[PATCH] module: Fix missing to hold module_mutex lock in module_kallsyms_lookup_name

[PATCH] module: Fix missing to hold module_mutex lock in module_kallsyms_lookup_name

[He Kuang]

Function find_module_all() searches for module by name and must be
called with module_mutex. module_kallsyms_lookup_name() calls it without
this mutex which emits a warning message (CONFIG_LOCKDEP=y) by failed
assertion for testing this module_mutex lock, as following:

  [  202.877152] ------------[ cut here ]------------
  [  202.881070] WARNING: CPU: 0 PID: 1010 at kernel/module.c:281
     module_assert_mutex+0x35/0x40()
  [  202.885446] Modules linked in: test_bpf
  [  202.886997] CPU: 0 PID: 1010 Comm: perf Tainted: G  W  4.2.0-rc3+ #5
  ...

This patch wraps this call with mutex_{lock,unlock} and fix the bug.

Signed-off-by: He Kuang <hekuang@huawei.com>
---
 kernel/module.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/kernel/module.c b/kernel/module.c
index 4d2b82e..43d728b 100644
--- a/kernel/module.c
+++ b/kernel/module.c
@@ -3802,8 +3802,10 @@ unsigned long module_kallsyms_lookup_name(const char *name)
 	/* Don't lock: we're in enough trouble already. */
 	preempt_disable();
 	if ((colon = strchr(name, ':')) != NULL) {
+		mutex_lock(&module_mutex);
 		if ((mod = find_module_all(name, colon - name, false)) != NULL)
 			ret = mod_find_symname(mod, colon+1);
+		mutex_unlock(&module_mutex);
 	} else {
 		list_for_each_entry_rcu(mod, &modules, list) {
 			if (mod->state == MODULE_STATE_UNFORMED)
-- 
1.8.5.2

Re: [PATCH] module: Fix missing to hold module_mutex lock in module_kallsyms_lookup_name

[Rusty Russell]

He Kuang <hekuang@huawei.com> writes:

> Function find_module_all() searches for module by name and must be
> called with module_mutex. module_kallsyms_lookup_name() calls it without
> this mutex which emits a warning message (CONFIG_LOCKDEP=y) by failed
> assertion for testing this module_mutex lock, as following:
>
>   [  202.877152] ------------[ cut here ]------------
>   [  202.881070] WARNING: CPU: 0 PID: 1010 at kernel/module.c:281
>      module_assert_mutex+0x35/0x40()
>   [  202.885446] Modules linked in: test_bpf
>   [  202.886997] CPU: 0 PID: 1010 Comm: perf Tainted: G  W  4.2.0-rc3+ #5
>   ...
>
> This patch wraps this call with mutex_{lock,unlock} and fix the bug.

Hi He!

        Thanks for this report!  This warning is overzealous; preempt
disabling should be enough to read the list.

Unfortunately, as you can see from the comment, taking a lock is a bad
idea here: it's called in the oops path (we don't want to risk
deadlock).

Peter?

Cheers,
Rusty.

module: weaken locking assertion for oops path.

We don't actually hold the module_mutex when calling find_module_all
from module_kallsyms_lookup_name: that's because it's used by the oops
code and we don't want to deadlock.

However, access to the list read-only is safe if preempt is disabled,
so we can weaken the assertion.  Keep a strong version for external
callers though.

Fixes: 0be964be0d45 ("module: Sanitize RCU usage and locking")
Reported-by: He Kuang <hekuang@huawei.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

diff --git a/kernel/module.c b/kernel/module.c
index 4d2b82e610e2..b86b7bf1be38 100644
--- a/kernel/module.c
+++ b/kernel/module.c
@@ -602,13 +602,16 @@ const struct kernel_symbol *find_symbol(const char *name,
 }
 EXPORT_SYMBOL_GPL(find_symbol);
 
-/* Search for module by name: must hold module_mutex. */
+/*
+ * Search for module by name: must hold module_mutex (or preempt disabled
+ * for read-only access).
+ */
 static struct module *find_module_all(const char *name, size_t len,
 				      bool even_unformed)
 {
 	struct module *mod;
 
-	module_assert_mutex();
+	module_assert_mutex_or_preempt();
 
 	list_for_each_entry(mod, &modules, list) {
 		if (!even_unformed && mod->state == MODULE_STATE_UNFORMED)
@@ -621,6 +624,7 @@ static struct module *find_module_all(const char *name, size_t len,
 
 struct module *find_module(const char *name)
 {
+	module_assert_mutex();
 	return find_module_all(name, strlen(name), false);
 }
 EXPORT_SYMBOL_GPL(find_module);

Re: [PATCH] module: Fix missing to hold module_mutex lock in module_kallsyms_lookup_name

[Peter Zijlstra]

On Tue, Jul 28, 2015 at 10:01:35AM +0930, Rusty Russell wrote:
> Peter?

> module: weaken locking assertion for oops path.
> 
> We don't actually hold the module_mutex when calling find_module_all
> from module_kallsyms_lookup_name: that's because it's used by the oops
> code and we don't want to deadlock.
> 
> However, access to the list read-only is safe if preempt is disabled,
> so we can weaken the assertion.  Keep a strong version for external
> callers though.
> 
> Fixes: 0be964be0d45 ("module: Sanitize RCU usage and locking")
> Reported-by: He Kuang <hekuang@huawei.com>
> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
> 
> diff --git a/kernel/module.c b/kernel/module.c
> index 4d2b82e610e2..b86b7bf1be38 100644
> --- a/kernel/module.c
> +++ b/kernel/module.c
> @@ -602,13 +602,16 @@ const struct kernel_symbol *find_symbol(const char *name,
>  }
>  EXPORT_SYMBOL_GPL(find_symbol);
>  
> -/* Search for module by name: must hold module_mutex. */
> +/*
> + * Search for module by name: must hold module_mutex (or preempt disabled
> + * for read-only access).
> + */
>  static struct module *find_module_all(const char *name, size_t len,
>  				      bool even_unformed)
>  {
>  	struct module *mod;
>  
> -	module_assert_mutex();
> +	module_assert_mutex_or_preempt();

Yeah, that should be fine indeed, I went by that comment you just
expanded.

The operation itself does indeed not modify data at all, so the
preempt_disable is perfectly adequate.

Thanks!

Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>