[ Upstream commit 917d80d376 ]
Use nf_msecs_to_jiffies64 and nf_jiffies64_to_msecs as provided by
8e1102d5a1 ("netfilter: nf_tables: support timeouts larger than 23
days"), otherwise ruleset listing breaks.
Fixes: a8b1e36d0d ("netfilter: nft_dynset: fix element timeout for HZ != 1000")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit ffe8923f10 ]
Pablo Neira found that after recent update of xt_IDLETIMER the
iptables-nft tests sometimes show an error.
He tracked this down to the delayed cleanup used by nf_tables core:
del rule (transaction A)
add rule (transaction B)
Its possible that by time transaction B (both in same netns) runs,
the xt target destructor has not been invoked yet.
For native nft expressions this is no problem because all expressions
that have such side effects make sure these are handled from the commit
phase, rather than async cleanup.
For nft_compat however this isn't true.
Instead of forcing synchronous behaviour for nft_compat, keep track
of the number of outstanding destructor calls.
When we attempt to create a new expression, flush the cleanup worker
to make sure destructors have completed.
With lots of help from Pablo Neira.
Reported-by: Pablo Neira Ayso <pablo@netfilter.org>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 3c78e9e0d3 upstream.
This patch adds nft_flow_rule_set_addr_type() to set the address type
from the nft_payload expression accordingly.
If the address type is not set in the control dissector then a rule that
matches either on source or destination IP address does not work.
After this patch, nft hardware offload generates the flow dissector
configuration as tc-flower does to match on an IP address.
This patch has been also tested functionally to make sure packets are
filtered out by the NIC.
This is also getting the code aligned with the existing netfilter flow
offload infrastructure which is also setting the control dissector.
Fixes: c9626a2cbd ("netfilter: nf_tables: add hardware offload support")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 2867e1eac6 ]
When adding support for propagating ECT(1) marking in IP headers it seems I
suffered from endianness-confusion in the checksum update calculation: In
fact the ECN field is in the *lower* bits of the first 16-bit word of the
IP header when calculating in network byte order. This means that the
addition performed to update the checksum field was wrong; let's fix that.
Fixes: b723748750 ("tunnel: Propagate ECT(1) when decapsulating as recommended by RFC6040")
Reported-by: Jonathan Morton <chromatix99@gmail.com>
Tested-by: Pete Heist <pete@heistp.net>
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Link: https://lore.kernel.org/r/20201130183705.17540-1-toke@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 025cc2fb6a ]
tls_device_offload_cleanup_rx doesn't clear tls_ctx->netdev after
calling tls_dev_del if TLX TX offload is also enabled. Clearing
tls_ctx->netdev gets postponed until tls_device_gc_task. It leaves a
time frame when tls_device_down may get called and call tls_dev_del for
RX one extra time, confusing the driver, which may lead to a crash.
This patch corrects this racy behavior by adding a flag to prevent
tls_device_down from calling tls_dev_del the second time.
Fixes: e8f6979981 ("net/tls: Add generic NIC offload infrastructure")
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Link: https://lore.kernel.org/r/20201125221810.69870-1-saeedm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 8cf8821e15 ]
Commit 58956317c8 ("neighbor: Improve garbage collection")
guarantees neighbour table entries a five-second lifetime. Processes
which make heavy use of multicast can fill the neighour table with
multicast addresses in five seconds. At that point, neighbour entries
can't be GC-ed because they aren't five seconds old yet, the kernel
log starts to fill up with "neighbor table overflow!" messages, and
sends start to fail.
This patch allows multicast addresses to be thrown out before they've
lived out their five seconds. This makes room for non-multicast
addresses and makes messages to all addresses more reliable in these
circumstances.
Fixes: 58956317c8 ("neighbor: Improve garbage collection")
Signed-off-by: Jeff Dike <jdike@akamai.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20201113015815.31397-1-jdike@akamai.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 31cc578ae2 upstream.
This patch fixes the issue due to:
BUG: KASAN: slab-out-of-bounds in nft_flow_rule_create+0x622/0x6a2
net/netfilter/nf_tables_offload.c:40
Read of size 8 at addr ffff888103910b58 by task syz-executor227/16244
The error happens when expr->ops is accessed early on before performing the boundary check and after nft_expr_next() moves the expr to go out-of-bounds.
This patch checks the boundary condition before expr->ops that fixes the slab-out-of-bounds Read issue.
Add nft_expr_more() and use it to fix this problem.
Signed-off-by: Saeed Mirzamohammadi <saeed.mirzamohammadi@oracle.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 0d9826bc18 ]
Dump vlan tag and proto for the usual vlan offload case if the
NF_LOG_MACDECODE flag is set on. Without this information the logging is
misleading as there is no reference to the VLAN header.
[12716.993704] test: IN=veth0 OUT= MACSRC=86:6c:92:ea:d6:73 MACDST=0e:3b:eb:86:73:76 VPROTO=8100 VID=10 MACPROTO=0800 SRC=192.168.10.2 DST=172.217.168.163 LEN=52 TOS=0x00 PREC=0x00 TTL=64 ID=2548 DF PROTO=TCP SPT=55848 DPT=80 WINDOW=501 RES=0x00 ACK FIN URGP=0
[12721.157643] test: IN=veth0 OUT= MACSRC=86:6c:92:ea:d6:73 MACDST=0e:3b:eb:86:73:76 VPROTO=8100 VID=10 MACPROTO=0806 ARP HTYPE=1 PTYPE=0x0800 OPCODE=2 MACSRC=86:6c:92:ea:d6:73 IPSRC=192.168.10.2 MACDST=0e:3b:eb:86:73:76 IPDST=192.168.10.1
Fixes: 83e96d443b ("netfilter: log: split family specific code to nf_log_{ip,ip6,common}.c files")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 02a1b175b0 ]
Documentation/networking/ip-sysctl.txt:46 says:
ip_forward_use_pmtu - BOOLEAN
By default we don't trust protocol path MTUs while forwarding
because they could be easily forged and can lead to unwanted
fragmentation by the router.
