[Linux-kernel-mentees] [PATCH] kernel: module: Pass lockdep expression to RCU lists

Jessica Yu jeyu at kernel.org
Thu Jan 23 12:10:10 UTC 2020


+++ Amol Grover [21/01/20 18:17 +0530]:
>modules is traversed using list_for_each_entry_rcu outside an
>RCU read-side critical section but under the protection
>of module_mutex or with preemption disabled.
>
>Hence, add corresponding lockdep expression to silence false-positive
>lockdep warnings, and harden RCU lists.
>
>list_for_each_entry_rcu when traversed inside a preempt disabled
>section, doesn't need an explicit lockdep expression since it is
>implicitly checked for.
>
>Add macro for the corresponding lockdep expression.
>
>Signed-off-by: Amol Grover <frextrite at gmail.com>

Hi Amol!

Masami already submitted a patch for this, it's been in linux-next for
a while. See commit bf08949cc8b9 ("modules: lockdep: Suppress
suspicious RCU usage warning").

Thanks!

Jessica

>---
> kernel/module.c | 12 +++++++-----
> 1 file changed, 7 insertions(+), 5 deletions(-)
>
>diff --git a/kernel/module.c b/kernel/module.c
>index b56f3224b161..2425f58159dd 100644
>--- a/kernel/module.c
>+++ b/kernel/module.c
>@@ -84,6 +84,8 @@
>  * 3) module_addr_min/module_addr_max.
>  * (delete and add uses RCU list operations). */
> DEFINE_MUTEX(module_mutex);
>+#define module_mutex_held() \
>+	lockdep_is_held(&module_mutex)
> EXPORT_SYMBOL_GPL(module_mutex);
> static LIST_HEAD(modules);
>
>@@ -214,7 +216,7 @@ static struct module *mod_find(unsigned long addr)
> {
> 	struct module *mod;
>
>-	list_for_each_entry_rcu(mod, &modules, list) {
>+	list_for_each_entry_rcu(mod, &modules, list, module_mutex_held()) {
> 		if (within_module(addr, mod))
> 			return mod;
> 	}
>@@ -448,7 +450,7 @@ bool each_symbol_section(bool (*fn)(const struct symsearch *arr,
> 	if (each_symbol_in_section(arr, ARRAY_SIZE(arr), NULL, fn, data))
> 		return true;
>
>-	list_for_each_entry_rcu(mod, &modules, list) {
>+	list_for_each_entry_rcu(mod, &modules, list, module_mutex_held()) {
> 		struct symsearch arr[] = {
> 			{ mod->syms, mod->syms + mod->num_syms, mod->crcs,
> 			  NOT_GPL_ONLY, false },
>@@ -616,7 +618,7 @@ static struct module *find_module_all(const char *name, size_t len,
>
> 	module_assert_mutex_or_preempt();
>
>-	list_for_each_entry_rcu(mod, &modules, list) {
>+	list_for_each_entry_rcu(mod, &modules, list, module_mutex_held()) {
> 		if (!even_unformed && mod->state == MODULE_STATE_UNFORMED)
> 			continue;
> 		if (strlen(mod->name) == len && !memcmp(mod->name, name, len))
>@@ -2040,7 +2042,7 @@ void set_all_modules_text_rw(void)
> 		return;
>
> 	mutex_lock(&module_mutex);
>-	list_for_each_entry_rcu(mod, &modules, list) {
>+	list_for_each_entry_rcu(mod, &modules, list, module_mutex_held()) {
> 		if (mod->state == MODULE_STATE_UNFORMED)
> 			continue;
>
>@@ -2059,7 +2061,7 @@ void set_all_modules_text_ro(void)
> 		return;
>
> 	mutex_lock(&module_mutex);
>-	list_for_each_entry_rcu(mod, &modules, list) {
>+	list_for_each_entry_rcu(mod, &modules, list, module_mutex_held()) {
> 		/*
> 		 * Ignore going modules since it's possible that ro
> 		 * protection has already been disabled, otherwise we'll
>-- 
>2.24.1
>


More information about the Linux-kernel-mentees mailing list