Re: [PATCH] kfence: Avoid stalling work queue task without allocations

[Steven Rostedt <rostedt@goodmis.org>]

On Thu, 19 Nov 2020 13:53:57 +0100
Marco Elver <elver@google.com> wrote:

> Running tests again, along with the function tracer
> Running tests on all trace events:
> Testing all events: 
> BUG: workqueue lockup - pool cpus=0 node=0 flags=0x0 nice=0 stuck for 12s!

The below patch might be noisy, but can you add it to the kernel that
crashes and see if a particular event causes the issue?

[ note I didn't even compile test. I hope it works ;) ]

Perhaps run it a couple of times to see if it crashes on the same set of
events each time.

-- Steve

diff --git a/kernel/trace/trace_events.c b/kernel/trace/trace_events.c
index 98d194d8460e..eb1dd9cf77a9 100644
--- a/kernel/trace/trace_events.c
+++ b/kernel/trace/trace_events.c
@@ -773,6 +773,8 @@ static void remove_event_file_dir(struct trace_event_file *file)
 	kmem_cache_free(file_cachep, file);
 }
 
+static int spam;
+
 /*
  * __ftrace_set_clr_event(NULL, NULL, NULL, set) will set/unset all events.
  */
@@ -808,6 +810,8 @@ __ftrace_set_clr_event_nolock(struct trace_array *tr, const char *match,
 		if (event && strcmp(event, name) != 0)
 			continue;
 
+		if (spam)
+			printk("%s event %s\n", set ? "enabling" : "disabling", name);
 		ret = ftrace_event_enable_disable(file, set);
 
 		/*
@@ -3647,6 +3651,7 @@ static __init void event_trace_self_tests(void)
 	pr_info("Running tests on all trace events:\n");
 	pr_info("Testing all events: ");
 
+	spam = 1;
 	ret = __ftrace_set_clr_event(tr, NULL, NULL, NULL, 1);
 	if (WARN_ON_ONCE(ret)) {
 		pr_warn("error enabling all events\n");
@@ -3661,6 +3666,7 @@ static __init void event_trace_self_tests(void)
 		pr_warn("error disabling all events\n");
 		return;
 	}
+	spam = 0;
 
 	pr_cont("OK\n");
 }