[v2] ethdev: add IPv4 checksum RSS offload type
Checks
Commit Message
This patch defines new RSS offload type for IPv4 checksum,
which is required when users want to distribute packets based
on the IPv4 checksum field.
For example "flow create 0 ingress pattern eth / ipv4 / end
actions rss types ipv4-chksum end queues end / end", this flow
causes all matching packets to be distributed to queues on
basis of IPv4 checksum.
Signed-off-by: Alvin Zhang <alvinx.zhang@intel.com>
---
app/test-pmd/cmdline.c | 2 ++
app/test-pmd/config.c | 1 +
lib/ethdev/rte_ethdev.h | 1 +
3 files changed, 4 insertions(+)
Comments
On 6/3/21 11:03 AM, Alvin Zhang wrote:
> This patch defines new RSS offload type for IPv4 checksum,
> which is required when users want to distribute packets based
> on the IPv4 checksum field.
>
> For example "flow create 0 ingress pattern eth / ipv4 / end
> actions rss types ipv4-chksum end queues end / end", this flow
> causes all matching packets to be distributed to queues on
> basis of IPv4 checksum.
>
> Signed-off-by: Alvin Zhang <alvinx.zhang@intel.com>
The patch LGTM now. I hope we'll have a PMD in the release
cycle which supports it.
In fact it is a funny idea. Let's consider a TCP connection
and routing change with one extra hop on the way while
the connection is alive. Or a UDP stream in the same
conditions. If so, TTL will different, checksum changes and
connection packets may be delivered to a different Rx queue.
May be sometimes it is not important and/or can't happen.
L4 checksum sounds better from this point of view since
it covers IP addresses and transport ports and less bits
to hash. The only concern is distribution fairness, but
I keep the topic to those who understand corresponding
math better.
> ---
> app/test-pmd/cmdline.c | 2 ++
> app/test-pmd/config.c | 1 +
> lib/ethdev/rte_ethdev.h | 1 +
> 3 files changed, 4 insertions(+)
>
> diff --git a/app/test-pmd/cmdline.c b/app/test-pmd/cmdline.c
> index 0268b18..32da066 100644
> --- a/app/test-pmd/cmdline.c
> +++ b/app/test-pmd/cmdline.c
> @@ -2254,6 +2254,8 @@ struct cmd_config_rss {
> rss_conf.rss_hf = ETH_RSS_ECPRI;
> else if (!strcmp(res->value, "mpls"))
> rss_conf.rss_hf = ETH_RSS_MPLS;
> + else if (!strcmp(res->value, "ipv4-chksum"))
> + rss_conf.rss_hf = ETH_RSS_IPV4_CHKSUM;
> else if (!strcmp(res->value, "none"))
> rss_conf.rss_hf = 0;
> else if (!strcmp(res->value, "level-default")) {
> diff --git a/app/test-pmd/config.c b/app/test-pmd/config.c
> index 43c79b5..2c0c415 100644
> --- a/app/test-pmd/config.c
> +++ b/app/test-pmd/config.c
> @@ -140,6 +140,7 @@
> { "gtpu", ETH_RSS_GTPU },
> { "ecpri", ETH_RSS_ECPRI },
> { "mpls", ETH_RSS_MPLS },
> + { "ipv4-chksum", ETH_RSS_IPV4_CHKSUM },
> { NULL, 0 },
> };
>
> diff --git a/lib/ethdev/rte_ethdev.h b/lib/ethdev/rte_ethdev.h
> index faf3bd9..f10d834 100644
> --- a/lib/ethdev/rte_ethdev.h
> +++ b/lib/ethdev/rte_ethdev.h
> @@ -537,6 +537,7 @@ struct rte_eth_rss_conf {
> #define ETH_RSS_PPPOE (1ULL << 31)
> #define ETH_RSS_ECPRI (1ULL << 32)
> #define ETH_RSS_MPLS (1ULL << 33)
> +#define ETH_RSS_IPV4_CHKSUM (1ULL << 34)
>
> /*
> * We use the following macros to combine with above ETH_RSS_* for
>
> > This patch defines new RSS offload type for IPv4 checksum, which is
> > required when users want to distribute packets based on the IPv4
> > checksum field.
> >
> > For example "flow create 0 ingress pattern eth / ipv4 / end actions
> > rss types ipv4-chksum end queues end / end", this flow causes all
> > matching packets to be distributed to queues on basis of IPv4
> > checksum.
> >
> > Signed-off-by: Alvin Zhang <alvinx.zhang@intel.com>
>
> The patch LGTM now. I hope we'll have a PMD in the release cycle which
> supports it.
