From: Kiran Kumar Kokkilagadda <kirankumark@marvell.com>
To: Robin Jarry <rjarry@redhat.com>, "dev@dpdk.org" <dev@dpdk.org>,
Jerin Jacob <jerinj@marvell.com>,
Nithin Kumar Dabilpuram <ndabilpuram@marvell.com>,
Zhirun Yan <yanzhirun_163@163.com>
Subject: RE: [EXTERNAL] [PATCH v2] graph: avoid accessing graph list when getting stats
Date: Wed, 3 Apr 2024 07:22:19 +0000 [thread overview]
Message-ID: <PH0PR18MB5071808B93878920C8FB0379AC3D2@PH0PR18MB5071.namprd18.prod.outlook.com> (raw)
In-Reply-To: <20240401203647.1909165-3-rjarry@redhat.com>
> -----Original Message-----
> From: Robin Jarry <rjarry@redhat.com>
> Sent: Tuesday, April 2, 2024 2:07 AM
> To: dev@dpdk.org; Jerin Jacob <jerinj@marvell.com>; Kiran Kumar
> Kokkilagadda <kirankumark@marvell.com>; Nithin Kumar Dabilpuram
> <ndabilpuram@marvell.com>; Zhirun Yan <yanzhirun_163@163.com>
> Subject: [EXTERNAL] [PATCH v2] graph: avoid accessing graph list when getting
> stats
>
> Prioritize security for external emails: Confirm sender and content safety
> before clicking links or opening attachments
>
> ----------------------------------------------------------------------
> In rte_graph_cluster_stats_get, the walk model of the first graph is checked
> to determine if multi-core dispatch specific counters should be updated or
> not. This global list is accessed without any locks.
>
> If the global list is modified by another thread while
> rte_graph_cluster_stats_get is called, it can result in undefined behaviour.
>
> Adding a lock would make it impossible to call rte_graph_cluster_stats_get in
> packet processing code paths. Avoid accessing the global list instead by
> storing a bool field in the private rte_graph_cluster_stats structure.
>
> Also update the default callback to avoid accessing the global list and use a
> different default callback depending on the graph model.
>
> Signed-off-by: Robin Jarry <rjarry@redhat.com>
> ---
Acked-by: Kiran Kumar Kokkilagadda <kirankumark@marvell.com>
>
> Notes:
> v2:
>
> * (kiran) removed unnecessary loop in stats_mem_init.
>
> lib/graph/graph_stats.c | 57 ++++++++++++++++++++++++++---------------
> 1 file changed, 36 insertions(+), 21 deletions(-)
>
> diff --git a/lib/graph/graph_stats.c b/lib/graph/graph_stats.c index
> 2fb808b21ec5..d71451a17b95 100644
> --- a/lib/graph/graph_stats.c
> +++ b/lib/graph/graph_stats.c
> @@ -34,6 +34,7 @@ struct __rte_cache_aligned rte_graph_cluster_stats {
> uint32_t cluster_node_size; /* Size of struct cluster_node */
> rte_node_t max_nodes;
> int socket_id;
> + bool dispatch;
> void *cookie;
> size_t sz;
>
> @@ -74,17 +75,16 @@ print_banner_dispatch(FILE *f) }
>
> static inline void
> -print_banner(FILE *f)
> +print_banner(FILE *f, bool dispatch)
> {
> - if
> (rte_graph_worker_model_get(STAILQ_FIRST(graph_list_head_get())->graph)
> ==
> - RTE_GRAPH_MODEL_MCORE_DISPATCH)
> + if (dispatch)
> print_banner_dispatch(f);
> else
> print_banner_default(f);
> }
>
> static inline void
> -print_node(FILE *f, const struct rte_graph_cluster_node_stats *stat)
> +print_node(FILE *f, const struct rte_graph_cluster_node_stats *stat,
> +bool dispatch)
> {
> double objs_per_call, objs_per_sec, cycles_per_call, ts_per_hz;
> const uint64_t prev_calls = stat->prev_calls; @@ -104,8 +104,7 @@
> print_node(FILE *f, const struct rte_graph_cluster_node_stats *stat)
> objs_per_sec = ts_per_hz ? (objs - prev_objs) / ts_per_hz : 0;
> objs_per_sec /= 1000000;
>
> - if
> (rte_graph_worker_model_get(STAILQ_FIRST(graph_list_head_get())->graph)
> ==
> - RTE_GRAPH_MODEL_MCORE_DISPATCH) {
> + if (dispatch) {
> fprintf(f,
> "|%-31s|%-15" PRIu64 "|%-15" PRIu64 "|%-15" PRIu64
> "|%-15" PRIu64 "|%-15" PRIu64
> @@ -123,20 +122,17 @@ print_node(FILE *f, const struct
> rte_graph_cluster_node_stats *stat) }
>
> static int
> -graph_cluster_stats_cb(bool is_first, bool is_last, void *cookie,
> +graph_cluster_stats_cb(bool dispatch, bool is_first, bool is_last, void
> +*cookie,
> const struct rte_graph_cluster_node_stats *stat) {
> FILE *f = cookie;
> - int model;
> -
> - model =
> rte_graph_worker_model_get(STAILQ_FIRST(graph_list_head_get())->graph);
>
