- 26 Oct, 2021 34 commits
-
-
Eric Dumazet authored
Aligning @size argument to 4 bytes is not needed. The header alignment has nothing to do with @size. It really depends on skb->head alignment and MAX_TCP_HEADER. Signed-off-by:
Eric Dumazet <edumazet@google.com> Acked-by:
Soheil Hassas Yeganeh <soheil@google.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Eric Dumazet authored
Both IPv4 and IPv6 uses same reserve, no need risking cache line misses to fetch its value. Signed-off-by:
-
Eric Dumazet authored
sk_stream_alloc_skb() is only used by TCP. Rename it to make this clear, and move its declaration to include/net/tcp.h Signed-off-by:
-
Eric Dumazet authored
neigh_output() reads n->nud_state and hh->hh_len locklessly. This is fine, but we need to add annotations and document this. We evaluate skip_cache first to avoid reading these fields if the cache has to by bypassed. syzbot report: BUG: KCSAN: data-race in __neigh_event_send / ip_finish_output2 write to 0xffff88810798a885 of 1 bytes by interrupt on cpu 1: __neigh_event_send+0x40d/0xac0 net/core/neighbour.c:1128 neigh_event_send include/net/neighbour.h:444 [inline] neigh_resolve_output+0x104/0x410 net/core/neighbour.c:1476 neigh_output include/net/neighbour.h:510 [inline] ip_finish_output2+0x80a/0xaa0 net/ipv4/ip_output.c:221 ip_finish_output+0x3b5/0x510 net/ipv4/ip_output.c:309 NF_HOOK_COND include/linux/netfilter.h:296 [inline] ip_output+0xf3/0x1a0 net/ipv4/ip_output.c:423 dst_output include/net/dst.h:450 [inline] ip_local_out+0x164/0x220 net/ipv4/ip_output.c:126 __ip_queue_xmit+0x9d3/0xa20 net/ipv4/ip_output.c:525 ip_queue_xmit+0x34/0x40 net/ipv4/ip_output.c:539 __tcp_transmit_skb+0x142a/0x1a00 net/ipv4/tcp_output.c:1405 tcp_transmit_skb net/ipv4/tcp_output.c:1423 [inline] tcp_xmit_probe_skb net/ipv4/tcp_output.c:4011 [inline] tcp_write_wakeup+0x4a9/0x810 net/ipv4/tcp_output.c:4064 tcp_send_probe0+0x2c/0x2b0 net/ipv4/tcp_output.c:4079 tcp_probe_timer net/ipv4/tcp_timer.c:398 [inline] tcp_write_timer_handler+0x394/0x520 net/ipv4/tcp_timer.c:626 tcp_write_timer+0xb9/0x180 net/ipv4/tcp_timer.c:642 call_timer_fn+0x2e/0x1d0 kernel/time/timer.c:1421 expire_timers+0x135/0x240 kernel/time/timer.c:1466 __run_timers+0x368/0x430 kernel/time/timer.c:1734 run_timer_softirq+0x19/0x30 kernel/time/timer.c:1747 __do_softirq+0x12c/0x26e kernel/softirq.c:558 invoke_softirq kernel/softirq.c:432 [inline] __irq_exit_rcu kernel/softirq.c:636 [inline] irq_exit_rcu+0x4e/0xa0 kernel/softirq.c:648 sysvec_apic_timer_interrupt+0x69/0x80 arch/x86/kernel/apic/apic.c:1097 asm_sysvec_apic_timer_interrupt+0x12/0x20 native_safe_halt arch/x86/include/asm/irqflags.h:51 [inline] arch_safe_halt arch/x86/include/asm/irqflags.h:89 [inline] acpi_safe_halt drivers/acpi/processor_idle.c:109 [inline] acpi_idle_do_entry drivers/acpi/processor_idle.c:553 [inline] acpi_idle_enter+0x258/0x2e0 drivers/acpi/processor_idle.c:688 cpuidle_enter_state+0x2b4/0x760 drivers/cpuidle/cpuidle.c:237 cpuidle_enter+0x3c/0x60 drivers/cpuidle/cpuidle.c:351 call_cpuidle kernel/sched/idle.c:158 [inline] cpuidle_idle_call kernel/sched/idle.c:239 [inline] do_idle+0x1a3/0x250 kernel/sched/idle.c:306 cpu_startup_entry+0x15/0x20 kernel/sched/idle.c:403 secondary_startup_64_no_verify+0xb1/0xbb read to 0xffff88810798a885 of 1 bytes by interrupt on cpu 0: neigh_output include/net/neighbour.h:507 [inline] ip_finish_output2+0x79a/0xaa0 net/ipv4/ip_output.c:221 ip_finish_output+0x3b5/0x510 net/ipv4/ip_output.c:309 NF_HOOK_COND include/linux/netfilter.h:296 [inline] ip_output+0xf3/0x1a0 net/ipv4/ip_output.c:423 dst_output include/net/dst.h:450 [inline] ip_local_out+0x164/0x220 net/ipv4/ip_output.c:126 __ip_queue_xmit+0x9d3/0xa20 net/ipv4/ip_output.c:525 ip_queue_xmit+0x34/0x40 net/ipv4/ip_output.c:539 __tcp_transmit_skb+0x142a/0x1a00 net/ipv4/tcp_output.c:1405 