| Message ID | 168003455815.3027256.7575362149566382055.stgit@firesoul (mailing list archive) |
|---|---|
| State | Superseded |
| Delegated to: | BPF |
| Headers | show |
| Series | XDP-hints: API change for RX-hash kfunc bpf_xdp_metadata_rx_hash | expand |
| Context | Check | Description |
|---|---|---|
| netdev/tree_selection | success | Clearly marked for bpf |
| netdev/apply | fail | Patch does not apply to bpf |
| bpf/vmtest-bpf-PR | fail | merge-conflict |
On 03/28, Jesper Dangaard Brouer wrote: > The RSS hash type specifies what portion of packet data NIC hardware used > when calculating RSS hash value. The RSS types are focused on Internet > traffic protocols at OSI layers L3 and L4. L2 (e.g. ARP) often get hash > value zero and no RSS type. For L3 focused on IPv4 vs. IPv6, and L4 > primarily TCP vs UDP, but some hardware supports SCTP. > Hardware RSS types are differently encoded for each hardware NIC. Most > hardware represent RSS hash type as a number. Determining L3 vs L4 often > requires a mapping table as there often isn't a pattern or sorting > according to ISO layer. > The patch introduce a XDP RSS hash type (xdp_rss_hash_type) that can both > be seen as a number that is ordered according by ISO layer, and can be bit > masked to separate IPv4 and IPv6 types for L4 protocols. Room is available > for extending later while keeping these properties. This maps and unifies > difference to hardware specific hashes. Looks good overall. Any reason we're making this specific layout? Why not simply the following? enum { XDP_RSS_TYPE_NONE = 0, XDP_RSS_TYPE_IPV4 = BIT(0), XDP_RSS_TYPE_IPV6 = BIT(1), /* IPv6 with extension header. */ /* let's note ^^^ it in the UAPI? */ XDP_RSS_TYPE_IPV6_EX = BIT(2), XDP_RSS_TYPE_UDP = BIT(3), XDP_RSS_TYPE_TCP = BIT(4), XDP_RSS_TYPE_SCTP = BIT(5), } And then using XDP_RSS_TYPE_IPV4|XDP_RSS_TYPE_UDP vs XDP_RSS_TYPE_IPV6|XXX ? > This proposal change the kfunc API bpf_xdp_metadata_rx_hash() to return > this RSS hash type on success. > Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com> > --- > include/net/xdp.h | 51 > +++++++++++++++++++++++++++++++++++++++++++++++++++ > net/core/xdp.c | 4 +++- > 2 files changed, 54 insertions(+), 1 deletion(-) > diff --git a/include/net/xdp.h b/include/net/xdp.h > index 5393b3ebe56e..63f462f5ea7f 100644 > --- a/include/net/xdp.h > +++ b/include/net/xdp.h > @@ -8,6 +8,7 @@ > #include <linux/skbuff.h> /* skb_shared_info */ > #include <uapi/linux/netdev.h> > +#include <linux/bitfield.h> > /** > * DOC: XDP RX-queue information > @@ -396,6 +397,56 @@ XDP_METADATA_KFUNC_xxx > MAX_XDP_METADATA_KFUNC, > }; > +/* For partitioning of xdp_rss_hash_type */ > +#define RSS_L3 GENMASK(2,0) /* 3-bits = values between 1-7 */ > +#define L4_BIT BIT(3) /* 1-bit - L4 indication */ > +#define RSS_L4_IPV4 GENMASK(6,4) /* 3-bits */ > +#define RSS_L4_IPV6 GENMASK(9,7) /* 3-bits */ > +#define RSS_L4 GENMASK(9,3) /* = 7-bits - covering L4 IPV4+IPV6 */ > +#define L4_IPV6_EX_BIT BIT(9) /* 1-bit - L4 IPv6 with Extension > hdr */ > + /* 11-bits in total */ > + > +/* The XDP RSS hash type (xdp_rss_hash_type) can both be seen as a > number that > + * is ordered according by ISO layer, and can be bit masked to separate > IPv4 and > + * IPv6 types for L4 protocols. Room is available for extending later > while > + * keeping above properties, as this need to cover NIC hardware RSS > types. > + */ > +enum xdp_rss_hash_type { > + XDP_RSS_TYPE_NONE = 0, > + XDP_RSS_TYPE_L2 = XDP_RSS_TYPE_NONE, > + > + XDP_RSS_TYPE_L3_MASK = RSS_L3, > + XDP_RSS_TYPE_L3_IPV4 = FIELD_PREP_CONST(RSS_L3, 1), > + XDP_RSS_TYPE_L3_IPV6 = FIELD_PREP_CONST(RSS_L3, 2), > + XDP_RSS_TYPE_L3_IPV6_EX = FIELD_PREP_CONST(RSS_L3, 4), > + > + XDP_RSS_TYPE_L4_MASK = RSS_L4, > + XDP_RSS_TYPE_L4_SHIFT = __bf_shf(RSS_L4), > + XDP_RSS_TYPE_L4_MASK_EX = RSS_L4 | L4_IPV6_EX_BIT, > + > + XDP_RSS_TYPE_L4_IPV4_MASK = RSS_L4_IPV4, > + XDP_RSS_TYPE_L4_BIT = L4_BIT, > + XDP_RSS_TYPE_L4_IPV4_TCP = L4_BIT|FIELD_PREP_CONST(RSS_L4_IPV4, 1), > + XDP_RSS_TYPE_L4_IPV4_UDP = L4_BIT|FIELD_PREP_CONST(RSS_L4_IPV4, 2), > + XDP_RSS_TYPE_L4_IPV4_SCTP = L4_BIT|FIELD_PREP_CONST(RSS_L4_IPV4, 3), > + > + XDP_RSS_TYPE_L4_IPV6_MASK = RSS_L4_IPV6, > + XDP_RSS_TYPE_L4_IPV6_TCP = L4_BIT|FIELD_PREP_CONST(RSS_L4_IPV6, 1), > + XDP_RSS_TYPE_L4_IPV6_UDP = L4_BIT|FIELD_PREP_CONST(RSS_L4_IPV6, 2), > + XDP_RSS_TYPE_L4_IPV6_SCTP = L4_BIT|FIELD_PREP_CONST(RSS_L4_IPV6, 3), > + > + XDP_RSS_TYPE_L4_IPV6_EX_MASK = L4_IPV6_EX_BIT, > + XDP_RSS_TYPE_L4_IPV6_TCP_EX = XDP_RSS_TYPE_L4_IPV6_TCP |L4_IPV6_EX_BIT, > + XDP_RSS_TYPE_L4_IPV6_UDP_EX = XDP_RSS_TYPE_L4_IPV6_UDP |L4_IPV6_EX_BIT, > + XDP_RSS_TYPE_L4_IPV6_SCTP_EX = XDP_RSS_TYPE_L4_IPV6_SCTP|L4_IPV6_EX_BIT, > +}; > +#undef RSS_L3 > +#undef L4_BIT > +#undef RSS_L4_IPV4 > +#undef RSS_L4_IPV6 > +#undef RSS_L4 > +#undef L4_IPV6_EX_BIT > + > #ifdef CONFIG_NET > u32 bpf_xdp_metadata_kfunc_id(int id); > bool bpf_dev_bound_kfunc_id(u32 btf_id); > diff --git a/net/core/xdp.c b/net/core/xdp.c > index 7133017bcd74..81d41df30695 100644 > --- a/net/core/xdp.c > +++ b/net/core/xdp.c > @@ -721,12 +721,14 @@ __bpf_kfunc int bpf_xdp_metadata_rx_timestamp(const > struct xdp_md *ctx, u64 *tim > * @hash: Return value pointer. > * > * Return: > - * * Returns 0 on success or ``-errno`` on error. > + * * Returns (positive) RSS hash **type** on success or ``-errno`` on > error. > + * * ``enum xdp_rss_hash_type`` : RSS hash type > * * ``-EOPNOTSUPP`` : means device driver doesn't implement kfunc > * * ``-ENODATA`` : means no RX-hash available for this frame > */ > __bpf_kfunc int bpf_xdp_metadata_rx_hash(const struct xdp_md *ctx, u32 > *hash) > { > + BTF_TYPE_EMIT(enum xdp_rss_hash_type); > return -EOPNOTSUPP; > }
On 28/03/2023 21:15, Jesper Dangaard Brouer wrote: > Hardware RSS types are differently encoded for each hardware NIC. Most > hardware represent RSS hash type as a number. Determining L3 vs L4 often > requires a mapping table as there often isn't a pattern or sorting > according to ISO layer. > > The patch introduce a XDP RSS hash type (xdp_rss_hash_type) that can both > be seen as a number that is ordered according by ISO layer, and can be bit > masked to separate IPv4 and IPv6 types for L4 protocols. Room is available > for extending later while keeping these properties. This maps and unifies > difference to hardware specific hashes. Would it be better to make use of the ETHTOOL_GRXFH defines (stuff like UDP_V6_FLOW, RXH_L4_B_0_1 etc.)? Seems like that could allow for some code reuse in drivers.
