) e(5);
if (WSTOPSIG(status) != SIGUSR1 && WSTOPSIG(status) != SIGUSR2) e(6);
sig = (WSTOPSIG(status) == SIGUSR1) ? SIGUSR2 : SIGUSR1;
if (ptrace(T_RESUME, pid, 0, WSTOPSIG(status)) != 0) e(7);
if (kill(pid, SIGTERM) != 0) e(8);
if (waitpid(pid, &status, 0) != pid) e(9);
if (!_WIFSTOPPED(status)) e(10);
if (WSTOPSIG(status) != sig && WSTOPSIG(status) != SIGTERM) e(11);
if (WSTOPSIG(status) == sig) sig = SIGTERM;
if (ptrace(T_RESUME, pid, 0, WSTOPSIG(status)) != 0) e(12);
if (kill(pid, SIGBUS) != 0) e(13);
if (waitpid(pid, &status, 0) != pid) e(14);
if (!_WIFSTOPPED(status)) e(15);
if (WSTOPSIG(status) != sig && WSTOPSIG(status) != SIGBUS) e(16);
if (ptrace(T_RESUME, pid, 0, sig) != 0) e(17);
if (WSTOPSIG(status) == sig) sig = SIGBUS;
if (waitpid(pid, &status, 0) != pid) e(18);
if (!_WIFSTOPPED(status)) e(19);
if (WSTOPSIG(status) != sig) e(20);
if (ptrace(T_RESUME, pid, 0, 0) != 0) e(21);
if (waitpid(pid, &status, 0) != pid) e(22);
if (!_WIFEXITED(status)) e(23);
if ((r = WEXITSTATUS(status)) != 42) e(r);
traced_wait();
}
void test_kill_child()
{
sigset_t set;
signal(SIGKILL, SIG_IGN);
sigfillset(&set);
sigprocmask(SIG_SETMASK, &set, NULL);
WRITE(0);
pause();
e(100);
}
void test_kill()
{
pid_t pid;
int status;
subtest = 8;
pid = traced_fork(test_kill_child);
if (READ() != 0) e(1);
/* SIGKILL must be unstoppable in every way. */
if (kill(pid, SIGKILL) != 0) e(2);
if (waitpid(pid, &status, 0) != pid) e(3);
if (!_WIFSIGNALED(status)) e(4);
if (WTERMSIG(status) != SIGKILL) e(5);
/* After termination, the child must no longer be visible to the tracer. */
if (waitpid(pid, &status, WNOHANG) != -1) e(6);
if (errno != ECHILD) e(7);
traced_wait();
}
void test_attach_child()
{
if (ptrace(T_OK, 0, 0, 0) != -1) e(100);
if (errno != EBUSY) e(101);
WRITE(0);
if (READ() != 0) e(102);
exit(42);
}
void test_attach()
{
pid_t pid;
int r, status;
subtest = 9;
/* Attaching to kernel processes is not allowed. */
if (ptrace(T_ATTACH, -1, 0, 0) != -1) e(1);
if (errno != ESRCH) e(2);
/* Attaching to self is not allowed. */
if (ptrace(T_ATTACH, getpid(), 0, 0) != -1) e(3);
if (errno != EPERM) e(4);
/* Attaching to PM is not allowed. */
2009-12-13 22:45:23 +01:00
#if 0
/* FIXME: disabled until we can reliably determine PM's pid */
Merge of David's ptrace branch. Summary: o Support for ptrace T_ATTACH/T_DETACH and T_SYSCALL o PM signal handling logic should now work properly, even with debuggers being present o Asynchronous PM/VFS protocol, full IPC support for senda(), and AMF_NOREPLY senda() flag DETAILS Process stop and delay call handling of PM: o Added sys_runctl() kernel call with sys_stop() and sys_resume() aliases, for PM to stop and resume a process o Added exception for sending/syscall-traced processes to sys_runctl(), and matching SIGKREADY pseudo-signal to PM o Fixed PM signal logic to deal with requests from a process after stopping it (so-called "delay calls"), using the SIGKREADY facility o Fixed various PM panics due to race conditions with delay calls versus VFS calls o Removed special PRIO_STOP priority value o Added SYS_LOCK RTS kernel flag, to stop an individual process from running while modifying its process structure Signal and debugger handling in PM: o Fixed debugger signals being dropped if a second signal arrives when the debugger has not retrieved the first one o Fixed debugger signals being sent to the debugger more than once o Fixed debugger signals unpausing process in VFS; removed PM_UNPAUSE_TR protocol message o Detached debugger signals from general signal logic and from being