From: "Dumitrescu, Cristian" <cristian.dumitrescu@intel.com>
To: WeiJie Zhuang <zhuangwj@gmail.com>
Cc: "dev@dpdk.org" <dev@dpdk.org>,
"Singh, Jasvinder" <jasvinder.singh@intel.com>,
"Yigit, Ferruh" <ferruh.yigit@intel.com>
Subject: Re: [dpdk-dev] [PATCH v3 1/3] port: add kni interface support
Date: Sat, 18 Jun 2016 16:44:52 +0000 [thread overview]
Message-ID: <3EB4FA525960D640B5BDFFD6A3D8912647A08FAC@IRSMSX108.ger.corp.intel.com> (raw)
In-Reply-To: <1466076423-8680-1-git-send-email-zhuangwj@gmail.com>
Hi Ethan,
Thank you, here are some comments inlined below.
Please reorganize this patch in a slightly different way to look similar to other DPDK patch sets and also ease up the integration work for Thomas:
Patch 0: I suggest adding a cover letter;
Patch 1: all librte_port changes (rte_port_kni.h, rte_port_kni.c, Makefile, rte_port_version.map), including the "nodrop" KNI port version
Patch 2: all ip_pipeline app changes
Patch 3: ip_pipeline app kni.cfg file
Patch 4: Documentation changes
> -----Original Message-----
> From: WeiJie Zhuang [mailto:zhuangwj@gmail.com]
> Sent: Thursday, June 16, 2016 12:27 PM
> To: Dumitrescu, Cristian <cristian.dumitrescu@intel.com>
> Cc: dev@dpdk.org; Singh, Jasvinder <jasvinder.singh@intel.com>; Yigit,
> Ferruh <ferruh.yigit@intel.com>; WeiJie Zhuang <zhuangwj@gmail.com>
> Subject: [PATCH v3 1/3] port: add kni interface support
>
> 1. add KNI port type to the packet framework
> 2. add KNI support to the IP Pipeline sample Application
> 3. some bug fix
>
> Signed-off-by: WeiJie Zhuang <zhuangwj@gmail.com>
> ---
> v2:
> * Fix check patch error.
> v3:
> * Fix code review comments.
> ---
> doc/api/doxy-api-index.md | 1 +
> examples/ip_pipeline/Makefile | 2 +-
> examples/ip_pipeline/app.h | 181 +++++++++++-
> examples/ip_pipeline/config/kni.cfg | 67 +++++
> examples/ip_pipeline/config_check.c | 26 +-
> examples/ip_pipeline/config_parse.c | 166 ++++++++++-
> examples/ip_pipeline/init.c | 132 ++++++++-
> examples/ip_pipeline/pipeline/pipeline_common_fe.c | 29 ++
> examples/ip_pipeline/pipeline/pipeline_master_be.c | 6 +
> examples/ip_pipeline/pipeline_be.h | 27 ++
> lib/librte_port/Makefile | 7 +
> lib/librte_port/rte_port_kni.c | 325 +++++++++++++++++++++
> lib/librte_port/rte_port_kni.h | 82 ++++++
> lib/librte_port/rte_port_version.map | 8 +
> 14 files changed, 1047 insertions(+), 12 deletions(-)
> create mode 100644 examples/ip_pipeline/config/kni.cfg
> create mode 100644 lib/librte_port/rte_port_kni.c
> create mode 100644 lib/librte_port/rte_port_kni.h
>
> diff --git a/doc/api/doxy-api-index.md b/doc/api/doxy-api-index.md
> index f626386..5e7f024 100644
> --- a/doc/api/doxy-api-index.md
> +++ b/doc/api/doxy-api-index.md
> @@ -118,6 +118,7 @@ There are many libraries, so their headers may be
> grouped by topics:
> [frag] (@ref rte_port_frag.h),
> [reass] (@ref rte_port_ras.h),
> [sched] (@ref rte_port_sched.h),
> + [kni] (@ref rte_port_kni.h),
> [src/sink] (@ref rte_port_source_sink.h)
> * [table] (@ref rte_table.h):
> [lpm IPv4] (@ref rte_table_lpm.h),
> diff --git a/examples/ip_pipeline/Makefile b/examples/ip_pipeline/Makefile
> index 5827117..6dc3f52 100644
> --- a/examples/ip_pipeline/Makefile
> +++ b/examples/ip_pipeline/Makefile
> @@ -1,6 +1,6 @@
> # BSD LICENSE
> #
> -# Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
> +# Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
> # All rights reserved.
> #
> # Redistribution and use in source and binary forms, with or without
> diff --git a/examples/ip_pipeline/app.h b/examples/ip_pipeline/app.h
> index 7611341..abbd6d4 100644
> --- a/examples/ip_pipeline/app.h
> +++ b/examples/ip_pipeline/app.h
> @@ -44,6 +44,9 @@
> #include <cmdline_parse.h>
>
> #include <rte_ethdev.h>
> +#ifdef RTE_LIBRTE_KNI
> +#include <rte_kni.h>
> +#endif
>
> #include "cpu_core_map.h"
> #include "pipeline.h"
> @@ -132,6 +135,20 @@ struct app_pktq_swq_params {
> uint32_t mempool_indirect_id;
> };
>
> +struct app_pktq_kni_params {
> + char *name;
> + uint32_t parsed;
> +
> + uint32_t socket_id;
> + uint32_t core_id;
> + uint32_t hyper_th_id;
> + uint32_t force_bind;
> +
> + uint32_t mempool_id; /* Position in the app->mempool_params */
> + uint32_t burst_read;
> + uint32_t burst_write;
> +};
> +
> #ifndef APP_FILE_NAME_SIZE
> #define APP_FILE_NAME_SIZE 256
> #endif
> @@ -185,6 +202,7 @@ enum app_pktq_in_type {
> APP_PKTQ_IN_HWQ,
> APP_PKTQ_IN_SWQ,
> APP_PKTQ_IN_TM,
> + APP_PKTQ_IN_KNI,
> APP_PKTQ_IN_SOURCE,
> };
>
> @@ -197,6 +215,7 @@ enum app_pktq_out_type {
> APP_PKTQ_OUT_HWQ,
> APP_PKTQ_OUT_SWQ,
> APP_PKTQ_OUT_TM,
> + APP_PKTQ_OUT_KNI,
> APP_PKTQ_OUT_SINK,
> };
>
> @@ -420,6 +439,8 @@ struct app_eal_params {
>
> #define APP_MAX_PKTQ_TM APP_MAX_LINKS
>
> +#define APP_MAX_PKTQ_KNI APP_MAX_LINKS
> +
> #ifndef APP_MAX_PKTQ_SOURCE
> #define APP_MAX_PKTQ_SOURCE 64
> #endif
> @@ -471,6 +492,7 @@ struct app_params {
> struct app_pktq_hwq_out_params
> hwq_out_params[APP_MAX_HWQ_OUT];
> struct app_pktq_swq_params swq_params[APP_MAX_PKTQ_SWQ];
> struct app_pktq_tm_params tm_params[APP_MAX_PKTQ_TM];
> + struct app_pktq_kni_params kni_params[APP_MAX_PKTQ_KNI];
> struct app_pktq_source_params
> source_params[APP_MAX_PKTQ_SOURCE];
> struct app_pktq_sink_params sink_params[APP_MAX_PKTQ_SINK];
> struct app_msgq_params msgq_params[APP_MAX_MSGQ];
> @@ -482,6 +504,7 @@ struct app_params {
> uint32_t n_pktq_hwq_out;
> uint32_t n_pktq_swq;
> uint32_t n_pktq_tm;
> + uint32_t n_pktq_kni;
> uint32_t n_pktq_source;
> uint32_t n_pktq_sink;
> uint32_t n_msgq;
> @@ -495,6 +518,9 @@ struct app_params {
> struct app_link_data link_data[APP_MAX_LINKS];
> struct rte_ring *swq[APP_MAX_PKTQ_SWQ];
> struct rte_sched_port *tm[APP_MAX_PKTQ_TM];
> +#ifdef RTE_LIBRTE_KNI
> + struct rte_kni *kni[APP_MAX_PKTQ_KNI];
> +#endif /* RTE_LIBRTE_KNI */
> struct rte_ring *msgq[APP_MAX_MSGQ];
> struct pipeline_type pipeline_type[APP_MAX_PIPELINE_TYPES];
> struct app_pipeline_data pipeline_data[APP_MAX_PIPELINES];
> @@ -667,11 +693,11 @@ app_swq_get_reader(struct app_params *app,
> struct app_pktq_swq_params *swq,
> uint32_t *pktq_in_id)
> {
> - struct app_pipeline_params *reader;
> + struct app_pipeline_params *reader = NULL;
> uint32_t pos = swq - app->swq_params;
> uint32_t n_pipelines = RTE_MIN(app->n_pipelines,
> RTE_DIM(app->pipeline_params));
> - uint32_t n_readers = 0, id, i;
> + uint32_t n_readers = 0, id = 0, i;
>
> for (i = 0; i < n_pipelines; i++) {
