diff options
-rw-r--r-- | drivers/net/ethernet/chelsio/cxgb4/t4_hw.c | 10 | ||||
-rw-r--r-- | drivers/net/ethernet/dec/tulip/tulip_core.c | 5 | ||||
-rw-r--r-- | drivers/net/ethernet/pensando/ionic/ionic_bus_pci.c | 3 | ||||
-rw-r--r-- | drivers/net/ethernet/sfc/efx_channels.c | 7 | ||||
-rw-r--r-- | drivers/net/ethernet/sfc/ptp.c | 14 | ||||
-rw-r--r-- | drivers/net/ethernet/sfc/ptp.h | 1 | ||||
-rw-r--r-- | drivers/net/phy/phy.c | 7 | ||||
-rw-r--r-- | drivers/ptp/ptp_ocp.c | 2 | ||||
-rw-r--r-- | lib/dim/net_dim.c | 44 | ||||
-rw-r--r-- | net/core/skbuff.c | 4 | ||||
-rw-r--r-- | net/ipv4/route.c | 1 | ||||
-rw-r--r-- | net/netlink/af_netlink.c | 1 | ||||
-rw-r--r-- | tools/testing/selftests/net/Makefile | 3 | ||||
-rw-r--r-- | tools/testing/selftests/net/bpf/Makefile | 14 | ||||
-rw-r--r-- | tools/testing/selftests/net/bpf/nat6to4.c | 285 | ||||
-rwxr-xr-x | tools/testing/selftests/net/udpgro_frglist.sh | 101 |
16 files changed, 467 insertions, 35 deletions
diff --git a/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c b/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c index e7b4e3ed056c..8d719f82854a 100644 --- a/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c +++ b/drivers/net/ethernet/chelsio/cxgb4/t4_hw.c @@ -2793,14 +2793,14 @@ int t4_get_raw_vpd_params(struct adapter *adapter, struct vpd_params *p) goto out; na = ret; - memcpy(p->id, vpd + id, min_t(int, id_len, ID_LEN)); + memcpy(p->id, vpd + id, min_t(unsigned int, id_len, ID_LEN)); strim(p->id); - memcpy(p->sn, vpd + sn, min_t(int, sn_len, SERNUM_LEN)); + memcpy(p->sn, vpd + sn, min_t(unsigned int, sn_len, SERNUM_LEN)); strim(p->sn); - memcpy(p->pn, vpd + pn, min_t(int, pn_len, PN_LEN)); + memcpy(p->pn, vpd + pn, min_t(unsigned int, pn_len, PN_LEN)); strim(p->pn); - memcpy(p->na, vpd + na, min_t(int, na_len, MACADDR_LEN)); - strim((char *)p->na); + memcpy(p->na, vpd + na, min_t(unsigned int, na_len, MACADDR_LEN)); + strim(p->na); out: vfree(vpd); diff --git a/drivers/net/ethernet/dec/tulip/tulip_core.c b/drivers/net/ethernet/dec/tulip/tulip_core.c index 79df5a72877b..0040dcaab945 100644 --- a/drivers/net/ethernet/dec/tulip/tulip_core.c +++ b/drivers/net/ethernet/dec/tulip/tulip_core.c @@ -1399,8 +1399,10 @@ static int tulip_init_one(struct pci_dev *pdev, const struct pci_device_id *ent) /* alloc_etherdev ensures aligned and zeroed private structures */ dev = alloc_etherdev (sizeof (*tp)); - if (!dev) + if (!dev) { + pci_disable_device(pdev); return -ENOMEM; + } SET_NETDEV_DEV(dev, &pdev->dev); if (pci_resource_len (pdev, 0) < tulip_tbl[chip_idx].io_size) { @@ -1785,6 +1787,7 @@ err_out_free_res: err_out_free_netdev: free_netdev (dev); + pci_disable_device(pdev); return -ENODEV; } diff --git a/drivers/net/ethernet/pensando/ionic/ionic_bus_pci.c b/drivers/net/ethernet/pensando/ionic/ionic_bus_pci.c index 6ffc62c41165..0a7a757494bc 100644 --- a/drivers/net/ethernet/pensando/ionic/ionic_bus_pci.c +++ b/drivers/net/ethernet/pensando/ionic/ionic_bus_pci.c @@ -256,7 +256,7 @@ static int ionic_probe(struct pci_dev *pdev, const struct pci_device_id *ent) err = ionic_map_bars(ionic); if (err) - goto err_out_pci_disable_device; + goto err_out_pci_release_regions; /* Configure the device */ err = ionic_setup(ionic); @@ -360,6 +360,7 @@ err_out_teardown: err_out_unmap_bars: ionic_unmap_bars(ionic); +err_out_pci_release_regions: pci_release_regions(pdev); err_out_pci_disable_device: pci_disable_device(pdev); diff --git a/drivers/net/ethernet/sfc/efx_channels.c b/drivers/net/ethernet/sfc/efx_channels.c index 377df8b7f015..40df910aa140 100644 --- a/drivers/net/ethernet/sfc/efx_channels.c +++ b/drivers/net/ethernet/sfc/efx_channels.c @@ -867,7 +867,9 @@ static void efx_set_xdp_channels(struct efx_nic *efx) int efx_realloc_channels(struct efx_nic *efx, u32 rxq_entries, u32 txq_entries) { - struct efx_channel *other_channel[EFX_MAX_CHANNELS], *channel; + struct efx_channel *other_channel[EFX_MAX_CHANNELS], *channel, + *ptp_channel = efx_ptp_channel(efx); + struct efx_ptp_data *ptp_data = efx->ptp_data; unsigned int i, next_buffer_table = 0; u32 old_rxq_entries, old_txq_entries; int rc, rc2; @@ -938,6 +940,7 @@ int efx_realloc_channels(struct efx_nic *efx, u32 rxq_entries, u32 txq_entries) efx_set_xdp_channels(efx); out: + efx->ptp_data = NULL; /* Destroy unused channel structures */ for (i = 0; i < efx->n_channels; i++) { channel = other_channel[i]; @@ -948,6 +951,7 @@ out: } } + efx->ptp_data = ptp_data; rc2 = efx_soft_enable_interrupts(efx); if (rc2) { rc = rc ? rc : rc2; @@ -966,6 +970,7 @@ rollback: efx->txq_entries = old_txq_entries; for (i = 0; i < efx->n_channels; i++) swap(efx->channel[i], other_channel[i]); + efx_ptp_update_channel(efx, ptp_channel); goto out; } diff --git a/drivers/net/ethernet/sfc/ptp.c b/drivers/net/ethernet/sfc/ptp.c index f0ef515e2ade..4625f85acab2 100644 --- a/drivers/net/ethernet/sfc/ptp.c +++ b/drivers/net/ethernet/sfc/ptp.c @@ -45,6 +45,7 @@ #include "farch_regs.h" #include "tx.h" #include "nic.h" /* indirectly includes ptp.h */ +#include "efx_channels.h" /* Maximum number of events expected to make up a PTP event */ #define MAX_EVENT_FRAGS 3 @@ -541,6 +542,12 @@ struct efx_channel *efx_ptp_channel(struct efx_nic *efx) return efx->ptp_data ? efx->ptp_data->channel : NULL; } +void efx_ptp_update_channel(struct efx_nic *efx, struct efx_channel *channel) +{ + if (efx->ptp_data) + efx->ptp_data->channel = channel; +} + static u32 last_sync_timestamp_major(struct efx_nic *efx) { struct efx_channel *channel = efx_ptp_channel(efx); @@ -1443,6 +1450,11 @@ int efx_ptp_probe(struct efx_nic *efx, struct efx_channel *channel) int rc = 0; unsigned int pos; + if (efx->ptp_data) { + efx->ptp_data->channel = channel; + return 0; + } + ptp = kzalloc(sizeof(struct efx_ptp_data), GFP_KERNEL); efx->ptp_data = ptp; if (!efx->ptp_data) @@ -2176,7 +2188,7 @@ static const struct efx_channel_type efx_ptp_channel_type = { .pre_probe = efx_ptp_probe_channel, .post_remove = efx_ptp_remove_channel, .get_name = efx_ptp_get_channel_name, - /* no copy operation; there is no need to reallocate this channel */ + .copy = efx_copy_channel, .receive_skb = efx_ptp_rx, .want_txqs = efx_ptp_want_txqs, .keep_eventq = false, diff --git a/drivers/net/ethernet/sfc/ptp.h b/drivers/net/ethernet/sfc/ptp.h index 9855e8c9e544..7b1ef7002b3f 100644 --- a/drivers/net/ethernet/sfc/ptp.h +++ b/drivers/net/ethernet/sfc/ptp.h @@ -16,6 +16,7 @@ struct ethtool_ts_info; int efx_ptp_probe(struct efx_nic *efx, struct efx_channel *channel); void efx_ptp_defer_probe_with_channel(struct efx_nic *efx); struct efx_channel *efx_ptp_channel(struct