[dpdk-dev] [dpdk-dev v2 3/4] app/testpmd: support GTP PDU type

Ori Kam orika at mellanox.com
Tue Apr 14 07:57:29 CEST 2020


Hi Jeff,
No problem about the review. I will see v3 right?

From: Jeff Guo <jia.guo at intel.com>
Sent: Tuesday, April 14, 2020 6:06 AM
To: Ori Kam <orika at mellanox.com>; xiaolong.ye at intel.com; qi.z.zhang at intel.com
Cc: dev at dpdk.org; jingjing.wu at intel.com; yahui.cao at intel.com; simei.su at intel.com
Subject: Re: [dpdk-dev] [dpdk-dev v2 3/4] app/testpmd: support GTP PDU type


hi, Ori


On 4/12/2020 5:58 PM, Ori Kam wrote:
Hi Jeff,

For some reason I got the mail in HTML format
So please see my comments marked by [Ori]

Thanks,
Ori

From: Jeff Guo <jia.guo at intel.com><mailto:jia.guo at intel.com>
Sent: Tuesday, April 7, 2020 8:37 AM
To: Ori Kam <orika at mellanox.com><mailto:orika at mellanox.com>; xiaolong.ye at intel.com<mailto:xiaolong.ye at intel.com>; qi.z.zhang at intel.com<mailto:qi.z.zhang at intel.com>
Cc: dev at dpdk.org<mailto:dev at dpdk.org>; jingjing.wu at intel.com<mailto:jingjing.wu at intel.com>; yahui.cao at intel.com<mailto:yahui.cao at intel.com>; simei.su at intel.com<mailto:simei.su at intel.com>
Subject: Re: [dpdk-dev] [dpdk-dev v2 3/4] app/testpmd: support GTP PDU type


hi, Ori
On 4/5/2020 11:56 PM, Ori Kam wrote:

Hi Jeff,



-----Original Message-----

From: Jeff Guo <jia.guo at intel.com><mailto:jia.guo at intel.com>

Sent: Tuesday, March 31, 2020 11:50 AM

To: Ori Kam <orika at mellanox.com><mailto:orika at mellanox.com>; xiaolong.ye at intel.com<mailto:xiaolong.ye at intel.com>;

qi.z.zhang at intel.com<mailto:qi.z.zhang at intel.com>

Cc: dev at dpdk.org<mailto:dev at dpdk.org>; jingjing.wu at intel.com<mailto:jingjing.wu at intel.com>; yahui.cao at intel.com<mailto:yahui.cao at intel.com>;

simei.su at intel.com<mailto:simei.su at intel.com>

Subject: Re: [dpdk-dev] [dpdk-dev v2 3/4] app/testpmd: support GTP PDU type



yes, Ori, please check the comment below.





On 3/30/2020 6:18 PM, Ori Kam wrote:

Hi Jeff,



My name is Ori 😊



I'm not an expert in GTP so this is just my thinking and maybe I'm

missing something, this is why a good explanation helps 😊



-----Original Message-----

From: Jeff Guo <jia.guo at intel.com><mailto:jia.guo at intel.com>

Sent: Monday, March 30, 2020 11:30 AM

To: Ori Kam <orika at mellanox.com><mailto:orika at mellanox.com>; xiaolong.ye at intel.com<mailto:xiaolong.ye at intel.com>;

qi.z.zhang at intel.com<mailto:qi.z.zhang at intel.com>

Cc: dev at dpdk.org<mailto:dev at dpdk.org>; jingjing.wu at intel.com<mailto:jingjing.wu at intel.com>; yahui.cao at intel.com<mailto:yahui.cao at intel.com>;

simei.su at intel.com<mailto:simei.su at intel.com>

Subject: Re: [dpdk-dev] [dpdk-dev v2 3/4] app/testpmd: support GTP PDU

type



hi, orika





On 3/29/2020 4:44 PM, Ori Kam wrote:

Hi Jeff,





-----Original Message-----

From: dev <dev-bounces at dpdk.org><mailto:dev-bounces at dpdk.org> On Behalf Of Jeff Guo

Sent: Thursday, March 26, 2020 6:41 PM

To: xiaolong.ye at intel.com<mailto:xiaolong.ye at intel.com>; qi.z.zhang at intel.com<mailto:qi.z.zhang at intel.com>

Cc: dev at dpdk.org<mailto:dev at dpdk.org>; jingjing.wu at intel.com<mailto:jingjing.wu at intel.com>; yahui.cao at intel.com<mailto:yahui.cao at intel.com>;

simei.su at intel.com<mailto:simei.su at intel.com>; jia.guo at intel.com<mailto:jia.guo at intel.com>

Subject: [dpdk-dev] [dpdk-dev v2 3/4] app/testpmd: support GTP PDU type



Add gtp pdu type configure in the cmdline.

Why not use ITEM_GTP_PSC_PDU?



