plugin-gen: add module for TCG-related code

We first inject empty instrumentation from translator_loop.
After translation, we go through the plugins to see what
they want to register for, filling in the empty instrumentation.
If if turns out that some instrumentation remains unused, we
remove it.

This approach supports the following features:

- Inlining TCG code for simple operations. Note that we do not
  export TCG ops to plugins. Instead, we give them a C API to
  insert inlined ops. So far we only support adding an immediate
  to a u64, e.g. to count events.

- "Direct" callbacks. These are callbacks that do not go via
  a helper. Instead, the helper is defined at run-time, so that
  the plugin code is directly called from TCG. This makes direct
  callbacks as efficient as possible; they are therefore used
  for very frequent events, e.g. memory callbacks.

- Passing the host address to memory callbacks. Most of this
  is implemented in a later patch though.

- Instrumentation of memory accesses performed from helpers.
  See the corresponding comment, as well as a later patch.

Signed-off-by: Emilio G. Cota <cota@braap.org>
[AJB: add alloc_tcg_plugin_context, use glib, rm hwaddr]
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

4 files changed, 56 insertions, 0 deletions

diff --git a/tcg/tcg-op.h b/tcg/tcg-op.h
index 7c778f9..4af272d 100644
--- a/tcg/tcg-op.h
+++ b/tcg/tcg-op.h
@@ -833,6 +833,17 @@ void tcg_gen_goto_tb(unsigned idx);
  */
 void tcg_gen_lookup_and_goto_ptr(void);
 
+static inline void tcg_gen_plugin_cb_start(unsigned from, unsigned type,
+                                           unsigned wr)
+{
+    tcg_gen_op3(INDEX_op_plugin_cb_start, from, type, wr);
+}
+
+static inline void tcg_gen_plugin_cb_end(void)
+{
+    tcg_emit_op(INDEX_op_plugin_cb_end);
+}
+
 #if TARGET_LONG_BITS == 32
 #define tcg_temp_new() tcg_temp_new_i32()
 #define tcg_global_reg_new tcg_global_reg_new_i32
diff --git a/tcg/tcg-opc.h b/tcg/tcg-opc.h
index 242d608..9288a04 100644
--- a/tcg/tcg-opc.h
+++ b/tcg/tcg-opc.h
@@ -198,6 +198,9 @@ DEF(goto_tb, 0, 0, 1, TCG_OPF_BB_EXIT | TCG_OPF_BB_END)
 DEF(goto_ptr, 0, 1, 0,
     TCG_OPF_BB_EXIT | TCG_OPF_BB_END | IMPL(TCG_TARGET_HAS_goto_ptr))
 
+DEF(plugin_cb_start, 0, 0, 3, TCG_OPF_NOT_PRESENT)
+DEF(plugin_cb_end, 0, 0, 0, TCG_OPF_NOT_PRESENT)
+
 DEF(qemu_ld_i32, 1, TLADDR_ARGS, 1,
     TCG_OPF_CALL_CLOBBER | TCG_OPF_SIDE_EFFECTS)
 DEF(qemu_st_i32, 0, TLADDR_ARGS + 1, 1,
diff --git a/tcg/tcg.c b/tcg/tcg.c
index 16b2d0e..5475d49 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -736,6 +736,15 @@ void tcg_region_init(void)
 #endif
 }
 
+static void alloc_tcg_plugin_context(TCGContext *s)
+{
+#ifdef CONFIG_PLUGIN
+    s->plugin_tb = g_new0(struct qemu_plugin_tb, 1);
+    s->plugin_tb->insns =
+        g_ptr_array_new_with_free_func(qemu_plugin_insn_cleanup_fn);
+#endif
+}
+
 /*
  * All TCG threads except the parent (i.e. the one that called tcg_context_init
  * and registered the target's TCG globals) must register with this function
@@ -780,6 +789,10 @@ void tcg_register_thread(void)
     g_assert(n < ms->smp.max_cpus);
     atomic_set(&tcg_ctxs[n], s);
 
+    if (n > 0) {
+        alloc_tcg_plugin_context(s);
+    }
+
     tcg_ctx = s;
     qemu_mutex_lock(&region.lock);
     err = tcg_region_initial_alloc__locked(tcg_ctx);
@@ -976,6 +989,8 @@ void tcg_context_init(TCGContext *s)
         indirect_reg_alloc_order[i] = tcg_target_reg_alloc_order[i];
     }
 
+    alloc_tcg_plugin_context(s);
+
     tcg_ctx = s;
     /*
      * In user-mode we simply share the init context among threads, since we
@@ -1681,6 +1696,13 @@ void tcg_gen_callN(void *func, TCGTemp *ret, int nargs, TCGTemp **args)
     flags = info->flags;
     sizemask = info->sizemask;
 
+#ifdef CONFIG_PLUGIN
+    /* detect non-plugin helpers */
+    if (tcg_ctx->plugin_insn && unlikely(strncmp(info->name, "plugin_", 7))) {
+        tcg_ctx->plugin_insn->calls_helpers = true;
+    }
+#endif
+
 #if defined(__sparc__) && !defined(__arch64__) \
     && !defined(CONFIG_TCG_INTERPRETER)
     /* We have 64-bit values in one register, but need to pass as two
diff --git a/tcg/tcg.h b/tcg/tcg.h
index a37181c..d234b4c 100644
--- a/tcg/tcg.h
+++ b/tcg/tcg.h
@@ -538,6 +538,9 @@ typedef struct TCGOp {
 
     /* Next and previous opcodes.  */
     QTAILQ_ENTRY(TCGOp) link;
+#ifdef CONFIG_PLUGIN
+    QSIMPLEQ_ENTRY(TCGOp) plugin_link;
+#endif
 
     /* Arguments for the opcode.  */
     TCGArg args[MAX_OPC_PARAM];
@@ -639,6 +642,23 @@ struct TCGContext {
 
     TCGLabel *exitreq_label;
 
+#ifdef CONFIG_PLUGIN
+    /*
+     * We keep one plugin_tb struct per TCGContext. Note that on every TB
+     * translation we clear but do not free its contents; this way we
+     * avoid a lot of malloc/free churn, since after a few TB's it's
+     * unlikely that we'll need to allocate either more instructions or more
+     * space for instructions (for variable-instruction-length ISAs).
+     */
+    struct qemu_plugin_tb *plugin_tb;
+
+    /* descriptor of the instruction being translated */
+    struct qemu_plugin_insn *plugin_insn;
+
+    /* list to quickly access the injected ops */
+    QSIMPLEQ_HEAD(, TCGOp) plugin_ops;
+#endif
+
     TCGTempSet free_temps[TCG_TYPE_COUNT * 2];
     TCGTemp temps[TCG_MAX_TEMPS]; /* globals first, temps after */