mptcp: More v5.13 fixes
Subsystems (4)
cover mptcp: More v5.13 fixes
Here's another batch of MPTCP fixes for v5.13.
Patch 1 cleans up memory accounting between the MPTCP-level socket and the subflows to more reliably transfer forward allocated memory under pressure.
Patch 2 wakes up socket readers more reliably.
Patch 3 changes a WARN_ONCE to a pr_debug.
Patch 4 changes the selftests to only use syncookies in test cases where they do not cause spurious failures.
Patch 5 modifies socket error reporting to avoid a possible soft lockup.
Paolo Abeni (5):
mptcp: try harder to borrow memory from subflow under pressure
mptcp: wake-up readers only for in sequence data
mptcp: do not warn on bad input from the network
selftests: mptcp: enable syncookie only in absence of reorders
mptcp: fix soft lookup in subflow_error_report()
net/mptcp/protocol.c | 52 +++++---- net/mptcp/protocol.h | 1 - net/mptcp/subflow.c | 108 +++++++++--------- .../selftests/net/mptcp/mptcp_connect.sh | 11 +- 4 files changed, 88 insertions(+), 84 deletions(-)
base-commit: 22488e45501eca74653b502b194eb0eb25d2ad00
Signature · 1 line
2.32.0
PATCH 1/5 mptcp: try harder to borrow memory from subflow under pressure
From: Paolo Abeni <[email protected]>
If the host is under sever memory pressure, and RX forward memory allocation for the msk fails, we try to borrow the required memory from the ingress subflow.
The current attempt is a bit flaky: if skb->truesize is less than SK_MEM_QUANTUM, the ssk will not release any memory, and the next schedule will fail again.
Instead, directly move the required amount of pages from the ssk to the msk, if available
--- net/mptcp/protocol.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-)
net/mptcp/protocol.c +6 -4
@@ -280,11 +280,13 @@ static bool __mptcp_move_skb(struct mptcp_sock *msk, struct sock *ssk,/* try to fetch required memory from subflow */ if (!sk_rmem_schedule (sk, skb, skb->truesize)) { -if (ssk->sk_forward_alloc < skb->truesize) - goto drop; -__sk_mem_reclaim (ssk, skb->truesize); -if (!sk_rmem_schedule (sk, skb, skb->truesize)) +int amount =sk_mem_pages (skb->truesize) <<SK_MEM_QUANTUM_SHIFT ; + +if (ssk->sk_forward_alloc < amount) goto drop; + + ssk->sk_forward_alloc -= amount; + sk->sk_forward_alloc += amount; }/* the skb map_seq accounts for the skb offset:
Signature · 1 line
2.32.0
PATCH 2/5 mptcp: wake-up readers only for in sequence data
In reply to Mat MartineauFrom: Paolo Abeni <[email protected]>
Currently we rely on the subflow->data_avail field, which is subject to races:
ssk1 skb len = 500 DSS(seq=1, len=1000, off=0) # data_avail == MPTCP_SUBFLOW_DATA_AVAIL ssk2 skb len = 500 DSS(seq = 501, len=1000) # data_avail == MPTCP_SUBFLOW_DATA_AVAIL ssk1 skb len = 500 DSS(seq = 1, len=1000, off =500) # still data_avail == MPTCP_SUBFLOW_DATA_AVAIL, # as the skb is covered by a pre-existing map, # which was in-sequence at reception time.
Instead we can explicitly check if some has been received in-sequence, propagating the info from __mptcp_move_skbs_from_subflow().
Additionally add the 'ONCE' annotation to the 'data_avail' memory access, as msk will read it outside the subflow socket lock.