I guess you mean ITEM_GTP_PSC_PDU_T, rihgt? We know  we have got

ITEM_GTP_PSC_QFI/ITEM_GTP_PSC_PDU_T but not define the



spec for them, so what i use is add the spec into the ITEM_GTP_PSC_PDU_T

to let the pdu type to be configured.



Yes you are correct, from rte_flow we have the

RTE_FLOW_ITEM_TYPE_GTP_PSC

Item that include pdu_type. This is the field you need right?



In testpmd we have the ITEM_GTP_PSC_PDU_T which should support adding

the pdu type.

Basically you just need to type the following cmd line:

flow create 0 ingress pattern gtp_psc pdu_t is xxx

if this command is not working we need to understand why.







please check the part before this patch as below:



         [ITEM_GTP_PSC_PDU_T] = {

                 .name = "pdu_t",

                 .help = "PDU type",

                .next = NEXT(item_gtp_psc, NEXT_ENTRY(UNSIGNED),

item_param),



sure, we got the ITEM_GTP_PSC_PDU_T at prior but the NEXT_ENTRY is

UNSIGNED, that means we still not implement



Sorry I don't understand your comment, what do you mean it is not implemented?

Yes it means that the parameter is should  be unsigned value.



I mean that if it is a unsigned value, user could not set the pdu_t to be a 0 or 1, or any other we

define for that.





the spec to let the pdu type to be configurable, so what the patch do is

to fix this issue.



What do you mean configurable?



Lets start at the beginning, maybe I'm just missing some key point.

What is the PDU type? What values can he hold?

How do you want the command to look like?



the command should be like as below

flow create 0 ingress pattern eth / ipv4 / udp / gtpu / gtp_psc pdu_t is 0 / ipv4 / end actions rss types ipv4-udp l3-dst-only end  key_len 0 queues end / end

It is eventually the same as you described about the command before.  User could set pdu_t to be 0 or 1, so what is need is modify

NEXT_ENTRY(UNSIGNED) to be "SIGNED" and defined.



[Ori]  I agree about the command look.

Can  pdu_t can be only 0 or 1?

Also I don’t understand why you need signed? Even if you add the GTP_PSC_PDU_T as int? and not unsinged?

All other items are unsinged. I also don’t see any reason to create the new type unless, you want the help line.

In any case, I feel that this patch already waste a lot of time.
(for both of us)

Please consider again if the new class is necessary, and if it should be signed.

If so you have my ack.

Acked-by: Ori Kam orika at mellanox.com<mailto:orika at mellanox.com>

Best,



I guess i know your meaning now, and after check it again, what we though is the same but the new class is absolutely not necessary,

you are totally right here Ori, thanks for your targeting it and good review.



Ori













Signed-off-by: Jeff Guo <jia.guo at intel.com><mailto:jia.guo at intel.com>

---

v1:

no change

---

   app/test-pmd/cmdline_flow.c | 11 ++++++++++-

   1 file changed, 10 insertions(+), 1 deletion(-)



diff --git a/app/test-pmd/cmdline_flow.c b/app/test-pmd/cmdline_flow.c

index a78154502..c1bd02919 100644

--- a/app/test-pmd/cmdline_flow.c

+++ b/app/test-pmd/cmdline_flow.c

@@ -49,6 +49,7 @@ enum index {

    PORT_ID,

    GROUP_ID,

    PRIORITY_LEVEL,

+ GTP_PSC_PDU_T,



         /* Top-level command. */

         SET,

@@ -1626,6 +1627,13 @@ static const struct token token_list[] = {

                 .call = parse_int,

                 .comp = comp_none,

         },

+ [GTP_PSC_PDU_T] = {

+        .name = "{GTPU pdu type}",

+        .type = "INTEGER",

+        .help = "gtpu pdu uplink/downlink identifier",

+        .call = parse_int,

+        .comp = comp_none,

+ },

Why is this created at this level?

This looks like is should be written totally differently.



As i said above,  the item we got but spec or say next token still need

to be added, do you mean it should not in the group of Common tokens? If

so, let me think about that, and please explicit your proposal if you

already have one.



Please see above response.



         /* Top-level command. */

         [FLOW] = {

                 .name = "flow",

@@ -2615,7 +2623,8 @@ static const struct token token_list[] = {

    [ITEM_GTP_PSC_PDU_T] = {

                 .name = "pdu_t",

                 .help = "PDU type",

-        .next = NEXT(item_gtp_psc, NEXT_ENTRY(UNSIGNED),

item_param),

+        .next = NEXT(item_gtp_psc, NEXT_ENTRY(GTP_PSC_PDU_T),

+                     item_param),

                 .args = ARGS(ARGS_ENTRY_HTON(struct

rte_flow_item_gtp_psc,

                                        pdu_type)),

         },

--

2.20.1


More information about the dev mailing list