With indirect blocks, a driver can register for callbacks from a device
that is does not 'own', for example, a tunnel device. When registering to
or unregistering from a new device, a callback is triggered to generate
a bind/unbind event. This, in turn, allows the driver to receive any
existing rules or to properly clean up installed rules.
When first added, it was assumed that all indirect block registrations
would be for ingress offloads. However, the NFP driver can, in some
instances, support clsact qdisc binds for egress offload.
Change the name of the indirect block callback command in flow_offload to
remove the 'ingress' identifier from it. While this does not change
functionality, a follow up patch will implement a more more generic
callback than just those currently just supporting ingress offload.
Fixes: 4d12ba4278 ("nfp: flower: allow offloading of matches on 'internal' ports")
Signed-off-by: John Hurley <john.hurley@netronome.com>
Acked-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull networking fixes from David Miller:
"This is mostly to fix the iwlwifi regression:
1) Flush GRO state properly in iwlwifi driver, from Alexander Lobakin.
2) Validate TIPC link name with properly length macro, from John
Rutherford.
3) Fix completion init and device query timeouts in ibmvnic, from
Thomas Falcon.
4) Fix SKB size calculation for netlink messages in psample, from
Nikolay Aleksandrov.
5) Similar kind of fix for OVS flow dumps, from Paolo Abeni.
6) Handle queue allocation failure unwind properly in gve driver, we
could try to release pages we didn't allocate. From Jeroen de
Borst.
7) Serialize TX queue SKB list accesses properly in mscc ocelot
driver. From Yangbo Lu"
* git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net:
net: usb: aqc111: Use the correct style for SPDX License Identifier
net: phy: Use the correct style for SPDX License Identifier
net: wireless: intel: iwlwifi: fix GRO_NORMAL packet stalling
net: mscc: ocelot: use skb queue instead of skbs list
net: mscc: ocelot: avoid incorrect consuming in skbs list
gve: Fix the queue page list allocated pages count
net: inet_is_local_reserved_port() port arg should be unsigned short
openvswitch: fix flow command message size
net: phy: dp83869: Fix return paths to return proper values
net: psample: fix skb_over_panic
net: usbnet: Fix -Wcast-function-type
net: hso: Fix -Wcast-function-type
net: port < inet_prot_sock(net) --> inet_port_requires_bind_service(net, port)
ibmvnic: Serialize device queries
ibmvnic: Bound waits for device queries
ibmvnic: Terminate waiting device threads after loss of service
ibmvnic: Fix completion structure initialization
net-sctp: replace some sock_net(sk) with just 'net'
net: Fix a documentation bug wrt. ip_unprivileged_port_start
tipc: fix link name length check
Pull RCU updates from Ingo Molnar:
"The main changes in this cycle were:
- Dynamic tick (nohz) updates, perhaps most notably changes to force
the tick on when needed due to lengthy in-kernel execution on CPUs
on which RCU is waiting.
- Linux-kernel memory consistency model updates.
- Replace rcu_swap_protected() with rcu_prepace_pointer().
- Torture-test updates.
- Documentation updates.
- Miscellaneous fixes"
* 'core-rcu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (51 commits)
security/safesetid: Replace rcu_swap_protected() with rcu_replace_pointer()
net/sched: Replace rcu_swap_protected() with rcu_replace_pointer()
net/netfilter: Replace rcu_swap_protected() with rcu_replace_pointer()
net/core: Replace rcu_swap_protected() with rcu_replace_pointer()
bpf/cgroup: Replace rcu_swap_protected() with rcu_replace_pointer()
fs/afs: Replace rcu_swap_protected() with rcu_replace_pointer()
drivers/scsi: Replace rcu_swap_protected() with rcu_replace_pointer()
drm/i915: Replace rcu_swap_protected() with rcu_replace_pointer()
x86/kvm/pmu: Replace rcu_swap_protected() with rcu_replace_pointer()
rcu: Upgrade rcu_swap_protected() to rcu_replace_pointer()
rcu: Suppress levelspread uninitialized messages
rcu: Fix uninitialized variable in nocb_gp_wait()
rcu: Update descriptions for rcu_future_grace_period tracepoint
rcu: Update descriptions for rcu_nocb_wake tracepoint
rcu: Remove obsolete descriptions for rcu_barrier tracepoint
rcu: Ensure that ->rcu_urgent_qs is set before resched IPI
workqueue: Convert for_each_wq to use built-in list check
rcu: Several rcu_segcblist functions can be static
rcu: Remove unused function hlist_bl_del_init_rcu()
Documentation: Rename rcu_node_context_switch() to rcu_note_context_switch()
...
