Skip to contentSkip to conversation

fix & prevent the missing preemption disabling

PATCH Superseded 王贇 2 participants 6 messages Oct 18, 2021 – Oct 26, 2021

Subsystems (8)
x86 PowerPC RISC-V Tracing +4 more subsystems

v4 of 7 Superseded Latest is v7 →

lore
Share & tools
lore

王贇

author
#0 lore raw

v4 cover fix & prevent the missing preemption disabling

The testing show that perf_ftrace_function_call() are using smp_processor_id() with preemption enabled, all the checking on CPU could be wrong after preemption.

As Peter point out, the section between ftrace_test_recursion_trylock/unlock() pair require the preemption to be disabled as 'Documentation/trace/ftrace-uses.rst' explained, but currently the work is done outside of the helpers.

And since the internal using of trace_test_and_set_recursion() and trace_clear_recursion() also require preemption to be disabled, we can just merge the logical together.

Patch 1/2 will make sure preemption disabled when recursion lock succeed, patch 2/2 will do smp_processor_id() checking after trylock() to address the issue.

v1: https://lore.kernel.org/all/[email protected]/
v2: https://lore.kernel.org/all/[email protected]/
v3: https://lore.kernel.org/all/[email protected]/

Michael Wang (2):
ftrace: disable preemption when recursion locked
ftrace: do CPU checking after preemption disabled

 arch/csky/kernel/probes/ftrace.c     |  2 --
 arch/parisc/kernel/ftrace.c          |  2 --
 arch/powerpc/kernel/kprobes-ftrace.c |  2 --
 arch/riscv/kernel/probes/ftrace.c    |  2 --
 arch/x86/kernel/kprobes/ftrace.c     |  2 --
 include/linux/trace_recursion.h      | 20 +++++++++++++++++++-
 kernel/livepatch/patch.c             | 13 +++++++------
 kernel/trace/ftrace.c                | 15 +++++----------
 kernel/trace/trace_event_perf.c      |  6 +++---
 kernel/trace/trace_functions.c       |  5 -----
 10 files changed, 34 insertions(+), 35 deletions(-)
Signature · 1 line
1.8.3.1
author
#1 lore raw

PATCH v4 1/2 ftrace: disable preemption when recursion locked

As the documentation explained, ftrace_test_recursion_trylock() and ftrace_test_recursion_unlock() were supposed to disable and enable preemption properly, however currently this work is done outside of the function, which could be missing by mistake.

And since the internal using of trace_test_and_set_recursion() and trace_clear_recursion() also require preemption disabled, we can just merge the logical.

This patch will make sure the preemption has been disabled when trace_test_and_set_recursion() return bit >= 0, and trace_clear_recursion() will enable the preemption if previously enabled.

---
 arch/csky/kernel/probes/ftrace.c     |  2 --
 arch/parisc/kernel/ftrace.c          |  2 --
 arch/powerpc/kernel/kprobes-ftrace.c |  2 --
 arch/riscv/kernel/probes/ftrace.c    |  2 --
 arch/x86/kernel/kprobes/ftrace.c     |  2 --
 include/linux/trace_recursion.h      | 20 +++++++++++++++++++-
 kernel/livepatch/patch.c             | 13 +++++++------
 kernel/trace/ftrace.c                | 15 +++++----------
 kernel/trace/trace_functions.c       |  5 -----
 9 files changed, 31 insertions(+), 32 deletions(-)
arch/csky/kernel/probes/ftrace.c +0 -2
index b388228..834cffc 100644
@@ -17,7 +17,6 @@ void kprobe_ftrace_handler(unsigned long ip, unsigned long parent_ip,
 		return;
 
 	regs = ftrace_get_regs(fregs);
-	preempt_disable_notrace();
 	p = get_kprobe((kprobe_opcode_t *)ip);
 	if (!p) {
 		p = get_kprobe((kprobe_opcode_t *)(ip - MCOUNT_INSN_SIZE));
@@ -57,7 +56,6 @@ void kprobe_ftrace_handler(unsigned long ip, unsigned long parent_ip,
 		__this_cpu_write(current_kprobe, NULL);
 	}
 out:
-	preempt_enable_notrace();
 	ftrace_test_recursion_unlock(bit);
 }
 NOKPROBE_SYMBOL(kprobe_ftrace_handler);
arch/parisc/kernel/ftrace.c +0 -2
index 0a1e75a..3543496 100644
@@ -216,7 +216,6 @@ void kprobe_ftrace_handler(unsigned long ip, unsigned long parent_ip,
 		return;
 
