[ Upstream commit a12c76a033 ]
This patch addresses issues with filter counting in block (tcf_block),
particularly for software bypass scenarios, by introducing a more
accurate mechanism using useswcnt.
Previously, filtercnt and skipswcnt were introduced by:
Commit 2081fd3445 ("net: sched: cls_api: add filter counter") and
Commit f631ef39d8 ("net: sched: cls_api: add skip_sw counter")
filtercnt tracked all tp (tcf_proto) objects added to a block, and
skipswcnt counted tp objects with the skipsw attribute set.
The problem is: a single tp can contain multiple filters, some with skipsw
and others without. The current implementation fails in the case:
When the first filter in a tp has skipsw, both skipswcnt and filtercnt
are incremented, then adding a second filter without skipsw to the same
tp does not modify these counters because tp->counted is already set.
This results in bypass software behavior based solely on skipswcnt
equaling filtercnt, even when the block includes filters without
skipsw. Consequently, filters without skipsw are inadvertently bypassed.
To address this, the patch introduces useswcnt in block to explicitly count
tp objects containing at least one filter without skipsw. Key changes
include:
Whenever a filter without skipsw is added, its tp is marked with usesw
and counted in useswcnt. tc_run() now uses useswcnt to determine software
bypass, eliminating reliance on filtercnt and skipswcnt.
This refined approach prevents software bypass for blocks containing
mixed filters, ensuring correct behavior in tc_run().
Additionally, as atomic operations on useswcnt ensure thread safety and
tp->lock guards access to tp->usesw and tp->counted, the broader lock
down_write(&block->cb_lock) is no longer required in tc_new_tfilter(),
and this resolves a performance regression caused by the filter counting
mechanism during parallel filter insertions.
The improvement can be demonstrated using the following script:
# cat insert_tc_rules.sh
tc qdisc add dev ens1f0np0 ingress
for i in $(seq 16); do
taskset -c $i tc -b rules_$i.txt &
done
wait
Each of rules_$i.txt files above includes 100000 tc filter rules to a
mlx5 driver NIC ens1f0np0.
Without this patch:
# time sh insert_tc_rules.sh
real 0m50.780s
user 0m23.556s
sys 4m13.032s
With this patch:
# time sh insert_tc_rules.sh
real 0m17.718s
user 0m7.807s
sys 3m45.050s
Fixes: 047f340b36 ("net: sched: make skip_sw actually skip software")
Reported-by: Shuang Li <shuali@redhat.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Reviewed-by: Asbjørn Sloth Tønnesen <ast@fiberby.net>
Tested-by: Asbjørn Sloth Tønnesen <ast@fiberby.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit bc50835e83 ]
Lion Ackermann was able to create a UAF which can be abused for privilege
escalation with the following script
Step 1. create root qdisc
tc qdisc add dev lo root handle 1:0 drr
step2. a class for packet aggregation do demonstrate uaf
tc class add dev lo classid 1:1 drr
step3. a class for nesting
tc class add dev lo classid 1:2 drr
step4. a class to graft qdisc to
tc class add dev lo classid 1:3 drr
step5.
tc qdisc add dev lo parent 1:1 handle 2:0 plug limit 1024
step6.
tc qdisc add dev lo parent 1:2 handle 3:0 drr
step7.
tc class add dev lo classid 3:1 drr
step 8.
tc qdisc add dev lo parent 3:1 handle 4:0 pfifo
step 9. Display the class/qdisc layout
tc class ls dev lo
class drr 1:1 root leaf 2: quantum 64Kb
class drr 1:2 root leaf 3: quantum 64Kb
class drr 3:1 root leaf 4: quantum 64Kb
tc qdisc ls
qdisc drr 1: dev lo root refcnt 2
qdisc plug 2: dev lo parent 1:1
qdisc pfifo 4: dev lo parent 3:1 limit 1000p
qdisc drr 3: dev lo parent 1:2
step10. trigger the bug <=== prevented by this patch
tc qdisc replace dev lo parent 1:3 handle 4:0
step 11. Redisplay again the qdiscs/classes
tc class ls dev lo
class drr 1:1 root leaf 2: quantum 64Kb
class drr 1:2 root leaf 3: quantum 64Kb
class drr 1:3 root leaf 4: quantum 64Kb
class drr 3:1 root leaf 4: quantum 64Kb
tc qdisc ls
qdisc drr 1: dev lo root refcnt 2
qdisc plug 2: dev lo parent 1:1
qdisc pfifo 4: dev lo parent 3:1 refcnt 2 limit 1000p
qdisc drr 3: dev lo parent 1:2
Observe that a) parent for 4:0 does not change despite the replace request.
There can only be one parent. b) refcount has gone up by two for 4:0 and
c) both class 1:3 and 3:1 are pointing to it.
Step 12. send one packet to plug
echo "" | socat -u STDIN UDP4-DATAGRAM:127.0.0.1:8888,priority=$((0x10001))
step13. send one packet to the grafted fifo
echo "" | socat -u STDIN UDP4-DATAGRAM:127.0.0.1:8888,priority=$((0x10003))
step14. lets trigger the uaf
tc class delete dev lo classid 1:3
tc class delete dev lo classid 1:1
The semantics of "replace" is for a del/add _on the same node_ and not
a delete from one node(3:1) and add to another node (1:3) as in step10.
While we could "fix" with a more complex approach there could be
consequences to expectations so the patch takes the preventive approach of
"disallow such config".
Joint work with Lion Ackermann <nnamrec@gmail.com>
Fixes: 1da177e4c3 ("Linux-2.6.12-rc2")
Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20250116013713.900000-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 10685681ba ]
The current implementation does not work correctly with a limit of
1. iproute2 actually checks for this and this patch adds the check in
kernel as well.