blocked on VFS calls, meaning that even VFS can now be traced o Fixed debugger being unable to receive more than one pending signal in one process stop o Fixed signal delivery being delayed needlessly when multiple signals are pending o Fixed wait test for tracer, which was returning for children that were not waited for o Removed second parallel pending call from PM to VFS for any process o Fixed process becoming runnable between exec() and debugger trap o Added support for notifying the debugger before the parent when a debugged child exits o Fixed debugger death causing child to remain stopped forever o Fixed consistently incorrect use of _NSIG Extensions to ptrace(): o Added T_ATTACH and T_DETACH ptrace request, to attach and detach a debugger to and from a process o Added T_SYSCALL ptrace request, to trace system calls o Added T_SETOPT ptrace request, to set trace options o Added TO_TRACEFORK trace option, to attach automatically to children of a traced process o Added TO_ALTEXEC trace option, to send SIGSTOP instead of SIGTRAP upon a successful exec() of the tracee o Extended T_GETUSER ptrace support to allow retrieving a process's priv structure o Removed T_STOP ptrace request again, as it does not help implementing debuggers properly o Added MINIX3-specific ptrace test (test42) o Added proper manual page for ptrace(2) Asynchronous PM/VFS interface: o Fixed asynchronous messages not being checked when receive() is called with an endpoint other than ANY o Added AMF_NOREPLY senda() flag, preventing such messages from satisfying the receive part of a sendrec() o Added asynsend3() that takes optional flags; asynsend() is now a #define passing in 0 as third parameter o Made PM/VFS protocol asynchronous; reintroduced tell_fs() o Made PM_BASE request/reply number range unique o Hacked in a horrible temporary workaround into RS to deal with newly revealed RS-PM-VFS race condition triangle until VFS is asynchronous System signal handling: o Fixed shutdown logic of device drivers; removed old SIGKSTOP signal o Removed is-superuser check from PM's do_procstat() (aka getsigset()) o Added sigset macros to allow system processes to deal with the full signal set, rather than just the POSIX subset Miscellaneous PM fixes: o Split do_getset into do_get and do_set, merging common code and making structure clearer o Fixed setpriority() being able to put to sleep processes using an invalid parameter, or revive zombie processes o Made find_proc() global; removed obsolete proc_from_pid() o Cleanup here and there Also included: o Fixed false-positive boot order kernel warning o Removed last traces of old NOTIFY_FROM code THINGS OF POSSIBLE INTEREST o It should now be possible to run PM at any priority, even lower than user processes o No assumptions are made about communication speed between PM and VFS, although communication must be FIFO o A debugger will now receive incoming debuggee signals at kill time only; the process may not yet be fully stopped o A first step has been made towards making the SYSTEM task preemptible
2009-09-30 11:57:22 +02:00
if (ptrace(T_ATTACH, 0, 0, 0) != -1) e(5);
if (errno != EPERM) e(6);
2009-12-13 22:45:23 +01:00
#endif
