[dpdk-dev] mlx5 FDIR rule comparison issue

DPDK patches and discussions
 help / color / mirror / Atom feed

* [dpdk-dev] mlx5 FDIR rule comparison issue
@ 2019-04-02  7:23 benli ye
  2019-04-02  7:23 ` benli ye
  2019-04-03  8:29 ` benli ye
  0 siblings, 2 replies; 8+ messages in thread
From: benli ye @ 2019-04-02  7:23 UTC (permalink / raw)
  To: dev

Hi Developers,

I am adding two FDIR rule (one is for UDP and the other is for TCP) for mlx5 pmd driver. The rules are listed below.
    struct rte_eth_fdir_filter filt[MAX_FDIR_PROTO] = {
        {
            .input.flow_type = RTE_ETH_FLOW_NONFRAG_IPV4_TCP,
            .input.flow.tcp4_flow.ip.dst_ip = dip,
            .input.flow.tcp4_flow.dst_port = dport,

            .action.behavior = RTE_ETH_FDIR_ACCEPT,
            .action.report_status = RTE_ETH_FDIR_REPORT_ID,
            .soft_id = 0,
        },
        {
            .input.flow_type = RTE_ETH_FLOW_NONFRAG_IPV4_UDP,
            .input.flow.udp4_flow.ip.dst_ip = dip,
            .input.flow.udp4_flow.dst_port = dport,

            .action.behavior = RTE_ETH_FDIR_ACCEPT,
            .action.report_status = RTE_ETH_FDIR_REPORT_ID,
            .soft_id = 1,
        },
    };

However, mlx5 lib prevent me to doing this as when it treats the two rules are the same.

I debugged for a while and found flow_fdir_cmp() didn’t compare the protocol type in field items of struct mlx5_fdir. So should this be a bug for mlx5?

flow_fdir_cmp(const struct mlx5_fdir *f1, const struct mlx5_fdir *f2)
{
	if (FLOW_FDIR_CMP(f1, f2, attr) ||
	    FLOW_FDIR_CMP(f1, f2, l2) ||
	    FLOW_FDIR_CMP(f1, f2, l2_mask) ||
	    FLOW_FDIR_CMP(f1, f2, l3) ||
	    FLOW_FDIR_CMP(f1, f2, l3_mask) ||
	    FLOW_FDIR_CMP(f1, f2, l4) ||
	    FLOW_FDIR_CMP(f1, f2, l4_mask) ||
	    FLOW_FDIR_CMP(f1, f2, actions[0].type))
		return 1;
	if (f1->actions[0].type == RTE_FLOW_ACTION_TYPE_QUEUE &&
	    FLOW_FDIR_CMP(f1, f2, queue))
		return 1;
	return 0;
}

Thanks,
Daniel

^ permalink raw reply	[flat|nested] 8+ messages in thread

* [dpdk-dev] mlx5 FDIR rule comparison issue
  2019-04-02  7:23 [dpdk-dev] mlx5 FDIR rule comparison issue benli ye
@ 2019-04-02  7:23 ` benli ye
  2019-04-03  8:29 ` benli ye
  1 sibling, 0 replies; 8+ messages in thread
From: benli ye @ 2019-04-02  7:23 UTC (permalink / raw)
  To: dev

Hi Developers,

I am adding two FDIR rule (one is for UDP and the other is for TCP) for mlx5 pmd driver. The rules are listed below.
    struct rte_eth_fdir_filter filt[MAX_FDIR_PROTO] = {
        {
            .input.flow_type = RTE_ETH_FLOW_NONFRAG_IPV4_TCP,
            .input.flow.tcp4_flow.ip.dst_ip = dip,
            .input.flow.tcp4_flow.dst_port = dport,

            .action.behavior = RTE_ETH_FDIR_ACCEPT,
            .action.report_status = RTE_ETH_FDIR_REPORT_ID,
            .soft_id = 0,
        },
        {
            .input.flow_type = RTE_ETH_FLOW_NONFRAG_IPV4_UDP,
            .input.flow.udp4_flow.ip.dst_ip = dip,
            .input.flow.udp4_flow.dst_port = dport,

            .action.behavior = RTE_ETH_FDIR_ACCEPT,
            .action.report_status = RTE_ETH_FDIR_REPORT_ID,
            .soft_id = 1,
        },
    };

However, mlx5 lib prevent me to doing this as when it treats the two rules are the same.

