On 6/9/2022 4:25 AM, lihuisong (C) wrote:
CAUTION: This message has originated from an External Source. Please use proper judgment and caution when opening attachments, clicking links, or responding to this email.在 2022/6/7 23:45, Ferruh Yigit 写道:On 6/7/2022 9:32 AM, Huisong Li wrote:And rte_eth_dev_info.flow_type_rss_offloads is populated in terms of RTE_ETH_RSS_* bits. If PMD sets RTE_ETH_RSS_L3_SRC_ONLY to dev_info->flow_type_rss_offloads. testpmd will display "user defined 63" when run 'show port info 0'. Because testpmd use flowtype_to_str() to display the supported RSS offload of PMD. In fact, the function is used to display flow type in FDIR commands. This patch uses the RTE_ETH_RSS_* bits to display supported RSS offload of PMD. Fixes: b12964f621dc ("ethdev: unification of RSS offload types") Cc: sta...@dpdk.org Signed-off-by: Huisong Li <lihuis...@huawei.com> Signed-off-by: Ferruh Yigit <ferruh.yi...@xilinx.com> --- app/test-pmd/config.c | 85 ++++++++++++++++++++++++++++++++++++++----- 1 file changed, 75 insertions(+), 10 deletions(-) diff --git a/app/test-pmd/config.c b/app/test-pmd/config.c index 72d2606d19..d290ca3a06 100644 --- a/app/test-pmd/config.c +++ b/app/test-pmd/config.c @@ -86,6 +86,56 @@ static const struct { }, }; +const struct rss_type_info rss_offload_table[] = { + {"ipv4", RTE_ETH_RSS_IPV4}, + {"ipv4-frag", RTE_ETH_RSS_FRAG_IPV4}, + {"ipv4-tcp", RTE_ETH_RSS_NONFRAG_IPV4_TCP}, + {"ipv4-udp", RTE_ETH_RSS_NONFRAG_IPV4_UDP}, + {"ipv4-sctp", RTE_ETH_RSS_NONFRAG_IPV4_SCTP}, + {"ipv4-other", RTE_ETH_RSS_NONFRAG_IPV4_OTHER}, + {"ipv6", RTE_ETH_RSS_IPV6}, + {"ipv6-frag", RTE_ETH_RSS_FRAG_IPV6}, + {"ipv6-tcp", RTE_ETH_RSS_NONFRAG_IPV6_TCP}, + {"ipv6-udp", RTE_ETH_RSS_NONFRAG_IPV6_UDP}, + {"ipv6-sctp", RTE_ETH_RSS_NONFRAG_IPV6_SCTP}, + {"ipv6-other", RTE_ETH_RSS_NONFRAG_IPV6_OTHER}, + {"l2_payload", RTE_ETH_RSS_L2_PAYLOAD}, + {"ipv6-ex", RTE_ETH_RSS_IPV6_EX}, + {"ipv6-tcp-ex", RTE_ETH_RSS_IPV6_TCP_EX}, + {"ipv6-udp-ex", RTE_ETH_RSS_IPV6_UDP_EX}, + {"port", RTE_ETH_RSS_PORT}, + {"vxlan", RTE_ETH_RSS_VXLAN}, + {"geneve", RTE_ETH_RSS_GENEVE}, + {"nvgre", RTE_ETH_RSS_NVGRE}, + {"gtpu", RTE_ETH_RSS_GTPU}, + {"eth", RTE_ETH_RSS_ETH}, + {"s-vlan", RTE_ETH_RSS_S_VLAN}, + {"c-vlan", RTE_ETH_RSS_C_VLAN}, + {"esp", RTE_ETH_RSS_ESP}, + {"ah", RTE_ETH_RSS_AH}, + {"l2tpv3", RTE_ETH_RSS_L2TPV3}, + {"pfcp", RTE_ETH_RSS_PFCP}, + {"pppoe", RTE_ETH_RSS_PPPOE}, + {"ecpri", RTE_ETH_RSS_ECPRI}, + {"mpls", RTE_ETH_RSS_MPLS}, + {"ipv4-chksum", RTE_ETH_RSS_IPV4_CHKSUM}, + {"l4-chksum", RTE_ETH_RSS_L4_CHKSUM}, + {"l2tpv2", RTE_ETH_RSS_L2TPV2}, + {"l3-pre96", RTE_ETH_RSS_L3_PRE96}, + {"l3-pre64", RTE_ETH_RSS_L3_PRE64}, + {"l3-pre56", RTE_ETH_RSS_L3_PRE56}, + {"l3-pre48", RTE_ETH_RSS_L3_PRE48}, + {"l3-pre40", RTE_ETH_RSS_L3_PRE40}, + {"l3-pre32", RTE_ETH_RSS_L3_PRE32}, + {"l2-dst-only", RTE_ETH_RSS_L2_DST_ONLY}, + {"l2-src-only", RTE_ETH_RSS_L2_SRC_ONLY}, + {"l4-dst-only", RTE_ETH_RSS_L4_DST_ONLY}, + {"l4-src-only", RTE_ETH_RSS_L4_SRC_ONLY}, + {"l3-dst-only", RTE_ETH_RSS_L3_DST_ONLY}, + {"l3-src-only", RTE_ETH_RSS_L3_SRC_ONLY}, + {NULL, 0}, +}; +Hi Huisong, Why not reusing existing 'rss_type_table[]', but adding a new one? Is it to have each individual RSS type instead of grouping 'rss_type_table[]' has? If so, since command "port config all rss ..." using the grouping, I think it makes sense to have grouping.The 'rss_offload_table[]' includes all defined RSS offloads in ethdev layer, every iterm has one bit, and is part of 'rss_type_table[]'. Some iterms in 'rss_type_table[]' consist of multiple bits. If we want to display all RSS offloads PMD supported and resolve undefined offload, we have to check 'flow_type_rss_offloads' bit by bit. However, "port config all rss ..." and "show port 0 rss-hash" commands supports the configuration and display of multiple offload bits at a time. So it is necessary to introduce this 'rss_offload_table[]' to display 'flow_type_rss_offloads'. That's what I think.
