[PATCH] misc: fastrpc: Fix race condition in MUNMAP

[PATCH] misc: fastrpc: Fix race condition in MUNMAP

[Ling Xu]

fastrpc_req_munmap involves two steps to unmap memory, first to locates
a matching fastrpc buf in the list and second is to send request to DSP
to unmap. There is a potential race condition between two operations
which can lead to user-after-free scenario.
Lock unmap request to avoid use-after-free.

Signed-off-by: Ling Xu <quic_lxu5@quicinc.com>
---
 drivers/misc/fastrpc.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c
index 7b7a22c91fe4..9ad092d3a705 100644
--- a/drivers/misc/fastrpc.c
+++ b/drivers/misc/fastrpc.c
@@ -2117,7 +2117,9 @@ 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);
-- 
2.34.1

Re: [PATCH] misc: fastrpc: Fix race condition in MUNMAP

[Greg KH]

On Wed, Jan 08, 2025 at 10:18:02AM +0530, Ling Xu wrote:
> fastrpc_req_munmap involves two steps to unmap memory, first to locates
> a matching fastrpc buf in the list and second is to send request to DSP
> to unmap. There is a potential race condition between two operations
> which can lead to user-after-free scenario.
> Lock unmap request to avoid use-after-free.
> 
> Signed-off-by: Ling Xu <quic_lxu5@quicinc.com>
> ---
>  drivers/misc/fastrpc.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c
> index 7b7a22c91fe4..9ad092d3a705 100644
> --- a/drivers/misc/fastrpc.c
> +++ b/drivers/misc/fastrpc.c
> @@ -2117,7 +2117,9 @@ 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);

Why do you grab a mutex here when this function grabs the spinlock too?
This feels like an odd fix, shouldn't the lock be in the function at the
very least?

thanks,

greg k-h