The Linux Kernel Mailing List
 help / color / mirror / Atom feed
* [PATCH] tracing: Report TP_printk double dereference with pr_warn()
@ 2026-08-06  5:07 David Carlier
  2026-08-06 21:41 ` Steven Rostedt
  0 siblings, 1 reply; 3+ messages in thread
From: David Carlier @ 2026-08-06  5:07 UTC (permalink / raw)
  To: Steven Rostedt
  Cc: Masami Hiramatsu, Mathieu Desnoyers, linux-trace-kernel,
	linux-kernel, David Carlier

test_double_dereference() uses WARN_ONCE(), which is per-call-site. Only
the first offending event in the kernel is ever reported, and the tree
still has six: ice_{rx,tx}_dim_template, two hfi1 txq events, mtu3_ep and
edma_log_io. Whichever registers first hides the rest, and it taints the
kernel for what is a source-level mistake.

Use pr_warn() instead, matching the "TRACE EVENT ERROR:" report that
handle_dereference_arg() already emits for the same class of problem.
All offenders are now listed on one boot.

Fixes: b5cc230af5e5 ("tracing: Warn when an event dereferences a pointer in TP_printk()")
Signed-off-by: David Carlier <devnexen@gmail.com>
---
 kernel/trace/trace_events.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/trace/trace_events.c b/kernel/trace/trace_events.c
index a8590d2394e3..53f32752edc0 100644
--- a/kernel/trace/trace_events.c
+++ b/kernel/trace/trace_events.c
@@ -415,8 +415,8 @@ static void test_double_dereference(const char *str, int len,
 		ptr += 5;
 		for (; ptr < end; ptr++) {
 			if (ptr[0] == '-' && ptr[1] == '>') {
-				WARN_ONCE(1, "Event %s has double dereference in TP_printk: %.*s\n",
-					  trace_event_name(call), len, str);
+				pr_warn("TRACE EVENT ERROR: Event %s has double dereference in TP_printk: %.*s\n",
+					trace_event_name(call), len, str);
 				return;
 			}
 			if (!isalnum(*ptr) && *ptr != '_')
-- 
2.55.0


^ permalink raw reply related	[flat|nested] 3+ messages in thread

* Re: [PATCH] tracing: Report TP_printk double dereference with pr_warn()
  2026-08-06  5:07 [PATCH] tracing: Report TP_printk double dereference with pr_warn() David Carlier
@ 2026-08-06 21:41 ` Steven Rostedt
  2026-08-06 21:52   ` [PATCH v2] tracing: Report every TP_printk double dereference David Carlier
  0 siblings, 1 reply; 3+ messages in thread
From: Steven Rostedt @ 2026-08-06 21:41 UTC (permalink / raw)
  To: David Carlier
  Cc: Masami Hiramatsu, Mathieu Desnoyers, linux-trace-kernel,
	linux-kernel

On Thu,  6 Aug 2026 06:07:16 +0100
David Carlier <devnexen@gmail.com> wrote:

> test_double_dereference() uses WARN_ONCE(), which is per-call-site. Only
> the first offending event in the kernel is ever reported, and the tree
> still has six: ice_{rx,tx}_dim_template, two hfi1 txq events, mtu3_ep and
> edma_log_io. Whichever registers first hides the rest, and it taints the
> kernel for what is a source-level mistake.
> 
> Use pr_warn() instead, matching the "TRACE EVENT ERROR:" report that
> handle_dereference_arg() already emits for the same class of problem.
> All offenders are now listed on one boot.

NAK!

> 
> Fixes: b5cc230af5e5 ("tracing: Warn when an event dereferences a pointer in TP_printk()")

This is by no way a fix!

I want a big splat when this is triggered. It is a bug and needs to be
fixed immediately. pr_warn() does not make tests fail. I want this to fail
tests. I want this to panic when panic_on_warn is set.

> Signed-off-by: David Carlier <devnexen@gmail.com>
> ---
>  kernel/trace/trace_events.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/kernel/trace/trace_events.c b/kernel/trace/trace_events.c
> index a8590d2394e3..53f32752edc0 100644
> --- a/kernel/trace/trace_events.c
> +++ b/kernel/trace/trace_events.c
> @@ -415,8 +415,8 @@ static void test_double_dereference(const char *str, int len,
>  		ptr += 5;
>  		for (; ptr < end; ptr++) {
>  			if (ptr[0] == '-' && ptr[1] == '>') {
> -				WARN_ONCE(1, "Event %s has double dereference in TP_printk: %.*s\n",
> -					  trace_event_name(call), len, str);
> +				pr_warn("TRACE EVENT ERROR: Event %s has double dereference in TP_printk: %.*s\n",
> +					trace_event_name(call), len, str);

Now I am OK with adding a pr_warn() with the WARN_ONCE() but the
WARN_ONCE() must stay. But leave off the fixes tag.

-- Steve


>  				return;
>  			}
>  			if (!isalnum(*ptr) && *ptr != '_')


^ permalink raw reply	[flat|nested] 3+ messages in thread

* [PATCH v2] tracing: Report every TP_printk double dereference
  2026-08-06 21:41 ` Steven Rostedt
@ 2026-08-06 21:52   ` David Carlier
  0 siblings, 0 replies; 3+ messages in thread
From: David Carlier @ 2026-08-06 21:52 UTC (permalink / raw)
  To: Steven Rostedt
  Cc: Masami Hiramatsu, Mathieu Desnoyers, linux-trace-kernel,
	linux-kernel, David Carlier

WARN_ONCE() splats once per call site, so only the first offending event
registered is ever reported. The tree currently has six:
ice_{rx,tx}_dim_template, two hfi1 txq events, mtu3_ep and edma_log_io.
Whichever registers first hides the rest, and each has to be found again
on the next boot.

Add a pr_warn() next to the WARN_ONCE() so every offender is listed, the
same way test_event_printk() already pairs WARN_ON_ONCE() with pr_warn()
for unsafe %p* dereferences. The WARN_ONCE() stays so the condition still
fails tests and panics under panic_on_warn.

Suggested-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: David Carlier <devnexen@gmail.com>
---
 kernel/trace/trace_events.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/kernel/trace/trace_events.c b/kernel/trace/trace_events.c
index a8590d2394e3..ce902482ec7c 100644
--- a/kernel/trace/trace_events.c
+++ b/kernel/trace/trace_events.c
@@ -415,6 +415,8 @@ static void test_double_dereference(const char *str, int len,
 		ptr += 5;
 		for (; ptr < end; ptr++) {
 			if (ptr[0] == '-' && ptr[1] == '>') {
+				pr_warn("TRACE EVENT ERROR: Event %s has double dereference in TP_printk: %.*s\n",
+					trace_event_name(call), len, str);
 				WARN_ONCE(1, "Event %s has double dereference in TP_printk: %.*s\n",
 					  trace_event_name(call), len, str);
 				return;
-- 
2.55.0


^ permalink raw reply related	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2026-08-06 21:53 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-08-06  5:07 [PATCH] tracing: Report TP_printk double dereference with pr_warn() David Carlier
2026-08-06 21:41 ` Steven Rostedt
2026-08-06 21:52   ` [PATCH v2] tracing: Report every TP_printk double dereference David Carlier

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox