All of lore.kernel.org
 help / color / mirror / Atom feed
From: Tejun Heo <tj@kernel.org>
To: oleg@redhat.com, roland@redhat.com, linux-kernel@vger.kernel.org,
	torvalds@linux-foundation.org, akpm@linux-foundation.org,
	rjw@sisk.pl, jan.kratochvil@redhat.com
Cc: Tejun Heo <tj@kernel.org>
Subject: [PATCH 14/16] ptrace: make SIGCONT notification reliable against ptrace
Date: Mon,  6 Dec 2010 17:57:02 +0100	[thread overview]
Message-ID: <1291654624-6230-15-git-send-email-tj@kernel.org> (raw)
In-Reply-To: <1291654624-6230-1-git-send-email-tj@kernel.org>

Currently, SIGCONT notifications which are pending on ptrace attach or
occur while ptraced are reported to the tracer and never make it to
the real parent.

This patch adds a new signal flag SIGNAL_NOTIFY_CONT which is set when
a task is woken up by SIGCONT and cleared once the event is notified
to the parent.  SIGNAL_CLD_MASK bits are no longer cleared after
notification.  Combined with clearing SIGNAL_CLD_MASK if
!SIGNAL_NOTIFY_CONT on ptrace attach, these bits are set on ptrace
detach iff the tracee owes a notification to the real parent.
__ptrace_unlink() is updated to check these bits and reschedule
SIGCONT notification if necessary.

As notification delivery should consider both the child's signal flags
and ptraced state, it should be done while holding both siglock and
tasklist_lock.  The delivery logic is moved into
do_notify_parent_cldstop() and performed while holding both
tasklist_lock and siglock.

This change puts the initial SIGNAL_NOTIFY_CONT test out of siglock.
This is safe as the bit set by the waker are guaranteed to be visible
to the woken up task.

Oleg spotted the above race condition caused by ptrace state and
signal state protected by different locks.

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Roland McGrath <roland@redhat.com>
---
 include/linux/sched.h |    1 +
 kernel/ptrace.c       |   36 ++++++++++++++++++++++++++++++++++++
 kernel/signal.c       |   49 ++++++++++++++++++++++++++++---------------------
 3 files changed, 65 insertions(+), 21 deletions(-)

diff --git a/include/linux/sched.h b/include/linux/sched.h
index 7a26e7d..ed7725b 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -654,6 +654,7 @@ struct signal_struct {
 #define SIGNAL_UNKILLABLE	0x00000040 /* for init: ignore fatal signals */
 
 #define SIGNAL_NOTIFY_STOP	0x00000100 /* notify parent of group stop */
+#define SIGNAL_NOTIFY_CONT	0x00000200 /* notify parent of continuation */
 
 /* If true, all threads except ->group_exit_task have pending SIGKILL */
 static inline int signal_group_exit(const struct signal_struct *sig)
diff --git a/kernel/ptrace.c b/kernel/ptrace.c
index 6ac12f4..dba6aeb 100644
--- a/kernel/ptrace.c
+++ b/kernel/ptrace.c
@@ -52,6 +52,7 @@ void __ptrace_link(struct task_struct *child, struct task_struct *new_parent)
 void __ptrace_unlink(struct task_struct *child)
 {
 	struct signal_struct *sig = child->signal;
+	bool woken_up = false;
 
 	BUG_ON(!child->ptrace);
 
@@ -66,6 +67,33 @@ void __ptrace_unlink(struct task_struct *child)
 		if (sig->flags & SIGNAL_STOP_STOPPED || sig->group_stop_count)
 			child->group_stop |= GROUP_STOP_PENDING;
 		signal_wake_up(child, 1);
+		woken_up = true;
+	}
+
+	/*
+	 * SIGNAL_CLD_MASK is cleared only on a stop signal or, if
+	 * notification isn't pending, ptrace attach.  If any bit is
+	 * set,
+	 *
+	 * - SIGCONT notification was pending before attach or there
+	 *   was one or more SIGCONT notifications while tracing.
+	 *
+	 * - And, there hasn't been any stop signal since the last
+	 *   pending SIGCONT notification.
+	 *
+	 * Combined, it means that the tracee owes a SIGCONT
+	 * notification to the real parent.
+	 */
+	if (sig->flags & SIGNAL_CLD_MASK) {
+		sig->flags |= SIGNAL_NOTIFY_CONT;
+		/*
+		 * Force the tracee into signal delivery path so that
+		 * the notification is delievered ASAP.  This wakeup
+		 * is unintrusive as SIGCONT delivery would have
+		 * caused the same effect.
+		 */
+		if (!woken_up)
+			signal_wake_up(child, 0);
 	}
 
 	child->ptrace = 0;
@@ -245,6 +273,14 @@ int ptrace_attach(struct task_struct *task)
 		signal_wake_up(task, 1);
 	}
 