>
> In fact it is a funny idea. Let's consider a TCP connection and routing change
> with one extra hop on the way while the connection is alive. Or a UDP stream in
> the same conditions. If so, TTL will different, checksum changes and connection
> packets may be delivered to a different Rx queue.
> May be sometimes it is not important and/or can't happen.
>
> L4 checksum sounds better from this point of view since it covers IP addresses
> and transport ports and less bits to hash. The only concern is distribution fairness,
> but I keep the topic to those who understand corresponding math better.
>
> > ---
We are preparing a PMD patch to support this feature, if ready, we will submit it ASAP.
Thanks,
Alvin
On Thu, Jun 3, 2021 at 1:04 AM Alvin Zhang <alvinx.zhang@intel.com> wrote:
>
> This patch defines new RSS offload type for IPv4 checksum,
> which is required when users want to distribute packets based
> on the IPv4 checksum field.
>
> For example "flow create 0 ingress pattern eth / ipv4 / end
> actions rss types ipv4-chksum end queues end / end", this flow
> causes all matching packets to be distributed to queues on
> basis of IPv4 checksum.
>
> Signed-off-by: Alvin Zhang <alvinx.zhang@intel.com>
Acked-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
> ---
> app/test-pmd/cmdline.c | 2 ++
> app/test-pmd/config.c | 1 +
> lib/ethdev/rte_ethdev.h | 1 +
> 3 files changed, 4 insertions(+)
>
> diff --git a/app/test-pmd/cmdline.c b/app/test-pmd/cmdline.c
> index 0268b18..32da066 100644
> --- a/app/test-pmd/cmdline.c
> +++ b/app/test-pmd/cmdline.c
> @@ -2254,6 +2254,8 @@ struct cmd_config_rss {
> rss_conf.rss_hf = ETH_RSS_ECPRI;
> else if (!strcmp(res->value, "mpls"))
> rss_conf.rss_hf = ETH_RSS_MPLS;
> + else if (!strcmp(res->value, "ipv4-chksum"))
> + rss_conf.rss_hf = ETH_RSS_IPV4_CHKSUM;
> else if (!strcmp(res->value, "none"))
> rss_conf.rss_hf = 0;
> else if (!strcmp(res->value, "level-default")) {
> diff --git a/app/test-pmd/config.c b/app/test-pmd/config.c
> index 43c79b5..2c0c415 100644
> --- a/app/test-pmd/config.c
> +++ b/app/test-pmd/config.c
> @@ -140,6 +140,7 @@
> { "gtpu", ETH_RSS_GTPU },
> { "ecpri", ETH_RSS_ECPRI },
> { "mpls", ETH_RSS_MPLS },
> + { "ipv4-chksum", ETH_RSS_IPV4_CHKSUM },
> { NULL, 0 },
> };
>
> diff --git a/lib/ethdev/rte_ethdev.h b/lib/ethdev/rte_ethdev.h
> index faf3bd9..f10d834 100644
> --- a/lib/ethdev/rte_ethdev.h
> +++ b/lib/ethdev/rte_ethdev.h
> @@ -537,6 +537,7 @@ struct rte_eth_rss_conf {
> #define ETH_RSS_PPPOE (1ULL << 31)
> #define ETH_RSS_ECPRI (1ULL << 32)
> #define ETH_RSS_MPLS (1ULL << 33)
> +#define ETH_RSS_IPV4_CHKSUM (1ULL << 34)
>
> /*
> * We use the following macros to combine with above ETH_RSS_* for
> --
> 1.8.3.1
>
@@ -2254,6 +2254,8 @@ struct cmd_config_rss {
rss_conf.rss_hf = ETH_RSS_ECPRI;
else if (!strcmp(res->value, "mpls"))
rss_conf.rss_hf = ETH_RSS_MPLS;
+ else if (!strcmp(res->value, "ipv4-chksum"))
+ rss_conf.rss_hf = ETH_RSS_IPV4_CHKSUM;
else if (!strcmp(res->value, "none"))
rss_conf.rss_hf = 0;
else if (!strcmp(res->value, "level-default")) {
@@ -140,6 +140,7 @@
{ "gtpu", ETH_RSS_GTPU },
{ "ecpri", ETH_RSS_ECPRI },
{ "mpls", ETH_RSS_MPLS },
+ { "ipv4-chksum", ETH_RSS_IPV4_CHKSUM },
{ NULL, 0 },
};
@@ -537,6 +537,7 @@ struct rte_eth_rss_conf {
#define ETH_RSS_PPPOE (1ULL << 31)
#define ETH_RSS_ECPRI (1ULL << 32)
#define ETH_RSS_MPLS (1ULL << 33)
+#define ETH_RSS_IPV4_CHKSUM (1ULL << 34)
/*
* We use the following macros to combine with above ETH_RSS_* for