efx_nic *efx); +void efx_ptp_update_channel(struct efx_nic *efx, struct efx_channel *channel); void efx_ptp_remove(struct efx_nic *efx); int efx_ptp_set_ts_config(struct efx_nic *efx, struct ifreq *ifr); int efx_ptp_get_ts_config(struct efx_nic *efx, struct ifreq *ifr); diff --git a/drivers/net/phy/phy.c b/drivers/net/phy/phy.c index beb2b66da132..f122026c4682 100644 --- a/drivers/net/phy/phy.c +++ b/drivers/net/phy/phy.c @@ -970,8 +970,13 @@ static irqreturn_t phy_interrupt(int irq, void *phy_dat) { struct phy_device *phydev = phy_dat; struct phy_driver *drv = phydev->drv; + irqreturn_t ret; - return drv->handle_interrupt(phydev); + mutex_lock(&phydev->lock); + ret = drv->handle_interrupt(phydev); + mutex_unlock(&phydev->lock); + + return ret; } /** diff --git a/drivers/ptp/ptp_ocp.c b/drivers/ptp/ptp_ocp.c index 0feaa4b45317..dd45471f6780 100644 --- a/drivers/ptp/ptp_ocp.c +++ b/drivers/ptp/ptp_ocp.c @@ -1557,7 +1557,7 @@ ptp_ocp_signal_set(struct ptp_ocp *bp, int gen, struct ptp_ocp_signal *s) start_ns = ktime_set(ts.tv_sec, ts.tv_nsec) + NSEC_PER_MSEC; if (!s->start) { /* roundup() does not work on 32-bit systems */ - s->start = DIV_ROUND_UP_ULL(start_ns, s->period); + s->start = DIV64_U64_ROUND_UP(start_ns, s->period); s->start = ktime_add(s->start, s->phase); } diff --git a/lib/dim/net_dim.c b/lib/dim/net_dim.c index 06811d866775..53f6b9c6e936 100644 --- a/lib/dim/net_dim.c +++ b/lib/dim/net_dim.c @@ -12,41 +12,41 @@ * Each profile size must be of NET_DIM_PARAMS_NUM_PROFILES */ #define NET_DIM_PARAMS_NUM_PROFILES 5 -#define NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE 256 -#define NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE 128 +#define NET_DIM_DEFAULT_RX_CQ_PKTS_FROM_EQE 256 +#define NET_DIM_DEFAULT_TX_CQ_PKTS_FROM_EQE 128 #define NET_DIM_DEF_PROFILE_CQE 1 #define NET_DIM_DEF_PROFILE_EQE 1 #define NET_DIM_RX_EQE_PROFILES { \ - {1, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \ - {8, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \ - {64, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \ - {128, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \ - {256, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \ + {.usec = 1, .pkts = NET_DIM_DEFAULT_RX_CQ_PKTS_FROM_EQE,}, \ + {.usec = 8, .pkts = NET_DIM_DEFAULT_RX_CQ_PKTS_FROM_EQE,}, \ + {.usec = 64, .pkts = NET_DIM_DEFAULT_RX_CQ_PKTS_FROM_EQE,}, \ + {.usec = 128, .pkts = NET_DIM_DEFAULT_RX_CQ_PKTS_FROM_EQE,}, \ + {.usec = 256, .pkts = NET_DIM_DEFAULT_RX_CQ_PKTS_FROM_EQE,} \ } #define NET_DIM_RX_CQE_PROFILES { \ - {2, 256}, \ - {8, 128}, \ - {16, 64}, \ - {32, 64}, \ - {64, 64} \ + {.usec = 2, .pkts = 256,}, \ + {.usec = 8, .pkts = 128,}, \ + {.usec = 16, .pkts = 64,}, \ + {.usec = 32, .pkts = 64,}, \ + {.usec = 64, .pkts = 64,} \ } #define NET_DIM_TX_EQE_PROFILES { \ - {1, NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE}, \ - {8, NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE}, \ - {32, NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE}, \ - {64, NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE}, \ - {128, NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE} \ + {.usec = 1, .pkts = NET_DIM_DEFAULT_TX_CQ_PKTS_FROM_EQE,}, \ + {.usec = 8, .pkts = NET_DIM_DEFAULT_TX_CQ_PKTS_FROM_EQE,}, \ + {.usec = 32, .pkts = NET_DIM_DEFAULT_TX_CQ_PKTS_FROM_EQE,}, \ + {.usec = 64, .pkts = NET_DIM_DEFAULT_TX_CQ_PKTS_FROM_EQE,}, \ + {.usec = 128, .pkts = NET_DIM_DEFAULT_TX_CQ_PKTS_FROM_EQE,} \ } #define NET_DIM_TX_CQE_PROFILES { \ - {5, 128}, \ - {8, 64}, \ - {16, 32}, \ - {32, 32}, \ - {64, 32} \ + {.usec = 5, .pkts = 128,}, \ + {.usec = 8, .pkts = 64,}, \ + {.usec = 16, .pkts = 32,}, \ + {.usec = 32, .pkts = 32,}, \ + {.usec = 64, .pkts = 32,} \ } static const struct dim_cq_moder diff --git a/net/core/skbuff.c b/net/core/skbuff.c index 30b523fa4ad2..c90c74de90d5 100644 --- a/net/core/skbuff.c +++ b/net/core/skbuff.c @@ -3897,7 +3897,7 @@ struct sk_buff *skb_segment_list(struct sk_buff *skb, unsigned int delta_len = 0; struct sk_buff *tail = NULL; struct sk_buff *nskb, *tmp; - int err; + int len_diff, err; skb_push(skb, -skb_network_offset(skb) + offset); @@ -3937,9 +3937,11 @@ struct sk_buff *skb_segment_list(struct sk_buff *skb, skb_push(nskb, -skb_network_offset(nskb) + offset); skb_release_head_state(nskb); + len_diff = skb_network_header_len(nskb) - skb_network_header_len(skb); __copy_skb_header(nskb, skb); skb_headers_offset_update(nskb, skb_headroom(nskb) - skb_headroom(skb)); + nskb->transport_header += len_diff; skb_copy_from_linear_data_offset(skb, -tnl_hlen, nskb->data - tnl_hlen, offset + tnl_hlen); diff --git a/net/ipv4/route.c b/net/ipv4/route.c index 98c6f3429593..57abd27e842c 100644 --- a/net/ipv4/route.c +++ b/net/ipv4/route.c @@ -1753,6 +1753,7 @@ static int ip_route_input_mc(struct sk_buff *skb, __be32 daddr, __be32 saddr, #endif RT_CACHE_STAT_INC(in_slow_mc); + skb_dst_drop(skb); skb_dst_set(skb, &rth->dst); return 0; } diff --git a/net/netlink/af_netlink.c b/net/netlink/af_netlink.c index 05a3795eac8e..73e9c0a9c187 100644 --- a/net/netlink/af_netlink.c +++ b/net/netlink/af_netlink.c @@ -1975,7 +1975,6 @@ static int netlink_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, copied = len; } - skb_reset_transport_header(data_skb); err = skb_copy_datagram_msg(data_skb, 0, msg, copied); if (msg->msg_name) { diff --git a/tools/testing/selftests/net/Makefile b/tools/testing/selftests/net/Makefile index 0f2ebc38d893..e1f998defd10 100644 --- a/tools/testing/selftests/net/Makefile +++ b/tools/testing/selftests/net/Makefile @@ -25,6 +25,7 @@ TEST_PROGS += bareudp.sh TEST_PROGS += amt.sh TEST_PROGS += unicast_extensions.sh TEST_PROGS += udpgro_fwd.sh +TEST_PROGS += udpgro_frglist.sh TEST_PROGS += veth.sh TEST_PROGS += ioam6.sh TEST_PROGS += gro.sh @@ -61,6 +62,8 @@ TEST_FILES := settings KSFT_KHDR_INSTALL := 1 include ../lib.mk +include bpf/Makefile + $(OUTPUT)/reuseport_bpf_numa: LDLIBS += -lnuma $(OUTPUT)/tcp_mmap: LDLIBS += -lpthread $(OUTPUT)/tcp_inq: LDLIBS += -lpthread diff --git a/tools/testing/selftests/net/bpf/Makefile b/tools/testing/selftests/net/bpf/Makefile new file mode 100644 index 000000000000..f91bf14bbee7 --- /dev/null +++ b/tools/testing/selftests/net/bpf/Makefile @@ -0,0 +1,14 @@ +# SPDX-License-Identifier: GPL-2.0 + +CLANG ?