Merge of David's ptrace branch. Summary: o Support for ptrace T_ATTACH/T_DETACH and T_SYSCALL o PM signal handling logic should now work properly, even with debuggers being present o Asynchronous PM/VFS protocol, full IPC support for senda(), and AMF_NOREPLY senda() flag DETAILS Process stop and delay call handling of PM: o Added sys_runctl() kernel call with sys_stop() and sys_resume() aliases, for PM to stop and resume a process o Added exception for sending/syscall-traced processes to sys_runctl(), and matching SIGKREADY pseudo-signal to PM o Fixed PM signal logic to deal with requests from a process after stopping it (so-called "delay calls"), using the SIGKREADY facility o Fixed various PM panics due to race conditions with delay calls versus VFS calls o Removed special PRIO_STOP priority value o Added SYS_LOCK RTS kernel flag, to stop an individual process from running while modifying its process structure Signal and debugger handling in PM: o Fixed debugger signals being dropped if a second signal arrives when the debugger has not retrieved the first one o Fixed debugger signals being sent to the debugger more than once o Fixed debugger signals unpausing process in VFS; removed PM_UNPAUSE_TR protocol message o Detached debugger signals from general signal logic and from being blocked on VFS calls, meaning that even VFS can now be traced o Fixed debugger being unable to receive more than one pending signal in one process stop o Fixed signal delivery being delayed needlessly when multiple signals are pending o Fixed wait test for tracer, which was returning for children that were not waited for o Removed second parallel pending call from PM to VFS for any process o Fixed process becoming runnable between exec() and debugger trap o Added support for notifying the debugger before the parent when a debugged child exits o Fixed debugger death causing child to remain stopped forever o Fixed consistently incorrect use of _NSIG Extensions to ptrace(): o Added T_ATTACH and T_DETACH ptrace request, to attach and detach a debugger to and from a process o Added T_SYSCALL ptrace request, to trace system calls o Added T_SETOPT ptrace request, to set trace options o Added TO_TRACEFORK trace option, to attach automatically to children of a traced process o Added TO_ALTEXEC trace option, to send SIGSTOP instead of SIGTRAP upon a successful exec() of the tracee o Extended T_GETUSER ptrace support to allow retrieving a process's priv structure o Removed T_STOP ptrace request again, as it does not help implementing debuggers properly o Added MINIX3-specific ptrace test (test42) o Added proper manual page for ptrace(2) Asynchronous PM/VFS interface: o Fixed asynchronous messages not being checked when receive() is called with an endpoint other than ANY o Added AMF_NOREPLY senda() flag, preventing such messages from satisfying the receive part of a sendrec() o Added asynsend3() that takes optional flags; asynsend() is now a #define passing in 0 as third parameter o Made PM/VFS protocol asynchronous; reintroduced tell_fs() o Made PM_BASE request/reply number range unique o Hacked in a horrible temporary workaround into RS to deal with newly revealed RS-PM-VFS race condition triangle until VFS is asynchronous System signal handling: o Fixed shutdown logic of device drivers; removed old SIGKSTOP signal o Removed is-superuser check from PM's do_procstat() (aka getsigset()) o Added sigset macros to allow system processes to deal with the full signal set, rather than just the POSIX subset Miscellaneous PM fixes: o Split do_getset into do_get and do_set, merging common code and making structure clearer o Fixed setpriority() being able to put to sleep processes using an invalid parameter, or revive zombie processes o Made find_proc() global; removed obsolete proc_from_pid() o Cleanup here and there Also included: o Fixed false-positive boot order kernel warning o Removed last traces of old NOTIFY_FROM code THINGS OF POSSIBLE INTEREST o It should now be possible to run PM at any priority, even lower than user processes o No assumptions are made about communication speed between PM and VFS, although communication must be FIFO o A debugger will now receive incoming debuggee signals at kill time only; the process may not yet be fully stopped o A first step has been made towards making the SYSTEM task preemptible