You only need to enable this if you have user-space software
which tries to discover path mtus by itself and depends on the
kernel honoring this information. This is normally not the case.
Default: 0 (disabled)
Possible values:
0 - disabled
1 - enabled
Which makes it pretty clear that setting it to 1 is a potential
security/safety/DoS issue, and yet it is entirely reasonable to want
forwarded traffic to honour explicitly administrator configured
route mtus (instead of defaulting to device mtu).
Indeed, I can't think of a single reason why you wouldn't want to.
Since you configured a route mtu you probably know better...
It is pretty common to have a higher device mtu to allow receiving
large (jumbo) frames, while having some routes via that interface
(potentially including the default route to the internet) specify
a lower mtu.
Note that ipv6 forwarding uses device mtu unless the route is locked
(in which case it will use the route mtu).
This approach is not usable for IPv4 where an 'mtu lock' on a route
also has the side effect of disabling TCP path mtu discovery via
disabling the IPv4 DF (don't frag) bit on all outgoing frames.
I'm not aware of a way to lock a route from an IPv6 RA, so that also
potentially seems wrong.
Signed-off-by: Maciej Żenczykowski <maze@google.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Willem de Bruijn <willemb@google.com>
Cc: Lorenzo Colitti <lorenzo@google.com>
Cc: Sunmeet Gill (Sunny) <sgill@quicinc.com>
Cc: Vinay Paradkar <vparadka@qti.qualcomm.com>
Cc: Tyler Wear <twear@quicinc.com>
Cc: David Ahern <dsahern@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 339ddaa626 upstream.
Starting with the upgrade to v5.8-rc3, I've noticed I wasn't able to
connect to my Bluetooth headset properly anymore. While connecting to
the device would eventually succeed, bluetoothd seemed to be confused
about the current connection state where the state was flapping hence
and forth. Bisecting this issue led to commit 3ca44c16b0 (Bluetooth:
Consolidate encryption handling in hci_encrypt_cfm, 2020-05-19), which
refactored `hci_encrypt_cfm` to also handle updating the connection
state.
The commit in question changed the code to call `hci_connect_cfm` inside
`hci_encrypt_cfm` and to change the connection state. But with the
conversion, we now only update the connection state if a status was set
already. In fact, the reverse should be true: the status should be
updated if no status is yet set. So let's fix the isuse by reversing the
condition.
Fixes: 3ca44c16b0 ("Bluetooth: Consolidate encryption handling in hci_encrypt_cfm")
Signed-off-by: Patrick Steinhardt <ps@pks.im>
Acked-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 3ca44c16b0 upstream.
This makes hci_encrypt_cfm calls hci_connect_cfm in case the connection
state is BT_CONFIG so callers don't have to check the state.
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Cc: Hans-Christian Noren Egtvedt <hegtvedt@cisco.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit f19425641c upstream.
Only sockets will have the chan->data set to an actual sk, channels
like A2MP would have its own data which would likely cause a crash when
calling sk_filter, in order to fix this a new callback has been
introduced so channels can implement their own filtering if necessary.
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit e49d8c22f1 upstream.
All TC actions call tcf_idr_insert() for new action at the end
of their ->init(), so we can actually move it to a central place
in tcf_action_init_1().
And once the action is inserted into the global IDR, other parallel
process could free it immediately as its refcnt is still 1, so we can
not fail after this, we need to move it after the goto action
validation to avoid handling the failure case after insertion.
This is found during code review, is not directly triggered by syzbot.
And this prepares for the next patch.
Cc: Vlad Buslov <vladbu@mellanox.com>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 91a46c6d1b ]
XFRMA_REPLAY_ESN_VAL was not cloned completely from the old to the new.
Migrate this attribute during XFRMA_MSG_MIGRATE
v1->v2:
- move curleft cloning to a separate patch
Fixes: af2f464e32 ("xfrm: Assign esn pointers when cloning a state")
Signed-off-by: Antony Antony <antony.antony@secunet.com>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit fe81d9f618 ]
When calculating ancestor_size with IPv6 enabled, simply using
sizeof(struct ipv6_pinfo) doesn't account for extra bytes needed for
alignment in the struct sctp6_sock. On x86, there aren't any extra
bytes, but on ARM the ipv6_pinfo structure is aligned on an 8-byte
boundary so there were 4 pad bytes that were omitted from the
ancestor_size calculation. This would lead to corruption of the
pd_lobby pointers, causing an oops when trying to free the sctp
structure on socket close.
Fixes: 636d25d557 ("sctp: not copy sctp_sock pd_lobby in sctp_copy_descendant")
Signed-off-by: Henry Ptasinski <hptasinski@google.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 1869e226a7 ]
flowi4_multipath_hash was added by the commit referenced below for
tunnels. Unfortunately, the patch did not initialize the new field
for several fast path lookups that do not initialize the entire flow
struct to 0. Fix those locations. Currently, flowi4_multipath_hash
is random garbage and affects the hash value computed by
fib_multipath_hash for multipath selection.
