Re: [dpdk-stable] [PATCH 18.11] net/cxgbe: fix L2T leak in filter error and free path

[Kevin Traynor <ktraynor@redhat.com>]

On 04/09/2020 15:33, Rahul Lakkireddy wrote:
> On Friday, September 09/04/20, 2020 at 15:10:43 +0100, Kevin Traynor wrote:
>> On 04/09/2020 13:34, Rahul Lakkireddy wrote:
>>> [ upstream commit e0e1496be5fcabd31807ef9981c421963a9f8578 ]
>>>
>>> Free up Layer 2 Table (L2T) entry properly during filter create
>>> failure and filter delete.
>>>
>>> Fixes: 1decc62b1cbe ("net/cxgbe: add flow operations to offload VLAN actions")
>>>
>>> Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
>>> ---
>>
>> Thanks Rahul. There was a bit of fuzz when applying because I had
>> rebased below locally.
>>
>> net/cxgbe: fix double MPS alloc by flow validate and create
>>
>> [ upstream commit be5f4d5ced10f9d7edc8ba4b3f1d98b9d476d367 ]
>>
>> Both commits are in the queue, please take a look and let me know if any
>> issue.
>>
>> https://github.com/kevintraynor/dpdk-stable-queue/commit/b82919008c7462a0f2e117e65210754ccb7470f1
>>
>> https://github.com/kevintraynor/dpdk-stable-queue/commit/7d482e8b14c7436f73e0497b6f1a0f00e0dc6eba
>>
>> Kevin.
>>
> 
> Hi Kevin,
> 
> Could you move the cxgbe_l2t_release() above the cxgbe_mpstcam_remove()?
> Like below:
> 

Sure

> if (f->l2t)
> 	cxgbe_l2t_release(f->l2t);
> 
> if (f->fs.mask.macidx)
> 	cxgbe_mpstcam_remove(pi, f->fs.val.macidx);
> 
> Let me know if you want me to resend the patch on top of
> dpdk-stable-queue tree with the fuzz fixed.
> 

No need. I've done it here:
https://github.com/kevintraynor/dpdk-stable-queue/commit/02647630a2f0be86f3daea2cc753c90357d46f85

> Thanks,
> Rahul
> 
>>>  drivers/net/cxgbe/cxgbe_filter.c | 29 ++++++++++++++++-------------
>>>  1 file changed, 16 insertions(+), 13 deletions(-)
>>>
>>> diff --git a/drivers/net/cxgbe/cxgbe_filter.c b/drivers/net/cxgbe/cxgbe_filter.c
>>> index 489c23249..0b6629895 100644
>>> --- a/drivers/net/cxgbe/cxgbe_filter.c
>>> +++ b/drivers/net/cxgbe/cxgbe_filter.c
>>> @@ -276,6 +276,9 @@ static void clear_filter(struct filter_entry *f)
>>>  	if (f->clipt)
>>>  		cxgbe_clip_release(f->dev, f->clipt);
>>>  
>>> +	if (f->l2t)
>>> +		cxgbe_l2t_release(f->l2t);
>>> +
>>>  	/* The zeroing of the filter rule below clears the filter valid,
>>>  	 * pending, locked flags etc. so it's all we need for
>>>  	 * this operation.
>>> @@ -711,19 +714,6 @@ static int set_filter_wr(struct rte_eth_dev *dev, unsigned int fidx)
>>>  	unsigned int port_id = ethdev2pinfo(dev)->port_id;
>>>  	int ret;
>>>  
>>> -	/*
>>> -	 * If the new filter requires loopback Destination MAC and/or VLAN
>>> -	 * rewriting then we need to allocate a Layer 2 Table (L2T) entry for
>>> -	 * the filter.
>>> -	 */
>>> -	if (f->fs.newvlan) {
>>> -		/* allocate L2T entry for new filter */
>>> -		f->l2t = cxgbe_l2t_alloc_switching(f->dev, f->fs.vlan,
>>> -						   f->fs.eport, f->fs.dmac);
>>> -		if (!f->l2t)
>>> -			return -ENOMEM;
>>> -	}
>>> -
>>>  	ctrlq = &adapter->sge.ctrlq[port_id];
>>>  	mbuf = rte_pktmbuf_alloc(ctrlq->mb_pool);
>>>  	if (!mbuf) {
>>> @@ -1015,6 +1005,19 @@ int cxgbe_set_filter(struct rte_eth_dev *dev, unsigned int filter_id,
>>>  		}
>>>  	}
>>>  
>>> +	/* If the new filter requires loopback Destination MAC and/or VLAN
>>> +	 * rewriting then we need to allocate a Layer 2 Table (L2T) entry for
>>> +	 * the filter.
>>> +	 */
>>> +	if (f->fs.newvlan) {
>>> +		f->l2t = cxgbe_l2t_alloc_switching(f->dev, f->fs.vlan,
>>> +						   f->fs.eport, f->fs.dmac);
>>> +		if (!f->l2t) {
>>> +			ret = -ENOMEM;
>>> +			goto free_tid;
>>> +		}
>>> +	}
>>> +
>>>  	/*
>>>  	 * Attempt to set the filter.  If we don't succeed, we clear
>>>  	 * it and return the failure.
>>>
>>
>