This fixes the following syzkaller reported crash:
UBSAN: array-index-out-of-bounds in net/sched/sch_sfq.c:210:6
index 65535 is out of range for type 'struct sfq_head[128]'
CPU: 0 PID: 2569 Comm: syz-executor101 Not tainted 5.10.0-smp-DEV #1
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 09/13/2024
Call Trace:
__dump_stack lib/dump_stack.c:79 [inline]
dump_stack+0x125/0x19f lib/dump_stack.c:120
ubsan_epilogue lib/ubsan.c:148 [inline]
__ubsan_handle_out_of_bounds+0xed/0x120 lib/ubsan.c:347
sfq_link net/sched/sch_sfq.c:210 [inline]
sfq_dec+0x528/0x600 net/sched/sch_sfq.c:238
sfq_dequeue+0x39b/0x9d0 net/sched/sch_sfq.c:500
sfq_reset+0x13/0x50 net/sched/sch_sfq.c:525
qdisc_reset+0xfe/0x510 net/sched/sch_generic.c:1026
tbf_reset+0x3d/0x100 net/sched/sch_tbf.c:319
qdisc_reset+0xfe/0x510 net/sched/sch_generic.c:1026
dev_reset_queue+0x8c/0x140 net/sched/sch_generic.c:1296
netdev_for_each_tx_queue include/linux/netdevice.h:2350 [inline]
dev_deactivate_many+0x6dc/0xc20 net/sched/sch_generic.c:1362
__dev_close_many+0x214/0x350 net/core/dev.c:1468
dev_close_many+0x207/0x510 net/core/dev.c:1506
unregister_netdevice_many+0x40f/0x16b0 net/core/dev.c:10738
unregister_netdevice_queue+0x2be/0x310 net/core/dev.c:10695
unregister_netdevice include/linux/netdevice.h:2893 [inline]
__tun_detach+0x6b6/0x1600 drivers/net/tun.c:689
tun_detach drivers/net/tun.c:705 [inline]
tun_chr_close+0x104/0x1b0 drivers/net/tun.c:3640
__fput+0x203/0x840 fs/file_table.c:280
task_work_run+0x129/0x1b0 kernel/task_work.c:185
exit_task_work include/linux/task_work.h:33 [inline]
do_exit+0x5ce/0x2200 kernel/exit.c:931
do_group_exit+0x144/0x310 kernel/exit.c:1046
__do_sys_exit_group kernel/exit.c:1057 [inline]
__se_sys_exit_group kernel/exit.c:1055 [inline]
__x64_sys_exit_group+0x3b/0x40 kernel/exit.c:1055
do_syscall_64+0x6c/0xd0
entry_SYSCALL_64_after_hwframe+0x61/0xcb
RIP: 0033:0x7fe5e7b52479
Code: Unable to access opcode bytes at RIP 0x7fe5e7b5244f.
RSP: 002b:00007ffd3c800398 EFLAGS: 00000246 ORIG_RAX: 00000000000000e7
RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007fe5e7b52479
RDX: 000000000000003c RSI: 00000000000000e7 RDI: 0000000000000000
RBP: 00007fe5e7bcd2d0 R08: ffffffffffffffb8 R09: 0000000000000014
R10: 0000000000000000 R11: 0000000000000246 R12: 00007fe5e7bcd2d0
R13: 0000000000000000 R14: 00007fe5e7bcdd20 R15: 00007fe5e7b24270
The crash can be also be reproduced with the following (with a tc
recompiled to allow for sfq limits of 1):
tc qdisc add dev dummy0 handle 1: root tbf rate 1Kbit burst 100b lat 1s
../iproute2-6.9.0/tc/tc qdisc add dev dummy0 handle 2: parent 1:10 sfq limit 1
ifconfig dummy0 up
ping -I dummy0 -f -c2 -W0.1 8.8.8.8
sleep 1
Scenario that triggers the crash:
* the first packet is sent and queued in TBF and SFQ; qdisc qlen is 1
* TBF dequeues: it peeks from SFQ which moves the packet to the
gso_skb list and keeps qdisc qlen set to 1. TBF is out of tokens so
it schedules itself for later.
* the second packet is sent and TBF tries to queues it to SFQ. qdisc
qlen is now 2 and because the SFQ limit is 1 the packet is dropped
by SFQ. At this point qlen is 1, and all of the SFQ slots are empty,
however q->tail is not NULL.
At this point, assuming no more packets are queued, when sch_dequeue
runs again it will decrement the qlen for the current empty slot
causing an underflow and the subsequent out of bounds access.
Reported-by: syzbot <syzkaller@googlegroups.com>
Fixes: 1da177e4c3 ("Linux-2.6.12-rc2")
Signed-off-by: Octavian Purdila <tavip@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20241204030520.2084663-2-tavip@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit e4650d7ae4 ]
SFQ has an assumption on dealing with packets smaller than 64KB.
Even before BIG TCP, TCA_STAB can provide arbitrary big values
in qdisc_pkt_len(skb)
It is time to switch (struct sfq_slot)->allot to a 32bit field.
sizeof(struct sfq_slot) is now 64 bytes, giving better cache locality.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Toke Høiland-Jørgensen <toke@redhat.com>
Link: https://patch.msgid.link/20241008111603.653140-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Stable-dep-of: 10685681ba ("net_sched: sch_sfq: don't allow 1 packet limit")
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 737d4d91d3 ]
Even though we fixed a logic error in the commit cited below, syzbot
still managed to trigger an underflow of the per-host bulk flow
counters, leading to an out of bounds memory access.
To avoid any such logic errors causing out of bounds memory accesses,
this commit factors out all accesses to the per-host bulk flow counters
to a series of helpers that perform bounds-checking before any
increments and decrements. This also has the benefit of improving
readability by moving the conditional checks for the flow mode into
these helpers, instead of having them spread out throughout the
code (which was the cause of the original logic error).
As part of this change, the flow quantum calculation is consolidated
into a helper function, which means that the dithering applied to the
ost load scaling is now applied both in the DRR rotation and when a
sparse flow's quantum is first initiated. The only user-visible effect
of this is that the maximum packet size that can be sent while a flow
stays sparse will now vary with +/- one byte in some cases. This should
not make a noticeable difference in practice, and thus it's not worth
complicating the code to preserve the old behaviour.