On 28/03/2023 23.58, Stanislav Fomichev wrote: > On 03/28, Jesper Dangaard Brouer wrote: >> The RSS hash type specifies what portion of packet data NIC hardware used >> when calculating RSS hash value. The RSS types are focused on Internet >> traffic protocols at OSI layers L3 and L4. L2 (e.g. ARP) often get hash >> value zero and no RSS type. For L3 focused on IPv4 vs. IPv6, and L4 >> primarily TCP vs UDP, but some hardware supports SCTP. > >> Hardware RSS types are differently encoded for each hardware NIC. Most >> hardware represent RSS hash type as a number. Determining L3 vs L4 often >> requires a mapping table as there often isn't a pattern or sorting >> according to ISO layer. > >> The patch introduce a XDP RSS hash type (xdp_rss_hash_type) that can both >> be seen as a number that is ordered according by ISO layer, and can be bit >> masked to separate IPv4 and IPv6 types for L4 protocols. Room is available >> for extending later while keeping these properties. This maps and unifies >> difference to hardware specific hashes. > > Looks good overall. Any reason we're making this specific layout? One important goal is to have a simple/fast way to determining L3 vs L4, because a L4 hash can be used for flow handling (e.g. load-balancing). We below layout you can: if (rss_type & XDP_RSS_TYPE_L4_MASK) bool hw_hash_do_LB = true; Or using it as a number: if (rss_type > XDP_RSS_TYPE_L4) bool hw_hash_do_LB = true; I'm very open to changes to my "specific" layout. I am in doubt if using it as a number is the right approach and worth the trouble. > Why not simply the following? > > enum { > XDP_RSS_TYPE_NONE = 0, > XDP_RSS_TYPE_IPV4 = BIT(0), > XDP_RSS_TYPE_IPV6 = BIT(1), > /* IPv6 with extension header. */ > /* let's note ^^^ it in the UAPI? */ > XDP_RSS_TYPE_IPV6_EX = BIT(2), > XDP_RSS_TYPE_UDP = BIT(3), > XDP_RSS_TYPE_TCP = BIT(4), > XDP_RSS_TYPE_SCTP = BIT(5), We know these bits for UDP, TCP, SCTP (and IPSEC) are exclusive, they cannot be set at the same time, e.g. as a packet cannot both be UDP and TCP. Thus, using these bits as a number make sense to me, and is more compact. This BIT() approach also have the issue of extending it later (forward compatibility). As mentioned a common task will be to check if hash-type is a L4 type. See mlx5 [patch 4/4] needed to extend with IPSEC. Notice how my XDP_RSS_TYPE_L4_MASK covers all the bits that this can be extended with new L4 types, such that existing progs will still work checking for L4 check. It can of-cause be solved in the same way for this BIT() approach by reserving some bits upfront in a mask. > } > > And then using XDP_RSS_TYPE_IPV4|XDP_RSS_TYPE_UDP vs > XDP_RSS_TYPE_IPV6|XXX ? Do notice, that I already does some level of or'ing ("|") in this proposal. The main difference is that I hide this from the driver, and kind of pre-combine the valid combination (enum's) drivers can select from. I do get the point, and I think I will come up with a combined solution based on your input. The RSS hashing types and combinations comes from M$ standards: [1] https://learn.microsoft.com/en-us/windows-hardware/drivers/network/rss-hashing-types#ipv4-hash-type-combinations >> This proposal change the kfunc API bpf_xdp_metadata_rx_hash() to return >> this RSS hash type on success. > >> Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com> >> --- >> include/net/xdp.h | 51 >> +++++++++++++++++++++++++++++++++++++++++++++++++++ >> net/core/xdp.c | 4 +++- >> 2 files changed, 54 insertions(+), 1 deletion(-) > >> diff --git a/include/net/xdp.h b/include/net/xdp.h >> index 5393b3ebe56e..63f462f5ea7f 100644 >> --- a/include/net/xdp.h >> +++ b/include/net/xdp.h >> @@ -8,6 +8,7 @@ > >> #include <linux/skbuff.h> /* skb_shared_info */ >> #include <uapi/linux/netdev.h> >> +#include <linux/bitfield.h> > >> /** >> * DOC: XDP RX-queue information >> @@ -396,6 +397,56 @@ XDP_METADATA_KFUNC_xxx >> MAX_XDP_METADATA_KFUNC, >> }; > >> +/* For partitioning of xdp_rss_hash_type */ >> +#define RSS_L3 GENMASK(2,0) /* 3-bits = values between 1-7 */ >> +#define L4_BIT BIT(3) /* 1-bit - L4 indication */ >> +#define RSS_L4_IPV4 GENMASK(6,4) /* 3-bits */ >> +#define RSS_L4_IPV6 GENMASK(9,7) /* 3-bits */ >> +#define RSS_L4 GENMASK(9,3) /* = 7-bits - covering L4 >> IPV4+IPV6 */ >> +#define L4_IPV6_EX_BIT BIT(9) /* 1-bit - L4 IPv6 with >> Extension hdr */ >> + /* 11-bits in total */ >> + >> +/* The XDP RSS hash type (xdp_rss_hash_type) can both be seen as a number that >> + * is ordered according by ISO layer, and can be bit masked to separate IPv4 and >> + * IPv6 types for L4 protocols. Room is available for extending later while >> + * keeping above properties, as this need to cover NIC hardware RSS types. >> + */ >> +enum xdp_rss_hash_type { >> + XDP_RSS_TYPE_NONE = 0, >> + XDP_RSS_TYPE_L2 = XDP_RSS_TYPE_NONE, >> + >> + XDP_RSS_TYPE_L3_MASK = RSS_L3, >> + XDP_RSS_TYPE_L3_IPV4 = FIELD_PREP_CONST(RSS_L3, 1), >> + XDP_RSS_TYPE_L3_IPV6 = FIELD_PREP_CONST(RSS_L3, 2), >> + XDP_RSS_TYPE_L3_IPV6_EX = FIELD_PREP_CONST(RSS_L3, 4), >> + >> + XDP_RSS_TYPE_L4_MASK = RSS_L4, >> + XDP_RSS_TYPE_L4_SHIFT = __bf_shf(RSS_L4), >> + XDP_RSS_TYPE_L4_MASK_EX = RSS_L4 | L4_IPV6_EX_BIT, >> + >> + XDP_RSS_TYPE_L4_IPV4_MASK = RSS_L4_IPV4, >> + XDP_RSS_TYPE_L4_BIT = L4_BIT, >> + XDP_RSS_TYPE_L4_IPV4_TCP = L4_BIT|FIELD_PREP_CONST(RSS_L4_IPV4, 1), >> + XDP_RSS_TYPE_L4_IPV4_UDP = L4_BIT|FIELD_PREP_CONST(RSS_L4_IPV4, 