'rss_type_table[]' has both group and individual types, so show functions displays both.
A sample of grouping is UDP, a few UDP flags compined under UDP,
#define RTE_ETH_RSS_UDP ( \
RTE_ETH_RSS_NONFRAG_IPV4_UDP | \
RTE_ETH_RSS_NONFRAG_IPV6_UDP | \
RTE_ETH_RSS_IPV6_UDP_EX)
Above mentioned commands use 'RTE_ETH_RSS_UDP' to set "udp" type RSS,
and when displying instead of displaying each above type separately, I
think displaying as "udp" makes sense.
And again I believe having two different arrays is prone to error for future.
Another thing is having two different array with very similar content is easy to confuse and can cause diversion on arrays and generate bugs.Add some notes about the use of the 'rss_offload_table[]'?As mentioned from "port config all rss ..." command, it seems it is also not using 'rss_type_table[]', but all string to RSS type matching done within the function ('cmd_config_rss_parsed()') duplicating what we have in 'rss_type_table[]'. Again this has concern to diverge between set and show functions. If you have time for it, can you make an additional patch to update 'cmd_config_rss_parsed()' to use new 'str_to_rsstypes()' function?All right. Let's fix it.
Thanks, and if you want it can be a separate patch on top of this one, it doesn't have to be part of this set if you don't have time for it.
Thanks, ferruhconst struct rss_type_info rss_type_table[] = { { "all", RTE_ETH_RSS_ETH | RTE_ETH_RSS_VLAN | RTE_ETH_RSS_IP | RTE_ETH_RSS_TCP | RTE_ETH_RSS_UDP | RTE_ETH_RSS_SCTP | RTE_ETH_RSS_L2_PAYLOAD | @@ -675,6 +725,19 @@ print_dev_capabilities(uint64_t capabilities) } } +static const char * +rss_offload_to_str(uint64_t rss_offload) +{ + uint16_t i; + + for (i = 0; rss_offload_table[i].str != NULL; i++) { + if (rss_offload_table[i].rss_type == rss_offload) + return rss_offload_table[i].str; + } + + return NULL; +} + void port_infos_display(portid_t port_id) { @@ -779,19 +842,21 @@ port_infos_display(portid_t port_id) if (!dev_info.flow_type_rss_offloads) printf("No RSS offload flow type is supported.\n"); else { + uint64_t rss_offload_types = dev_info.flow_type_rss_offloads; uint16_t i; - char *p; printf("Supported RSS offload flow types:\n"); - for (i = RTE_ETH_FLOW_UNKNOWN + 1; - i < sizeof(dev_info.flow_type_rss_offloads) * CHAR_BIT; i++) { - if (!(dev_info.flow_type_rss_offloads & (1ULL << i))) - continue; - p = flowtype_to_str(i); - if (p) - printf(" %s\n", p); - else - printf(" user defined %d\n", i); + for (i = 0; i < sizeof(rss_offload_types) * CHAR_BIT; i++) { + uint64_t rss_offload = RTE_BIT64(i); + if ((rss_offload_types & rss_offload) != 0) { + const char *p = + rss_offload_to_str(rss_offload); + if (p) + printf(" %s\n", p); + else + printf(" user defined 0x%"PRIx64"\n", + rss_offload); + }If you go with 'rss_type_table[]', you need to change above logic as you did in your v2.} } -- 2.33.0.