2009-09-30 11:57:22 +02:00
pid = traced_fork(test_attach_child);
/* Attaching more than once is not allowed. */
if (ptrace(T_ATTACH, pid, 0, 0) != -1) e(7);
if (errno != EBUSY) e(8);
if (READ() != 0) e(9);
/* Detaching a running child should not succeed. */
if (ptrace(T_DETACH, pid, 0, 0) == 0) e(10);
if (errno != EBUSY) e(11);
detach_running(pid);
WRITE(0);
traced_wait();
}
void test_detach_child()
{
struct sigaction sa;
sigset_t set, sset, oset;
sa.sa_handler = catch_handler;
sigemptyset(&sa.sa_mask);
sa.sa_flags = SA_NODEFER;
sigaction(SIGUSR1, &sa, NULL);
sigaction(SIGUSR2, &sa, NULL);
sigaction(SIGTERM, &sa, NULL);
sigfillset(&set);
sigprocmask(SIG_SETMASK, &set, &oset);
sigfillset(&sset);
sigdelset(&sset, SIGUSR1);
caught = 0;
WRITE(0);
if (sigsuspend(&sset) != -1) e(102);
if (errno != EINTR) e(103);
if (caught != 1) e(104);
if (READ() != 0) e(105);
while (caught != 7) sigsuspend(&oset);
exit(42);
}
void test_detach()
{
pid_t pid;
int r, status;
/* Can't use traced_fork(), so simplify a bit */
if (attach != 0) return;
subtest = 10;
pid = traced_pfork(test_detach_child);
if (READ() != 0) e(1);
/* The tracer should not see signals sent to the process before attaching. */
if (kill(pid, SIGUSR2) != 0) e(2);
if (ptrace(T_ATTACH, pid, 0, 0) != 0) e(3);
if (waitpid(pid, &status, 0) != pid) e(4);
if (!_WIFSTOPPED(status)) e(5);
if (WSTOPSIG(status) != SIGSTOP) e(6);
if (ptrace(T_RESUME, pid, 0, 0) != 0) e(7);
if (kill(pid, SIGUSR1) != 0) e(8);
if (waitpid(pid, &status, 0) != pid) e(9);
if (!_WIFSTOPPED(status)) e(10);
if (WSTOPSIG(status) != SIGUSR1) e(11);
/* Signals pending at the tracer should be passed on after detaching. */
if (kill(pid, SIGTERM) != 0) e(12);
/* A signal may be passed with the detach request. */
if (ptrace(T_DETACH, pid, 0, SIGUSR1) != 0) e(13);
WRITE(0);
if (waitpid(pid, &status, 0) != pid) e(14);
if (!_WIFEXITED(status)) e(15);
if ((r = WEXITSTATUS(status)) != 42) e(r);
traced_wait();
}
void test_death_child()
{
pid_t pid;
pid = fork();
if (pid < 0) e(100);
if (pid == 0) {
ptrace(T_OK, 0, 0, 0);
WRITE(getpid());
for (;;) pause();
}
if (READ() != 0) e(101);
kill(getpid(), SIGKILL);
e(102);
}
void test_death()
{
pid_t pid, cpid;
int status;
subtest = 11;
pid = traced_fork(test_death_child);
cpid = READ();
if (kill(cpid, 0) != 0) e(1);
WRITE(0);
if (waitpid(pid, &status, 0) != pid) e(2);
if (!_WIFSIGNALED(status)) e(3);
if (WTERMSIG(status) != SIGKILL) e(4);
/* The children of killed tracers should be terminated. */
while (kill(cpid, 0) == 0) sleep(1);
if (errno != ESRCH) e(5);
traced_wait();
}
void test_zdeath_child()
{
if (READ() != 0) e(100);
exit(42);
}
void test_zdeath()
{
pid_t pid, tpid;
int r, status;
/* Can't use traced_fork(), so simplify a bit */
if (attach != 0) return;
subtest = 12;
pid = traced_pfork(test_zdeath_child);
tpid = fork();
if (tpid < 0) e(1);
if (tpid == 0) {
if (ptrace(T_ATTACH, pid, 0, 0) != 0) exit(101);
if (waitpid(pid, &status, 0) != pid) exit(102);
if (!_WIFSTOPPED(status)) exit(103);
if (WSTOPSIG(status) != SIGSTOP) exit(104);
if (ptrace(T_RESUME, pid, 0, 0) != 0) exit(105);
WRITE(0);