I debugged for a while and found flow_fdir_cmp() didn’t compare the protocol type in field items of struct mlx5_fdir. So should this be a bug for mlx5?

flow_fdir_cmp(const struct mlx5_fdir *f1, const struct mlx5_fdir *f2)
{
	if (FLOW_FDIR_CMP(f1, f2, attr) ||
	    FLOW_FDIR_CMP(f1, f2, l2) ||
	    FLOW_FDIR_CMP(f1, f2, l2_mask) ||
	    FLOW_FDIR_CMP(f1, f2, l3) ||
	    FLOW_FDIR_CMP(f1, f2, l3_mask) ||
	    FLOW_FDIR_CMP(f1, f2, l4) ||
	    FLOW_FDIR_CMP(f1, f2, l4_mask) ||
	    FLOW_FDIR_CMP(f1, f2, actions[0].type))
		return 1;
	if (f1->actions[0].type == RTE_FLOW_ACTION_TYPE_QUEUE &&
	    FLOW_FDIR_CMP(f1, f2, queue))
		return 1;
	return 0;
}

Thanks,
Daniel

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [dpdk-dev] mlx5 FDIR rule comparison issue
  2019-04-02  7:23 [dpdk-dev] mlx5 FDIR rule comparison issue benli ye
  2019-04-02  7:23 ` benli ye
@ 2019-04-03  8:29 ` benli ye
  2019-04-03  8:29   ` benli ye
  2019-04-04 11:18   ` Dekel Peled
  1 sibling, 2 replies; 8+ messages in thread
From: benli ye @ 2019-04-03  8:29 UTC (permalink / raw)
  To: dekelp; +Cc: dev

+Dekel

Add Dekel to see if this is an issue.

Thanks,
Daniel

> On Apr 2, 2019, at 3:23 PM, benli ye <danielbenliye@gmail.com> wrote:
> 
> Hi Developers,
> 
> I am adding two FDIR rule (one is for UDP and the other is for TCP) for mlx5 pmd driver. The rules are listed below.
>    struct rte_eth_fdir_filter filt[MAX_FDIR_PROTO] = {
>        {
>            .input.flow_type = RTE_ETH_FLOW_NONFRAG_IPV4_TCP,
>            .input.flow.tcp4_flow.ip.dst_ip = dip,
>            .input.flow.tcp4_flow.dst_port = dport,
> 
>            .action.behavior = RTE_ETH_FDIR_ACCEPT,
>            .action.report_status = RTE_ETH_FDIR_REPORT_ID,
>            .soft_id = 0,
>        },
>        {
>            .input.flow_type = RTE_ETH_FLOW_NONFRAG_IPV4_UDP,
>            .input.flow.udp4_flow.ip.dst_ip = dip,
>            .input.flow.udp4_flow.dst_port = dport,
> 
>            .action.behavior = RTE_ETH_FDIR_ACCEPT,
>            .action.report_status = RTE_ETH_FDIR_REPORT_ID,
>            .soft_id = 1,
>        },
>    };
> 
> However, mlx5 lib prevent me to doing this as when it treats the two rules are the same.
> 
> I debugged for a while and found flow_fdir_cmp() didn’t compare the protocol type in field items of struct mlx5_fdir. So should this be a bug for mlx5?
> 
> flow_fdir_cmp(const struct mlx5_fdir *f1, const struct mlx5_fdir *f2)
> {
> 	if (FLOW_FDIR_CMP(f1, f2, attr) ||
> 	    FLOW_FDIR_CMP(f1, f2, l2) ||
> 	    FLOW_FDIR_CMP(f1, f2, l2_mask) ||
> 	    FLOW_FDIR_CMP(f1, f2, l3) ||
> 	    FLOW_FDIR_CMP(f1, f2, l3_mask) ||
> 	    FLOW_FDIR_CMP(f1, f2, l4) ||
> 	    FLOW_FDIR_CMP(f1, f2, l4_mask) ||
> 	    FLOW_FDIR_CMP(f1, f2, actions[0].type))
> 		return 1;
> 	if (f1->actions[0].type == RTE_FLOW_ACTION_TYPE_QUEUE &&
> 	    FLOW_FDIR_CMP(f1, f2, queue))
> 		return 1;
> 	return 0;
> }
> 
> Thanks,
> Daniel