Fixes: 546ea84d07 ("sched: sch_cake: fix bulk flow accounting logic for host fairness")
Reported-by: syzbot+f63600d288bfb7057424@syzkaller.appspotmail.com
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Acked-by: Dave Taht <dave.taht@gmail.com>
Link: https://patch.msgid.link/20250107120105.70685-1-toke@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 5eb7de8cd5 upstream.
Changes to sch->q.qlen around qdisc_tree_reduce_backlog() need to happen
_before_ a call to said function because otherwise it may fail to notify
parent qdiscs when the child is about to become empty.
Signed-off-by: Lion Ackermann <nnamrec@gmail.com>
Acked-by: Toke Høiland-Jørgensen <toke@toke.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>
Cc: Artem Metla <ametla@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit f8d4bc4550 ]
In general, 'qlen' of any classful qdisc should keep track of the
number of packets that the qdisc itself and all of its children holds.
In case of netem, 'qlen' only accounts for the packets in its internal
tfifo. When netem is used with a child qdisc, the child qdisc can use
'qdisc_tree_reduce_backlog' to inform its parent, netem, about created
or dropped SKBs. This function updates 'qlen' and the backlog statistics
of netem, but netem does not account for changes made by a child qdisc.
'qlen' then indicates the wrong number of packets in the tfifo.
If a child qdisc creates new SKBs during enqueue and informs its parent
about this, netem's 'qlen' value is increased. When netem dequeues the
newly created SKBs from the child, the 'qlen' in netem is not updated.
If 'qlen' reaches the configured sch->limit, the enqueue function stops
working, even though the tfifo is not full.
Reproduce the bug:
Ensure that the sender machine has GSO enabled. Configure netem as root
qdisc and tbf as its child on the outgoing interface of the machine
as follows:
$ tc qdisc add dev <oif> root handle 1: netem delay 100ms limit 100
$ tc qdisc add dev <oif> parent 1:0 tbf rate 50Mbit burst 1542 latency 50ms
Send bulk TCP traffic out via this interface, e.g., by running an iPerf3
client on the machine. Check the qdisc statistics:
$ tc -s qdisc show dev <oif>
Statistics after 10s of iPerf3 TCP test before the fix (note that
netem's backlog > limit, netem stopped accepting packets):
qdisc netem 1: root refcnt 2 limit 1000 delay 100ms
Sent 2767766 bytes 1848 pkt (dropped 652, overlimits 0 requeues 0)
backlog 4294528236b 1155p requeues 0
qdisc tbf 10: parent 1:1 rate 50Mbit burst 1537b lat 50ms
Sent 2767766 bytes 1848 pkt (dropped 327, overlimits 7601 requeues 0)
backlog 0b 0p requeues 0
Statistics after the fix:
qdisc netem 1: root refcnt 2 limit 1000 delay 100ms
Sent 37766372 bytes 24974 pkt (dropped 9, overlimits 0 requeues 0)
backlog 0b 0p requeues 0
qdisc tbf 10: parent 1:1 rate 50Mbit burst 1537b lat 50ms
Sent 37766372 bytes 24974 pkt (dropped 327, overlimits 96017 requeues 0)
backlog 0b 0p requeues 0
tbf segments the GSO SKBs (tbf_segment) and updates the netem's 'qlen'.
The interface fully stops transferring packets and "locks". In this case,
the child qdisc and tfifo are empty, but 'qlen' indicates the tfifo is at
its limit and no more packets are accepted.
This patch adds a counter for the entries in the tfifo. Netem's 'qlen' is
only decreased when a packet is returned by its dequeue function, and not
during enqueuing into the child qdisc. External updates to 'qlen' are thus
accounted for and only the behavior of the backlog statistics changes. As
in other qdiscs, 'qlen' then keeps track of how many packets are held in
netem and all of its children. As before, sch->limit remains as the
maximum number of packets in the tfifo. The same applies to netem's
backlog statistics.
Fixes: 50612537e9 ("netem: fix classful handling")
Signed-off-by: Martin Ottens <martin.ottens@fau.de>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Link: https://patch.msgid.link/20241210131412.1837202-1-martin.ottens@fau.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 397006ba5d ]
The subsequent calculation of port_rate = speed * 1000 * BYTES_PER_KBIT,
where the BYTES_PER_KBIT is of type LL, may cause an overflow.
At least when speed = SPEED_20000, the expression to the left of port_rate
will be greater than INT_MAX.
Found by Linux Verification Center (linuxtesting.org) with SVACE.
Signed-off-by: Elena Salomatkina <esalomatkina@ispras.ru>
Link: https://patch.msgid.link/20241013124529.1043-1-esalomatkina@ispras.ru
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 2922078094 ]
When matching erspan_opt in cls_flower, only the (version, dir, hwid)
fields are relevant. However, in fl_set_erspan_opt() it initializes
all bits of erspan_opt and its mask to 1. This inadvertently requires
packets to match not only the (version, dir, hwid) fields but also the
other fields that are unexpectedly set to 1.
This patch resolves the issue by ensuring that only the (version, dir,
hwid) fields are configured in fl_set_erspan_opt(), leaving the other
fields to 0 in erspan_opt.
Fixes: 79b1011cb3 ("net: sched: allow flower to match erspan options")
Reported-by: Shuang Li <shuali@redhat.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Reviewed-by: Cong Wang <cong.wang@bytedance.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 1596a135e3 ]
When the length of a GSO packet in the tbf qdisc is larger than the burst
size configured the packet will be segmented by the tbf_segment function.
Whenever this function is used to enqueue SKBs, the backlog statistic of
the tbf is not increased correctly. This can lead to underflows of the
'backlog' byte-statistic value when these packets are dequeued from tbf.
