svn commit: r314852 - in head: sys/kern tests/sys/kern
Eric Badger
badger at FreeBSD.org
Tue Mar 7 13:41:03 UTC 2017
Author: badger
Date: Tue Mar 7 13:41:01 2017
New Revision: 314852
URL: https://svnweb.freebsd.org/changeset/base/314852
Log:
don't stop in issignal() if P_SINGLE_EXIT is set
Suppose a traced process is stopped in ptracestop() due to receipt of a
SIGSTOP signal, and is awaiting orders from the tracing process on how
to handle the signal. Before sending any such orders, the tracing
process exits. This should kill the traced process. But suppose a second
thread handles the SIGKILL and proceeds to exit1(), calling
thread_single(). The first thread will now awaken and will have a chance
to check once more if it should go to sleep due to the SIGSTOP. It must
not sleep after P_SINGLE_EXIT has been set; this would prevent the
SIGKILL from taking effect, leaving a stopped orphan behind after the
tracing process dies.
Also add new tests for this condition.
Reviewed by: kib
MFC after: 2 weeks
Sponsored by: Dell EMC
Differential Revision: https://reviews.freebsd.org/D9890
Modified:
head/sys/kern/kern_sig.c
head/tests/sys/kern/ptrace_test.c
Modified: head/sys/kern/kern_sig.c
==============================================================================
--- head/sys/kern/kern_sig.c Tue Mar 7 13:38:25 2017 (r314851)
+++ head/sys/kern/kern_sig.c Tue Mar 7 13:41:01 2017 (r314852)
@@ -2859,14 +2859,15 @@ issignal(struct thread *td)
break; /* == ignore */
}
/*
- * If there is a pending stop signal to process
- * with default action, stop here,
- * then clear the signal. However,
- * if process is member of an orphaned
- * process group, ignore tty stop signals.
+ * If there is a pending stop signal to process with
+ * default action, stop here, then clear the signal.
+ * Traced or exiting processes should ignore stops.
+ * Additionally, a member of an orphaned process group
+ * should ignore tty stops.
*/
if (prop & SIGPROP_STOP) {
- if (p->p_flag & (P_TRACED|P_WEXIT) ||
+ if (p->p_flag &
+ (P_TRACED | P_WEXIT | P_SINGLE_EXIT) ||
(p->p_pgrp->pg_jobc == 0 &&
prop & SIGPROP_TTYSTOP))
break; /* == ignore */
Modified: head/tests/sys/kern/ptrace_test.c
==============================================================================
--- head/tests/sys/kern/ptrace_test.c Tue Mar 7 13:38:25 2017 (r314851)
+++ head/tests/sys/kern/ptrace_test.c Tue Mar 7 13:41:01 2017 (r314852)
@@ -31,6 +31,7 @@ __FBSDID("$FreeBSD$");
#include <sys/cpuset.h>
#include <sys/event.h>
#include <sys/time.h>
+#include <sys/procctl.h>
#include <sys/ptrace.h>
#include <sys/syscall.h>
#include <sys/sysctl.h>
@@ -2785,6 +2786,139 @@ ATF_TC_BODY(ptrace__PT_CONTINUE_with_sig
ATF_REQUIRE(errno == ECHILD);
}
+static void *
+raise_sigstop_thread(void *arg __unused)
+{
+
+ raise(SIGSTOP);
+ return NULL;
+}
+
+static void *
+sleep_thread(void *arg __unused)
+{
+
+ sleep(60);
+ return NULL;
+}
+
+static void
+terminate_with_pending_sigstop(bool sigstop_from_main_thread)
+{
+ pid_t fpid, wpid;
+ int status, i;
+ cpuset_t setmask;
+ cpusetid_t setid;
+ pthread_t t;
+
+ /*
+ * Become the reaper for this process tree. We need to be able to check
+ * that both child and grandchild have died.
+ */
+ ATF_REQUIRE(procctl(P_PID, getpid(), PROC_REAP_ACQUIRE, NULL) == 0);
+
+ fpid = fork();
+ ATF_REQUIRE(fpid >= 0);
+ if (fpid == 0) {
+ fpid = fork();
+ CHILD_REQUIRE(fpid >= 0);
+ if (fpid == 0) {
+ trace_me();
+
+ /* Pin to CPU 0 to serialize thread execution. */
+ CPU_ZERO(&setmask);
+ CPU_SET(0, &setmask);
+ CHILD_REQUIRE(cpuset(&setid) == 0);
+ CHILD_REQUIRE(cpuset_setaffinity(CPU_LEVEL_CPUSET,
+ CPU_WHICH_CPUSET, setid,
+ sizeof(setmask), &setmask) == 0);
+
+ if (sigstop_from_main_thread) {
+ /*
+ * We expect the SIGKILL sent when our parent
+ * dies to be delivered to the new thread.
