tracing/kprobes: Use global event perf buffers in kprobe tracer

Use new percpu global event buffer instead of stack in kprobe
tracer while tracing through perf.
Signed-off-by: NMasami Hiramatsu <mhiramat@redhat.com>
Acked-by: NSteven Rostedt <rostedt@goodmis.org>
Acked-by: NIngo Molnar <mingo@elte.hu>
Cc: Jim Keniston <jkenisto@us.ibm.com>
Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Frank Ch. Eigler <fche@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jason Baron <jbaron@redhat.com>
Cc: K.Prasad <prasad@linux.vnet.ibm.com>
Cc: Lai Jiangshan <laijs@cn.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <20090925182011.10157.60140.stgit@omoto>
Signed-off-by: NFrederic Weisbecker <fweisbec@gmail.com>

kernel/trace/trace_kprobe.c

@@ -1149,23 +1149,36 @@ static __kprobes int kprobe_profile_func(struct kprobe *kp,
 	struct trace_probe *tp = container_of(kp, struct trace_probe, rp.kp);
 	struct ftrace_event_call *call = &tp->call;
 	struct kprobe_trace_entry *entry;
-	int size, __size, i, pc;
+	struct trace_entry *ent;
+	int size, __size, i, pc, __cpu;
 	unsigned long irq_flags;
+	char *raw_data;
 
-	local_save_flags(irq_flags);
 	pc = preempt_count();
-
 	__size = SIZEOF_KPROBE_TRACE_ENTRY(tp->nr_args);
 	size = ALIGN(__size + sizeof(u32), sizeof(u64));
 	size -= sizeof(u32);
+	if (WARN_ONCE(size > FTRACE_MAX_PROFILE_SIZE,
+		     "profile buffer not large enough"))
+		return 0;
 
-	do {
-		char raw_data[size];
-		struct trace_entry *ent;
 	/*
-		 * Zero dead bytes from alignment to avoid stack leak
-		 * to userspace
+	 * Protect the non nmi buffer
+	 * This also protects the rcu read side
 	 */
+	local_irq_save(irq_flags);
+	__cpu = smp_processor_id();
+
+	if (in_nmi())
+		raw_data = rcu_dereference(trace_profile_buf_nmi);
+	else
+		raw_data = rcu_dereference(trace_profile_buf);
+
+	if (!raw_data)
+		goto end;
+
+	raw_data = per_cpu_ptr(raw_data, __cpu);
+	/* Zero dead bytes from alignment to avoid buffer leak to userspace */
 	*(u64 *)(&raw_data[size - sizeof(u64)]) = 0ULL;
 	entry = (struct kprobe_trace_entry *)raw_data;
 	ent = &entry->ent;
@@ -1177,7 +1190,8 @@ static __kprobes int kprobe_profile_func(struct kprobe *kp,
 	for (i = 0; i < tp->nr_args; i++)
 		entry->args[i] = call_fetch(&tp->args[i].fetch, regs);
 	perf_tp_event(call->id, entry->ip, 1, entry, size);
-	} while (0);
+end:
+	local_irq_restore(irq_flags);
 	return 0;
 }
 
@@ -1188,20 +1202,36 @@ static __kprobes int kretprobe_profile_func(struct kretprobe_instance *ri,
 	struct trace_probe *tp = container_of(ri->rp, struct trace_probe, rp);
 	struct ftrace_event_call *call = &tp->call;
 	struct kretprobe_trace_entry *entry;
-	int size, __size, i, pc;
+	struct trace_entry *ent;
+	int size, __size, i, pc, __cpu;
 	unsigned long irq_flags;
+	char *raw_data;
 
-	local_save_flags(irq_flags);
 	pc = preempt_count();
-
 	__size = SIZEOF_KRETPROBE_TRACE_ENTRY(tp->nr_args);
 	size = ALIGN(__size + sizeof(u32), sizeof(u64));
 	size -= sizeof(u32);
+	if (WARN_ONCE(size > FTRACE_MAX_PROFILE_SIZE,
+		     "profile buffer not large enough"))
+		return 0;
 
-	do {
-		char raw_data[size];
-		struct trace_entry *ent;
+	/*
+	 * Protect the non nmi buffer
+	 * This also protects the rcu read side
+	 */
+	local_irq_save(irq_flags);
+	__cpu = smp_processor_id();
 
+	if (in_nmi())
+		raw_data = rcu_dereference(trace_profile_buf_nmi);
+	else
+		raw_data = rcu_dereference(trace_profile_buf);
+
+	if (!raw_data)
+		goto end;
+
+	raw_data = per_cpu_ptr(raw_data, __cpu);
+	/* Zero dead bytes from alignment to avoid buffer leak to userspace */
 	*(u64 *)(&raw_data[size - sizeof(u64)]) = 0ULL;
 	entry = (struct kretprobe_trace_entry *)raw_data;
 	ent = &entry->ent;
@@ -1214,7 +1244,8 @@ static __kprobes int kretprobe_profile_func(struct kretprobe_instance *ri,
 	for (i = 0; i < tp->nr_args; i++)
 		entry->args[i] = call_fetch(&tp->args[i].fetch, regs);
 	perf_tp_event(call->id, entry->ret_ip, 1, entry, size);
-	} while (0);
+end:
+	local_irq_restore(irq_flags);
 	return 0;
 }