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [dpdk-dev] mlx5 FDIR rule comparison issue
  2019-04-03  8:29 ` benli ye
@ 2019-04-03  8:29   ` benli ye
  2019-04-04 11:18   ` Dekel Peled
  1 sibling, 0 replies; 8+ messages in thread
From: benli ye @ 2019-04-03  8:29 UTC (permalink / raw)
  To: dekelp; +Cc: dev

+Dekel

Add Dekel to see if this is an issue.

Thanks,
Daniel

> On Apr 2, 2019, at 3:23 PM, benli ye <danielbenliye@gmail.com> wrote:
> 
> Hi Developers,
> 
> I am adding two FDIR rule (one is for UDP and the other is for TCP) for mlx5 pmd driver. The rules are listed below.
>    struct rte_eth_fdir_filter filt[MAX_FDIR_PROTO] = {
>        {
>            .input.flow_type = RTE_ETH_FLOW_NONFRAG_IPV4_TCP,
>            .input.flow.tcp4_flow.ip.dst_ip = dip,
>            .input.flow.tcp4_flow.dst_port = dport,
> 
>            .action.behavior = RTE_ETH_FDIR_ACCEPT,
>            .action.report_status = RTE_ETH_FDIR_REPORT_ID,
>            .soft_id = 0,
>        },
>        {
>            .input.flow_type = RTE_ETH_FLOW_NONFRAG_IPV4_UDP,
>            .input.flow.udp4_flow.ip.dst_ip = dip,
>            .input.flow.udp4_flow.dst_port = dport,
> 
>            .action.behavior = RTE_ETH_FDIR_ACCEPT,
>            .action.report_status = RTE_ETH_FDIR_REPORT_ID,
>            .soft_id = 1,
>        },
>    };
> 
> However, mlx5 lib prevent me to doing this as when it treats the two rules are the same.
> 
> I debugged for a while and found flow_fdir_cmp() didn’t compare the protocol type in field items of struct mlx5_fdir. So should this be a bug for mlx5?
> 
> flow_fdir_cmp(const struct mlx5_fdir *f1, const struct mlx5_fdir *f2)
> {
> 	if (FLOW_FDIR_CMP(f1, f2, attr) ||
> 	    FLOW_FDIR_CMP(f1, f2, l2) ||
> 	    FLOW_FDIR_CMP(f1, f2, l2_mask) ||
> 	    FLOW_FDIR_CMP(f1, f2, l3) ||
> 	    FLOW_FDIR_CMP(f1, f2, l3_mask) ||
> 	    FLOW_FDIR_CMP(f1, f2, l4) ||
> 	    FLOW_FDIR_CMP(f1, f2, l4_mask) ||
> 	    FLOW_FDIR_CMP(f1, f2, actions[0].type))
> 		return 1;
> 	if (f1->actions[0].type == RTE_FLOW_ACTION_TYPE_QUEUE &&
> 	    FLOW_FDIR_CMP(f1, f2, queue))
> 		return 1;
> 	return 0;
> }
> 
> Thanks,
> Daniel


^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [dpdk-dev] mlx5 FDIR rule comparison issue
  2019-04-03  8:29 ` benli ye
  2019-04-03  8:29   ` benli ye
@ 2019-04-04 11:18   ` Dekel Peled
  2019-04-04 11:18     ` Dekel Peled
  2019-04-04 12:12     ` benli ye
  1 sibling, 2 replies; 8+ messages in thread
From: Dekel Peled @ 2019-04-04 11:18 UTC (permalink / raw)
  To: benli ye; +Cc: dev

Hi Daniel,

The flow_director API will be deprecated in the near future.
It was replaced by rte_flow API, which contains much more features, and is fully supported by Mellanox.

