Re: [dpdk-dev] [PATCH] net/hns3: add runtime config to set MBX limit time

[Min Hu (Connor)]

Hi, Ferruh,

在 2021/9/9 21:20, Ferruh Yigit 写道:

On 8/30/2021 4:48 AM, Min Hu (Connor) wrote:

From: Chengchang Tang <tangchengch...@huawei.com>

Current, the max waiting time for MBX response is 500ms, but in
some scenarios, it is not enough. Since it depends on the response
of the kernel mode driver, and its response time is related to the
scheduling of the system. In this special scenario, most of the
cores are isolated, and only a few cores are used for system
scheduling. When a large number of services are started, the
scheduling of the system will be very busy, and the reply of the
mbx message will time out, which will cause our PMD initialization
to fail.

This patch add a runtime config to set the max wait time. For the
above scenes, users can adjust the waiting time to a suitable value
by themselves.

Fixes: 463e748964f5 ("net/hns3: support mailbox")
Cc: sta...@dpdk.org

Signed-off-by: Chengchang Tang <tangchengch...@huawei.com>
Signed-off-by: Min Hu (Connor) <humi...@huawei.com>
---
  drivers/net/hns3/hns3_ethdev.c    | 26 +++++++++++++++++++++++++-
  drivers/net/hns3/hns3_ethdev.h    |  3 +++
  drivers/net/hns3/hns3_ethdev_vf.c |  3 ++-
  drivers/net/hns3/hns3_mbx.c       | 10 +++++++---
  drivers/net/hns3/hns3_mbx.h       |  1 +
  5 files changed, 38 insertions(+), 5 deletions(-)

diff --git a/drivers/net/hns3/hns3_ethdev.c b/drivers/net/hns3/hns3_ethdev.c
index 7d37004..4c1a6f1 100644
--- a/drivers/net/hns3/hns3_ethdev.c
+++ b/drivers/net/hns3/hns3_ethdev.c
@@ -7357,9 +7357,24 @@ hns3_parse_dev_caps_mask(const char *key, const char 
*value, void *extra_args)
        return 0;
  }

+static int

+hns3_parse_mbx_time_limit(const char *key, const char *value, void *extra_args)
+{
+       uint32_t val;
+
+       RTE_SET_USED(key);
+
+       val = strtoul(value, NULL, 10);
+       if (val > HNS3_MBX_DEF_TIME_LIMIT_MS && val <= UINT16_MAX)

This check adds a restriction that timeout can't be set less the 500ms but this
restriction is not documented in the help string, comment or commit log etc...
Can you either remove the restriction or document it?

Fixed in v2, please check it out.

+               *(uint16_t *)extra_args = val;
+
+       return 0;
+}
+
  void
  hns3_parse_devargs(struct rte_eth_dev *dev)
  {
+       uint16_t mbx_time_limit_ms = HNS3_MBX_DEF_TIME_LIMIT_MS;
        struct hns3_adapter *hns = dev->data->dev_private;
        uint32_t rx_func_hint = HNS3_IO_FUNC_HINT_NONE;
        uint32_t tx_func_hint = HNS3_IO_FUNC_HINT_NONE;
@@ -7380,6 +7395,9 @@ hns3_parse_devargs(struct rte_eth_dev *dev)
                           &hns3_parse_io_hint_func, &tx_func_hint);
        (void)rte_kvargs_process(kvlist, HNS3_DEVARG_DEV_CAPS_MASK,
                           &hns3_parse_dev_caps_mask, &dev_caps_mask);
+       (void)rte_kvargs_process(kvlist, HNS3_DEVARG_MBX_TIME_LIMIT_MS,
+                          &hns3_parse_mbx_time_limit, &mbx_time_limit_ms);
+
        rte_kvargs_free(kvlist);

if (rx_func_hint != HNS3_IO_FUNC_HINT_NONE)

@@ -7395,6 +7413,11 @@ hns3_parse_devargs(struct rte_eth_dev *dev)
                hns3_warn(hw, "parsed %s = 0x%" PRIx64 ".",
                          HNS3_DEVARG_DEV_CAPS_MASK, dev_caps_mask);
        hns->dev_caps_mask = dev_caps_mask;
+
+       if (mbx_time_limit_ms != HNS3_MBX_DEF_TIME_LIMIT_MS)
+               hns3_warn(hw, "parsed %s = %u.", HNS3_DEVARG_MBX_TIME_LIMIT_MS,
+                               mbx_time_limit_ms);
+       hns->mbx_time_limit_ms = mbx_time_limit_ms;
  }

static const struct eth_dev_ops hns3_eth_dev_ops = {

@@ -7651,6 +7674,7 @@ RTE_PMD_REGISTER_KMOD_DEP(net_hns3, "* igb_uio | 
vfio-pci");
  RTE_PMD_REGISTER_PARAM_STRING(net_hns3,
                HNS3_DEVARG_RX_FUNC_HINT "=vec|sve|simple|common "
                HNS3_DEVARG_TX_FUNC_HINT "=vec|sve|simple|common "
-               HNS3_DEVARG_DEV_CAPS_MASK "=<1-65535> ");
+               HNS3_DEVARG_DEV_CAPS_MASK "=<1-65535> "
+               HNS3_DEVARG_MBX_TIME_LIMIT_MS "=<uint16> ");
  RTE_LOG_REGISTER_SUFFIX(hns3_logtype_init, init, NOTICE);
  RTE_LOG_REGISTER_SUFFIX(hns3_logtype_driver, driver, NOTICE);
diff --git a/drivers/net/hns3/hns3_ethdev.h b/drivers/net/hns3/hns3_ethdev.h
index 0e4e426..6476ad5 100644
--- a/drivers/net/hns3/hns3_ethdev.h
+++ b/drivers/net/hns3/hns3_ethdev.h
@@ -852,6 +852,7 @@ struct hns3_adapter {
        uint32_t tx_func_hint;

uint64_t dev_caps_mask;

