MJIT: Cancel all on disastrous situations (#7019)

I noticed this while running test_yjit with --mjit-call-threshold=1, 
which redefines `Integer#<`. When Ruby is monkey-patched, 
MJIT itself could be broken.

Similarly, Ruby scripts could break MJIT in many different ways. I
prepared the same set of hooks as YJIT so that we could possibly
override it and disable it on those moments. Every constant under
RubyVM::MJIT is private and thus it's an unsupported behavior though.

1 files changed, 32 insertions, 0 deletions

diff --git a/lib/ruby_vm/mjit/hooks.rb b/lib/ruby_vm/mjit/hooks.rb
new file mode 100644
index 0000000000..3fb1004111
--- /dev/null
+++ b/lib/ruby_vm/mjit/hooks.rb
@@ -0,0 +1,32 @@
+module RubyVM::MJIT::Hooks # :nodoc: all
+  C = RubyVM::MJIT.const_get(:C, false)
+
+  def self.on_bop_redefined(_redefined_flag, _bop)
+    C.mjit_cancel_all("BOP is redefined")
+  end
+
+  def self.on_cme_invalidate(_cme)
+    # to be used later
+  end
+
+  def self.on_ractor_spawn
+    C.mjit_cancel_all("Ractor is spawned")
+  end
+
+  def self.on_constant_state_changed(_id)
+    # to be used later
+  end
+
+  def self.on_constant_ic_update(_iseq, _ic, _insn_idx)
+    # to be used later
+  end
+
+  def self.on_tracing_invalidate_all(new_iseq_events)
+    # Stop calling all JIT-ed code. We can't rewrite existing JIT-ed code to trace_ insns for now.
+    # :class events are triggered only in ISEQ_TYPE_CLASS, but mjit_target_iseq_p ignores such iseqs.
+    # Thus we don't need to cancel JIT-ed code for :class events.
+    if new_iseq_events != C.RUBY_EVENT_CLASS
+      C.mjit_cancel_all("TracePoint is enabled")
+    end
+  end
+end