Pull networking fixes from Jakub Kicinski: "Cross-tree/merge window issues: - rtl8150: don't incorrectly assign random MAC addresses; fix late in the 5.9 cycle started depending on a return code from a function which changed with the 5.10 PR from the usb subsystem Current release regressions: - Revert "virtio-net: ethtool configurable RXCSUM", it was causing crashes at probe when control vq was not negotiated/available Previous release regressions: - ixgbe: fix probing of multi-port 10 Gigabit Intel NICs with an MDIO bus, only first device would be probed correctly - nexthop: Fix performance regression in nexthop deletion by effectively switching from recently added synchronize_rcu() to synchronize_rcu_expedited() - netsec: ignore 'phy-mode' device property on ACPI systems; the property is not populated correctly by the firmware, but firmware configures the PHY so just keep boot settings Previous releases - always broken: - tcp: fix to update snd_wl1 in bulk receiver fast path, addressing bulk transfers getting "stuck" - icmp: randomize the global rate limiter to prevent attackers from getting useful signal - r8169: fix operation under forced interrupt threading, make the driver always use hard irqs, even on RT, given the handler is light and only wants to schedule napi (and do so through a _irqoff() variant, preferably) - bpf: Enforce pointer id generation for all may-be-null register type to avoid pointers erroneously getting marked as null-checked - tipc: re-configure queue limit for broadcast link - net/sched: act_tunnel_key: fix OOB write in case of IPv6 ERSPAN tunnels - fix various issues in chelsio inline tls driver Misc: - bpf: improve just-added bpf_redirect_neigh() helper api to support supplying nexthop by the caller - in case BPF program has already done a lookup we can avoid doing another one - remove unnecessary break statements - make MCTCP not select IPV6, but rather depend on it" * tag 'net-5.10-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (62 commits) tcp: fix to update snd_wl1 in bulk receiver fast path net: Properly typecast int values to set sk_max_pacing_rate netfilter: nf_fwd_netdev: clear timestamp in forwarding path ibmvnic: save changed mac address to adapter->mac_addr selftests: mptcp: depends on built-in IPv6 Revert "virtio-net: ethtool configurable RXCSUM" rtnetlink: fix data overflow in rtnl_calcit() net: ethernet: mtk-star-emac: select REGMAP_MMIO net: hdlc_raw_eth: Clear the IFF_TX_SKB_SHARING flag after calling ether_setup net: hdlc: In hdlc_rcv, check to make sure dev is an HDLC device bpf, libbpf: Guard bpf inline asm from bpf_tail_call_static bpf, selftests: Extend test_tc_redirect to use modified bpf_redirect_neigh() bpf: Fix bpf_redirect_neigh helper api to support supplying nexthop mptcp: depends on IPV6 but not as a module sfc: move initialisation of efx->filter_sem to efx_init_struct() mpls: load mpls_gso after mpls_iptunnel net/sched: act_tunnel_key: fix OOB write in case of IPv6 ERSPAN tunnels net/sched: act_gate: Unlock ->tcfa_lock in tc_setup_flow_action() net: dsa: bcm_sf2: make const array static, makes object smaller mptcp: MPTCP_IPV6 should depend on IPV6 instead of selecting it ...
Showing the changes to kernel (2 of 79 files). Show all
Unified · Split
| @@ -2913,7 +2913,6 @@ attach_type_to_prog_type(enum bpf_attach_type attach_type) · 2912 unmodified lines | |||
| 2913 | case BPF_CGROUP_INET_INGRESS: | 2913 | case BPF_CGROUP_INET_INGRESS: |
| 2914 | case BPF_CGROUP_INET_EGRESS: | 2914 | case BPF_CGROUP_INET_EGRESS: |
| 2915 | return BPF_PROG_TYPE_CGROUP_SKB; | 2915 | return BPF_PROG_TYPE_CGROUP_SKB; |
| 2916 | break; | ||
| 2917 | case BPF_CGROUP_INET_SOCK_CREATE: | 2916 | case BPF_CGROUP_INET_SOCK_CREATE: |
| 2918 | case BPF_CGROUP_INET_SOCK_RELEASE: | 2917 | case BPF_CGROUP_INET_SOCK_RELEASE: |
| 2919 | case BPF_CGROUP_INET4_POST_BIND: | 2918 | case BPF_CGROUP_INET4_POST_BIND: |