= clang +CCINCLUDE += -I../../bpf +CCINCLUDE += -I../../../../../usr/include/ + +TEST_CUSTOM_PROGS = $(OUTPUT)/bpf/nat6to4.o +all: $(TEST_CUSTOM_PROGS) + +$(OUTPUT)/%.o: %.c + $(CLANG) -O2 -target bpf -c $< $(CCINCLUDE) -o $@ + +clean: + rm -f $(TEST_CUSTOM_PROGS) diff --git a/tools/testing/selftests/net/bpf/nat6to4.c b/tools/testing/selftests/net/bpf/nat6to4.c new file mode 100644 index 000000000000..ac54c36b25fc --- /dev/null +++ b/tools/testing/selftests/net/bpf/nat6to4.c @@ -0,0 +1,285 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * This code is taken from the Android Open Source Project and the author + * (Maciej Żenczykowski) has gave permission to relicense it under the + * GPLv2. Therefore this program is free software; + * You can redistribute it and/or modify it under the terms of the GNU + * General Public License version 2 as published by the Free Software + * Foundation + + * The original headers, including the original license headers, are + * included below for completeness. + * + * Copyright (C) 2019 The Android Open Source Project + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include <linux/bpf.h> +#include <linux/if.h> +#include <linux/if_ether.h> +#include <linux/if_packet.h> +#include <linux/in.h> +#include <linux/in6.h> +#include <linux/ip.h> +#include <linux/ipv6.h> +#include <linux/pkt_cls.h> +#include <linux/swab.h> +#include <stdbool.h> +#include <stdint.h> + + +#include <linux/udp.h> + +#include <bpf/bpf_helpers.h> +#include <bpf/bpf_endian.h> + +#define IP_DF 0x4000 // Flag: "Don't Fragment" + +SEC("schedcls/ingress6/nat_6") +int sched_cls_ingress6_nat_6_prog(struct __sk_buff *skb) +{ + const int l2_header_size = sizeof(struct ethhdr); + void *data = (void *)(long)skb->data; + const void *data_end = (void *)(long)skb->data_end; + const struct ethhdr * const eth = data; // used iff is_ethernet + const struct ipv6hdr * const ip6 = (void *)(eth + 1); + + // Require ethernet dst mac address to be our unicast address. + if (skb->pkt_type != PACKET_HOST) + return TC_ACT_OK; + + // Must be meta-ethernet IPv6 frame + if (skb->protocol != bpf_htons(ETH_P_IPV6)) + return TC_ACT_OK; + + // Must have (ethernet and) ipv6 header + if (data + l2_header_size + sizeof(*ip6) > data_end) + return TC_ACT_OK; + + // Ethertype - if present - must be IPv6 + if (eth->h_proto != bpf_htons(ETH_P_IPV6)) + return TC_ACT_OK; + + // IP version must be 6 + if (ip6->version != 6) + return TC_ACT_OK; + // Maximum IPv6 payload length that can be translated to IPv4 + if (bpf_ntohs(ip6->payload_len) > 0xFFFF - sizeof(struct iphdr)) + return TC_ACT_OK; + switch (ip6->nexthdr) { + case IPPROTO_TCP: // For TCP & UDP the checksum neutrality of the chosen IPv6 + case IPPROTO_UDP: // address means there is no need to update their checksums. + case IPPROTO_GRE: // We do not need to bother looking at GRE/ESP headers, + case IPPROTO_ESP: // since there is never a checksum to update. + break; + default: // do not know how to handle anything else + return TC_ACT_OK; + } + + struct ethhdr eth2; // used iff is_ethernet + + eth2 = *eth; // Copy over the ethernet header (src/dst mac) + eth2.h_proto = bpf_htons(ETH_P_IP); // But replace the ethertype + + struct iphdr ip = { + .version = 4, // u4 + .ihl = sizeof(struct iphdr) / sizeof(__u32), // u4 + .tos = (ip6->priority << 4) + (ip6->flow_lbl[0] >> 4), // u8 + .tot_len = bpf_htons(bpf_ntohs(ip6->payload_len) + sizeof(struct iphdr)), // u16 + .id = 0, // u16 + .frag_off = bpf_htons(IP_DF), // u16 + .ttl = ip6->hop_limit, // u8 + .protocol = ip6->nexthdr, // u8 + .check = 0, // u16 + .saddr = 0x0201a8c0, // u32 + .daddr = 0x0101a8c0, // u32 + }; + + // Calculate the IPv4 one's complement checksum of the IPv4 header. + __wsum sum4 = 0; + + for (int i = 0; i < sizeof(ip) / sizeof(__u16); ++i) + sum4 += ((__u16 *)&ip)[i]; + + // Note that sum4 is guaranteed to be non-zero by virtue of ip.version == 4 + sum4 = (sum4 & 0xFFFF) + (sum4 >> 16); // collapse u32 into range 1 .. 