[PATCH] ethernet/realtek/r8169: Fix a double free in rtl8169_start_xmit

Subsystems: 8169 10/100/1000 gigabit ethernet driver, networking drivers, the rest

STALE1982d

3 messages, 2 authors, 2021-03-29 · open the first message on its own page

[PATCH] ethernet/realtek/r8169: Fix a double free in rtl8169_start_xmit

From: Lv Yunlong <hidden>
Date: 2021-03-29 09:08:17

In rtl8169_start_xmit, it calls rtl8169_tso_csum_v2(tp, skb, opts) and
rtl8169_tso_csum_v2() calls __skb_put_padto(skb, padto, false). If
__skb_put_padto() failed, it will free the skb in the first time and
return error. Then rtl8169_start_xmit will goto err_dma_0.

But in err_dma_0 label, the skb is freed by dev_kfree_skb_any(skb) in
the second time.

My patch adds a new label inside the old err_dma_0 label to avoid the
double free and renames the error labels to keep the origin function
unchanged.

Fixes: b8447abc4c8fb ("r8169: factor out rtl8169_tx_map")
Signed-off-by: Lv Yunlong <redacted>
---
 drivers/net/ethernet/realtek/r8169_main.c | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index f704da3f214c..91c43399712b 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c

@@ -4217,13 +4217,13 @@ static netdev_tx_t rtl8169_start_xmit(struct sk_buff *skb,
 
 	if (unlikely(rtl8169_tx_map(tp, opts, skb_headlen(skb), skb->data,
 				    entry, false)))
-		goto err_dma_0;
+		goto err_dma_1;
 
 	txd_first = tp->TxDescArray + entry;
 
 	if (frags) {
 		if (rtl8169_xmit_frags(tp, skb, opts, entry))
-			goto err_dma_1;
+			goto err_dma_2;
 		entry = (entry + frags) % NUM_TX_DESC;
 	}

@@ -4270,10 +4270,11 @@ static netdev_tx_t rtl8169_start_xmit(struct sk_buff *skb,
 
 	return NETDEV_TX_OK;
 
-err_dma_1:
+err_dma_2:
 	rtl8169_unmap_tx_skb(tp, entry);
-err_dma_0:
+err_dma_1:
 	dev_kfree_skb_any(skb);
+err_dma_0:
 	dev->stats.tx_dropped++;
 	return NETDEV_TX_OK;

-- 
2.25.1

Re: [PATCH] ethernet/realtek/r8169: Fix a double free in rtl8169_start_xmit

From: Heiner Kallweit <hkallweit1@gmail.com>
Date: 2021-03-29 09:52:41

On 29.03.2021 11:02, Lv Yunlong wrote:

In rtl8169_start_xmit, it calls rtl8169_tso_csum_v2(tp, skb, opts) and
rtl8169_tso_csum_v2() calls __skb_put_padto(skb, padto, false). If
__skb_put_padto() failed, it will free the skb in the first time and
return error. Then rtl8169_start_xmit will goto err_dma_0.

No, the skb isn't freed here in case of error. Have a look at the
implementation of __skb_put_padto() and see also cc6528bc9a0c
("r8169: fix potential skb double free in an error path").

quoted hunk

But in err_dma_0 label, the skb is freed by dev_kfree_skb_any(skb) in
the second time.

My patch adds a new label inside the old err_dma_0 label to avoid the
double free and renames the error labels to keep the origin function
unchanged.

