BPF Archive on lore.kernel.org
 help / color / Atom feed
From: Masami Hiramatsu <mhiramat@kernel.org>
To: Brendan Gregg <brendan.d.gregg@gmail.com>,
	Steven Rostedt <rostedt@goodmis.org>,
	Alexei Starovoitov <ast@kernel.org>
Cc: mhiramat@kernel.org, Ingo Molnar <mingo@kernel.org>,
	bpf@vger.kernel.org, linux-kernel@vger.kernel.org,
	Daniel Borkmann <daniel@iogearbox.net>,
	Arnaldo Carvalho de Melo <acme@kernel.org>,
	"David S . Miller" <davem@davemloft.net>,
	paulmck@kernel.org, joel@joelfernandes.org,
	"Naveen N . Rao" <naveen.n.rao@linux.ibm.com>,
	Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com>
Subject: [RFT PATCH 05/13] tracing/kprobe: Use call_rcu to defer freeing event_file_link
Date: Thu, 16 Jan 2020 23:45:02 +0900
Message-ID: <157918590192.29301.6909688694265698678.stgit@devnote2> (raw)
In-Reply-To: <157918584866.29301.6941815715391411338.stgit@devnote2>

Use call_rcu() to defer freeing event_file_link data
structure. This removes RCU synchronization from
per-probe event disabling operation.

Since unregistering kprobe event requires all handlers to
be disabled and have finished, this also introduces a
gatekeeper to ensure that. If there is any disabled event
which is not finished, the unregister process can synchronize
RCU once (IOW, may sleep a while.)

Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
---
 kernel/trace/trace_kprobe.c |   35 +++++++++++++++++++++++++++++------
 kernel/trace/trace_probe.c  |   10 ++++++++--
 kernel/trace/trace_probe.h  |    1 +
 3 files changed, 38 insertions(+), 8 deletions(-)

diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c
index cbdc4f4e64c7..906af1ffe2b2 100644
--- a/kernel/trace/trace_kprobe.c
+++ b/kernel/trace/trace_kprobe.c
@@ -328,10 +328,25 @@ static inline int __enable_trace_kprobe(struct trace_kprobe *tk)
 	return ret;
 }
 
+atomic_t trace_kprobe_disabled_finished;
+
+static void trace_kprobe_disabled_handlers_finish(void)
+{
+	if (atomic_read(&trace_kprobe_disabled_finished))
+		synchronize_rcu();
+}
+
+static void trace_kprobe_disabled_finish_cb(struct rcu_head *head)
+{
+	atomic_dec(&trace_kprobe_disabled_finished);
+	kfree(head);
+}
+
 static void __disable_trace_kprobe(struct trace_probe *tp)
 {
 	struct trace_probe *pos;
 	struct trace_kprobe *tk;
+	struct rcu_head *head;
 
 	list_for_each_entry(pos, trace_probe_probe_list(tp), list) {
 		tk = container_of(pos, struct trace_kprobe, tp);
@@ -342,6 +357,13 @@ static void __disable_trace_kprobe(struct trace_probe *tp)
 		else
 			disable_kprobe(&tk->rp.kp);
 	}
+
+	/* Handler exit gatekeeper */
+	head = kzalloc(sizeof(*head), GFP_KERNEL);
+	if (WARN_ON(!head))
+		return;
+	atomic_inc(&trace_kprobe_disabled_finished);
+	call_rcu(head, trace_kprobe_disabled_finish_cb);
 }
 
 /*
@@ -422,13 +444,11 @@ static int disable_trace_kprobe(struct trace_event_call *call,
 
  out:
 	if (file)
-		/*
-		 * Synchronization is done in below function. For perf event,
-		 * file == NULL and perf_trace_event_unreg() calls
-		 * tracepoint_synchronize_unregister() to ensure synchronize
-		 * event. We don't need to care about it.
-		 */
 		trace_probe_remove_file(tp, file);
+	/*
+	 * We have no RCU synchronization here. Caller must wait for the
+	 * completion of disabling.
+	 */
 
 	return 0;
 }
@@ -542,6 +562,9 @@ static int unregister_trace_kprobe(struct trace_kprobe *tk)
 	if (trace_probe_is_enabled(&tk->tp))
 		return -EBUSY;
 
+	/* Make sure all disabled trace_kprobe handlers finished */
+	trace_kprobe_disabled_handlers_finish();
+
 	/* Will fail if probe is being used by ftrace or perf */
 	if (unregister_kprobe_event(tk))
 		return -EBUSY;
diff --git a/kernel/trace/trace_probe.c b/kernel/trace/trace_probe.c
index 905b10af5d5c..b18df8e1b2d6 100644
--- a/kernel/trace/trace_probe.c
+++ b/kernel/trace/trace_probe.c
@@ -1067,6 +1067,13 @@ struct event_file_link *trace_probe_get_file_link(struct trace_probe *tp,
 	return NULL;
 }
 
