Capture inlined iseq's cc entries in root iseq's

jit_unit to avoid marking wrong cc entries when inlined iseq is compiled
multiple times, resolving the TODO added by daf7c48d88.

This obviates pseudo jit_unit in inlined iseq introduced by 7ec2359374
and fixes memory leak of the adhoc unit.

mjit.c

@@ -26,7 +26,6 @@
 #include "internal/warnings.h"
 
 #include "mjit_worker.c"
-#include "vm_callinfo.h"
 
 // Copy ISeq's states so that race condition does not happen on compilation.
 static void
@@ -53,18 +52,6 @@ mjit_copy_job_handler(void *data)
     }
 
     const struct rb_iseq_constant_body *body = job->iseq->body;
-    const unsigned int ci_size = body->ci_size;
-    if (ci_size > 0) {
-        VM_ASSERT(body->jit_unit != NULL);
-        VM_ASSERT(body->jit_unit->cc_entries != NULL);
-
-        const struct rb_callcache **cc_entries = body->jit_unit->cc_entries;
-
-        for (unsigned int i=0; i<ci_size; i++) {
-            cc_entries[i] = body->call_data[i].cc;
-        }
-    }
-
     if (job->is_entries) {
         memcpy(job->is_entries, body->is_entries, sizeof(union iseq_inline_storage_entry) * body->is_size);
     }
@@ -293,9 +280,6 @@ create_unit(const rb_iseq_t *iseq)
 
     unit->id = current_unit_num++;
     unit->iseq = (rb_iseq_t *)iseq;
-    if (iseq->body->ci_size > 0) {
-        unit->cc_entries = ZALLOC_N(const struct rb_callcache *, iseq->body->ci_size);
-    }
     iseq->body->jit_unit = unit;
 }
 

mjit_compile.c

@@ -32,12 +32,6 @@
 #define NOT_COMPILED_STACK_SIZE -1
 #define ALREADY_COMPILED_P(status, pos) (status->stack_size_for_pos[pos] != NOT_COMPILED_STACK_SIZE)
 
-static size_t
-call_data_index(CALL_DATA cd, const struct rb_iseq_constant_body *body)
-{
-    return cd - body->call_data;
-}
-
 // For propagating information needed for lazily pushing a frame.
 struct inlined_call_context {
     int orig_argc; // ci->orig_argc
@@ -55,8 +49,12 @@ struct compile_status {
     // If true, JIT-ed code will use local variables to store pushed values instead of
     // using VM's stack and moving stack pointer.
     bool local_stack_p;
-    // Safely-accessible cache entries copied from main thread.
+    // Safely-accessible ivar cache entries copied from main thread.
     union iseq_inline_storage_entry *is_entries;
+    // Safely-accessible call cache entries captured to compiled_iseq to be marked on GC
+    const struct rb_callcache **cc_entries;
+    // A pointer to root (i.e. not inlined) iseq being compiled.
+    const struct rb_iseq_constant_body *compiled_iseq;
     // Mutated optimization levels
     struct rb_mjit_compile_info *compile_info;
     // If `inlined_iseqs[pos]` is not NULL, `mjit_compile_body` tries to inline ISeq there.
@@ -78,6 +76,12 @@ struct case_dispatch_var {
     VALUE last_value;
 };
 
+static size_t
+call_data_index(CALL_DATA cd, const struct rb_iseq_constant_body *body)
+{
+    return cd - body->call_data;
+}
+
 // Returns true if call cache is still not obsoleted and vm_cc_cme(cc)->def->type is available.
 static bool
 has_valid_method_type(CALL_CACHE cc)
@@ -273,6 +277,9 @@ compile_cancel_handler(FILE *f, const struct rb_iseq_constant_body *body, struct
     fprintf(f, "    return Qundef;\n");
 }
 
+extern const struct rb_callcache **
+mjit_capture_cc_entries(const struct rb_iseq_constant_body *compiled_iseq, const struct rb_iseq_constant_body *captured_iseq);
+
 extern bool mjit_copy_cache_from_main_thread(const rb_iseq_t *iseq,
                                              union iseq_inline_storage_entry *is_entries);
 