> struct app_pipeline_params *p = &app->pipeline_params[i];
> @@ -727,11 +753,11 @@ app_tm_get_reader(struct app_params *app,
> struct app_pktq_tm_params *tm,
> uint32_t *pktq_in_id)
> {
> - struct app_pipeline_params *reader;
> + struct app_pipeline_params *reader = NULL;
> uint32_t pos = tm - app->tm_params;
> uint32_t n_pipelines = RTE_MIN(app->n_pipelines,
> RTE_DIM(app->pipeline_params));
> - uint32_t n_readers = 0, id, i;
> + uint32_t n_readers = 0, id = 0, i;
>
> for (i = 0; i < n_pipelines; i++) {
> struct app_pipeline_params *p = &app->pipeline_params[i];
> @@ -758,6 +784,66 @@ app_tm_get_reader(struct app_params *app,
> }
>
> static inline uint32_t
> +app_kni_get_readers(struct app_params *app, struct
> app_pktq_kni_params *kni)
> +{
> + uint32_t pos = kni - app->kni_params;
> + uint32_t n_pipelines = RTE_MIN(app->n_pipelines,
> + RTE_DIM(app->pipeline_params));
> + uint32_t n_readers = 0, i;
> +
> + for (i = 0; i < n_pipelines; i++) {
> + struct app_pipeline_params *p = &app->pipeline_params[i];
> + uint32_t n_pktq_in = RTE_MIN(p->n_pktq_in, RTE_DIM(p-
> >pktq_in));
> + uint32_t j;
> +
> + for (j = 0; j < n_pktq_in; j++) {
> + struct app_pktq_in_params *pktq = &p->pktq_in[j];
> +
> + if ((pktq->type == APP_PKTQ_IN_KNI) &&
> + (pktq->id == pos))
> + n_readers++;
> + }
> + }
> +
> + return n_readers;
> +}
> +
> +static inline struct app_pipeline_params *
> +app_kni_get_reader(struct app_params *app,
> + struct app_pktq_kni_params *kni,
> + uint32_t *pktq_in_id)
> +{
> + struct app_pipeline_params *reader = NULL;
> + uint32_t pos = kni - app->kni_params;
> + uint32_t n_pipelines = RTE_MIN(app->n_pipelines,
> + RTE_DIM(app->pipeline_params));
> + uint32_t n_readers = 0, id = 0, i;
> +
> + for (i = 0; i < n_pipelines; i++) {
> + struct app_pipeline_params *p = &app->pipeline_params[i];
> + uint32_t n_pktq_in = RTE_MIN(p->n_pktq_in, RTE_DIM(p-
> >pktq_in));
> + uint32_t j;
> +
> + for (j = 0; j < n_pktq_in; j++) {
> + struct app_pktq_in_params *pktq = &p->pktq_in[j];
> +
> + if ((pktq->type == APP_PKTQ_IN_KNI) &&
> + (pktq->id == pos)) {
> + n_readers++;
> + reader = p;
> + id = j;
> + }
> + }
> + }
> +
> + if (n_readers != 1)
> + return NULL;
> +
> + *pktq_in_id = id;
> + return reader;
> +}
> +
> +static inline uint32_t
> app_source_get_readers(struct app_params *app,
> struct app_pktq_source_params *source)
> {
> @@ -861,11 +947,11 @@ app_swq_get_writer(struct app_params *app,
> struct app_pktq_swq_params *swq,
> uint32_t *pktq_out_id)
> {
> - struct app_pipeline_params *writer;
> + struct app_pipeline_params *writer = NULL;
> uint32_t pos = swq - app->swq_params;
> uint32_t n_pipelines = RTE_MIN(app->n_pipelines,
> RTE_DIM(app->pipeline_params));
> - uint32_t n_writers = 0, id, i;
> + uint32_t n_writers = 0, id = 0, i;
>
> for (i = 0; i < n_pipelines; i++) {
> struct app_pipeline_params *p = &app->pipeline_params[i];
> @@ -923,11 +1009,11 @@ app_tm_get_writer(struct app_params *app,
> struct app_pktq_tm_params *tm,
> uint32_t *pktq_out_id)
> {
> - struct app_pipeline_params *writer;
> + struct app_pipeline_params *writer = NULL;
> uint32_t pos = tm - app->tm_params;
> uint32_t n_pipelines = RTE_MIN(app->n_pipelines,
> RTE_DIM(app->pipeline_params));
> - uint32_t n_writers = 0, id, i;
> + uint32_t n_writers = 0, id = 0, i;
>
> for (i = 0; i < n_pipelines; i++) {
> struct app_pipeline_params *p = &app->pipeline_params[i];
> @@ -939,10 +1025,73 @@ app_tm_get_writer(struct app_params *app,
> struct app_pktq_out_params *pktq = &p-
> >pktq_out[j];
>
> if ((pktq->type == APP_PKTQ_OUT_TM) &&
> + (pktq->id == pos)) {
> + n_writers++;
> + writer = p;
> + id = j;
> + }
> + }
> + }
> +
> + if (n_writers != 1)
> + return NULL;
> +
> + *pktq_out_id = id;
> + return writer;
> +}
> +
> +static inline uint32_t
> +app_kni_get_writers(struct app_params *app, struct
> app_pktq_kni_params *kni)
> +{
> + uint32_t pos = kni - app->kni_params;
> + uint32_t n_pipelines = RTE_MIN(app->n_pipelines,
> + RTE_DIM(app->pipeline_params));
> + uint32_t n_writers = 0, i;
> +
> + for (i = 0; i < n_pipelines; i++) {
> + struct app_pipeline_params *p = &app->pipeline_params[i];
> + uint32_t n_pktq_out = RTE_MIN(p->n_pktq_out,
> + RTE_DIM(p->pktq_out));
> + uint32_t j;
> +
> + for (j = 0; j < n_pktq_out; j++) {
> + struct app_pktq_out_params *pktq = &p-
> >pktq_out[j];
> +
> + if ((pktq->type == APP_PKTQ_OUT_KNI) &&
> (pktq->id == pos))
> n_writers++;
> + }
> + }
> +
> + return n_writers;
> +}
> +
> +static inline struct app_pipeline_params *
> +app_kni_get_writer(struct app_params *app,
> + struct app_pktq_kni_params *kni,
> + uint32_t *pktq_out_id)
> +{
> + struct app_pipeline_params *writer = NULL;
> + uint32_t pos = kni - app->kni_params;
> + uint32_t n_pipelines = RTE_MIN(app->n_pipelines,
> + RTE_DIM(app->pipeline_params));
> + uint32_t n_writers = 0, id = 0, i;
> +
> + for (i = 0; i < n_pipelines; i++) {
> + struct app_pipeline_params *p = &app->pipeline_params[i];
> + uint32_t n_pktq_out = RTE_MIN(p->n_pktq_out,
> + RTE_DIM(p->pktq_out));
> + uint32_t j;
> +
> + for (j = 0; j < n_pktq_out; j++) {
> + struct app_pktq_out_params *pktq = &p-
> >pktq_out[j];
> +
> + if ((pktq->type == APP_PKTQ_OUT_KNI) &&
> + (pktq->id == pos)) {
> + n_writers++;
> writer = p;
> id = j;
> + }
> }
> }
>
> @@ -1051,6 +1200,22 @@ app_get_link_for_tm(struct app_params *app,
> struct app_pktq_tm_params *p_tm)
> return &app->link_params[link_param_idx];
> }
>
> +static inline struct app_link_params *
> +app_get_link_for_kni(struct app_params *app, struct
> app_pktq_kni_params *p_kni)
> +{
> + char link_name[APP_PARAM_NAME_SIZE];
> + uint32_t link_id;
> + ssize_t link_param_idx;
> +
> + sscanf(p_kni->name, "KNI%" PRIu32, &link_id);
> + sprintf(link_name, "LINK%" PRIu32, link_id);
> + link_param_idx = APP_PARAM_FIND(app->link_params, link_name);
> + APP_CHECK((link_param_idx >= 0),
> + "Cannot find %s for %s", link_name, p_kni->name);
> +
> + return &app->link_params[link_param_idx];
> +}
> +
> void app_pipeline_params_get(struct app_params *app,
> struct app_pipeline_params *p_in,
> struct pipeline_params *p_out);
> diff --git a/examples/ip_pipeline/config/kni.cfg
> b/examples/ip_pipeline/config/kni.cfg
> new file mode 100644
> index 0000000..c339080
> --- /dev/null
> +++ b/examples/ip_pipeline/config/kni.cfg
> @@ -0,0 +1,67 @@
> +; BSD LICENSE
> +;
> +; Copyright(c) 2016 Intel Corporation.
