Message-ID: <9bfef283-e2ac-b2ba-386c-6833e9cb1283@linaro.org>
Date:   Tue, 21 Mar 2023 09:27:46 +0000
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.8.0
Subject: Re: [PATCH] misc: fastrpc: Fix a Use after-free-bug by race condition
Content-Language: en-US
To:     Sangsup Lee <k1rh4.lee@gmail.com>
Cc:     Amol Maheshwari <amahesh@qti.qualcomm.com>,
        Arnd Bergmann <arnd@arndb.de>,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        linux-arm-msm@vger.kernel.org, linux-kernel@vger.kernel.org
References: <20230216014120.3110-1-k1rh4.lee@gmail.com>
From:   Srinivas Kandagatla <srinivas.kandagatla@linaro.org>
In-Reply-To: <20230216014120.3110-1-k1rh4.lee@gmail.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

Thanks Sangsup for reporting the issue and sharing the patch,

Sorry, for some reason I missed this patch.

On 16/02/2023 01:41, Sangsup Lee wrote:
> This patch adds mutex_lock for fixing an Use-after-free bug.
> fastrpc_req_munmap_impl can be called concurrently in multi-threded environments.
> The buf which is allocated by list_for_each_safe can be used after another thread frees it.
>
Commit log can be improved here to something like:

fastrcp_munmap takes two steps to unmap the memory, first to find a 
matching fastrpc buf in the list and second is to send request to DSP to 
unmap it.
There is a potentially window of race between these two operations, 
which can lead to user-after-free.

Fix this by adding locking around this two operations.

> Signed-off-by: Sangsup Lee <k1rh4.lee@gmail.com>
> ---
>   drivers/misc/fastrpc.c | 7 ++++++-
>   1 file changed, 6 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c
> index 5310606113fe..c4b5fa4a50a6 100644
> --- a/drivers/misc/fastrpc.c
> +++ b/drivers/misc/fastrpc.c
> @@ -1806,10 +1806,12 @@ static int fastrpc_req_munmap(struct fastrpc_user *fl, char __user *argp)
>   	struct fastrpc_buf *buf = NULL, *iter, *b;
>   	struct fastrpc_req_munmap req;
>   	struct device *dev = fl->sctx->dev;
> +	int err;
>   
>   	if (copy_from_user(&req, argp, sizeof(req)))
>   		return -EFAULT;
>   
> +	mutex_lock(&fl->mutex);
>   	spin_lock(&fl->lock);
>   	list_for_each_entry_safe(iter, b, &fl->mmaps, node) {
>   		if ((iter->raddr == req.vaddrout) && (iter->size == req.size)) {
> @@ -1822,10 +1824,13 @@ static int fastrpc_req_munmap(struct fastrpc_user *fl, char __user *argp)
>   	if (!buf) {
>   		dev_err(dev, "mmap\t\tpt 0x%09llx [len 0x%08llx] not in list\n",
>   			req.vaddrout, req.size);
> +		mutex_unlock(&fl->mutex);
>   		return -EINVAL;
>   	}
>   
> -	return fastrpc_req_munmap_impl(fl, buf);
> +	err = fastrpc_req_munmap_impl(fl, buf);
> +	mutex_unlock(&fl->mutex);
> +	return err;

How about moving the locking to ioctl:

diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c
index a701132638cf..2f217071a6c3 100644
--- a/drivers/misc/fastrpc.c
+++ b/drivers/misc/fastrpc.c
@@ -2087,7 +2087,9 @@ static long fastrpc_device_ioctl(struct file 
*file, unsigned int cmd,
                 err = fastrpc_req_mmap(fl, argp);
                 break;
         case FASTRPC_IOCTL_MUNMAP:
+               mutex_lock(&fl->mutex);
                 err = fastrpc_req_munmap(fl, argp);
+               mutex_unlock(&fl->mutex);
                 break;
         case FASTRPC_IOCTL_MEM_MAP:
                 err = fastrpc_req_mem_map(fl, argp);


thanks,
srini
>   }
>   
>   static int fastrpc_req_mmap(struct fastrpc_user *fl, char __user *argp)