2), >> + XDP_RSS_TYPE_L4_IPV4_SCTP = L4_BIT|FIELD_PREP_CONST(RSS_L4_IPV4, 3), >> + >> + XDP_RSS_TYPE_L4_IPV6_MASK = RSS_L4_IPV6, >> + XDP_RSS_TYPE_L4_IPV6_TCP = L4_BIT|FIELD_PREP_CONST(RSS_L4_IPV6, 1), >> + XDP_RSS_TYPE_L4_IPV6_UDP = L4_BIT|FIELD_PREP_CONST(RSS_L4_IPV6, 2), >> + XDP_RSS_TYPE_L4_IPV6_SCTP = L4_BIT|FIELD_PREP_CONST(RSS_L4_IPV6, 3), >> + >> + XDP_RSS_TYPE_L4_IPV6_EX_MASK = L4_IPV6_EX_BIT, >> + XDP_RSS_TYPE_L4_IPV6_TCP_EX = XDP_RSS_TYPE_L4_IPV6_TCP|L4_IPV6_EX_BIT, >> + XDP_RSS_TYPE_L4_IPV6_UDP_EX = XDP_RSS_TYPE_L4_IPV6_UDP|L4_IPV6_EX_BIT, >> + XDP_RSS_TYPE_L4_IPV6_SCTP_EX = XDP_RSS_TYPE_L4_IPV6_SCTP|L4_IPV6_EX_BIT, >> +}; >> +#undef RSS_L3 >> +#undef L4_BIT >> +#undef RSS_L4_IPV4 >> +#undef RSS_L4_IPV6 >> +#undef RSS_L4 >> +#undef L4_IPV6_EX_BIT >> + >> #ifdef CONFIG_NET >> u32 bpf_xdp_metadata_kfunc_id(int id); >> bool bpf_dev_bound_kfunc_id(u32 btf_id); >> diff --git a/net/core/xdp.c b/net/core/xdp.c >> index 7133017bcd74..81d41df30695 100644 >> --- a/net/core/xdp.c >> +++ b/net/core/xdp.c >> @@ -721,12 +721,14 @@ __bpf_kfunc int >> bpf_xdp_metadata_rx_timestamp(const struct xdp_md *ctx, u64 *tim >> * @hash: Return value pointer. >> * >> * Return: >> - * * Returns 0 on success or ``-errno`` on error. >> + * * Returns (positive) RSS hash **type** on success or ``-errno`` on >> error. >> + * * ``enum xdp_rss_hash_type`` : RSS hash type >> * * ``-EOPNOTSUPP`` : means device driver doesn't implement kfunc >> * * ``-ENODATA`` : means no RX-hash available for this frame >> */ >> __bpf_kfunc int bpf_xdp_metadata_rx_hash(const struct xdp_md *ctx, >> u32 *hash) >> { >> + BTF_TYPE_EMIT(enum xdp_rss_hash_type); >> return -EOPNOTSUPP; >> } > > >
On 29/03/2023 10.10, Edward Cree wrote: > On 28/03/2023 21:15, Jesper Dangaard Brouer wrote: >> Hardware RSS types are differently encoded for each hardware NIC. Most >> hardware represent RSS hash type as a number. Determining L3 vs L4 often >> requires a mapping table as there often isn't a pattern or sorting >> according to ISO layer. >> >> The patch introduce a XDP RSS hash type (xdp_rss_hash_type) that can both >> be seen as a number that is ordered according by ISO layer, and can be bit >> masked to separate IPv4 and IPv6 types for L4 protocols. Room is available >> for extending later while keeping these properties. This maps and unifies >> difference to hardware specific hashes. > > Would it be better to make use of the ETHTOOL_GRXFH defines (stuff > like UDP_V6_FLOW, RXH_L4_B_0_1 etc.)? Seems like that could allow > for some code reuse in drivers. Thanks for the point to ethtool defines. I can see that these are used when configuring the hardware RSS hash the NIC should calculate. From: include/uapi/linux/ethtool.h /* L3-L4 network traffic flow hash options */ #define RXH_L2DA (1 << 1) #define RXH_VLAN (1 << 2) #define RXH_L3_PROTO (1 << 3) #define RXH_IP_SRC (1 << 4) #define RXH_IP_DST (1 << 5) #define RXH_L4_B_0_1 (1 << 6) /* src port in case of TCP/UDP/SCTP */ #define RXH_L4_B_2_3 (1 << 7) /* dst port in case of TCP/UDP/SCTP */ #define RXH_DISCARD (1 << 31) I notice that I forgot about VLAN tag (RXH_VLAN) also can be part of the hash calc in my proposed design. It is interpreting to follow the possible ethool cmd->flow_type's: /* L2-L4 network traffic flow types */ #define TCP_V4_FLOW 0x01 /* hash or spec (tcp_ip4_spec) */ #define UDP_V4_FLOW 0x02 /* hash or spec (udp_ip4_spec) */ #define SCTP_V4_FLOW 0x03 /* hash or spec (sctp_ip4_spec) */ #define AH_ESP_V4_FLOW 0x04 /* hash only */ #define TCP_V6_FLOW 0x05 /* hash or spec (tcp_ip6_spec; nfc only) */ #define UDP_V6_FLOW 0x06 /* hash or spec (udp_ip6_spec; nfc only) */ #define SCTP_V6_FLOW 0x07 /* hash or spec (sctp_ip6_spec; nfc only) */ #define AH_ESP_V6_FLOW 0x08 /* hash only */ #define AH_V4_FLOW 0x09 /* hash or spec (ah_ip4_spec) */ #define ESP_V4_FLOW 0x0a /* hash or spec (esp_ip4_spec) */ #define AH_V6_FLOW 0x0b /* hash or spec (ah_ip6_spec; nfc only) */ #define ESP_V6_FLOW 0x0c /* hash or spec (esp_ip6_spec; nfc only) */ #define IPV4_USER_FLOW 0x0d /* spec only (usr_ip4_spec) */ #define IP_USER_FLOW IPV4_USER_FLOW #define IPV6_USER_FLOW 0x0e /* spec only (usr_ip6_spec; nfc only) */ #define IPV4_FLOW 0x10 /* hash only */ #define IPV6_FLOW 0x11 /* hash only */ #define ETHER_FLOW 0x12 /* spec only (ether_spec) */ /* Flag to enable additional fields in struct ethtool_rx_flow_spec */ #define FLOW_EXT 0x80000000 #define FLOW_MAC_EXT 0x40000000 /* Flag to enable RSS spreading of traffic matching rule (nfc only) */ #define FLOW_RSS 0x20000000 It is clear that we need to support TCP+UDP+SCTP. I assume the IPSEC is AH (Authentication Header) and ESP ( Encapsulating Security Payload. Thus, (like I found with mlx5) we also need IPSET and maybe a bit (or number) for each protocol AH or ESP. Both ah_ip4_spec and esp_ip4_spec points to ethtool.h struct: /** * struct ethtool_ah_espip4_spec - flow specification for IPsec/IPv4 * @ip4src: Source host * @ip4dst: Destination host * @spi: Security parameters index * @tos: Type-of-service * * This can be used to specify an IPsec transport or tunnel over IPv4. */ struct ethtool_ah_espip4_spec { __be32 ip4src; __be32 ip4dst; __be32 spi; __u8 tos; }; Which confirms that it is the SPI that is the extra part of the hash. --Jesper