> +; All rights reserved.
> +;
> +; Redistribution and use in source and binary forms, with or without
> +; modification, are permitted provided that the following conditions
> +; are met:
> +;
> +; * Redistributions of source code must retain the above copyright
> +; notice, this list of conditions and the following disclaimer.
> +; * Redistributions in binary form must reproduce the above copyright
> +; notice, this list of conditions and the following disclaimer in
> +; the documentation and/or other materials provided with the
> +; distribution.
> +; * Neither the name of Intel Corporation nor the names of its
> +; contributors may be used to endorse or promote products derived
> +; from this software without specific prior written permission.
> +;
> +; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
> CONTRIBUTORS
> +; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT
> NOT
> +; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
> FITNESS FOR
> +; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
> COPYRIGHT
> +; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
> INCIDENTAL,
> +; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
> NOT
> +; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
> OF USE,
> +; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
> AND ON ANY
> +; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
> +; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF
> THE USE
> +; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH
> DAMAGE.
> +
> +; ______________ ______________________
> +; | | KNI0 | |
> +; RXQ0.0 --->| |------->|--+ |
> +; | | KNI1 | | br0 |
> +; TXQ0.0 <---| |<-------|<-+ |
> +; | Pass-through | | Linux Kernel |
> +; | (P1) | | Network Stack |
> +; | | KNI0 | |
> +; RXQ0.0 --->| |------->|--+ |
> +; | | KNI1 | | br0 |
> +; TXQ0.0 <---| |<-------|<-+ |
> +; |______________| |______________________|
> +;
There are some errors in the diagram I previously sent you, here is the corrected version (which matches the PIPELINE1 configuration):
; ______________ ______________________
; | | KNI0 | |
; RXQ0.0 --->| |------->|--+ |
; | | KNI1 | | br0 |
; TXQ1.0 <---| |<-------|<-+ |
; | Pass-through | | Linux Kernel |
; | (P1) | | Network Stack |
; | | KNI1 | |
; RXQ1.0 --->| |------->|--+ |
; | | KNI0 | | br0 |
; TXQ0.0 <---| |<-------|<-+ |
; |______________| |______________________|
> +; Insert Linux kernel KNI module:
> +; [Linux]$ insmod rte_kni.ko
> +;
> +; Configure Linux kernel bridge between KNI0 and KNI1 interfaces:
> +; [Linux]$ ifconfig KNI0 up
> +; [Linux]$ ifconfig KNI1 up
> +; [Linux]$ brctl addbr "br0"
> +; [Linux]$ brctl addif br0 KNI0
> +; [Linux]$ brctl addif br0 KNI1
> +; [Linux]$ ifconfig br0 up
> +
> +[EAL]
> +log_level = 0
> +
> +[PIPELINE0]
> +type = MASTER
> +core = 0
> +
> +[PIPELINE1]
> +type = PASS-THROUGH
> +core = 1
> +pktq_in = RXQ0.0 KNI1 RXQ1.0 KNI0
> +pktq_out = KNI0 TXQ1.0 KNI1 TXQ0.0
> diff --git a/examples/ip_pipeline/config_check.c
> b/examples/ip_pipeline/config_check.c
> index 18f57be..7e90342 100644
> --- a/examples/ip_pipeline/config_check.c
> +++ b/examples/ip_pipeline/config_check.c
> @@ -1,7 +1,7 @@
> /*-
> * BSD LICENSE
> *
> - * Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
> + * Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
> * All rights reserved.
> *
> * Redistribution and use in source and binary forms, with or without
> @@ -444,6 +444,29 @@ check_pipelines(struct app_params *app)
> }
> }
>
> +static void
> +check_knis(struct app_params *app) {
> + uint32_t i;
> +
> + for (i = 0; i < app->n_pktq_kni; i++) {
> + struct app_pktq_kni_params *p = &app->kni_params[i];
> + uint32_t n_readers = app_kni_get_readers(app, p);
> + uint32_t n_writers = app_kni_get_writers(app, p);
> +
> + APP_CHECK((n_readers != 0),
> + "%s has no reader\n", p->name);
> +
> + APP_CHECK((n_readers == 1),
> + "%s has more than one reader\n", p->name);
> +
> + APP_CHECK((n_writers != 0),
> + "%s has no writer\n", p->name);
> +
> + APP_CHECK((n_writers == 1),
> + "%s has more than one writer\n", p->name);
> + }
> +}
I suggest you place this function just after the check_tms() function, in order to keep consistency with our conventions.
> +
> int
> app_config_check(struct app_params *app)
> {
> @@ -453,6 +476,7 @@ app_config_check(struct app_params *app)
> check_txqs(app);
> check_swqs(app);
> check_tms(app);
> + check_knis(app);
> check_sources(app);
> check_sinks(app);
> check_msgqs(app);
> diff --git a/examples/ip_pipeline/config_parse.c
> b/examples/ip_pipeline/config_parse.c
> index 504018e..c55be31 100644
> --- a/examples/ip_pipeline/config_parse.c
> +++ b/examples/ip_pipeline/config_parse.c
> @@ -189,6 +189,18 @@ struct app_pktq_tm_params default_tm_params = {
> .burst_write = 32,
> };
>
> +struct app_pktq_kni_params default_kni_params = {
> + .parsed = 0,
> + .socket_id = 0,
> + .core_id = 0,
> + .hyper_th_id = 0,
> + .force_bind = 0,
> +
> + .mempool_id = 0,
> + .burst_read = 32,
> + .burst_write = 32,
> +};
> +
> struct app_pktq_source_params default_source_params = {
> .parsed = 0,
> .mempool_id = 0,
> @@ -300,6 +312,18 @@ app_print_usage(char *prgname)
> link_param_pos;
> \
> })
>
> +#define APP_PARAM_ADD_LINK_FOR_KNI(app, kni_name)
> \
> +({ \
> + char link_name[APP_PARAM_NAME_SIZE];
> \
> + ssize_t link_param_pos;
> \
> + uint32_t link_id; \
> + \
> + sscanf((kni_name), "KNI%" SCNu32, &link_id); \
> + sprintf(link_name, "LINK%" PRIu32, link_id); \
> + link_param_pos = APP_PARAM_ADD((app)->link_params,
> link_name); \
> + link_param_pos;
> \
> +})
> +
> #define PARSE_CHECK_DUPLICATE_SECTION(obj)
> \
> do { \
> APP_CHECK(((obj)->parsed == 0),
> \
> @@ -826,6 +850,10 @@ parse_pipeline_pktq_in(struct app_params *app,
> type = APP_PKTQ_IN_TM;
> id = APP_PARAM_ADD(app->tm_params, name);
> APP_PARAM_ADD_LINK_FOR_TM(app, name);
> + } else if (validate_name(name, "KNI", 1) == 0) {
> + type = APP_PKTQ_IN_KNI;
> + id = APP_PARAM_ADD(app->kni_params, name);
> + APP_PARAM_ADD_LINK_FOR_KNI(app, name);
> } else if (validate_name(name, "SOURCE", 1) == 0) {
> type = APP_PKTQ_IN_SOURCE;
> id = APP_PARAM_ADD(app->source_params, name);
> @@ -871,6 +899,10 @@ parse_pipeline_pktq_out(struct app_params *app,
> type = APP_PKTQ_OUT_TM;
> id = APP_PARAM_ADD(app->tm_params, name);
> APP_PARAM_ADD_LINK_FOR_TM(app, name);
