[dpdk-dev] [PATCH v3 13/13] vdpa/mlx5: disable ROCE

[Maxime Coquelin]

On 2/3/20 1:44 PM, Matan Azrad wrote:
> 
> 
> From: Maxime Coquelin
>> Sent: Monday, February 3, 2020 1:00 PM
>> To: Matan Azrad <matan at mellanox.com>; dev at dpdk.org; Slava Ovsiienko
>> <viacheslavo at mellanox.com>
>> Subject: Re: [PATCH v3 13/13] vdpa/mlx5: disable ROCE
>>
>>
>>
>> On 2/3/20 10:27 AM, Maxime Coquelin wrote:
>>> Hi Matan,
>>>
>>> On 2/2/20 5:03 PM, Matan Azrad wrote:
>>>> In order to support virtio queue creation by the FW, ROCE mode should
>>>> be disabled in the device.
>>>>
>>>> Do it by netlink which is like the devlink tool commands:
>>>> 	1. devlink dev param set pci/[pci] name enable_roce value false
>>>> 	   cmode driverinit
>>>>     	2. devlink dev reload pci/[pci]
>>>> Or by sysfs which is like:
>>>> 	echo 0 >  /sys/bus/pci/devices/[pci]/roce_enable
>>>>
>>>> The IB device is matched again after ROCE disabling.
>>>>
>>>> Signed-off-by: Matan Azrad <matan at mellanox.com>
>>>> Acked-by: Viacheslav Ovsiienko <viacheslavo at mellanox.com>
>>>> Acked-by: Maxime Coquelin <maxime.coquelin at redhat.com>
>>>> ---
>>>>  drivers/vdpa/mlx5/Makefile    |   2 +-
>>>>  drivers/vdpa/mlx5/meson.build |   2 +-
>>>>  drivers/vdpa/mlx5/mlx5_vdpa.c | 191
>>>> ++++++++++++++++++++++++++++++++++--------
>>>>  3 files changed, 160 insertions(+), 35 deletions(-)
>>> ...
>>>> diff --git a/drivers/vdpa/mlx5/mlx5_vdpa.c
>>>> b/drivers/vdpa/mlx5/mlx5_vdpa.c index 57619d2..710f305 100644
>>>> --- a/drivers/vdpa/mlx5/mlx5_vdpa.c
>>>> +++ b/drivers/vdpa/mlx5/mlx5_vdpa.c
>>>
>>> ...
>>>
>>>> @@ -246,8 +389,7 @@
>>>>  mlx5_vdpa_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
>>>>  		    struct rte_pci_device *pci_dev __rte_unused)  {
>>>> -	struct ibv_device **ibv_list;
>>>> -	struct ibv_device *ibv_match = NULL;
>>>> +	struct ibv_device *ibv;
>>>>  	struct mlx5_vdpa_priv *priv = NULL;
>>>>  	struct ibv_context *ctx = NULL;
>>>>  	struct mlx5_hca_attr attr;
>>>> @@ -258,42 +400,25 @@
>>>>  			" driver.");
>>>>  		return 1;
>>>>  	}
>>>> -	errno = 0;
>>>> -	ibv_list = mlx5_glue->get_device_list(&ret);
>>>> -	if (!ibv_list) {
>>>> -		rte_errno = ENOSYS;
>>>> -		DRV_LOG(ERR, "Failed to get device list, is ib_uverbs
>> loaded?");
>>>> +	ibv = mlx5_vdpa_get_ib_device_match(&pci_dev->addr);
>>>> +	if (!ibv) {
>>>> +		DRV_LOG(ERR, "No matching IB device for PCI slot "
>>>> +			PCI_PRI_FMT ".", pci_dev->addr.domain,
>>>> +			pci_dev->addr.bus, pci_dev->addr.devid,
>>>> +			pci_dev->addr.function);
>>>>  		return -rte_errno;
>>>> -	}
>>>> -	while (ret-- > 0) {
>>>> -		struct rte_pci_addr pci_addr;
>>>> -
>>>> -		DRV_LOG(DEBUG, "Checking device \"%s\"..", ibv_list[ret]-
>>> name);
>>>> -		if (mlx5_dev_to_pci_addr(ibv_list[ret]->ibdev_path,
>> &pci_addr))
>>>> -			continue;
>>>> -		if (pci_dev->addr.domain != pci_addr.domain ||
>>>> -		    pci_dev->addr.bus != pci_addr.bus ||
>>>> -		    pci_dev->addr.devid != pci_addr.devid ||
>>>> -		    pci_dev->addr.function != pci_addr.function)
>>>> -			continue;
>>>> +	} else {
>>>>  		DRV_LOG(INFO, "PCI information matches for device
>> \"%s\".",
>>>> -			ibv_list[ret]->name);
>>>> -		ibv_match = ibv_list[ret];
>>>> -		break;
>>>> +			ibv->name);
>>>>  	}
>>>> -	mlx5_glue->free_device_list(ibv_list);
>>>> -	if (!ibv_match) {
>>>> -		DRV_LOG(ERR, "No matching IB device for PCI slot "
>>>> -			"%" SCNx32 ":%" SCNx8 ":%" SCNx8 ".%" SCNx8 ".",
>>>> -			pci_dev->addr.domain, pci_dev->addr.bus,
>>>> -			pci_dev->addr.devid, pci_dev->addr.function);
>>>> -		rte_errno = ENOENT;
>>>> -		return -rte_errno;
>>>> +	if (mlx5_vdpa_roce_disable(&pci_dev->addr, &ibv) != 0) {
>>>> +		DRV_LOG(WARNING, "Failed to disable ROCE for \"%s\".",
>>>> +			ibv->name);
>>>> +		//return -rte_errno;
>>>>  	}
>>>
>>> Is that commented return expected?
>>>
>>
>> Please let me know if I should remove the comment, or remove the return.
> 
> Sorry, forgot the comment , good catch!
> It should not be comment just need to remove "//".

Thanks Matan.

> Can you do it in integration?
Sure, will do now.

Maxime

>>
>> Thanks,
>> Maxime
>