Fixes: b8447abc4c8fb ("r8169: factor out rtl8169_tx_map")
Signed-off-by: Lv Yunlong <redacted>
---
 drivers/net/ethernet/realtek/r8169_main.c | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index f704da3f214c..91c43399712b 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c

@@ -4217,13 +4217,13 @@ static netdev_tx_t rtl8169_start_xmit(struct sk_buff *skb,
 
 	if (unlikely(rtl8169_tx_map(tp, opts, skb_headlen(skb), skb->data,
 				    entry, false)))
-		goto err_dma_0;
+		goto err_dma_1;
 
 	txd_first = tp->TxDescArray + entry;
 
 	if (frags) {
 		if (rtl8169_xmit_frags(tp, skb, opts, entry))
-			goto err_dma_1;
+			goto err_dma_2;
 		entry = (entry + frags) % NUM_TX_DESC;
 	}

@@ -4270,10 +4270,11 @@ static netdev_tx_t rtl8169_start_xmit(struct sk_buff *skb,
 
 	return NETDEV_TX_OK;
 
-err_dma_1:
+err_dma_2:
 	rtl8169_unmap_tx_skb(tp, entry);
-err_dma_0:
+err_dma_1:
 	dev_kfree_skb_any(skb);
+err_dma_0:
 	dev->stats.tx_dropped++;
 	return NETDEV_TX_OK;

Re: Re: [PATCH] ethernet/realtek/r8169: Fix a double free in rtl8169_start_xmit

From: <hidden>
Date: 2021-03-29 10:12:14

-----原始邮件-----
发件人: "Heiner Kallweit" [off-list ref]
发送时间: 2021-03-29 17:51:30 (星期一)
收件人: "Lv Yunlong" [off-list ref], nic_swsd@realtek.com, kuba@kernel.org
抄送: netdev@vger.kernel.org, linux-kernel@vger.kernel.org
主题: Re: [PATCH] ethernet/realtek/r8169: Fix a double free in rtl8169_start_xmit

On 29.03.2021 11:02, Lv Yunlong wrote:

quoted

In rtl8169_start_xmit, it calls rtl8169_tso_csum_v2(tp, skb, opts) and
rtl8169_tso_csum_v2() calls __skb_put_padto(skb, padto, false). If
__skb_put_padto() failed, it will free the skb in the first time and
return error. Then rtl8169_start_xmit will goto err_dma_0.

No, the skb isn't freed here in case of error. Have a look at the
implementation of __skb_put_padto() and see also cc6528bc9a0c
("r8169: fix potential skb double free in an error path").

quoted

But in err_dma_0 label, the skb is freed by dev_kfree_skb_any(skb) in
the second time.

My patch adds a new label inside the old err_dma_0 label to avoid the
double free and renames the error labels to keep the origin function
unchanged.

Fixes: b8447abc4c8fb ("r8169: factor out rtl8169_tx_map")
Signed-off-by: Lv Yunlong <redacted>
---
 drivers/net/ethernet/realtek/r8169_main.c | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index f704da3f214c..91c43399712b 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c

@@ -4217,13 +4217,13 @@ static netdev_tx_t rtl8169_start_xmit(struct sk_buff *skb,
 
 	if (unlikely(rtl8169_tx_map(tp, opts, skb_headlen(skb), skb->data,
 				    entry, false)))
-		goto err_dma_0;
+		goto err_dma_1;
 
 	txd_first = tp->TxDescArray + entry;
 
 	if (frags) {
 		if (rtl8169_xmit_frags(tp, skb, opts, entry))
-			goto err_dma_1;
+			goto err_dma_2;
 		entry = (entry + frags) % NUM_TX_DESC;
 	}

@@ -4270,10 +4270,11 @@ static netdev_tx_t rtl8169_start_xmit(struct sk_buff *skb,
 
 	return NETDEV_TX_OK;
 
-err_dma_1:
+err_dma_2:
 	rtl8169_unmap_tx_skb(tp, entry);
-err_dma_0:
+err_dma_1:
 	dev_kfree_skb_any(skb);
+err_dma_0:
 	dev->stats.tx_dropped++;
 	return NETDEV_TX_OK;

Ok, i see taht __skb_put_padto(skb, padto, false) isn't free skb on error.

Thanks.

`h`	back out one level
`j`	next message in thread
`k`	previous message in thread
`l`	drill in
`Esc`	close help / fold thread tree
`?`	toggle this help