linux-kernel - [PATCH v2] tracing: Optimize event type allocation with IDA

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-ID: <20221110020319.1259291-1-zhengyejian1@huawei.com>
Date:   Thu, 10 Nov 2022 10:03:19 +0800
From:   Zheng Yejian <zhengyejian1@...wei.com>
To:     <rostedt@...dmis.org>, <mhiramat@...nel.org>
CC:     <linux-kernel@...r.kernel.org>, <bpf@...r.kernel.org>,
        <zhengyejian1@...wei.com>
Subject: [PATCH v2] tracing: Optimize event type allocation with IDA

After commit 060fa5c83e67 ("tracing/events: reuse trace event ids after
 overflow"), trace events with dynamic type are linked up in list
'ftrace_event_list' through field 'trace_event.list'. Then when max
event type number used up, it's possible to reuse type number of some
freed one by traversing 'ftrace_event_list'.

As instead, using IDA to manage available type numbers can make codes
simpler and then the field 'trace_event.list' can be dropped.

Since 'struct trace_event' is used in static tracepoints, drop
'trace_event.list' can make vmlinux smaller. Local test with about 2000
tracepoints, vmlinux reduced about 64KB:
  before：-rwxrwxr-x 1 root root 76669448 Nov  8 17:14 vmlinux
  after： -rwxrwxr-x 1 root root 76604176 Nov  8 17:15 vmlinux

Signed-off-by: Zheng Yejian <zhengyejian1@...wei.com>
---
 include/linux/trace_events.h |  1 -
 kernel/trace/trace_output.c  | 66 +++++++++---------------------------
 2 files changed, 16 insertions(+), 51 deletions(-)

Changes since v1:
  - Explicitly include linux/idr.h as suggested by Masami Hiramatsu
    Link: https://lore.kernel.org/lkml/20221109222650.ce6c22e231345f6852f6956f@kernel.org/#t

diff --git a/include/linux/trace_events.h b/include/linux/trace_events.h
index 20749bd9db71..bb2053246d6a 100644
--- a/include/linux/trace_events.h
+++ b/include/linux/trace_events.h
@@ -136,7 +136,6 @@ struct trace_event_functions {
 
 struct trace_event {
 	struct hlist_node		node;
-	struct list_head		list;
 	int				type;
 	struct trace_event_functions	*funcs;
 };
diff --git a/kernel/trace/trace_output.c b/kernel/trace/trace_output.c
index 67f47ea27921..f0ba97121345 100644
--- a/kernel/trace/trace_output.c
+++ b/kernel/trace/trace_output.c
@@ -11,6 +11,7 @@
 #include <linux/kprobes.h>
 #include <linux/sched/clock.h>
 #include <linux/sched/mm.h>
+#include <linux/idr.h>
 
 #include "trace_output.h"
 
@@ -21,8 +22,6 @@ DECLARE_RWSEM(trace_event_sem);
 
 static struct hlist_head event_hash[EVENT_HASHSIZE] __read_mostly;
 
-static int next_event_type = __TRACE_LAST_TYPE;
-
 enum print_line_t trace_print_bputs_msg_only(struct trace_iterator *iter)
 {
 	struct trace_seq *s = &iter->seq;
@@ -688,38 +687,23 @@ struct trace_event *ftrace_find_event(int type)
 	return NULL;
 }
 
-static LIST_HEAD(ftrace_event_list);
+static DEFINE_IDA(trace_event_ida);
 
-static int trace_search_list(struct list_head **list)
+static void free_trace_event_type(int type)
 {
-	struct trace_event *e = NULL, *iter;
-	int next = __TRACE_LAST_TYPE;
-
-	if (list_empty(&ftrace_event_list)) {
-		*list = &ftrace_event_list;
-		return next;
-	}
+	if (type >= __TRACE_LAST_TYPE)
+		ida_free(&trace_event_ida, type);
+}
 
-	/*
-	 * We used up all possible max events,
-	 * lets see if somebody freed one.
-	 */
-	list_for_each_entry(iter, &ftrace_event_list, list) {
-		if (iter->type != next) {
-			e = iter;
-			break;
-		}
-		next++;
-	}
+static int alloc_trace_event_type(void)
+{
+	int next;
 
-	/* Did we used up all 65 thousand events??? */
-	if (next > TRACE_EVENT_TYPE_MAX)
+	/* Skip static defined type numbers */
+	next = ida_alloc_range(&trace_event_ida, __TRACE_LAST_TYPE,
+			       TRACE_EVENT_TYPE_MAX, GFP_KERNEL);
+	if (next < 0)
 		return 0;
-
-	if (e)
-		*list = &e->list;
-	else
-		*list = &ftrace_event_list;
 	return next;
 }
 
@@ -761,28 +745,10 @@ int register_trace_event(struct trace_event *event)
 	if (WARN_ON(!event->funcs))
 		goto out;
 
-	INIT_LIST_HEAD(&event->list);
-
 	if (!event->type) {
-		struct list_head *list = NULL;
-
-		if (next_event_type > TRACE_EVENT_TYPE_MAX) {
-
-			event->type = trace_search_list(&list);
-			if (!event->type)
-				goto out;
-
-		} else {
-
-			event->type = next_event_type++;
-			list = &ftrace_event_list;
-		}
-
-		if (WARN_ON(ftrace_find_event(event->type)))
+		event->type = alloc_trace_event_type();
+		if (!event->type)
 			goto out;
-
-		list_add_tail(&event->list, list);
-
 	} else if (WARN(event->type > __TRACE_LAST_TYPE,
 			"Need to add type to trace.h")) {
 		goto out;
@@ -819,7 +785,7 @@ EXPORT_SYMBOL_GPL(register_trace_event);
 int __unregister_trace_event(struct trace_event *event)
 {
 	hlist_del(&event->node);
-	list_del(&event->list);
+	free_trace_event_type(event->type);
 	return 0;
 }
 
-- 
2.25.1