Note that the sysctl write accessor functions guarantee that:
net->ipv4.sysctl_ip_prot_sock <= net->ipv4.ip_local_ports.range[0]
invariant is maintained, and as such the max() in selinux hooks is actually spurious.
ie. even though
if (snum < max(inet_prot_sock(sock_net(sk)), low) || snum > high) {
per logic is the same as
if ((snum < inet_prot_sock(sock_net(sk)) && snum < low) || snum > high) {
it is actually functionally equivalent to:
if (snum < low || snum > high) {
which is equivalent to:
if (snum < inet_prot_sock(sock_net(sk)) || snum < low || snum > high) {
even though the first clause is spurious.
But we want to hold on to it in case we ever want to change what what
inet_port_requires_bind_service() means (for example by changing
it from a, by default, [0..1024) range to some sort of set).
Test: builds, git 'grep inet_prot_sock' finds no other references
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: Maciej Żenczykowski <maze@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Match on h_vlan_encapsulated_proto and set up protocol dependency. Check
for protocol dependency before accessing the tci field. Allow to match
on the encapsulated ethertype too.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Match on ethertype and set up protocol dependency. Check for protocol
dependency before accessing the tci field. Allow to match on the
encapsulated ethertype too.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Hardware offload support at this stage assumes an ethernet device in
place. The flow dissector provides the intermediate representation to
express this selector, so extend it to allow to store the interface
type. Flower does not uses this, so skb_flow_dissect_meta() is not
extended to match on this new field.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Unbind flowtable callback if hook is unregistered.
This patch is implicitly fixing the error path of
nf_tables_newflowtable() and nft_flowtable_event().
Fixes: 8bb69f3b29 ("netfilter: nf_tables: add flowtable offload control plane")
Reported-by: wenxu <wenxu@ucloud.cn>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Undo the callback binding before unregistering the existing hooks. This
should also check for error of the bind setup call.
Fixes: c29f74e0df ("netfilter: nf_flow_table: hardware offload support")
Signed-off-by: wenxu <wenxu@ucloud.cn>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
The nft_flow_rule_offload_commit() function might fail after several
successful commands, thus, leaving the hardware filtering policy in
inconsistent state.
This patch adds nft_flow_rule_offload_abort() function which undoes the
updates that have been already processed if one command in this
transaction fails. Hence, the hardware ruleset is left as it was before
this aborted transaction.
The deletion path needs to create the flow_rule object too, in case that
an existing rule needs to be re-added from the abort path.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
The ct object is already in the flow_offload structure, remove it.
Signed-off-by: wenxu <wenxu@ucloud.cn>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
It should check the ndo_setup_tc in the nf_flow_table_offload_setup.
Fixes: c29f74e0df ("netfilter: nf_flow_table: hardware offload support")
Signed-off-by: wenxu <wenxu@ucloud.cn>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Add nf_flow_rule_route_ipv6() and use it from the IPv6 and the inet
flowtable type definitions. Rename the nf_flow_rule_route() function to
nf_flow_rule_route_ipv4().
Adjust maximum number of actions, which now becomes 16 to leave
sufficient room for the IPv6 address mangling for NAT.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
On 32-bit architectures, get_seconds() returns an unsigned 32-bit
time value, which also matches the type used in the nft_meta
code. This will not overflow in year 2038 as a time_t would, but
it still suffers from the overflow problem later on in year 2106.
Change this instance to use the time64_t type consistently
and avoid the deprecated get_seconds().
