[PATCH v6 7/9] module: Make the mod_tree stuff conditional on PERF_EVENTS || TRACING

[Peter Zijlstra]

Andrew worried about the overhead on small systems; only use the fancy
code when either perf or tracing is enabled.

Cc: Rusty Russell <rusty@xxxxxxxxxxxxxxx>
Cc: Steven Rostedt <rostedt@xxxxxxxxxxx>
Requested-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
Signed-off-by: Peter Zijlstra (Intel) <peterz@xxxxxxxxxxxxx>
---
 include/linux/module.h |    4 +++-
 init/Kconfig           |    4 ++++
 kernel/module.c        |   30 ++++++++++++++++++++++++++++--
 3 files changed, 35 insertions(+), 3 deletions(-)

--- a/include/linux/module.h
+++ b/include/linux/module.h
@@ -282,7 +282,7 @@ struct module {
 	 *
 	 * Cacheline align here, such that:
 	 *   module_init, module_core, init_size, core_size,
-	 *   init_text_size, core_text_size and ltn_core.node[0]
+	 *   init_text_size, core_text_size and mtn_core::{mod,node[0]}
 	 * are on the same cacheline.
 	 */
 	void *module_init	____cacheline_aligned;
@@ -296,6 +296,7 @@ struct module {
 	/* The size of the executable code in each section.  */
 	unsigned int init_text_size, core_text_size;
 
+#ifdef CONFIG_MODULES_TREE_LOOKUP
 	/*
 	 * We want mtn_core::{mod,node[0]} to be in the same cacheline as the
 	 * above entries such that a regular lookup will only touch one
@@ -303,6 +304,7 @@ struct module {
 	 */
 	struct mod_tree_node	mtn_core;
 	struct mod_tree_node	mtn_init;
+#endif
 
 	/* Size of RO sections of the module (text+rodata) */
 	unsigned int init_ro_size, core_ro_size;
--- a/init/Kconfig
+++ b/init/Kconfig
@@ -1978,6 +1978,10 @@ endchoice
 
 endif # MODULES
 
+config MODULES_TREE_LOOKUP
+	def_bool y
+	depends on PERF_EVENTS || TRACING
+
 config INIT_ALL_POSSIBLE
 	bool
 	help
--- a/kernel/module.c
+++ b/kernel/module.c
@@ -102,6 +102,8 @@ DEFINE_MUTEX(module_mutex);
 EXPORT_SYMBOL_GPL(module_mutex);
 static LIST_HEAD(modules);
 
+#ifdef CONFIG_MODULES_TREE_LOOKUP
+
 /*
  * Use a latched RB-tree for __module_address(); this allows us to use
  * RCU-sched lookups of the address from any context.
@@ -112,6 +114,10 @@ static LIST_HEAD(modules);
  *
  * Because init ranges are short lived we mark them unlikely and have placed
  * them outside the critical cacheline in struct module.
+ *
+ * This is conditional on PERF_EVENTS || TRACING because those can really hit
+ * __module_address() hard by doing a lot of stack unwinding; potentially from
+ * NMI context.
  */
 
 static __always_inline unsigned long __mod_tree_val(struct latch_tree_node *n)
@@ -192,7 +198,7 @@ static void mod_tree_remove(struct modul
 	mod_tree_remove_init(mod);
 }
 
-static struct module *mod_tree_find(unsigned long addr)
+static struct module *mod_find(unsigned long addr)
 {
 	struct latch_tree_node *ltn;
 
@@ -203,6 +209,26 @@ static struct module *mod_tree_find(unsi
 	return container_of(ltn, struct mod_tree_node, node)->mod;
 }
 
+#else /* MODULES_TREE_LOOKUP */
+
+static void mod_tree_insert(struct module *mod) { }
+static void mod_tree_remove_init(struct module *mod) { }
+static void mod_tree_remove(struct module *mod) { }
+
+static struct module *mod_find(unsigned long addr)
+{
+	struct module *mod;
+
+	list_for_each_entry_rcu(mod, &modules, list) {
+		if (within_module(addr, mod))
+			return mod;
+	}
+
+	return NULL;
+}
+
+#endif /* MODULES_TREE_LOOKUP */
+
 #ifdef CONFIG_KGDB_KDB
 struct list_head *kdb_modules = &modules; /* kdb needs the list of modules */
 #endif /* CONFIG_KGDB_KDB */
@@ -3963,7 +3989,7 @@ struct module *__module_address(unsigned
 	if (addr < module_addr_min || addr > module_addr_max)
 		return NULL;
 
-	mod = mod_tree_find(addr);
+	mod = mod_find(addr);
 	if (mod) {
 		BUG_ON(!within_module(addr, mod));
 		if (mod->state == MODULE_STATE_UNFORMED)


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/