Reproduce the bug:
Ensure that the sender machine has GSO enabled. Configured the tbf on
the outgoing interface of the machine as follows (burstsize = 1 MTU):
$ tc qdisc add dev <oif> root handle 1: tbf rate 50Mbit burst 1514 latency 50ms
Send bulk TCP traffic out via this interface, e.g., by running an iPerf3
client on this machine. Check the qdisc statistics:
$ tc -s qdisc show dev <oif>
The 'backlog' byte-statistic has incorrect values while traffic is
transferred, e.g., high values due to u32 underflows. When the transfer
is stopped, the value is != 0, which should never happen.
This patch fixes this bug by updating the statistics correctly, even if
single SKBs of a GSO SKB cannot be enqueued.
Fixes: e43ac79a4b ("sch_tbf: segment too big GSO packets")
Signed-off-by: Martin Ottens <martin.ottens@fau.de>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20241125174608.1484356-1-martin.ottens@fau.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 122aba8c80 upstream.
Recent kernels cause a lot of TCP retransmissions
[ ID] Interval Transfer Bitrate Retr Cwnd
[ 5] 0.00-1.00 sec 2.24 GBytes 19.2 Gbits/sec 2767 442 KBytes
[ 5] 1.00-2.00 sec 2.23 GBytes 19.1 Gbits/sec 2312 350 KBytes
^^^^
Replacing the qdisc with pfifo makes retransmissions go away.
It appears that a flow may have a delayed packet with a very near
Tx time. Later, we may get busy processing Rx and the target Tx time
will pass, but we won't service Tx since the CPU is busy with Rx.
If Rx sees an ACK and we try to push more data for the delayed flow
we may fastpath the skb, not realizing that there are already "ready
to send" packets for this flow sitting in the qdisc.
Don't trust the fastpath if we are "behind" according to the projected
Tx time for next flow waiting in the Qdisc. Because we consider anything
within the offload window to be okay for fastpath we must consider
the entire offload window as "now".
Qdisc config:
qdisc fq 8001: dev eth0 parent 1234:1 limit 10000p flow_limit 100p \
buckets 32768 orphan_mask 1023 bands 3 \
priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1 \
weights 589824 196608 65536 quantum 3028b initial_quantum 15140b \
low_rate_threshold 550Kbit \
refill_delay 40ms timer_slack 10us horizon 10s horizon_drop
For iperf this change seems to do fine, the reordering is gone.
The fastpath still gets used most of the time:
gc 0 highprio 0 fastpath 142614 throttled 418309 latency 19.1us
xx_behind 2731
where "xx_behind" counts how many times we hit the new "return false".
CC: stable@vger.kernel.org
Fixes: 076433bd78 ("net_sched: sch_fq: add fast path for mostly idle qdisc")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20241124022148.3126719-1-kuba@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
[stable: drop the offload horizon, it's not supported / 0]
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
To generate hnode handles (in gen_new_htid()), u32 uses IDR and
encodes the returned small integer into a structured 32-bit
word. Unfortunately, at disposal time, the needed decoding
is not done. As a result, idr_remove() fails, and the IDR
fills up. Since its size is 2048, the following script ends up
with "Filter already exists":
tc filter add dev myve $FILTER1
tc filter add dev myve $FILTER2
for i in {1..2048}
do
echo $i
tc filter del dev myve $FILTER2
tc filter add dev myve $FILTER2
done
This patch adds the missing decoding logic for handles that
deserve it.
Fixes: e7614370d6 ("net_sched: use idr to allocate u32 filter handles")
Reviewed-by: Eric Dumazet <edumazet@google.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Alexandre Ferrieux <alexandre.ferrieux@orange.com>
Tested-by: Victor Nogueira <victor@mojatatu.com>
Link: https://patch.msgid.link/20241110172836.331319-1-alexandre.ferrieux@orange.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This command:
$ tc qdisc replace dev eth0 ingress_block 1 egress_block 1 clsact
Error: block dev insert failed: -EBUSY.
fails because user space requests the same block index to be set for
both ingress and egress.
[ side note, I don't think it even failed prior to commit 913b47d342
("net/sched: Introduce tc block netdev tracking infra"), because this
is a command from an old set of notes of mine which used to work, but
alas, I did not scientifically bisect this ]
The problem is not that it fails, but rather, that the second time
around, it fails differently (and irrecoverably):
$ tc qdisc replace dev eth0 ingress_block 1 egress_block 1 clsact
Error: dsa_core: Flow block cb is busy.
[ another note: the extack is added by me for illustration purposes.
the context of the problem is that clsact_init() obtains the same
&q->ingress_block pointer as &q->egress_block, and since we call
tcf_block_get_ext() on both of them, "dev" will be added to the
block->ports xarray twice, thus failing the operation: once through
the ingress block pointer, and once again through the egress block
pointer. the problem itself is that when xa_insert() fails, we have
emitted a FLOW_BLOCK_BIND command through ndo_setup_tc(), but the
offload never sees a corresponding FLOW_BLOCK_UNBIND. ]
Even correcting the bad user input, we still cannot recover:
$ tc qdisc replace dev swp3 ingress_block 1 egress_block 2 clsact
Error: dsa_core: Flow block cb is busy.
Basically the only way to recover is to reboot the system, or unbind and
rebind the net device driver.
To fix the bug, we need to fill the correct error teardown path which
was missed during code movement, and call tcf_block_offload_unbind()
when xa_insert() fails.
[ last note, fundamentally I blame the label naming convention in
tcf_block_get_ext() for the bug. The labels should be named after what
they do, not after the error path that jumps to them. This way, it is
obviously wrong that two labels pointing to the same code mean
something is wrong, and checking the code correctness at the goto site
is also easier ]
Fixes: 94e2557d08 ("net: sched: move block device tracking into tcf_block_get/put_ext()")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Link: https://patch.msgid.link/20241023100541.974362-1-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In qdisc_tree_reduce_backlog, Qdiscs with major handle ffff: are assumed
to be either root or ingress. This assumption is bogus since it's valid
to create egress qdiscs with major handle ffff:
Budimir Markovic found that for qdiscs like DRR that maintain an active
class list, it will cause a UAF with a dangling class pointer.
