[dpdk-dev] [PATCH] net/mlx5: allow install more meter actions
Suanming Mou
suanmingm at mellanox.com
Wed Jan 8 14:28:13 CET 2020
Hi,
> -----Original Message-----
> From: Tonghao Zhang <xiangxia.m.yue at gmail.com>
> Sent: Monday, January 6, 2020 9:47 AM
> To: Suanming Mou <suanmingm at mellanox.com>
> Cc: dev at dpdk.org; stable at dpdk.org; Matan Azrad <matan at mellanox.com>;
> Slava Ovsiienko <viacheslavo at mellanox.com>
> Subject: Re: [PATCH] net/mlx5: allow install more meter actions
>
> On Fri, Jan 3, 2020 at 11:38 AM Suanming Mou <suanmingm at mellanox.com>
> wrote:
> >
> > Hello Tonghao,
> >
> > Could you please explain much detail about your issue scenario?
> >
> > If I understand correctly, you are trying to create two flows with the two same
> match criteria?
> >
> > Example from testpmd just like that:
> > add port meter profile srtcm_rfc2697 0 24 65536 32768 0 create port
> > meter 0 0 24 yes G Y D 0xffff 1 0 flow create 0 ingress pattern eth /
> > end actions jump group 1 / end flow create 0 priority 3 group 1
> > ingress pattern eth / end actions meter mtr_id 0 / queue index 0 / end
> > flow create 0 priority 3 group 1 ingress pattern eth / end actions
> > meter mtr_id 0 / queue index 1 / end
> >
> > Then the third flow will report "hardware refuses to create flow: Invalid
> argument".
> >
> > Please correct me if I'm wrong.
> > And better to give the issue reproduce method via testpmd cmdline.
> ./testpmd -l 0-3 -n 4 -w 0000:82:00.0 -w 0000:82:00.1 -- -i
> --portmask=0x3 --rxq=2 --nb-cores=2
>
> add port meter profile srtcm_rfc2697 0 100 1024 1024 0 add port meter profile
> srtcm_rfc2697 0 101 2048 2048 0 create port meter 0 100 100 yes G Y D 0xffff 1
> 0 create port meter 0 101 101 yes G Y D 0xffff 1 0
>
> flow create 0 ingress pattern eth / end actions jump group 1 / end flow create 0
> group 1 priority 3 ingress pattern eth / ipv4 dst is
> 1.1.1.100 / end actions meter mtr_id 100 / queue index 0 / end
>
> The one of commands below will run fail:
> flow create 0 group 1 priority 3 ingress pattern eth / ipv4 dst is
> 1.1.1.200 / end actions meter mtr_id 101 / queue index 1 / end flow create 0
> group 1 priority 3 ingress pattern eth / ipv4 dst is
> 1.1.1.200 / tcp dst is 80 / end actions meter mtr_id 101 / queue index 1 / end
> flow create 0 group 1 priority 3 ingress pattern eth / ipv4 / tcp dst is 80 / end
> actions meter mtr_id 101 / queue index 1 / end
>
> Caught error type 1 (cause unspecified): hardware refuses to create
> flow: Invalid argument
>
> And with this patch, it works fine.
> > Thanks
> > SuanmingMou
> >
> > > -----Original Message-----
> > > From: Tonghao Zhang <xiangxia.m.yue at gmail.com>
> > > Sent: Thursday, December 19, 2019 6:00 PM
> > > To: Suanming Mou <suanmingm at mellanox.com>
> > > Cc: dev at dpdk.org; stable at dpdk.org
> > > Subject: Re: [PATCH] net/mlx5: allow install more meter actions
> > >
> > > ping
> > >
> > > On Tue, Dec 17, 2019 at 3:29 PM <xiangxia.m.yue at gmail.com> wrote:
> > > >
> > > > From: Tonghao Zhang <xiangxia.m.yue at gmail.com>
> > > >
> > > > When creating the dr rule of meter, the matcher which struct is
> > > > "struct mlx5dv_dr_matcher" should not be shared, if shared,
> > > > mlx5dv_dr_rule_create will return NULL. We can't install more
> > > > metering offload actions.
> > > >
> > > > The call tree (rdma-core-47mlnx1 OFED 4.7-3.2.9):
> > > > * dr_rule_handle_ste_branch
> > > > * dr_rule_create_rule_nic
> > > > * dr_rule_create_rule_fdb
> > > > * dr_rule_create_rule
> > > > * mlx5dv_dr_rule_create
> > > >
> > > > In the dr_rule_handle_ste_branch, if ste is not used,
> > > > mlx5dv_dr_rule_create will return rule, if the ste is used, and
> > > > the ste is the last in the rule, mlx5dv_dr_rule_create will return NULL.
