[PATCH] fsl-edma: tracing: no ptr dereference during log output

[PATCH] fsl-edma: tracing: no ptr dereference during log output

[Martin Kaiser]

The fsl edma events store a pointer to a struct fsl_edma_engine in the
ringbuffer and dereference it when a log entry is printed. At this time,
the pointer may no longer be valid.

Event injection can be used to trigger a crash:

$ cd /sys/kernel/tracing
$ echo 'value = 0' > events/fsl_edma/edma_writeb/inject
$ cat trace

The log output needs only edma->membase. Add a membase field at the end
of the event and use the new field for log output. Keep the existing
fields for backward compatibility.

Fixes: 11102d0c343b ("dmaengine: fsl-edma: add trace event support")
Signed-off-by: Martin Kaiser <martin@kaiser.cx>
---
 drivers/dma/fsl-edma-trace.h | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/dma/fsl-edma-trace.h b/drivers/dma/fsl-edma-trace.h
index d3541301a247..45d964a3726d 100644
--- a/drivers/dma/fsl-edma-trace.h
+++ b/drivers/dma/fsl-edma-trace.h
@@ -19,14 +19,16 @@ DECLARE_EVENT_CLASS(edma_log_io,
 		__field(struct fsl_edma_engine *, edma)
 		__field(void __iomem *, addr)
 		__field(u32, value)
+		__field(void __iomem *, membase)
 	),
 	TP_fast_assign(
 		__entry->edma = edma;
 		__entry->addr = addr;
 		__entry->value = value;
+		__entry->membase = edma->membase;
 	),
 	TP_printk("offset %08x: value %08x",
-		(u32)(__entry->addr - __entry->edma->membase), __entry->value)
+		(u32)(__entry->addr - __entry->membase), __entry->value)
 );
 
 DEFINE_EVENT(edma_log_io, edma_readl,
-- 
2.43.7

Re: [PATCH] fsl-edma: tracing: no ptr dereference during log output

[Steven Rostedt]

On Tue, 30 Jun 2026 22:00:11 +0200
Martin Kaiser <martin@kaiser.cx> wrote:

> The fsl edma events store a pointer to a struct fsl_edma_engine in the
> ringbuffer and dereference it when a log entry is printed. At this time,
> the pointer may no longer be valid.

Nice catch.

> 
> Event injection can be used to trigger a crash:
> 
> $ cd /sys/kernel/tracing
> $ echo 'value = 0' > events/fsl_edma/edma_writeb/inject
> $ cat trace
> 
> The log output needs only edma->membase. Add a membase field at the end
> of the event and use the new field for log output. Keep the existing
> fields for backward compatibility.
> 

Cc: stable@vger.kernel.org

> Fixes: 11102d0c343b ("dmaengine: fsl-edma: add trace event support")
> Signed-off-by: Martin Kaiser <martin@kaiser.cx>

Reviewed-by: Steven Rostedt <rostedt@goodmis.org>

> ---
>  drivers/dma/fsl-edma-trace.h | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/dma/fsl-edma-trace.h b/drivers/dma/fsl-edma-trace.h
> index d3541301a247..45d964a3726d 100644
> --- a/drivers/dma/fsl-edma-trace.h
> +++ b/drivers/dma/fsl-edma-trace.h
> @@ -19,14 +19,16 @@ DECLARE_EVENT_CLASS(edma_log_io,
>  		__field(struct fsl_edma_engine *, edma)
>  		__field(void __iomem *, addr)
>  		__field(u32, value)
> +		__field(void __iomem *, membase)
>  	),
>  	TP_fast_assign(
>  		__entry->edma = edma;
>  		__entry->addr = addr;
>  		__entry->value = value;
> +		__entry->membase = edma->membase;
>  	),
>  	TP_printk("offset %08x: value %08x",
> -		(u32)(__entry->addr - __entry->edma->membase), __entry->value)
> +		(u32)(__entry->addr - __entry->membase), __entry->value)

Hmm, I think I should update the TP_printk checks at boot to cover this too.

