From: Jeff Guo <jia.guo@intel.com>
To: "Xie, WeiX" <weix.xie@intel.com>,
"Wang, ShougangX" <shougangx.wang@intel.com>,
"dev@dpdk.org" <dev@dpdk.org>
Cc: "Xing, Beilei" <beilei.xing@intel.com>,
"stable@dpdk.org" <stable@dpdk.org>
Subject: Re: [dpdk-dev] [PATCH v2] net/i40e: fix incorrect hash look up table
Date: Wed, 22 Jul 2020 13:50:46 +0800 [thread overview]
Message-ID: <334f916e-1aad-264f-e073-6e737e78c767@intel.com> (raw)
In-Reply-To: <6FD6A7610D20924F885A4ECF34E8AC910463D264@CDSMSX102.ccr.corp.intel.com>
hi, shougang
On 7/21/2020 2:41 PM, Xie, WeiX wrote:
> Tested-by: Zhang, XiX <xix.zhang@intel.com>
>
> Regards,
> Xie Wei
>
>
> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Shougang Wang
> Sent: Tuesday, July 21, 2020 1:49 PM
> To: dev@dpdk.org
> Cc: Xing, Beilei <beilei.xing@intel.com>; Guo, Jia <jia.guo@intel.com>; Wang, ShougangX <shougangx.wang@intel.com>; stable@dpdk.org
> Subject: [dpdk-dev] [PATCH v2] net/i40e: fix incorrect hash look up table
>
> The hash look up table(LUT) will not be initializing when starting testpmd with --disable-rss. So that some invalid queue indexes may still in the LUT. When enable RSS by creating RSS rule, some packets will not be into the valid queues.
> This patch fixes this issue by initializing the LUT when creating an RSS rule.
>
> Fixes: feaae285b342 ("net/i40e: support hash configuration in RSS flow")
> Cc: stable@dpdk.org
>
> Signed-off-by: Shougang Wang <shougangx.wang@intel.com>
> ---
> drivers/net/i40e/i40e_ethdev.c | 134 ++++++++++++++++-----------------
> 1 file changed, 63 insertions(+), 71 deletions(-)
>
> diff --git a/drivers/net/i40e/i40e_ethdev.c b/drivers/net/i40e/i40e_ethdev.c index 393b5320f..e56543393 100644
> --- a/drivers/net/i40e/i40e_ethdev.c
> +++ b/drivers/net/i40e/i40e_ethdev.c
> @@ -13070,6 +13070,55 @@ i40e_rss_conf_init(struct i40e_rte_flow_rss_conf *out,
> return 0;
> }
>
> +/* If conf is NULL, function will init hash LUT with default
> +configration*/
Please fix the checkpatch issue here.
> static int i40e_rss_set_lut(struct i40e_pf *pf,
In order to eliminate any confuse with current i40e_set_rss_lut, please
change the name, such as "i40e_rss_lut_init" or other better naming.
> + struct i40e_rte_flow_rss_conf *conf) {
> + struct i40e_hw *hw = I40E_PF_TO_HW(pf);
> + uint32_t lut = 0;
> + uint16_t j, num;
> + uint32_t i;
> +
> + /* If both VMDQ and RSS enabled, not all of PF queues are configured.
> + * It's necessary to calculate the actual PF queues that are configured.
> + */
> + if (pf->dev_data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_VMDQ_FLAG)
> + num = i40e_pf_calc_configured_queues_num(pf);
> + else
> + num = pf->dev_data->nb_rx_queues;
> +
> + if (conf == NULL)
> + num = RTE_MIN(num, I40E_MAX_Q_PER_TC);
> + else
> + num = RTE_MIN(num, conf->conf.queue_num);
> + PMD_DRV_LOG(INFO, "Max of contiguous %u PF queues are configured",
> + num);
Alignment should match open parenthesis.