In 066a3b5b23, the concern was to avoid iterating over the ingress
qdisc since its parent is itself. The proper fix is to stop when parent
TC_H_ROOT is reached because the only way to retrieve ingress is when a
hierarchy which does not contain a ffff: major handle call into
qdisc_lookup with TC_H_MAJ(TC_H_ROOT).
In the scenario where major ffff: is an egress qdisc in any of the tree
levels, the updates will also propagate to TC_H_ROOT, which then the
iteration must stop.
Fixes: 066a3b5b23 ("[NET_SCHED] sch_api: fix qdisc_tree_decrease_qlen() loop")
Reported-by: Budimir Markovic <markovicbudimir@gmail.com>
Suggested-by: Jamal Hadi Salim <jhs@mojatatu.com>
Tested-by: Victor Nogueira <victor@mojatatu.com>
Signed-off-by: Pedro Tammela <pctammela@mojatatu.com>
Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
net/sched/sch_api.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20241024165547.418570-1-jhs@mojatatu.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In 'taprio_change()', 'admin' pointer may become dangling due to sched
switch / removal caused by 'advance_sched()', and critical section
protected by 'q->current_entry_lock' is too small to prevent from such
a scenario (which causes use-after-free detected by KASAN). Fix this
by prefer 'rcu_replace_pointer()' over 'rcu_assign_pointer()' to update
'admin' immediately before an attempt to schedule freeing.
Fixes: a3d43c0d56 ("taprio: Add support adding an admin schedule")
Reported-by: syzbot+b65e0af58423fc8a73aa@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=b65e0af58423fc8a73aa
Acked-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Signed-off-by: Dmitry Antipov <dmantipov@yandex.ru>
Link: https://patch.msgid.link/20241018051339.418890-1-dmantipov@yandex.ru
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
tcf_action_init() has logic for checking mismatches between action and
filter offload flags (skip_sw/skip_hw). AFAIU, this is intended to run
on the transition between the new tc_act_bind(flags) returning true (aka
now gets bound to classifier) and tc_act_bind(act->tcfa_flags) returning
false (aka action was not bound to classifier before). Otherwise, the
check is skipped.
For the case where an action is not standalone, but rather it was
created by a classifier and is bound to it, tcf_action_init() skips the
check entirely, and this means it allows mismatched flags to occur.
Taking the matchall classifier code path as an example (with mirred as
an action), the reason is the following:
1 | mall_change()
2 | -> mall_replace_hw_filter()
3 | -> tcf_exts_validate_ex()
4 | -> flags |= TCA_ACT_FLAGS_BIND;
5 | -> tcf_action_init()
6 | -> tcf_action_init_1()
7 | -> a_o->init()
8 | -> tcf_mirred_init()
9 | -> tcf_idr_create_from_flags()
10 | -> tcf_idr_create()
11 | -> p->tcfa_flags = flags;
12 | -> tc_act_bind(flags))
13 | -> tc_act_bind(act->tcfa_flags)
When invoked from tcf_exts_validate_ex() like matchall does (but other
classifiers validate their extensions as well), tcf_action_init() runs
in a call path where "flags" always contains TCA_ACT_FLAGS_BIND (set by
line 4). So line 12 is always true, and line 13 is always true as well.
No transition ever takes place, and the check is skipped.
The code was added in this form in commit c86e0209dc ("flow_offload:
validate flags of filter and actions"), but I'm attributing the blame
even earlier in that series, to when TCA_ACT_FLAGS_SKIP_HW and
TCA_ACT_FLAGS_SKIP_SW were added to the UAPI.
Following the development process of this change, the check did not
always exist in this form. A change took place between v3 [1] and v4 [2],
AFAIU due to review feedback that it doesn't make sense for action flags
to be different than classifier flags. I think I agree with that
feedback, but it was translated into code that omits enforcing this for
"classic" actions created at the same time with the filters themselves.
There are 3 more important cases to discuss. First there is this command:
$ tc qdisc add dev eth0 clasct
$ tc filter add dev eth0 ingress matchall skip_sw \
action mirred ingress mirror dev eth1
which should be allowed, because prior to the concept of dedicated
action flags, it used to work and it used to mean the action inherited
the skip_sw/skip_hw flags from the classifier. It's not a mismatch.
Then we have this command:
$ tc qdisc add dev eth0 clasct
$ tc filter add dev eth0 ingress matchall skip_sw \
action mirred ingress mirror dev eth1 skip_hw
where there is a mismatch and it should be rejected.
Finally, we have:
$ tc qdisc add dev eth0 clasct
$ tc filter add dev eth0 ingress matchall skip_sw \
action mirred ingress mirror dev eth1 skip_sw
where the offload flags coincide, and this should be treated the same as
the first command based on inheritance, and accepted.
[1]: https://lore.kernel.org/netdev/20211028110646.13791-9-simon.horman@corigine.com/
[2]: https://lore.kernel.org/netdev/20211118130805.23897-10-simon.horman@corigine.com/
Fixes: 7adc576512 ("flow_offload: add skip_hw and skip_sw to control if offload the action")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Tested-by: Ido Schimmel <idosch@nvidia.com>
Link: https://patch.msgid.link/20241017161049.3570037-1-vladimir.oltean@nxp.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Some workloads hit the infamous dev_watchdog() message:
"NETDEV WATCHDOG: eth0 (xxxx): transmit queue XX timed out"
It seems possible to hit this even for perfectly normal
BQL enabled drivers:
1) Assume a TX queue was idle for more than dev->watchdog_timeo
(5 seconds unless changed by the driver)
2) Assume a big packet is sent, exceeding current BQL limit.
3) Driver ndo_start_xmit() puts the packet in TX ring,
and netdev_tx_sent_queue() is called.
4) QUEUE_STATE_STACK_XOFF could be set from netdev_tx_sent_queue()
before txq->trans_start has been written.
5) txq->trans_start is written later, from netdev_start_xmit()
if (rc == NETDEV_TX_OK)
txq_trans_update(txq)
dev_watchdog() running on another cpu could read the old
txq->trans_start, and then see QUEUE_STATE_STACK_XOFF, because 5)
did not happen yet.
