The testing show that perf_ftrace_function_call() are using smp_processor_id() with preemption enabled, all the checking on CPU could be wrong after preemption. As Peter point out, the section between ftrace_test_recursion_trylock/unlock() pair require the preemption to be disabled as 'Documentation/trace/ftrace-uses.rst' explained, but currently the work is done outside of the helpers. And since the internal using of trace_test_and_set_recursion() and trace_clear_recursion() also require preemption to be disabled, we can just merge the logical together. Patch 1/2 will make sure preemption disabled when recursion lock succeed, patch 2/2 will do smp_processor_id() checking after trylock() to address the issue. v1: https://lore.kernel.org/all/8c7de46d-9869-aa5e-2bb9-5dbc2eda395e@xxxxxxxxxxxxxxxxx/ v2: https://lore.kernel.org/all/b1d7fe43-ce84-0ed7-32f7-ea1d12d0b716@xxxxxxxxxxxxxxxxx/ v3: https://lore.kernel.org/all/609b565a-ed6e-a1da-f025-166691b5d994@xxxxxxxxxxxxxxxxx/ Michael Wang (2): ftrace: disable preemption when recursion locked ftrace: do CPU checking after preemption disabled arch/csky/kernel/probes/ftrace.c | 2 -- arch/parisc/kernel/ftrace.c | 2 -- arch/powerpc/kernel/kprobes-ftrace.c | 2 -- arch/riscv/kernel/probes/ftrace.c | 2 -- arch/x86/kernel/kprobes/ftrace.c | 2 -- include/linux/trace_recursion.h | 20 +++++++++++++++++++- kernel/livepatch/patch.c | 13 +++++++------ kernel/trace/ftrace.c | 15 +++++---------- kernel/trace/trace_event_perf.c | 6 +++--- kernel/trace/trace_functions.c | 5 ----- 10 files changed, 34 insertions(+), 35 deletions(-) -- 1.8.3.1