Fixes: 24ba14406c ("route: Add multipath_hash in flowi_common to make user-define hash")
Signed-off-by: David Ahern <dsahern@gmail.com>
Cc: wenxu <wenxu@ucloud.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 1e105e6afa ]
Following bug was reported via irc:
nft list ruleset
set knock_candidates_ipv4 {
type ipv4_addr . inet_service
size 65535
elements = { 127.0.0.1 . 123,
127.0.0.1 . 123 }
}
..
udp dport 123 add @knock_candidates_ipv4 { ip saddr . 123 }
udp dport 123 add @knock_candidates_ipv4 { ip saddr . udp dport }
It should not have been possible to add a duplicate set entry.
After some debugging it turned out that the problem is the immediate
value (123) in the second-to-last rule.
Concatenations use 32bit registers, i.e. the elements are 8 bytes each,
not 6 and it turns out the kernel inserted
inet firewall @knock_candidates_ipv4
element 0100007f ffff7b00 : 0 [end]
element 0100007f 00007b00 : 0 [end]
Note the non-zero upper bits of the first element. It turns out that
nft_immediate doesn't zero the destination register, but this is needed
when the length isn't a multiple of 4.
Furthermore, the zeroing in nft_payload is broken. We can't use
[len / 4] = 0 -- if len is a multiple of 4, index is off by one.
Skip zeroing in this case and use a conditional instead of (len -1) / 4.
Fixes: 49499c3e6e ("netfilter: nf_tables: switch registers to 32 bit addressing")
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 1d4adfaf65 ]
Fix rxrpc_kernel_get_srtt() to indicate the validity of the returned
smoothed RTT. If we haven't had any valid samples yet, the SRTT isn't
useful.
Fixes: c410bf0193 ("rxrpc: Fix the excessive initial retransmission timeout")
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit d9539752d2 upstream.
Add missed sock updates to compat path via a new helper, which will be
used more in coming patches. (The net/core/scm.c code is left as-is here
to assist with -stable backports for the compat path.)
Cc: Christoph Hellwig <hch@lst.de>
Cc: Sargun Dhillon <sargun@sargun.me>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: stable@vger.kernel.org
Fixes: 48a87cc26c ("net: netprio: fd passed in SCM_RIGHTS datagram not set correctly")
Fixes: d84295067f ("net: net_cls: fd passed in SCM_RIGHTS datagram not set correctly")
Acked-by: Christian Brauner <christian.brauner@ubuntu.com>
Signed-off-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 62ffc589ab ]
Refactor the fastreuse update code in inet_csk_get_port into a small
helper function that can be called from other places.
Acked-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Tim Froidcoeur <tim.froidcoeur@tessares.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit f19008e676 ]
When TFO keys are read back on big endian systems either via the global
sysctl interface or via getsockopt() using TCP_FASTOPEN_KEY, the values
don't match what was written.
For example, on s390x:
# echo "1-2-3-4" > /proc/sys/net/ipv4/tcp_fastopen_key
# cat /proc/sys/net/ipv4/tcp_fastopen_key
02000000-01000000-04000000-03000000
Instead of:
# cat /proc/sys/net/ipv4/tcp_fastopen_key
00000001-00000002-00000003-00000004
Fix this by converting to the correct endianness on read. This was
reported by Colin Ian King when running the 'tcp_fastopen_backup_key' net
selftest on s390x, which depends on the read value matching what was
written. I've confirmed that the test now passes on big and little endian
systems.
Signed-off-by: Jason Baron <jbaron@akamai.com>
Fixes: 438ac88009 ("net: fastopen: robustness and endianness fixes for SipHash")
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Eric Dumazet <edumazet@google.com>
Reported-and-tested-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit f0a5e4d7a5 ]
YangYuxi is reporting that connection reuse
is causing one-second delay when SYN hits
existing connection in TIME_WAIT state.
Such delay was added to give time to expire
both the IPVS connection and the corresponding
conntrack. This was considered a rare case
at that time but it is causing problem for
some environments such as Kubernetes.
As nf_conntrack_tcp_packet() can decide to
release the conntrack in TIME_WAIT state and
to replace it with a fresh NEW conntrack, we
can use this to allow rescheduling just by
tuning our check: if the conntrack is
confirmed we can not schedule it to different
real server and the one-second delay still
applies but if new conntrack was created,
we are free to select new real server without
any delays.
YangYuxi lists some of the problem reports:
- One second connection delay in masquerading mode:
https://marc.info/?t=151683118100004&r=1&w=2
- IPVS low throughput #70747
https://github.com/kubernetes/kubernetes/issues/70747
- Apache Bench can fill up ipvs service proxy in seconds #544https://github.com/cloudnativelabs/kube-router/issues/544
- Additional 1s latency in `host -> service IP -> pod`
https://github.com/kubernetes/kubernetes/issues/90854
Fixes: f719e3754e ("ipvs: drop first packet to redirect conntrack")
Co-developed-by: YangYuxi <yx.atom1@gmail.com>
Signed-off-by: YangYuxi <yx.atom1@gmail.com>
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Reviewed-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 8c0de6e96c ]
IPV6_ADDRFORM causes resource leaks when converting an IPv6 socket
to IPv4, particularly struct ipv6_ac_socklist. Similar to
struct ipv6_mc_socklist, we should just close it on this path.