To solve the issue, write txq->trans_start right before one XOFF bit
is set :
- _QUEUE_STATE_DRV_XOFF from netif_tx_stop_queue()
- __QUEUE_STATE_STACK_XOFF from netdev_tx_sent_queue()
From dev_watchdog(), we have to read txq->state before txq->trans_start.
Add memory barriers to enforce correct ordering.
In the future, we could avoid writing over txq->trans_start for normal
operations, and rename this field to txq->xoff_start_time.
Fixes: bec251bc8b ("net: no longer stop all TX queues in dev_watchdog()")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Toke Høiland-Jørgensen <toke@redhat.com>
Link: https://patch.msgid.link/20241015194118.3951657-1-edumazet@google.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
asm/unaligned.h is always an include of asm-generic/unaligned.h;
might as well move that thing to linux/unaligned.h and include
that - there's nothing arch-specific in that header.
auto-generated by the following:
for i in `git grep -l -w asm/unaligned.h`; do
sed -i -e "s/asm\/unaligned.h/linux\/unaligned.h/" $i
done
for i in `git grep -l -w asm-generic/unaligned.h`; do
sed -i -e "s/asm-generic\/unaligned.h/linux\/unaligned.h/" $i
done
git mv include/asm-generic/unaligned.h include/linux/unaligned.h
git mv tools/include/asm-generic/unaligned.h tools/include/linux/unaligned.h
sed -i -e "/unaligned.h/d" include/asm-generic/Kbuild
sed -i -e "s/__ASM_GENERIC/__LINUX/" include/linux/unaligned.h tools/include/linux/unaligned.h
sch_cake uses a cache of the first 16 values of the inverse square root
calculation for the Cobalt AQM to save some cycles on the fast path.
This cache is populated when the qdisc is first loaded, but there's
really no reason why it can't just be pre-populated. So change it to be
pre-populated with constants, which also makes it possible to constify
it.
This gives a modest space saving for the module (not counting debug data):
.text: -224 bytes
.rodata: +80 bytes
.bss: -64 bytes
Total: -192 bytes
Signed-off-by: Dave Taht <dave.taht@gmail.com>
[ fixed up comment, rewrote commit message ]
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Link: https://patch.msgid.link/20240909091630.22177-1-toke@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
According to Vinicius (and carefully looking through the whole
https://syzkaller.appspot.com/bug?extid=b65e0af58423fc8a73aa
once again), txtime branch of 'taprio_change()' is not going to
race against 'advance_sched()'. But using 'rcu_replace_pointer()'
in the former may be a good idea as well.
Suggested-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Signed-off-by: Dmitry Antipov <dmantipov@yandex.ru>
Acked-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In sch_cake, we keep track of the count of active bulk flows per host,
when running in dst/src host fairness mode, which is used as the
round-robin weight when iterating through flows. The count of active
bulk flows is updated whenever a flow changes state.
This has a peculiar interaction with the hash collision handling: when a
hash collision occurs (after the set-associative hashing), the state of
the hash bucket is simply updated to match the new packet that collided,
and if host fairness is enabled, that also means assigning new per-host
state to the flow. For this reason, the bulk flow counters of the
host(s) assigned to the flow are decremented, before new state is
assigned (and the counters, which may not belong to the same host
anymore, are incremented again).
Back when this code was introduced, the host fairness mode was always
enabled, so the decrement was unconditional. When the configuration
flags were introduced the *increment* was made conditional, but
the *decrement* was not. Which of course can lead to a spurious
decrement (and associated wrap-around to U16_MAX).
AFAICT, when host fairness is disabled, the decrement and wrap-around
happens as soon as a hash collision occurs (which is not that common in
itself, due to the set-associative hashing). However, in most cases this
is harmless, as the value is only used when host fairness mode is
enabled. So in order to trigger an array overflow, sch_cake has to first
be configured with host fairness disabled, and while running in this
mode, a hash collision has to occur to cause the overflow. Then, the
qdisc has to be reconfigured to enable host fairness, which leads to the
array out-of-bounds because the wrapped-around value is retained and
used as an array index. It seems that syzbot managed to trigger this,
which is quite impressive in its own right.
This patch fixes the issue by introducing the same conditional check on
decrement as is used on increment.
The original bug predates the upstreaming of cake, but the commit listed
in the Fixes tag touched that code, meaning that this patch won't apply
before that.
Fixes: 7126399299 ("sch_cake: Make the dual modes fairer")
Reported-by: syzbot+7fe7b81d602cc1e6b94d@syzkaller.appspotmail.com
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Link: https://patch.msgid.link/20240903160846.20909-1-toke@redhat.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
If netem_dequeue() enqueues packet to inner qdisc and that qdisc
returns __NET_XMIT_STOLEN. The packet is dropped but
qdisc_tree_reduce_backlog() is not called to update the parent's
q.qlen, leading to the similar use-after-free as Commit
e04991a48dbaf382 ("netem: fix return value if duplicate enqueue
fails")
Commands to trigger KASAN UaF:
ip link add type dummy
ip link set lo up
ip link set dummy0 up
tc qdisc add dev lo parent root handle 1: drr
tc filter add dev lo parent 1: basic classid 1:1
tc class add dev lo classid 1:1 drr
tc qdisc add dev lo parent 1:1 handle 2: netem
tc qdisc add dev lo parent 2: handle 3: drr
tc filter add dev lo parent 3: basic classid 3:1 action mirred egress
redirect dev dummy0
tc class add dev lo classid 3:1 drr
ping -c1 -W0.01 localhost # Trigger bug
tc class del dev lo classid 1:1
tc class add dev lo classid 1:1 drr
ping -c1 -W0.01 localhost # UaF
Fixes: 50612537e9 ("netem: fix classful handling")
Reported-by: Budimir Markovic <markovicbudimir@gmail.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Link: https://patch.msgid.link/20240901182438.4992-1-stephen@networkplumber.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
fq_dequeue() has a complex logic to find packets in one of the 3 bands.