On 03/29, Jesper Dangaard Brouer wrote: > On 28/03/2023 23.58, Stanislav Fomichev wrote: > > On 03/28, Jesper Dangaard Brouer wrote: > > > The RSS hash type specifies what portion of packet data NIC hardware > used > > > when calculating RSS hash value. The RSS types are focused on Internet > > > traffic protocols at OSI layers L3 and L4. L2 (e.g. ARP) often get > hash > > > value zero and no RSS type. For L3 focused on IPv4 vs. IPv6, and L4 > > > primarily TCP vs UDP, but some hardware supports SCTP. > > > > > Hardware RSS types are differently encoded for each hardware NIC. Most > > > hardware represent RSS hash type as a number. Determining L3 vs L4 > often > > > requires a mapping table as there often isn't a pattern or sorting > > > according to ISO layer. > > > > > The patch introduce a XDP RSS hash type (xdp_rss_hash_type) that can > both > > > be seen as a number that is ordered according by ISO layer, and can > be bit > > > masked to separate IPv4 and IPv6 types for L4 protocols. Room is > available > > > for extending later while keeping these properties. This maps and > unifies > > > difference to hardware specific hashes. > > > > Looks good overall. Any reason we're making this specific layout? > One important goal is to have a simple/fast way to determining L3 vs L4, > because a L4 hash can be used for flow handling (e.g. load-balancing). > We below layout you can: > if (rss_type & XDP_RSS_TYPE_L4_MASK) > bool hw_hash_do_LB = true; > Or using it as a number: > if (rss_type > XDP_RSS_TYPE_L4) > bool hw_hash_do_LB = true; Why is it strictly better then the following? if (rss_type & (TYPE_UDP | TYPE_TCP | TYPE_SCTP)) {} If we add some new L4 format, the bpf programs can be updated to support it? > I'm very open to changes to my "specific" layout. I am in doubt if > using it as a number is the right approach and worth the trouble. > > Why not simply the following? > > > > enum { > > ����XDP_RSS_TYPE_NONE = 0, > > ����XDP_RSS_TYPE_IPV4 = BIT(0), > > ����XDP_RSS_TYPE_IPV6 = BIT(1), > > ����/* IPv6 with extension header. */ > > ����/* let's note ^^^ it in the UAPI? */ > > ����XDP_RSS_TYPE_IPV6_EX = BIT(2), > > ����XDP_RSS_TYPE_UDP = BIT(3), > > ����XDP_RSS_TYPE_TCP = BIT(4), > > ����XDP_RSS_TYPE_SCTP = BIT(5), > We know these bits for UDP, TCP, SCTP (and IPSEC) are exclusive, they > cannot be set at the same time, e.g. as a packet cannot both be UDP and > TCP. Thus, using these bits as a number make sense to me, and is more > compact. [..] > This BIT() approach also have the issue of extending it later (forward > compatibility). As mentioned a common task will be to check if > hash-type is a L4 type. See mlx5 [patch 4/4] needed to extend with > IPSEC. Notice how my XDP_RSS_TYPE_L4_MASK covers all the bits that this > can be extended with new L4 types, such that existing progs will still > work checking for L4 check. It can of-cause be solved in the same way > for this BIT() approach by reserving some bits upfront in a mask. We're using 6 bits out of 64, we should be good for awhile? If there is ever a forward compatibility issue, we can always come up with a new kfunc. One other related question I have is: should we export the type over some additional new kfunc argument? (instead of abusing the return type) Maybe that will let us drop the explicit BTF_TYPE_EMIT as well? > > } > > > > And then using XDP_RSS_TYPE_IPV4|XDP_RSS_TYPE_UDP vs > > XDP_RSS_TYPE_IPV6|XXX ? > Do notice, that I already does some level of or'ing ("|") in this > proposal. The main difference is that I hide this from the driver, and > kind of pre-combine the valid combination (enum's) drivers can select > from. I do get the point, and I think I will come up with a combined > solution based on your input. > The RSS hashing types and combinations comes from M$ standards: > [1] > https://learn.microsoft.com/en-us/windows-hardware/drivers/network/rss-hashing-types#ipv4-hash-type-combinations My main concern here is that we're over-complicating it with the masks and the format. With the explicit bits we can easily map to that spec you mention. For example, for forward compat, I'm not sure we can assume that the people will do: "rss_type & XDP_RSS_TYPE_L4_MASK" instead of something like: "rss_type & (XDP_RSS_TYPE_L4_IPV4_TCP|XDP_RSS_TYPE_L4_IPV4_UDP)" > > > This proposal change the kfunc API bpf_xdp_metadata_rx_hash() to > return > > > this RSS hash type on success. > > > > > Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com> > > > --- > > > � include/net/xdp.h |�� 51 > > > +++++++++++++++++++++++++++++++++++++++++++++++++++ > > > � net/core/xdp.c��� |��� 4 +++- > > > � 2 files changed, 54 insertions(+), 1 deletion(-) > > > > > diff --git a/include/net/xdp.h b/include/net/xdp.h > > > index 5393b3ebe56e..63f462f5ea7f 100644 > > > --- a/include/net/xdp.h > > > +++ b/include/net/xdp.h > > > @@ -8,6 +8,7 @@ > > > > > � #include <linux/skbuff.h> /* skb_shared_info */ > > > � #include <uapi/linux/netdev.h> > > > +#include <linux/bitfield.h> > > > > > � /** > > > �� * DOC: XDP RX-queue information > > > @@ -396,6 +397,56 @@ XDP_METADATA_KFUNC_xxx > > > � MAX_XDP_METADATA_KFUNC, > > > � }; > > > > > +/* For partitioning of xdp_rss_hash_type */ > > > +#define RSS_L3������� GENMASK(2,0) /* 3-bits = values between 1-7 */ > > > +#define L4_BIT������� BIT(3)������ /* 1-bit - L4 indication */ > > > +#define RSS_L4_IPV4��� GENMASK(6,4) /* 3-bits */ > > > +#define RSS_L4_IPV6��� GENMASK(9,7) /* 3-bits */ > > > +#define RSS_L4������� GENMASK(9,3) /* = 7-bits - covering L4 > > > IPV4+IPV6 */ > > > +#define L4_IPV6_EX_BIT��� BIT(9)������ /* 1-bit - L4 IPv6 with > > > Extension hdr */ > > > +�������������������� /* 11-bits in total */ > > > + > > > +/* The XDP RSS hash type (xdp_rss_hash_type) can both be seen as a > number that > > > + * is ordered according by ISO layer, and can be bit masked to > separate IPv4 and > > > + * IPv6 types for L4 protocols. Room is available for extending > later while > > > + * keeping above properties, as this need to cover NIC hardware RSS > types. > > > + */ > > > +enum xdp_rss_hash_type { > > > +��� XDP_RSS_TYPE_NONE����������� = 0, > > > +��� XDP_RSS_TYPE_L2������������� = XDP_RSS_TYPE_NONE, > > > + > > > +��� XDP_RSS_TYPE_L3_MASK�������� = RSS_L3, > > > +��� XDP_RSS_TYPE_L3_IPV4�������� = FIELD_PREP_CONST(RSS_L3, 1), > > > +��� XDP_RSS_TYPE_L3_IPV6�������� = FIELD_PREP_CONST(RSS_L3, 2), > > > +��� XDP_RSS_TYPE_L3_IPV6_EX����� = FIELD_PREP_CONST(RSS_L3, 4), > > > + > > > +��� XDP_RSS_TYPE_L4_MASK�������� = RSS_L4, > > > +��� XDP_RSS_TYPE_L4_SHIFT������� = __bf_shf(RSS_L4), > > > +��� XDP_RSS_TYPE_L4_MASK_EX����� = RSS_L4 | L4_IPV6_EX_BIT, > > > + > > > +��� XDP_RSS_TYPE_L4_IPV4_MASK��� = RSS_L4_IPV4, > > > +��� XDP_RSS_TYPE_L4_BIT��������� = L4_BIT, > > > +��� XDP_RSS_TYPE_L4_IPV4_TCP���� = L4_BIT| > FIELD_PREP_CONST(RSS_L4_IPV4, 1), > > > +��� XDP_RSS_TYPE_L4_IPV4_UDP���� = L4_BIT| > FIELD_PREP_CONST(RSS_L4_IPV4, 2), > > > +��� XDP_RSS_TYPE_L4_IPV4_SCTP��� = L4_BIT| > FIELD_PREP_CONST(RSS_L4_IPV4, 3), > > > + > > > +��� XDP_RSS_TYPE_L4_IPV6_MASK��� = RSS_L4_IPV6, > > > +��� XDP_RSS_TYPE_L4_IPV6_TCP���� = L4_BIT| > FIELD_PREP_CONST(RSS_L4_IPV6, 1), > > > +��� XDP_RSS_TYPE_L4_IPV6_UDP���� = L4_BIT| > FIELD_PREP_CONST(RSS_L4_IPV6, 2), > > > +��� XDP_RSS_TYPE_L4_IPV6_SCTP��� = L4_BIT| > FIELD_PREP_CONST(RSS_L4_IPV6, 3), > > > + > > > +��� XDP_RSS_TYPE_L4_IPV6_EX_MASK = L4_IPV6_EX_BIT, > > > +��� XDP_RSS_TYPE_L4_IPV6_TCP_EX� = XDP_RSS_TYPE_L4_IPV6_TCP| > L4_IPV6_EX_BIT, > > > +��� XDP_RSS_TYPE_L4_IPV6_UDP_EX� = XDP_RSS_TYPE_L4_IPV6_UDP| > L4_IPV6_EX_BIT, > > > +��� XDP_RSS_TYPE_L4_IPV6_SCTP_EX = XDP_RSS_TYPE_L4_IPV6_SCTP| > L4_IPV6_EX_BIT, > > > +}; > > > +#undef RSS_L3 > > > +#undef L4_BIT > > > +#undef RSS_L4_IPV4 > > > +#undef RSS_L4_IPV6 > > > +#undef RSS_L4 > > > +#undef L4_IPV6_EX_BIT > > > + > > > � #ifdef CONFIG_NET > > > � u32 bpf_xdp_metadata_kfunc_id(int id); > > > � bool bpf_dev_bound_kfunc_id(u32 btf_id); > > > diff --git a/net/core/xdp.c b/net/core/xdp.c > > > index 7133017bcd74..81d41df30695 100644 > > > --- a/net/core/xdp.c > > > +++ b/net/core/xdp.c > > > @@ -721,12 +721,14 @@ __bpf_kfunc int > > > bpf_xdp_metadata_rx_timestamp(const struct xdp_md *ctx, u64 *tim > > > �� * @hash: Return value pointer. > > > �� * > > > �� * Return: > > > - * * Returns 0 on success or ``-errno`` on error. > > > + * * Returns (positive) RSS hash **type** on success or ``-errno`` > > > on error. > > > + * * ``enum xdp_rss_hash_type`` : RSS hash type > > > �� * * ``-EOPNOTSUPP`` : means device driver doesn't implement kfunc > > > �� * * ``-ENODATA``��� : means no RX-hash available for this frame > > > �� */ > > > � __bpf_kfunc int bpf_xdp_metadata_rx_hash(const struct xdp_md *ctx, > > > u32 *hash) > > > � { > > > +��� BTF_TYPE_EMIT(enum xdp_rss_hash_type); > > > ����� return -EOPNOTSUPP; > > > � } > > > > > >
On 29/03/2023 19.18, Stanislav Fomichev wrote: > On 03/29, Jesper Dangaard Brouer wrote: > >> On 28/03/2023 23.58, Stanislav Fomichev wrote: >> > On 03/28, Jesper Dangaard Brouer wrote: >> > > The RSS hash type specifies what portion of packet data NIC hardware used >> > > when calculating RSS hash value. The RSS types are focused on Internet >> > > traffic protocols at OSI layers L3 and L4. L2 (e.g. ARP) often get hash >> > > value zero and no RSS type. For L3 focused on IPv4 vs. IPv6, and L4 >> > > primarily TCP vs UDP, but some hardware supports SCTP. >> > >> > > Hardware RSS types are differently encoded for each hardware NIC. Most >> > > hardware represent RSS hash type as a number. Determining L3 vs L4 often >> > > requires a mapping table as there often isn't a pattern or sorting >> > > according to ISO layer. >> > >> > > The patch introduce a XDP RSS hash type (xdp_rss_hash_type) that can both >> > > be seen as a number that is ordered according by ISO layer, and can be bit >> > > masked to separate IPv4 and IPv6 types for L4 protocols. Room is available >> > > for extending later while keeping these properties. This maps and unifies >> > > difference to hardware specific hashes. >> > >> > Looks good overall. Any reason we're making this specific layout? > >> One important goal is to have a simple/fast way to determining L3 vs L4, >> because a L4 hash can be used for flow handling (e.g. load-balancing). > >> We below layout you can: > >> if (rss_type & XDP_RSS_TYPE_L4_MASK) >> bool hw_hash_do_LB = true; > >> Or using it as a number: > >> if (rss_type > XDP_RSS_TYPE_L4) >> bool hw_hash_do_LB = true; > > Why is it strictly better then the following? > > if (rss_type & (TYPE_UDP | TYPE_TCP | TYPE_SCTP)) {} > See V2 I dropped the idea of this being a number (that idea was not a good idea). > If we add some new L4 format, the bpf programs can be updated to support > it? > >> I'm very open to changes to my "specific" layout. I am in doubt if >> using it as a number is the right approach and worth the trouble. > >> > Why not simply the following? >> > >> > enum { >> > ����XDP_RSS_TYPE_NONE = 0, >> > ����XDP_RSS_TYPE_IPV4 = BIT(0), >> > ����XDP_RSS_TYPE_IPV6 = BIT(1), >> > ����/* IPv6 with extension header. */ >> > ����/* let's note ^^^ it in the UAPI? */ >> > ����XDP_RSS_TYPE_IPV6_EX = BIT(2), >> > ����XDP_RSS_TYPE_UDP = BIT(3), >> > ����XDP_RSS_TYPE_TCP = BIT(4), >> > ����XDP_RSS_TYPE_SCTP = BIT(5), > >> We know these bits for UDP, TCP, SCTP (and IPSEC) are exclusive, they >> cannot be set at the same time, e.g. as a packet cannot both be UDP and >> TCP. Thus, using these bits as a number make sense to me, and is more >> compact. > > [..] > >> This BIT() approach also have the issue of extending it later (forward >> compatibility). As mentioned a common task will be to check if >> hash-type is a L4 type. See mlx5 [patch 4/4] needed to extend with >> IPSEC. Notice how my XDP_RSS_TYPE_L4_MASK covers all the bits that this >> can be extended with new L4 types, such that existing progs will still >> work checking for L4 check. It can of-cause be solved in the same way >> for this BIT() approach by reserving some bits upfront in a mask. > > We're using 6 bits out of 64, we should be good for awhile? If there > is ever a forward compatibility issue, we can always come up with > a new kfunc. I want/need store the RSS-type in the xdp_frame, for XDP_REDIRECT and SKB use-cases. Thus, I don't want to use 64-bit/8-bytes, as xdp_frame size is limited (given it reduces headroom expansion). > > One other related question I have is: should we export the type > over some additional new kfunc argument? (instead of abusing the return > type) Good question. I was also wondering if it wouldn't be better to add another kfunc argument with the rss_hash_type? That will change the call signature, so that will not be easy to handle between kernel releases. > Maybe that will let us drop the explicit BTF_TYPE_EMIT as well? Sure, if we define it as an argument, then it will automatically exported as BTF. >> > } >> > >> > And then using XDP_RSS_TYPE_IPV4|XDP_RSS_TYPE_UDP vs >> > XDP_RSS_TYPE_IPV6|XXX ? > >> Do notice, that I already does some level of or'ing ("|") in this >> proposal. The main difference is that I hide this from the driver, and >> kind of pre-combine the valid combination (enum's) drivers can select >> from. I do get the point, and I think I will come up with a combined >> solution based on your input. > > >> The RSS hashing types and combinations comes from M$ standards: >> [1] >> https://learn.microsoft.com/en-us/windows-hardware/drivers/network/rss-hashing-types#ipv4-hash-type-combinations > > My main concern here is that we're over-complicating it with the masks > and the format. With the explicit bits we can easily map to that > spec you mention. See if you like my RFC-V2 proposal better. It should go more in your direction. > > For example, for forward compat, I'm not sure we can assume that the people > will do: > "rss_type & XDP_RSS_TYPE_L4_MASK" > instead of something like: > "rss_type & (XDP_RSS_TYPE_L4_IPV4_TCP|XDP_RSS_TYPE_L4_IPV4_UDP)" > This code is allowed in V2 and should be. It is a choice of BPF-programmer in line-2 to not be forward compatible with newer L4 types. >> > > This proposal change the kfunc API bpf_xdp_metadata_rx_hash() to return >> > > this RSS hash type on success. This is the real question (as also raised above)... Should we use return value or add an argument for type? --Jesper
On 03/29, Jesper Dangaard Brouer wrote: > On 29/03/2023 19.18, Stanislav Fomichev wrote: > > On 03/29, Jesper Dangaard Brouer wrote: > > > > > On 28/03/2023 23.58, Stanislav Fomichev wrote: > > > > On 03/28, Jesper Dangaard Brouer wrote: > > > > > The RSS hash type specifies what portion of packet data NIC > hardware used > > > > > when calculating RSS hash value. The RSS types are focused on > Internet > > > > > traffic protocols at OSI layers L3 and L4. L2 (e.g. ARP) often > get hash > > > > > value zero and no RSS type. For L3 focused on IPv4 vs. IPv6, and > L4 > > > > > primarily TCP vs UDP, but some hardware supports SCTP. > > > > > > > > > Hardware RSS types are differently encoded for each hardware NIC. > Most > > > > > hardware represent RSS hash type as a number. Determining L3 vs > L4 often > > > > > requires a mapping table as there often isn't a pattern or sorting > > > > > according to ISO layer. > > > > > > > > > The patch introduce a XDP RSS hash type (xdp_rss_hash_type) that > can both > > > > > be seen as a number that is ordered according by ISO layer, and > can be bit > > > > > masked to separate IPv4 and IPv6 types for L4 protocols. Room is > available > > > > > for extending later while keeping these properties. This maps and > unifies > > > > > difference to hardware specific hashes. > > > > > > > > Looks good overall. Any reason we're making this specific layout? > > > > > One important goal is to have a simple/fast way to determining L3 vs > L4, > > > because a L4 hash can be used for flow handling (e.g. load-balancing). > > > > > We below layout you can: > > > > > if (rss_type & XDP_RSS_TYPE_L4_MASK) > > > bool hw_hash_do_LB = true; > > > > > Or using it as a number: > > > > > if (rss_type > XDP_RSS_TYPE_L4) > > > bool hw_hash_do_LB = true; > > > > Why is it strictly better then the following? > > > > if (rss_type & (TYPE_UDP | TYPE_TCP | TYPE_SCTP)) {} > > > See V2 I dropped the idea of this being a number (that idea was not a > good idea).