+	/*
+	 * Clear SIGNAL_CLD_MASK if NOTIFY_CONT is not set.  This is
+	 * used to preserve SIGCONT notification across ptrace
+	 * attach/detach.  Read the comment in __ptrace_unlink().
+	 */
+	if (!(task->signal->flags & SIGNAL_NOTIFY_CONT))
+		task->signal->flags &= ~SIGNAL_CLD_MASK;
+
 	spin_unlock(&task->sighand->siglock);
 
 	retval = 0;
diff --git a/kernel/signal.c b/kernel/signal.c
index 3196367..7b6f972 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -832,7 +832,8 @@ static int prepare_signal(int sig, struct task_struct *p, int from_ancestor_ns)
 			 * will take ->siglock, notice SIGNAL_CLD_MASK, and
 			 * notify its parent. See get_signal_to_deliver().
 			 */
-			signal->flags = why | SIGNAL_STOP_CONTINUED;
+			why |= SIGNAL_STOP_CONTINUED | SIGNAL_NOTIFY_CONT;
+			signal->flags = why;
 			signal->group_stop_count = 0;
 			signal->group_exit_code = 0;
 		} else {
@@ -1603,6 +1604,8 @@ int do_notify_parent(struct task_struct *tsk, int sig)
  * Depending on @why and other conditions, the notification might be
  * skipped.
  *
+ * CLD_CONTINUED	: Notify once if SIGNAL_NOTIFY_CONT is set.
+ *
  * CLD_STOPPED		: If ptraced, always notify; otherwise, notify
  *			  once if SIGNAL_NOTIFY_STOP is set.
  *
@@ -1639,7 +1642,24 @@ static void do_notify_parent_cldstop(struct task_struct *tsk, int why)
 
 	switch (why) {
 	case CLD_CONTINUED:
-		notify = why;
+		/*
+		 * Notify once if NOTIFY_CONT is set regardless of ptrace.
+		 * NOTIFY_CONT will be reinstated on detach if necessary.
+		 */
+		if (!(sig->flags & SIGNAL_NOTIFY_CONT))
+			break;
+
+		/*
+		 * If ptraced, always report CLD_CONTINUED; otherwise,
+		 * prepare_signal(SIGCONT) encodes the CLD_ si_code into
+		 * SIGNAL_CLD_MASK bits.
+		 */
+		if (task_ptrace(tsk) || (sig->flags & SIGNAL_CLD_CONTINUED))
+			notify = CLD_CONTINUED;
+		else
+			notify = CLD_STOPPED;
+
+		sig->flags &= ~SIGNAL_NOTIFY_CONT;
 		break;
 
 	case CLD_STOPPED:
@@ -2015,31 +2035,18 @@ relock:
 	 */
 	try_to_freeze();
 
-	spin_lock_irq(&sighand->siglock);
 	/*
-	 * Every stopped thread goes here after wakeup. Check to see if
-	 * we should notify the parent, prepare_signal(SIGCONT) encodes
-	 * the CLD_ si_code into SIGNAL_CLD_MASK bits.
+	 * Every stopped thread should go through this function after
+	 * waking up.  Check to see if we should notify the parent.
 	 */
-	if (unlikely(signal->flags & SIGNAL_CLD_MASK)) {
-		int why;
-
-		if (task_ptrace(current) ||
-		    (signal->flags & SIGNAL_CLD_CONTINUED))
-			why = CLD_CONTINUED;
-		else
-			why = CLD_STOPPED;
-
-		signal->flags &= ~SIGNAL_CLD_MASK;
-
-		spin_unlock_irq(&sighand->siglock);
-
+	if (unlikely(current->signal->flags & SIGNAL_NOTIFY_CONT)) {
 		read_lock(&tasklist_lock);
-		do_notify_parent_cldstop(current, why);
+		do_notify_parent_cldstop(current, CLD_CONTINUED);
 		read_unlock(&tasklist_lock);
-		goto relock;
 	}
 
+	spin_lock_irq(&sighand->siglock);
+
 	for (;;) {
 		struct k_sigaction *ka;
 		/*
-- 
1.7.1


  parent reply	other threads:[~2010-12-06 16:58 UTC|newest]