As Neal found out, it is possible that one band has a deficit smaller
than its weight. fq_dequeue() can return NULL while some packets are
elligible for immediate transmit.
In this case, more than one iteration is needed to refill pband->credit.
With default parameters (weights 589824 196608 65536) bug can trigger
if large BIG TCP packets are sent to the lowest priority band.
Bisected-by: John Sperbeck <jsperbeck@google.com>
Diagnosed-by: Neal Cardwell <ncardwell@google.com>
Fixes: 29f834aa32 ("net_sched: sch_fq: add 3 bands and WRR scheduling")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Neal Cardwell <ncardwell@google.com>
Link: https://patch.msgid.link/20240824181901.953776-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
<tldr>
skb network header of the single-tagged vlan packet continues to point the
vlan payload (e.g. IP) after second vlan tag is pushed by tc act_vlan. This
causes problem at the dissector which expects double-tagged packet network
header to point to the inner vlan.
The fix is to adjust network header in tcf_act_vlan.c but requires
refactoring of skb_vlan_push function.
</tldr>
Consider the following shell script snippet configuring TC rules on the
veth interface:
ip link add veth0 type veth peer veth1
ip link set veth0 up
ip link set veth1 up
tc qdisc add dev veth0 clsact
tc filter add dev veth0 ingress pref 10 chain 0 flower \
num_of_vlans 2 cvlan_ethtype 0x800 action goto chain 5
tc filter add dev veth0 ingress pref 20 chain 0 flower \
num_of_vlans 1 action vlan push id 100 \
protocol 0x8100 action goto chain 5
tc filter add dev veth0 ingress pref 30 chain 5 flower \
num_of_vlans 2 cvlan_ethtype 0x800 action simple sdata "success"
Sending double-tagged vlan packet with the IP payload inside:
cat <<ENDS | text2pcap - - | tcpreplay -i veth1 -
0000 00 00 00 00 00 11 00 00 00 00 00 22 81 00 00 64 ..........."...d
0010 81 00 00 14 08 00 45 04 00 26 04 d2 00 00 7f 11 ......E..&......
0020 18 ef 0a 00 00 01 14 00 00 02 00 00 00 00 00 12 ................
0030 e1 c7 00 00 00 00 00 00 00 00 00 00 ............
ENDS
will match rule 10, goto rule 30 in chain 5 and correctly emit "success" to
the dmesg.
OTOH, sending single-tagged vlan packet:
cat <<ENDS | text2pcap - - | tcpreplay -i veth1 -
0000 00 00 00 00 00 11 00 00 00 00 00 22 81 00 00 14 ..........."....
0010 08 00 45 04 00 2a 04 d2 00 00 7f 11 18 eb 0a 00 ..E..*..........
0020 00 01 14 00 00 02 00 00 00 00 00 16 e1 bf 00 00 ................
0030 00 00 00 00 00 00 00 00 00 00 00 00 ............
ENDS
will match rule 20, will push the second vlan tag but will *not* match
rule 30. IOW, the match at rule 30 fails if the second vlan was freshly
pushed by the kernel.
Lets look at __skb_flow_dissect working on the double-tagged vlan packet.
Here is the relevant code from around net/core/flow_dissector.c:1277
copy-pasted here for convenience:
if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX &&
skb && skb_vlan_tag_present(skb)) {
proto = skb->protocol;
} else {
vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan),
data, hlen, &_vlan);
if (!vlan) {
fdret = FLOW_DISSECT_RET_OUT_BAD;
break;
}
proto = vlan->h_vlan_encapsulated_proto;
nhoff += sizeof(*vlan);
}
The "else" clause above gets the protocol of the encapsulated packet from
the skb data at the network header location. printk debugging has showed
that in the good double-tagged packet case proto is
htons(0x800 == ETH_P_IP) as expected. However in the single-tagged packet
case proto is garbage leading to the failure to match tc filter 30.
proto is being set from the skb header pointed by nhoff parameter which is
defined at the beginning of __skb_flow_dissect
(net/core/flow_dissector.c:1055 in the current version):
nhoff = skb_network_offset(skb);
Therefore the culprit seems to be that the skb network offset is different
between double-tagged packet received from the interface and single-tagged
packet having its vlan tag pushed by TC.
Lets look at the interesting points of the lifetime of the single/double
tagged packets as they traverse our packet flow.
Both of them will start at __netif_receive_skb_core where the first vlan
tag will be stripped:
if (eth_type_vlan(skb->protocol)) {
skb = skb_vlan_untag(skb);
if (unlikely(!skb))
goto out;
}
At this stage in double-tagged case skb->data points to the second vlan tag
while in single-tagged case skb->data points to the network (eg. IP)
header.
Looking at TC vlan push action (net/sched/act_vlan.c) we have the following
code at tcf_vlan_act (interesting points are in square brackets):
if (skb_at_tc_ingress(skb))
[1] skb_push_rcsum(skb, skb->mac_len);
....
case TCA_VLAN_ACT_PUSH:
err = skb_vlan_push(skb, p->tcfv_push_proto, p->tcfv_push_vid |
(p->tcfv_push_prio << VLAN_PRIO_SHIFT),
0);
if (err)
goto drop;
break;
....
out:
if (skb_at_tc_ingress(skb))
[3] skb_pull_rcsum(skb, skb->mac_len);
And skb_vlan_push (net/core/skbuff.c:6204) function does:
err = __vlan_insert_tag(skb, skb->vlan_proto,
skb_vlan_tag_get(skb));
if (err)
return err;
skb->protocol = skb->vlan_proto;
[2] skb->mac_len += VLAN_HLEN;
in the case of pushing the second tag. Lets look at what happens with
skb->data of the single-tagged packet at each of the above points:
1. As a result of the skb_push_rcsum, skb->data is moved back to the start
of the packet.
2. First VLAN tag is moved from the skb into packet buffer, skb->mac_len is
incremented, skb->data still points to the start of the packet.
3. As a result of the skb_pull_rcsum, skb->data is moved forward by the
modified skb->mac_len, thus pointing to the network header again.