The nft_meta_weekday() calculation potentially gets a little slower
on 32-bit architectures, but now it has the same behavior as on
64-bit architectures and does not overflow.
Fixes: 63d10e12b0 ("netfilter: nft_meta: support for time matching")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Phil Sutter <phil@nwl.cc>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
The current xt_time driver suffers from the y2038 overflow on 32-bit
architectures, when the time of day calculations break.
Also, on both 32-bit and 64-bit architectures, there is a problem with
info->date_start/stop, which is part of the user ABI and overflows in
in 2106.
Fix the first issue by using time64_t and explicit calls to div_u64()
and div_u64_rem(), and document the seconds issue.
The explicit 64-bit division is unfortunately slower on 32-bit
architectures, but doing it as unsigned lets us use the optimized
division-through-multiplication path in most configurations. This should
be fine, as the code already does not allow any negative time of day
values.
Using u32 seconds values consistently would probably also work and
be a little more efficient, but that doesn't feel right as it would
propagate the y2106 overflow to more place rather than fewer.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Jozsef Kadlecsik says:
====================
ipset patches for nf-next
- Add wildcard support to hash:net,iface which makes possible to
match interface prefixes besides complete interfaces names, from
Kristian Evensen.
====================
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
If the encapsulated ethertype announces another inner VLAN header and
the offset falls within the boundaries of the inner VLAN header, then
adjust arithmetics to include the extra VLAN header length and fetch the
bytes from the vlan header in the skbuff data area that represents this
inner VLAN header.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Otherwise this leads to a stack corruption.
Fixes: c5d275276f ("netfilter: nf_tables_offload: add nft_flow_cls_offload_setup()")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Wrap the code to rebuild the ethernet + vlan header into a function.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Acked-by: Florian Westphal <fw@strlen.de>
If the offset is within the ethernet + vlan header size boundary, then
rebuild the ethernet + vlan header and use it to copy the bytes to the
register. Otherwise, subtract the vlan header size from the offset and
fall back to use skb_copy_bits().
There is one corner case though: If the offset plus the length of the
payload instruction goes over the ethernet + vlan header boundary, then,
fetch as many bytes as possible from the rebuilt ethernet + vlan header
and fall back to copy the remaining bytes through skb_copy_bits().
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Acked-by: Florian Westphal <fw@strlen.de>
This patch adds the dataplane hardware offload to the flowtable
infrastructure. Three new flags represent the hardware state of this
flow:
* FLOW_OFFLOAD_HW: This flow entry resides in the hardware.
* FLOW_OFFLOAD_HW_DYING: This flow entry has been scheduled to be remove
from hardware. This might be triggered by either packet path (via TCP
RST/FIN packet) or via aging.
* FLOW_OFFLOAD_HW_DEAD: This flow entry has been already removed from
the hardware, the software garbage collector can remove it from the
software flowtable.
This patch supports for:
* IPv4 only.
* Aging via FLOW_CLS_STATS, no packet and byte counter synchronization
at this stage.
This patch also adds the action callback that specifies how to convert
the flow entry into the flow_rule object that is passed to the driver.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds the NFTA_FLOWTABLE_FLAGS attribute that allows users to
specify the NF_FLOWTABLE_HW_OFFLOAD flag. This patch also adds a new
setup interface for the flowtable type to perform the flowtable offload
block callback configuration.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds the infrastructure to support for flow entry types.
The initial type is NF_FLOW_OFFLOAD_ROUTE that stores the routing
information into the flow entry to define a fastpath for the classic
forwarding path.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move rcu_head to struct flow_offload, then remove the flow_offload_entry
structure definition.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Simplify this code by storing the pointer to conntrack object in the
flow_offload structure.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
One conflict in the BPF samples Makefile, some fixes in 'net' whilst
we were converting over to Makefile.target rules in 'net-next'.
Signed-off-by: David S. Miller <davem@davemloft.net>
The same code which recognizes ICMP error packets is duplicated several
times. Use the icmp_is_err() and icmpv6_is_err() helpers instead, which
do the same thing.
ip_multipath_l3_keys() and tcf_nat_act() didn't check for all the error types,
assume that they should instead.
