linux/include/net
Jakub Kicinski 8c48eea3ad page_pool: allow caching from safely localized NAPI
Recent patches to mlx5 mentioned a regression when moving from
driver local page pool to only using the generic page pool code.
Page pool has two recycling paths (1) direct one, which runs in
safe NAPI context (basically consumer context, so producing
can be lockless); and (2) via a ptr_ring, which takes a spin
lock because the freeing can happen from any CPU; producer
and consumer may run concurrently.

Since the page pool code was added, Eric introduced a revised version
of deferred skb freeing. TCP skbs are now usually returned to the CPU
which allocated them, and freed in softirq context. This places the
freeing (producing of pages back to the pool) enticingly close to
the allocation (consumer).

If we can prove that we're freeing in the same softirq context in which
the consumer NAPI will run - lockless use of the cache is perfectly fine,
no need for the lock.

Let drivers link the page pool to a NAPI instance. If the NAPI instance
is scheduled on the same CPU on which we're freeing - place the pages
in the direct cache.

With that and patched bnxt (XDP enabled to engage the page pool, sigh,
bnxt really needs page pool work :() I see a 2.6% perf boost with
a TCP stream test (app on a different physical core than softirq).

The CPU use of relevant functions decreases as expected:

  page_pool_refill_alloc_cache   1.17% -> 0%
  _raw_spin_lock                 2.41% -> 0.98%

Only consider lockless path to be safe when NAPI is scheduled
- in practice this should cover majority if not all of steady state
workloads. It's usually the NAPI kicking in that causes the skb flush.

The main case we'll miss out on is when application runs on the same
CPU as NAPI. In that case we don't use the deferred skb free path.

Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Tested-by: Dragos Tatulea <dtatulea@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2023-04-14 18:56:12 -07:00
..
9p net/9p: distinguish zero-copy requests 2022-12-06 07:30:55 +09:00
bluetooth Bluetooth: Fix printing errors if LE Connection times out 2023-04-10 10:21:33 -07:00
caif net: remove the caif_hsi driver 2021-07-01 13:19:48 -07:00
iucv net/af_iucv: Use struct_group() to zero struct iucv_sock region 2021-11-19 11:52:25 +00:00
mana net: mana: Add support for jumbo frame 2023-04-14 08:56:19 +01:00
netfilter netfilter: nft_redir: use struct nf_nat_range2 throughout and deduplicate eval call-backs 2023-03-22 21:48:59 +01:00
netns Merge git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf 2023-02-22 21:25:23 -08:00
nfc NFC: add NCI_UNREG flag to eliminate the race 2021-11-17 20:17:05 -08:00
phonet net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
sctp sctp: add weighted fair queueing stream scheduler 2023-03-09 11:31:44 +01:00
tc_act net/sched: act_connmark: transition to percpu stats and rcu 2023-02-16 10:39:28 +01:00
6lowpan.h
act_api.h net/sched: Rename user cookie and act cookie 2023-02-20 16:46:10 -08:00
addrconf.h ipv6: constify inet6_mc_check() 2023-03-17 08:56:37 +00:00
af_ieee802154.h
af_rxrpc.h rxrpc: Tidy up abort generation infrastructure 2023-01-06 09:43:32 +00:00
af_unix.h af_unix: preserve const qualifier in unix_sk() 2023-03-18 12:23:33 +00:00
af_vsock.h vsock: support sockmap 2023-03-29 08:19:38 +01:00
ah.h
amt.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
arp.h neighbour: switch to standard rcu, instead of rcu_bh 2023-03-21 21:32:18 -07:00
atmclip.h
ax25.h x25: preserve const qualifier in [a]x25_sk() 2023-03-18 12:23:34 +00:00
ax88796.h ax88796: Fix some typo in a comment 2022-08-09 22:14:02 -07:00
bareudp.h bareudp: Move definition of struct bareudp_conf to bareudp.c 2021-12-13 12:34:09 +00:00
bond_3ad.h net: bonding: Share lacpdu_mcast_addr definition 2022-09-16 14:34:01 +01:00
bond_alb.h bonding (gcc13): synchronize bond_{a,t}lb_xmit() types 2022-11-02 20:38:13 -07:00
bond_options.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
bonding.h bonding: fix ns validation on backup slaves 2023-04-07 08:47:20 +01:00
bpf_sk_storage.h
busy_poll.h net: Fix a data-race around sysctl_net_busy_poll. 2022-08-24 13:46:58 +01:00
calipso.h
cfg80211-wext.h wifi: cfg80211: Avoid clashing function prototypes 2022-11-16 11:31:47 +02:00
cfg80211.h wifi: nl80211: make nl80211_send_chandef non-static 2023-03-24 11:45:51 +01:00
cfg802154.h ieee802154: Add support for user beaconing requests 2023-01-28 13:51:22 +01:00
checksum.h net: checksum: drop the linux/uaccess.h include 2023-01-27 11:19:46 +00:00
cipso_ipv4.h
cls_cgroup.h
codel.h codel: remove unnecessary pkt_sched.h include 2021-12-22 15:03:51 -08:00
codel_impl.h codel: remove unnecessary sock.h include 2021-12-22 15:03:47 -08:00
codel_qdisc.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
compat.h net: copy from user before calling __get_compat_msghdr 2022-07-24 18:39:17 -06:00
datalink.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
dcbevent.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
dcbnl.h net: dcb: add helper functions to retrieve PCP and DSCP rewrite maps 2023-01-20 09:33:22 +00:00
devlink.h devlink: fix the name of value arg of devl_param_driverinit_value_get() 2023-02-13 09:49:14 +00:00
dropreason.h ipv6: icmp6: add SKB_DROP_REASON_IPV6_NDISC_NS_OTHERHOST 2023-02-20 08:54:23 +00:00
dsa.h net: create a netdev notifier for DSA to reject PTP on DSA master 2023-04-03 10:04:27 +01:00
dsa_stubs.h net: dsa: replace NETDEV_PRE_CHANGE_HWTSTAMP notifier with a stub 2023-04-09 15:35:49 +01:00