> +
> + if (num == 0) {
> + PMD_DRV_LOG(ERR,
> + "No PF queues are configured to enable RSS for port %u",
> + pf->dev_data->port_id);
> + return -ENOTSUP;
> + }
> +
> + /* Fill in redirection table */
> + for (i = 0, j = 0; i < hw->func_caps.rss_table_size; i++, j++) {
> + if (j == num)
> + j = 0;
> + if (conf == NULL)
> + lut = (lut << 8) | (j & ((0x1 <<
> + hw->func_caps.rss_table_entry_width) - 1));
> + else
> + lut = (lut << 8) | (conf->conf.queue[j] & ((0x1 <<
> + hw->func_caps.rss_table_entry_width) - 1));
> + if ((i & 3) == 3)
> + I40E_WRITE_REG(hw, I40E_PFQF_HLUT(i >> 2), lut);
> + }
> +
> + return 0;
> +}
> +
> /* Write HENA register to enable hash */ static int i40e_rss_hash_set(struct i40e_pf *pf, struct i40e_rte_flow_rss_conf *rss_conf) @@ -13318,12 +13367,24 @@ static int i40e_rss_enable_hash(struct i40e_pf *pf,
> struct i40e_rte_flow_rss_conf *conf)
> {
> + enum rte_eth_rx_mq_mode mq_mode =
> +pf->dev_data->dev_conf.rxmode.mq_mode;
> struct i40e_rte_flow_rss_conf *rss_info = &pf->rss_info;
> struct i40e_rte_flow_rss_conf rss_conf;
> + int ret;
Suggest set the 0 to ret and return ret at the end of this function, so
ret could be use in all part.
>
> if (!(conf->conf.types & pf->adapter->flow_types_mask))
> return -ENOTSUP;
>
> + /* If the RSS is disabled before this, the LUT is uninitialized.
> + * So it is necessary to initialize it here.
> + */
> + if (!(mq_mode & ETH_MQ_RX_RSS_FLAG) && !pf->rss_info.conf.queue_num &&
> + !pf->adapter->rss_reta_updated) {
> + ret = i40e_rss_set_lut(pf, NULL);
> + if (ret)
> + return ret;
> + }
> +
> memset(&rss_conf, 0, sizeof(rss_conf));
> rte_memcpy(&rss_conf, conf, sizeof(rss_conf));
>
> @@ -13362,39 +13423,7 @@ static int
> i40e_rss_config_queue_region(struct i40e_pf *pf,
> struct i40e_rte_flow_rss_conf *conf)
> {
> - struct i40e_hw *hw = I40E_PF_TO_HW(pf);
> - uint32_t lut = 0;
> - uint16_t j, num;
> - uint32_t i;
> -
> - /* If both VMDQ and RSS enabled, not all of PF queues are configured.
> - * It's necessary to calculate the actual PF queues that are configured.
> - */
> - if (pf->dev_data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_VMDQ_FLAG)
> - num = i40e_pf_calc_configured_queues_num(pf);
> - else
> - num = pf->dev_data->nb_rx_queues;
> -
> - num = RTE_MIN(num, conf->conf.queue_num);
> - PMD_DRV_LOG(INFO, "Max of contiguous %u PF queues are configured",
> - num);
> -
> - if (num == 0) {
> - PMD_DRV_LOG(ERR,
> - "No PF queues are configured to enable RSS for port %u",
> - pf->dev_data->port_id);
> - return -ENOTSUP;
> - }
> -
> - /* Fill in redirection table */
> - for (i = 0, j = 0; i < hw->func_caps.rss_table_size; i++, j++) {
> - if (j == num)
> - j = 0;
> - lut = (lut << 8) | (conf->conf.queue[j] & ((0x1 <<
> - hw->func_caps.rss_table_entry_width) - 1));
> - if ((i & 3) == 3)
> - I40E_WRITE_REG(hw, I40E_PFQF_HLUT(i >> 2), lut);
> - }
> + i40e_rss_set_lut(pf, conf);
>
> i40e_rss_mark_invalid_rule(pf, conf);
>
> @@ -13491,46 +13520,9 @@ i40e_rss_disable_hash(struct i40e_pf *pf, static int i40e_rss_clear_queue_region(struct i40e_pf *pf) {
> - struct i40e_hw *hw = I40E_PF_TO_HW(pf);
> struct i40e_rte_flow_rss_conf *rss_info = &pf->rss_info;
> - uint16_t queue[I40E_MAX_Q_PER_TC];
> - uint32_t num_rxq, i;
> - uint32_t lut = 0;
> - uint16_t j, num;
> -
> - num_rxq = RTE_MIN(pf->dev_data->nb_rx_queues, I40E_MAX_Q_PER_TC);
>
> - for (j = 0; j < num_rxq; j++)
> - queue[j] = j;
> -
> - /* If both VMDQ and RSS enabled, not all of PF queues are configured.