This bug can be easily reproduced with the following C program:
#include <stdio.h>
#include <string.h>
#include <sys/types.h>
#include <sys/socket.h>
#include <arpa/inet.h>
int main()
{
int s, value;
struct sockaddr_in6 addr;
struct ipv6_mreq m6;
s = socket(AF_INET6, SOCK_DGRAM, 0);
addr.sin6_family = AF_INET6;
addr.sin6_port = htons(5000);
inet_pton(AF_INET6, "::ffff:192.168.122.194", &addr.sin6_addr);
connect(s, (struct sockaddr *)&addr, sizeof(addr));
inet_pton(AF_INET6, "fe80::AAAA", &m6.ipv6mr_multiaddr);
m6.ipv6mr_interface = 5;
setsockopt(s, SOL_IPV6, IPV6_JOIN_ANYCAST, &m6, sizeof(m6));
value = AF_INET;
setsockopt(s, SOL_IPV6, IPV6_ADDRFORM, &value, sizeof(value));
close(s);
return 0;
}
Reported-by: ch3332xr@gmail.com
Fixes: 1da177e4c3 ("Linux-2.6.12-rc2")
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 101dde4207 ]
The commits "xfrm: Move dst->path into struct xfrm_dst"
and "net: Create and use new helper xfrm_dst_child()."
changed xfrm bundle handling under the assumption
that xdst->path and dst->child are not a NULL pointer
only if dst->xfrm is not a NULL pointer. That is true
with one exception. If the xfrm hold queue is used
to wait until a SA is installed by the key manager,
we create a dummy bundle without a valid dst->xfrm
pointer. The current xfrm bundle handling crashes
in that case. Fix this by extending the NULL check
of dst->xfrm with a test of the DST_XFRM_QUEUE flag.
Fixes: 0f6c480f23 ("xfrm: Move dst->path into struct xfrm_dst")
Fixes: b92cf4aab8 ("net: Create and use new helper xfrm_dst_child().")
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 4f47e8ab6a ]
In commit ed17b8d377 ("xfrm: fix a warning in xfrm_policy_insert_list"),
it would take 'priority' to make a policy unique, and allow duplicated
policies with different 'priority' to be added, which is not expected
by userland, as Tobias reported in strongswan.
To fix this duplicated policies issue, and also fix the issue in
commit ed17b8d377 ("xfrm: fix a warning in xfrm_policy_insert_list"),
when doing add/del/get/update on user interfaces, this patch is to change
to look up a policy with both mark and mask by doing:
mark.v == pol->mark.v && mark.m == pol->mark.m
and leave the check:
(mark & pol->mark.m) == pol->mark.v
for tx/rx path only.
As the userland expects an exact mark and mask match to manage policies.
v1->v2:
- make xfrm_policy_mark_match inline and fix the changelog as
Tobias suggested.
Fixes: 295fae5688 ("xfrm: Allow user space manipulation of SPD mark")
Fixes: ed17b8d377 ("xfrm: fix a warning in xfrm_policy_insert_list")
Reported-by: Tobias Brunner <tobias@strongswan.org>
Tested-by: Tobias Brunner <tobias@strongswan.org>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit d7bf2ebebc ]
There are a couple of places in net/sched/ that check skb->protocol and act
on the value there. However, in the presence of VLAN tags, the value stored
in skb->protocol can be inconsistent based on whether VLAN acceleration is
enabled. The commit quoted in the Fixes tag below fixed the users of
skb->protocol to use a helper that will always see the VLAN ethertype.
However, most of the callers don't actually handle the VLAN ethertype, but
expect to find the IP header type in the protocol field. This means that
things like changing the ECN field, or parsing diffserv values, stops
working if there's a VLAN tag, or if there are multiple nested VLAN
tags (QinQ).
To fix this, change the helper to take an argument that indicates whether
the caller wants to skip the VLAN tags or not. When skipping VLAN tags, we
make sure to skip all of them, so behaviour is consistent even in QinQ
mode.
To make the helper usable from the ECN code, move it to if_vlan.h instead
of pkt_sched.h.
v3:
- Remove empty lines
- Move vlan variable definitions inside loop in skb_protocol()
- Also use skb_protocol() helper in IP{,6}_ECN_decapsulate() and
bpf_skb_ecn_set_ce()
v2:
- Use eth_type_vlan() helper in skb_protocol()
- Also fix code that reads skb->protocol directly
- Change a couple of 'if/else if' statements to switch constructs to avoid
calling the helper twice
Reported-by: Ilya Ponetayev <i.ponetaev@ndmsystems.com>
Fixes: d8b9605d26 ("net: sched: fix skb->protocol use in case of accelerated vlan path")
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 1e82a62fec ]
A potential deadlock can occur during registering or unregistering a
new generic netlink family between the main nl_table_lock and the
cb_lock where each thread wants the lock held by the other, as
demonstrated below.
1) Thread 1 is performing a netlink_bind() operation on a socket. As part
of this call, it will call netlink_lock_table(), incrementing the
nl_table_users count to 1.
2) Thread 2 is registering (or unregistering) a genl_family via the
genl_(un)register_family() API. The cb_lock semaphore will be taken for
writing.
3) Thread 1 will call genl_bind() as part of the bind operation to handle
subscribing to GENL multicast groups at the request of the user. It will
attempt to take the cb_lock semaphore for reading, but it will fail and
be scheduled away, waiting for Thread 2 to finish the write.
