2018-03-29 07:25:07

by Ji-Hun Kim

Subject: [PATCH v2] staging: vt6655: check for memory allocation failures

There are no null pointer checking on rd_info and td_info values which
are allocated by kzalloc. It has potential null pointer dereferencing
issues. Add return when allocation is failed.

Signed-off-by: Ji-Hun Kim <[email protected]>
---

Change: since v1:

- Delete WARN_ON which can makes crashes on some machines.
- Instead of return directly, goto freeing function for freeing previously
allocated memory in the for loop after kzalloc() failed.
- In the freeing function, if td_info and rd_info are not allocated, no
needs to free.

drivers/staging/vt6655/device_main.c | 64 +++++++++++++++++++++++++-----------
1 file changed, 44 insertions(+), 20 deletions(-)

diff --git a/drivers/staging/vt6655/device_main.c b/drivers/staging/vt6655/device_main.c
index fbc4bc6..ecbba43 100644
--- a/drivers/staging/vt6655/device_main.c
+++ b/drivers/staging/vt6655/device_main.c
@@ -539,7 +539,8 @@ static void device_init_rd0_ring(struct vnt_private *priv)
i ++, curr += sizeof(struct vnt_rx_desc)) {
desc = &priv->aRD0Ring[i];
desc->rd_info = kzalloc(sizeof(*desc->rd_info), GFP_KERNEL);
-
+ if (!desc->rd_info)
+ goto error;
if (!device_alloc_rx_buf(priv, desc))
dev_err(&priv->pcid->dev, "can not alloc rx bufs\n");

@@ -550,6 +551,10 @@ static void device_init_rd0_ring(struct vnt_private *priv)
if (i > 0)
priv->aRD0Ring[i-1].next_desc = cpu_to_le32(priv->rd0_pool_dma);
priv->pCurrRD[0] = &priv->aRD0Ring[0];
+
+ return;
+error:
+ device_free_rd0_ring(priv);
}

static void device_init_rd1_ring(struct vnt_private *priv)
@@ -563,7 +568,8 @@ static void device_init_rd1_ring(struct vnt_private *priv)
i ++, curr += sizeof(struct vnt_rx_desc)) {
desc = &priv->aRD1Ring[i];
desc->rd_info = kzalloc(sizeof(*desc->rd_info), GFP_KERNEL);
-
+ if (!desc->rd_info)
+ goto error;
if (!device_alloc_rx_buf(priv, desc))
dev_err(&priv->pcid->dev, "can not alloc rx bufs\n");

@@ -574,6 +580,10 @@ static void device_init_rd1_ring(struct vnt_private *priv)
if (i > 0)
priv->aRD1Ring[i-1].next_desc = cpu_to_le32(priv->rd1_pool_dma);
priv->pCurrRD[1] = &priv->aRD1Ring[0];
+
+ return;
+error:
+ device_free_rd1_ring(priv);
}

static void device_free_rd0_ring(struct vnt_private *priv)
@@ -584,12 +594,12 @@ static void device_free_rd0_ring(struct vnt_private *priv)
struct vnt_rx_desc *desc = &priv->aRD0Ring[i];
struct vnt_rd_info *rd_info = desc->rd_info;

- dma_unmap_single(&priv->pcid->dev, rd_info->skb_dma,
- priv->rx_buf_sz, DMA_FROM_DEVICE);
-
- dev_kfree_skb(rd_info->skb);
-
- kfree(desc->rd_info);
+ if (rd_info) {
+ dma_unmap_single(&priv->pcid->dev, rd_info->skb_dma,
+ priv->rx_buf_sz, DMA_FROM_DEVICE);
+ dev_kfree_skb(rd_info->skb);
+ kfree(desc->rd_info);
+ }
}
}

@@ -601,12 +611,12 @@ static void device_free_rd1_ring(struct vnt_private *priv)
struct vnt_rx_desc *desc = &priv->aRD1Ring[i];
struct vnt_rd_info *rd_info = desc->rd_info;