0x1FFFE + sum4 = (sum4 & 0xFFFF) + (sum4 >> 16); // collapse any potential carry into u16 + ip.check = (__u16)~sum4; // sum4 cannot be zero, so this is never 0xFFFF + + // Calculate the *negative* IPv6 16-bit one's complement checksum of the IPv6 header. + __wsum sum6 = 0; + // We'll end up with a non-zero sum due to ip6->version == 6 (which has '0' bits) + for (int i = 0; i < sizeof(*ip6) / sizeof(__u16); ++i) + sum6 += ~((__u16 *)ip6)[i]; // note the bitwise negation + + // Note that there is no L4 checksum update: we are relying on the checksum neutrality + // of the ipv6 address chosen by netd's ClatdController. + + // Packet mutations begin - point of no return, but if this first modification fails + // the packet is probably still pristine, so let clatd handle it. + if (bpf_skb_change_proto(skb, bpf_htons(ETH_P_IP), 0)) + return TC_ACT_OK; + bpf_csum_update(skb, sum6); + + data = (void *)(long)skb->data; + data_end = (void *)(long)skb->data_end; + if (data + l2_header_size + sizeof(struct iphdr) > data_end) + return TC_ACT_SHOT; + + struct ethhdr *new_eth = data; + + // Copy over the updated ethernet header + *new_eth = eth2; + + // Copy over the new ipv4 header. + *(struct iphdr *)(new_eth + 1) = ip; + return bpf_redirect(skb->ifindex, BPF_F_INGRESS); +} + +SEC("schedcls/egress4/snat4") +int sched_cls_egress4_snat4_prog(struct __sk_buff *skb) +{ + const int l2_header_size = sizeof(struct ethhdr); + void *data = (void *)(long)skb->data; + const void *data_end = (void *)(long)skb->data_end; + const struct ethhdr *const eth = data; // used iff is_ethernet + const struct iphdr *const ip4 = (void *)(eth + 1); + + // Must be meta-ethernet IPv4 frame + if (skb->protocol != bpf_htons(ETH_P_IP)) + return TC_ACT_OK; + + // Must have ipv4 header + if (data + l2_header_size + sizeof(struct ipv6hdr) > data_end) + return TC_ACT_OK; + + // Ethertype - if present - must be IPv4 + if (eth->h_proto != bpf_htons(ETH_P_IP)) + return TC_ACT_OK; + + // IP version must be 4 + if (ip4->version != 4) + return TC_ACT_OK; + + // We cannot handle IP options, just standard 20 byte == 5 dword minimal IPv4 header + if (ip4->ihl != 5) + return TC_ACT_OK; + + // Maximum IPv6 payload length that can be translated to IPv4 + if (bpf_htons(ip4->tot_len) > 0xFFFF - sizeof(struct ipv6hdr)) + return TC_ACT_OK; + + // Calculate the IPv4 one's complement checksum of the IPv4 header. + __wsum sum4 = 0; + + for (int i = 0; i < sizeof(*ip4) / sizeof(__u16); ++i) + sum4 += ((__u16 *)ip4)[i]; + + // Note that sum4 is guaranteed to be non-zero by virtue of ip4->version == 4 + sum4 = (sum4 & 0xFFFF) + (sum4 >> 16); // collapse u32 into range 1 .. 