ftrace: Fix recursive locking direct_mutex in ftrace_modify_direct_caller
authorSong Liu <song@kernel.org>
Tue, 27 Sep 2022 00:41:46 +0000 (17:41 -0700)
committerSteven Rostedt (Google) <rostedt@goodmis.org>
Tue, 27 Sep 2022 18:48:27 +0000 (14:48 -0400)
Naveen reported recursive locking of direct_mutex with sample
ftrace-direct-modify.ko:

[   74.762406] WARNING: possible recursive locking detected
[   74.762887] 6.0.0-rc6+ #33 Not tainted
[   74.763216] --------------------------------------------
[   74.763672] event-sample-fn/1084 is trying to acquire lock:
[   74.764152] ffffffff86c9d6b0 (direct_mutex){+.+.}-{3:3}, at: \
    register_ftrace_function+0x1f/0x180
[   74.764922]
[   74.764922] but task is already holding lock:
[   74.765421] ffffffff86c9d6b0 (direct_mutex){+.+.}-{3:3}, at: \
    modify_ftrace_direct+0x34/0x1f0
[   74.766142]
[   74.766142] other info that might help us debug this:
[   74.766701]  Possible unsafe locking scenario:
[   74.766701]
[   74.767216]        CPU0
[   74.767437]        ----
[   74.767656]   lock(direct_mutex);
[   74.767952]   lock(direct_mutex);
[   74.768245]
[   74.768245]  *** DEADLOCK ***
[   74.768245]
[   74.768750]  May be due to missing lock nesting notation
[   74.768750]
[   74.769332] 1 lock held by event-sample-fn/1084:
[   74.769731]  #0: ffffffff86c9d6b0 (direct_mutex){+.+.}-{3:3}, at: \
    modify_ftrace_direct+0x34/0x1f0
[   74.770496]
[   74.770496] stack backtrace:
[   74.770884] CPU: 4 PID: 1084 Comm: event-sample-fn Not tainted ...
[   74.771498] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), ...
[   74.772474] Call Trace:
[   74.772696]  <TASK>
[   74.772896]  dump_stack_lvl+0x44/0x5b
[   74.773223]  __lock_acquire.cold.74+0xac/0x2b7
[   74.773616]  lock_acquire+0xd2/0x310
[   74.773936]  ? register_ftrace_function+0x1f/0x180
[   74.774357]  ? lock_is_held_type+0xd8/0x130
[   74.774744]  ? my_tramp2+0x11/0x11 [ftrace_direct_modify]
[   74.775213]  __mutex_lock+0x99/0x1010
[   74.775536]  ? register_ftrace_function+0x1f/0x180
[   74.775954]  ? slab_free_freelist_hook.isra.43+0x115/0x160
[   74.776424]  ? ftrace_set_hash+0x195/0x220
[   74.776779]  ? register_ftrace_function+0x1f/0x180
[   74.777194]  ? kfree+0x3e1/0x440
[   74.777482]  ? my_tramp2+0x11/0x11 [ftrace_direct_modify]
[   74.777941]  ? __schedule+0xb40/0xb40
[   74.778258]  ? register_ftrace_function+0x1f/0x180
[   74.778672]  ? my_tramp1+0xf/0xf [ftrace_direct_modify]
[   74.779128]  register_ftrace_function+0x1f/0x180
[   74.779527]  ? ftrace_set_filter_ip+0x33/0x70
[   74.779910]  ? __schedule+0xb40/0xb40
[   74.780231]  ? my_tramp1+0xf/0xf [ftrace_direct_modify]
[   74.780678]  ? my_tramp2+0x11/0x11 [ftrace_direct_modify]
[   74.781147]  ftrace_modify_direct_caller+0x5b/0x90
[   74.781563]  ? 0xffffffffa0201000
[   74.781859]  ? my_tramp1+0xf/0xf [ftrace_direct_modify]
[   74.782309]  modify_ftrace_direct+0x1b2/0x1f0
[   74.782690]  ? __schedule+0xb40/0xb40
[   74.783014]  ? simple_thread+0x2a/0xb0 [ftrace_direct_modify]
[   74.783508]  ? __schedule+0xb40/0xb40
[   74.783832]  ? my_tramp2+0x11/0x11 [ftrace_direct_modify]
[   74.784294]  simple_thread+0x76/0xb0 [ftrace_direct_modify]
[   74.784766]  kthread+0xf5/0x120
[   74.785052]  ? kthread_complete_and_exit+0x20/0x20
[   74.785464]  ret_from_fork+0x22/0x30
[   74.785781]  </TASK>

Fix this by using register_ftrace_function_nolock in
ftrace_modify_direct_caller.

Link: https://lkml.kernel.org/r/20220927004146.1215303-1-song@kernel.org
Fixes: 53cd885bc5c3 ("ftrace: Allow IPMODIFY and DIRECT ops on the same function")
Reported-and-tested-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Signed-off-by: Song Liu <song@kernel.org>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
kernel/trace/ftrace.c

index 5a1ec7e..406d059 100644 (file)
@@ -5427,6 +5427,8 @@ static struct ftrace_ops stub_ops = {
  * it is safe to modify the ftrace record, where it should be
  * currently calling @old_addr directly, to call @new_addr.
  *
+ * This is called with direct_mutex locked.
+ *
  * Safety checks should be made to make sure that the code at
  * @rec->ip is currently calling @old_addr. And this must
  * also update entry->direct to @new_addr.
@@ -5439,6 +5441,8 @@ int __weak ftrace_modify_direct_caller(struct ftrace_func_entry *entry,
        unsigned long ip = rec->ip;
        int ret;
 
+       lockdep_assert_held(&direct_mutex);
+
        /*
         * The ftrace_lock was used to determine if the record
         * had more than one registered user to it. If it did,
@@ -5461,7 +5465,7 @@ int __weak ftrace_modify_direct_caller(struct ftrace_func_entry *entry,
        if (ret)
                goto out_lock;
 
-       ret = register_ftrace_function(&stub_ops);
+       ret = register_ftrace_function_nolock(&stub_ops);
        if (ret) {
                ftrace_set_filter_ip(&stub_ops, ip, 1, 0);
                goto out_lock;