+ * Raise the SIGSTOP in this thread so the
+ * threads compete.
+ */
+ CHILD_REQUIRE(pthread_create(&t, NULL,
+ sleep_thread, NULL) == 0);
+ raise(SIGSTOP);
+ } else {
+ /*
+ * We expect the SIGKILL to be delivered to
+ * this thread. After creating the new thread,
+ * just get off the CPU so the other thread can
+ * raise the SIGSTOP.
+ */
+ CHILD_REQUIRE(pthread_create(&t, NULL,
+ raise_sigstop_thread, NULL) == 0);
+ sleep(60);
+ }
+
+ exit(0);
+ }
+ /* First stop is trace_me() immediately after fork. */
+ wpid = waitpid(fpid, &status, 0);
+ CHILD_REQUIRE(wpid == fpid);
+ CHILD_REQUIRE(WIFSTOPPED(status));
+ CHILD_REQUIRE(WSTOPSIG(status) == SIGSTOP);
+
+ CHILD_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0);
+
+ /* Second stop is from the raise(SIGSTOP). */
+ wpid = waitpid(fpid, &status, 0);
+ CHILD_REQUIRE(wpid == fpid);
+ CHILD_REQUIRE(WIFSTOPPED(status));
+ CHILD_REQUIRE(WSTOPSIG(status) == SIGSTOP);
+
+ /*
+ * Terminate tracing process without detaching. Our child
+ * should be killed.
+ */
+ exit(0);
+ }
+
+ /*
+ * We should get a normal exit from our immediate child and a SIGKILL
+ * exit from our grandchild. The latter case is the interesting one.
+ * Our grandchild should not have stopped due to the SIGSTOP that was
+ * left dangling when its parent died.
+ */
+ for (i = 0; i < 2; ++i) {
+ wpid = wait(&status);
+ if (wpid == fpid)
+ ATF_REQUIRE(WIFEXITED(status) &&
+ WEXITSTATUS(status) == 0);
+ else
+ ATF_REQUIRE(WIFSIGNALED(status) &&
+ WTERMSIG(status) == SIGKILL);
+ }
+}
+
+/*
+ * These two tests ensure that if the tracing process exits without detaching
+ * just after the child received a SIGSTOP, the child is cleanly killed and
+ * doesn't go to sleep due to the SIGSTOP. The parent's death will send a
+ * SIGKILL to the child. If the SIGKILL and the SIGSTOP are handled by
+ * different threads, the SIGKILL must win. There are two variants of this
+ * test, designed to catch the case where the SIGKILL is delivered to the
+ * younger thread (the first test) and the case where the SIGKILL is delivered
+ * to the older thread (the second test). This behavior has changed in the
+ * past, so make no assumption.
+ */
+ATF_TC_WITHOUT_HEAD(ptrace__parent_terminate_with_pending_sigstop1);
+ATF_TC_BODY(ptrace__parent_terminate_with_pending_sigstop1, tc)
+{
+
+ terminate_with_pending_sigstop(true);
+}
+ATF_TC_WITHOUT_HEAD(ptrace__parent_terminate_with_pending_sigstop2);
+ATF_TC_BODY(ptrace__parent_terminate_with_pending_sigstop2, tc)
+{
+
+ terminate_with_pending_sigstop(false);
+}
+
ATF_TP_ADD_TCS(tp)
{
@@ -2829,6 +2963,8 @@ ATF_TP_ADD_TCS(tp)
ATF_TP_ADD_TC(tp, ptrace__PT_CONTINUE_with_signal_mix);
ATF_TP_ADD_TC(tp, ptrace__PT_CONTINUE_with_signal_kqueue);
ATF_TP_ADD_TC(tp, ptrace__PT_CONTINUE_with_signal_thread_sigmask);
+ ATF_TP_ADD_TC(tp, ptrace__parent_terminate_with_pending_sigstop1);
+ ATF_TP_ADD_TC(tp, ptrace__parent_terminate_with_pending_sigstop2);
return (atf_no_error());
}
More information about the svn-src-all
mailing list