It is recommended that you use rte_flow API in your application.

Regards,
Dekel

> -----Original Message-----
> From: benli ye <danielbenliye@gmail.com>
> Sent: Wednesday, April 3, 2019 11:30 AM
> To: Dekel Peled <dekelp@mellanox.com>
> Cc: dev@dpdk.org
> Subject: Re: mlx5 FDIR rule comparison issue
> 
> +Dekel
> 
> Add Dekel to see if this is an issue.
> 
> Thanks,
> Daniel
> 
> > On Apr 2, 2019, at 3:23 PM, benli ye <danielbenliye@gmail.com> wrote:
> >
> > Hi Developers,
> >
> > I am adding two FDIR rule (one is for UDP and the other is for TCP) for mlx5
> pmd driver. The rules are listed below.
> >    struct rte_eth_fdir_filter filt[MAX_FDIR_PROTO] = {
> >        {
> >            .input.flow_type = RTE_ETH_FLOW_NONFRAG_IPV4_TCP,
> >            .input.flow.tcp4_flow.ip.dst_ip = dip,
> >            .input.flow.tcp4_flow.dst_port = dport,
> >
> >            .action.behavior = RTE_ETH_FDIR_ACCEPT,
> >            .action.report_status = RTE_ETH_FDIR_REPORT_ID,
> >            .soft_id = 0,
> >        },
> >        {
> >            .input.flow_type = RTE_ETH_FLOW_NONFRAG_IPV4_UDP,
> >            .input.flow.udp4_flow.ip.dst_ip = dip,
> >            .input.flow.udp4_flow.dst_port = dport,
> >
> >            .action.behavior = RTE_ETH_FDIR_ACCEPT,
> >            .action.report_status = RTE_ETH_FDIR_REPORT_ID,
> >            .soft_id = 1,
> >        },
> >    };
> >
> > However, mlx5 lib prevent me to doing this as when it treats the two rules
> are the same.
> >
> > I debugged for a while and found flow_fdir_cmp() didn’t compare the
> protocol type in field items of struct mlx5_fdir. So should this be a bug for
> mlx5?
> >
> > flow_fdir_cmp(const struct mlx5_fdir *f1, const struct mlx5_fdir *f2)
> > {
> > 	if (FLOW_FDIR_CMP(f1, f2, attr) ||
> > 	    FLOW_FDIR_CMP(f1, f2, l2) ||
> > 	    FLOW_FDIR_CMP(f1, f2, l2_mask) ||
> > 	    FLOW_FDIR_CMP(f1, f2, l3) ||
> > 	    FLOW_FDIR_CMP(f1, f2, l3_mask) ||
> > 	    FLOW_FDIR_CMP(f1, f2, l4) ||
> > 	    FLOW_FDIR_CMP(f1, f2, l4_mask) ||
> > 	    FLOW_FDIR_CMP(f1, f2, actions[0].type))
> > 		return 1;
> > 	if (f1->actions[0].type == RTE_FLOW_ACTION_TYPE_QUEUE &&
> > 	    FLOW_FDIR_CMP(f1, f2, queue))
> > 		return 1;
> > 	return 0;
> > }
> >
> > Thanks,
> > Daniel


^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [dpdk-dev] mlx5 FDIR rule comparison issue
  2019-04-04 11:18   ` Dekel Peled
@ 2019-04-04 11:18     ` Dekel Peled
  2019-04-04 12:12     ` benli ye
  1 sibling, 0 replies; 8+ messages in thread
From: Dekel Peled @ 2019-04-04 11:18 UTC (permalink / raw)
  To: benli ye; +Cc: dev

Hi Daniel,

The flow_director API will be deprecated in the near future.
It was replaced by rte_flow API, which contains much more features, and is fully supported by Mellanox.