/* Unwaited-for traced zombies should be passed to their parent. */
sleep(2);
exit(84);
}
sleep(1);
/* However, that should only happen once the tracer has actually died. */
if (waitpid(pid, &status, WNOHANG) != 0) e(2);
if (waitpid(tpid, &status, 0) != tpid) e(3);
if (!_WIFEXITED(status)) e(4);
if ((r = WEXITSTATUS(status)) != 84) e(r);
if (waitpid(pid, &status, 0) != pid) e(5);
if (!_WIFEXITED(status)) e(6);
if ((r = WEXITSTATUS(status)) != 42) e(r);
traced_wait();
}
void test_syscall_child()
{
signal(SIGUSR1, count_handler);
signal(SIGUSR2, count_handler);
sigs = 0;
WRITE(0);
if (READ() != 0) e(100);
/* Three calls (may fail) */
setuid(0);
close(123);
getpid();
if (sigs != 2) e(101);
exit(42);
}
void test_syscall()
{
pid_t pid;
int i, r, sig, status;
subtest = 13;
pid = traced_fork(test_syscall_child);
if (READ() != 0) e(1);
if (kill(pid, SIGSTOP) != 0) e(2);
if (waitpid(pid, &status, 0) != pid) e(3);
if (!_WIFSTOPPED(status)) e(4);
if (WSTOPSIG(status) != SIGSTOP) e(5);
WRITE(0);
/* Upon resuming a first system call, no syscall leave event must be sent. */
if (ptrace(T_SYSCALL, pid, 0, 0) != 0) e(6);
if (waitpid(pid, &status, 0) != pid) e(7);
for (i = 0; _WIFSTOPPED(status); i++) {
if (WSTOPSIG(status) != SIGTRAP) e(8);
/* Signals passed via T_SYSCALL should arrive, on enter and exit. */
if (i == 3) sig = SIGUSR1;
else if (i == 6) sig = SIGUSR2;
else sig = 0;
if (ptrace(T_SYSCALL, pid, 0, sig) != 0) e(9);
if (waitpid(pid, &status, 0) != pid) e(10);
}
if (!_WIFEXITED(status)) e(11);
if ((r = WEXITSTATUS(status)) != 42) e(r);
/* The number of events seen is deterministic but libc-dependent. */
if (i < 10 || i > 100) e(12);
/* The last system call event must be for entering exit(). */
if (!(i % 2)) e(13);
traced_wait();
}
void test_tracefork_child()
{
pid_t pid;
signal(SIGHUP, SIG_IGN);
pid = setsid();
WRITE(pid);
if (READ() != 0) e(100);
if ((pid = fork()) < 0) e(101);
exit(pid > 0 ? 42 : 84);
}
void test_tracefork()
{
pid_t pgrp, ppid, cpid, wpid;
int r, status, gotstop, ptraps, ctraps;
subtest = 14;
ppid = traced_fork(test_tracefork_child);
if ((pgrp = READ()) <= 0) e(1);
if (kill(ppid, SIGSTOP) != 0) e(2);
if (waitpid(ppid, &status, 0) != ppid) e(3);
if (!_WIFSTOPPED(status)) e(4);
if (WSTOPSIG(status) != SIGSTOP) e(5);
if (ptrace(T_SETOPT, ppid, 0, TO_TRACEFORK) != 0) e(6);
WRITE(0);
if (ptrace(T_SYSCALL, ppid, 0, 0) != 0) e(7);
cpid = -1;
gotstop = -1;
/* Count how many traps we get for parent and child, until they both exit. */
for (ptraps = ctraps = 0; ppid || cpid; ) {
wpid = waitpid(-pgrp, &status, 0);
if (wpid <= 0) e(8);
if (cpid < 0 && wpid != ppid) {
cpid = wpid;
gotstop = 0;
}
if (wpid != ppid && wpid != cpid) e(9);
if (_WIFEXITED(status)) {
if (wpid == ppid) {
if ((r = WEXITSTATUS(status)) != 42) e(r);
ppid = 0;
}
else {
if ((r = WEXITSTATUS(status)) != 84) e(r);
cpid = 0;
}
}
else {
if (!_WIFSTOPPED(status)) e(10);
switch (WSTOPSIG(status)) {
case SIGCHLD:
case SIGHUP:
break;
case SIGSTOP:
if (wpid != cpid) e(11);
if (gotstop) e(12);
gotstop = 1;
break;
case SIGTRAP:
if (wpid == ppid) ptraps++;
else ctraps++;
break;
default:
e(13);
}
if (ptrace(T_SYSCALL, wpid, 0, 0) != 0) e(14);
}
}
/* The parent should get an odd number of traps: the first one is a syscall
* enter trap (typically for the fork()), the last one is the syscall enter
* trap for its exit().
*/
if (ptraps < 3) e(15);
if (!(ptraps % 2)) e(16);
/* The child should get an even number of traps: the first one is a syscall
* leave trap from the fork(), the last one is the syscall enter trap for
* its exit().