> > > >
> > > > dr_rule_handle_ste_branch:
> > > > if dr_ste_not_used_ste
> > > > dr_rule_handle_empty_entry
> > > > else
> > > > dr_rule_find_ste_in_miss_list
> > > > dr_ste_is_last_in_rule: if so return NULL and set errno =
> > > > EINVAL;
> > > >
> > > > Fixes: 9ea9b049a960 ("net/mlx5: split meter flow")
> > > > Cc: Suanming Mou <suanmingm at mellanox.com>
> > > > Cc: stable at dpdk.org
> > > >
> > > > Signed-off-by: Tonghao Zhang <xiangxia.m.yue at gmail.com>
NACK, as we have another fix about that:
https://patches.dpdk.org/patch/64288/
> > > > ---
> > > > drivers/net/mlx5/mlx5_flow.c | 20 +++++++++++++-------
> > > > drivers/net/mlx5/mlx5_flow.h | 2 ++
> > > > drivers/net/mlx5/mlx5_flow_dv.c | 5 +++++
> > > > 3 files changed, 20 insertions(+), 7 deletions(-)
> > > >
> > > > diff --git a/drivers/net/mlx5/mlx5_flow.c
> > > > b/drivers/net/mlx5/mlx5_flow.c index 0087163..f8cdc25 100644
> > > > --- a/drivers/net/mlx5/mlx5_flow.c
> > > > +++ b/drivers/net/mlx5/mlx5_flow.c
> > > > @@ -3421,7 +3421,9 @@ uint32_t mlx5_flow_adjust_priority(struct
> > > rte_eth_dev *dev, int32_t priority,
> > > > const struct rte_flow_attr *attr,
> > > > const struct rte_flow_item items[],
> > > > const struct rte_flow_action actions[],
> > > > - bool external, struct rte_flow_error *error)
> > > > + bool external,
> > > > + bool shared,
> > > > + struct rte_flow_error *error)
> > > > {
> > > > struct mlx5_flow *dev_flow;
> > > >
> > > > @@ -3434,6 +3436,7 @@ uint32_t mlx5_flow_adjust_priority(struct
> > > rte_eth_dev *dev, int32_t priority,
> > > > LIST_INSERT_HEAD(&flow->dev_flows, dev_flow, next);
> > > > if (sub_flow)
> > > > *sub_flow = dev_flow;
> > > > + dev_flow->matcher_shared = shared;
> > > > return flow_drv_translate(dev, dev_flow, attr, items,
> > > > actions, error); }
> > > >
> > > > @@ -3741,7 +3744,9 @@ uint32_t mlx5_flow_adjust_priority(struct
> > > rte_eth_dev *dev, int32_t priority,
> > > > const struct rte_flow_attr *attr,
> > > > const struct rte_flow_item items[],
> > > > const struct rte_flow_action actions[],
> > > > - bool external, struct rte_flow_error *error)
> > > > + bool external,
> > > > + bool shared,
> > > > + struct rte_flow_error *error)
> > > > {
> > > > struct mlx5_priv *priv = dev->data->dev_private;
> > > > struct mlx5_dev_config *config = &priv->config; @@ -3759,7
> > > > +3764,8 @@ uint32_t mlx5_flow_adjust_priority(struct rte_eth_dev
> > > > +*dev,
> > > int32_t priority,
> > > > config->dv_xmeta_en == MLX5_XMETA_MODE_LEGACY ||
> > > > !mlx5_flow_ext_mreg_supported(dev))
> > > > return flow_create_split_inner(dev, flow, NULL, attr, items,
> > > > - actions, external, error);
> > > > + actions, external,
> > > > + shared, error);
> > > > actions_n = flow_parse_qrss_action(actions, &qrss);
> > > > if (qrss) {
> > > > /* Exclude hairpin flows from splitting. */ @@
> > > > -3842,7
> > > > +3848,7 @@ uint32_t mlx5_flow_adjust_priority(struct rte_eth_dev
> > > > +*dev,
> > > int32_t priority,
> > > > /* Add the unmodified original or prefix subflow. */
> > > > ret = flow_create_split_inner(dev, flow, &dev_flow, attr, items,
> > > > ext_actions ? ext_actions : actions,
> > > > - external, error);
> > > > + external, shared, error);
> > > > if (ret < 0)
> > > > goto exit;
> > > > assert(dev_flow);
> > > > @@ -3906,7 +3912,7 @@ uint32_t mlx5_flow_adjust_priority(struct
> > > rte_eth_dev *dev, int32_t priority,
> > > > ret = flow_create_split_inner(dev, flow, &dev_flow,
> > > > &q_attr, mtr_sfx ? items :
> > > > q_items, q_actions,