-- Steve

>  );
>  
>  DEFINE_EVENT(edma_log_io, edma_readl,

Re: [PATCH] fsl-edma: tracing: no ptr dereference during log output

[Steven Rostedt]

On Tue, 30 Jun 2026 16:05:44 -0400
Steven Rostedt <rostedt@goodmis.org> wrote:

> >  	TP_printk("offset %08x: value %08x",
> > -		(u32)(__entry->addr - __entry->edma->membase), __entry->value)
> > +		(u32)(__entry->addr - __entry->membase), __entry->value)  
> 
> Hmm, I think I should update the TP_printk checks at boot to cover this too.

I created the following to catch this:

diff --git a/kernel/trace/trace_events.c b/kernel/trace/trace_events.c
index c46e623e7e0d..2da3c02bea54 100644
--- a/kernel/trace/trace_events.c
+++ b/kernel/trace/trace_events.c
@@ -400,10 +400,37 @@ static bool process_string(const char *fmt, int len, struct trace_event_call *ca
 	return true;
 }
 
+static void test_double_dereference(const char *str, int len,
+				    struct trace_event_call *call)
+{
+	const char *ptr;
+	const char *end = str + len;
+
+	ptr = strstr(str, "REC->");
+
+	while (ptr && ptr < end) {
+
+		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);
+				return;
+			}
+			if (!isalnum(*ptr) && *ptr != '_')
+				break;
+		}
+
+		ptr = strstr(ptr, "REC->");
+	}
+}
+
 static void handle_dereference_arg(const char *arg_str, u64 string_flags, int len,
 				   u64 *dereference_flags, int arg,
 				   struct trace_event_call *call)
 {
+	test_double_dereference(arg_str, len, call);
+
 	if (string_flags & (1ULL << arg)) {
 		if (process_string(arg_str, len, call))
 			*dereference_flags &= ~(1ULL << arg);

Enabled this event to see if it would trigger, but instead it found *another* BUG!

[    0.719012][    T0] ------------[ cut here ]------------
[    0.720850][    T0] Event ufshcd_exception_event has double dereference in TP_printk: dev_name(REC->hba->dev), REC->status
[    0.724646][    T0] WARNING: kernel/trace/trace_events.c:416 at handle_dereference_arg+0x342/0x5a0, CPU#0: swapper/0/0


I'll go make a fix for the ufshcd_exception_event event, and then I will
definitely add this patch to make sure this bug isn't in other places.

-- Steve

Re: [PATCH] fsl-edma: tracing: no ptr dereference during log output

[Frank Li]

On Tue, Jun 30, 2026 at 10:00:11PM +0200, Martin Kaiser wrote:

subject need a tag to identify subsystem

dmaengine: fsl-edma: tracing: remove ptr dereference during log output

>
> The fsl edma events store a pointer to a struct fsl_edma_engine in the
> ringbuffer and dereference it when a log entry is printed. At this time,
> the pointer may no longer be valid.
>
> Event injection can be used to trigger a crash:
>
> $ cd /sys/kernel/tracing
> $ echo 'value = 0' > events/fsl_edma/edma_writeb/inject
> $ cat trace
>
> The log output needs only edma->membase. Add a membase field at the end
> of the event and use the new field for log output. Keep the existing
> fields for backward compatibility.

thanks you for your fix.

Frank
>
> Fixes: 11102d0c343b ("dmaengine: fsl-edma: add trace event support")
> Signed-off-by: Martin Kaiser <martin@kaiser.cx>
> ---
>  drivers/dma/fsl-edma-trace.h | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/dma/fsl-edma-trace.h b/drivers/dma/fsl-edma-trace.h
> index d3541301a247..45d964a3726d 100644
> --- a/drivers/dma/fsl-edma-trace.h
> +++ b/drivers/dma/fsl-edma-trace.h
> @@ -19,14 +19,16 @@ DECLARE_EVENT_CLASS(edma_log_io,
>                 __field(struct fsl_edma_engine *, edma)
>                 __field(void __iomem *, addr)
>                 __field(u32, value)
> +               __field(void __iomem *, membase)
>         ),
>         TP_fast_assign(
>                 __entry->edma = edma;
>                 __entry->addr = addr;
>                 __entry->value = value;
> +               __entry->membase = edma->membase;
>         ),
>         TP_printk("offset %08x: value %08x",
> -               (u32)(__entry->addr - __entry->edma->membase), __entry->value)
> +               (u32)(__entry->addr - __entry->membase), __entry->value)
>  );
>
>  DEFINE_EVENT(edma_log_io, edma_readl,
> --
> 2.43.7
>
>