--- net/mptcp/protocol.c | 33 ++++++++++++--------------------- net/mptcp/protocol.h | 1 - net/mptcp/subflow.c | 23 +++++++++-------------- 3 files changed, 21 insertions(+), 36 deletions(-)
net/mptcp/protocol.c +12 -21
@@ -670,15 +670,13 @@ static bool __mptcp_ofo_queue(struct mptcp_sock *msk)/* In most cases we will be able to lock the mptcp socket. If its already * owned, we need to defer to the work queue to avoid ABBA deadlock. */ -static void move_skbs_to_msk (struct mptcp_sock *msk,struct sock *ssk) +static bool move_skbs_to_msk (struct mptcp_sock *msk,struct sock *ssk) {struct sock *sk = (struct sock *)msk;unsigned int moved =0 ;if (inet_sk_state_load (sk) ==TCP_CLOSE ) -return ; - -mptcp_data_lock (sk); +return false;__mptcp_move_skbs_from_subflow (msk, ssk, &moved);__mptcp_ofo_queue (msk);
@@ -690,7 +688,7 @@ static void move_skbs_to_msk(struct mptcp_sock *msk, struct sock *ssk)*/ if (mptcp_pending_data_fin (sk,NULL ))mptcp_schedule_work (sk); -mptcp_data_unlock (sk); +return moved >0 ; }void mptcp_data_ready (struct sock *sk,struct sock *ssk)
@@ -698,7 +696,6 @@ void mptcp_data_ready(struct sock *sk, struct sock *ssk)struct mptcp_subflow_context *subflow =mptcp_subflow_ctx (ssk);struct mptcp_sock *msk =mptcp_sk (sk);int sk_rbuf, ssk_rbuf; -bool wake;/* The peer can send data while we are shutting down this * subflow at msk destruction time, but we must avoid enqueuing
@@ -707,28 +704,22 @@ void mptcp_data_ready(struct sock *sk, struct sock *ssk)if (unlikely (subflow->disposable))return ; -/* move_skbs_to_msk below can legitly clear the data_avail flag, -* but we will need later to properly woke the reader, cache its -* value -*/ - wake = subflow->data_avail ==MPTCP_SUBFLOW_DATA_AVAIL ; -if (wake) -set_bit (MPTCP_DATA_READY , &msk->flags); - ssk_rbuf =READ_ONCE (ssk->sk_rcvbuf); sk_rbuf =READ_ONCE (sk->sk_rcvbuf);if (unlikely (ssk_rbuf > sk_rbuf)) sk_rbuf = ssk_rbuf; -/* over limit? can't append more skbs to msk */ +/* over limit? can't append more skbs to msk, Also, no need to wake-up*/ if (atomic_read (&sk->sk_rmem_alloc) > sk_rbuf) - goto wake; - -move_skbs_to_msk (msk, ssk); +return ; -wake: -if (wake) +/* Wake-up the reader only for in-sequence data */ +mptcp_data_lock (sk); +if (move_skbs_to_msk (msk, ssk)) { +set_bit (MPTCP_DATA_READY , &msk->flags); sk->sk_data_ready (sk); + } +mptcp_data_unlock (sk); }static bool mptcp_do_flush_join_list (struct mptcp_sock *msk)
@@ -860,7 +851,7 @@ static struct sock *mptcp_subflow_recv_lookup(const struct mptcp_sock *msk)sock_owned_by_me (sk);mptcp_for_each_subflow (msk, subflow) { -if (subflow->data_avail) +if (READ_ONCE (subflow->data_avail))return mptcp_subflow_tcp_sock (subflow); }
net/mptcp/protocol.h +0 -1
@@ -362,7 +362,6 @@ mptcp_subflow_rsk(const struct request_sock *rsk)enum mptcp_data_avail {MPTCP_SUBFLOW_NODATA ,MPTCP_SUBFLOW_DATA_AVAIL , -MPTCP_SUBFLOW_OOO_DATA };struct mptcp_delegated_action {
net/mptcp/subflow.c +9 -14
@@ -1000,7 +1000,7 @@ static bool subflow_check_data_avail(struct sock *ssk)struct sk_buff *skb;if (!skb_peek (&ssk->sk_receive_queue)) - subflow->data_avail =0 ; +WRITE_ONCE (subflow->data_avail,0 );if (subflow->data_avail)return true;
@@ -1039,18 +1039,13 @@ static bool subflow_check_data_avail(struct sock *ssk) ack_seq =mptcp_subflow_get_mapped_dsn (subflow);pr_debug ("msk ack_seq=%llx subflow ack_seq=%llx" , old_ack, ack_seq); -if (ack_seq == old_ack) { - subflow->data_avail =MPTCP_SUBFLOW_DATA_AVAIL ; -break ; - }else if (after64 (ack_seq, old_ack)) { - subflow->data_avail =MPTCP_SUBFLOW_OOO_DATA ; -break ; +if (unlikely (before64 (ack_seq, old_ack))) { +mptcp_subflow_discard_data (ssk, skb, old_ack - ack_seq); +continue ; } -/* only accept in-sequence mapping. Old values are spurious -* retransmission -*/ -mptcp_subflow_discard_data (ssk, skb, old_ack - ack_seq); +WRITE_ONCE (subflow->data_avail,MPTCP_SUBFLOW_DATA_AVAIL ); +break ; }return true;
@@ -1070,7 +1065,7 @@ static bool subflow_check_data_avail(struct sock *ssk) subflow->reset_transient =0 ; subflow->reset_reason =MPTCP_RST_EMPTCP ;tcp_send_active_reset (ssk,GFP_ATOMIC ); - subflow->data_avail =0 ; +WRITE_ONCE (subflow->data_avail,0 );return false; }
@@ -1080,7 +1075,7 @@ static bool subflow_check_data_avail(struct sock *ssk) subflow->map_seq =READ_ONCE (msk->ack_seq); subflow->map_data_len = skb->len; subflow->map_subflow_seq =tcp_sk (ssk)->copied_seq - subflow->ssn_offset; - subflow->data_avail =MPTCP_SUBFLOW_DATA_AVAIL ; +WRITE_ONCE (subflow->data_avail,MPTCP_SUBFLOW_DATA_AVAIL );return true; }
@@ -1092,7 +1087,7 @@ bool mptcp_subflow_data_available(struct sock *sk)if (subflow->map_valid &&mptcp_subflow_get_map_offset (subflow) >= subflow->map_data_len) { subflow->map_valid =0 ; - subflow->data_avail =0 ; +WRITE_ONCE (subflow->data_avail,0 ); pr_debug("Done with mapping: seq=%u data_len=%u", subflow->map_subflow_seq,
Signature · 1 line
2.32.0
PATCH 3/5 mptcp: do not warn on bad input from the network
In reply to Mat MartineauFrom: Paolo Abeni <[email protected]>
warn_bad_map() produces a kernel WARN on bad input coming from the network. Use pr_debug() to avoid spamming the system log.