On 30/03/2023 01.19, Stanislav Fomichev wrote: > On 03/29, Jesper Dangaard Brouer wrote: > >> On 29/03/2023 19.18, Stanislav Fomichev wrote: >> > On 03/29, Jesper Dangaard Brouer wrote: >> > >> > > On 28/03/2023 23.58, Stanislav Fomichev wrote: >> > > > On 03/28, Jesper Dangaard Brouer wrote: >> > > > > The RSS hash type specifies what portion of packet data NIC hardware used >> > > > > when calculating RSS hash value. The RSS types are focused on Internet >> > > > > traffic protocols at OSI layers L3 and L4. L2 (e.g. ARP) often get hash >> > > > > value zero and no RSS type. For L3 focused on IPv4 vs. IPv6, and L4 >> > > > > primarily TCP vs UDP, but some hardware supports SCTP. >> > > > >> > > > > Hardware RSS types are differently encoded for each hardware NIC. Most >> > > > > hardware represent RSS hash type as a number. Determining L3 vs L4 often >> > > > > requires a mapping table as there often isn't a pattern or sorting >> > > > > according to ISO layer. >> > > > >> > > > > The patch introduce a XDP RSS hash type (xdp_rss_hash_type) that can both >> > > > > be seen as a number that is ordered according by ISO layer, and can be bit >> > > > > masked to separate IPv4 and IPv6 types for L4 protocols. Room is available >> > > > > for extending later while keeping these properties. This maps and unifies >> > > > > difference to hardware specific hashes. >> > > > >> > > > Looks good overall. Any reason we're making this specific layout? >> > >> > > One important goal is to have a simple/fast way to determining L3 vs L4, >> > > because a L4 hash can be used for flow handling (e.g. load-balancing). >> > >> > > We below layout you can: >> > >> > > if (rss_type & XDP_RSS_TYPE_L4_MASK) >> > > bool hw_hash_do_LB = true; >> > >> > > Or using it as a number: >> > >> > > if (rss_type > XDP_RSS_TYPE_L4) >> > > bool hw_hash_do_LB = true; >> > >> > Why is it strictly better then the following? >> > >> > if (rss_type & (TYPE_UDP | TYPE_TCP | TYPE_SCTP)) {} >> > > >> See V2 I dropped the idea of this being a number (that idea was not a >> good idea). > >
On 03/30, Jesper Dangaard Brouer wrote: > On 30/03/2023 01.19, Stanislav Fomichev wrote: > > On 03/29, Jesper Dangaard Brouer wrote: > > > > > On 29/03/2023 19.18, Stanislav Fomichev wrote: > > > > On 03/29, Jesper Dangaard Brouer wrote: > > > > > > > > > On 28/03/2023 23.58, Stanislav Fomichev wrote: > > > > > > On 03/28, Jesper Dangaard Brouer wrote: > > > > > > > The RSS hash type specifies what portion of packet data NIC > hardware used > > > > > > > when calculating RSS hash value. The RSS types are focused on > Internet > > > > > > > traffic protocols at OSI layers L3 and L4. L2 (e.g. ARP) > often get hash > > > > > > > value zero and no RSS type. For L3 focused on IPv4 vs. IPv6, > and L4 > > > > > > > primarily TCP vs UDP, but some hardware supports SCTP. > > > > > > > > > > > > > Hardware RSS types are differently encoded for each hardware > NIC. Most > > > > > > > hardware represent RSS hash type as a number. Determining L3 > vs L4 often > > > > > > > requires a mapping table as there often isn't a pattern or > sorting > > > > > > > according to ISO layer. > > > > > > > > > > > > > The patch introduce a XDP RSS hash type (xdp_rss_hash_type) > that can both > > > > > > > be seen as a number that is ordered according by ISO layer, > and can be bit > > > > > > > masked to separate IPv4 and IPv6 types for L4 protocols. Room > is available > > > > > > > for extending later while keeping these properties. This maps > and unifies > > > > > > > difference to hardware specific hashes. > > > > > > > > > > > > Looks good overall. Any reason we're making this specific > layout? > > > > > > > > > One important goal is to have a simple/fast way to determining L3 > vs L4, > > > > > because a L4 hash can be used for flow handling (e.g. > load-balancing). > > > > > > > > > We below layout you can: > > > > > > > > > if (rss_type & XDP_RSS_TYPE_L4_MASK) > > > > > bool hw_hash_do_LB = true; > > > > > > > > > Or using it as a number: > > > > > > > > > if (rss_type > XDP_RSS_TYPE_L4) > > > > > bool hw_hash_do_LB = true; > > > > > > > > Why is it strictly better then the following? > > > > > > > > if (rss_type & (TYPE_UDP | TYPE_TCP | TYPE_SCTP)) {} > > > > > > > > > See V2 I dropped the idea of this being a number (that idea was not a > > > good idea). > > > >
On 30/03/2023 19.11, Stanislav Fomichev wrote: > On 03/30, Jesper Dangaard Brouer wrote: > >> On 30/03/2023 01.19, Stanislav Fomichev wrote: >> > On 03/29, Jesper Dangaard Brouer wrote: >> > >> > > On 29/03/2023 19.18, Stanislav Fomichev wrote: >> > > > On 03/29, Jesper Dangaard Brouer wrote: >> > > > >> > > > > On 28/03/2023 23.58, Stanislav Fomichev wrote: >> > > > > > On 03/28, Jesper Dangaard Brouer wrote: >> > > > > > > The RSS hash type specifies what portion of packet data NIC hardware used >> > > > > > > when calculating RSS hash value. The RSS types are focused on Internet >> > > > > > > traffic protocols at OSI layers L3 and L4. L2 (e.g. ARP) often get hash >> > > > > > > value zero and no RSS type. For L3 focused on IPv4 vs. IPv6, and L4 >> > > > > > > primarily TCP vs UDP, but some hardware supports SCTP. >> > > > > > >> > > > > > > Hardware RSS types are differently encoded for each hardware NIC. Most >> > > > > > > hardware represent RSS hash type as a number. Determining L3 vs L4 often >> > > > > > > requires a mapping table as there often isn't a pattern or sorting >> > > > > > > according to ISO layer. >> > > > > > [...] >> > Any reason it's not a XDP_RSS_L3_IPV6_EX within XDP_RSS_L3_MASK? >> > > >> Hmm... I guess it belongs with L3. > >> Do notice that both IPv4 and IPv6 have a flexible header called either >> options/extensions headers, after their fixed header. (Mlx4 HW contains this >> info for IPv4, but I didn't extend xdp_rss_hash_type in that patch). >> Thus, we could have a single BIT that is valid for both IPv4 and IPv6. >> (This can help speedup packet parsing having this info). > > A separate bit for both v4/v6 sounds good. But thinking more about it, > not sure what the users are supposed to do with it. Whether the flow is > hashed over the extension header should a config option, not a per-packet signal? > Microsoft defines which part of the IPv6 Extensions headers will be used for replacing either the Source (Home address) and Dest (Routing-Header-Type-2) IPv6 Addresses, in the hash calc, here[1]: [1] https://learn.microsoft.com/en-us/windows-hardware/drivers/network/rss-hashing-types#ndis_hash_ipv6_ex The igc/i225 chip returns per-packet the RSS Type's with _EX added. Thus, I implemented this per-packet basis. >> [...] >> > >> > > > For example, for forward compat, I'm not sure we can assume that >> the people >> > > > will do: >> > > > "rss_type & XDP_RSS_TYPE_L4_MASK" >> > > > instead of something like: >> > > > "rss_type & >> (XDP_RSS_TYPE_L4_IPV4_TCP|XDP_RSS_TYPE_L4_IPV4_UDP)" >> > > > >> > >> > > This code is allowed in V2 and should be. It is a choice of >> > > BPF-programmer in line-2 to not be forward compatible with newer L4 >> > > types. >> > > >> The above code made me realize, I was wrong and you are right, we should >> represent the L4 types as BITs (and not as numbers). >> Even-though a single packet cannot be both UDP and TCP at the same time, >> then it is reasonable to have a code path that want to match both UDP >> and TCP. If L4 types are BITs then code can do a single compare (via >> ORing), while if they are numbers then we need more compares. >> Thus, I'll change scheme in V3 to use BITs. > > So you are saying that the following: > if (rss_type & (TCP|UDP) > > is much faster than the following: > proto = rss_type & L4_MASK; > if (proto == TCP || proto == UDP) > > ? For XDP every instruction/cycle counts. Just to make sure, I tested it with godbolt.org, 3 vs 4 inst. > > idk, as long as we have enough bits to represent everything, I'm fine > with either way, up to you. (not sure how much you want to constrain the > data > to fit it into xdp_frame; assuming u16 is fine?) Yes, u16 is fine. > > >> > > > > > > This proposal change the kfunc API >> > > bpf_xdp_metadata_rx_hash() > > > > to return this RSS hash type on >> > > success. >> > >> > > This is the real question (as also raised above)... >> > > Should we use return value or add an argument for type? >> > >> > Let's fix the prototype while it's still early in the rc? > >> Okay, in V3 I will propose adding an argument for the type then. > > SG, thx! > >> > Maybe also extend the tests to drop/decode/verify the mask? > >> Yes, I/we obviously need to update the selftests. > >> One problem with selftests is that it's using veth SKB-based mode, and >> SKB's have lost the RSS hash info and converted this into a single BIT >> telling us if this was L4 based. Thus, its hard to do some e.g. UDP >> type verification, but I guess we can check if expected UDP packet is >> RSS type L4. > > Yeah, sounds fair. > >> In xdp_hw_metadata, I will add something that uses the RSS type bits. I >> was thinking to match against L4-UDP RSS type as program only AF_XDP >> redirect UDP packets, so we can verify it was a UDP packet by HW info. > > Or maybe just dump it, idk.
