[ Upstream commit 0ec986ed7bab6801faed1440e8839dcc710331ff ]
Loss recovery undo_retrans bookkeeping had a long-standing bug where a
DSACK from a spurious TLP retransmit packet could cause an erroneous
undo of a fast recovery or RTO recovery that repaired a single
really-lost packet (in a sequence range outside that of the TLP
retransmit). Basically, because the loss recovery state machine didn't
account for the fact that it sent a TLP retransmit, the DSACK for the
TLP retransmit could erroneously be implicitly be interpreted as
corresponding to the normal fast recovery or RTO recovery retransmit
that plugged a real hole, thus resulting in an improper undo.
For example, consider the following buggy scenario where there is a
real packet loss but the congestion control response is improperly
undone because of this bug:
+ send packets P1, P2, P3, P4
+ P1 is really lost
+ send TLP retransmit of P4
+ receive SACK for original P2, P3, P4
+ enter fast recovery, fast-retransmit P1, increment undo_retrans to 1
+ receive DSACK for TLP P4, decrement undo_retrans to 0, undo (bug!)
+ receive cumulative ACK for P1-P4 (fast retransmit plugged real hole)
The fix: when we initialize undo machinery in tcp_init_undo(), if
there is a TLP retransmit in flight, then increment tp->undo_retrans
so that we make sure that we receive a DSACK corresponding to the TLP
retransmit, as well as DSACKs for all later normal retransmits, before
triggering a loss recovery undo. Note that we also have to move the
line that clears tp->tlp_high_seq for RTO recovery, so that upon RTO
we remember the tp->tlp_high_seq value until tcp_init_undo() and clear
it only afterward.
Also note that the bug dates back to the original 2013 TLP
implementation, commit 6ba8a3b19e76 ("tcp: Tail loss probe (TLP)").
However, this patch will only compile and work correctly with kernels
that have tp->tlp_retrans, which was added only in v5.8 in 2020 in
commit 76be93fc0702 ("tcp: allow at most one TLP probe per flight").
So we associate this fix with that later commit.
Fixes: 76be93fc0702 ("tcp: allow at most one TLP probe per flight")
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Kevin Yang <yyd@google.com>
Link: https://patch.msgid.link/20240703171246.1739561-1-ncardwell.sw@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit a6458ab7fd4f427d4f6f54380453ad255b7fde83 ]
In some production workloads we noticed that connections could
sometimes close extremely prematurely with ETIMEDOUT after
transmitting only 1 TLP and RTO retransmission (when we would normally
expect roughly tcp_retries2 = TCP_RETR2 = 15 RTOs before a connection
closes with ETIMEDOUT).
From tracing we determined that these workloads can suffer from a
scenario where in fast recovery, after some retransmits, a DSACK undo
can happen at a point where the scoreboard is totally clear (we have
retrans_out == sacked_out == lost_out == 0). In such cases, calling
tcp_try_keep_open() means that we do not execute any code path that
clears tp->retrans_stamp to 0. That means that tp->retrans_stamp can
remain erroneously set to the start time of the undone fast recovery,
even after the fast recovery is undone. If minutes or hours elapse,
and then a TLP/RTO/RTO sequence occurs, then the start_ts value in
retransmits_timed_out() (which is from tp->retrans_stamp) will be
erroneously ancient (left over from the fast recovery undone via
DSACKs). Thus this ancient tp->retrans_stamp value can cause the
connection to die very prematurely with ETIMEDOUT via
tcp_write_err().
The fix: we change DSACK undo in fast recovery (TCP_CA_Recovery) to
call tcp_try_to_open() instead of tcp_try_keep_open(). This ensures
that if no retransmits are in flight at the time of DSACK undo in fast
recovery then we properly zero retrans_stamp. Note that calling
tcp_try_to_open() is more consistent with other loss recovery
behavior, since normal fast recovery (CA_Recovery) and RTO recovery
(CA_Loss) both normally end when tp->snd_una meets or exceeds
tp->high_seq and then in tcp_fastretrans_alert() the "default" switch
case executes tcp_try_to_open(). Also note that by inspection this
change to call tcp_try_to_open() implies at least one other nice bug
fix, where now an ECE-marked DSACK that causes an undo will properly
invoke tcp_enter_cwr() rather than ignoring the ECE mark.
Fixes: c7d9d6a185a7 ("tcp: undo on DSACK during recovery")
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 3d501dd326fb1c73f1b8206d4c6e1d7b15c07e27 ]
This patch is based on a detailed report and ideas from Yepeng Pan
and Christian Rossow.
ACK seq validation is currently following RFC 5961 5.2 guidelines:
The ACK value is considered acceptable only if
it is in the range of ((SND.UNA - MAX.SND.WND) <= SEG.ACK <=
SND.NXT). All incoming segments whose ACK value doesn't satisfy the
above condition MUST be discarded and an ACK sent back. It needs to
be noted that RFC 793 on page 72 (fifth check) says: "If the ACK is a
duplicate (SEG.ACK < SND.UNA), it can be ignored. If the ACK
acknowledges something not yet sent (SEG.ACK > SND.NXT) then send an
ACK, drop the segment, and return". The "ignored" above implies that
the processing of the incoming data segment continues, which means
the ACK value is treated as acceptable. This mitigation makes the
ACK check more stringent since any ACK < SND.UNA wouldn't be
accepted, instead only ACKs that are in the range ((SND.UNA -
MAX.SND.WND) <= SEG.ACK <= SND.NXT) get through.