Additionally, when the right bound check fails, warn_bad_map() reports the wrong ssn value, let's fix it.
--- net/mptcp/subflow.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-)
net/mptcp/subflow.c +5 -5
@@ -784,10 +784,10 @@ static u64 expand_seq(u64 old_seq, u16 old_data_len, u64 seq)return seq | ((old_seq + old_data_len +1 ) &GENMASK_ULL (63 ,32 )); } -static void warn_bad_map (struct mptcp_subflow_context *subflow,u32 ssn) +static void dbg_bad_map (struct mptcp_subflow_context *subflow,u32 ssn) { -WARN_ONCE (1 ,"Bad mapping: ssn=%d map_seq=%d map_data_len=%d" , - ssn, subflow->map_subflow_seq, subflow->map_data_len); +pr_debug ("Bad mapping: ssn=%d map_seq=%d map_data_len=%d" , + ssn, subflow->map_subflow_seq, subflow->map_data_len); }static bool skb_is_fully_mapped (struct sock *ssk,struct sk_buff *skb)
@@ -812,13 +812,13 @@ static bool validate_mapping(struct sock *ssk, struct sk_buff *skb)/* Mapping covers data later in the subflow stream, * currently unsupported. */ -warn_bad_map (subflow, ssn); +dbg_bad_map (subflow, ssn);return false; }if (unlikely (!before (ssn, subflow->map_subflow_seq + subflow->map_data_len))) {/* Mapping does covers past subflow data, invalid */ -warn_bad_map (subflow, ssn + skb->len); +dbg_bad_map (subflow, ssn);return false; }return true;
Signature · 1 line
2.32.0
PATCH 4/5 selftests: mptcp: enable syncookie only in absence of reorders
In reply to Mat MartineauFrom: Paolo Abeni <[email protected]>
Syncookie validation may fail for OoO packets, causing spurious resets and self-tests failures, so let's force syncookie only for tests iteration with no OoO.