tcp_transmit_skb net/ipv4/tcp_output.c:1423 [inline] tcp_xmit_probe_skb net/ipv4/tcp_output.c:4011 [inline] tcp_write_wakeup+0x4a9/0x810 net/ipv4/tcp_output.c:4064 tcp_send_probe0+0x2c/0x2b0 net/ipv4/tcp_output.c:4079 tcp_probe_timer net/ipv4/tcp_timer.c:398 [inline] tcp_write_timer_handler+0x394/0x520 net/ipv4/tcp_timer.c:626 tcp_write_timer+0xb9/0x180 net/ipv4/tcp_timer.c:642 call_timer_fn+0x2e/0x1d0 kernel/time/timer.c:1421 expire_timers+0x135/0x240 kernel/time/timer.c:1466 __run_timers+0x368/0x430 kernel/time/timer.c:1734 run_timer_softirq+0x19/0x30 kernel/time/timer.c:1747 __do_softirq+0x12c/0x26e kernel/softirq.c:558 invoke_softirq kernel/softirq.c:432 [inline] __irq_exit_rcu kernel/softirq.c:636 [inline] irq_exit_rcu+0x4e/0xa0 kernel/softirq.c:648 sysvec_apic_timer_interrupt+0x69/0x80 arch/x86/kernel/apic/apic.c:1097 asm_sysvec_apic_timer_interrupt+0x12/0x20 native_safe_halt arch/x86/include/asm/irqflags.h:51 [inline] arch_safe_halt arch/x86/include/asm/irqflags.h:89 [inline] acpi_safe_halt drivers/acpi/processor_idle.c:109 [inline] acpi_idle_do_entry drivers/acpi/processor_idle.c:553 [inline] acpi_idle_enter+0x258/0x2e0 drivers/acpi/processor_idle.c:688 cpuidle_enter_state+0x2b4/0x760 drivers/cpuidle/cpuidle.c:237 cpuidle_enter+0x3c/0x60 drivers/cpuidle/cpuidle.c:351 call_cpuidle kernel/sched/idle.c:158 [inline] cpuidle_idle_call kernel/sched/idle.c:239 [inline] do_idle+0x1a3/0x250 kernel/sched/idle.c:306 cpu_startup_entry+0x15/0x20 kernel/sched/idle.c:403 rest_init+0xee/0x100 init/main.c:734 arch_call_rest_init+0xa/0xb start_kernel+0x5e4/0x669 init/main.c:1142 secondary_startup_64_no_verify+0xb1/0xbb value changed: 0x20 -> 0x01 Reported by Kernel Concurrency Sanitizer on: CPU: 0 PID: 0 Comm: swapper/0 Not tainted 5.15.0-rc6-syzkaller #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011 Signed-off-by:
syzbot <syzkaller@googlegroups.com> Signed-off-by:
-
David S. Miller authored
Ido Schimmel says: ==================== mlxsw: Support multiple RIF MAC prefixes Currently, mlxsw enforces that all the netdevs used as router interfaces (RIFs) have the same MAC prefix (e.g., same 38 MSBs in Spectrum-1). Otherwise, an error is returned to user space with extack. This patchset relaxes the limitation through the use of RIF MAC profiles. A RIF MAC profile is a hardware entity that represents a particular MAC prefix which multiple RIFs can reference. Therefore, the number of possible MAC prefixes is no longer one, but the number of profiles supported by the device. The ability to change the MAC of a particular netdev is useful, for example, for users who use the netdev to connect to an upstream provider that performs MAC filtering. Currently, such users are either forced to negotiate with the provider or change the MAC address of all other netdevs so that they share the same prefix. Patchset overview: Patches #1-#3 are preparations. Patch #4 adds actual support for RIF MAC profiles. Patch #5 exposes RIF MAC profiles as a devlink resource, so that user space has visibility into the maximum number of profiles and current occupancy. Useful for debugging and testing (next 3 patches). Patches #6-#8 add both scale and functional tests. Patch #9 removes tests that validated the previous limitation. It is now covered by patch #6 for devices that support a single profile. ==================== Signed-off-by: -
Danielle Ratson authored
After adding the previous patches, the constraint that all the router interface MAC addresses have the same prefix is no longer relevant. Remove the test cases that validated that this constraint is honored. Signed-off-by:
Danielle Ratson <danieller@nvidia.com> Signed-off-by:
Ido Schimmel <idosch@nvidia.com> Signed-off-by:
-
Danielle Ratson authored
