答复: [PATCH] gpu/drm: fix potential memleak in error branch

[赵军奎]

-----邮件原件-----
发件人: bernard at vivo.com <bernard at vivo.com> 代表 Jani Nikula
发送时间: 2021年12月31日 19:09
收件人: 赵军奎 <bernard at vivo.com>; Maarten Lankhorst <maarten.lankhorst at linux.intel.com>; Maxime Ripard <mripard at kernel.org>; Thomas Zimmermann <tzimmermann at suse.de>; David Airlie <airlied at linux.ie>; Daniel Vetter <daniel at ffwll.ch>; dri-devel at lists.freedesktop.org; linux-kernel at vger.kernel.org
抄送: 赵军奎 <bernard at vivo.com>
主题: Re: [PATCH] gpu/drm: fix potential memleak in error branch

On Tue, 16 Nov 2021, Bernard Zhao <bernard at vivo.com> wrote:
> This patch try to fix potential memleak in error branch.

>Please elaborate.

Hi Jani:

This patch try to fix potential memleak in error branch.
For example:
nv50_sor_create ->nv50_mstm_new-> drm_dp_mst_topology_mgr_init
In function drm_dp_mst_topology_mgr_init, there are five error branches, error branch just return error code, no free called. 
And we see that the caller didn`t do the drm_dp_mst_topology_mgr_destroy job.
I am not sure if there some gap, I think this may bring in the risk of memleak issue.
Thanks!

BR//Bernard

>BR,
>Jani.


>
> Signed-off-by: Bernard Zhao <bernard at vivo.com>
> ---
>  drivers/gpu/drm/drm_dp_mst_topology.c | 22 ++++++++++++++++------
>  1 file changed, 16 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/gpu/drm/drm_dp_mst_topology.c 
> b/drivers/gpu/drm/drm_dp_mst_topology.c
> index f3d79eda94bb..f73b180dee73 100644
> --- a/drivers/gpu/drm/drm_dp_mst_topology.c
> +++ b/drivers/gpu/drm/drm_dp_mst_topology.c
> @@ -5501,7 +5501,10 @@ int drm_dp_mst_topology_mgr_init(struct drm_dp_mst_topology_mgr *mgr,
>  				 int max_lane_count, int max_link_rate,
>  				 int conn_base_id)
>  {
> -	struct drm_dp_mst_topology_state *mst_state;
> +	struct drm_dp_mst_topology_state *mst_state = NULL;
> +
> +	mgr->payloads = NULL;
> +	mgr->proposed_vcpis = NULL;
>  
>  	mutex_init(&mgr->lock);
>  	mutex_init(&mgr->qlock);
> @@ -5523,7 +5526,7 @@ int drm_dp_mst_topology_mgr_init(struct drm_dp_mst_topology_mgr *mgr,
>  	 */
>  	mgr->delayed_destroy_wq = alloc_ordered_workqueue("drm_dp_mst_wq", 0);
>  	if (mgr->delayed_destroy_wq == NULL)
> -		return -ENOMEM;
> +		goto out;
>  
>  	INIT_WORK(&mgr->work, drm_dp_mst_link_probe_work);
>  	INIT_WORK(&mgr->tx_work, drm_dp_tx_work); @@ -5539,18 +5542,18 @@ 
> int drm_dp_mst_topology_mgr_init(struct drm_dp_mst_topology_mgr *mgr,
>  	mgr->conn_base_id = conn_base_id;
>  	if (max_payloads + 1 > sizeof(mgr->payload_mask) * 8 ||
>  	    max_payloads + 1 > sizeof(mgr->vcpi_mask) * 8)
> -		return -EINVAL;
> +		goto failed;
>  	mgr->payloads = kcalloc(max_payloads, sizeof(struct drm_dp_payload), GFP_KERNEL);
>  	if (!mgr->payloads)
> -		return -ENOMEM;
> +		goto failed;
>  	mgr->proposed_vcpis = kcalloc(max_payloads, sizeof(struct drm_dp_vcpi *), GFP_KERNEL);
>  	if (!mgr->proposed_vcpis)
> -		return -ENOMEM;
> +		goto failed;
>  	set_bit(0, &mgr->payload_mask);
>  
>  	mst_state = kzalloc(sizeof(*mst_state), GFP_KERNEL);
>  	if (mst_state == NULL)
> -		return -ENOMEM;
> +		goto failed;
>  
>  	mst_state->total_avail_slots = 63;
>  	mst_state->start_slot = 1;
> @@ -5563,6 +5566,13 @@ int drm_dp_mst_topology_mgr_init(struct drm_dp_mst_topology_mgr *mgr,
>  				    &drm_dp_mst_topology_state_funcs);
>  
>  	return 0;
> +
> +failed:
> +	kfree(mgr->proposed_vcpis);
> +	kfree(mgr->payloads);
> +	destroy_workqueue(mgr->delayed_destroy_wq);
> +out:
> +	return -ENOMEM;
>  }
>  EXPORT_SYMBOL(drm_dp_mst_topology_mgr_init);

--
Jani Nikula, Intel Open Source Graphics Center