--- tools/testing/selftests/net/mptcp/mptcp_connect.sh | 11 ++++++++--- 1 file changed, 8 insertions(+), 3 deletions(-)
tools/testing/selftests/net/mptcp/mptcp_connect.sh +8 -3
@@ -197,9 +197,6 @@ ip -net "$ns4" link set ns4eth3 up ip -net "$ns4" route add default via 10.0.3.2 ip -net "$ns4" route add default via dead:beef:3::2 -# use TCP syn cookies, even if no flooding was detected. -ip netns exec "$ns2" sysctl -q net.ipv4.tcp_syncookies=2 - set_ethtool_flags() { local ns="$1" local dev="$2"
@@ -737,6 +734,14 @@ for sender in $ns1 $ns2 $ns3 $ns4;do exit $ret fi + # ns1<->ns2 is not subject to reordering/tc delays. Use it to test + # mptcp syncookie support. + if [ $sender = $ns1 ]; then + ip netns exec "$ns2" sysctl -q net.ipv4.tcp_syncookies=2 + else + ip netns exec "$ns2" sysctl -q net.ipv4.tcp_syncookies=1 + fi + run_tests "$ns2" $sender 10.0.1.2 run_tests "$ns2" $sender dead:beef:1::2 run_tests "$ns2" $sender 10.0.2.1
Signature · 1 line
2.32.0
PATCH 5/5 mptcp: fix soft lookup in subflow_error_report()
In reply to Mat MartineauFrom: Paolo Abeni <[email protected]>
Maxim reported a soft lookup in subflow_error_report():
watchdog: BUG: soft lockup - CPU#0 stuck for 22s! [swapper/0:0] RIP: 0010:native_queued_spin_lock_slowpath RSP: 0018:ffffa859c0003bc0 EFLAGS: 00000202 RAX: 0000000000000101 RBX: 0000000000000001 RCX: 0000000000000000 RDX: ffff9195c2772d88 RSI: 0000000000000000 RDI: ffff9195c2772d88 RBP: ffff9195c2772d00 R08: 00000000000067b0 R09: c6e31da9eb1e44f4 R10: ffff9195ef379700 R11: ffff9195edb50710 R12: ffff9195c2772d88 R13: ffff9195f500e3d0 R14: ffff9195ef379700 R15: ffff9195ef379700 FS: 0000000000000000(0000) GS:ffff91961f400000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 000000c000407000 CR3: 0000000002988000 CR4: 00000000000006f0 Call Trace: <IRQ> _raw_spin_lock_bh subflow_error_report mptcp_subflow_data_available __mptcp_move_skbs_from_subflow mptcp_data_ready tcp_data_queue tcp_rcv_established tcp_v4_do_rcv tcp_v4_rcv ip_protocol_deliver_rcu ip_local_deliver_finish __netif_receive_skb_one_core netif_receive_skb rtl8139_poll 8139too __napi_poll net_rx_action __do_softirq __irq_exit_rcu common_interrupt </IRQ>
The calling function - mptcp_subflow_data_available() - can be invoked from different contexts:
- plain ssk socket lock
- ssk socket lock + mptcp_data_lock
- ssk socket lock + mptcp_data_lock + msk socket lock.
Since subflow_error_report() tries to acquire the mptcp_data_lock, the latter two call chains will cause soft lookup.
This change addresses the issue moving the error reporting call to outer functions, where the held locks list is known and the we can acquire only the needed one.
--- net/mptcp/protocol.c | 9 ++++++ net/mptcp/subflow.c | 75 +++++++++++++++++++++++--------------------- 2 files changed, 48 insertions(+), 36 deletions(-)
net/mptcp/protocol.c +9 -0
@@ -680,6 +680,12 @@ static bool move_skbs_to_msk(struct mptcp_sock *msk, struct sock *ssk)__mptcp_move_skbs_from_subflow (msk, ssk, &moved);__mptcp_ofo_queue (msk); +if (unlikely (ssk->sk_err)) { +if (!sock_owned_by_user (sk)) +__mptcp_error_report (sk); +else +set_bit (MPTCP_ERROR_REPORT , &msk->flags); + }/* If the moves have caught up with the DATA_FIN sequence number * it's time to ack the DATA_FIN and change socket state, but
@@ -1948,6 +1954,9 @@ static bool __mptcp_move_skbs(struct mptcp_sock *msk) done =__mptcp_move_skbs_from_subflow (msk, ssk, &moved);mptcp_data_unlock (sk);tcp_cleanup_rbuf (ssk, moved); + +if (unlikely (ssk->sk_err)) +__mptcp_error_report (sk);unlock_sock_fast (ssk, slowpath); }while (!done);
net/mptcp/subflow.c +39 -36
99 lines not on this page · raw message
Signature · 1 line
2.32.0
This series was applied to netdev/net.git (refs/heads/master):
Hello:
This series was applied to netdev/net.git (refs/heads/master):
Quoted 9 lines · On Thu, 10 Jun 2021 15:59:39 -0700 you wrote:
Here's another batch of MPTCP fixes for v5.13.
Patch 1 cleans up memory accounting between the MPTCP-level socket and the subflows to more reliably transfer forward allocated memory under pressure.
Patch 2 wakes up socket readers more reliably.
[...]
Here is the summary with links:
- [net,1/5] mptcp: try harder to borrow memory from subflow under pressure
https://git.kernel.org/netdev/net/c/72f961320d5d - [net,2/5] mptcp: wake-up readers only for in sequence data https://git.kernel.org/netdev/net/c/99d1055ce246 - [net,3/5] mptcp: do not warn on bad input from the network https://git.kernel.org/netdev/net/c/61e710227e97 - [net,4/5] selftests: mptcp: enable syncookie only in absence of reorders https://git.kernel.org/netdev/net/c/2395da0e1793 - [net,5/5] mptcp: fix soft lookup in subflow_error_report() https://git.kernel.org/netdev/net/c/499ada507336
You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html