It is recommended that you use rte_flow API in your application.

Regards,
Dekel

> -----Original Message-----
> From: benli ye <danielbenliye@gmail.com>
> Sent: Wednesday, April 3, 2019 11:30 AM
> To: Dekel Peled <dekelp@mellanox.com>
> Cc: dev@dpdk.org
> Subject: Re: mlx5 FDIR rule comparison issue
> 
> +Dekel
> 
> Add Dekel to see if this is an issue.
> 
> Thanks,
> Daniel
> 
> > On Apr 2, 2019, at 3:23 PM, benli ye <danielbenliye@gmail.com> wrote:
> >
> > Hi Developers,
> >
> > I am adding two FDIR rule (one is for UDP and the other is for TCP) for mlx5
> pmd driver. The rules are listed below.
> >    struct rte_eth_fdir_filter filt[MAX_FDIR_PROTO] = {
> >        {
> >            .input.flow_type = RTE_ETH_FLOW_NONFRAG_IPV4_TCP,
> >            .input.flow.tcp4_flow.ip.dst_ip = dip,
> >            .input.flow.tcp4_flow.dst_port = dport,
> >
> >            .action.behavior = RTE_ETH_FDIR_ACCEPT,
> >            .action.report_status = RTE_ETH_FDIR_REPORT_ID,
> >            .soft_id = 0,
> >        },
> >        {
> >            .input.flow_type = RTE_ETH_FLOW_NONFRAG_IPV4_UDP,
> >            .input.flow.udp4_flow.ip.dst_ip = dip,
> >            .input.flow.udp4_flow.dst_port = dport,
> >
> >            .action.behavior = RTE_ETH_FDIR_ACCEPT,
> >            .action.report_status = RTE_ETH_FDIR_REPORT_ID,
> >            .soft_id = 1,
> >        },
> >    };
> >
> > However, mlx5 lib prevent me to doing this as when it treats the two rules
> are the same.
> >
> > I debugged for a while and found flow_fdir_cmp() didn’t compare the
> protocol type in field items of struct mlx5_fdir. So should this be a bug for
> mlx5?
> >
> > flow_fdir_cmp(const struct mlx5_fdir *f1, const struct mlx5_fdir *f2)
> > {
> > 	if (FLOW_FDIR_CMP(f1, f2, attr) ||
> > 	    FLOW_FDIR_CMP(f1, f2, l2) ||
> > 	    FLOW_FDIR_CMP(f1, f2, l2_mask) ||
> > 	    FLOW_FDIR_CMP(f1, f2, l3) ||
> > 	    FLOW_FDIR_CMP(f1, f2, l3_mask) ||
> > 	    FLOW_FDIR_CMP(f1, f2, l4) ||
> > 	    FLOW_FDIR_CMP(f1, f2, l4_mask) ||
> > 	    FLOW_FDIR_CMP(f1, f2, actions[0].type))
> > 		return 1;
> > 	if (f1->actions[0].type == RTE_FLOW_ACTION_TYPE_QUEUE &&
> > 	    FLOW_FDIR_CMP(f1, f2, queue))
> > 		return 1;
> > 	return 0;
> > }
> >
> > Thanks,
> > Daniel


^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [dpdk-dev] mlx5 FDIR rule comparison issue
  2019-04-04 11:18   ` Dekel Peled
  2019-04-04 11:18     ` Dekel Peled
@ 2019-04-04 12:12     ` benli ye
  2019-04-04 12:12       ` benli ye
  1 sibling, 1 reply; 8+ messages in thread
From: benli ye @ 2019-04-04 12:12 UTC (permalink / raw)
  To: Dekel Peled; +Cc: dev

Thanks Dekel for your information. I will have a try.