4) Thread 2 will call netlink_table_grab() during the (un)registration
call. However, as Thread 1 has incremented nl_table_users, it will not
be able to proceed, and both threads will be stuck waiting for the
other.
genl_bind() is a noop, unless a genl_family implements the mcast_bind()
function to handle setting up family-specific multicast operations. Since
no one in-tree uses this functionality as Cong pointed out, simply removing
the genl_bind() function will remove the possibility for deadlock, as there
is no attempt by Thread 1 above to take the cb_lock semaphore.
Fixes: c380d9a7af ("genetlink: pass multicast bind/unbind to families")
Suggested-by: Cong Wang <xiyou.wangcong@gmail.com>
Acked-by: Johannes Berg <johannes.berg@intel.com>
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Sean Tranchetti <stranche@codeaurora.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 94579ac3f6 ]
During IPsec performance testing, we see bad ICMP checksum. The error packet
has duplicated ESP trailer due to double validate_xmit_xfrm calls. The first call
is from ip_output, but the packet cannot be sent because
netif_xmit_frozen_or_stopped is true and the packet gets dev_requeue_skb. The second
call is from NET_TX softirq. However after the first call, the packet already
has the ESP trailer.
Fix by marking the skb with XFRM_XMIT bit after the packet is handled by
validate_xmit_xfrm to avoid duplicate ESP trailer insertion.
Fixes: f6e27114a6 ("net: Add a xfrm validate function to validate_xmit_skb")
Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Reviewed-by: Boris Pismenny <borisp@mellanox.com>
Reviewed-by: Raed Salem <raeds@mellanox.com>
Reviewed-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 471e39df96 ]
If a socket is set ipv6only, it will still send IPv4 addresses in the
INIT and INIT_ACK packets. This potentially misleads the peer into using
them, which then would cause association termination.
The fix is to not add IPv4 addresses to ipv6only sockets.
Fixes: 1da177e4c3 ("Linux-2.6.12-rc2")
Reported-by: Corey Minyard <cminyard@mvista.com>
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Tested-by: Corey Minyard <cminyard@mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 41b14fb872 ]
Clearing the sock TX queue in sk_set_socket() might cause unexpected
out-of-order transmit when called from sock_orphan(), as outstanding
packets can pick a different TX queue and bypass the ones already queued.
This is undesired in general. More specifically, it breaks the in-order
scheduling property guarantee for device-offloaded TLS sockets.
Remove the call to sk_tx_queue_clear() in sk_set_socket(), and add it
explicitly only where needed.
Fixes: e022f0b4a0 ("net: Introduce sk_tx_queue_mapping")
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Boris Pismenny <borisp@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit e91de6afa8 ]
KTLS uses a stream parser to collect TLS messages and send them to
the upper layer tls receive handler. This ensures the tls receiver
has a full TLS header to parse when it is run. However, when a
socket has BPF_SK_SKB_STREAM_VERDICT program attached before KTLS
is enabled we end up with two stream parsers running on the same
socket.
The result is both try to run on the same socket. First the KTLS
stream parser runs and calls read_sock() which will tcp_read_sock
which in turn calls tcp_rcv_skb(). This dequeues the skb from the
sk_receive_queue. When this is done KTLS code then data_ready()
callback which because we stacked KTLS on top of the bpf stream
verdict program has been replaced with sk_psock_start_strp(). This
will in turn kick the stream parser again and eventually do the
same thing KTLS did above calling into tcp_rcv_skb() and dequeuing
a skb from the sk_receive_queue.
At this point the data stream is broke. Part of the stream was
handled by the KTLS side some other bytes may have been handled
by the BPF side. Generally this results in either missing data
or more likely a "Bad Message" complaint from the kTLS receive
handler as the BPF program steals some bytes meant to be in a
TLS header and/or the TLS header length is no longer correct.
We've already broke the idealized model where we can stack ULPs
in any order with generic callbacks on the TX side to handle this.
So in this patch we do the same thing but for RX side. We add
a sk_psock_strp_enabled() helper so TLS can learn a BPF verdict
program is running and add a tls_sw_has_ctx_rx() helper so BPF
side can learn there is a TLS ULP on the socket.
Then on BPF side we omit calling our stream parser to avoid
breaking the data stream for the KTLS receiver. Then on the
KTLS side we call BPF_SK_SKB_STREAM_VERDICT once the KTLS
receiver is done with the packet but before it posts the
msg to userspace. This gives us symmetry between the TX and
RX halfs and IMO makes it usable again. On the TX side we
process packets in this order BPF -> TLS -> TCP and on
the receive side in the reverse order TCP -> TLS -> BPF.
Discovered while testing OpenSSL 3.0 Alpha2.0 release.
Fixes: d829e9c411 ("tls: convert to generic sk_msg interface")
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/159079361946.5745.605854335665044485.stgit@john-Precision-5820-Tower
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit c96b6acc8f ]
There are some memory leaks in dccp_init() and dccp_fini().
In dccp_fini() and the error handling path in dccp_init(), free lhash2
is missing. Add inet_hashinfo2_free_mod() to do it.
If inet_hashinfo2_init_mod() failed in dccp_init(),
percpu_counter_destroy() should be called to destroy dccp_orphan_count.
It need to goto out_free_percpu when inet_hashinfo2_init_mod() failed.