*/
if (ctraps < 2) e(17);
if (ctraps % 2) e(18);
traced_wait();
}
void altexec(setflag, traps, stop)
int setflag;
int *traps;
int *stop;
{
pid_t pid;
int r, status;
pid = traced_fork(test_exec_child);
if (kill(pid, SIGSTOP) != 0) e(1);
if (waitpid(pid, &status, 0) != pid) e(2);
if (!_WIFSTOPPED(status)) e(3);
if (WSTOPSIG(status) != SIGSTOP) e(4);
if (setflag && ptrace(T_SETOPT, pid, 0, TO_ALTEXEC) != 0) e(5);
WRITE(0);
if (ptrace(T_SYSCALL, pid, 0, 0) != 0) e(6);
*traps = 0;
*stop = -1;
for (;;) {
if (waitpid(pid, &status, 0) != pid) e(7);
if (_WIFEXITED(status)) break;
if (!_WIFSTOPPED(status)) e(8);
switch (WSTOPSIG(status)) {
case SIGTRAP:
(*traps)++;
break;
case SIGSTOP:
if (*stop >= 0) e(9);
*stop = *traps;
break;
default:
e(10);
}
if (ptrace(T_SYSCALL, pid, 0, 0) != 0) e(11);
}
if ((r = WEXITSTATUS(status)) != 42) e(r);
traced_wait();
}
void test_altexec()
{
int traps, stop;
subtest = 15;
altexec(1, &traps, &stop);
/* The exec causes a SIGSTOP. This gives us an odd number of traps: a pair
* for each system call, plus one for the final exit(). The stop must have
* taken place after a syscall enter event, i.e. must be odd as well.
*/
if (traps < 3) e(12);
if (!(traps % 2)) e(13);
if (stop < 0) e(14);
if (!(stop % 2)) e(15);
}
void test_noaltexec()
{
int traps, stop;
subtest = 16;
altexec(0, &traps, &stop);
/* The exec does not cause a SIGSTOP. This gives us an even number of traps;
* as above, but plus the exec()'s extra SIGTRAP. This trap is
* indistinguishable from a syscall trap, especially when considering failed
* exec() calls and immediately following signal handler invocations.
*/
if (traps < 4) e(12);
if (traps % 2) e(13);
if (stop >= 0) e(14);
}
void test_reattach_child()
{
struct timeval tv;
if (READ() != 0) e(100);
tv.tv_sec = 2;
tv.tv_usec = 0;
if (select(0, NULL, NULL, NULL, &tv) != 0) e(101);
exit(42);
}
void test_reattach()
{
pid_t pid;
int r, status, count;
subtest = 17;
pid = traced_fork(test_reattach_child);
if (kill(pid, SIGSTOP) != 0) e(1);
if (waitpid(pid, &status, 0) != pid) e(2);
if (!_WIFSTOPPED(status)) e(3);
if (WSTOPSIG(status) != SIGSTOP) e(4);
WRITE(0);
signal(SIGALRM, dummy_handler);
alarm(1);
/* Start tracing system calls. We don't know how many there will be until
* we reach the child's select(), so we have to interrupt ourselves.
*/
if (ptrace(T_SYSCALL, pid, 0, 0) != 0) e(5);
for (count = 0; (r = waitpid(pid, &status, 0)) == pid; count++) {
if (!_WIFSTOPPED(status)) e(6);
if (WSTOPSIG(status) != SIGTRAP) e(7);
if (ptrace(T_SYSCALL, pid, 0, 0) != 0) e(8);
}
if (r != -1 || errno != EINTR) e(9);
/* We always start with syscall enter event; the last event we should have
* seen before the alarm was entering the select() call.
*/
if (!(count % 2)) e(10);
/* Detach, and immediately attach again. */
detach_running(pid);
if (ptrace(T_ATTACH, pid, 0, 0) != 0) e(11);
if (waitpid(pid, &status, 0) != pid) e(12);
if (!_WIFSTOPPED(status)) e(13);
if (WSTOPSIG(status) != SIGSTOP) e(14);
if (ptrace(T_SYSCALL, pid, 0, 0) != 0) e(15);
if (waitpid(pid, &status, 0) != pid) e(16);
for (count = 0; _WIFSTOPPED(status); count++) {
if (WSTOPSIG(status) != SIGTRAP) e(17);
if (ptrace(T_SYSCALL, pid, 0, 0) != 0) e(18);
if (waitpid(pid, &status, 0) != pid) e(19);
}
if (!_WIFEXITED(status)) e(20);
if ((r = WEXITSTATUS(status)) != 42) e(r);
/* We must not have seen the select()'s syscall leave event, and the last
* event will be the syscall enter for the exec().