Signed-off-by: Matteo Croce <mcroce@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jozsef Kadlecsik says:
====================
ipset patches for nf
- Fix the error code in ip_set_sockfn_get() when copy_to_user() is used,
from Dan Carpenter.
- The IPv6 part was missed when fixing copying the right MAC address
in the patch "netfilter: ipset: Copy the right MAC address in bitmap:ip,mac
and hash:ip,mac sets", it is completed now by Stefano Brivio.
- ipset nla_policies are fixed to fully support NL_VALIDATE_STRICT and
the code is converted from deprecated parsings to verified ones.
====================
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Do not try to bind a chain again if it exists, otherwise the driver
returns EBUSY.
Fixes: c9626a2cbd ("netfilter: nf_tables: add hardware offload support")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Userspace never includes the NFT_BASE_CHAIN flag, this flag is inferred
from the NFTA_CHAIN_HOOK atribute. The chain update path does not allow
to update flags at this stage, the existing sanity check bogusly hits
EOPNOTSUPP in the basechain case if the offload flag is set on.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
If the object type doesn't implement an update operation and the user tries to
update it will silently ignore the update operation.
Fixes: aa4095a156 ("netfilter: nf_tables: fix possible null-pointer dereference in object update")
Signed-off-by: Fernando Fernandez Mancera <ffmancera@riseup.net>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Since v5.2 (commit "netlink: re-add parse/validate functions in strict
mode") NL_VALIDATE_STRICT is enabled. Fix the ipset nla_policies which did
not support strict mode and convert from deprecated parsings to verified ones.
Signed-off-by: Jozsef Kadlecsik <kadlec@netfilter.org>
Same as commit 1b4a75108d ("netfilter: ipset: Copy the right MAC
address in bitmap:ip,mac and hash:ip,mac sets"), another copy and paste
went wrong in commit 8cc4ccf583 ("netfilter: ipset: Allow matching on
destination MAC address for mac and ipmac sets").
When I fixed this for IPv4 in 1b4a75108d, I didn't realise that
hash:ip,mac sets also support IPv6 as family, and this is covered by a
separate function, hash_ipmac6_kadt().
In hash:ip,mac sets, the first dimension is the IP address, and the
second dimension is the MAC address: check the IPSET_DIM_TWO_SRC flag
in flags while deciding which MAC address to copy, destination or
source.
This way, mixing source and destination matches for the two dimensions
of ip,mac hash type works as expected, also for IPv6. With this setup:
ip netns add A
ip link add veth1 type veth peer name veth2 netns A
ip addr add 2001:db8::1/64 dev veth1
ip -net A addr add 2001:db8::2/64 dev veth2
ip link set veth1 up
ip -net A link set veth2 up
dst=$(ip netns exec A cat /sys/class/net/veth2/address)
ip netns exec A ipset create test_hash hash:ip,mac family inet6
ip netns exec A ipset add test_hash 2001:db8::1,${dst}
ip netns exec A ip6tables -A INPUT -p icmpv6 --icmpv6-type 135 -j ACCEPT
ip netns exec A ip6tables -A INPUT -m set ! --match-set test_hash src,dst -j DROP
ipset now correctly matches a test packet:
# ping -c1 2001:db8::2 >/dev/null
# echo $?
0
Reported-by: Chen, Yi <yiche@redhat.com>
Fixes: 8cc4ccf583 ("netfilter: ipset: Allow matching on destination MAC address for mac and ipmac sets")
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Signed-off-by: Jozsef Kadlecsik <kadlec@netfilter.org>
The copy_to_user() function returns the number of bytes remaining to be
copied. In this code, that positive return is checked at the end of the
function and we return zero/success. What we should do instead is
return -EFAULT.
Fixes: a7b4f989a6 ("netfilter: ipset: IP set core support")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Jozsef Kadlecsik <kadlec@netfilter.org>
The net,iface equal functions currently compares the full interface
names. In several cases, wildcard (or prefix) matching is useful. For
example, when converting a large iptables rule-set to make use of ipset,
I was able to significantly reduce the number of set elements by making
use of wildcard matching.