dsfield.h
dst.h net: dst: Switch to rcuref_t reference counting 2023-03-28 18:52:28 -07:00
dst_cache.h wireguard: device: reset peer src endpoint when netns exits 2021-11-29 19:50:45 -08:00
dst_metadata.h xfrm: interface: Add unstable helpers for setting/getting XFRM metadata from TC-BPF 2022-12-05 21:58:27 -08:00
dst_ops.h ipv6: remove max_size check inline with ipv4 2023-01-13 20:59:14 -08:00
erspan.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
esp.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
espintcp.h
ethoc.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
failover.h net: failover: add net device refcount tracker 2021-12-06 16:06:02 -08:00
fib_notifier.h
fib_rules.h fib: expand fib_rule_policy 2021-12-16 07:18:35 -08:00
firewire.h firewire: net: Make use of get_unaligned_be48(), put_unaligned_be48() 2022-07-28 22:21:54 -07:00
flow.h net: remove unnecessary includes from net/flow.h 2023-01-27 11:19:46 +00:00
flow_dissector.h flow_dissector: Add L2TPv3 dissectors 2022-09-20 09:13:38 +02:00
flow_offload.h net/sched: cls_api: Support hardware miss to tc action 2023-02-20 16:46:10 -08:00
fou.h bpf,fou: Add bpf_skb_{set,get}_fou_encap kfuncs 2023-04-12 16:40:39 -07:00
fq.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
fq_impl.h wifi: mac80211: add support for restricting netdev features per vif 2022-12-01 15:09:10 +01:00
garp.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
gen_stats.h net: sched: Remove Qdisc::running sequence counter 2021-10-18 12:54:41 +01:00
genetlink.h mptcp: more detailed error reporting on endpoint creation 2022-11-21 13:09:07 +00:00
geneve.h net: geneve: fix array of flexible structures warnings 2022-10-31 10:43:04 +00:00
gre.h
gro.h Merge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net 2022-08-25 16:07:42 -07:00
gro_cells.h
gtp.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
gue.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
hwbm.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
icmp.h ipv6: ICMPV6: add response to ICMPV6 RFC 8335 PROBE messages 2021-06-28 14:29:45 -07:00
ieee80211_radiotap.h wifi: radiotap: separate vendor TLV into header/content 2023-03-07 20:11:36 +01:00
ieee802154_netdev.h mac802154: Handle basic beaconing 2023-01-28 13:55:10 +01:00
if_inet6.h ipv6: fix locking issues with loops over idev->addr_list 2022-04-06 22:09:39 -07:00
ife.h
ila.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
inet6_connection_sock.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
inet6_hashtables.h net: allow unbound socket for packets in VRF when tcp_l3mdev_accept set 2022-07-29 11:58:54 +01:00
inet_common.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
inet_connection_sock.h net: Add a bhash2 table hashed by port and address 2022-08-24 19:30:07 -07:00
inet_dscp.h ipv6: Define dscp_t and stop taking ECN bits into account in fib6-rules 2022-02-07 20:12:45 -08:00
inet_ecn.h net: add skb_get_dsfield() helper 2021-10-15 11:33:08 +01:00
inet_frag.h net: dropreason: add SKB_DROP_REASON_FRAG_REASM_TIMEOUT 2022-10-31 20:14:27 -07:00
inet_hashtables.h tcp: Add TIME_WAIT sockets in bhash2. 2022-12-30 07:25:52 +00:00
inet_sock.h inet: preserve const qualifier in inet_sk() 2023-03-17 08:56:37 +00:00
inet_timewait_sock.h tcp: Add TIME_WAIT sockets in bhash2. 2022-12-30 07:25:52 +00:00
inetpeer.h
ioam6.h treewide: Replace zero-length arrays with flexible-array members 2022-02-17 07:00:39 -06:00
ip.h inet: Add IP_LOCAL_PORT_RANGE socket option 2023-01-25 22:45:00 -08:00
ip6_checksum.h net: move gro definitions to include/net/gro.h 2021-11-16 13:16:54 +00:00
ip6_fib.h ipv6: Remove in6addr_any alternatives. 2023-03-29 08:22:52 +01:00