@@ -368,6 +375,9 @@ inlinable_iseq_p(const struct rb_iseq_constant_body *body)
             alloca(sizeof(const struct rb_iseq_constant_body *) * body->iseq_size) : NULL, \
         .is_entries = (body->is_size > 0) ? \
             alloca(sizeof(union iseq_inline_storage_entry) * body->is_size) : NULL, \
+        .cc_entries = (body->ci_size > 0) ? \
+            mjit_capture_cc_entries(status.compiled_iseq, body) : NULL, \
+        .compiled_iseq = status.compiled_iseq, \
         .compile_info = compile_root_p ? \
             rb_mjit_iseq_compile_info(body) : alloca(sizeof(struct rb_mjit_compile_info)) \
     }; \
@@ -393,7 +403,7 @@ precompile_inlinable_iseqs(FILE *f, const rb_iseq_t *iseq, struct compile_status
         if (insn == BIN(opt_send_without_block)) { // `compile_inlined_cancel_handler` supports only `opt_send_without_block`
             CALL_DATA cd = (CALL_DATA)body->iseq_encoded[pos + 1];
             const struct rb_callinfo *ci = cd->ci;
-            const struct rb_callcache *cc = mjit_iseq_cc_entries(iseq->body)[call_data_index(cd, body)]; // use copy to avoid race condition
+            const struct rb_callcache *cc = status->cc_entries[call_data_index(cd, body)]; // use copy to avoid race condition
 
             const rb_iseq_t *child_iseq;
             if (has_valid_method_type(cc) &&
@@ -411,7 +421,7 @@ precompile_inlinable_iseqs(FILE *f, const rb_iseq_t *iseq, struct compile_status
                             RSTRING_PTR(child_iseq->body->location.label),
                             RSTRING_PTR(rb_iseq_path(child_iseq)), FIX2INT(child_iseq->body->location.first_lineno));
 
-                struct compile_status child_status;
+                struct compile_status child_status = { .compiled_iseq = status->compiled_iseq };
                 INIT_COMPILE_STATUS(child_status, child_iseq->body, false);
                 child_status.inline_context = (struct inlined_call_context){
                     .orig_argc = vm_ci_argc(ci),
@@ -419,9 +429,10 @@ precompile_inlinable_iseqs(FILE *f, const rb_iseq_t *iseq, struct compile_status
                     .param_size = child_iseq->body->param.size,
                     .local_size = child_iseq->body->local_table_size
                 };
-                if ((child_iseq->body->ci_size > 0 || child_status.is_entries != NULL)
+                if ((child_iseq->body->ci_size > 0 && child_status.cc_entries == NULL)
-                        && !mjit_copy_cache_from_main_thread(child_iseq, child_status.is_entries))
+                    || (child_status.is_entries != NULL && !mjit_copy_cache_from_main_thread(child_iseq, child_status.is_entries))) {
                     return false;
+                }
 
                 fprintf(f, "ALWAYS_INLINE(static VALUE _mjit_inlined_%d(rb_execution_context_t *ec, rb_control_frame_t *reg_cfp, const VALUE orig_self, const rb_iseq_t *original_iseq));\n", pos);
                 fprintf(f, "static inline VALUE\n_mjit_inlined_%d(rb_execution_context_t *ec, rb_control_frame_t *reg_cfp, const VALUE orig_self, const rb_iseq_t *original_iseq)\n{\n", pos);
@@ -449,10 +460,10 @@ mjit_compile(FILE *f, const rb_iseq_t *iseq, const char *funcname)
         fprintf(f, "#define OPT_CHECKED_RUN 0\n\n");
     }
 
-    struct compile_status status;
+    struct compile_status status = { .compiled_iseq = iseq->body };
     INIT_COMPILE_STATUS(status, iseq->body, true);
-    if ((iseq->body->ci_size > 0 || status.is_entries != NULL)
+    if ((iseq->body->ci_size > 0 && status.cc_entries == NULL)
-        && !mjit_copy_cache_from_main_thread(iseq, status.is_entries)) {
+        || (status.is_entries != NULL && !mjit_copy_cache_from_main_thread(iseq, status.is_entries))) {
         return false;
     }
 

mjit_worker.c

@@ -86,6 +86,7 @@
 #endif
 
 #include "vm_core.h"
+#include "vm_callinfo.h"
 #include "mjit.h"
 #include "gc.h"
 #include "ruby_assert.h"
@@ -160,7 +161,8 @@ struct rb_mjit_unit {
     // mjit_compile's optimization switches
     struct rb_mjit_compile_info compile_info;
     // captured CC values, they should be marked with iseq.
-    const struct rb_callcache **cc_entries; // size: iseq->body->ci_size
+    const struct rb_callcache **cc_entries;
+    unsigned int cc_entries_size; // iseq->body->ci_size + ones of inlined iseqs
 };
 
 // Linked list of struct rb_mjit_unit.
@@ -1149,6 +1151,40 @@ static void mjit_copy_job_handler(void *data);
 // vm_trace.c
 int rb_workqueue_register(unsigned flags, rb_postponed_job_func_t , void *);
 
