[PATCH v2 0/2] ptrace: DEBUGCTLMSR_BTF fixes

[PATCH v2 0/2] ptrace: DEBUGCTLMSR_BTF fixes

[Oleg Nesterov]

Hello.

The subject says "ptrace:", but we need these changes for uprobes.

Although I think this fix makes sense by itself. I am really
surprised how many problems this simple code has, or I am
totally confused (this is very possible).

Could someone review 2/2 ?


The next step will change uprobes to use the new helper and avoid
user_enable_single_step().

Oleg.

[PATCH v2 1/2] ptrace: introduce set_task_blockstep() helper

[Oleg Nesterov]

No functional changes, preparation for the next fix and for uprobes
single-step fixes.

Move the code playing with TIF_BLOCKSTEP/DEBUGCTLMSR_BTF into the
new helper, set_task_blockstep().

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
---
 arch/x86/kernel/step.c |   41 +++++++++++++++++++++--------------------
 1 files changed, 21 insertions(+), 20 deletions(-)

diff --git a/arch/x86/kernel/step.c b/arch/x86/kernel/step.c
index c346d11..7a51498 100644
--- a/arch/x86/kernel/step.c
+++ b/arch/x86/kernel/step.c
@@ -157,6 +157,21 @@ static int enable_single_step(struct task_struct *child)
 	return 1;
 }
 
+static void set_task_blockstep(struct task_struct *task, bool on)
+{
+	unsigned long debugctl;
+
+	debugctl = get_debugctlmsr();
+	if (on) {
+		debugctl |= DEBUGCTLMSR_BTF;
+		set_tsk_thread_flag(task, TIF_BLOCKSTEP);
+	} else {
+		debugctl &= ~DEBUGCTLMSR_BTF;
+		clear_tsk_thread_flag(task, TIF_BLOCKSTEP);
+	}
+	update_debugctlmsr(debugctl);
+}
+
 /*
  * Enable single or block step.
  */
@@ -169,19 +184,10 @@ static void enable_step(struct task_struct *child, bool block)
 	 * So no one should try to use debugger block stepping in a program
 	 * that uses user-mode single stepping itself.
 	 */
-	if (enable_single_step(child) && block) {
-		unsigned long debugctl = get_debugctlmsr();
-
-		debugctl |= DEBUGCTLMSR_BTF;
-		update_debugctlmsr(debugctl);
-		set_tsk_thread_flag(child, TIF_BLOCKSTEP);
-	} else if (test_tsk_thread_flag(child, TIF_BLOCKSTEP)) {
-		unsigned long debugctl = get_debugctlmsr();
-
-		debugctl &= ~DEBUGCTLMSR_BTF;
-		update_debugctlmsr(debugctl);
-		clear_tsk_thread_flag(child, TIF_BLOCKSTEP);
-	}
+	if (enable_single_step(child) && block)
+		set_task_blockstep(child, true);
+	else if (test_tsk_thread_flag(child, TIF_BLOCKSTEP))
+		set_task_blockstep(child, false);
 }
 
 void user_enable_single_step(struct task_struct *child)
@@ -199,13 +205,8 @@ void user_disable_single_step(struct task_struct *child)
 	/*
 	 * Make sure block stepping (BTF) is disabled.
 	 */
-	if (test_tsk_thread_flag(child, TIF_BLOCKSTEP)) {
-		unsigned long debugctl = get_debugctlmsr();
-
-		debugctl &= ~DEBUGCTLMSR_BTF;
-		update_debugctlmsr(debugctl);
-		clear_tsk_thread_flag(child, TIF_BLOCKSTEP);
-	}
+	if (test_tsk_thread_flag(child, TIF_BLOCKSTEP))
+		set_task_blockstep(child, false);
 
 	/* Always clear TIF_SINGLESTEP... */
 	clear_tsk_thread_flag(child, TIF_SINGLESTEP);
-- 
1.5.5.1

[PATCH v2 2/2] [PATCH] ptrace: partly fix set_task_blockstep()->update_debugctlmsr() logic

[Oleg Nesterov]

Afaics the usage of update_debugctlmsr() and TIF_BLOCKSTEP in
step.c was always very wrong.

1. update_debugctlmsr() was simply unneeded. The child sleeps
   TASK_TRACED, __switch_to_xtra(next_p => child) should notice
   TIF_BLOCKSTEP and set/clear DEBUGCTLMSR_BTF after resume if
   needed.

2. It is wrong. The state of DEBUGCTLMSR_BTF bit in CPU register
   should always match the state of current's TIF_BLOCKSTEP bit.

3. Even get_debugctlmsr() + update_debugctlmsr() itself does not
   look right. Irq can change other bits in MSR_IA32_DEBUGCTLMSR
   register or the caller can be preempted in between.

4. It is not safe to play with TIF_BLOCKSTEP if task != current.
   DEBUGCTLMSR_BTF and TIF_BLOCKSTEP should always match each
   other if the task is running. The tracee is stopped but it
   can be SIGKILL'ed right before set/clear_tsk_thread_flag().

However, now that uprobes uses user_enable_single_step(current)
we can't simply remove update_debugctlmsr(). So this patch adds
the additional "task == current" check and disables irqs to avoid
the race with interrupts/preemption.

Unfortunately this patch doesn't solve the last problem, we need
another fix. Probably we should teach ptrace_stop() to set/clear
single/block stepping after resume.

And afaics there is yet another problem: perf can play with
MSR_IA32_DEBUGCTLMSR from nmi, this obviously means that even
__switch_to_xtra() has problems.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
---
 arch/x86/kernel/step.c |   14 +++++++++++++-
 1 files changed, 13 insertions(+), 1 deletions(-)

diff --git a/arch/x86/kernel/step.c b/arch/x86/kernel/step.c
index 7a51498..f89cdc6 100644
--- a/arch/x86/kernel/step.c
+++ b/arch/x86/kernel/step.c
@@ -161,6 +161,16 @@ static void set_task_blockstep(struct task_struct *task, bool on)
 {
 	unsigned long debugctl;
 
+	/*
+	 * 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().
+	 */
+	local_irq_disable();
 	debugctl = get_debugctlmsr();
 	if (on) {
 		debugctl |= DEBUGCTLMSR_BTF;
@@ -169,7 +179,9 @@ static void set_task_blockstep(struct task_struct *task, bool on)
 		debugctl &= ~DEBUGCTLMSR_BTF;
 		clear_tsk_thread_flag(task, TIF_BLOCKSTEP);
 	}
-	update_debugctlmsr(debugctl);
+	if (task == current)
+		update_debugctlmsr(debugctl);
+	local_irq_enable();
 }
 
 /*
-- 
1.5.5.1