ip6_route.h net: dst: Prevent false sharing vs. dst_entry:: __refcnt 2023-03-28 18:52:22 -07:00
ip6_tunnel.h ip_gre, ip6_gre: Fix race condition on o_seqno in collect_md mode 2022-04-25 11:40:45 +01:00
ip_fib.h ipv4: Use dscp_t in struct fib_entry_notifier_info 2022-04-11 17:37:50 -07:00
ip_tunnels.h bpf-next-for-netdev 2023-04-13 16:43:38 -07:00
ip_vs.h ipvs: avoid kfree_rcu without 2nd arg 2023-02-02 14:02:01 +01:00
ipcomp.h xfrm: ipcomp: add extack to ipcomp{4,6}_init_state 2022-09-29 07:18:00 +02:00
ipconfig.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
ipv6.h ipv6: icmp6: add drop reason support to icmpv6_notify() 2023-02-13 19:55:32 -08:00
ipv6_frag.h net: dropreason: add SKB_DROP_REASON_FRAG_REASM_TIMEOUT 2022-10-31 20:14:27 -07:00
ipv6_stubs.h bpf: Change bpf_getsockopt(SOL_IPV6) to reuse do_ipv6_getsockopt() 2022-09-02 20:34:32 -07:00
iw_handler.h
kcm.h
l3mdev.h
lag.h
lapb.h
lib80211.h
llc.h llc: fix out-of-bound array index in llc_sk_dev_hash() 2021-11-07 19:25:29 +00:00
llc_c_ac.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
llc_c_ev.h
llc_c_st.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
llc_conn.h llc: add net device refcount tracker 2021-12-07 20:44:59 -08:00
llc_if.h llc/snap: constify dev_addr passing 2021-10-13 09:40:46 -07:00
llc_pdu.h net: llc: fix skb_over_panic 2021-07-27 13:05:56 +01:00
llc_s_ac.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
llc_s_ev.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
llc_s_st.h add missing includes and forward declarations to networking includes under linux/ 2022-07-28 11:29:36 +02:00
llc_sap.h
lwtunnel.h netfilter: add netfilter hooks to SRv6 data plane 2021-08-30 01:51:36 +02:00
mac80211.h wifi: mac80211: generate EMA beacons in AP mode 2023-03-22 13:54:57 +01:00
mac802154.h mac802154: Drop IEEE802154_HW_RX_DROP_BAD_CKSUM 2022-10-12 12:57:19 +02:00
macsec.h net: macsec: remove the prepare flag from the MACsec offloading context 2022-09-23 06:56:08 -07:00
mctp.h mctp: Use output netdev to allocate skb headroom 2022-04-01 12:04:15 +01:00
mctpdevice.h mctp: Pass flow data & flow release events to drivers 2021-10-29 13:23:51 +01:00
mip6.h
mld.h
mpls.h
mpls_iptunnel.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
mptcp.h mptcp: remove MPTCP 'ifdef' in TCP SYN cookies 2022-12-12 13:11:24 -08:00
mrp.h mrp: introduce active flags to prevent UAF when applicant uninit 2022-11-18 12:14:55 +00:00
ncsi.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
ndisc.h neighbour: switch to standard rcu, instead of rcu_bh 2023-03-21 21:32:18 -07:00
neighbour.h neighbour: switch to standard rcu, instead of rcu_bh 2023-03-21 21:32:18 -07:00
net_debug.h net: add CONFIG_DEBUG_NET 2022-05-11 12:43:10 +01:00
net_failover.h
net_namespace.h net: add a refcount tracker for kernel sockets 2022-10-24 11:04:43 +01:00
net_ratelimit.h
net_trackers.h net: add networking namespace refcount tracker 2021-12-10 06:38:26 -08:00
netdev_queues.h net: piggy back on the memory barrier in bql when waking queues 2023-04-10 17:56:18 -07:00
netevent.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
netlabel.h
netlink.h net: netlink: recommend policy range validation 2023-01-28 00:33:51 -08:00
netprio_cgroup.h
netrom.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
nexthop.h neighbour: switch to standard rcu, instead of rcu_bh 2023-03-21 21:32:18 -07:00
nl802154.h ieee802154: Add support for user beaconing requests 2023-01-28 13:51:22 +01:00
nsh.h
p8022.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
page_pool.h page_pool: allow caching from safely localized NAPI 2023-04-14 18:56:12 -07:00
pie.h
ping.h inet: ping: use hlist_nulls rcu iterator during lookup 2022-12-01 12:42:46 +01:00
