From: Jakub Kicinski <kuba@kernel.org>
To: Michael Chan <michael.chan@broadcom.com>
Cc: davem@davemloft.net, netdev@vger.kernel.org,
Sriharsha Basavapatna <sriharsha.basavapatna@broadcom.com>
Subject: Re: [PATCH net-next 01/16] bnxt_en: Support ingress rate limiting with TC-offload.
Date: Sun, 26 Jan 2020 16:12:56 -0800 [thread overview]
Message-ID: <20200126161256.6cf37aad@cakuba> (raw)
In-Reply-To: <1580029390-32760-2-git-send-email-michael.chan@broadcom.com>
On Sun, 26 Jan 2020 04:02:55 -0500, Michael Chan wrote:
> From: Sriharsha Basavapatna <sriharsha.basavapatna@broadcom.com>
>
> This patch enables offloading of ingress rate limiting TC-action
> on a VF. The driver processes "cls_matchall" filter callbacks to
> add and remove ingress rate limiting actions. The driver parses
> police action parameter and sends the command to FW to configure
> rate limiting for the VF.
>
> For example, to configure rate limiting offload on a VF using OVS,
> use the below command on the corresponding VF-rep port. The example
> below configures min and max tx rates of 200 and 600 Mbps.
>
> # ovs-vsctl set interface bnxt0_pf0vf0 \
> ingress_policing_rate=600000 ingress_policing_burst=200000
>
> Signed-off-by: Sriharsha Basavapatna <sriharsha.basavapatna@broadcom.com>
> Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Does the device drop or back-pressure when VF goes over the rate?
> diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.h b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
> index f143354..534bc9e 100644
> --- a/drivers/net/ethernet/broadcom/bnxt/bnxt.h
> +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
> @@ -1069,6 +1069,7 @@ struct bnxt_vf_info {
> u32 max_tx_rate;
> void *hwrm_cmd_req_addr;
> dma_addr_t hwrm_cmd_req_dma_addr;
> + unsigned long police_id;
> };
> #endif
>
> diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_tc.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_tc.c
> index 0cc6ec5..2dfb650 100644
> --- a/drivers/net/ethernet/broadcom/bnxt/bnxt_tc.c
> +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_tc.c
> @@ -10,6 +10,7 @@
> #include <linux/netdevice.h>
> #include <linux/inetdevice.h>
> #include <linux/if_vlan.h>
> +#include <linux/pci.h>
> #include <net/flow_dissector.h>
> #include <net/pkt_cls.h>
> #include <net/tc_act/tc_gact.h>
> @@ -1983,6 +1984,95 @@ static int bnxt_tc_indr_block_event(struct notifier_block *nb,
> return NOTIFY_DONE;
> }
>
> +static inline int bnxt_tc_find_vf_by_fid(struct bnxt *bp, u16 fid)
No static inlines in C files
> +{
> + int num_vfs = pci_num_vf(bp->pdev);
> + int i;
> +
> + for (i = 0; i < num_vfs; i++) {
> + if (bp->pf.vf[i].fw_fid == fid)
return i;
> + break;
> + }
return -EINVAL;
> + if (i >= num_vfs)
> + return -EINVAL;
> + return i;
> +}
> +static int bnxt_tc_add_matchall(struct bnxt *bp, u16 src_fid,
> + struct tc_cls_matchall_offload *matchall_cmd)
> +{
> + struct flow_action_entry *action;
> + int vf_idx;
> + s64 burst;
> + u64 rate;
> + int rc;
> +
> + vf_idx = bnxt_tc_find_vf_by_fid(bp, src_fid);
> + if (vf_idx < 0)
> + return vf_idx;
You need to check this is the only action, check priority, check
shared, etc. Just look as one of the drivers which do this right :/
> + action = &matchall_cmd->rule->action.entries[0];
> + if (action->id != FLOW_ACTION_POLICE) {
> + netdev_err(bp->dev, "%s: Unsupported matchall action: %d",
> + __func__, action->id);
> + return -EOPNOTSUPP;
> + }
> + if (bp->pf.vf[vf_idx].police_id && bp->pf.vf[vf_idx].police_id !=
> + matchall_cmd->cookie) {
> + netdev_err(bp->dev,
> + "%s: Policer is already configured for VF: %d",
> + __func__, vf_idx);
> + return -EEXIST;
> + }
> +
> + rate = (u32)div_u64(action->police.rate_bytes_ps, 1024 * 1000) * 8;
> + burst = (u32)div_u64(action->police.rate_bytes_ps *
> + PSCHED_NS2TICKS(action->police.burst),
> + PSCHED_TICKS_PER_SEC);
> + burst = (u32)PSCHED_TICKS2NS(burst) / (1 << 20);
> +