Bests,
Daniel

> On Apr 4, 2019, at 7:18 PM, Dekel Peled <dekelp@mellanox.com> wrote:
> 
> Hi Daniel,
> 
> The flow_director API will be deprecated in the near future.
> It was replaced by rte_flow API, which contains much more features, and is fully supported by Mellanox.
> 
> It is recommended that you use rte_flow API in your application.
> 
> Regards,
> Dekel
> 
>> -----Original Message-----
>> From: benli ye <danielbenliye@gmail.com>
>> Sent: Wednesday, April 3, 2019 11:30 AM
>> To: Dekel Peled <dekelp@mellanox.com>
>> Cc: dev@dpdk.org
>> Subject: Re: mlx5 FDIR rule comparison issue
>> 
>> +Dekel
>> 
>> Add Dekel to see if this is an issue.
>> 
>> Thanks,
>> Daniel
>> 
>>> On Apr 2, 2019, at 3:23 PM, benli ye <danielbenliye@gmail.com> wrote:
>>> 
>>> Hi Developers,
>>> 
>>> I am adding two FDIR rule (one is for UDP and the other is for TCP) for mlx5
>> pmd driver. The rules are listed below.
>>>   struct rte_eth_fdir_filter filt[MAX_FDIR_PROTO] = {
>>>       {
>>>           .input.flow_type = RTE_ETH_FLOW_NONFRAG_IPV4_TCP,
>>>           .input.flow.tcp4_flow.ip.dst_ip = dip,
>>>           .input.flow.tcp4_flow.dst_port = dport,
>>> 
>>>           .action.behavior = RTE_ETH_FDIR_ACCEPT,
>>>           .action.report_status = RTE_ETH_FDIR_REPORT_ID,
>>>           .soft_id = 0,
>>>       },
>>>       {
>>>           .input.flow_type = RTE_ETH_FLOW_NONFRAG_IPV4_UDP,
>>>           .input.flow.udp4_flow.ip.dst_ip = dip,
>>>           .input.flow.udp4_flow.dst_port = dport,
>>> 
>>>           .action.behavior = RTE_ETH_FDIR_ACCEPT,
>>>           .action.report_status = RTE_ETH_FDIR_REPORT_ID,
>>>           .soft_id = 1,
>>>       },
>>>   };
>>> 
>>> However, mlx5 lib prevent me to doing this as when it treats the two rules
>> are the same.
>>> 
>>> I debugged for a while and found flow_fdir_cmp() didn’t compare the
>> protocol type in field items of struct mlx5_fdir. So should this be a bug for
>> mlx5?
>>> 
>>> flow_fdir_cmp(const struct mlx5_fdir *f1, const struct mlx5_fdir *f2)
>>> {
>>> 	if (FLOW_FDIR_CMP(f1, f2, attr) ||
>>> 	    FLOW_FDIR_CMP(f1, f2, l2) ||
>>> 	    FLOW_FDIR_CMP(f1, f2, l2_mask) ||
>>> 	    FLOW_FDIR_CMP(f1, f2, l3) ||
>>> 	    FLOW_FDIR_CMP(f1, f2, l3_mask) ||
>>> 	    FLOW_FDIR_CMP(f1, f2, l4) ||
>>> 	    FLOW_FDIR_CMP(f1, f2, l4_mask) ||
>>> 	    FLOW_FDIR_CMP(f1, f2, actions[0].type))
>>> 		return 1;
>>> 	if (f1->actions[0].type == RTE_FLOW_ACTION_TYPE_QUEUE &&
>>> 	    FLOW_FDIR_CMP(f1, f2, queue))
>>> 		return 1;
>>> 	return 0;
>>> }
>>> 
>>> Thanks,
>>> Daniel
> 

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [dpdk-dev] mlx5 FDIR rule comparison issue
  2019-04-04 12:12     ` benli ye
@ 2019-04-04 12:12       ` benli ye
  0 siblings, 0 replies; 8+ messages in thread
From: benli ye @ 2019-04-04 12:12 UTC (permalink / raw)
  To: Dekel Peled; +Cc: dev

Thanks Dekel for your information. I will have a try.