> > > > - external, error);
> > > > + external, shared,
> > > > + error);
> > > > if (ret < 0)
> > > > goto exit;
> > > > assert(dev_flow);
> > > > @@ -3999,7 +4005,7 @@ uint32_t mlx5_flow_adjust_priority(struct
> > > rte_eth_dev *dev, int32_t priority,
> > > > }
> > > > /* Add the prefix subflow. */
> > > > ret = flow_create_split_inner(dev, flow, &dev_flow, attr, items,
> > > > - pre_actions, external, error);
> > > > + pre_actions,
> > > > + external, false, error);
> > > > if (ret) {
> > > > ret = -rte_errno;
> > > > goto exit; @@ -4035,7 +4041,7 @@ uint32_t
> > > > mlx5_flow_adjust_priority(struct
> > > rte_eth_dev *dev, int32_t priority,
> > > > ret = flow_create_split_metadata(dev, flow, &sfx_attr,
> > > > sfx_items ? sfx_items : items,
> > > > sfx_actions ? sfx_actions : actions,
> > > > - external, error);
> > > > + external, sfx_items ==
> > > > + NULL, error);
> > > > exit:
> > > > if (sfx_actions)
> > > > rte_free(sfx_actions); diff --git
> > > > a/drivers/net/mlx5/mlx5_flow.h b/drivers/net/mlx5/mlx5_flow.h
> > > > index 3fff5dd..84636be 100644
> > > > --- a/drivers/net/mlx5/mlx5_flow.h
> > > > +++ b/drivers/net/mlx5/mlx5_flow.h
> > > > @@ -338,6 +338,7 @@ struct mlx5_flow_dv_matcher {
> > > > /**< Pointer to the table(group) the matcher associated with. */
> > > > rte_atomic32_t refcnt; /**< Reference counter. */
> > > > void *matcher_object; /**< Pointer to DV matcher */
> > > > + bool shared;
> > > > uint16_t crc; /**< CRC of key. */
> > > > uint16_t priority; /**< Priority of matcher. */
> > > > struct mlx5_flow_dv_match_params mask; /**< Matcher mask.
> > > > */ @@ -532,6 +533,7 @@ struct mlx5_flow {
> > > > uint32_t mtr_flow_id; /**< Unique meter match flow id. */
> > > > };
> > > > bool external; /**< true if the flow is created external
> > > > to PMD. */
> > > > + bool matcher_shared;
> > > > };
> > > >
> > > > /* Flow meter state. */
> > > > diff --git a/drivers/net/mlx5/mlx5_flow_dv.c
> > > > b/drivers/net/mlx5/mlx5_flow_dv.c index 7528556..362c36c 100644
> > > > --- a/drivers/net/mlx5/mlx5_flow_dv.c
> > > > +++ b/drivers/net/mlx5/mlx5_flow_dv.c
> > > > @@ -6324,10 +6324,13 @@ struct field_modify_info modify_tcp[] = {
> > > > if (!tbl)
> > > > return -rte_errno; /* No need to refill the error info */
> > > > tbl_data = container_of(tbl, struct
> > > > mlx5_flow_tbl_data_entry, tbl);
> > > > + if (!dev_flow->matcher_shared)
> > > > + goto create_matcher;
> > > > /* Lookup from cache. */
> > > > LIST_FOREACH(cache_matcher, &tbl_data->matchers, next) {
> > > > if (matcher->crc == cache_matcher->crc &&
> > > > matcher->priority == cache_matcher->priority
> > > > &&
> > > > + cache_matcher->shared &&
> > > > !memcmp((const void *)matcher->mask.buf,
> > > > (const void *)cache_matcher->mask.buf,
> > > > cache_matcher->mask.size)) { @@
> > > > -6346,6
> > > > +6349,7 @@ struct field_modify_info modify_tcp[] = {
> > > > return 0;
> > > > }
> > > > }
> > > > +create_matcher:
> > > > /* Register new matcher. */
> > > > cache_matcher = rte_calloc(__func__, 1, sizeof(*cache_matcher), 0);
> > > > if (!cache_matcher) {
> > > > @@ -6355,6 +6359,7 @@ struct field_modify_info modify_tcp[] = {
> > > > "cannot allocate matcher memory");
> > > > }
> > > > *cache_matcher = *matcher;
> > > > + cache_matcher->shared = dev_flow->matcher_shared;
> > > > dv_attr.match_criteria_enable =
> > > > flow_dv_matcher_enable(cache_matcher->mask.buf);
> > > > dv_attr.priority = matcher->priority;
> > > > --
> > > > 1.8.3.1
> > > >
More information about the dev
mailing list