 	regs = ftrace_get_regs(fregs);
-	preempt_disable_notrace();
 	p = get_kprobe((kprobe_opcode_t *)ip);
 	if (unlikely(!p) || kprobe_disabled(p))
 		goto out;
@@ -245,7 +244,6 @@ void kprobe_ftrace_handler(unsigned long ip, unsigned long parent_ip,
 	}
 	__this_cpu_write(current_kprobe, NULL);
 out:
-	preempt_enable_notrace();
 	ftrace_test_recursion_unlock(bit);
 }
 NOKPROBE_SYMBOL(kprobe_ftrace_handler);
arch/powerpc/kernel/kprobes-ftrace.c +0 -2
index 7154d58..072ebe7 100644
@@ -26,7 +26,6 @@ void kprobe_ftrace_handler(unsigned long nip, unsigned long parent_nip,
 		return;
 
 	regs = ftrace_get_regs(fregs);
-	preempt_disable_notrace();
 	p = get_kprobe((kprobe_opcode_t *)nip);
 	if (unlikely(!p) || kprobe_disabled(p))
 		goto out;
@@ -61,7 +60,6 @@ void kprobe_ftrace_handler(unsigned long nip, unsigned long parent_nip,
 		__this_cpu_write(current_kprobe, NULL);
 	}
 out:
-	preempt_enable_notrace();
 	ftrace_test_recursion_unlock(bit);
 }
 NOKPROBE_SYMBOL(kprobe_ftrace_handler);
arch/riscv/kernel/probes/ftrace.c +0 -2
index aab85a8..7142ec4 100644
@@ -15,7 +15,6 @@ void kprobe_ftrace_handler(unsigned long ip, unsigned long parent_ip,
 	if (bit < 0)
 		return;
 
-	preempt_disable_notrace();
 	p = get_kprobe((kprobe_opcode_t *)ip);
 	if (unlikely(!p) || kprobe_disabled(p))
 		goto out;
@@ -52,7 +51,6 @@ void kprobe_ftrace_handler(unsigned long ip, unsigned long parent_ip,
 		__this_cpu_write(current_kprobe, NULL);
 	}
 out:
-	preempt_enable_notrace();
 	ftrace_test_recursion_unlock(bit);
 }
 NOKPROBE_SYMBOL(kprobe_ftrace_handler);
arch/x86/kernel/kprobes/ftrace.c +0 -2
index 596de2f..dd2ec14 100644
@@ -25,7 +25,6 @@ void kprobe_ftrace_handler(unsigned long ip, unsigned long parent_ip,
 	if (bit < 0)
 		return;
 
-	preempt_disable_notrace();
 	p = get_kprobe((kprobe_opcode_t *)ip);
 	if (unlikely(!p) || kprobe_disabled(p))
 		goto out;
@@ -59,7 +58,6 @@ void kprobe_ftrace_handler(unsigned long ip, unsigned long parent_ip,
 		__this_cpu_write(current_kprobe, NULL);
 	}
 out:
-	preempt_enable_notrace();
 	ftrace_test_recursion_unlock(bit);
 }
 NOKPROBE_SYMBOL(kprobe_ftrace_handler);
include/linux/trace_recursion.h +19 -1
index a9f9c57..f13a9f9 100644
@@ -159,6 +159,9 @@ static __always_inline int trace_get_context_bit(void)
 # define do_ftrace_record_recursion(ip, pip)	do { } while (0)
 #endif
 
+/*
+ * Preemption is promised to be disabled when return bit >= 0.
+ */
 static __always_inline int trace_test_and_set_recursion(unsigned long ip, unsigned long pip,
 							int start, int max)
 {
@@ -189,14 +192,29 @@ static __always_inline int trace_test_and_set_recursion(unsigned long ip, unsign
 	current->trace_recursion = val;
 	barrier();
 