+// Capture cc entries of `captured_iseq` and append them to `compiled_iseq->jit_unit->cc_entries`.
+// This is needed when `captured_iseq` is inlined by `compiled_iseq` and GC needs to mark inlined cc.
+//
+// This assumes that it's safe to reference cc without acquiring GVL.
+const struct rb_callcache **
+mjit_capture_cc_entries(const struct rb_iseq_constant_body *compiled_iseq, const struct rb_iseq_constant_body *captured_iseq)
+{
+    struct rb_mjit_unit *unit = compiled_iseq->jit_unit;
+    unsigned int new_entries_size = unit->cc_entries_size + captured_iseq->ci_size;
+    VM_ASSERT(captured_iseq->ci_size > 0);
+
+    // Allocate new cc_entries and append them to unit->cc_entries
+    const struct rb_callcache **cc_entries;
+    if (unit->cc_entries_size == 0) {
+        VM_ASSERT(unit->cc_entries == NULL);
+        unit->cc_entries = cc_entries = malloc(sizeof(struct rb_callcache *) * new_entries_size);
+        if (cc_entries == NULL) return NULL;
+    }
+    else {
+        cc_entries = realloc(unit->cc_entries, sizeof(struct rb_callcache *) * new_entries_size);
+        if (cc_entries == NULL) return NULL;
+        unit->cc_entries = cc_entries;
+        cc_entries += unit->cc_entries_size;
+    }
+    unit->cc_entries_size = new_entries_size;
+
+    // Capture cc to cc_enties
+    for (unsigned int i = 0; i < captured_iseq->ci_size; i++) {
+        cc_entries[i] = captured_iseq->call_data[i].cc;
+    }
+
+    return cc_entries;
+}
+
 // Copy inline cache values of `iseq` to `cc_entries` and `is_entries`.
 // These buffers should be pre-allocated properly prior to calling this function.
 // Return true if copy succeeds or is not needed.
@@ -1176,25 +1212,6 @@ mjit_copy_cache_from_main_thread(const rb_iseq_t *iseq, union iseq_inline_storag
     CRITICAL_SECTION_FINISH(3, "in mjit_copy_cache_from_main_thread");
 
     if (UNLIKELY(mjit_opts.wait)) {
-        // setup pseudo jit_unit
-        //
-        // Usually jit_unit is created in `rb_mjit_add_iseq_to_process`.
-        // However, this copy job can be used for inlined ISeqs too, and
-        // inlined ISeq doesn't have a jit_unit.
-        // TODO: Manage the cc in outer ISeq's jit_unit.
-        if (iseq->body->jit_unit == NULL) {
-            // This function is invoked in mjit worker thread, so GC should not be invoked.
-            // To prevent GC with xmalloc(), use malloc() directly here.
-            // However, mixing xmalloc() and malloc() will cause another issue.
-            // TODO: fix this allocation code.
-            iseq->body->jit_unit = (struct rb_mjit_unit *)malloc(sizeof(struct rb_mjit_unit));
-            if (iseq->body->jit_unit == NULL) rb_fatal("malloc failed");
-            if (iseq->body->ci_size > 0) {
-                iseq->body->jit_unit->cc_entries =
-                  (const struct rb_callcache **)calloc(iseq->body->ci_size, sizeof(const struct rb_callcache *));
-                if (iseq->body->jit_unit->cc_entries == NULL) rb_fatal("malloc failed");
-            }
-        }
         mjit_copy_job_handler((void *)job);
     }
     else if (rb_workqueue_register(0, mjit_copy_job_handler, (void *)job)) {

tool/ruby_vm/views/_mjit_compile_send.erb

@@ -14,7 +14,7 @@
     MAYBE_UNUSED(<%= ope.fetch(:decl) %>) = (<%= ope.fetch(:type) %>)operands[<%= i %>];
 % end
 % # compiler: Use copied cc to avoid race condition
-    const struct rb_callcache *captured_cc = mjit_iseq_cc_entries(body)[call_data_index(cd, body)];
+    const struct rb_callcache *captured_cc = status->cc_entries[call_data_index(cd, body)];
 %
     if (!status->compile_info->disable_send_cache && has_valid_method_type(captured_cc)) {
         const rb_iseq_t *iseq;

tool/ruby_vm/views/mjit_compile.inc.erb

@@ -57,7 +57,7 @@ switch (insn) {
 %   when *send_compatible_opt_insns
 %     # To avoid cancel, just emit `opt_send_without_block` instead of `opt_*` insn if call cache is populated.
 %     cd_index = insn.opes.index { |o| o.fetch(:type) == 'CALL_DATA' }
-      if (has_valid_method_type(mjit_iseq_cc_entries(body)[call_data_index((CALL_DATA)operands[<%= cd_index %>], body)])) {
+      if (has_valid_method_type(status->cc_entries[call_data_index((CALL_DATA)operands[<%= cd_index %>], body)])) {
 <%=       render 'mjit_compile_send', locals: { insn: opt_send_without_block } -%>
 <%=       render 'mjit_compile_insn', locals: { insn: opt_send_without_block } -%>
           break;