tracing: Fix snapshot counter going between two tracers that use it
authorSteven Rostedt (Google) <rostedt@goodmis.org>
Fri, 23 Feb 2024 01:33:24 +0000 (20:33 -0500)
committerSteven Rostedt (Google) <rostedt@goodmis.org>
Mon, 18 Mar 2024 14:33:05 +0000 (10:33 -0400)
commitcca990c7b565af0dc61a8f647c00833453cf5bff
tree4a0af7f7b91086e09ff48c5fe7e7790751bf673f
parent70a6ed553f7d3504febac467cb4a0bae621ba3c6
tracing: Fix snapshot counter going between two tracers that use it

Running the ftrace selftests caused the ring buffer mapping test to fail.
Investigating, I found that the snapshot counter would be incremented
every time a tracer that uses the snapshot is enabled even if the snapshot
was used by the previous tracer.

That is:

 # cd /sys/kernel/tracing
 # echo wakeup_rt > current_tracer
 # echo wakeup_dl > current_tracer
 # echo nop > current_tracer

would leave the snapshot counter at 1 and not zero. That's because the
enabling of wakeup_dl would increment the counter again but the setting
the tracer to nop would only decrement it once.

Do not arm the snapshot for a tracer if the previous tracer already had it
armed.

Link: https://lore.kernel.org/linux-trace-kernel/20240223013344.570525723@goodmis.org
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Vincent Donnefort <vdonnefort@google.com>
Fixes: 16f7e48ffc53a ("tracing: Add snapshot refcount")
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
kernel/trace/trace.c