Thread overview: 62+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2010-12-06 16:56 [PATCHSET] ptrace,signal: sane interaction between ptrace and job control signals, take#2 Tejun Heo
2010-12-06 16:56 ` [PATCH 01/16] signal: fix SIGCONT notification code Tejun Heo
2010-12-06 16:56 ` [PATCH 02/16] signal: fix CLD_CONTINUED notification target Tejun Heo
2010-12-20 14:58   ` Oleg Nesterov
2010-12-21 16:31     ` Tejun Heo
2010-12-06 16:56 ` [PATCH 03/16] signal: remove superflous try_to_freeze() loop in do_signal_stop() Tejun Heo
2010-12-20 14:59   ` Oleg Nesterov
2010-12-06 16:56 ` [PATCH 04/16] ptrace: kill tracehook_notify_jctl() Tejun Heo
2010-12-20 14:59   ` Oleg Nesterov
2010-12-21 17:00     ` Tejun Heo
2010-12-06 16:56 ` [PATCH 05/16] ptrace: add @why to ptrace_stop() Tejun Heo
2010-12-06 16:56 ` [PATCH 06/16] signal: fix premature completion of group stop when interfered by ptrace Tejun Heo
2010-12-20 15:00   ` Oleg Nesterov
2010-12-21 17:04     ` Tejun Heo
2010-12-06 16:56 ` [PATCH 07/16] signal: use GROUP_STOP_PENDING to stop once for a single group stop Tejun Heo
2010-12-06 16:56 ` [PATCH 08/16] ptrace: participate in group stop from ptrace_stop() iff the task is trapping for " Tejun Heo
2010-12-06 16:56 ` [PATCH 09/16] ptrace: make do_signal_stop() use ptrace_stop() if the task is being ptraced Tejun Heo
2010-12-23 12:26   ` Oleg Nesterov
2010-12-23 13:53     ` Tejun Heo
2010-12-23 16:06       ` Oleg Nesterov
2010-12-23 16:33         ` Tejun Heo
2011-01-17 22:09     ` Roland McGrath
2011-01-27 13:56       ` Tejun Heo
2011-01-28 20:30         ` Roland McGrath
2011-01-31 14:39           ` Tejun Heo
2010-12-06 16:56 ` [PATCH 10/16] ptrace: clean transitions between TASK_STOPPED and TRACED Tejun Heo
2010-12-20 15:00   ` Oleg Nesterov
2010-12-21 17:31     ` Tejun Heo
2010-12-21 17:32       ` Tejun Heo
2010-12-22 10:54       ` Tejun Heo
2010-12-22 11:39       ` Oleg Nesterov
2010-12-22 15:14         ` Tejun Heo
2010-12-22 16:00           ` Oleg Nesterov
2010-12-22 16:21             ` Tejun Heo
2010-12-06 16:56 ` [PATCH 11/16] signal: prepare for CLD_* notification changes Tejun Heo
2010-12-20 16:21   ` Oleg Nesterov
2010-12-20 16:23     ` Oleg Nesterov
2010-12-21 17:35     ` Tejun Heo
2010-12-06 16:57 ` [PATCH 12/16] ptrace: make group stop notification reliable against ptrace Tejun Heo
2010-12-20 17:34   ` Oleg Nesterov
2010-12-21 17:43     ` Tejun Heo
2010-12-22 11:54       ` Oleg Nesterov
2010-12-22 15:26         ` Tejun Heo
2010-12-22 16:02           ` Oleg Nesterov
2010-12-06 16:57 ` [PATCH 13/16] ptrace: reorganize __ptrace_unlink() and ptrace_untrace() Tejun Heo
2010-12-20 18:15   ` Oleg Nesterov
2010-12-21 17:54     ` Tejun Heo
2010-12-06 16:57 ` Tejun Heo [this message]
2010-12-20 19:43   ` [PATCH 14/16] ptrace: make SIGCONT notification reliable against ptrace Oleg Nesterov
2010-12-21 17:48     ` Tejun Heo
2010-12-22 12:16       ` Oleg Nesterov
2010-12-21 17:25   ` Oleg Nesterov
2010-12-22 10:35     ` Tejun Heo
2010-12-06 16:57 ` [PATCH 15/16] ptrace: make sure SIGNAL_NOTIFY_CONT is checked after ptrace_signal() Tejun Heo
2010-12-06 16:57 ` [PATCH 16/16] ptrace: remove the extra wake_up_process() from ptrace_detach() Tejun Heo
2010-12-07  0:10   ` Roland McGrath
2010-12-07 13:43     ` Tejun Heo
2010-12-21 17:54   ` Oleg Nesterov
2010-12-22 10:36     ` Tejun Heo
2010-12-14 17:36 ` [PATCHSET] ptrace,signal: sane interaction between ptrace and job control signals, take#2 Oleg Nesterov
2010-12-14 17:46   ` Tejun Heo
2010-12-22 15:20 ` Oleg Nesterov

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1291654624-6230-15-git-send-email-tj@kernel.org \
    --to=tj@kernel.org \
    --cc=akpm@linux-foundation.org \
    --cc=jan.kratochvil@redhat.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=oleg@redhat.com \
    --cc=rjw@sisk.pl \
    --cc=roland@redhat.com \
    --cc=torvalds@linux-foundation.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.