When all the RIF MAC profiles are in use, test that it is possible to change the MAC of a netdev (i.e., a RIF) when its MAC profile is not shared with other RIFs. Test that replacement fails when the MAC profile is shared. Signed-off-by:
-
Danielle Ratson authored
Verify that MAC profile changes are indeed applied and that packets are forwarded with the correct source MAC. Output example: $ ./rif_mac_profiles.sh TEST: h1->h2: new mac profile [ OK ] TEST: h2->h1: new mac profile [ OK ] TEST: h1->h2: edit mac profile [ OK ] TEST: h2->h1: edit mac profile [ OK ] Signed-off-by:
-
Danielle Ratson authored
Query the maximum number of supported RIF MAC profiles using devlink-resource and verify that all available MAC profiles can be utilized and that an error is generated when user space tries to exceed this number. Output example in Spectrum-2: $ TESTS='rif_mac_profile' ./resource_scale.sh TEST: 'rif_mac_profile' 4 [ OK ] TEST: 'rif_mac_profile' overflow 5 [ OK ] Signed-off-by:
-
Danielle Ratson authored
Expose via devlink-resource the maximum number of RIF MAC profiles and their current occupancy, so it can be used for debug and writing generic tests, like in the next patch. Example for Spectrum-2 output: $ devlink resource show pci/0000:06:00.0 ... name rif_mac_profiles size 4 occ 0 unit entry dpipe_tables none Signed-off-by:
-
Danielle Ratson authored
Currently, mlxsw enforces that all the router interfaces (RIFs) have the same MAC prefix. Relax this limitation by using RIF MAC profiles. Each profile is associated with a particular MAC prefix and multiple RIFs can use the same profile. Therefore, the number of possible MAC prefixes is no longer one, but the number of profiles supported by the device. Store the profiles in an IDR and reference count them according to the number of RIFs using them. Associate a RIF with a profile when the RIF is created and remove the association when the RIF is deleted. Change the association following 'NETDEV_CHANGEADDR' events, except when only one RIF is using the profile. In which case, change the MAC prefix of the profile itself instead of associating the RIF with a new profile. Signed-off-by:
-
Danielle Ratson authored
The next patch will set the MAC profile of a router interface (RIF) as part of its configure() callback. The operation can fail in case the maximum number of profiles was exceeded. Add extack to mlxsw_sp_rif_ops::configure() in order to communicate such failures to user space. In addition, the MAC profile of a RIF can change following a 'NETDEV_CHANGEADDR' notification. Propagate extack to mlxsw_sp_router_port_change_event() so that failures could be communicated in this path as well. No functional changes intended. Signed-off-by:
-
Danielle Ratson authored
Add a resource identifier for maximum RIF MAC profiles so that it could be later used to query the information from firmware. Signed-off-by:
-
Danielle Ratson authored
Add MAC profile ID field to RITR register so that it could be used for associating a RIF with a MAC profile ID by a later patch. Signed-off-by:
-
David S. Miller authored
Florian Westphal says: ==================== vrf: rework interaction with netfilter/conntrack V2: - fix 'plain integer as null pointer' warning - reword commit message in patch 2 to clarify loss of 'ct set untracked' This patch series aims to solve the to-be-reverted change 09e856d5 ("vrf: Reset skb conntrack connection on VRF rcv") in a different way. Rather than have skbs pass through conntrack and nat hooks twice, suppress conntrack invocation if the conntrack/nat hook is called from the vrf driver. First patch deals with 'incoming connection' case: 1. suppress NAT transformations 2. skip conntrack confirmation NAT and conntrack confirmation is done when ip/ipv6 stack calls the postrouting hook. Second patch deals with local packets: in vrf driver, mark the skbs as 'untracked', so conntrack