> + } else if (validate_name(name, "KNI", 1) == 0) {
> + type = APP_PKTQ_OUT_KNI;
> + id = APP_PARAM_ADD(app->kni_params, name);
> + APP_PARAM_ADD_LINK_FOR_KNI(app, name);
> } else if (validate_name(name, "SINK", 1) == 0) {
> type = APP_PKTQ_OUT_SINK;
> id = APP_PARAM_ADD(app->sink_params, name);
> @@ -1816,7 +1848,7 @@ parse_tm(struct app_params *app,
> param = &app->tm_params[param_idx];
> PARSE_CHECK_DUPLICATE_SECTION(param);
>
> - APP_PARAM_ADD_LINK_FOR_TXQ(app, section_name);
> + APP_PARAM_ADD_LINK_FOR_TM(app, section_name);
>
> for (i = 0; i < n_entries; i++) {
> struct rte_cfgfile_entry *ent = &entries[i];
> @@ -1853,6 +1885,85 @@ parse_tm(struct app_params *app,
> }
>
> static void
> +parse_kni(struct app_params *app,
> + const char *section_name,
> + struct rte_cfgfile *cfg)
> +{
> + struct app_pktq_kni_params *param;
> + struct rte_cfgfile_entry *entries;
> + int n_entries, i;
> + ssize_t param_idx;
> +
> + n_entries = rte_cfgfile_section_num_entries(cfg, section_name);
> + PARSE_ERROR_SECTION_NO_ENTRIES((n_entries > 0),
> section_name);
> +
> + entries = malloc(n_entries * sizeof(struct rte_cfgfile_entry));
> + PARSE_ERROR_MALLOC(entries != NULL);
> +
> + rte_cfgfile_section_entries(cfg, section_name, entries, n_entries);
> +
> + param_idx = APP_PARAM_ADD(app->kni_params, section_name);
> + param = &app->kni_params[param_idx];
> + PARSE_CHECK_DUPLICATE_SECTION(param);
> +
> + APP_PARAM_ADD_LINK_FOR_KNI(app, section_name);
> +
> +
> + for (i = 0; i < n_entries; i++) {
> + struct rte_cfgfile_entry *ent = &entries[i];
> +
> + if (strcmp(ent->name, "core") == 0) {
> + int status = parse_pipeline_core(
> + ¶m->socket_id,
> + ¶m->core_id,
> + ¶m->hyper_th_id,
> + ent->value);
> +
> + PARSE_ERROR((status == 0), section_name,
> + ent->name);
> + param->force_bind = 1;
> + continue;
> + }
> +
> + if (strcmp(ent->name, "mempool") == 0) {
> + int status = validate_name(ent->value,
> +
> "MEMPOOL", 1);
> + ssize_t idx;
> +
> + PARSE_ERROR((status == 0), section_name,
> + ent->name);
> +
> + idx = APP_PARAM_ADD(app->mempool_params,
> ent->value);
> + param->mempool_id = idx;
> + continue;
> + }
> +
> + if (strcmp(ent->name, "burst_read") == 0) {
> + int status = parser_read_uint32(¶m-
> >burst_read,
> +
> ent->value);
> +
> + PARSE_ERROR((status == 0), section_name,
> + ent->name);
> + continue;
> + }
> +
> + if (strcmp(ent->name, "burst_write") == 0) {
> + int status = parser_read_uint32(¶m-
> >burst_write,
> +
> ent->value);
> +
> + PARSE_ERROR((status == 0), section_name,
> + ent->name);
> + continue;
> + }
> +
> + /* unrecognized */
> + PARSE_ERROR_INVALID(0, section_name, ent->name);
> + }
> +
Here is one bug for you, you need to make sure you add the following line here:
param->parsed = 1;
> + free(entries);
> +}
> +
> +static void
> parse_source(struct app_params *app,
> const char *section_name,
> struct rte_cfgfile *cfg)
> @@ -2147,6 +2258,7 @@ static const struct config_section cfg_file_scheme[]
> = {
> {"TXQ", 2, parse_txq},
> {"SWQ", 1, parse_swq},
> {"TM", 1, parse_tm},
> + {"KNI", 1, parse_kni},
> {"SOURCE", 1, parse_source},
> {"SINK", 1, parse_sink},
> {"MSGQ-REQ-PIPELINE", 1, parse_msgq_req_pipeline},
> @@ -2285,6 +2397,7 @@ app_config_parse(struct app_params *app, const
> char *file_name)
> APP_PARAM_COUNT(app->hwq_out_params, app-
> >n_pktq_hwq_out);
> APP_PARAM_COUNT(app->swq_params, app->n_pktq_swq);
> APP_PARAM_COUNT(app->tm_params, app->n_pktq_tm);
> + APP_PARAM_COUNT(app->kni_params, app->n_pktq_kni);
> APP_PARAM_COUNT(app->source_params, app->n_pktq_source);
> APP_PARAM_COUNT(app->sink_params, app->n_pktq_sink);
> APP_PARAM_COUNT(app->msgq_params, app->n_msgq);
> @@ -2647,6 +2760,45 @@ save_tm_params(struct app_params *app, FILE
> *f)
> }
>
> static void
> +save_kni_params(struct app_params *app, FILE *f)
> +{
> + struct app_pktq_kni_params *p;
> + size_t i, count;
> +
> + count = RTE_DIM(app->kni_params);
> + for (i = 0; i < count; i++) {
> + p = &app->kni_params[i];
> + if (!APP_PARAM_VALID(p))
> + continue;
> +
> + /* section name */
> + fprintf(f, "[%s]\n", p->name);
> +
> + /* core */
> + if (p->force_bind) {
> + fprintf(f, "; force_bind = 1\n");
> + fprintf(f, "core = s%" PRIu32 "c%" PRIu32 "%s\n",
> + p->socket_id,
> + p->core_id,
> + (p->hyper_th_id) ? "h" : "");
> + } else
> + fprintf(f, "; force_bind = 0\n");
> +
> + /* mempool */
> + fprintf(f, "%s = %s\n", "mempool",
> + app->mempool_params[p-
> >mempool_id].name);
> +
> + /* burst_read */
> + fprintf(f, "%s = %" PRIu32 "\n", "burst_read", p-
> >burst_read);
> +
> + /* burst_write */
> + fprintf(f, "%s = %" PRIu32 "\n", "burst_write", p-
> >burst_write);
> +
> + fputc('\n', f);
> + }
> +}
> +
> +static void
> save_source_params(struct app_params *app, FILE *f)
> {
> struct app_pktq_source_params *p;
> @@ -2753,6 +2905,9 @@ save_pipeline_params(struct app_params *app,
> FILE *f)
> case APP_PKTQ_IN_TM:
> name = app->tm_params[pp-
> >id].name;
> break;
> + case APP_PKTQ_IN_KNI:
> + name = app->kni_params[pp-
> >id].name;
> + break;
> case APP_PKTQ_IN_SOURCE:
> name = app->source_params[pp-
> >id].name;
> break;
> @@ -2787,6 +2942,9 @@ save_pipeline_params(struct app_params *app,
> FILE *f)
> case APP_PKTQ_OUT_TM:
> name = app->tm_params[pp-
> >id].name;
> break;
> + case APP_PKTQ_OUT_KNI:
> + name = app->kni_params[pp-
> >id].name;
> + break;
> case APP_PKTQ_OUT_SINK:
> name = app->sink_params[pp-
> >id].name;
> break;
> @@ -2872,6 +3030,7 @@ app_config_save(struct app_params *app, const
> char *file_name)
> save_txq_params(app, file);
> save_swq_params(app, file);
> save_tm_params(app, file);
> + save_kni_params(app, file);
> save_source_params(app, file);
> save_sink_params(app, file);
> save_msgq_params(app, file);
> @@ -2921,6 +3080,11 @@ app_config_init(struct app_params *app)
> &default_tm_params,
> sizeof(default_tm_params));
>
> + for (i = 0; i < RTE_DIM(app->kni_params); i++)
> + memcpy(&app->kni_params[i],
> + &default_kni_params,
> + sizeof(default_kni_params));
> +
> for (i = 0; i < RTE_DIM(app->source_params); i++)
> memcpy(&app->source_params[i],
> &default_source_params,
> diff --git a/examples/ip_pipeline/init.c b/examples/ip_pipeline/init.c
In order to keep up with our existing code conventions, please move the KNI code (functions kni_change_mtu(), kni_config_network_interface(), app_init_kni()) just after function app_init_tm().