This can be refined for new (and possibly spoofed) flows,
by not accepting ACK for bytes that were never sent.
This greatly improves TCP security at a little cost.
I added a Fixes: tag to make sure this patch will reach stable trees,
even if the 'blamed' patch was adhering to the RFC.
tp->bytes_acked was added in linux-4.2
Following packetdrill test (courtesy of Yepeng Pan) shows
the issue at hand:
0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+0 bind(3, ..., ...) = 0
+0 listen(3, 1024) = 0
// ---------------- Handshake ------------------- //
// when window scale is set to 14 the window size can be extended to
// 65535 * (2^14) = 1073725440. Linux would accept an ACK packet
// with ack number in (Server_ISN+1-1073725440. Server_ISN+1)
// ,though this ack number acknowledges some data never
// sent by the server.
+0 < S 0:0(0) win 65535 <mss 1400,nop,wscale 14>
+0 > S. 0:0(0) ack 1 <...>
+0 < . 1:1(0) ack 1 win 65535
+0 accept(3, ..., ...) = 4
// For the established connection, we send an ACK packet,
// the ack packet uses ack number 1 - 1073725300 + 2^32,
// where 2^32 is used to wrap around.
// Note: we used 1073725300 instead of 1073725440 to avoid possible
// edge cases.
// 1 - 1073725300 + 2^32 = 3221241997
// Oops, old kernels happily accept this packet.
+0 < . 1:1001(1000) ack 3221241997 win 65535
// After the kernel fix the following will be replaced by a challenge ACK,
// and prior malicious frame would be dropped.
+0 > . 1:1(0) ack 1001
Fixes: 354e4aa391ed ("tcp: RFC 5961 5.2 Blind Data Injection Attack Mitigation")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Yepeng Pan <yepeng.pan@cispa.de>
Reported-by: Christian Rossow <rossow@cispa.de>
Acked-by: Neal Cardwell <ncardwell@google.com>
Link: https://lore.kernel.org/r/20231205161841.2702925-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit e326578a21414738de45f77badd332fb00bd0f58 ]
For passive TCP Fast Open sockets that had SYN/ACK timeout and did not
send more data in SYN_RECV, upon receiving the final ACK in 3WHS, the
congestion state may awkwardly stay in CA_Loss mode unless the CA state
was undone due to TCP timestamp checks. However, if
tcp_rcv_synrecv_state_fastopen() decides not to undo, then we should
enter CA_Open, because at that point we have received an ACK covering
the retransmitted SYNACKs. Currently, the icsk_ca_state is only set to
CA_Open after we receive an ACK for a data-packet. This is because
tcp_ack does not call tcp_fastretrans_alert (and tcp_process_loss) if
!prior_packets
Note that tcp_process_loss() calls tcp_try_undo_recovery(), so having
tcp_rcv_synrecv_state_fastopen() decide that if we're in CA_Loss we
should call tcp_try_undo_recovery() is consistent with that, and
low risk.
Fixes: dad8cea7add9 ("tcp: fix TFO SYNACK undo to avoid double-timestamp-undo")
Signed-off-by: Aananth V <aananthv@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit d2a0fc372aca561556e765d0a9ec365c7c12f0ad ]
This commit fix wrong RTO timeout when received SACK reneging.
When an ACK arrived pointing to a SACK reneging, tcp_check_sack_reneging()
will rearm the RTO timer for min(1/2*srtt, 10ms) into to the future.
But since the commit 62d9f1a6945b ("tcp: fix TLP timer not set when
CA_STATE changes from DISORDER to OPEN") merged, the tcp_set_xmit_timer()
is moved after tcp_fastretrans_alert()(which do the SACK reneging check),
so the RTO timeout will be overwrited by tcp_set_xmit_timer() with
icsk_rto instead of 1/2*srtt.
Here is a packetdrill script to check this bug:
0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+0 bind(3, ..., ...) = 0
+0 listen(3, 1) = 0
// simulate srtt to 100ms
+0 < S 0:0(0) win 32792 <mss 1000, sackOK,nop,nop,nop,wscale 7>
+0 > S. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK,nop,wscale 7>
+.1 < . 1:1(0) ack 1 win 1024
+0 accept(3, ..., ...) = 4
+0 write(4, ..., 10000) = 10000
+0 > P. 1:10001(10000) ack 1
// inject sack
+.1 < . 1:1(0) ack 1 win 257 <sack 1001:10001,nop,nop>
+0 > . 1:1001(1000) ack 1
// inject sack reneging
+.1 < . 1:1(0) ack 1001 win 257 <sack 9001:10001,nop,nop>
// we expect rto fired in 1/2*srtt (50ms)
+.05 > . 1001:2001(1000) ack 1
This fix remove the FLAG_SET_XMIT_TIMER from ack_flag when
tcp_check_sack_reneging() set RTO timer with 1/2*srtt to avoid
being overwrited later.
Fixes: 62d9f1a6945b ("tcp: fix TLP timer not set when CA_STATE changes from DISORDER to OPEN")
Signed-off-by: Fred Chen <fred.chenchen03@gmail.com>
Reviewed-by: Neal Cardwell <ncardwell@google.com>
Tested-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Neal Cardwell