+static void event_file_link_free_cb(struct rcu_head *head)
+{
+	struct event_file_link *link = container_of(head, typeof(*link), rcu);
+
+	kfree(link);
+}
+
 int trace_probe_remove_file(struct trace_probe *tp,
 			    struct trace_event_file *file)
 {
@@ -1077,8 +1084,7 @@ int trace_probe_remove_file(struct trace_probe *tp,
 		return -ENOENT;
 
 	list_del_rcu(&link->list);
-	synchronize_rcu();
-	kfree(link);
+	call_rcu(&link->rcu, event_file_link_free_cb);
 
 	if (list_empty(&tp->event->files))
 		trace_probe_clear_flag(tp, TP_FLAG_TRACE);
diff --git a/kernel/trace/trace_probe.h b/kernel/trace/trace_probe.h
index 4ee703728aec..71ac01a50815 100644
--- a/kernel/trace/trace_probe.h
+++ b/kernel/trace/trace_probe.h
@@ -243,6 +243,7 @@ struct trace_probe {
 struct event_file_link {
 	struct trace_event_file		*file;
 	struct list_head		list;
+	struct rcu_head			rcu;
 };
 
 static inline bool trace_probe_test_flag(struct trace_probe *tp,


  parent reply index

Thread overview: 26+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-01-16 14:44 [RFT PATCH 00/13] tracing: kprobes: Introduce async unregistration Masami Hiramatsu
2020-01-16 14:44 ` [RFT PATCH 01/13] kprobes: Fix to protect kick_kprobe_optimizer() by kprobe_mutex Masami Hiramatsu
2020-01-16 14:44 ` [RFT PATCH 02/13] kprobes: Remove redundant arch_disarm_kprobe() call Masami Hiramatsu
2020-01-16 14:44 ` [RFT PATCH 03/13] kprobes: Postpone optimizer until a bunch of probes (un)registered Masami Hiramatsu
2020-01-16 14:44 ` [RFT PATCH 04/13] kprobes: Make optimizer delay to 1 second Masami Hiramatsu
2020-01-22  0:29   ` Steven Rostedt
2020-01-22  7:23     ` Masami Hiramatsu
2020-01-22 12:11       ` Steven Rostedt
2020-01-22 13:12         ` Masami Hiramatsu
2020-01-22 16:54           ` Paul E. McKenney
2020-01-23  1:33             ` Masami Hiramatsu
2020-01-23  2:26               ` Paul E. McKenney
2020-01-23  6:13                 ` Masami Hiramatsu
2020-01-23 10:03                   ` Paul E. McKenney
2020-01-16 14:45 ` Masami Hiramatsu [this message]
2020-01-27 15:02   ` [RFT PATCH 05/13] tracing/kprobe: Use call_rcu to defer freeing event_file_link kbuild test robot
2020-01-27 15:02   ` [RFC PATCH] tracing/kprobe: trace_kprobe_disabled_finished can be static kbuild test robot
2020-01-28 15:02     ` Masami Hiramatsu
2020-01-16 14:45 ` [RFT PATCH 06/13] kprobes: Enable kprobe-booster with CONFIG_PREEMPT=y Masami Hiramatsu
2020-01-16 14:45 ` [RFT PATCH 07/13] kprobes: Use normal list traversal API if a mutex is held Masami Hiramatsu
2020-01-16 14:45 ` [RFT PATCH 08/13] kprobes: Use workqueue for reclaiming kprobe insn cache pages Masami Hiramatsu
2020-01-16 14:45 ` [RFT PATCH 09/13] kprobes: Free kprobe_insn_page asynchronously Masami Hiramatsu
2020-01-16 14:45 ` [RFT PATCH 10/13] kprobes: Make free_*insn_slot() mutex-less Masami Hiramatsu
2020-01-16 14:46 ` [RFT PATCH 11/13] kprobes: Add asynchronous unregistration APIs Masami Hiramatsu
2020-01-16 14:46 ` [RFT PATCH 12/13] tracing/kprobe: Free probe event asynchronously Masami Hiramatsu
2020-01-16 14:46 ` [RFT PATCH 13/13] tracing/kprobe: perf_event: Remove local kprobe " Masami Hiramatsu

Reply instructions:

You may reply publically to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=157918590192.29301.6909688694265698678.stgit@devnote2 \
    --to=mhiramat@kernel.org \
    --cc=acme@kernel.org \
    --cc=anil.s.keshavamurthy@intel.com \
    --cc=ast@kernel.org \
    --cc=bpf@vger.kernel.org \
    --cc=brendan.d.gregg@gmail.com \
    --cc=daniel@iogearbox.net \
    --cc=davem@davemloft.net \
    --cc=joel@joelfernandes.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mingo@kernel.org \
    --cc=naveen.n.rao@linux.ibm.com \
    --cc=paulmck@kernel.org \
    --cc=rostedt@goodmis.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

BPF Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/bpf/0 bpf/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 bpf bpf/ https://lore.kernel.org/bpf \
		bpf@vger.kernel.org
	public-inbox-index bpf

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.bpf


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git