linux-kernel - Re: [PATCH] kernel/kprobes: add re-register safe check for register

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-Id: <20171128215520.21846e0b2243f6654c9a63f0@kernel.org>
Date:   Tue, 28 Nov 2017 21:55:20 +0900
From:   Masami Hiramatsu <mhiramat@...nel.org>
To:     JianKang Chen <chenjiankang1@...wei.com>
Cc:     <mingo@...nel.org>, <ananth@...ux.vnet.ibm.com>,
        <anil.s.keshavamurthy@...el.com>, <linux-kernel@...r.kernel.org>,
        <xieyisheng1@...wei.com>, <wangkefeng.wang@...wei.com>
Subject: Re: [PATCH] kernel/kprobes: add re-register safe check for
 register_kretprobe()

On Mon, 27 Nov 2017 20:25:25 +0800
JianKang Chen <chenjiankang1@...wei.com> wrote:

> From: Chen Jiankang <chenjiankang1@...wei.com>
> 
> When there are two same struct kretprobe rp, the INIT_HLIST_HEAD()
> will result in a empty list table rp->free_instances. The memory leak
> will happen. So it needs to add re-register safe check by
> __get_valid_kprobe().
> 
> However, current this is not safe for multi-threadings, because
> there is still a chance to re-register kretprobe concurrently.
> So I add a kretprobe_mutex lock to protect the INIT_LIST_HEAD;
> 
> And we use rcu read lock to protect the rcu list for __get_valid_kprobe
> 
> Signed-off-by: Chen Jiankang <chenjiankang1@...wei.com>
> ---
>  kernel/kprobes.c | 14 +++++++++++++-
>  1 file changed, 13 insertions(+), 1 deletion(-)
> 
> diff --git a/kernel/kprobes.c b/kernel/kprobes.c
> index a1606a4..6e4e657 100644
> --- a/kernel/kprobes.c
> +++ b/kernel/kprobes.c
> @@ -67,6 +67,7 @@
>  
>  /* This protects kprobe_table and optimizing_list */
>  static DEFINE_MUTEX(kprobe_mutex);
> +static DEFINE_MUTEX(kretprobe_mutex);
>  static DEFINE_PER_CPU(struct kprobe *, kprobe_instance) = NULL;
>  static struct {
>  	raw_spinlock_t lock ____cacheline_aligned_in_smp;
> @@ -1947,6 +1948,13 @@ int register_kretprobe(struct kretprobe *rp)
>  		rp->maxactive = num_possible_cpus();
>  #endif
>  	}
> +
> +	mutex_lock(&kretprobe_mutex);
> +	rcu_read_lock();
> +	if (__get_valid_kprobe(&rp->kp)) {
> +		ret = -EINVAL;
> +		goto out;
> +	}
>  	raw_spin_lock_init(&rp->lock);

You can not use rcu_read_lock() as this way, since below block
including mutex_lock() (means it may sleep) and also it updates
rcu list.

So, instead, I meant 

	mutex_lock(&kretprobe_mutex);
	rcu_read_lock();
	kp = __get_valid_kprobe(&rp->kp);
	rcu_read_unlock();
	if (kp) {
		ret = -EINVAL;
		goto out;
	}

Which is enough, because we only take care the case that
same "kretprobe" is reused, which is protected by
kretprobe_mutex. rcu_read_lock() is only for accessing
kprobe_table.

Thank you,


>  	INIT_HLIST_HEAD(&rp->free_instances);
>  	for (i = 0; i < rp->maxactive; i++) {
> @@ -1954,7 +1962,8 @@ int register_kretprobe(struct kretprobe *rp)
>  			       rp->data_size, GFP_KERNEL);
>  		if (inst == NULL) {
>  			free_rp_inst(rp);
> -			return -ENOMEM;
> +			ret = -ENOMEM;
> +			goto out;
>  		}
>  		INIT_HLIST_NODE(&inst->hlist);
>  		hlist_add_head(&inst->hlist, &rp->free_instances);
> @@ -1965,6 +1974,9 @@ int register_kretprobe(struct kretprobe *rp)
>  	ret = register_kprobe(&rp->kp);
>  	if (ret != 0)
>  		free_rp_inst(rp);
> +out:
> +	rcu_read_unlock();
> +	mutex_unlock(&kretprobe_mutex);
>  	return ret;
>  }
>  EXPORT_SYMBOL_GPL(register_kretprobe);
> -- 
> 1.7.12.4
> 


-- 
Masami Hiramatsu <mhiramat@...nel.org>