Fixes: c92c81df93 ("net: dccp: fix kernel crash on module load")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Wang Hai <wanghai38@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 1fd1c768f3 upstream.
Similar to the last path, need to fix fib_info_nh_uses_dev for
external nexthops to avoid referencing multiple nh_grp structs.
Move the device check in fib_info_nh_uses_dev to a helper and
create a nexthop version that is called if the fib_info uses an
external nexthop.
Fixes: 430a049190 ("nexthop: Add support for nexthop groups")
Signed-off-by: David Ahern <dsahern@gmail.com>
Acked-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 0b5e2e3973 upstream.
I got too fancy consolidating checks on multipath type. The result
is that path lookups can access 2 different nh_grp structs as exposed
by Nik's torture tests. Expand nexthop_is_multipath within nexthop.h to
avoid multiple, nh_grp dereferences and make decisions based on the
consistent struct.
Only 2 places left using nexthop_is_multipath are within IPv6, both
only check that the nexthop is a multipath for a branching decision
which are acceptable.
Fixes: 430a049190 ("nexthop: Add support for nexthop groups")
Signed-off-by: David Ahern <dsahern@gmail.com>
Acked-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 90f33bffa3 upstream.
We must avoid modifying published nexthop groups while they might be
in use, otherwise we might see NULL ptr dereferences. In order to do
that we allocate 2 nexthoup group structures upon nexthop creation
and swap between them when we have to delete an entry. The reason is
that we can't fail nexthop group removal, so we can't handle allocation
failure thus we move the extra allocation on creation where we can
safely fail and return ENOMEM.
Fixes: 430a049190 ("nexthop: Add support for nexthop groups")
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 0cada33241 ]
tls_sw_recvmsg() and tls_decrypt_done() can be run concurrently.
// tls_sw_recvmsg()
if (atomic_read(&ctx->decrypt_pending))
crypto_wait_req(-EINPROGRESS, &ctx->async_wait);
else
reinit_completion(&ctx->async_wait.completion);
//tls_decrypt_done()
pending = atomic_dec_return(&ctx->decrypt_pending);
if (!pending && READ_ONCE(ctx->async_notify))
complete(&ctx->async_wait.completion);
Consider the scenario tls_decrypt_done() is about to run complete()
if (!pending && READ_ONCE(ctx->async_notify))
and tls_sw_recvmsg() reads decrypt_pending == 0, does reinit_completion(),
then tls_decrypt_done() runs complete(). This sequence of execution
results in wrong completion. Consequently, for next decrypt request,
it will not wait for completion, eventually on connection close, crypto
resources freed, there is no way to handle pending decrypt response.
This race condition can be avoided by having atomic_read() mutually
exclusive with atomic_dec_return(),complete().Intoduced spin lock to
ensure the mutual exclution.
Addressed similar problem in tx direction.
v1->v2:
- More readable commit message.
- Corrected the lock to fix new race scenario.
- Removed barrier which is not needed now.
Fixes: a42055e8d2 ("net/tls: Add support for async encryption of records for performance")
Signed-off-by: Vinay Kumar Yadav <vinay.yadav@chelsio.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit b15e62631c ]
When a new action is installed, firstuse field of 'tcf_t' is explicitly set
to 0. Value of zero means "new action, not yet used"; as a packet hits the
action, 'firstuse' is stamped with the current jiffies value.
tcf_tm_dump() should return 0 for firstuse if action has not yet been hit.
Fixes: 48d8ee1694 ("net sched actions: aggregate dumping of actions timeinfo")
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Roman Mashak <mrv@mojatatu.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 41b4bd986f ]
In case we can't find a ->dumpit callback for the requested
(family,type) pair, we fall back to (PF_UNSPEC,type). In effect, we're
in the same situation as if userspace had requested a PF_UNSPEC
dump. For RTM_GETROUTE, that handler is rtnl_dump_all, which calls all
the registered RTM_GETROUTE handlers.
The requested table id may or may not exist for all of those
families. commit ae677bbb44 ("net: Don't return invalid table id
error when dumping all families") fixed the problem when userspace
explicitly requests a PF_UNSPEC dump, but missed the fallback case.
For example, when we pass ipv6.disable=1 to a kernel with
CONFIG_IP_MROUTE=y and CONFIG_IP_MROUTE_MULTIPLE_TABLES=y,
the (PF_INET6, RTM_GETROUTE) handler isn't registered, so we end up in
rtnl_dump_all, and listing IPv6 routes will unexpectedly print:
# ip -6 r
Error: ipv4: MR table does not exist.
Dump terminated
commit ae677bbb44 introduced the dump_all_families variable, which
gets set when userspace requests a PF_UNSPEC dump. However, we can't
simply set the family to PF_UNSPEC in rtnetlink_rcv_msg in the
fallback case to get dump_all_families == true, because some messages
types (for example RTM_GETRULE and RTM_GETNEIGH) only register the
PF_UNSPEC handler and use the family to filter in the kernel what is
dumped to userspace. We would then export more entries, that userspace
would have to filter. iproute does that, but other programs may not.
Instead, this patch removes dump_all_families and updates the
RTM_GETROUTE handlers to check if the family that is being dumped is
their own. When it's not, which covers both the intentional PF_UNSPEC
dumps (as dump_all_families did) and the fallback case, ignore the
missing table id error.