+	/*
+	 * Disable preemption to fulfill the promise.
+	 *
+	 * Don't worry about the bit 0 cases, they indicate
+	 * the disabling behaviour has already been done by
+	 * internal call previously.
+	 */
+	preempt_disable_notrace();
+
 	return bit + 1;
 }
 
+/*
+ * Preemption will be enabled (if it was previously enabled).
+ */
 static __always_inline void trace_clear_recursion(int bit)
 {
 	if (!bit)
 		return;
 
+	if (bit > 0)
+		preempt_enable_notrace();
+
 	barrier();
 	bit--;
 	trace_recursion_clear(bit);
@@ -209,7 +227,7 @@ static __always_inline void trace_clear_recursion(int bit)
  * tracing recursed in the same context (normal vs interrupt),
  *
  * Returns: -1 if a recursion happened.
- *           >= 0 if no recursion
+ *           >= 0 if no recursion.
  */
 static __always_inline int ftrace_test_recursion_trylock(unsigned long ip,
 							 unsigned long parent_ip)
kernel/livepatch/patch.c +7 -6
index e8029ae..b8d75fb 100644
@@ -49,14 +49,16 @@ static void notrace klp_ftrace_handler(unsigned long ip,
 
 	ops = container_of(fops, struct klp_ops, fops);
 
+	/*
+	 *
+	 * The ftrace_test_recursion_trylock() will disable preemption,
+	 * which is required for the variant of synchronize_rcu() that is
+	 * used to allow patching functions where RCU is not watching.
+	 * See klp_synchronize_transition() for more details.
+	 */
 	bit = ftrace_test_recursion_trylock(ip, parent_ip);
 	if (WARN_ON_ONCE(bit < 0))
 		return;
-	/*
-	 * A variant of synchronize_rcu() is used to allow patching functions
-	 * where RCU is not watching, see klp_synchronize_transition().
-	 */
-	preempt_disable_notrace();
 
 	func = list_first_or_null_rcu(&ops->func_stack, struct klp_func,
 				      stack_node);
@@ -120,7 +122,6 @@ static void notrace klp_ftrace_handler(unsigned long ip,
 	klp_arch_set_pc(fregs, (unsigned long)func->new_func);
 
 unlock:
-	preempt_enable_notrace();
 	ftrace_test_recursion_unlock(bit);
 }
 
kernel/trace/ftrace.c +5 -10
index 7d0fd89..35cccef 100644
@@ -6977,16 +6977,15 @@ void ftrace_reset_array_ops(struct trace_array *tr)
 	struct ftrace_ops *op;
 	int bit;
 
+	/*
+	 * The ftrace_test_and_set_recursion() will disable preemption,
+	 * which is required since some of the ops may be dynamically
+	 * allocated, they must be freed after a synchronize_rcu().
+	 */
 	bit = trace_test_and_set_recursion(ip, parent_ip, TRACE_LIST_START, TRACE_LIST_MAX);
 	if (bit < 0)
 		return;
 
-	/*
-	 * Some of the ops may be dynamically allocated,
-	 * they must be freed after a synchronize_rcu().
-	 */
-	preempt_disable_notrace();
-
 	do_for_each_ftrace_op(op, ftrace_ops_list) {
 		/* Stub functions don't need to be called nor tested */
 		if (op->flags & FTRACE_OPS_FL_STUB)
@@ -7010,7 +7009,6 @@ void ftrace_reset_array_ops(struct trace_array *tr)
 		}
 	} while_for_each_ftrace_op(op);
 out:
-	preempt_enable_notrace();
 	trace_clear_recursion(bit);
 }
 
@@ -7056,12 +7054,9 @@ static void ftrace_ops_assist_func(unsigned long ip, unsigned long parent_ip,
 	if (bit < 0)
 		return;
 
-	preempt_disable_notrace();
-
 	if (!(op->flags & FTRACE_OPS_FL_RCU) || rcu_is_watching())
 		op->func(ip, parent_ip, op, fregs);
 