Wildcard matching is enabled by adding "wildcard" when adding an element
to a set. Internally, this causes the IPSET_FLAG_IFACE_WILDCARD-flag to
be set. When this flag is set, only the initial part of the interface
name is used for comparison.
Wildcard matching is done per element and not per set, as there are many
cases where mixing wildcard and non-wildcard elements are useful. This
means that is up to the user to handle (avoid) overlapping interface
names.
Signed-off-by: Kristian Evensen <kristian.evensen@gmail.com>
Signed-off-by: Jozsef Kadlecsik <kadlec@netfilter.org>
Make sure register data length does not mismatch immediate data length,
otherwise hit EOPNOTSUPP.
Fixes: c9626a2cbd ("netfilter: nf_tables: add hardware offload support")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
The only slightly tricky merge conflict was the netdevsim because the
mutex locking fix overlapped a lot of driver reload reorganization.
The rest were (relatively) trivial in nature.
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull RCU and LKMM changes from Paul E. McKenney:
- Documentation updates.
- Miscellaneous fixes.
- Dynamic tick (nohz) updates, perhaps most notably changes to
force the tick on when needed due to lengthy in-kernel execution
on CPUs on which RCU is waiting.
- Replace rcu_swap_protected() with rcu_prepace_pointer().
- Torture-test updates.
- Linux-kernel memory consistency model updates.
Signed-off-by: Ingo Molnar <mingo@kernel.org>
This commit replaces the use of rcu_swap_protected() with the more
intuitively appealing rcu_replace_pointer() as a step towards removing
rcu_swap_protected().
Link: https://lore.kernel.org/lkml/CAHk-=wiAsJLw1egFEE=Z7-GGtM6wcvtyytXZA1+BHqta4gg6Hw@mail.gmail.com/
Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
[ paulmck: From rcu_replace() to rcu_replace_pointer() per Ingo Molnar. ]
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Acked-by: Pablo Neira Ayuso <pablo@netfilter.org>
Cc: Jozsef Kadlecsik <kadlec@netfilter.org>
Cc: Florian Westphal <fw@strlen.de>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: <netfilter-devel@vger.kernel.org>
Cc: <coreteam@netfilter.org>
Cc: <netdev@vger.kernel.org>
Fix misspellings of "disconnect", "disconnecting", "connections", and
"disconnected".
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Acked-by: Kalle Valo <kvalo@codeaurora.org>
Acked-by: Simon Horman <horms@verge.net.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pablo Neira Ayuso says:
====================
Netfilter/IPVS updates for net-next
The following patchset contains Netfilter/IPVS updates for net-next,
more specifically:
* Updates for ipset:
1) Coding style fix for ipset comment extension, from Jeremy Sowden.
2) De-inline many functions in ipset, from Jeremy Sowden.
3) Move ipset function definition from header to source file.
4) Move ip_set_put_flags() to source, export it as a symbol, remove
inline.
5) Move range_to_mask() to the source file where this is used.
6) Move ip_set_get_ip_port() to the source file where this is used.
* IPVS selftests and netns improvements:
7) Two patches to speedup ipvs netns dismantle, from Haishuang Yan.
8) Three patches to add selftest script for ipvs, also from
Haishuang Yan.
* Conntrack updates and new nf_hook_slow_list() function:
9) Document ct ecache extension, from Florian Westphal.
10) Skip ct extensions from ctnetlink dump, from Florian.
11) Free ct extension immediately, from Florian.
12) Skip access to ecache extension from nf_ct_deliver_cached_events()
this is not correct as reported by Syzbot.
13) Add and use nf_hook_slow_list(), from Florian.
* Flowtable infrastructure updates:
14) Move priority to nf_flowtable definition.
15) Dynamic allocation of per-device hooks in flowtables.
16) Allow to include netdevice only once in flowtable definitions.
17) Rise maximum number of devices per flowtable.