0x1FFFE + sum4 = (sum4 & 0xFFFF) + (sum4 >> 16); // collapse any potential carry into u16 + // for a correct checksum we should get *a* zero, but sum4 must be positive, ie 0xFFFF + if (sum4 != 0xFFFF) + return TC_ACT_OK; + + // Minimum IPv4 total length is the size of the header + if (bpf_ntohs(ip4->tot_len) < sizeof(*ip4)) + return TC_ACT_OK; + + // We are incapable of dealing with IPv4 fragments + if (ip4->frag_off & ~bpf_htons(IP_DF)) + return TC_ACT_OK; + + switch (ip4->protocol) { + case IPPROTO_TCP: // For TCP & UDP the checksum neutrality of the chosen IPv6 + case IPPROTO_GRE: // address means there is no need to update their checksums. + case IPPROTO_ESP: // We do not need to bother looking at GRE/ESP headers, + break; // since there is never a checksum to update. + + case IPPROTO_UDP: // See above comment, but must also have UDP header... + if (data + sizeof(*ip4) + sizeof(struct udphdr) > data_end) + return TC_ACT_OK; + const struct udphdr *uh = (const struct udphdr *)(ip4 + 1); + // If IPv4/UDP checksum is 0 then fallback to clatd so it can calculate the + // checksum. Otherwise the network or more likely the NAT64 gateway might + // drop the packet because in most cases IPv6/UDP packets with a zero checksum + // are invalid. See RFC 6935. TODO: calculate checksum via bpf_csum_diff() + if (!uh->check) + return TC_ACT_OK; + break; + + default: // do not know how to handle anything else + return TC_ACT_OK; + } + struct ethhdr eth2; // used iff is_ethernet + + eth2 = *eth; // Copy over the ethernet header (src/dst mac) + eth2.h_proto = bpf_htons(ETH_P_IPV6); // But replace the ethertype + + struct ipv6hdr ip6 = { + .version = 6, // __u8:4 + .priority = ip4->tos >> 4, // __u8:4 + .flow_lbl = {(ip4->tos & 0xF) << 4, 0, 0}, // __u8[3] + .payload_len = bpf_htons(bpf_ntohs(ip4->tot_len) - 20), // __be16 + .nexthdr = ip4->protocol, // __u8 + .hop_limit = ip4->ttl, // __u8 + }; + ip6.saddr.in6_u.u6_addr32[0] = bpf_htonl(0x20010db8); + ip6.saddr.in6_u.u6_addr32[1] = 0; + ip6.saddr.in6_u.u6_addr32[2] = 0; + ip6.saddr.in6_u.u6_addr32[3] = bpf_htonl(1); + ip6.daddr.in6_u.u6_addr32[0] = bpf_htonl(0x20010db8); + ip6.daddr.in6_u.u6_addr32[1] = 0; + ip6.daddr.in6_u.u6_addr32[2] = 0; + ip6.daddr.in6_u.u6_addr32[3] = bpf_htonl(2); + + // Calculate the IPv6 16-bit one's complement checksum of the IPv6 header. + __wsum sum6 = 0; + // We'll end up with a non-zero sum due to ip6.version == 6 + for (int i = 0; i < sizeof(ip6) / sizeof(__u16); ++i) + sum6 += ((__u16 *)&ip6)[i]; + + // Packet mutations begin - point of no return, but if this first modification fails + // the packet is probably still pristine, so let clatd handle it. + if (bpf_skb_change_proto(skb, bpf_htons(ETH_P_IPV6), 0)) + return TC_ACT_OK; + + // This takes care of updating the skb->csum field for a CHECKSUM_COMPLETE packet. + // In such a case, skb->csum is a 16-bit one's complement sum of the entire payload, + // thus we need to subtract out the ipv4 header's sum, and add in the ipv6 header's sum. + // However, we've already verified the ipv4 checksum is correct and thus 0. + // Thus we only need to add the ipv6 header's sum. + // + // bpf_csum_update() always succeeds if the skb is CHECKSUM_COMPLETE and returns an error + // (-ENOTSUPP) if it isn't. So we just ignore the return code (see above for more details). + bpf_csum_update(skb, sum6); + + // bpf_skb_change_proto() invalidates all pointers - reload them. + data = (void *)(long)skb->data; + data_end = (void *)(long)skb->data_end; + + // I cannot think of any valid way for this