-	preempt_enable_notrace();
 	trace_clear_recursion(bit);
 }
 NOKPROBE_SYMBOL(ftrace_ops_assist_func);
kernel/trace/trace_functions.c +0 -5
index 1f0e63f..9f1bfbe 100644
@@ -186,7 +186,6 @@ static void function_trace_start(struct trace_array *tr)
 		return;
 
 	trace_ctx = tracing_gen_ctx();
-	preempt_disable_notrace();
 
 	cpu = smp_processor_id();
 	data = per_cpu_ptr(tr->array_buffer.data, cpu);
@@ -194,7 +193,6 @@ static void function_trace_start(struct trace_array *tr)
 		trace_function(tr, ip, parent_ip, trace_ctx);
 
 	ftrace_test_recursion_unlock(bit);
-	preempt_enable_notrace();
 }
 
 #ifdef CONFIG_UNWINDER_ORC
@@ -298,8 +296,6 @@ static inline void process_repeats(struct trace_array *tr,
 	if (bit < 0)
 		return;
 
-	preempt_disable_notrace();
-
 	cpu = smp_processor_id();
 	data = per_cpu_ptr(tr->array_buffer.data, cpu);
 	if (atomic_read(&data->disabled))
@@ -324,7 +320,6 @@ static inline void process_repeats(struct trace_array *tr,
 
 out:
 	ftrace_test_recursion_unlock(bit);
-	preempt_enable_notrace();
 }
 
 static void
Signature · 1 line
1.8.3.1
author
#2 lore raw

PATCH v4 2/2 ftrace: do CPU checking after preemption disabled

In reply to 王贇

With CONFIG_DEBUG_PREEMPT we observed reports like:

  BUG: using smp_processor_id() in preemptible
  caller is perf_ftrace_function_call+0x6f/0x2e0
  CPU: 1 PID: 680 Comm: a.out Not tainted
  Call Trace:
   <TASK>
   dump_stack_lvl+0x8d/0xcf
   check_preemption_disabled+0x104/0x110
   ? optimize_nops.isra.7+0x230/0x230
   ? text_poke_bp_batch+0x9f/0x310
   perf_ftrace_function_call+0x6f/0x2e0
   ...
   __text_poke+0x5/0x620
   text_poke_bp_batch+0x9f/0x310

This telling us the CPU could be changed after task is preempted, and the checking on CPU before preemption will be invalid.

Since now ftrace_test_recursion_trylock() will help to disable the preemption, this patch just do the checking after trylock() to address the issue.

---
 kernel/trace/trace_event_perf.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
kernel/trace/trace_event_perf.c +3 -3
index 6aed10e..fba8cb7 100644
@@ -441,13 +441,13 @@ void perf_trace_buf_update(void *record, u16 type)
 	if (!rcu_is_watching())
 		return;
 
-	if ((unsigned long)ops->private != smp_processor_id())
-		return;
-
 	bit = ftrace_test_recursion_trylock(ip, parent_ip);
 	if (bit < 0)
 		return;
 
+	if ((unsigned long)ops->private != smp_processor_id())
+		goto out;
+
 	event = container_of(ops, struct perf_event, ftrace_ops);
 
 	/*
Signature · 1 line
1.8.3.1
author
#3 lore raw
In reply to 王贇

Just a ping, to see if there are any more comments :-P

Regards,
Michael Wang

Quoted 35 lines · 王贇

The testing show that perf_ftrace_function_call() are using smp_processor_id() with preemption enabled, all the checking on CPU could be wrong after preemption.

As Peter point out, the section between ftrace_test_recursion_trylock/unlock() pair require the preemption to be disabled as 'Documentation/trace/ftrace-uses.rst' explained, but currently the work is done outside of the helpers.

And since the internal using of trace_test_and_set_recursion() and trace_clear_recursion() also require preemption to be disabled, we can just merge the logical together.