output hook ignores them. This skips all nat hooks as well. Afterwards, remove the untracked state again so the second round will pick them up. One alternative to the chosen implementation would be to add a 'caller id' field to 'struct nf_hook_state' and then use that, these patches use the more straightforward check of VRF flag on the state->out device. The two patches apply to both net and net-next, i am targeting -next because I think that since snat did not work correctly for so long that we can take the longer route. If you disagree, apply to net at your discretion. The patches apply both with 09e856d5 reverted or still in-place, but only with the revert in place ingress conntrack settings (zone, notrack etc) start working again. I've already submitted selftests for vrf+nfqueue and conntrack+vrf. ==================== Signed-off-by:
-
Florian Westphal authored
The VRF driver invokes netfilter for output+postrouting hooks so that users can create rules that check for 'oif $vrf' rather than lower device name. This is a problem when NAT rules are configured. To avoid any conntrack involvement in round 1, tag skbs as 'untracked' to prevent conntrack from picking them up. This gets cleared before the packet gets handed to the ip stack so conntrack will be active on the second iteration. One remaining issue is that a rule like output ... oif $vrfname notrack won't propagate to the second round because we can't tell 'notrack set via ruleset' and 'notrack set by vrf driver' apart. However, this isn't a regression: the 'notrack' removal happens instead of unconditional nf_reset_ct(). I'd also like to avoid leaking more vrf specific conditionals into the netfilter infra. For ingress, conntrack has already been done before the packet makes it to the vrf driver, with this patch egress does connection tracking with lower/physical device as well. Signed-off-by:
Florian Westphal <fw@strlen.de> Acked-by:
David Ahern <dsahern@kernel.org> Signed-off-by:
-
Florian Westphal authored
The VRF driver invokes netfilter for output+postrouting hooks so that users can create rules that check for 'oif $vrf' rather than lower device name. Afterwards, ip stack calls those hooks again. This is a problem when conntrack is used with IP masquerading. masquerading has an internal check that re-validates the output interface to account for route changes. This check will trigger in the vrf case. If the -j MASQUERADE rule matched on the first iteration, then round 2 finds state->out->ifindex != nat->masq_index: the latter is the vrf index, but out->ifindex is the lower device. The packet gets dropped and the conntrack entry is invalidated. This change makes conntrack postrouting skip the nat hooks. Also skip confirmation. This allows the second round (postrouting invocation from ipv4/ipv6) to create nat bindings. This also prevents the second round from seeing packets that had their source address changed by the nat hook. Signed-off-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linuxDavid S. Miller authored
Saeed Mahameed says: ==================== mlx5-updates-2021-10-25 Misc updates for mlx5 driver: 1) Misc updates and cleanups: - Don't write directly to netdev->dev_addr, From Jakub Kicinski - Remove unnecessary checks for slow path flag in tc module - Fix unused function warning of mlx5i_flow_type_mask - Bridge, support replacing existing FDB entry 2) Sub Functions, Reduction in memory usage: - Reduce flow counters bulk query buffer size - Implement max_macs devlink parameter - Add devlink vendor params to control Event Queue sizes - Added SF life cycle trace points by Parav/ 3) From Aya, Firmware health buffer reporting improvements - Print health buffer by log level and more missing information - Periodic update of host time to firmware ==================== Signed-off-by: -
