[PATCH v3 2/3] ptrace: ensure arch_ptrace/ptrace_request can neverrace with SIGKILL

[Oleg Nesterov]

putreg() assumes that the tracee is not running and pt_regs_access()
can safely play with its stack. However a killed tracee can return
from ptrace_stop() to the low-level asm code and do RESTORE_REST,
this means that debugger can actually read/modify the kernel stack
until the tracee does SAVE_REST again.

set_task_blockstep() can race with SIGKILL too and in some sense
this race is even worse, the very fact the tracee can be woken up
breaks the logic.

As Linus suggested we can clear TASK_WAKEKILL around the arch_ptrace()
call, this ensures that nobody can ever wakeup the tracee while the
debugger looks at it. Not only this fixes the mentioned problems,
we can do some cleanups/simplifications in arch_ptrace() paths.

Probably ptrace_unfreeze_traced() needs more callers, for example
it makes sense to make the tracee killable for oom-killer before
access_process_vm().

While at it, add the comment into may_ptrace_stop() to explain why
ptrace_stop() still can't rely on SIGKILL and signal_pending_state().

Reported-by: Salman Qazi <sqazi@xxxxxxxxxx>
Reported-by: Suleiman Souhlal <suleiman@xxxxxxxxxx>
Suggested-by: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
Signed-off-by: Oleg Nesterov <oleg@xxxxxxxxxx>
---
 arch/x86/kernel/step.c |    9 +++---
 kernel/ptrace.c        |   64 ++++++++++++++++++++++++++++++++++++++++-------
 kernel/signal.c        |    5 +++
 3 files changed, 64 insertions(+), 14 deletions(-)

diff --git a/arch/x86/kernel/step.c b/arch/x86/kernel/step.c
index cd3b243..9b4d51d 100644
--- a/arch/x86/kernel/step.c
+++ b/arch/x86/kernel/step.c
@@ -165,10 +165,11 @@ void set_task_blockstep(struct task_struct *task, bool on)
 	 * Ensure irq/preemption can't change debugctl in between.
 	 * Note also that both TIF_BLOCKSTEP and debugctl should
 	 * be changed atomically wrt preemption.
-	 * FIXME: this means that set/clear TIF_BLOCKSTEP is simply
-	 * wrong if task != current, SIGKILL can wakeup the stopped
-	 * tracee and set/clear can play with the running task, this
-	 * can confuse the next __switch_to_xtra().
+	 *
+	 * NOTE: this means that set/clear TIF_BLOCKSTEP is only safe if
+	 * task is current or it can't be running, otherwise we can race
+	 * with __switch_to_xtra(). We rely on ptrace_freeze_traced() but
+	 * PTRACE_KILL is not safe.
 	 */
 	local_irq_disable();
 	debugctl = get_debugctlmsr();
diff --git a/kernel/ptrace.c b/kernel/ptrace.c
index 62f7c27..b6c22b5 100644
--- a/kernel/ptrace.c
+++ b/kernel/ptrace.c
@@ -122,6 +122,40 @@ void __ptrace_unlink(struct task_struct *child)
 	spin_unlock(&child->sighand->siglock);
 }
 
