linux/net
Yunsheng Lin a90c57f2ce net: sched: fix packet stuck problem for lockless qdisc
Lockless qdisc has below concurrent problem:
    cpu0                 cpu1
     .                     .
q->enqueue                 .
     .                     .
qdisc_run_begin()          .
     .                     .
dequeue_skb()              .
     .                     .
sch_direct_xmit()          .
     .                     .
     .                q->enqueue
     .             qdisc_run_begin()
     .            return and do nothing
     .                     .
qdisc_run_end()            .

cpu1 enqueue a skb without calling __qdisc_run() because cpu0
has not released the lock yet and spin_trylock() return false
for cpu1 in qdisc_run_begin(), and cpu0 do not see the skb
enqueued by cpu1 when calling dequeue_skb() because cpu1 may
enqueue the skb after cpu0 calling dequeue_skb() and before
cpu0 calling qdisc_run_end().

Lockless qdisc has below another concurrent problem when
tx_action is involved:

cpu0(serving tx_action)     cpu1             cpu2
          .                   .                .
          .              q->enqueue            .
          .            qdisc_run_begin()       .
          .              dequeue_skb()         .
          .                   .            q->enqueue
          .                   .                .
          .             sch_direct_xmit()      .
          .                   .         qdisc_run_begin()
          .                   .       return and do nothing
          .                   .                .
 clear __QDISC_STATE_SCHED    .                .
 qdisc_run_begin()            .                .
 return and do nothing        .                .
          .                   .                .
          .            qdisc_run_end()         .

This patch fixes the above data race by:
1. If the first spin_trylock() return false and STATE_MISSED is
   not set, set STATE_MISSED and retry another spin_trylock() in
   case other CPU may not see STATE_MISSED after it releases the
   lock.
2. reschedule if STATE_MISSED is set after the lock is released
   at the end of qdisc_run_end().

For tx_action case, STATE_MISSED is also set when cpu1 is at the
end if qdisc_run_end(), so tx_action will be rescheduled again
to dequeue the skb enqueued by cpu2.

Clear STATE_MISSED before retrying a dequeuing when dequeuing
returns NULL in order to reduce the overhead of the second
spin_trylock() and __netif_schedule() calling.

Also clear the STATE_MISSED before calling __netif_schedule()
at the end of qdisc_run_end() to avoid doing another round of
dequeuing in the pfifo_fast_dequeue().

The performance impact of this patch, tested using pktgen and
dummy netdev with pfifo_fast qdisc attached:

 threads  without+this_patch   with+this_patch      delta
    1        2.61Mpps            2.60Mpps           -0.3%
    2        3.97Mpps            3.82Mpps           -3.7%
    4        5.62Mpps            5.59Mpps           -0.5%
    8        2.78Mpps            2.77Mpps           -0.3%
   16        2.22Mpps            2.22Mpps           -0.0%

