[PATCH tracing/kprobes 1/3] tracing/kprobes: Use global event profile buffer in kprobe tracer

[Masami Hiramatsu <mhiramat at redhat dot com>]

Use new percpu global event profile buffer instead of stack in kprobe
tracer.

Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
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: Frederic Weisbecker <fweisbec@gmail.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Ingo Molnar <mingo@elte.hu>
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>
---

 kernel/trace/trace_kprobe.c |  115 +++++++++++++++++++++++++++----------------
 1 files changed, 73 insertions(+), 42 deletions(-)

diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c
index 09cba27..97309d4 100644
--- a/kernel/trace/trace_kprobe.c
+++ b/kernel/trace/trace_kprobe.c
@@ -1149,35 +1149,49 @@ 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
-		 */
-		*(u64 *)(&raw_data[size - sizeof(u64)]) = 0ULL;
-		entry = (struct kprobe_trace_entry *)raw_data;
-		ent = &entry->ent;
-
-		tracing_generic_entry_update(ent, irq_flags, pc);
-		ent->type = call->id;
-		entry->nargs = tp->nr_args;
-		entry->ip = (unsigned long)kp->addr;
-		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);
+	/*
+	 * 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;
+
+	tracing_generic_entry_update(ent, irq_flags, pc);
+	ent->type = call->id;
+	entry->nargs = tp->nr_args;
+	entry->ip = (unsigned long)kp->addr;
+	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);
+end:
+	local_irq_restore(irq_flags);
 	return 0;
 }
 
@@ -1188,33 +1202,50 @@ 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;
+
+	/*
+	 * 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;
 
-	do {
-		char raw_data[size];
-		struct trace_entry *ent;
-
-		*(u64 *)(&raw_data[size - sizeof(u64)]) = 0ULL;
-		entry = (struct kretprobe_trace_entry *)raw_data;
-		ent = &entry->ent;
-
-		tracing_generic_entry_update(ent, irq_flags, pc);
-		ent->type = call->id;
-		entry->nargs = tp->nr_args;
-		entry->func = (unsigned long)tp->rp.kp.addr;
-		entry->ret_ip = (unsigned long)ri->ret_addr;
-		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);
+	tracing_generic_entry_update(ent, irq_flags, pc);
+	ent->type = call->id;
+	entry->nargs = tp->nr_args;
+	entry->func = (unsigned long)tp->rp.kp.addr;
+	entry->ret_ip = (unsigned long)ri->ret_addr;
+	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);
+end:
+	local_irq_restore(irq_flags);
 	return 0;
 }
 


-- 
Masami Hiramatsu

Software Engineer
Hitachi Computer Products (America), Inc.
Software Solutions Division

e-mail: mhiramat@redhat.com