> index 7120bab..d522de4 100644
> --- a/examples/ip_pipeline/init.c
> +++ b/examples/ip_pipeline/init.c
> @@ -1281,10 +1281,21 @@ void app_pipeline_params_get(struct
> app_params *app,
> break;
> }
> case APP_PKTQ_IN_TM:
> + {
> out->type = PIPELINE_PORT_IN_SCHED_READER;
> out->params.sched.sched = app->tm[in->id];
> out->burst_size = app->tm_params[in-
> >id].burst_read;
> break;
> + }
> +#ifdef RTE_LIBRTE_KNI
> + case APP_PKTQ_IN_KNI:
> + {
> + out->type = PIPELINE_PORT_IN_KNI_READER;
> + out->params.kni.kni = app->kni[in->id];
> + out->burst_size = app->kni_params[in-
> >id].burst_read;
> + break;
> + }
> +#endif /* RTE_LIBRTE_KNI */
> case APP_PKTQ_IN_SOURCE:
> {
> uint32_t mempool_id =
> @@ -1409,7 +1420,8 @@ void app_pipeline_params_get(struct app_params
> *app,
> }
> break;
> }
> - case APP_PKTQ_OUT_TM: {
> + case APP_PKTQ_OUT_TM:
> + {
> struct rte_port_sched_writer_params *params =
> &out->params.sched;
>
> @@ -1419,6 +1431,16 @@ void app_pipeline_params_get(struct
> app_params *app,
> app->tm_params[in->id].burst_write;
> break;
> }
> +#ifdef RTE_LIBRTE_KNI
> + case APP_PKTQ_OUT_KNI:
> + {
> + out->type = PIPELINE_PORT_OUT_KNI_WRITER;
> + out->params.kni.kni = app->kni[in->id];
> + out->params.kni.tx_burst_sz =
> + app->kni_params[in->id].burst_write;
> + break;
> + }
> +#endif /* RTE_LIBRTE_KNI */
> case APP_PKTQ_OUT_SINK:
> {
> out->type = PIPELINE_PORT_OUT_SINK;
> @@ -1452,6 +1474,113 @@ void app_pipeline_params_get(struct
> app_params *app,
> }
> }
>
> +#ifdef RTE_LIBRTE_KNI
> +static int
> +kni_config_network_interface(uint8_t port_id, uint8_t if_up) {
> + int ret = 0;
> +
> + if (port_id >= rte_eth_dev_count())
> + return -EINVAL;
> +
> + if (if_up) {
> + rte_eth_dev_stop(port_id);
> + ret = rte_eth_dev_start(port_id);
> + } else
> + rte_eth_dev_stop(port_id);
> +
> + return ret;
> +}
> +
Here is a critical issue that we found out recently during testing the KNI code: the correct PMD API functions to use here are rte_eth_dev_set_link_up/down(); stopping and restarting the device while the device is running can lead to segfault and nondeterministic behaviour.
Here is my proposal:
static int
kni_config_network_interface(uint8_t port_id, uint8_t if_up) {
int ret = 0;
if (port_id >= rte_eth_dev_count())
return -EINVAL;
ret = (if_up)?
rte_eth_dev_set_link_up(port_id) :
rte_eth_dev_set_link_down(port_id);
return ret;
}
I tested this function successfully.
> +static int
> +kni_change_mtu(uint8_t port_id, unsigned new_mtu) {
> + int ret;
> +
> + if (port_id >= rte_eth_dev_count())
> + return -EINVAL;
> +
> + if (new_mtu > ETHER_MAX_LEN)
> + return -EINVAL;
> +
> + /* Stop specific port */
> + rte_eth_dev_stop(port_id);
> +
> + /* Set new MTU */
> + ret = rte_eth_dev_set_mtu(port_id, new_mtu);
> + if (ret < 0)
> + return ret;
> +
> + /* Restart specific port */
> + ret = rte_eth_dev_start(port_id);
> + if (ret < 0)
> + return ret;
> +
> + return 0;
> +}
Same issue about using the rte_eth_dev_start/stop() functions. Based on function port_mtu_set() of the app/test-pmd application, it looks like rte_eth_dev_set_mtu() function could be called directly with no start/stop or up/down preparations.
Here is my version, although I did not fully test this one:
static int
kni_change_mtu(uint8_t port_id, unsigned new_mtu) {
int ret;
if (port_id >= rte_eth_dev_count())
return -EINVAL;
if (new_mtu > ETHER_MAX_LEN)
return -EINVAL;
/* Set new MTU */
ret = rte_eth_dev_set_mtu(port_id, new_mtu);
if (ret < 0)
return ret;
return 0;
}
> +#endif /* RTE_LIBRTE_KNI */
> +
> +static void
> +app_init_kni(struct app_params *app) {
> + if (app->n_pktq_kni == 0)
> + return;
> +#ifndef RTE_LIBRTE_KNI
> + else
> + rte_panic("Can not init KNI without librte_kni support.\n");
> +#else
I would like to avoid #ifdefs in the middle of the function code as much as possible. I suggest we do the following:
#ifndef RTE_LIBRTE_KNI
static void
app_init_kni(struct app_params *app)
{
if (app->n_pktq_kni == 0)
return;
rte_panic("Cannot init KNI without librte_kni support.\n");
}
#else
static void
app_init_kni(struct app_params *app)
{
//the real function code
}
#endif
> + rte_kni_init(app->n_pktq_kni);
> +
> + for (uint32_t i = 0; i < app->n_pktq_kni; i++) {
Please declare the "i" variable at the start of the function, as usual. Although I personally like this idea, this pattern is not used anywhere in DPDK and it might trigger warnings/errors from some C compilers. Btw, any idea whether this is mentioned in any way by the ANSI C/C99 standards?
> + struct app_pktq_kni_params *p_kni = &app->kni_params[i];
> + struct app_link_params *p_link;
> + struct rte_eth_dev_info dev_info;
> + struct app_mempool_params *mempool_params;
> + struct rte_mempool *mempool;
> + struct rte_kni_conf conf;
> + struct rte_kni_ops ops;
> +
> + /* LINK */
> + p_link = app_get_link_for_kni(app, p_kni);
> + memset(&dev_info, 0, sizeof(dev_info));
> + rte_eth_dev_info_get(p_link->pmd_id, &dev_info);
> +
> + /* MEMPOOL */
> + mempool_params = &app->mempool_params[p_kni-
> >mempool_id];
> + mempool = app->mempool[p_kni->mempool_id];
> +
> + /* KNI */
> + memset(&conf, 0, sizeof(conf));
> + snprintf(conf.name, RTE_KNI_NAMESIZE, "%s", p_kni-
> >name);
> + conf.force_bind = p_kni->force_bind;
> + if (conf.force_bind) {
> + int lcore_id;
> +
> + lcore_id = cpu_core_map_get_lcore_id(app-
> >core_map,
> + p_kni->socket_id,
> + p_kni->core_id,
> + p_kni->hyper_th_id);
> +
> + if (lcore_id < 0)
> + rte_panic("%s invalid CPU core\n", p_kni-
> >name);
> +
> + conf.core_id = (uint32_t) lcore_id;
> + }
> + conf.group_id = p_link->pmd_id;
> + conf.mbuf_size = mempool_params->buffer_size;
> + conf.addr = dev_info.pci_dev->addr;
> + conf.id = dev_info.pci_dev->id;
> +
> + memset(&ops, 0, sizeof(ops));
> + ops.port_id = (uint8_t) p_link->pmd_id;
> + ops.change_mtu = kni_change_mtu;
> + ops.config_network_if = kni_config_network_interface;
> +
> + APP_LOG(app, HIGH, "Initializing %s ...", p_kni->name);
> + app->kni[i] = rte_kni_alloc(mempool, &conf, &ops);
> + if (!app->kni[i])
> + rte_panic("%s init error\n", p_kni->name);
> + }
> +#endif /* RTE_LIBRTE_KNI */
> +}
> +
> static void
> app_init_pipelines(struct app_params *app)
> {
> @@ -1607,6 +1736,7 @@ int app_init(struct app_params *app)
> app_init_link(app);
> app_init_swq(app);
> app_init_tm(app);
> + app_init_kni(app);
> app_init_msgq(app);
>
> app_pipeline_common_cmd_push(app);
> diff --git a/examples/ip_pipeline/pipeline/pipeline_common_fe.c
> b/examples/ip_pipeline/pipeline/pipeline_common_fe.c
> index 70c57e4..63723cd 100644
> --- a/examples/ip_pipeline/pipeline/pipeline_common_fe.c
> +++ b/examples/ip_pipeline/pipeline/pipeline_common_fe.c
> @@ -130,6 +130,35 @@ app_pipeline_track_pktq_out_to_link(struct
> app_params *app,
> break;
> }
>
> +#ifdef RTE_LIBRTE_KNI
As in any other place that just deals with KNI configuration data rather than KNI init data (e.g. in app.h or config_parse.c), we do not need to #ifdef the code fragment below. Please remove this instance of #ifdef RTE_LIBRTE_KNI.
> + case APP_PKTQ_OUT_KNI:
> + {
> + struct pipeline_params pp;
> + struct pipeline_type *ptype;
> + struct app_pktq_kni_params *kni;
> + uint32_t pktq_in_id;
> + int status;
> +
> + kni = &app->kni_params[pktq_out->id];
> + p = app_kni_get_reader(app, kni, &pktq_in_id);
> + if (p == NULL)
> + return NULL;
> +
> + ptype = app_pipeline_type_find(app, p->type);
> + if ((ptype == NULL) || (ptype->fe_ops->f_track ==
> NULL))
> + return NULL;
> +
> + app_pipeline_params_get(app, p, &pp);
> + status = ptype->fe_ops->f_track(&pp,
> + pktq_in_id,
> + &pktq_out_id);
> + if (status)
> + return NULL;
> +
> + break;
> + }
> +#endif
> +
> case APP_PKTQ_OUT_SINK:
> default:
> return NULL;
> diff --git a/examples/ip_pipeline/pipeline/pipeline_master_be.c
> b/examples/ip_pipeline/pipeline/pipeline_master_be.c
> index 79869a4..32d4635 100644
> --- a/examples/ip_pipeline/pipeline/pipeline_master_be.c
> +++ b/examples/ip_pipeline/pipeline/pipeline_master_be.c
> @@ -144,6 +144,12 @@ pipeline_run(void *pipeline)
> rte_exit(0, "Bye!\n");
> }
>
> +#ifdef RTE_LIBRTE_KNI
> + /* Handle KNI requests from Linux kernel */
> + for (uint32_t i = 0; i < app->n_pktq_kni; i++)
Same comment as above on definition of the "i" variable.
