Re: [PATCH 02/15] drivers, net, ethernet: convert mtk_eth.dma_refcnt from atomic_t to refcount_t

[Sean Wang]

On Fri, 2017-10-20 at 10:23 +0300, Elena Reshetova wrote:
> atomic_t variables are currently used to implement reference
> counters with the following properties:
>  - counter is initialized to 1 using atomic_set()
>  - a resource is freed upon counter reaching zero
>  - once counter reaches zero, its further
>    increments aren't allowed
>  - counter schema uses basic atomic operations
>    (set, inc, inc_not_zero, dec_and_test, etc.)
> 
> Such atomic variables should be converted to a newly provided
> refcount_t type and API that prevents accidental counter overflows
> and underflows. This is important since overflows and underflows
> can lead to use-after-free situation and be exploitable.
> 
> The variable mtk_eth.dma_refcnt is used as pure reference counter.
> Convert it to refcount_t and fix up the operations.
> 
> Suggested-by: Kees Cook <keescook@xxxxxxxxxxxx>
> Reviewed-by: David Windsor <dwindsor@xxxxxxxxx>
> Reviewed-by: Hans Liljestrand <ishkamiel@xxxxxxxxx>
> Signed-off-by: Elena Reshetova <elena.reshetova@xxxxxxxxx>
> ---
>  drivers/net/ethernet/mediatek/mtk_eth_soc.c | 8 +++++---
>  drivers/net/ethernet/mediatek/mtk_eth_soc.h | 4 +++-
>  2 files changed, 8 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
> index 5e81a72..54adfd9 100644
> --- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
> +++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
> @@ -1817,7 +1817,7 @@ static int mtk_open(struct net_device *dev)
>  	struct mtk_eth *eth = mac->hw;
>  
>  	/* we run 2 netdevs on the same dma ring so we only bring it up once */
> -	if (!atomic_read(&eth->dma_refcnt)) {
> +	if (!refcount_read(&eth->dma_refcnt)) {
>  		int err = mtk_start_dma(eth);
>  
>  		if (err)
> @@ -1827,8 +1827,10 @@ static int mtk_open(struct net_device *dev)
>  		napi_enable(&eth->rx_napi);
>  		mtk_tx_irq_enable(eth, MTK_TX_DONE_INT);
>  		mtk_rx_irq_enable(eth, MTK_RX_DONE_INT);
> +		refcount_set(&eth->dma_refcnt, 1);

the existing driver seems to have a missing initial atomic_set for the
eth->dma_refcnt. 

>  	}
> -	atomic_inc(&eth->dma_refcnt);
> +	else
> +		refcount_inc(&eth->dma_refcnt);
>  

how about add the initial refcount_set into probe handler, and keep
logic else unchanged ? 

>  	phy_start(dev->phydev);
>  	netif_start_queue(dev);
> @@ -1868,7 +1870,7 @@ static int mtk_stop(struct net_device *dev)
>  	phy_stop(dev->phydev);
>  
>  	/* only shutdown DMA if this is the last user */
> -	if (!atomic_dec_and_test(&eth->dma_refcnt))
> +	if (!refcount_dec_and_test(&eth->dma_refcnt))
>  		return 0;
>  
>  	mtk_tx_irq_disable(eth, MTK_TX_DONE_INT);
> diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.h b/drivers/net/ethernet/mediatek/mtk_eth_soc.h
> index 3d3c24a..a3af466 100644
> --- a/drivers/net/ethernet/mediatek/mtk_eth_soc.h
> +++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.h
> @@ -15,6 +15,8 @@
>  #ifndef MTK_ETH_H
>  #define MTK_ETH_H
>  
> +#include <linux/refcount.h>
> +
>  #define MTK_QDMA_PAGE_SIZE	2048
>  #define	MTK_MAX_RX_LENGTH	1536
>  #define MTK_TX_DMA_BUF_LEN	0x3fff
> @@ -632,7 +634,7 @@ struct mtk_eth {
>  	struct regmap			*pctl;
>  	u32				chip_id;
>  	bool				hwlro;
> -	atomic_t			dma_refcnt;
> +	refcount_t			dma_refcnt;
>  	struct mtk_tx_ring		tx_ring;
>  	struct mtk_rx_ring		rx_ring[MTK_MAX_RX_RING_NUM];
>  	struct mtk_rx_ring		rx_ring_qdma;