error condition to trigger, however I do + // believe the explicit check is required to keep the in kernel ebpf verifier happy. + if (data + l2_header_size + sizeof(ip6) > data_end) + return TC_ACT_SHOT; + + struct ethhdr *new_eth = data; + + // Copy over the updated ethernet header + *new_eth = eth2; + // Copy over the new ipv4 header. + *(struct ipv6hdr *)(new_eth + 1) = ip6; + return TC_ACT_OK; +} + +char _license[] SEC("license") = ("GPL"); diff --git a/tools/testing/selftests/net/udpgro_frglist.sh b/tools/testing/selftests/net/udpgro_frglist.sh new file mode 100755 index 000000000000..807b74c8fd80 --- /dev/null +++ b/tools/testing/selftests/net/udpgro_frglist.sh @@ -0,0 +1,101 @@ +#!/bin/bash +# SPDX-License-Identifier: GPL-2.0 +# +# Run a series of udpgro benchmarks + +readonly PEER_NS="ns-peer-$(mktemp -u XXXXXX)" + +cleanup() { + local -r jobs="$(jobs -p)" + local -r ns="$(ip netns list|grep $PEER_NS)" + + [ -n "${jobs}" ] && kill -INT ${jobs} 2>/dev/null + [ -n "$ns" ] && ip netns del $ns 2>/dev/null +} +trap cleanup EXIT + +run_one() { + # use 'rx' as separator between sender args and receiver args + local -r all="$@" + local -r tx_args=${all%rx*} + local rx_args=${all#*rx} + + + + ip netns add "${PEER_NS}" + ip -netns "${PEER_NS}" link set lo up + ip link add type veth + ip link set dev veth0 up + ip addr add dev veth0 192.168.1.2/24 + ip addr add dev veth0 2001:db8::2/64 nodad + + ip link set dev veth1 netns "${PEER_NS}" + ip -netns "${PEER_NS}" addr add dev veth1 192.168.1.1/24 + ip -netns "${PEER_NS}" addr add dev veth1 2001:db8::1/64 nodad + ip -netns "${PEER_NS}" link set dev veth1 up + ip netns exec "${PEER_NS}" ethtool -K veth1 rx-gro-list on + + + ip -n "${PEER_NS}" link set veth1 xdp object ../bpf/xdp_dummy.o section xdp_dummy + tc -n "${PEER_NS}" qdisc add dev veth1 clsact + tc -n "${PEER_NS}" filter add dev veth1 ingress prio 4 protocol ipv6 bpf object-file ../bpf/nat6to4.o section schedcls/ingress6/nat_6 direct-action + tc -n "${PEER_NS}" filter add dev veth1 egress prio 4 protocol ip bpf object-file ../bpf/nat6to4.o section schedcls/egress4/snat4 direct-action + echo ${rx_args} + ip netns exec "${PEER_NS}" ./udpgso_bench_rx ${rx_args} -r & + + # Hack: let bg programs complete the startup + sleep 0.1 + ./udpgso_bench_tx ${tx_args} +} + +run_in_netns() { + local -r args=$@ + echo ${args} + ./in_netns.sh $0 __subprocess ${args} +} + +run_udp() { + local -r args=$@ + + echo "udp gso - over veth touching data" + run_in_netns ${args} -u -S 0 rx -4 -v + + echo "udp gso and gro - over veth touching data" + run_in_netns ${args} -S 0 rx -4 -G +} + +run_tcp() { + local -r args=$@ + + echo "tcp - over veth touching data" + run_in_netns ${args} -t rx -4 -t +} + +run_all() { + local -r core_args="-l 4" + local -r ipv4_args="${core_args} -4 -D 192.168.1.1" + local -r ipv6_args="${core_args} -6 -D 2001:db8::1" + + echo "ipv6" + run_tcp "${ipv6_args}" + run_udp "${ipv6_args}" +} + +if [ ! -f ../bpf/xdp_dummy.o ]; then + echo "Missing xdp_dummy helper. Build bpf selftest first" + exit -1 +fi + +if [ ! -f bpf/nat6to4.o ]; then + echo "Missing nat6to4 helper. Build bpfnat6to4.o selftest first" + exit -1 +fi + +if [[ $# -eq 0 ]]; then + run_all +elif [[ $1 == "__subprocess" ]]; then + shift + run_one $@ +else + run_in_netns $@ +fi |