> if (unlikely(is_first))
> - print_banner(f);
> + print_banner(f, dispatch);
> if (stat->objs)
> - print_node(f, stat);
> + print_node(f, stat, dispatch);
> if (unlikely(is_last)) {
> - if (model == RTE_GRAPH_MODEL_MCORE_DISPATCH)
> + if (dispatch)
> boarder_model_dispatch();
> else
> boarder();
> @@ -145,6 +141,20 @@ graph_cluster_stats_cb(bool is_first, bool is_last,
> void *cookie,
> return 0;
> };
>
> +static int
> +graph_cluster_stats_cb_rtc(bool is_first, bool is_last, void *cookie,
> + const struct rte_graph_cluster_node_stats *stat) {
> + return graph_cluster_stats_cb(false, is_first, is_last, cookie, stat);
> +};
> +
> +static int
> +graph_cluster_stats_cb_dispatch(bool is_first, bool is_last, void *cookie,
> + const struct rte_graph_cluster_node_stats
> *stat) {
> + return graph_cluster_stats_cb(true, is_first, is_last, cookie, stat);
> +};
> +
> static struct rte_graph_cluster_stats * stats_mem_init(struct cluster *cluster,
> const struct rte_graph_cluster_stats_param *prm) @@ -157,8
> +167,13 @@ stats_mem_init(struct cluster *cluster,
>
> /* Fix up callback */
> fn = prm->fn;
> - if (fn == NULL)
> - fn = graph_cluster_stats_cb;
> + if (fn == NULL) {
> + const struct rte_graph *graph = cluster->graphs[0]->graph;
> + if (graph->model == RTE_GRAPH_MODEL_MCORE_DISPATCH)
> + fn = graph_cluster_stats_cb_dispatch;
> + else
> + fn = graph_cluster_stats_cb_rtc;
> + }
>
> cluster_node_size = sizeof(struct cluster_node);
> /* For a given cluster, max nodes will be the max number of graphs
> */ @@ -350,6 +365,8 @@ rte_graph_cluster_stats_create(const struct
> rte_graph_cluster_stats_param *prm)
> if (stats_mem_populate(&stats, graph_fp,
> graph_node))
> goto realloc_fail;
> }
> + if (graph->graph->model ==
> RTE_GRAPH_MODEL_MCORE_DISPATCH)
> + stats->dispatch = true;
> }
>
> /* Finally copy to hugepage memory to avoid pressure on rte_realloc
> */ @@ -375,20 +392,18 @@ rte_graph_cluster_stats_destroy(struct
> rte_graph_cluster_stats *stat) }
>
> static inline void
> -cluster_node_arregate_stats(struct cluster_node *cluster)
> +cluster_node_arregate_stats(struct cluster_node *cluster, bool
> +dispatch)
> {
> uint64_t calls = 0, cycles = 0, objs = 0, realloc_count = 0;
> struct rte_graph_cluster_node_stats *stat = &cluster->stat;
> uint64_t sched_objs = 0, sched_fail = 0;
> struct rte_node *node;
> rte_node_t count;
> - int model;
>
> - model =
> rte_graph_worker_model_get(STAILQ_FIRST(graph_list_head_get())->graph);
> for (count = 0; count < cluster->nb_nodes; count++) {
> node = cluster->nodes[count];
>
> - if (model == RTE_GRAPH_MODEL_MCORE_DISPATCH) {
> + if (dispatch) {
> sched_objs += node->dispatch.total_sched_objs;
> sched_fail += node->dispatch.total_sched_fail;
> }
> @@ -403,7 +418,7 @@ cluster_node_arregate_stats(struct cluster_node
> *cluster)
> stat->objs = objs;
> stat->cycles = cycles;
>
> - if (model == RTE_GRAPH_MODEL_MCORE_DISPATCH) {
> + if (dispatch) {
> stat->dispatch.sched_objs = sched_objs;
> stat->dispatch.sched_fail = sched_fail;
> }
> @@ -433,7 +448,7 @@ rte_graph_cluster_stats_get(struct
> rte_graph_cluster_stats *stat, bool skip_cb)
> cluster = stat->clusters;
>
> for (count = 0; count < stat->max_nodes; count++) {
> - cluster_node_arregate_stats(cluster);
> + cluster_node_arregate_stats(cluster, stat->dispatch);
> if (!skip_cb)
> rc = stat->fn(!count, (count == stat->max_nodes - 1),
> stat->cookie, &cluster->stat);
> --
> 2.44.0
next prev parent reply other threads:[~2024-04-03 7:22 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-03-25 15:53 [PATCH] " Robin Jarry
2024-03-29 7:44 ` [EXTERNAL] " Kiran Kumar Kokkilagadda
2024-03-29 9:32 ` Robin Jarry
2024-04-01 20:36 ` [PATCH v2] " Robin Jarry
2024-04-03 7:22 ` Kiran Kumar Kokkilagadda [this message]
2024-06-18 13:15 ` David Marchand
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=PH0PR18MB5071808B93878920C8FB0379AC3D2@PH0PR18MB5071.namprd18.prod.outlook.com \
--to=kirankumark@marvell.com \
--cc=dev@dpdk.org \
--cc=jerinj@marvell.com \
--cc=ndabilpuram@marvell.com \
--cc=rjarry@redhat.com \
--cc=yanzhirun_163@163.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).