[PATCH v3 1/2] kprobes: avoid the kprobe being re-registered

[PATCH v3 1/2] kprobes: avoid the kprobe being re-registered

[Zhou Chengming]

Changes:
- We should put the modifies of the kprobe after the re-reg check.
- And then the address_safe check.

Old code use check_kprobe_rereg() to check if the kprobe has been
registered already, but check_kprobe_rereg() will release the
kprobe_mutex then, so maybe two paths will pass the check and
register the same kprobe. This patch put the check inside the mutex.

Suggested-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Zhou Chengming <zhouchengming1@huawei.com>
---
 kernel/kprobes.c | 27 ++++++++-------------------
 1 file changed, 8 insertions(+), 19 deletions(-)

diff --git a/kernel/kprobes.c b/kernel/kprobes.c
index a1606a4..1eeedac 100644
--- a/kernel/kprobes.c
+++ b/kernel/kprobes.c
@@ -1443,19 +1443,6 @@ static struct kprobe *__get_valid_kprobe(struct kprobe *p)
 	return ap;
 }
 
-/* Return error if the kprobe is being re-registered */
-static inline int check_kprobe_rereg(struct kprobe *p)
-{
-	int ret = 0;
-
-	mutex_lock(&kprobe_mutex);
-	if (__get_valid_kprobe(p))
-		ret = -EINVAL;
-	mutex_unlock(&kprobe_mutex);
-
-	return ret;
-}
-
 int __weak arch_check_ftrace_location(struct kprobe *p)
 {
 	unsigned long ftrace_addr;
@@ -1536,9 +1523,13 @@ int register_kprobe(struct kprobe *p)
 		return PTR_ERR(addr);
 	p->addr = addr;
 
-	ret = check_kprobe_rereg(p);
-	if (ret)
-		return ret;
+	mutex_lock(&kprobe_mutex);
+
+	/* Return error if the kprobe is being re-registered */
+	if (__get_valid_kprobe(p)) {
+		ret = -EINVAL;
+		goto out;
+	}
 
 	/* User can pass only KPROBE_FLAG_DISABLED to register_kprobe */
 	p->flags &= KPROBE_FLAG_DISABLED;
@@ -1547,9 +1538,7 @@ int register_kprobe(struct kprobe *p)
 
 	ret = check_kprobe_address_safe(p, &probed_mod);
 	if (ret)
-		return ret;
-
-	mutex_lock(&kprobe_mutex);
+		goto out;
 
 	old_p = get_kprobe(p->addr);
 	if (old_p) {
-- 
1.8.3.1

[PATCH v3 2/2] kprobes: initialize probed_mod to NULL

[Zhou Chengming]

When check_kprobe_address_safe() return fail, the probed_mod
should be set to NULL, because no module refcount held. And we
initialize probed_mod to NULL in register_kprobe() for the same reason.

Signed-off-by: Zhou Chengming <zhouchengming1@huawei.com>
---
 kernel/kprobes.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/kernel/kprobes.c b/kernel/kprobes.c
index 1eeedac..a04588c 100644
--- a/kernel/kprobes.c
+++ b/kernel/kprobes.c
@@ -1488,6 +1488,7 @@ static int check_kprobe_address_safe(struct kprobe *p,
 		 * its code to prohibit unexpected unloading.
 		 */
 		if (unlikely(!try_module_get(*probed_mod))) {
+			*probed_mod = NULL;
 			ret = -ENOENT;
 			goto out;
 		}
@@ -1514,7 +1515,7 @@ int register_kprobe(struct kprobe *p)
 {
 	int ret;
 	struct kprobe *old_p;
-	struct module *probed_mod;
+	struct module *probed_mod = NULL;
 	kprobe_opcode_t *addr;
 
 	/* Adjust probe address from symbol */
-- 
1.8.3.1

Re: [PATCH v3 1/2] kprobes: avoid the kprobe being re-registered

[Masami Hiramatsu]

On Fri, 27 Oct 2017 17:23:14 +0800
Zhou Chengming <zhouchengming1@huawei.com> wrote:

> Changes:
> - We should put the modifies of the kprobe after the re-reg check.
> - And then the address_safe check.
> 
> Old code use check_kprobe_rereg() to check if the kprobe has been
> registered already, but check_kprobe_rereg() will release the
> kprobe_mutex then, so maybe two paths will pass the check and
> register the same kprobe. This patch put the check inside the mutex.
> 
> Suggested-by: Masami Hiramatsu <mhiramat@kernel.org>
> Signed-off-by: Zhou Chengming <zhouchengming1@huawei.com>

Looks good to me :)

Acked-by: Masami Hiramatsu <mhiramat@kernel.org>

Thanks!

