[PATCH] dmaengine: Add a null pointer check to the dma_request_chan

[PATCH] dmaengine: Add a null pointer check to the dma_request_chan

[Kunwu Chan]

kasprintf() returns a pointer to dynamically allocated memory
which can be NULL upon failure. Ensure the allocation was successful
by checking the pointer validity.

Signed-off-by: Kunwu Chan <chentao@kylinos.cn>
---
 drivers/dma/dmaengine.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c
index 491b22240221..a6f808d13aa4 100644
--- a/drivers/dma/dmaengine.c
+++ b/drivers/dma/dmaengine.c
@@ -856,6 +856,8 @@ struct dma_chan *dma_request_chan(struct device *dev, const char *name)
 #ifdef CONFIG_DEBUG_FS
 	chan->dbg_client_name = kasprintf(GFP_KERNEL, "%s:%s", dev_name(dev),
 					  name);
+	if (!chan->dbg_client_name)
+		return chan;
 #endif
 
 	chan->name = kasprintf(GFP_KERNEL, "dma:%s", name);
-- 
2.39.2

Re: [PATCH] dmaengine: Add a null pointer check to the dma_request_chan

[Vinod Koul]

On 18-01-24, 11:30, Kunwu Chan wrote:
> kasprintf() returns a pointer to dynamically allocated memory
> which can be NULL upon failure. Ensure the allocation was successful
> by checking the pointer validity.
> 
> Signed-off-by: Kunwu Chan <chentao@kylinos.cn>
> ---
>  drivers/dma/dmaengine.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c
> index 491b22240221..a6f808d13aa4 100644
> --- a/drivers/dma/dmaengine.c
> +++ b/drivers/dma/dmaengine.c
> @@ -856,6 +856,8 @@ struct dma_chan *dma_request_chan(struct device *dev, const char *name)
>  #ifdef CONFIG_DEBUG_FS
>  	chan->dbg_client_name = kasprintf(GFP_KERNEL, "%s:%s", dev_name(dev),
>  					  name);
> +	if (!chan->dbg_client_name)
> +		return chan;

That is wrong, you cant return a half done channel here
Pls see rest of the code for reference

>  #endif
>  
>  	chan->name = kasprintf(GFP_KERNEL, "dma:%s", name);
> -- 
> 2.39.2

-- 
~Vinod

Re: [PATCH] dmaengine: Add a null pointer check to the dma_request_chan

[Kunwu Chan]

Thanks for your reply.

On 2024/1/22 19:25, Vinod Koul wrote:
> On 18-01-24, 11:30, Kunwu Chan wrote:
>> kasprintf() returns a pointer to dynamically allocated memory
>> which can be NULL upon failure. Ensure the allocation was successful
>> by checking the pointer validity.
>>
>> Signed-off-by: Kunwu Chan <chentao@kylinos.cn>
>> ---
>>   drivers/dma/dmaengine.c | 2 ++
>>   1 file changed, 2 insertions(+)
>>
>> diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c
>> index 491b22240221..a6f808d13aa4 100644
>> --- a/drivers/dma/dmaengine.c
>> +++ b/drivers/dma/dmaengine.c
>> @@ -856,6 +856,8 @@ struct dma_chan *dma_request_chan(struct device *dev, const char *name)
>>   #ifdef CONFIG_DEBUG_FS
>>   	chan->dbg_client_name = kasprintf(GFP_KERNEL, "%s:%s", dev_name(dev),
>>   					  name);
>> +	if (!chan->dbg_client_name)
>> +		return chan;
> 
> That is wrong, you cant return a half done channel here
> Pls see rest of the code for reference
I handle the error path like line 862 in 'dma_request_chan', though i 
think it's not so perfect. Why it just return chan in line 862 when in 
error? I would be very grateful if it could help me solve this puzzle

dma_request_chan find and return the ``name`` DMA channel associated 
with the 'dev' device. The association is done via DT, ACPI or board 
file based dma_slave_map matching table.

So when error occurs, just do the opposite, like in 'dma_release_channel'.

look likes:
---

         chan->dbg_client_name = kasprintf(GFP_KERNEL, "%s:%s", 
dev_name(dev),
                                           name);
+       if (!chan->dbg_client_name) {
+               dma_cap_clear(DMA_SLAVE, chan->device->cap_mask);
+               dma_cap_zero(chan->device->cap_mask);
+               dma_chan_put(chan);
+               chan->slave = NULL;
+               chan = NULL;
+               return chan;
+       }

---
It's that right by clear the cap and set chan to NULL？


> 
>>   #endif
>>   
>>   	chan->name = kasprintf(GFP_KERNEL, "dma:%s", name);
>> -- 
>> 2.39.2
> 
-- 
Thanks,
   Kunwu