> - * It's necessary to calculate the actual PF queues that are configured.
> - */
> - if (pf->dev_data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_VMDQ_FLAG)
> - num = i40e_pf_calc_configured_queues_num(pf);
> - else
> - num = pf->dev_data->nb_rx_queues;
> -
> - num = RTE_MIN(num, num_rxq);
> - PMD_DRV_LOG(INFO, "Max of contiguous %u PF queues are configured",
> - num);
> -
> - if (num == 0) {
> - PMD_DRV_LOG(ERR,
> - "No PF queues are configured to enable RSS for port %u",
> - pf->dev_data->port_id);
> - return -ENOTSUP;
> - }
> -
> - /* Fill in redirection table */
> - for (i = 0, j = 0; i < hw->func_caps.rss_table_size; i++, j++) {
> - if (j == num)
> - j = 0;
> - lut = (lut << 8) | (queue[j] & ((0x1 <<
> - hw->func_caps.rss_table_entry_width) - 1));
> - if ((i & 3) == 3)
> - I40E_WRITE_REG(hw, I40E_PFQF_HLUT(i >> 2), lut);
> - }
> + i40e_rss_set_lut(pf, NULL);
Need to check return value.
>
> rss_info->conf.queue_num = 0;
> memset(&rss_info->conf.queue, 0, sizeof(uint16_t));
> --
> 2.17.1
>
next prev parent reply other threads:[~2020-07-22 5:50 UTC|newest]
Thread overview: 23+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-07-15 6:35 [dpdk-dev] [PATCH] " Shougang Wang
2020-07-15 9:07 ` Chen, BoX C
2020-07-18 4:07 ` Jeff Guo
2020-07-21 5:49 ` [dpdk-dev] [PATCH v2] " Shougang Wang
2020-07-21 6:41 ` Xie, WeiX
2020-07-22 5:50 ` Jeff Guo [this message]
2020-07-22 6:39 ` Wang, ShougangX
2020-07-22 5:31 ` Zhang, Qi Z
2020-07-22 6:20 ` Wang, ShougangX
2020-07-22 8:15 ` [dpdk-dev] [PATCH v3] " Shougang Wang
2020-07-23 1:57 ` Yang, Qiming
2020-07-23 12:53 ` Zhang, Qi Z
2020-07-24 2:34 ` Wang, ShougangX
2020-07-24 2:47 ` [dpdk-dev] [PATCH v4] " Shougang Wang
2020-07-24 3:57 ` Jeff Guo
2020-07-24 4:49 ` Wang, ShougangX
2020-07-24 5:07 ` Yang, Qiming
2020-07-24 5:42 ` Wang, ShougangX
2020-07-24 8:12 ` [dpdk-dev] [PATCH v5] " Shougang Wang
2020-07-24 9:00 ` Xie, WeiX
2020-07-24 9:38 ` [dpdk-dev] [PATCH v6] " Shougang Wang
2020-07-24 9:54 ` Jeff Guo
2020-07-24 9:58 ` Zhang, Qi Z
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=334f916e-1aad-264f-e073-6e737e78c767@intel.com \
--to=jia.guo@intel.com \
--cc=beilei.xing@intel.com \
--cc=dev@dpdk.org \
--cc=shougangx.wang@intel.com \
--cc=stable@dpdk.org \
--cc=weix.xie@intel.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).