+       uint16_t mbx_time_limit_ms; /* wait time for mbx message */

struct hns3_ptype_table ptype_tbl __rte_cache_aligned;

  };
@@ -869,6 +870,8 @@ enum {

#define HNS3_DEVARG_DEV_CAPS_MASK "dev_caps_mask" +#define HNS3_DEVARG_MBX_TIME_LIMIT_MS "mbx_time_limit_ms"

+
  enum {
        HNS3_DEV_SUPPORT_DCB_B,
        HNS3_DEV_SUPPORT_COPPER_B,
diff --git a/drivers/net/hns3/hns3_ethdev_vf.c 
b/drivers/net/hns3/hns3_ethdev_vf.c
index 8d9b797..7ed9cb2 100644
--- a/drivers/net/hns3/hns3_ethdev_vf.c
+++ b/drivers/net/hns3/hns3_ethdev_vf.c
@@ -3115,4 +3115,5 @@ RTE_PMD_REGISTER_KMOD_DEP(net_hns3_vf, "* igb_uio | 
vfio-pci");
  RTE_PMD_REGISTER_PARAM_STRING(net_hns3_vf,
                HNS3_DEVARG_RX_FUNC_HINT "=vec|sve|simple|common "
                HNS3_DEVARG_TX_FUNC_HINT "=vec|sve|simple|common "
-               HNS3_DEVARG_DEV_CAPS_MASK "=<1-65535> ");
+               HNS3_DEVARG_DEV_CAPS_MASK "=<1-65535> "
+               HNS3_DEVARG_MBX_TIME_LIMIT_MS "=<uint16_t> ");
diff --git a/drivers/net/hns3/hns3_mbx.c b/drivers/net/hns3/hns3_mbx.c
index f36cb10..a1556f7 100644
--- a/drivers/net/hns3/hns3_mbx.c
+++ b/drivers/net/hns3/hns3_mbx.c
@@ -61,8 +61,9 @@ static int
  hns3_get_mbx_resp(struct hns3_hw *hw, uint16_t code, uint16_t subcode,
                  uint8_t *resp_data, uint16_t resp_len)
  {
-#define HNS3_MAX_RETRY_US      500000
  #define HNS3_WAIT_RESP_US     100
+#define US_PER_MS              1000
+       uint32_t mbx_time_limit = HNS3_MBX_DEF_TIME_LIMIT_MS * US_PER_MS;
        struct hns3_adapter *hns = HNS3_DEV_HW_TO_ADAPTER(hw);
        struct hns3_mbx_resp_status *mbx_resp;
        uint32_t wait_time = 0;
@@ -74,7 +75,10 @@ hns3_get_mbx_resp(struct hns3_hw *hw, uint16_t code, 
uint16_t subcode,
                return -EINVAL;
        }

- while (wait_time < HNS3_MAX_RETRY_US) {

+       if (hns->mbx_time_limit_ms > HNS3_MBX_DEF_TIME_LIMIT_MS)
+               mbx_time_limit = (uint32_t)hns->mbx_time_limit_ms * US_PER_MS;
+

Why calculating the 'mbx_time_limit' twice in this function, isn't
'hns->mbx_time_limit_ms' has already either default value or user provided
value, why not just keep this calculation even with removing above check?

Fixed in v2, please check it out.

+       while (wait_time < mbx_time_limit) {
                if (__atomic_load_n(&hw->reset.disable_cmd, __ATOMIC_RELAXED)) {
                        hns3_err(hw, "Don't wait for mbx respone because of "
                                 "disable_cmd");
@@ -103,7 +107,7 @@ hns3_get_mbx_resp(struct hns3_hw *hw, uint16_t code, 
uint16_t subcode,
                wait_time += HNS3_WAIT_RESP_US;
        }
        hw->mbx_resp.req_msg_data = 0;
-       if (wait_time >= HNS3_MAX_RETRY_US) {
+       if (wait_time >= mbx_time_limit) {
                hns3_mbx_proc_timeout(hw, code, subcode);
                return -ETIME;
        }
diff --git a/drivers/net/hns3/hns3_mbx.h b/drivers/net/hns3/hns3_mbx.h
index f868e33..d637bd2 100644
--- a/drivers/net/hns3/hns3_mbx.h
+++ b/drivers/net/hns3/hns3_mbx.h
@@ -87,6 +87,7 @@ enum hns3_mbx_link_fail_subcode {

#define HNS3_MBX_MAX_MSG_SIZE 16

  #define HNS3_MBX_MAX_RESP_DATA_SIZE   8
+#define HNS3_MBX_DEF_TIME_LIMIT_MS     500

enum {

        HNS3_MBX_RESP_MATCHING_SCHEME_OF_ORIGINAL = 0,

.