> + rte_kni_handle_request(app->kni[i]);
> +#endif /* RTE_LIBRTE_KNI */
> +
> return 0;
> }
>
> diff --git a/examples/ip_pipeline/pipeline_be.h
> b/examples/ip_pipeline/pipeline_be.h
> index 5501ab7..0ee208c 100644
> --- a/examples/ip_pipeline/pipeline_be.h
> +++ b/examples/ip_pipeline/pipeline_be.h
> @@ -40,6 +40,9 @@
> #include <rte_port_ras.h>
> #include <rte_port_sched.h>
> #include <rte_port_source_sink.h>
> +#ifdef RTE_LIBRTE_KNI
> +#include <rte_port_kni.h>
> +#endif
> #include <rte_pipeline.h>
>
> enum pipeline_port_in_type {
> @@ -49,6 +52,7 @@ enum pipeline_port_in_type {
> PIPELINE_PORT_IN_RING_READER_IPV4_FRAG,
> PIPELINE_PORT_IN_RING_READER_IPV6_FRAG,
> PIPELINE_PORT_IN_SCHED_READER,
> + PIPELINE_PORT_IN_KNI_READER,
> PIPELINE_PORT_IN_SOURCE,
> };
>
> @@ -61,6 +65,9 @@ struct pipeline_port_in_params {
> struct rte_port_ring_reader_ipv4_frag_params
> ring_ipv4_frag;
> struct rte_port_ring_reader_ipv6_frag_params
> ring_ipv6_frag;
> struct rte_port_sched_reader_params sched;
> +#ifdef RTE_LIBRTE_KNI
> + struct rte_port_kni_reader_params kni;
> +#endif
> struct rte_port_source_params source;
> } params;
> uint32_t burst_size;
> @@ -82,6 +89,10 @@ pipeline_port_in_params_convert(struct
> pipeline_port_in_params *p)
> return (void *) &p->params.ring_ipv6_frag;
> case PIPELINE_PORT_IN_SCHED_READER:
> return (void *) &p->params.sched;
> +#ifdef RTE_LIBRTE_KNI
> + case PIPELINE_PORT_IN_KNI_READER:
> + return (void *) &p->params.kni;
> +#endif
> case PIPELINE_PORT_IN_SOURCE:
> return (void *) &p->params.source;
> default:
> @@ -105,6 +116,10 @@ pipeline_port_in_params_get_ops(struct
> pipeline_port_in_params *p)
> return &rte_port_ring_reader_ipv6_frag_ops;
> case PIPELINE_PORT_IN_SCHED_READER:
> return &rte_port_sched_reader_ops;
> +#ifdef RTE_LIBRTE_KNI
> + case PIPELINE_PORT_IN_KNI_READER:
> + return &rte_port_kni_reader_ops;
> +#endif
> case PIPELINE_PORT_IN_SOURCE:
> return &rte_port_source_ops;
> default:
> @@ -122,6 +137,7 @@ enum pipeline_port_out_type {
> PIPELINE_PORT_OUT_RING_WRITER_IPV4_RAS,
> PIPELINE_PORT_OUT_RING_WRITER_IPV6_RAS,
> PIPELINE_PORT_OUT_SCHED_WRITER,
> + PIPELINE_PORT_OUT_KNI_WRITER,
> PIPELINE_PORT_OUT_SINK,
> };
>
> @@ -137,6 +153,9 @@ struct pipeline_port_out_params {
> struct rte_port_ring_writer_ipv4_ras_params ring_ipv4_ras;
> struct rte_port_ring_writer_ipv6_ras_params ring_ipv6_ras;
> struct rte_port_sched_writer_params sched;
> +#ifdef RTE_LIBRTE_KNI
> + struct rte_port_kni_writer_params kni;
> +#endif
> struct rte_port_sink_params sink;
> } params;
> };
> @@ -163,6 +182,10 @@ pipeline_port_out_params_convert(struct
> pipeline_port_out_params *p)
> return (void *) &p->params.ring_ipv6_ras;
> case PIPELINE_PORT_OUT_SCHED_WRITER:
> return (void *) &p->params.sched;
> +#ifdef RTE_LIBRTE_KNI
> + case PIPELINE_PORT_OUT_KNI_WRITER:
> + return (void *) &p->params.kni;
> +#endif
> case PIPELINE_PORT_OUT_SINK:
> return (void *) &p->params.sink;
> default:
> @@ -192,6 +215,10 @@ pipeline_port_out_params_get_ops(struct
> pipeline_port_out_params *p)
> return &rte_port_ring_writer_ipv6_ras_ops;
> case PIPELINE_PORT_OUT_SCHED_WRITER:
> return &rte_port_sched_writer_ops;
> +#ifdef RTE_LIBRTE_KNI
> + case PIPELINE_PORT_OUT_KNI_WRITER:
> + return &rte_port_kni_writer_ops;
> +#endif
> case PIPELINE_PORT_OUT_SINK:
> return &rte_port_sink_ops;
> default:
> diff --git a/lib/librte_port/Makefile b/lib/librte_port/Makefile
> index dc6a601..0fc929b 100644
> --- a/lib/librte_port/Makefile
> +++ b/lib/librte_port/Makefile
> @@ -56,6 +56,9 @@ SRCS-$(CONFIG_RTE_LIBRTE_PORT) += rte_port_frag.c
> SRCS-$(CONFIG_RTE_LIBRTE_PORT) += rte_port_ras.c
> endif
> SRCS-$(CONFIG_RTE_LIBRTE_PORT) += rte_port_sched.c
> +ifeq ($(CONFIG_RTE_LIBRTE_KNI),y)
> +SRCS-$(CONFIG_RTE_LIBRTE_PORT) += rte_port_kni.c
> +endif
> SRCS-$(CONFIG_RTE_LIBRTE_PORT) += rte_port_source_sink.c
>
> # install includes
> @@ -67,6 +70,9 @@ SYMLINK-$(CONFIG_RTE_LIBRTE_PORT)-include +=
> rte_port_frag.h
> SYMLINK-$(CONFIG_RTE_LIBRTE_PORT)-include += rte_port_ras.h
> endif
> SYMLINK-$(CONFIG_RTE_LIBRTE_PORT)-include += rte_port_sched.h
> +ifeq ($(CONFIG_RTE_LIBRTE_KNI),y)
> +SYMLINK-$(CONFIG_RTE_LIBRTE_PORT)-include += rte_port_kni.h
> +endif
> SYMLINK-$(CONFIG_RTE_LIBRTE_PORT)-include += rte_port_source_sink.h
>
> # this lib depends upon:
> @@ -76,5 +82,6 @@ DEPDIRS-$(CONFIG_RTE_LIBRTE_PORT) +=
> lib/librte_mempool
> DEPDIRS-$(CONFIG_RTE_LIBRTE_PORT) += lib/librte_ether
> DEPDIRS-$(CONFIG_RTE_LIBRTE_PORT) += lib/librte_ip_frag
> DEPDIRS-$(CONFIG_RTE_LIBRTE_PORT) += lib/librte_sched
> +DEPDIRS-$(CONFIG_RTE_LIBRTE_PORT) += lib/librte_kni
>
> include $(RTE_SDK)/mk/rte.lib.mk
> diff --git a/lib/librte_port/rte_port_kni.c b/lib/librte_port/rte_port_kni.c
> new file mode 100644
> index 0000000..4cbc345
> --- /dev/null
> +++ b/lib/librte_port/rte_port_kni.c
> @@ -0,0 +1,325 @@
> +/*-
> + * BSD LICENSE
> + *
> + * Copyright(c) 2016 Ethan Zhuang <zhuangwj@gmail.com>.
> + * Copyright(c) 2016 Intel Corporation.
> + * All rights reserved.
> + *
> + * Redistribution and use in source and binary forms, with or without
> + * modification, are permitted provided that the following conditions
> + * are met:
> + *
> + * * Redistributions of source code must retain the above copyright
> + * notice, this list of conditions and the following disclaimer.