* Netfilter hardware offload infrastructure updates:
18) Add nft_flow_block_chain() helper function.
19) Pass callback list to nft_setup_cb_call().
20) Add nft_flow_cls_offload_setup() helper function.
21) Remove rules for the unregistered device via netdevice event.
22) Support for multiple devices in a basechain definition at the
ingress hook.
22) Add nft_chain_offload_cmd() helper function.
23) Add nft_flow_block_offload_init() helper function.
24) Rewind in case of failing to bind multiple devices to hook.
25) Typo in IPv6 tproxy module description, from Norman Rasmussen.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Simon Horman says:
====================
IPVS fixes for v5.4
* Eric Dumazet resolves a race condition in switching the defense level
* Davide Caratti resolves a race condition in module removal
====================
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
nft_flow_block_chain() needs to unbind in case of error when performing
the multi-device binding.
Fixes: d54725cd11 ("netfilter: nf_tables: support for multiple devices per netdev hook")
Reported-by: wenxu <wenxu@ucloud.cn>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
This patch adds the nft_flow_block_offload_init() helper function to
initialize the flow_block_offload object.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
syzbot reported following splat:
BUG: KASAN: use-after-free in __nf_ct_ext_exist
include/net/netfilter/nf_conntrack_extend.h:53 [inline]
BUG: KASAN: use-after-free in nf_ct_deliver_cached_events+0x5c3/0x6d0
net/netfilter/nf_conntrack_ecache.c:205
nf_conntrack_confirm include/net/netfilter/nf_conntrack_core.h:65 [inline]
nf_confirm+0x3d8/0x4d0 net/netfilter/nf_conntrack_proto.c:154
[..]
While there is no reproducer yet, the syzbot report contains one
interesting bit of information:
Freed by task 27585:
[..]
kfree+0x10a/0x2c0 mm/slab.c:3757
nf_ct_ext_destroy+0x2ab/0x2e0 net/netfilter/nf_conntrack_extend.c:38
nf_conntrack_free+0x8f/0xe0 net/netfilter/nf_conntrack_core.c:1418
destroy_conntrack+0x1a2/0x270 net/netfilter/nf_conntrack_core.c:626
nf_conntrack_put include/linux/netfilter/nf_conntrack_common.h:31 [inline]
nf_ct_resolve_clash net/netfilter/nf_conntrack_core.c:915 [inline]
^^^^^^^^^^^^^^^^^^^
__nf_conntrack_confirm+0x21ca/0x2830 net/netfilter/nf_conntrack_core.c:1038
nf_conntrack_confirm include/net/netfilter/nf_conntrack_core.h:63 [inline]
nf_confirm+0x3e7/0x4d0 net/netfilter/nf_conntrack_proto.c:154
This is whats happening:
1. a conntrack entry is about to be confirmed (added to hash table).
2. a clash with existing entry is detected.
3. nf_ct_resolve_clash() puts skb->nfct (the "losing" entry).
4. this entry now has a refcount of 0 and is freed to SLAB_TYPESAFE_BY_RCU
kmem cache.
skb->nfct has been replaced by the one found in the hash.
Problem is that nf_conntrack_confirm() uses the old ct:
static inline int nf_conntrack_confirm(struct sk_buff *skb)
{
struct nf_conn *ct = (struct nf_conn *)skb_nfct(skb);
int ret = NF_ACCEPT;
if (ct) {
if (!nf_ct_is_confirmed(ct))
ret = __nf_conntrack_confirm(skb);
if (likely(ret == NF_ACCEPT))
nf_ct_deliver_cached_events(ct); /* This ct has refcount 0! */
}
return ret;
}
As of "netfilter: conntrack: free extension area immediately", we can't
access conntrack extensions in this case.
To fix this, make sure we check the dying bit presence before attempting
to get the eache extension.
Reported-by: syzbot+c7aabc9fe93e7f3637ba@syzkaller.appspotmail.com
Fixes: 2ad9d7747c ("netfilter: conntrack: free extension area immediately")
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
John Hurley