+/* Ensure that nothing can wake it up, even SIGKILL */
+static bool ptrace_freeze_traced(struct task_struct *task)
+{
+	bool ret = false;
+
+	/* Lockless, nobody but us can set this flag */
+	if (task->jobctl & JOBCTL_LISTENING)
+		return ret;
+
+	spin_lock_irq(&task->sighand->siglock);
+	if (task_is_traced(task) && !__fatal_signal_pending(task)) {
+		task->state = __TASK_TRACED;
+		ret = true;
+	}
+	spin_unlock_irq(&task->sighand->siglock);
+
+	return ret;
+}
+
+static void ptrace_unfreeze_traced(struct task_struct *task)
+{
+	if (task->state != __TASK_TRACED)
+		return;
+
+	WARN_ON(!task->ptrace || task->parent != current);
+
+	spin_lock_irq(&task->sighand->siglock);
+	if (__fatal_signal_pending(task))
+		wake_up_state(task, __TASK_TRACED);
+	else
+		task->state = TASK_TRACED;
+	spin_unlock_irq(&task->sighand->siglock);
+}
+
 /**
  * ptrace_check_attach - check whether ptracee is ready for ptrace operation
  * @child: ptracee to check for
@@ -151,24 +185,29 @@ static int ptrace_check_attach(struct task_struct *child, bool ignore_state)
 	 * be changed by us so it's not changing right after this.
 	 */
 	read_lock(&tasklist_lock);
-	if ((child->ptrace & PT_PTRACED) && child->parent == current) {
+	if (child->ptrace && child->parent == current) {
+		WARN_ON(child->state == __TASK_TRACED);
 		/*
 		 * child->sighand can't be NULL, release_task()
 		 * does ptrace_unlink() before __exit_signal().
 		 */
-		spin_lock_irq(&child->sighand->siglock);
-		WARN_ON_ONCE(task_is_stopped(child));
-		if (ignore_state || (task_is_traced(child) &&
-				     !(child->jobctl & JOBCTL_LISTENING)))
+		if (ignore_state || ptrace_freeze_traced(child))
 			ret = 0;
-		spin_unlock_irq(&child->sighand->siglock);
 	}
 	read_unlock(&tasklist_lock);
 
-	if (!ret && !ignore_state)
-		ret = wait_task_inactive(child, TASK_TRACED) ? 0 : -ESRCH;
+	if (!ret && !ignore_state) {
+		if (!wait_task_inactive(child, __TASK_TRACED)) {
+			/*
+			 * This can only happen if may_ptrace_stop() fails and
+			 * ptrace_stop() changes ->state back to TASK_RUNNING,
+			 * so we should not worry about leaking __TASK_TRACED.
+			 */
+			WARN_ON(child->state == __TASK_TRACED);
+			ret = -ESRCH;
+		}
+	}
 
-	/* All systems go.. */
 	return ret;
 }
 
@@ -900,6 +939,8 @@ SYSCALL_DEFINE4(ptrace, long, request, long, pid, unsigned long, addr,
 		goto out_put_task_struct;
 
 	ret = arch_ptrace(child, request, addr, data);
+	if (request != PTRACE_DETACH)
+		ptrace_unfreeze_traced(child);
 
  out_put_task_struct:
 	put_task_struct(child);
@@ -1039,8 +1080,11 @@ asmlinkage long compat_sys_ptrace(compat_long_t request, compat_long_t pid,
 
 	ret = ptrace_check_attach(child, request == PTRACE_KILL ||
 				  request == PTRACE_INTERRUPT);
-	if (!ret)
+	if (!ret) {
 		ret = compat_arch_ptrace(child, request, addr, data);
+		if (request != PTRACE_DETACH)
+			ptrace_unfreeze_traced(child);
+	}
 
  out_put_task_struct:
 	put_task_struct(child);
diff --git a/kernel/signal.c b/kernel/signal.c
index 2bf2731..980bcf3 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -1794,6 +1794,10 @@ static inline int may_ptrace_stop(void)
 	 * If SIGKILL was already sent before the caller unlocked
 	 * ->siglock we must see ->core_state != NULL. Otherwise it
 	 * is safe to enter schedule().
+	 *
+	 * This is almost outdated, a task with the pending SIGKILL can't
+	 * block in TASK_TRACED. But PTRACE_EVENT_EXIT can be reported
+	 * after SIGKILL was already dequeued.
 	 */
 	if (unlikely(current->mm->core_state) &&
 	    unlikely(current->mm == current->parent->mm))
@@ -1919,6 +1923,7 @@ static void ptrace_stop(int exit_code, int why, int clear_code, siginfo_t *info)
 		if (gstop_done)
 			do_notify_parent_cldstop(current, false, why);
 
+		/* tasklist protects us from ptrace_freeze_traced() */
 		__set_current_state(TASK_RUNNING);
 		if (clear_code)
 			current->exit_code = 0;
-- 
1.5.5.1


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/