Fixes: cb167893f4 ("net: Plumb support for filtering ipv4 and ipv6 multicast route dumps")
Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Reviewed-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit c410bf0193 upstream.
rxrpc currently uses a fixed 4s retransmission timeout until the RTT is
sufficiently sampled. This can cause problems with some fileservers with
calls to the cache manager in the afs filesystem being dropped from the
fileserver because a packet goes missing and the retransmission timeout is
greater than the call expiry timeout.
Fix this by:
(1) Copying the RTT/RTO calculation code from Linux's TCP implementation
and altering it to fit rxrpc.
(2) Altering the various users of the RTT to make use of the new SRTT
value.
(3) Replacing the use of rxrpc_resend_timeout to use the calculated RTO
value instead (which is needed in jiffies), along with a backoff.
Notes:
(1) rxrpc provides RTT samples by matching the serial numbers on outgoing
DATA packets that have the RXRPC_REQUEST_ACK set and PING ACK packets
against the reference serial number in incoming REQUESTED ACK and
PING-RESPONSE ACK packets.
(2) Each packet that is transmitted on an rxrpc connection gets a new
per-connection serial number, even for retransmissions, so an ACK can
be cross-referenced to a specific trigger packet. This allows RTT
information to be drawn from retransmitted DATA packets also.
(3) rxrpc maintains the RTT/RTO state on the rxrpc_peer record rather than
on an rxrpc_call because many RPC calls won't live long enough to
generate more than one sample.
(4) The calculated SRTT value is in units of 8ths of a microsecond rather
than nanoseconds.
The (S)RTT and RTO values are displayed in /proc/net/rxrpc/peers.
Fixes: 17926a7932 ([AF_RXRPC]: Provide secure RxRPC sockets for use by userspace and kernel both"")
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 1cd9b3abf5 ]
In net/Kconfig, NET_DEVLINK implies NET_DROP_MONITOR.
The original behavior of the 'imply' keyword prevents NET_DROP_MONITOR
from being 'm' when NET_DEVLINK=y.
With the planned Kconfig change that relaxes the 'imply', the
combination of NET_DEVLINK=y and NET_DROP_MONITOR=m would be allowed.
Use IS_REACHABLE() to avoid the vmlinux link error for this case.
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 2c407aca64 ]
gcc-10 warns around a suspicious access to an empty struct member:
net/netfilter/nf_conntrack_core.c: In function '__nf_conntrack_alloc':
net/netfilter/nf_conntrack_core.c:1522:9: warning: array subscript 0 is outside the bounds of an interior zero-length array 'u8[0]' {aka 'unsigned char[0]'} [-Wzero-length-bounds]
1522 | memset(&ct->__nfct_init_offset[0], 0,
| ^~~~~~~~~~~~~~~~~~~~~~~~~~
In file included from net/netfilter/nf_conntrack_core.c:37:
include/net/netfilter/nf_conntrack.h:90:5: note: while referencing '__nfct_init_offset'
90 | u8 __nfct_init_offset[0];
| ^~~~~~~~~~~~~~~~~~
The code is correct but a bit unusual. Rework it slightly in a way that
does not trigger the warning, using an empty struct instead of an empty
array. There are probably more elegant ways to do this, but this is the
smallest change.
Fixes: c41884ce05 ("netfilter: conntrack: avoid zeroing timer")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 24adbc1676 ]
We autotune rcvbuf whenever SO_RCVLOWAT is set to account for 100%
overhead in tcp_set_rcvlowat()
This works well when skb->len/skb->truesize ratio is bigger than 0.5
But if we receive packets with small MSS, we can end up in a situation
where not enough bytes are available in the receive queue to satisfy
RCVLOWAT setting.
As our sk_rcvbuf limit is hit, we send zero windows in ACK packets,
preventing remote peer from sending more data.
Even autotuning does not help, because it only triggers at the time
user process drains the queue. If no EPOLLIN is generated, this
can not happen.
Note poll() has a similar issue, after commit
c7004482e8 ("tcp: Respect SO_RCVLOWAT in tcp_poll().")
Fixes: 03f45c883c ("tcp: avoid extra wakeups for SO_RCVLOWAT users")
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>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit a7df4870d7 ]
When we tell kernel to dump filters from root (ffff:ffff),
those filters on ingress (ffff:0000) are matched, but their
true parents must be dumped as they are. However, kernel
dumps just whatever we tell it, that is either ffff:ffff
or ffff:0000:
$ nl-cls-list --dev=dummy0 --parent=root
cls basic dev dummy0 id none parent root prio 49152 protocol ip match-all
cls basic dev dummy0 id :1 parent root prio 49152 protocol ip match-all
$ nl-cls-list --dev=dummy0 --parent=ffff:
cls basic dev dummy0 id none parent ffff: prio 49152 protocol ip match-all
cls basic dev dummy0 id :1 parent ffff: prio 49152 protocol ip match-all
This is confusing and misleading, more importantly this is
a regression since 4.15, so the old behavior must be restored.
And, when tc filters are installed on a tc class, the parent
should be the classid, rather than the qdisc handle. Commit
edf6711c98 ("net: sched: remove classid and q fields from tcf_proto")
removed the classid we save for filters, we can just restore
this classid in tcf_block.