- dma_unmap_single(&priv->pcid->dev, rd_info->skb_dma,
- priv->rx_buf_sz, DMA_FROM_DEVICE);
-
- dev_kfree_skb(rd_info->skb);
-
- kfree(desc->rd_info);
+ if (rd_info) {
+ dma_unmap_single(&priv->pcid->dev, rd_info->skb_dma,
+ priv->rx_buf_sz, DMA_FROM_DEVICE);
+ dev_kfree_skb(rd_info->skb);
+ kfree(desc->rd_info);
+ }
}
}

@@ -621,7 +631,8 @@ static void device_init_td0_ring(struct vnt_private *priv)
i++, curr += sizeof(struct vnt_tx_desc)) {
desc = &priv->apTD0Rings[i];
desc->td_info = kzalloc(sizeof(*desc->td_info), GFP_KERNEL);
-
+ if (!desc->td_info)
+ goto error;
desc->td_info->buf = priv->tx0_bufs + i * PKT_BUF_SZ;
desc->td_info->buf_dma = priv->tx_bufs_dma0 + i * PKT_BUF_SZ;

@@ -632,6 +643,10 @@ static void device_init_td0_ring(struct vnt_private *priv)
if (i > 0)
priv->apTD0Rings[i-1].next_desc = cpu_to_le32(priv->td0_pool_dma);
priv->apTailTD[0] = priv->apCurrTD[0] = &priv->apTD0Rings[0];
+
+ return;
+error:
+ device_free_td0_ring(priv);
}

static void device_init_td1_ring(struct vnt_private *priv)
@@ -646,7 +661,8 @@ static void device_init_td1_ring(struct vnt_private *priv)
i++, curr += sizeof(struct vnt_tx_desc)) {
desc = &priv->apTD1Rings[i];
desc->td_info = kzalloc(sizeof(*desc->td_info), GFP_KERNEL);
-
+ if (!desc->td_info)
+ goto error;
desc->td_info->buf = priv->tx1_bufs + i * PKT_BUF_SZ;
desc->td_info->buf_dma = priv->tx_bufs_dma1 + i * PKT_BUF_SZ;

@@ -657,6 +673,10 @@ static void device_init_td1_ring(struct vnt_private *priv)
if (i > 0)
priv->apTD1Rings[i-1].next_desc = cpu_to_le32(priv->td1_pool_dma);
priv->apTailTD[1] = priv->apCurrTD[1] = &priv->apTD1Rings[0];
+
+ return;
+error:
+ device_free_td1_ring(priv);
}

static void device_free_td0_ring(struct vnt_private *priv)
@@ -667,8 +687,10 @@ static void device_free_td0_ring(struct vnt_private *priv)
struct vnt_tx_desc *desc = &priv->apTD0Rings[i];
struct vnt_td_info *td_info = desc->td_info;

- dev_kfree_skb(td_info->skb);
- kfree(desc->td_info);
+ if (td_info) {
+ dev_kfree_skb(td_info->skb);
+ kfree(desc->td_info);
+ }
}
}

@@ -680,8 +702,10 @@ static void device_free_td1_ring(struct vnt_private *priv)
struct vnt_tx_desc *desc = &priv->apTD1Rings[i];
struct vnt_td_info *td_info = desc->td_info;

- dev_kfree_skb(td_info->skb);
- kfree(desc->td_info);
+ if (td_info) {
+ dev_kfree_skb(td_info->skb);
+ kfree(desc->td_info);
+ }
}
}