| @@ -5133,24 +5133,19 @@ static int check_helper_call(struct bpf_verifier_env *env, int func_id, int insn · 5132 unmodified lines | |||
| 5133 | regs[BPF_REG_0].id = ++env->id_gen; | 5133 | regs[BPF_REG_0].id = ++env->id_gen; |
| 5134 | } else { | 5134 | } else { |
| 5135 | regs[BPF_REG_0].type = PTR_TO_MAP_VALUE_OR_NULL; | 5135 | regs[BPF_REG_0].type = PTR_TO_MAP_VALUE_OR_NULL; |
| 5136 | regs[BPF_REG_0].id = ++env->id_gen; | ||
| 5137 | } | 5136 | } |
| 5138 | } else if (fn->ret_type == RET_PTR_TO_SOCKET_OR_NULL) { | 5137 | } else if (fn->ret_type == RET_PTR_TO_SOCKET_OR_NULL) { |
| 5139 | mark_reg_known_zero(env, regs, BPF_REG_0); | 5138 | mark_reg_known_zero(env, regs, BPF_REG_0); |
| 5140 | regs[BPF_REG_0].type = PTR_TO_SOCKET_OR_NULL; | 5139 | regs[BPF_REG_0].type = PTR_TO_SOCKET_OR_NULL; |
| 5141 | regs[BPF_REG_0].id = ++env->id_gen; | ||
| 5142 | } else if (fn->ret_type == RET_PTR_TO_SOCK_COMMON_OR_NULL) { | 5140 | } else if (fn->ret_type == RET_PTR_TO_SOCK_COMMON_OR_NULL) { |
| 5143 | mark_reg_known_zero(env, regs, BPF_REG_0); | 5141 | mark_reg_known_zero(env, regs, BPF_REG_0); |
| 5144 | regs[BPF_REG_0].type = PTR_TO_SOCK_COMMON_OR_NULL; | 5142 | regs[BPF_REG_0].type = PTR_TO_SOCK_COMMON_OR_NULL; |
| 5145 | regs[BPF_REG_0].id = ++env->id_gen; | ||
| 5146 | } else if (fn->ret_type == RET_PTR_TO_TCP_SOCK_OR_NULL) { | 5143 | } else if (fn->ret_type == RET_PTR_TO_TCP_SOCK_OR_NULL) { |
| 5147 | mark_reg_known_zero(env, regs, BPF_REG_0); | 5144 | mark_reg_known_zero(env, regs, BPF_REG_0); |
| 5148 | regs[BPF_REG_0].type = PTR_TO_TCP_SOCK_OR_NULL; | 5145 | regs[BPF_REG_0].type = PTR_TO_TCP_SOCK_OR_NULL; |
| 5149 | regs[BPF_REG_0].id = ++env->id_gen; | ||
| 5150 | } else if (fn->ret_type == RET_PTR_TO_ALLOC_MEM_OR_NULL) { | 5146 | } else if (fn->ret_type == RET_PTR_TO_ALLOC_MEM_OR_NULL) { |
| 5151 | mark_reg_known_zero(env, regs, BPF_REG_0); | 5147 | mark_reg_known_zero(env, regs, BPF_REG_0); |
| 5152 | regs[BPF_REG_0].type = PTR_TO_MEM_OR_NULL; | 5148 | regs[BPF_REG_0].type = PTR_TO_MEM_OR_NULL; |
| 5153 | regs[BPF_REG_0].id = ++env->id_gen; | ||
| 5154 | regs[BPF_REG_0].mem_size = meta.mem_size; | 5149 | regs[BPF_REG_0].mem_size = meta.mem_size; |
| 5155 | } else if (fn->ret_type == RET_PTR_TO_MEM_OR_BTF_ID_OR_NULL || | 5150 | } else if (fn->ret_type == RET_PTR_TO_MEM_OR_BTF_ID_OR_NULL || |
| 5156 | fn->ret_type == RET_PTR_TO_MEM_OR_BTF_ID) { | 5151 | fn->ret_type == RET_PTR_TO_MEM_OR_BTF_ID) { |
| @@ -5199,6 +5194,9 @@ static int check_helper_call(struct bpf_verifier_env *env, int func_id, int insn · 42 unmodified lines | |||
| 5199 | return -EINVAL; | 5194 | return -EINVAL; |
| 5200 | } | 5195 | } |
| 5201 | 5196 | ||
| 5197 | if (reg_type_may_be_null(regs[BPF_REG_0].type)) | ||
| 5198 | regs[BPF_REG_0].id = ++env->id_gen; | ||
| 5199 | |||
| 5202 | if (is_ptr_cast_function(func_id)) { | 5200 | if (is_ptr_cast_function(func_id)) { |
| 5203 | /* For release_reference() */ | 5201 | /* For release_reference() */ |
| 5204 | regs[BPF_REG_0].ref_obj_id = meta.ref_obj_id; | 5202 | regs[BPF_REG_0].ref_obj_id = meta.ref_obj_id; |
| @@ -7212,7 +7210,8 @@ static void mark_ptr_or_null_reg(struct bpf_func_state *state, · 2007 unmodified lines | |||
| 7212 | struct bpf_reg_state *reg, u32 id, | 7210 | struct bpf_reg_state *reg, u32 id, |
| 7213 | bool is_null) | 7211 | bool is_null) |
| 7214 | { | 7212 | { |
| 7215 | if (reg_type_may_be_null(reg->type) && reg->id == id) { | 7213 | if (reg_type_may_be_null(reg->type) && reg->id == id && |
| 7214 | !WARN_ON_ONCE(!reg->id)) { | ||
| 7216 | /* Old offset (both fixed and variable parts) should | 7215 | /* Old offset (both fixed and variable parts) should |
| 7217 | * have been known-zero, because we don't allow pointer | 7216 | * have been known-zero, because we don't allow pointer |
| 7218 | * arithmetic on pointers that might be NULL. | 7217 | * arithmetic on pointers that might be NULL. |