*/
if (!(count % 2)) e(21);
traced_wait();
}
Merge of David's ptrace branch. Summary: o Support for ptrace T_ATTACH/T_DETACH and T_SYSCALL o PM signal handling logic should now work properly, even with debuggers being present o Asynchronous PM/VFS protocol, full IPC support for senda(), and AMF_NOREPLY senda() flag DETAILS Process stop and delay call handling of PM: o Added sys_runctl() kernel call with sys_stop() and sys_resume() aliases, for PM to stop and resume a process o Added exception for sending/syscall-traced processes to sys_runctl(), and matching SIGKREADY pseudo-signal to PM o Fixed PM signal logic to deal with requests from a process after stopping it (so-called "delay calls"), using the SIGKREADY facility o Fixed various PM panics due to race conditions with delay calls versus VFS calls o Removed special PRIO_STOP priority value o Added SYS_LOCK RTS kernel flag, to stop an individual process from running while modifying its process structure Signal and debugger handling in PM: o Fixed debugger signals being dropped if a second signal arrives when the debugger has not retrieved the first one o Fixed debugger signals being sent to the debugger more than once o Fixed debugger signals unpausing process in VFS; removed PM_UNPAUSE_TR protocol message o Detached debugger signals from general signal logic and from being blocked on VFS calls, meaning that even VFS can now be traced o Fixed debugger being unable to receive more than one pending signal in one process stop o Fixed signal delivery being delayed needlessly when multiple signals are pending o Fixed wait test for tracer, which was returning for children that were not waited for o Removed second parallel pending call from PM to VFS for any process o Fixed process becoming runnable between exec() and debugger trap o Added support for notifying the debugger before the parent when a debugged child exits o Fixed debugger death causing child to remain stopped forever o Fixed consistently incorrect use of _NSIG Extensions to ptrace(): o Added T_ATTACH and T_DETACH ptrace request, to attach and detach a debugger to and from a process o Added T_SYSCALL ptrace request, to trace system calls o Added T_SETOPT ptrace request, to set trace options o Added TO_TRACEFORK trace option, to attach automatically to children of a traced process o Added TO_ALTEXEC trace option, to send SIGSTOP instead of SIGTRAP upon a successful exec() of the tracee o Extended T_GETUSER ptrace support to allow retrieving a process's priv structure o Removed T_STOP ptrace request again, as it does not help implementing debuggers properly o Added MINIX3-specific ptrace test (test42) o Added proper manual page for ptrace(2) Asynchronous PM/VFS interface: o Fixed asynchronous messages not being checked when receive() is called with an endpoint other than ANY o Added AMF_NOREPLY senda() flag, preventing such messages from satisfying the receive part of a sendrec() o Added asynsend3() that takes optional flags; asynsend() is now a #define passing in 0 as third parameter o Made PM/VFS protocol asynchronous; reintroduced tell_fs() o Made PM_BASE request/reply number range unique o Hacked in a horrible temporary workaround into RS to deal with newly revealed RS-PM-VFS race condition triangle until VFS is asynchronous System signal handling: o Fixed shutdown logic of device drivers; removed old SIGKSTOP signal o Removed is-superuser check from PM's do_procstat() (aka getsigset()) o Added sigset macros to allow system processes to deal with the full signal set, rather than just the POSIX subset Miscellaneous PM fixes: o Split do_getset into do_get and do_set, merging common code and making structure clearer o Fixed setpriority() being able to put to sleep processes using an invalid parameter, or revive zombie processes o Made find_proc() global; removed obsolete proc_from_pid() o Cleanup here and there Also included: o Fixed false-positive boot order kernel warning o Removed last traces of old NOTIFY_FROM code THINGS OF POSSIBLE INTEREST o It should now be possible to run PM at any priority, even lower than user processes o No assumptions are made about communication speed between PM and VFS, although communication must be FIFO o A debugger will now receive incoming debuggee signals at kill time only; the process may not yet be fully stopped o A first step has been made towards making the SYSTEM task preemptible
2009-09-30 11:57:22 +02:00
void e(n)
int n;
{
if (child) exit(n);
printf("Subtest %d, attach type %d, error %d, errno %d: %s\n",
subtest, attach, n, errno, strerror(errno));
if (errct++ > MAX_ERROR) {
printf("Too many errors; test aborted\n");
exit(1);
}
}
void quit()
{
if (errct == 0) {
printf("ok\n");
exit(0);
} else {
printf("%d errors\n", errct);
exit(1);
}
}