> + rc = bnxt_set_vf_bw(bp->dev, vf_idx, burst, rate);
> + if (rc) {
> + netdev_err(bp->dev,
> + "Error: %s: VF: %d rate: %llu burst: %llu rc: %d",
> + __func__, vf_idx, rate, burst, rc);
> + return rc;
> + }
> +
> + bp->pf.vf[vf_idx].police_id = matchall_cmd->cookie;
> + return 0;
> +}
> +
> +int bnxt_tc_setup_matchall(struct bnxt *bp, u16 src_fid,
> + struct tc_cls_matchall_offload *cls_matchall)
> +{
> + switch (cls_matchall->command) {
> + case TC_CLSMATCHALL_REPLACE:
> + return bnxt_tc_add_matchall(bp, src_fid, cls_matchall);
> + case TC_CLSMATCHALL_DESTROY:
> + return bnxt_tc_del_matchall(bp, src_fid, cls_matchall);
> + default:
> + return -EOPNOTSUPP;
> + }
> +}
> +
> static const struct rhashtable_params bnxt_tc_flow_ht_params = {
> .head_offset = offsetof(struct bnxt_tc_flow_node, node),
> .key_offset = offsetof(struct bnxt_tc_flow_node, cookie),
> diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_tc.h b/drivers/net/ethernet/broadcom/bnxt/bnxt_tc.h
> index 10c62b0..963788e 100644
> --- a/drivers/net/ethernet/broadcom/bnxt/bnxt_tc.h
> +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_tc.h
> @@ -220,6 +220,9 @@ int bnxt_tc_setup_flower(struct bnxt *bp, u16 src_fid,
> int bnxt_init_tc(struct bnxt *bp);
> void bnxt_shutdown_tc(struct bnxt *bp);
> void bnxt_tc_flow_stats_work(struct bnxt *bp);
> +int bnxt_tc_setup_matchall(struct bnxt *bp, u16 src_fid,
> + struct tc_cls_matchall_offload *cls_matchall);
> +
>
Spurious new line
> static inline bool bnxt_tc_flower_enabled(struct bnxt *bp)
> {
next prev parent reply other threads:[~2020-01-27 0:14 UTC|newest]
Thread overview: 23+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-01-26 9:02 [PATCH net-next 00/16] bnxt_en: Updates for net-next Michael Chan
2020-01-26 9:02 ` [PATCH net-next 01/16] bnxt_en: Support ingress rate limiting with TC-offload Michael Chan
2020-01-27 0:12 ` Jakub Kicinski [this message]
2020-01-26 9:02 ` [PATCH net-next 02/16] bnxt_en: Improve link up detection Michael Chan
2020-01-26 9:02 ` [PATCH net-next 03/16] bnxt_en: Improve bnxt_probe_phy() Michael Chan
2020-01-26 9:02 ` [PATCH net-next 04/16] bnxt_en: Remove the setting of dev_port Michael Chan
2020-01-26 9:02 ` [PATCH net-next 05/16] bnxt_en: Support UDP RSS hashing on 575XX chips Michael Chan
2020-01-26 9:03 ` [PATCH net-next 06/16] bnxt_en: Do not accept fragments for aRFS flow steering Michael Chan
2020-01-26 9:03 ` [PATCH net-next 07/16] bnxt_en: Periodically check and remove aged-out ntuple filters Michael Chan
2020-01-26 9:03 ` [PATCH net-next 08/16] bnxt_en: Disable workaround for lost interrupts on 575XX B0 and newer chips Michael Chan
2020-01-26 9:03 ` [PATCH net-next 09/16] bnxt_en: Refactor bnxt_dl_register() Michael Chan
2020-01-26 11:32 ` Jiri Pirko
2020-01-27 4:51 ` Vasundhara Volam
2020-01-26 9:03 ` [PATCH net-next 10/16] bnxt_en: Register devlink irrespective of firmware spec version Michael Chan
2020-01-26 9:03 ` [PATCH net-next 11/16] bnxt_en: Move devlink_register before registering netdev Michael Chan
2020-01-26 9:03 ` [PATCH net-next 12/16] bnxt_en: Add support to update progress of flash update Michael Chan
2020-01-26 9:03 ` [PATCH net-next 13/16] bnxt_en: Rename switch_id to dsn Michael Chan
2020-01-26 9:03 ` [PATCH net-next 14/16] devlink: add macros for "fw.roce" and "board.nvm_cfg" Michael Chan
2020-01-27 0:18 ` Jakub Kicinski
2020-01-27 5:42 ` Vasundhara Volam
2020-01-27 14:25 ` Jakub Kicinski
2020-01-26 9:03 ` [PATCH net-next 15/16] bnxt_en: Add support for devlink info command Michael Chan
2020-01-26 9:03 ` [PATCH net-next 16/16] devlink: document devlink info versions reported by bnxt_en driver Michael Chan
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=20200126161256.6cf37aad@cakuba \
--to=kuba@kernel.org \
--cc=davem@davemloft.net \
--cc=michael.chan@broadcom.com \
--cc=netdev@vger.kernel.org \
--cc=sriharsha.basavapatna@broadcom.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).