diff --git a/include/net/xdp.h b/include/net/xdp.h index 5393b3ebe56e..63f462f5ea7f 100644 --- a/include/net/xdp.h +++ b/include/net/xdp.h @@ -8,6 +8,7 @@ #include <linux/skbuff.h> /* skb_shared_info */ #include <uapi/linux/netdev.h> +#include <linux/bitfield.h> /** * DOC: XDP RX-queue information @@ -396,6 +397,56 @@ XDP_METADATA_KFUNC_xxx MAX_XDP_METADATA_KFUNC, }; +/* For partitioning of xdp_rss_hash_type */ +#define RSS_L3 GENMASK(2,0) /* 3-bits = values between 1-7 */ +#define L4_BIT BIT(3) /* 1-bit - L4 indication */ +#define RSS_L4_IPV4 GENMASK(6,4) /* 3-bits */ +#define RSS_L4_IPV6 GENMASK(9,7) /* 3-bits */ +#define RSS_L4 GENMASK(9,3) /* = 7-bits - covering L4 IPV4+IPV6 */ +#define L4_IPV6_EX_BIT BIT(9) /* 1-bit - L4 IPv6 with Extension hdr */ + /* 11-bits in total */ + +/* The XDP RSS hash type (xdp_rss_hash_type) can both be seen as a number that + * is ordered according by ISO layer, and can be bit masked to separate IPv4 and + * IPv6 types for L4 protocols. Room is available for extending later while + * keeping above properties, as this need to cover NIC hardware RSS types. + */ +enum xdp_rss_hash_type { + XDP_RSS_TYPE_NONE = 0, + XDP_RSS_TYPE_L2 = XDP_RSS_TYPE_NONE, + + XDP_RSS_TYPE_L3_MASK = RSS_L3, + XDP_RSS_TYPE_L3_IPV4 = FIELD_PREP_CONST(RSS_L3, 1), + XDP_RSS_TYPE_L3_IPV6 = FIELD_PREP_CONST(RSS_L3, 2), + XDP_RSS_TYPE_L3_IPV6_EX = FIELD_PREP_CONST(RSS_L3, 4), + + XDP_RSS_TYPE_L4_MASK = RSS_L4, + XDP_RSS_TYPE_L4_SHIFT = __bf_shf(RSS_L4), + XDP_RSS_TYPE_L4_MASK_EX = RSS_L4 | L4_IPV6_EX_BIT, + + XDP_RSS_TYPE_L4_IPV4_MASK = RSS_L4_IPV4, + XDP_RSS_TYPE_L4_BIT = L4_BIT, + XDP_RSS_TYPE_L4_IPV4_TCP = L4_BIT|FIELD_PREP_CONST(RSS_L4_IPV4, 1), + XDP_RSS_TYPE_L4_IPV4_UDP = L4_BIT|FIELD_PREP_CONST(RSS_L4_IPV4, 2), + XDP_RSS_TYPE_L4_IPV4_SCTP = L4_BIT|FIELD_PREP_CONST(RSS_L4_IPV4, 3), + + XDP_RSS_TYPE_L4_IPV6_MASK = RSS_L4_IPV6, + XDP_RSS_TYPE_L4_IPV6_TCP = L4_BIT|FIELD_PREP_CONST(RSS_L4_IPV6, 1), + XDP_RSS_TYPE_L4_IPV6_UDP = L4_BIT|FIELD_PREP_CONST(RSS_L4_IPV6, 2), + XDP_RSS_TYPE_L4_IPV6_SCTP = L4_BIT|FIELD_PREP_CONST(RSS_L4_IPV6, 3), + + XDP_RSS_TYPE_L4_IPV6_EX_MASK = L4_IPV6_EX_BIT, + XDP_RSS_TYPE_L4_IPV6_TCP_EX = XDP_RSS_TYPE_L4_IPV6_TCP |L4_IPV6_EX_BIT, + XDP_RSS_TYPE_L4_IPV6_UDP_EX = XDP_RSS_TYPE_L4_IPV6_UDP |L4_IPV6_EX_BIT, + XDP_RSS_TYPE_L4_IPV6_SCTP_EX = XDP_RSS_TYPE_L4_IPV6_SCTP|L4_IPV6_EX_BIT, +}; +#undef RSS_L3 +#undef L4_BIT +#undef RSS_L4_IPV4 +#undef RSS_L4_IPV6 +#undef RSS_L4 +#undef L4_IPV6_EX_BIT + #ifdef CONFIG_NET u32 bpf_xdp_metadata_kfunc_id(int id); bool bpf_dev_bound_kfunc_id(u32 btf_id); diff --git a/net/core/xdp.c b/net/core/xdp.c index 7133017bcd74..81d41df30695 100644 --- a/net/core/xdp.c +++ b/net/core/xdp.c @@ -721,12 +721,14 @@ __bpf_kfunc int bpf_xdp_metadata_rx_timestamp(const struct xdp_md *ctx, u64 *tim * @hash: Return value pointer. * * Return: - * * Returns 0 on success or ``-errno`` on error. + * * Returns (positive) RSS hash **type** on success or ``-errno`` on error. + * * ``enum xdp_rss_hash_type`` : RSS hash type * * ``-EOPNOTSUPP`` : means device driver doesn't implement kfunc * * ``-ENODATA`` : means no RX-hash available for this frame */ __bpf_kfunc int bpf_xdp_metadata_rx_hash(const struct xdp_md *ctx, u32 *hash) { + BTF_TYPE_EMIT(enum xdp_rss_hash_type); return -EOPNOTSUPP; }
The RSS hash type specifies what portion of packet data NIC hardware used when calculating RSS hash value. The RSS types are focused on Internet traffic protocols at OSI layers L3 and L4. L2 (e.g. ARP) often get hash value zero and no RSS type. For L3 focused on IPv4 vs. IPv6, and L4 primarily TCP vs UDP, but some hardware supports SCTP. Hardware RSS types are differently encoded for each hardware NIC. Most hardware represent RSS hash type as a number. Determining L3 vs L4 often requires a mapping table as there often isn't a pattern or sorting according to ISO layer. The patch introduce a XDP RSS hash type (xdp_rss_hash_type) that can both be seen as a number that is ordered according by ISO layer, and can be bit masked to separate IPv4 and IPv6 types for L4 protocols. Room is available for extending later while keeping these properties. This maps and unifies difference to hardware specific hashes. This proposal change the kfunc API bpf_xdp_metadata_rx_hash() to return this RSS hash type on success. Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com> --- include/net/xdp.h | 51 +++++++++++++++++++++++++++++++++++++++++++++++++++ net/core/xdp.c | 4 +++- 2 files changed, 54 insertions(+), 1 deletion(-)