Steps to reproduce this:
ip li set dev dummy0 up
tc qd add dev dummy0 ingress
tc filter add dev dummy0 parent ffff: protocol arp basic action pass
tc filter show dev dummy0 root
Before this patch:
filter protocol arp pref 49152 basic
filter protocol arp pref 49152 basic handle 0x1
action order 1: gact action pass
random type none pass val 0
index 1 ref 1 bind 1
After this patch:
filter parent ffff: protocol arp pref 49152 basic
filter parent ffff: protocol arp pref 49152 basic handle 0x1
action order 1: gact action pass
random type none pass val 0
index 1 ref 1 bind 1
Fixes: a10fa20101 ("net: sched: propagate q and parent from caller down to tcf_fill_node")
Fixes: edf6711c98 ("net: sched: remove classid and q fields from tcf_proto")
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit b723748750 ]
RFC 6040 recommends propagating an ECT(1) mark from an outer tunnel header
to the inner header if that inner header is already marked as ECT(0). When
RFC 6040 decapsulation was implemented, this case of propagation was not
added. This simply appears to be an oversight, so let's fix that.
Fixes: eccc1bb8d4 ("tunnel: drop packet if ECN present with not-ECT")
Reported-by: Bob Briscoe <ietf@bobbriscoe.net>
Reported-by: Olivier Tilmans <olivier.tilmans@nokia-bell-labs.com>
Cc: Dave Taht <dave.taht@gmail.com>
Cc: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 8f34e53b60 ]
Nik reported a bug with pcpu dst cache when nexthop objects are
used illustrated by the following:
$ ip netns add foo
$ ip -netns foo li set lo up
$ ip -netns foo addr add 2001:db8:11::1/128 dev lo
$ ip netns exec foo sysctl net.ipv6.conf.all.forwarding=1
$ ip li add veth1 type veth peer name veth2
$ ip li set veth1 up
$ ip addr add 2001:db8:10::1/64 dev veth1
$ ip li set dev veth2 netns foo
$ ip -netns foo li set veth2 up
$ ip -netns foo addr add 2001:db8:10::2/64 dev veth2
$ ip -6 nexthop add id 100 via 2001:db8:10::2 dev veth1
$ ip -6 route add 2001:db8:11::1/128 nhid 100
Create a pcpu entry on cpu 0:
$ taskset -a -c 0 ip -6 route get 2001:db8:11::1
Re-add the route entry:
$ ip -6 ro del 2001:db8:11::1
$ ip -6 route add 2001:db8:11::1/128 nhid 100
Route get on cpu 0 returns the stale pcpu:
$ taskset -a -c 0 ip -6 route get 2001:db8:11::1
RTNETLINK answers: Network is unreachable
While cpu 1 works:
$ taskset -a -c 1 ip -6 route get 2001:db8:11::1
2001:db8:11::1 from :: via 2001:db8:10::2 dev veth1 src 2001:db8:10::1 metric 1024 pref medium
Conversion of FIB entries to work with external nexthop objects
missed an important difference between IPv4 and IPv6 - how dst
entries are invalidated when the FIB changes. IPv4 has a per-network
namespace generation id (rt_genid) that is bumped on changes to the FIB.
Checking if a dst_entry is still valid means comparing rt_genid in the
rtable to the current value of rt_genid for the namespace.
IPv6 also has a per network namespace counter, fib6_sernum, but the
count is saved per fib6_node. With the per-node counter only dst_entries
based on fib entries under the node are invalidated when changes are
made to the routes - limiting the scope of invalidations. IPv6 uses a
reference in the rt6_info, 'from', to track the corresponding fib entry
used to create the dst_entry. When validating a dst_entry, the 'from'
is used to backtrack to the fib6_node and check the sernum of it to the
cookie passed to the dst_check operation.
With the inline format (nexthop definition inline with the fib6_info),
dst_entries cached in the fib6_nh have a 1:1 correlation between fib
entries, nexthop data and dst_entries. With external nexthops, IPv6
looks more like IPv4 which means multiple fib entries across disparate
fib6_nodes can all reference the same fib6_nh. That means validation
of dst_entries based on external nexthops needs to use the IPv4 format
- the per-network namespace counter.
Add sernum to rt6_info and set it when creating a pcpu dst entry. Update
rt6_get_cookie to return sernum if it is set and update dst_check for
IPv6 to look for sernum set and based the check on it if so. Finally,
rt6_get_pcpu_route needs to validate the cached entry before returning
a pcpu entry (similar to the rt_cache_valid calls in __mkroute_input and
__mkroute_output for IPv4).
This problem only affects routes using the new, external nexthops.
Thanks to the kbuild test robot for catching the IS_ENABLED needed
around rt_genid_ipv6 before I sent this out.
Fixes: 5b98324ebe ("ipv6: Allow routes to use nexthop objects")
Reported-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David Ahern <dsahern@kernel.org>
Reviewed-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Tested-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Pablo Neira Ayuso
Florian Westphal
Toke Høiland-Jørgensen
Jamie Iles
Maxim Mikityanskiy
Yi-Hung Wei
Jeff Dike
Saeed Mirzamohammadi
Maciej Żenczykowski
Luiz Augusto von Dentz
Patrick Steinhardt
Cong Wang
Antony Antony
Eric Dumazet
Henry Ptasinski
David Ahern
David Howells
Kees Cook
Tim Froidcoeur
Jason Baron
Julian Anastasov
Steffen Klassert
Xin Long
Martin Varghese
Sean Tranchetti
Huy Nguyen
Marcelo Ricardo Leitner
Tariq Toukan
John Fastabend
Wang Hai
Nikolay Aleksandrov
Vinay Kumar Yadav
Roman Mashak
Sabrina Dubroca
Masahiro Yamada
Arnd Bergmann
David Ahern