pkt_cls.h net/sched: cls_api: Support hardware miss to tc action 2023-02-20 16:46:10 -08:00
pkt_sched.h net/sched: mqprio: allow per-TC user input of FP adminStatus 2023-04-13 22:22:10 -07:00
pptp.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
protocol.h tcp/udp: Make early_demux back namespacified. 2022-07-15 18:50:35 -07:00
psample.h psample: Add a fwd declaration for skbuff 2021-08-09 15:34:21 -07:00
psnap.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
raw.h Merge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net 2023-04-06 12:01:20 -07:00
rawv6.h ipv6: raw: constify raw_v6_match() socket argument 2023-03-17 08:56:37 +00:00
red.h treewide: use get_random_u32() when possible 2022-10-11 17:42:58 -06:00
regulatory.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
request_sock.h tcp: Use BPF timeout setting for SYN ACK RTO 2022-02-02 14:45:18 +00:00
rose.h net: rose: add netdev ref tracker to 'struct rose_sock' 2022-08-01 11:59:23 -07:00
route.h net: dst: Prevent false sharing vs. dst_entry:: __refcnt 2023-03-28 18:52:22 -07:00
rpl.h
rsi_91x.h
rtnetlink.h rtnetlink: Honour NLM_F_ECHO flag in rtnl_delete_link 2022-10-31 18:10:21 -07:00
rtnh.h
sch_generic.h net/sched: cls_api: Support hardware miss to tc action 2023-02-20 16:46:10 -08:00
scm.h scm: fix MSG_CTRUNC setting condition for SO_PASSSEC 2023-03-15 08:20:12 +00:00
secure_seq.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
seg6.h udp6: Use Segment Routing Header for dest address if present 2022-01-04 12:17:35 +00:00
seg6_hmac.h
seg6_local.h
selftests.h net: selftest: fix build issue if INET is disabled 2021-04-28 14:06:45 -07:00
slhc_vj.h
smc.h net/smc: Introduce explicit check for v2 support 2023-03-15 08:18:35 +00:00
snmp.h
sock.h net: dst: Switch to rcuref_t reference counting 2023-03-28 18:52:28 -07:00
sock_reuseport.h soreuseport: Fix socket selection for SO_INCOMING_CPU. 2022-10-25 11:35:16 +02:00
Space.h wan: remove sbni/granch driver 2021-08-03 13:05:26 +01:00
stp.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
strparser.h tls: rx: remove the message decrypted tracking 2022-07-18 11:24:10 +01:00
switchdev.h bridge: switchdev: Allow device drivers to install locked FDB entries 2022-11-09 19:06:13 -08:00
tc_wrapper.h net/sched: Retire rsvp classifier 2023-02-16 09:27:07 +01:00
tcp.h bpf-next-for-netdev 2023-04-13 16:43:38 -07:00
tcp_states.h
timewait_sock.h
tipc.h
tls.h net/tls: Describe ciphers sizes by const structs 2022-09-22 17:27:41 -07:00
tls_toe.h
transp_v6.h inet6: Remove inet6_destroy_sock(). 2022-10-24 09:40:39 +01:00
tso.h net: tso: inline tso_count_descs() 2022-12-12 15:04:39 -08:00
tun_proto.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
udp.h udp: track the forward memory release threshold in an hot cacheline 2022-10-24 10:52:50 +01:00
udp_tunnel.h net: Change the udp encap_err_rcv to allow use of {ip,ipv6}_icmp_error() 2022-11-08 16:42:28 +00:00
udplite.h tcp/udp: Call inet6_destroy_sock() in IPv6 sk->sk_destruct(). 2022-10-12 17:50:37 -07:00
vsock_addr.h
vxlan.h vxlan: Expose helper vxlan_build_gbp_hdr 2023-03-17 22:41:16 -07:00
wext.h
x25.h x25: preserve const qualifier in [a]x25_sk() 2023-03-18 12:23:34 +00:00
x25device.h
xdp.h bpf-next-for-netdev 2023-04-13 16:43:38 -07:00
xdp_priv.h net: add missing includes and forward declarations under net/ 2022-07-22 12:53:22 +01:00
xdp_sock.h bpf, net: xskmap memory usage 2023-03-07 09:33:43 -08:00
xdp_sock_drv.h xsk: Remove unused xsk_buff_discard 2022-09-30 07:55:46 -07:00
xfrm.h xfrm: add new device offload acquire flag 2023-03-20 11:29:33 +02:00
xsk_buff_pool.h xsk: Fix unaligned descriptor validation 2023-04-06 09:53:05 -07:00