Bests,
Daniel

> On Apr 4, 2019, at 7:18 PM, Dekel Peled <dekelp@mellanox.com> wrote:
> 
> Hi Daniel,
> 
> The flow_director API will be deprecated in the near future.
> It was replaced by rte_flow API, which contains much more features, and is fully supported by Mellanox.
> 
> It is recommended that you use rte_flow API in your application.
> 
> Regards,
> Dekel
> 
>> -----Original Message-----
>> From: benli ye <danielbenliye@gmail.com>
>> Sent: Wednesday, April 3, 2019 11:30 AM
>> To: Dekel Peled <dekelp@mellanox.com>
>> Cc: dev@dpdk.org
>> Subject: Re: mlx5 FDIR rule comparison issue
>> 
>> +Dekel
>> 
>> Add Dekel to see if this is an issue.
>> 
>> Thanks,
>> Daniel
>> 
>>> On Apr 2, 2019, at 3:23 PM, benli ye <danielbenliye@gmail.com> wrote:
>>> 
>>> Hi Developers,
>>> 
>>> I am adding two FDIR rule (one is for UDP and the other is for TCP) for mlx5
>> pmd driver. The rules are listed below.
>>>   struct rte_eth_fdir_filter filt[MAX_FDIR_PROTO] = {
>>>       {
>>>           .input.flow_type = RTE_ETH_FLOW_NONFRAG_IPV4_TCP,
>>>           .input.flow.tcp4_flow.ip.dst_ip = dip,
>>>           .input.flow.tcp4_flow.dst_port = dport,
>>> 
>>>           .action.behavior = RTE_ETH_FDIR_ACCEPT,
>>>           .action.report_status = RTE_ETH_FDIR_REPORT_ID,
>>>           .soft_id = 0,
>>>       },
>>>       {
>>>           .input.flow_type = RTE_ETH_FLOW_NONFRAG_IPV4_UDP,
>>>           .input.flow.udp4_flow.ip.dst_ip = dip,
>>>           .input.flow.udp4_flow.dst_port = dport,
>>> 
>>>           .action.behavior = RTE_ETH_FDIR_ACCEPT,
>>>           .action.report_status = RTE_ETH_FDIR_REPORT_ID,
>>>           .soft_id = 1,
>>>       },
>>>   };
>>> 
>>> However, mlx5 lib prevent me to doing this as when it treats the two rules
>> are the same.
>>> 
>>> I debugged for a while and found flow_fdir_cmp() didn’t compare the
>> protocol type in field items of struct mlx5_fdir. So should this be a bug for
>> mlx5?
>>> 
>>> flow_fdir_cmp(const struct mlx5_fdir *f1, const struct mlx5_fdir *f2)
>>> {
>>> 	if (FLOW_FDIR_CMP(f1, f2, attr) ||
>>> 	    FLOW_FDIR_CMP(f1, f2, l2) ||
>>> 	    FLOW_FDIR_CMP(f1, f2, l2_mask) ||
>>> 	    FLOW_FDIR_CMP(f1, f2, l3) ||
>>> 	    FLOW_FDIR_CMP(f1, f2, l3_mask) ||
>>> 	    FLOW_FDIR_CMP(f1, f2, l4) ||
>>> 	    FLOW_FDIR_CMP(f1, f2, l4_mask) ||
>>> 	    FLOW_FDIR_CMP(f1, f2, actions[0].type))
>>> 		return 1;
>>> 	if (f1->actions[0].type == RTE_FLOW_ACTION_TYPE_QUEUE &&
>>> 	    FLOW_FDIR_CMP(f1, f2, queue))
>>> 		return 1;
>>> 	return 0;
>>> }
>>> 
>>> Thanks,
>>> Daniel
> 


^ permalink raw reply	[flat|nested] 8+ messages in thread

end of thread, other threads:[~2019-04-04 12:12 UTC | newest]

Thread overview: 8+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-04-02  7:23 [dpdk-dev] mlx5 FDIR rule comparison issue benli ye
2019-04-02  7:23 ` benli ye
2019-04-03  8:29 ` benli ye
2019-04-03  8:29   ` benli ye
2019-04-04 11:18   ` Dekel Peled
2019-04-04 11:18     ` Dekel Peled
2019-04-04 12:12     ` benli ye
2019-04-04 12:12       ` benli ye

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).