> + * * Redistributions in binary form must reproduce the above copyright
> + * notice, this list of conditions and the following disclaimer in
> + * the documentation and/or other materials provided with the
> + * distribution.
> + * * Neither the name of Intel Corporation nor the names of its
> + * contributors may be used to endorse or promote products derived
> + * from this software without specific prior written permission.
> + *
> + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
> CONTRIBUTORS
> + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT
> NOT
> + * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
> FITNESS FOR
> + * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
> COPYRIGHT
> + * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
> INCIDENTAL,
> + * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
> NOT
> + * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
> OF USE,
> + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
> AND ON ANY
> + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR
> TORT
> + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF
> THE USE
> + * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH
> DAMAGE.
> + */
> +#include <string.h>
> +
> +#include <rte_common.h>
> +#include <rte_malloc.h>
> +#include <rte_kni.h>
> +
> +#include "rte_port_kni.h"
> +
> +/*
> + * Port KNI Reader
> + */
> +#ifdef RTE_PORT_STATS_COLLECT
> +
> +#define RTE_PORT_KNI_READER_STATS_PKTS_IN_ADD(port, val) \
> + port->stats.n_pkts_in += val
> +#define RTE_PORT_KNI_READER_STATS_PKTS_DROP_ADD(port, val) \
> + port->stats.n_pkts_drop += val
> +
> +#else
> +
> +#define RTE_PORT_KNI_READER_STATS_PKTS_IN_ADD(port, val)
> +#define RTE_PORT_KNI_READER_STATS_PKTS_DROP_ADD(port, val)
> +
> +#endif
> +
> +struct rte_port_kni_reader {
> + struct rte_port_in_stats stats;
> +
> + struct rte_kni *kni;
> +};
> +
> +static void *
> +rte_port_kni_reader_create(void *params, int socket_id)
> +{
> + struct rte_port_kni_reader_params *conf =
> + (struct rte_port_kni_reader_params *) params;
> + struct rte_port_kni_reader *port;
> +
> + /* Check input parameters */
> + if (conf == NULL) {
> + RTE_LOG(ERR, PORT, "%s: params is NULL\n", __func__);
> + return NULL;
> + }
> +
> + /* Memory allocation */
> + port = rte_zmalloc_socket("PORT", sizeof(*port),
> + RTE_CACHE_LINE_SIZE, socket_id);
> + if (port == NULL) {
> + RTE_LOG(ERR, PORT, "%s: Failed to allocate port\n",
> __func__);
> + return NULL;
> + }
> +
> + /* Initialization */
> + port->kni = conf->kni;
> +
> + return port;
> +}
> +
> +static int
> +rte_port_kni_reader_rx(void *port, struct rte_mbuf **pkts, uint32_t
> n_pkts)
> +{
> + struct rte_port_kni_reader *p =
> + (struct rte_port_kni_reader *) port;
> + uint16_t rx_pkt_cnt;
> +
> + rx_pkt_cnt = rte_kni_rx_burst(p->kni, pkts, n_pkts);
> + RTE_PORT_KNI_READER_STATS_PKTS_IN_ADD(p, rx_pkt_cnt);
> + return rx_pkt_cnt;
> +}
> +
> +static int
> +rte_port_kni_reader_free(void *port)
> +{
> + if (port == NULL) {
> + RTE_LOG(ERR, PORT, "%s: port is NULL\n", __func__);
> + return -EINVAL;
> + }
> +
> + rte_free(port);
> +
> + return 0;
> +}
> +
> +static int rte_port_kni_reader_stats_read(void *port,
> + struct rte_port_in_stats *stats, int clear)
> +{
> + struct rte_port_kni_reader *p =
> + (struct rte_port_kni_reader *) port;
> +
> + if (stats != NULL)
> + memcpy(stats, &p->stats, sizeof(p->stats));
> +
> + if (clear)
> + memset(&p->stats, 0, sizeof(p->stats));
> +
> + return 0;
> +}
> +
> +/*
> + * Port KNI Writer
> + */
> +#ifdef RTE_PORT_STATS_COLLECT
> +
> +#define RTE_PORT_KNI_WRITER_STATS_PKTS_IN_ADD(port, val) \
> + port->stats.n_pkts_in += val
> +#define RTE_PORT_KNI_WRITER_STATS_PKTS_DROP_ADD(port, val) \
> + port->stats.n_pkts_drop += val
> +
> +#else
> +
> +#define RTE_PORT_KNI_WRITER_STATS_PKTS_IN_ADD(port, val)
> +#define RTE_PORT_KNI_WRITER_STATS_PKTS_DROP_ADD(port, val)
> +
> +#endif
> +
> +struct rte_port_kni_writer {
> + struct rte_port_out_stats stats;
> +
> + struct rte_mbuf *tx_buf[2 * RTE_PORT_IN_BURST_SIZE_MAX];
> + uint32_t tx_burst_sz;
> + uint32_t tx_buf_count;
> + uint64_t bsz_mask;
> + struct rte_kni *kni;
> +};
> +
> +static void *
> +rte_port_kni_writer_create(void *params, int socket_id)
> +{
> + struct rte_port_kni_writer_params *conf =
> + (struct rte_port_kni_writer_params *) params;
> + struct rte_port_kni_writer *port;
> +
> + /* Check input parameters */
> + if ((conf == NULL) ||
> + (conf->tx_burst_sz == 0) ||
> + (conf->tx_burst_sz > RTE_PORT_IN_BURST_SIZE_MAX) ||
> + (!rte_is_power_of_2(conf->tx_burst_sz))) {
> + RTE_LOG(ERR, PORT, "%s: Invalid input parameters\n",
> __func__);
> + return NULL;
> + }
> +
> + /* Memory allocation */
> + port = rte_zmalloc_socket("PORT", sizeof(*port),
> + RTE_CACHE_LINE_SIZE, socket_id);
> + if (port == NULL) {
> + RTE_LOG(ERR, PORT, "%s: Failed to allocate port\n",
> __func__);
> + return NULL;
> + }
> +
> + /* Initialization */
> + port->kni = conf->kni;
> + port->tx_burst_sz = conf->tx_burst_sz;
> + port->tx_buf_count = 0;
> + port->bsz_mask = 1LLU << (conf->tx_burst_sz - 1);
> +
> + return port;
> +}
> +
> +static inline void
> +send_burst(struct rte_port_kni_writer *p)
> +{
> + uint32_t nb_tx;
> +
> + nb_tx = rte_kni_tx_burst(p->kni, p->tx_buf, p->tx_buf_count);
> +
> + RTE_PORT_KNI_WRITER_STATS_PKTS_DROP_ADD(p, p-
> >tx_buf_count - nb_tx);
> + for (; nb_tx < p->tx_buf_count; nb_tx++)
> + rte_pktmbuf_free(p->tx_buf[nb_tx]);
> +
> + p->tx_buf_count = 0;
> +}
> +
> +static int
> +rte_port_kni_writer_tx(void *port, struct rte_mbuf *pkt)
> +{
> + struct rte_port_kni_writer *p =
> + (struct rte_port_kni_writer *) port;
> +
> + p->tx_buf[p->tx_buf_count++] = pkt;
> + RTE_PORT_KNI_WRITER_STATS_PKTS_IN_ADD(p, 1);
> + if (p->tx_buf_count >= p->tx_burst_sz)
> + send_burst(p);
> +
> + return 0;
> +}
> +
> +static int
> +rte_port_kni_writer_tx_bulk(void *port,
> + struct rte_mbuf **pkts,
> + uint64_t pkts_mask)
> +{
> + struct rte_port_kni_writer *p =
> + (struct rte_port_kni_writer *) port;
> + uint64_t bsz_mask = p->bsz_mask;
> + uint32_t tx_buf_count = p->tx_buf_count;
> + uint64_t expr = (pkts_mask & (pkts_mask + 1)) |
> + ((pkts_mask & bsz_mask) ^
> bsz_mask);
> +
> + if (expr == 0) {
> + uint64_t n_pkts = __builtin_popcountll(pkts_mask);
> + uint32_t n_pkts_ok;
> +
> + if (tx_buf_count)
> + send_burst(p);
> +
> + RTE_PORT_KNI_WRITER_STATS_PKTS_IN_ADD(p, n_pkts);
> + n_pkts_ok = rte_kni_tx_burst(p->kni, pkts, n_pkts);
> +
> + RTE_PORT_KNI_WRITER_STATS_PKTS_DROP_ADD(p, n_pkts
> - n_pkts_ok);
> + for (; n_pkts_ok < n_pkts; n_pkts_ok++) {
> + struct rte_mbuf *pkt = pkts[n_pkts_ok];
> +
> + rte_pktmbuf_free(pkt);
> + }
> + } else {
> + for (; pkts_mask;) {
> + uint32_t pkt_index = __builtin_ctzll(pkts_mask);
> + uint64_t pkt_mask = 1LLU << pkt_index;
> + struct rte_mbuf *pkt = pkts[pkt_index];
> +
> + p->tx_buf[tx_buf_count++] = pkt;
