On Sat, 14 Jul 2018 01:28:15 +0900
Masami Hiramatsu <mhiramat@xxxxxxxxxx> wrote:
> Inherit the tracing on/off setting on ring_buffer to next
> trace buffer when taking a snapshot.
>
> Taking a snapshot is done by swapping with backup ring buffer
> (max_tr_buffer). But since the tracing on/off setting is set
> in the ring buffer, when swapping it, tracing on/off setting
> can also be changed. This causes a strange result like below;
>
> /sys/kernel/debug/tracing # cat tracing_on
> 1
> /sys/kernel/debug/tracing # echo 0 > tracing_on
> /sys/kernel/debug/tracing # echo 1 > snapshot
> /sys/kernel/debug/tracing # cat tracing_on
> 1
> /sys/kernel/debug/tracing # echo 1 > snapshot
> /sys/kernel/debug/tracing # cat tracing_on
> 0
>
> We don't touch tracing_on, but snapshot changes tracing_on
> setting each time. This must be a bug, because user never know
> that each "ring_buffer" stores tracing-enable state and
> snapshot is done by swapping ring buffers.
>
> This patch fixes above strange behavior.
>
OK, so the issue is that the tracing_on state is saved with the buffer
itself. And when we swap the buffer, we also swap the state.
> Fixes: commit debdd57f5145 ("tracing: Make a snapshot feature available from userspace")
> Signed-off-by: Masami Hiramatsu <mhiramat@xxxxxxxxxx>
> Cc: Steven Rostedt <rostedt@xxxxxxxxxxx>
> Cc: Ingo Molnar <mingo@xxxxxxxxxx>
> Cc: Hiraku Toyooka <hiraku.toyooka@xxxxxxxxxxxxxxxx>
> Cc: stable@xxxxxxxxxxxxxxx
> ---
> include/linux/ring_buffer.h | 1 +
> kernel/trace/ring_buffer.c | 12 ++++++++++++
> kernel/trace/trace.c | 6 ++++++
> 3 files changed, 19 insertions(+)
>
> diff --git a/include/linux/ring_buffer.h b/include/linux/ring_buffer.h
> index b72ebdff0b77..003d09ab308d 100644
> --- a/include/linux/ring_buffer.h
> +++ b/include/linux/ring_buffer.h
> @@ -165,6 +165,7 @@ void ring_buffer_record_enable(struct ring_buffer *buffer);
> void ring_buffer_record_off(struct ring_buffer *buffer);
> void ring_buffer_record_on(struct ring_buffer *buffer);
> int ring_buffer_record_is_on(struct ring_buffer *buffer);
> +int ring_buffer_record_is_set_on(struct ring_buffer *buffer);
> void ring_buffer_record_disable_cpu(struct ring_buffer *buffer, int cpu);
> void ring_buffer_record_enable_cpu(struct ring_buffer *buffer, int cpu);
>
> diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c
> index 6a46af21765c..4038ed74ab95 100644
> --- a/kernel/trace/ring_buffer.c
> +++ b/kernel/trace/ring_buffer.c
> @@ -3227,6 +3227,18 @@ int ring_buffer_record_is_on(struct ring_buffer *buffer)
> }
>
> /**
> + * ring_buffer_record_is_set_on - return true if the ring buffer is set writable
> + * @buffer: The ring buffer to see if write is set enabled
> + *
> + * Returns true if the ring buffer is set writable by ring_buffer_record_on().
> + * Note that this does NOT mean it is in a writable state.
I'd add a little more context here. Something like:
* It may return true when the ring buffer has been disabled by
* ring_buffer_record_disable(), as that is a temporary disabling of
* the ring buffer.
-- Steve
> + */
> +int ring_buffer_record_is_set_on(struct ring_buffer *buffer)
> +{
> + return !(atomic_read(&buffer->record_disabled) & RB_BUFFER_OFF);
> +}
> +
> +/**
> * ring_buffer_record_disable_cpu - stop all writes into the cpu_buffer
> * @buffer: The ring buffer to stop writes to.
> * @cpu: The CPU buffer to stop
> diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
> index 2556d8c097d2..bbd5a94a7ef1 100644
> --- a/kernel/trace/trace.c
> +++ b/kernel/trace/trace.c
> @@ -1378,6 +1378,12 @@ update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu)
>
> arch_spin_lock(&tr->max_lock);
>
> + /* Inherit the recordable setting from trace_buffer */
> + if (ring_buffer_record_is_set_on(tr->trace_buffer.buffer))
> + ring_buffer_record_on(tr->max_buffer.buffer);
> + else
> + ring_buffer_record_off(tr->max_buffer.buffer);
> +
> swap(tr->trace_buffer.buffer, tr->max_buffer.buffer);
>
> __update_max_tr(tr, tsk, cpu);