> ---
>  kernel/kprobes.c | 27 ++++++++-------------------
>  1 file changed, 8 insertions(+), 19 deletions(-)
> 
> diff --git a/kernel/kprobes.c b/kernel/kprobes.c
> index a1606a4..1eeedac 100644
> --- a/kernel/kprobes.c
> +++ b/kernel/kprobes.c
> @@ -1443,19 +1443,6 @@ static struct kprobe *__get_valid_kprobe(struct kprobe *p)
>  	return ap;
>  }
>  
> -/* Return error if the kprobe is being re-registered */
> -static inline int check_kprobe_rereg(struct kprobe *p)
> -{
> -	int ret = 0;
> -
> -	mutex_lock(&kprobe_mutex);
> -	if (__get_valid_kprobe(p))
> -		ret = -EINVAL;
> -	mutex_unlock(&kprobe_mutex);
> -
> -	return ret;
> -}
> -
>  int __weak arch_check_ftrace_location(struct kprobe *p)
>  {
>  	unsigned long ftrace_addr;
> @@ -1536,9 +1523,13 @@ int register_kprobe(struct kprobe *p)
>  		return PTR_ERR(addr);
>  	p->addr = addr;
>  
> -	ret = check_kprobe_rereg(p);
> -	if (ret)
> -		return ret;
> +	mutex_lock(&kprobe_mutex);
> +
> +	/* Return error if the kprobe is being re-registered */
> +	if (__get_valid_kprobe(p)) {
> +		ret = -EINVAL;
> +		goto out;
> +	}
>  
>  	/* User can pass only KPROBE_FLAG_DISABLED to register_kprobe */
>  	p->flags &= KPROBE_FLAG_DISABLED;
> @@ -1547,9 +1538,7 @@ int register_kprobe(struct kprobe *p)
>  
>  	ret = check_kprobe_address_safe(p, &probed_mod);
>  	if (ret)
> -		return ret;
> -
> -	mutex_lock(&kprobe_mutex);
> +		goto out;
>  
>  	old_p = get_kprobe(p->addr);
>  	if (old_p) {
> -- 
> 1.8.3.1
> 


-- 
Masami Hiramatsu <mhiramat@kernel.org>

Re: [PATCH v3 2/2] kprobes: initialize probed_mod to NULL

[Masami Hiramatsu]

On Fri, 27 Oct 2017 17:23:15 +0800
Zhou Chengming <zhouchengming1@huawei.com> wrote:

> When check_kprobe_address_safe() return fail, the probed_mod
> should be set to NULL, because no module refcount held. And we
> initialize probed_mod to NULL in register_kprobe() for the same reason.
> 

Ok, it's a kind of hardening code but looks good to me.

Acked-by: Masami Hiramatsu <mhiramat@kernel.org>

Thank you!


> Signed-off-by: Zhou Chengming <zhouchengming1@huawei.com>
> ---
>  kernel/kprobes.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/kernel/kprobes.c b/kernel/kprobes.c
> index 1eeedac..a04588c 100644
> --- a/kernel/kprobes.c
> +++ b/kernel/kprobes.c
> @@ -1488,6 +1488,7 @@ static int check_kprobe_address_safe(struct kprobe *p,
>  		 * its code to prohibit unexpected unloading.
>  		 */
>  		if (unlikely(!try_module_get(*probed_mod))) {
> +			*probed_mod = NULL;
>  			ret = -ENOENT;
>  			goto out;
>  		}
> @@ -1514,7 +1515,7 @@ int register_kprobe(struct kprobe *p)
>  {
>  	int ret;
>  	struct kprobe *old_p;
> -	struct module *probed_mod;
> +	struct module *probed_mod = NULL;
>  	kprobe_opcode_t *addr;
>  
>  	/* Adjust probe address from symbol */
> -- 
> 1.8.3.1
> 


-- 
Masami Hiramatsu <mhiramat@kernel.org>