--
1.9.1

2018-03-29 08:02:37

by Jia-Ju Bai

Subject: Re: [PATCH v2] staging: vt6655: check for memory allocation failures

On 2018/3/29 15:22, Ji-Hun Kim wrote:
> There are no null pointer checking on rd_info and td_info values which
> are allocated by kzalloc. It has potential null pointer dereferencing
> issues. Add return when allocation is failed.
>
> Signed-off-by: Ji-Hun Kim <[email protected]>
> ---
>
> Change: since v1:
>
> - Delete WARN_ON which can makes crashes on some machines.
> - Instead of return directly, goto freeing function for freeing previously
> allocated memory in the for loop after kzalloc() failed.
> - In the freeing function, if td_info and rd_info are not allocated, no
> needs to free.
>
> drivers/staging/vt6655/device_main.c | 64 +++++++++++++++++++++++++-----------
> 1 file changed, 44 insertions(+), 20 deletions(-)
>
> diff --git a/drivers/staging/vt6655/device_main.c b/drivers/staging/vt6655/device_main.c
> index fbc4bc6..ecbba43 100644
> --- a/drivers/staging/vt6655/device_main.c
> +++ b/drivers/staging/vt6655/device_main.c
> @@ -539,7 +539,8 @@ static void device_init_rd0_ring(struct vnt_private *priv)
> i ++, curr += sizeof(struct vnt_rx_desc)) {
> desc = &priv->aRD0Ring[i];
> desc->rd_info = kzalloc(sizeof(*desc->rd_info), GFP_KERNEL);
> -
> + if (!desc->rd_info)
> + goto error;
> if (!device_alloc_rx_buf(priv, desc))
> dev_err(&priv->pcid->dev, "can not alloc rx bufs\n");
>
> @@ -550,6 +551,10 @@ static void device_init_rd0_ring(struct vnt_private *priv)
> if (i > 0)
> priv->aRD0Ring[i-1].next_desc = cpu_to_le32(priv->rd0_pool_dma);
> priv->pCurrRD[0] = &priv->aRD0Ring[0];
> +
> + return;
> +error:
> + device_free_rd0_ring(priv);
> }
>

I think you should return an error number here, because
device_init_rd0_ring() is called by vnt_start().
You should also implement error handling code in vnt_start(), and let
vnt_start() returns an error number too.
The same for device_init_rd1_ring(), device_init_td0_ring() and
device_init_td1_ring().


Best wishes,
Jia-Ju Bai

2018-03-29 10:55:00

by ji-hun Kim

Subject: Re: [PATCH v2] staging: vt6655: check for memory allocation failures

2018-03-29 17:00 GMT+09:00 Jia-Ju Bai <[email protected]>:
>
>
>
> On 2018/3/29 15:22, Ji-Hun Kim wrote:
>>
>> There are no null pointer checking on rd_info and td_info values which
>> are allocated by kzalloc. It has potential null pointer dereferencing
>> issues. Add return when allocation is failed.
>>
>> Signed-off-by: Ji-Hun Kim <[email protected]>
>> ---
>>
>> Change: since v1:
>>
>> - Delete WARN_ON which can makes crashes on some machines.
>> - Instead of return directly, goto freeing function for freeing previously
>> allocated memory in the for loop after kzalloc() failed.
>> - In the freeing function, if td_info and rd_info are not allocated, no
>> needs to free.
>>
>> drivers/staging/vt6655/device_main.c | 64 +++++++++++++++++++++++++-----------
>> 1 file changed, 44 insertions(+), 20 deletions(-)
>>
>> diff --git a/drivers/staging/vt6655/device_main.c b/drivers/staging/vt6655/device_main.c
>> index fbc4bc6..ecbba43 100644
>> --- a/drivers/staging/vt6655/device_main.c
>> +++ b/drivers/staging/vt6655/device_main.c
>> @@ -539,7 +539,8 @@ static void device_init_rd0_ring(struct vnt_private *priv)
>> i ++, curr += sizeof(struct vnt_rx_desc)) {
>> desc = &priv->aRD0Ring[i];
>> desc->rd_info = kzalloc(sizeof(*desc->rd_info), GFP_KERNEL);
>> -
>> + if (!desc->rd_info)
>> + goto error;
>> if (!device_alloc_rx_buf(priv, desc))
>> dev_err(&priv->pcid->dev, "can not alloc rx bufs\n");
>> @@ -550,6 +551,10 @@ static void device_init_rd0_ring(struct vnt_private *priv)
>> if (i > 0)
>> priv->aRD0Ring[i-1].next_desc = cpu_to_le32(priv->rd0_pool_dma);
>> priv->pCurrRD[0] = &priv->aRD0Ring[0];
>> +
>> + return;
>> +error:
>> + device_free_rd0_ring(priv);
>> }
>>
>
>
> I think you should return an error number here, because device_init_rd0_ring() is called by vnt_start().
> You should also implement error handling code in vnt_start(), and let vnt_start() returns an error number too.
> The same for device_init_rd1_ring(), device_init_td0_ring() and device_init_td1_ring().
>

Hi Jia-Ju, Thanks for the feedback. All right, those function looks
that needs returns
an error number. I will implement error handling code and send a patch
v3 tomorrow

Best regards,
Ji-Hun