Jon Maxwell authored
v1: Implement a more general statement as recommended by Eric Dumazet. The sequence number will be advanced, so this check will fix the FIN case and other cases. A customer reported sockets stuck in the CLOSING state. A Vmcore revealed that the write_queue was not empty as determined by tcp_write_queue_empty() but the sk_buff containing the FIN flag had been freed and the socket was zombied in that state. Corresponding pcaps show no FIN from the Linux kernel on the wire. Some instrumentation was added to the kernel and it was found that there is a timing window where tcp_sendmsg() can run after tcp_send_fin(). tcp_sendmsg() will hit an error, for example: 1269 ▹ if (sk->sk_err || (sk->sk_shutdown & SEND_SHUTDOWN))
↩ 1270 ▹ ▹ goto do_error;↩ tcp_remove_empty_skb() will then free the FIN sk_buff as "skb->len == 0". The TCP socket is now wedged in the FIN-WAIT-1 state because the FIN is never sent. If the other side sends a FIN packet the socket will transition to CLOSING and remain that way until the system is rebooted. Fix this by checking for the FIN flag in the sk_buff and don't free it if that is the case. Testing confirmed that fixed the issue. Fixes: fdfc5c85 ("tcp: remove empty skb from write queue in error cases") Signed-off-by:Jon Maxwell <jmaxwell37@gmail.com> Reported-by:
Monir Zouaoui <Monir.Zouaoui@mail.schwarz> Reported-by:
Simon Stier <simon.stier@mail.schwarz> Reviewed-by:
-
Jakub Kicinski authored
Jean Sacren says: ==================== Small fixes for true expression checks This series fixes checks of true !rc expression. ==================== Link: https://lore.kernel.org/r/cover.1634974124.git.sakiwit@gmail.comSigned-off-by:
Jakub Kicinski <kuba@kernel.org>
-
Jean Sacren authored
Remove the check of !rc in (!rc && !resc_lock_params.b_granted) since it is always true. Signed-off-by:
Jean Sacren <sakiwit@gmail.com> Signed-off-by:
-
Jean Sacren authored
Remove the check of !rc in (!rc && !params.b_granted) since it is always true. We should also use constant 0 for return. Signed-off-by:
-
Jakub Kicinski authored
Eric Dumazet says: ==================== tcp: receive path optimizations This series aims to reduce cache line misses in RX path. I am still working on better cache locality in tcp_sock but this will wait few more weeks. ==================== Link: https://lore.kernel.org/r/20211025164825.259415-1-eric.dumazet@gmail.comSigned-off-by:
-
Eric Dumazet authored
Two kfree_skb() calls must be replaced by consume_skb() for skbs that are not technically dropped. Signed-off-by:
-
Eric Dumazet authored
RFC 5082 IP_MINTTL option is rarely used on hosts. Add a static key to remove from TCP fast path useless code, and potential cache line miss to fetch inet_sk(sk)->min_ttl Note that once ip4_min_ttl static key has been enabled, it stays enabled until next boot. Signed-off-by:
-
Eric Dumazet authored
No report yet from KCSAN, yet worth documenting the races. Signed-off-by:
-
Eric Dumazet authored
RFC 5082 IPV6_MINHOPCOUNT is rarely used on hosts. Add a static key to remove from TCP fast path useless code, and potential cache line miss to fetch tcp_inet6_sk(sk)->min_hopcount Note that once ip6_min_hopcount static key has been enabled, it stays enabled until next boot. Signed-off-by:
-
Eric Dumazet authored
No report yet from KCSAN, yet worth documenting the races. Signed-off-by:
-
Eric Dumazet authored
sk->sk_rx_queue_mapping can be modified locklessly, add a couple of READ_ONCE()/WRITE_ONCE() to document this fact. Signed-off-by:
-
Eric Dumazet authored
sk_rx_queue_mapping is located in a cache line that should be kept read mostly. Signed-off-by:
-
Eric Dumazet authored
sk_napi_id is located in a cache line that can be kept read mostly. Signed-off-by:
-
Eric Dumazet authored