> + RTE_PORT_KNI_WRITER_STATS_PKTS_IN_ADD(p, 1);
> + pkts_mask &= ~pkt_mask;
> + }
> +
> + p->tx_buf_count = tx_buf_count;
> + if (tx_buf_count >= p->tx_burst_sz)
> + send_burst(p);
> + }
> +
> + return 0;
> +}
> +
> +static int
> +rte_port_kni_writer_flush(void *port)
> +{
> + struct rte_port_kni_writer *p =
> + (struct rte_port_kni_writer *) port;
> +
> + if (p->tx_buf_count > 0)
> + send_burst(p);
> +
> + return 0;
> +}
> +
> +static int
> +rte_port_kni_writer_free(void *port)
> +{
> + if (port == NULL) {
> + RTE_LOG(ERR, PORT, "%s: Port is NULL\n", __func__);
> + return -EINVAL;
> + }
> +
> + rte_port_kni_writer_flush(port);
> + rte_free(port);
> +
> + return 0;
> +}
> +
> +static int rte_port_kni_writer_stats_read(void *port,
> + struct rte_port_out_stats *stats, int clear)
> +{
> + struct rte_port_kni_writer *p =
> + (struct rte_port_kni_writer *) port;
> +
> + if (stats != NULL)
> + memcpy(stats, &p->stats, sizeof(p->stats));
> +
> + if (clear)
> + memset(&p->stats, 0, sizeof(p->stats));
> +
> + return 0;
> +}
> +
> +/*
> + * Summary of port operations
> + */
> +struct rte_port_in_ops rte_port_kni_reader_ops = {
> + .f_create = rte_port_kni_reader_create,
> + .f_free = rte_port_kni_reader_free,
> + .f_rx = rte_port_kni_reader_rx,
> + .f_stats = rte_port_kni_reader_stats_read,
> +};
> +
> +struct rte_port_out_ops rte_port_kni_writer_ops = {
> + .f_create = rte_port_kni_writer_create,
> + .f_free = rte_port_kni_writer_free,
> + .f_tx = rte_port_kni_writer_tx,
> + .f_tx_bulk = rte_port_kni_writer_tx_bulk,
> + .f_flush = rte_port_kni_writer_flush,
> + .f_stats = rte_port_kni_writer_stats_read,
> +};
> diff --git a/lib/librte_port/rte_port_kni.h b/lib/librte_port/rte_port_kni.h
> new file mode 100644
> index 0000000..d4de8c4
> --- /dev/null
> +++ b/lib/librte_port/rte_port_kni.h
> @@ -0,0 +1,82 @@
> +/*-
> + * BSD LICENSE
> + *
> + * Copyright(c) 2016 Ethan Zhuang <zhuangwj@gmail.com>.
> + * Copyright(c) 2016 Intel Corporation.
> + * All rights reserved.
> + *
> + * Redistribution and use in source and binary forms, with or without
> + * modification, are permitted provided that the following conditions
> + * are met:
> + *
> + * * Redistributions of source code must retain the above copyright
> + * notice, this list of conditions and the following disclaimer.
> + * * Redistributions in binary form must reproduce the above copyright
> + * notice, this list of conditions and the following disclaimer in
> + * the documentation and/or other materials provided with the
> + * distribution.
> + * * Neither the name of Intel Corporation nor the names of its
> + * contributors may be used to endorse or promote products derived
> + * from this software without specific prior written permission.
> + *
> + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
> CONTRIBUTORS
> + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT
> NOT
> + * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
> FITNESS FOR
> + * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
> COPYRIGHT
> + * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
> INCIDENTAL,
> + * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
> NOT
> + * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
> OF USE,
> + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
> AND ON ANY
> + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR
> TORT
> + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF
> THE USE
> + * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH
> DAMAGE.
> + */
> +
> +#ifndef __INCLUDE_RTE_PORT_KNI_H__
> +#define __INCLUDE_RTE_PORT_KNI_H__
> +
> +#ifdef __cplusplus
> +extern "C" {
> +#endif
> +
> +/**
> + * @file
> + * RTE Port KNI Interface
> + *
> + * kni_reader: input port built on top of pre-initialized KNI interface
> + * kni_writer: output port built on top of pre-initialized KNI interface
> + *
> + ***/
> +
> +#include <stdint.h>
> +
> +#include <rte_kni.h>
> +
> +#include "rte_port.h"
> +
> +/** kni_reader port parameters */
> +struct rte_port_kni_reader_params {
> + /** KNI interface reference */
> + struct rte_kni *kni;
> +};
> +
> +/** kni_reader port operations */
> +extern struct rte_port_in_ops rte_port_kni_reader_ops;
> +
> +
> +/** kni_writer port parameters */
> +struct rte_port_kni_writer_params {
> + /** KNI interface reference */
> + struct rte_kni *kni;
> + /** Burst size to KNI interface. */
> + uint32_t tx_burst_sz;
> +};
> +
> +/** kni_writer port operations */
> +extern struct rte_port_out_ops rte_port_kni_writer_ops;
> +
> +#ifdef __cplusplus
> +}
> +#endif
> +
> +#endif
> diff --git a/lib/librte_port/rte_port_version.map
> b/lib/librte_port/rte_port_version.map
> index 7a0b34d..e61b3fa 100644
> --- a/lib/librte_port/rte_port_version.map
> +++ b/lib/librte_port/rte_port_version.map
> @@ -35,3 +35,11 @@ DPDK_2.2 {
> rte_port_ring_multi_writer_nodrop_ops;
>
> } DPDK_2.1;
> +
> +DPDK_16.07 {
> + global:
> +
> + rte_port_kni_reader_ops;
> + rte_port_kni_writer_ops;
> +
> +} DPDK_2.2;
> --
> 2.7.4
Regards,
Cristian
next prev parent reply other threads:[~2016-06-18 16:45 UTC|newest]
Thread overview: 22+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-05-27 5:07 [dpdk-dev] [PATCH] port: add KNI interface support 1. add KNI port type to the packet framework 2. add KNI support to the IP Pipeline sample Application WeiJie Zhuang
2016-05-28 11:25 ` [dpdk-dev] [PATCH] port: add kni interface support WeiJie Zhuang
2016-05-30 14:40 ` Dumitrescu, Cristian
2016-06-01 4:18 ` Ethan
2016-06-09 23:42 ` Dumitrescu, Cristian
2016-06-13 10:25 ` Dumitrescu, Cristian
2016-06-13 10:47 ` Ethan
2016-06-13 13:18 ` Dumitrescu, Cristian
2016-06-16 11:34 ` Ethan
2016-06-16 11:27 ` [dpdk-dev] [PATCH v3 1/3] " WeiJie Zhuang
2016-06-16 11:27 ` [dpdk-dev] [PATCH v3 2/3] port: add kni nodrop writer WeiJie Zhuang
2016-06-18 21:47 ` Dumitrescu, Cristian
2016-06-16 11:27 ` [dpdk-dev] [PATCH v3 3/3] port: document update WeiJie Zhuang
2016-06-18 16:44 ` Dumitrescu, Cristian [this message]
2016-06-21 11:10 ` [dpdk-dev] [PATCH v3 1/3] port: add kni interface support Ethan
2016-06-21 11:31 ` Dumitrescu, Cristian
2016-06-21 10:55 ` [dpdk-dev] [PATCH v4 1/4] port: " Ethan Zhuang
2016-06-21 10:55 ` [dpdk-dev] [PATCH v4 2/4] examples/ip_pipeline: " Ethan Zhuang
2016-06-21 10:55 ` [dpdk-dev] [PATCH v4 3/4] examples/ip_pipeline: kni example configuration Ethan Zhuang
2016-06-21 10:55 ` [dpdk-dev] [PATCH v4 4/4] doc: kni port support in the packet framework Ethan Zhuang
2016-06-21 12:03 ` [dpdk-dev] [PATCH v4 1/4] port: kni interface support Dumitrescu, Cristian
2016-06-21 16:09 ` Thomas Monjalon
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=3EB4FA525960D640B5BDFFD6A3D8912647A08FAC@IRSMSX108.ger.corp.intel.com \
--to=cristian.dumitrescu@intel.com \
--cc=dev@dpdk.org \
--cc=ferruh.yigit@intel.com \
--cc=jasvinder.singh@intel.com \
--cc=zhuangwj@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).