Patch 1/2 will make sure preemption disabled when recursion lock succeed, patch 2/2 will do smp_processor_id() checking after trylock() to address the issue.

v1: https://lore.kernel.org/all/[email protected]/
v2: https://lore.kernel.org/all/[email protected]/
v3: https://lore.kernel.org/all/[email protected]/

Michael Wang (2):
ftrace: disable preemption when recursion locked
ftrace: do CPU checking after preemption disabled

 arch/csky/kernel/probes/ftrace.c     |  2 --
 arch/parisc/kernel/ftrace.c          |  2 --
 arch/powerpc/kernel/kprobes-ftrace.c |  2 --
 arch/riscv/kernel/probes/ftrace.c    |  2 --
 arch/x86/kernel/kprobes/ftrace.c     |  2 --
 include/linux/trace_recursion.h      | 20 +++++++++++++++++++-
 kernel/livepatch/patch.c             | 13 +++++++------
 kernel/trace/ftrace.c                | 15 +++++----------
 kernel/trace/trace_event_perf.c      |  6 +++---
 kernel/trace/trace_functions.c       |  5 -----
 10 files changed, 34 insertions(+), 35 deletions(-)
maintainer
#4 lore raw
Quoted 1 line · On Tue, 26 Oct 2021 10:09:12 +0800 王贇 <[email protected]> wrote:

Just a ping, to see if there are any more comments :-P

I guess you missed what went into mainline (and your name found a bug in my perl script for importing patches ;-)

https://lore.kernel.org/all/[email protected]/

Which means patch 1 needs to change:

Quoted 13 lines
(quoted diff) +10 -0
+	/*
+	 * Disable preemption to fulfill the promise.
+	 *
+	 * Don't worry about the bit 0 cases, they indicate
+	 * the disabling behaviour has already been done by
+	 * internal call previously.
+	 */
+	preempt_disable_notrace();
+
 	return bit + 1;
 }
 
+/*

Review

Quoted 10 lines
(quoted diff) +5 -0
+ * Preemption will be enabled (if it was previously enabled).
+ */
 static __always_inline void trace_clear_recursion(int bit)
 {
 	if (!bit)
 		return;
 
+	if (bit > 0)
+		preempt_enable_notrace();
+

Where this wont work anymore.

Need to preempt disable and enable always.

-- Steve

Quoted 6 lines
(quoted diff) +0 -0
 	barrier();
 	bit--;
 	trace_recursion_clear(bit);
@@ -209,7 +227,7 @@ static __always_inline void trace_clear_recursion(int bit)
  * tracing recursed in the same context (normal vs interrupt),
  *
author
#5 lore raw
Quoted 9 lines · Steven Rostedt
Quoted 1 line · On Tue, 26 Oct 2021 10:09:12 +0800 王贇 <[email protected]> wrote:

Just a ping, to see if there are any more comments :-P

I guess you missed what went into mainline (and your name found a bug in my perl script for importing patches ;-)

https://lore.kernel.org/all/[email protected]/

Cool~ Missing some chinese font maybe, that's fine :-)

Quoted 22 lines

Which means patch 1 needs to change:

Quoted 20 lines
(quoted diff) +12 -0
+	/*
+	 * Disable preemption to fulfill the promise.
+	 *
+	 * Don't worry about the bit 0 cases, they indicate
+	 * the disabling behaviour has already been done by
+	 * internal call previously.
+	 */
+	preempt_disable_notrace();
+
 	return bit + 1;
 }
 
+/*
+ * Preemption will be enabled (if it was previously enabled).
+ */
 static __always_inline void trace_clear_recursion(int bit)
 {
 	if (!bit)
 		return;
Quoted 7 lines

Review

Quoted 3 lines
(quoted diff) +3 -0
+	if (bit > 0)
+		preempt_enable_notrace();
+

Where this wont work anymore.

Need to preempt disable and enable always.

Yup, will rebase on the latest changes~

Regards,
Michael Wang

Quoted 10 lines

-- Steve

Quoted 6 lines
(quoted diff) +0 -0
 	barrier();
 	bit--;
 	trace_recursion_clear(bit);
@@ -209,7 +227,7 @@ static __always_inline void trace_clear_recursion(int bit)
  * tracing recursed in the same context (normal vs interrupt),
  *