plugins: scoreboard API

We introduce a cpu local storage, automatically managed (and extended)
by QEMU itself. Plugin allocate a scoreboard, and don't have to deal
with how many cpus are launched.

This API will be used by new inline functions but callbacks can benefit
from this as well. This way, they can operate without a global lock for
simple operations.

At any point during execution, any scoreboard will be dimensioned with
at least qemu_plugin_num_vcpus entries.

New functions:
- qemu_plugin_scoreboard_find
- qemu_plugin_scoreboard_free
- qemu_plugin_scoreboard_new

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Message-Id: <20240304130036.124418-2-pierrick.bouvier@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20240305121005.3528075-15-alex.bennee@linaro.org>

plugins/api.c

@@ -465,3 +465,22 @@ int qemu_plugin_read_register(struct qemu_plugin_register *reg, GByteArray *buf)
 
     return gdb_read_register(current_cpu, buf, GPOINTER_TO_INT(reg));
 }
+
+struct qemu_plugin_scoreboard *qemu_plugin_scoreboard_new(size_t element_size)
+{
+    return plugin_scoreboard_new(element_size);
+}
+
+void qemu_plugin_scoreboard_free(struct qemu_plugin_scoreboard *score)
+{
+    plugin_scoreboard_free(score);
+}
+
+void *qemu_plugin_scoreboard_find(struct qemu_plugin_scoreboard *score,
+                                  unsigned int vcpu_index)
+{
+    g_assert(vcpu_index < qemu_plugin_num_vcpus());
+    /* we can't use g_array_index since entry size is not statically known */
+    char *base_ptr = score->data->data;
+    return base_ptr + vcpu_index * g_array_get_element_size(score->data);
+}

plugins/core.c

@@ -18,6 +18,7 @@
 #include "qemu/lockable.h"
 #include "qemu/option.h"
 #include "qemu/plugin.h"
+#include "qemu/queue.h"
 #include "qemu/rcu_queue.h"
 #include "qemu/xxhash.h"
 #include "qemu/rcu.h"
@@ -215,6 +216,35 @@ CPUPluginState *qemu_plugin_create_vcpu_state(void)
     return g_new0(CPUPluginState, 1);
 }
 
+static void plugin_grow_scoreboards__locked(CPUState *cpu)
+{
+    if (cpu->cpu_index < plugin.scoreboard_alloc_size) {
+        return;
+    }
+
+    bool need_realloc = FALSE;
+    while (cpu->cpu_index >= plugin.scoreboard_alloc_size) {
+        plugin.scoreboard_alloc_size *= 2;
+        need_realloc = TRUE;
+    }
+
+
+    if (!need_realloc || QLIST_EMPTY(&plugin.scoreboards)) {
+        /* nothing to do, we just updated sizes for future scoreboards */
+        return;
+    }
+
+    /* cpus must be stopped, as tb might still use an existing scoreboard. */
+    start_exclusive();
+    struct qemu_plugin_scoreboard *score;
+    QLIST_FOREACH(score, &plugin.scoreboards, entry) {
+        g_array_set_size(score->data, plugin.scoreboard_alloc_size);
+    }
+    /* force all tb to be flushed, as scoreboard pointers were changed. */
+    tb_flush(cpu);
+    end_exclusive();
+}
+
 void qemu_plugin_vcpu_init_hook(CPUState *cpu)
 {
     bool success;
@@ -225,6 +255,7 @@ void qemu_plugin_vcpu_init_hook(CPUState *cpu)
     success = g_hash_table_insert(plugin.cpu_ht, &cpu->cpu_index,
                                   &cpu->cpu_index);
     g_assert(success);
+    plugin_grow_scoreboards__locked(cpu);
     qemu_rec_mutex_unlock(&plugin.lock);
 
     plugin_vcpu_cb__simple(cpu, QEMU_PLUGIN_EV_VCPU_INIT);
@@ -578,6 +609,8 @@ static void __attribute__((__constructor__)) plugin_init(void)
     qemu_rec_mutex_init(&plugin.lock);
     plugin.id_ht = g_hash_table_new(g_int64_hash, g_int64_equal);
     plugin.cpu_ht = g_hash_table_new(g_int_hash, g_int_equal);
+    QLIST_INIT(&plugin.scoreboards);
+    plugin.scoreboard_alloc_size = 16; /* avoid frequent reallocation */
     QTAILQ_INIT(&plugin.ctxs);
     qht_init(&plugin.dyn_cb_arr_ht, plugin_dyn_cb_arr_cmp, 16,
              QHT_MODE_AUTO_RESIZE);
@@ -588,3 +621,27 @@ int plugin_num_vcpus(void)
 {
     return plugin.num_vcpus;
 }
+
+struct qemu_plugin_scoreboard *plugin_scoreboard_new(size_t element_size)
+{
+    struct qemu_plugin_scoreboard *score =
+        g_malloc0(sizeof(struct qemu_plugin_scoreboard));
+    score->data = g_array_new(FALSE, TRUE, element_size);
+    g_array_set_size(score->data, plugin.scoreboard_alloc_size);
+
+    qemu_rec_mutex_lock(&plugin.lock);
+    QLIST_INSERT_HEAD(&plugin.scoreboards, score, entry);
+    qemu_rec_mutex_unlock(&plugin.lock);
+
+    return score;
+}
+
+void plugin_scoreboard_free(struct qemu_plugin_scoreboard *score)
+{
+    qemu_rec_mutex_lock(&plugin.lock);
+    QLIST_REMOVE(score, entry);
+    qemu_rec_mutex_unlock(&plugin.lock);
+
+    g_array_free(score->data, TRUE);
+    g_free(score);
+}

plugins/plugin.h

@@ -31,6 +31,8 @@ struct qemu_plugin_state {
      * but with the HT we avoid adding a field to CPUState.
      */
     GHashTable *cpu_ht;
+    QLIST_HEAD(, qemu_plugin_scoreboard) scoreboards;
+    size_t scoreboard_alloc_size;
     DECLARE_BITMAP(mask, QEMU_PLUGIN_EV_MAX);
     /*
      * @lock protects the struct as well as ctx->uninstalling.
@@ -101,4 +103,8 @@ void exec_inline_op(struct qemu_plugin_dyn_cb *cb);
 
 int plugin_num_vcpus(void);
 
+struct qemu_plugin_scoreboard *plugin_scoreboard_new(size_t element_size);
+
+void plugin_scoreboard_free(struct qemu_plugin_scoreboard *score);
+
 #endif /* PLUGIN_H */

plugins/qemu-plugins.symbols

@@ -37,6 +37,9 @@
   qemu_plugin_register_vcpu_tb_exec_inline;
   qemu_plugin_register_vcpu_tb_trans_cb;
   qemu_plugin_reset;
+  qemu_plugin_scoreboard_free;
+  qemu_plugin_scoreboard_find;
+  qemu_plugin_scoreboard_new;
   qemu_plugin_start_code;
   qemu_plugin_tb_get_insn;
   qemu_plugin_tb_n_insns;