Increase cache locality by moving rx_dst_coookie next to sk->sk_rx_dst This removes one or two cache line misses in IPv6 early demux (TCP/UDP) Signed-off-by:
-
Eric Dumazet authored
Increase cache locality by moving rx_dst_ifindex next to sk->sk_rx_dst This is part of an effort to reduce cache line misses in TCP fast path. This removes one cache line miss in early demux. Signed-off-by:
-
Alexander Lobakin authored
rx_dropped, tx_dropped, rx_frame_errors and rx_crc_errors are being wrongly fetched from the target container rather than source percpu ones. No idea if that goes from the vendor driver or was brainoed during the refactoring, but fix it either way. Fixes: a97c69ba ("net: ax88796c: ASIX AX88796C SPI Ethernet Adapter Driver") Signed-off-by:
Alexander Lobakin <alobakin@pm.me> Acked-by:
Łukasz Stelmach <l.stelmach@samsung.com> Link: https://lore.kernel.org/r/20211023121148.113466-1-alobakin@pm.meSigned-off-by:
-
- 25 Oct, 2021 6 commits
-
-
Parav Pandit authored
Add SF device add and delete specific trace points. echo mlx5:mlx5_sf_dev_add >> /sys/kernel/debug/tracing/set_event echo mlx5:mlx5_sf_dev_del >> /sys/kernel/debug/tracing/set_event echo mlx5:mlx5_sf_vhca_event >> /sys/kernel/debug/tracing/set_event Signed-off-by:
Parav Pandit <parav@nvidia.com> Signed-off-by:
Saeed Mahameed <saeedm@nvidia.com>
-
Parav Pandit authored
Add support for trace events for SFs to improve debugging. This covers (a) port add and free trace points (b) device level trace points (c) SF hardware context add, free trace points. (d) SF function activate/deacticate and state trace points SF events examples: echo mlx5:mlx5_sf_add >> /sys/kernel/debug/tracing/set_event echo mlx5:mlx5_sf_free >> /sys/kernel/debug/tracing/set_event echo mlx5:mlx5_sf_hwc_alloc >> /sys/kernel/debug/tracing/set_event echo mlx5:mlx5_sf_hwc_free >> /sys/kernel/debug/tracing/set_event echo mlx5:mlx5_sf_hwc_deferred_free >> /sys/kernel/debug/tracing/set_event echo mlx5:mlx5_sf_update_state >> /sys/kernel/debug/tracing/set_event echo mlx5:mlx5_sf_activate >> /sys/kernel/debug/tracing/set_event echo mlx5:mlx5_sf_deactivate >> /sys/kernel/debug/tracing/set_event Signed-off-by:
-
Shay Drory authored
Currently, max_macs is taking 70Kbytes of memory per function. This size is not needed in all use cases, and is critical with large scale. Hence, allow user to configure the number of max_macs. For example, to reduce the number of max_macs to 1, execute:: $ devlink dev param set pci/0000:00:0b.0 name max_macs value 1 \ cmode driverinit $ devlink dev reload pci/0000:00:0b.0 Signed-off-by:Shay Drory <shayd@nvidia.com> Reviewed-by:
Moshe Shemesh <moshe@nvidia.com> Reviewed-by:
-
Shay Drory authored
Event EQ is an EQ which received the notification of almost all the events generated by the NIC. Currently, each event EQ is taking 512KB of memory. This size is not needed in most use cases, and is critical with large scale. Hence, allow user to configure the size of the event EQ. For example to reduce event EQ size to 64, execute:: $ devlink resource set pci/0000:00:0b.0 path /event_eq_size/ size 64 $ devlink dev reload pci/0000:00:0b.0 Signed-off-by:
-
Shay Drory authored
Currently, each I/O EQ is taking 128KB of memory. This size is not needed in all use cases, and is critical with large scale. Hence, allow user to configure the size of I/O EQs. For example, to reduce I/O EQ size to 64, execute: $ devlink resource set pci/0000:00:0b.0 path /io_eq_size/ size 64 $ devlink dev reload pci/0000:00:0b.0 Signed-off-by:
-
Vlad Buslov authored
The SWITCHDEV_FDB_ADD_TO_DEVICE is used for both adding new and replacing existing entry. Implement support for replacing existing FDB entries in mlx5 offload code. Signed-off-by:
Vlad Buslov <vladbu@nvidia.com> Reviewed-by:
Paul Blakey <paulb@nvidia.com> Reviewed-by:
Roi Dayan <roid@nvidia.com> Signed-off-by:
-