Fixes: 6b3ba9146f ("net: sched: allow qdiscs to handle locking")
Acked-by: Jakub Kicinski <kuba@kernel.org>
Tested-by: Juergen Gross <jgross@suse.com>
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2021-05-14 15:05:46 -07:00
..
6lowpan 6lowpan: Fix some typos in nhc_udp.c 2021-03-24 17:52:11 -07:00
9p net: 9p: Correct function names in the kerneldoc comments 2021-03-28 17:56:56 -07:00
802
8021q Merge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net 2021-04-26 12:00:00 -07:00
appletalk appletalk: Fix skb allocation size in loopback case 2021-02-12 16:40:28 -08:00
atm
ax25 net/ax25: Delete obsolete TODO file 2021-03-30 16:54:50 -07:00
batman-adv Merge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net 2021-04-09 20:48:35 -07:00
bluetooth Networking changes for 5.13. 2021-04-29 11:57:23 -07:00
bpf bpf: selftests: Add kfunc_call test 2021-03-26 20:41:52 -07:00
bpfilter
bridge bridge: Fix possible races between assigning rx_handler_data and setting IFF_BRIDGE_PORT bit 2021-04-29 15:33:17 -07:00
caif net: caif: Use netif_rx_any_context(). 2021-02-15 13:21:48 -08:00
can can: isotp: prevent race between isotp_bind() and isotp_setsockopt() 2021-05-12 08:52:47 +02:00
ceph Notable items here are a series to take advantage of David Howells' 2021-05-06 10:27:02 -07:00
core net: really orphan skbs tied to closing sk 2021-05-12 13:48:58 -07:00
dcb
dccp net: dccp: use net_generic storage 2021-04-09 16:34:56 -07:00
decnet net/decnet: Delete obsolete TODO file 2021-03-30 16:54:50 -07:00
dns_resolver
dsa net: dsa: fix error code getting shifted with 4 in dsa_slave_get_sset_count 2021-05-10 14:36:59 -07:00
ethernet of: net: pass the dst buffer to of_get_mac_address() 2021-04-13 14:35:02 -07:00
ethtool ethtool: fix missing NLM_F_MULTI flag when dumping 2021-05-05 12:41:10 -07:00
hsr net: hsr: check skb can contain struct hsr_ethhdr in fill_frame_info 2021-05-03 13:33:54 -07:00
ieee802154 net: remove the new_ifindex argument from dev_change_net_namespace 2021-04-07 14:43:28 -07:00
ife
ipv4 Merge git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf 2021-05-11 16:05:56 -07:00
ipv6 net: Remove redundant assignment to err 2021-04-29 15:34:15 -07:00
iucv iucv: af_iucv.c: Couple of typo fixes 2021-03-28 17:31:13 -07:00
kcm kcm: kcmsock.c: Couple of typo fixes 2021-03-28 17:31:13 -07:00
key
l2tp net: fix a concurrency bug in l2tp_tunnel_register() 2021-04-27 14:23:13 -07:00
l3mdev l3mdev: Correct function names in the kerneldoc comments 2021-03-28 17:56:55 -07:00
lapb net: lapb: Make "lapb_t1timer_running" able to detect an already running timer 2021-03-23 14:14:50 -07:00
llc llc2: Remove redundant assignment to rc 2021-04-27 14:16:14 -07:00
mac80211 mac80211: extend protection against mixed key and fragment cache attacks 2021-05-11 20:14:50 +02:00
mac802154 net: mac802154: Fix general protection fault 2021-04-06 22:42:16 +02:00
mpls mpls: Remove redundant assignment to err 2021-04-27 14:17:00 -07:00
mptcp mptcp: fix data stream corruption 2021-05-11 16:19:17 -07:00
ncsi Merge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net 2021-04-09 20:48:35 -07:00
netfilter netfilter: nft_set_pipapo_avx2: Add irq_fpu_usable() check, fallback to non-AVX2 version 2021-05-14 01:42:52 +02:00
netlabel Networking changes for 5.13. 2021-04-29 11:57:23 -07:00
netlink netlink: don't call ->netlink_bind with table lock held 2021-04-16 17:01:04 -07:00
netrom net: netrom: nr_in: Remove redundant assignment to ns 2021-04-28 13:59:08 -07:00
nfc net/nfc/rawsock.c: fix a permission check bug 2021-05-10 14:21:02 -07:00
nsh
openvswitch openvswitch: meter: fix race when getting now_ms. 2021-05-13 15:54:59 -07:00
packet net: packetmmap: fix only tx timestamp on request 2021-05-12 14:00:04 -07:00
phonet
psample psample: Add additional metadata attributes 2021-03-14 15:00:43 -07:00
qrtr Merge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net 2021-04-26 12:00:00 -07:00
rds RDMA merge window pull request 2021-05-01 09:15:05 -07:00
rfkill Another set of updates, all over the map: 2021-04-20 16:44:04 -07:00
rose net: rose: Fix fall-through warnings for Clang 2021-03-10 12:45:15 -08:00
rxrpc Networking changes for 5.13. 2021-04-29 11:57:23 -07:00
sched net: sched: fix packet stuck problem for lockless qdisc 2021-05-14 15:05:46 -07:00
sctp sctp: delay auto_asconf init until binding the first addr 2021-05-03 13:36:21 -07:00
smc smc: disallow TCP_ULP in smc_setsockopt() 2021-05-05 12:52:45 -07:00
strparser
sunrpc NFS client updates for Linux 5.13 2021-05-07 11:23:41 -07:00
switchdev net: bridge: propagate extack through switchdev_port_attr_set 2021-02-14 17:38:11 -08:00
tipc Revert "net:tipc: Fix a double free in tipc_sk_mcast_rcv" 2021-05-14 15:01:58 -07:00
tls tls splice: check SPLICE_F_NONBLOCK instead of MSG_DONTWAIT 2021-05-14 15:03:25 -07:00
unix
vmw_vsock vsock/vmci: Remove redundant assignment to err 2021-04-30 15:00:59 -07:00
wireless cfg80211: mitigate A-MSDU aggregation attacks 2021-05-11 20:13:13 +02:00
x25 af_x25.c: Fix a spello 2021-03-28 17:31:13 -07:00
xdp xsk: Fix for xp_aligned_validate_desc() when len == chunk_size 2021-05-04 00:28:06 +02:00
xfrm xfrm: ipcomp: remove unnecessary get_cpu() 2021-04-19 12:49:29 +02:00
compat.c
devres.c
Kconfig bpf, kconfig: Add consolidated menu entry for bpf with core options 2021-05-11 13:56:16 -07:00
Makefile
socket.c net: Fix a misspell in socket.c 2021-03-25 16:56:27 -07:00
sysctl_net.c net: Ensure net namespace isolation of sysctls 2021-04-12 13:27:11 -07:00