Then __skb_flow_dissect will get confused by having double-tagged vlan
packet with the skb->data at the network header.
The solution for the bug is to preserve "skb->data at second vlan header"
semantics in the skb_vlan_push function. We do this by manipulating
skb->network_header rather than skb->mac_len. skb_vlan_push callers are
updated to do skb_reset_mac_len.
Signed-off-by: Boris Sukholitko <boris.sukholitko@broadcom.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Cross-merge networking fixes after downstream PR.
No conflicts.
Adjacent changes:
drivers/net/ethernet/broadcom/bnxt/bnxt.h
c948c0973d ("bnxt_en: Don't clear ntuple filters and rss contexts during ethtool ops")
f2878cdeb7 ("bnxt_en: Add support to call FW to update a VNIC")
Link: https://patch.msgid.link/20240822210125.1542769-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
There is a bug in netem_enqueue() introduced by
commit 5845f70638 ("net: netem: fix skb length BUG_ON in __skb_to_sgvec")
that can lead to a use-after-free.
This commit made netem_enqueue() always return NET_XMIT_SUCCESS
when a packet is duplicated, which can cause the parent qdisc's q.qlen
to be mistakenly incremented. When this happens qlen_notify() may be
skipped on the parent during destruction, leaving a dangling pointer
for some classful qdiscs like DRR.
There are two ways for the bug happen:
- If the duplicated packet is dropped by rootq->enqueue() and then
the original packet is also dropped.
- If rootq->enqueue() sends the duplicated packet to a different qdisc
and the original packet is dropped.
In both cases NET_XMIT_SUCCESS is returned even though no packets
are enqueued at the netem qdisc.
The fix is to defer the enqueue of the duplicate packet until after
the original packet has been guaranteed to return NET_XMIT_SUCCESS.
Fixes: 5845f70638 ("net: netem: fix skb length BUG_ON in __skb_to_sgvec")
Reported-by: Budimir Markovic <markovicbudimir@gmail.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20240819175753.5151-1-stephen@networkplumber.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
-Wflex-array-member-not-at-end was introduced in GCC-14, and we are
getting ready to enable it, globally.
Remove unnecessary flex-array member `pad[]` and refactor the related
code a bit.
Fix the following warning:
net/sched/act_ct.c:57:29: warning: structure containing a flexible array member is not at the end of another structure [-Wflex-array-member-not-at-end]
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Link: https://patch.msgid.link/ZrY0JMVsImbDbx6r@cute
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Make sure to set encapsulated control flags also for non-IP
packets, such that it's possible to allow matching on e.g.
TUNNEL_OAM on a geneve packet carrying a non-IP packet.
Suggested-by: Davide Caratti <dcaratti@redhat.com>
Signed-off-by: Asbjørn Sloth Tønnesen <ast@fiberby.net>
Tested-by: Davide Caratti <dcaratti@redhat.com>
Reviewed-by: Davide Caratti <dcaratti@redhat.com>
Link: https://patch.msgid.link/20240713021911.1631517-13-ast@fiberby.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Now that TCA_FLOWER_KEY_ENC_FLAGS is unused, as it's
former data is stored behind TCA_FLOWER_KEY_ENC_CONTROL,
then remove the last bits of FLOW_DISSECTOR_KEY_ENC_FLAGS.
FLOW_DISSECTOR_KEY_ENC_FLAGS is unreleased, and have been
in net-next since 2024-06-04.
Signed-off-by: Asbjørn Sloth Tønnesen <ast@fiberby.net>
Tested-by: Davide Caratti <dcaratti@redhat.com>
Reviewed-by: Davide Caratti <dcaratti@redhat.com>
Link: https://patch.msgid.link/20240713021911.1631517-12-ast@fiberby.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This patch changes how TCA_FLOWER_KEY_ENC_FLAGS is used, so that
it is used with TCA_FLOWER_KEY_FLAGS_* flags, in the same way as
TCA_FLOWER_KEY_FLAGS is currently used.
Where TCA_FLOWER_KEY_FLAGS uses {key,mask}->control.flags, then
TCA_FLOWER_KEY_ENC_FLAGS now uses {key,mask}->enc_control.flags,
therefore {key,mask}->enc_flags is now unused.
As the generic fl_set_key_flags/fl_dump_key_flags() is used with
encap set to true, then fl_{set,dump}_key_enc_flags() is removed.
This breaks unreleased userspace API (net-next since 2024-06-04).
Signed-off-by: Asbjørn Sloth Tønnesen <ast@fiberby.net>
Tested-by: Davide Caratti <dcaratti@redhat.com>
Reviewed-by: Davide Caratti <dcaratti@redhat.com>
Link: https://patch.msgid.link/20240713021911.1631517-10-ast@fiberby.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Prepare to set and dump the tunnel flags.
This code won't see any of these flags yet, as these flags
aren't allowed by the NLA_POLICY_MASK, and the functions
doesn't get called with encap set to true yet.
Signed-off-by: Asbjørn Sloth Tønnesen <ast@fiberby.net>
Tested-by: Davide Caratti <dcaratti@redhat.com>
Reviewed-by: Davide Caratti <dcaratti@redhat.com>
Link: https://patch.msgid.link/20240713021911.1631517-9-ast@fiberby.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This policy guards fl_set_key_flags() from seeing flags
not used in the context of TCA_FLOWER_KEY_FLAGS.
In order For the policy check to be performed with the
correct endianness, then we also needs to change the
attribute type to NLA_BE32 (Thanks Davide).
TCA_FLOWER_KEY_FLAGS{,_MASK} already has a be32 comment
in include/uapi/linux/pkt_cls.h.
Signed-off-by: Asbjørn Sloth Tønnesen <ast@fiberby.net>
Tested-by: Davide Caratti <dcaratti@redhat.com>
Reviewed-by: Davide Caratti <dcaratti@redhat.com>
Link: https://patch.msgid.link/20240713021911.1631517-6-ast@fiberby.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Prepare fl_set_key_flags/fl_dump_key_flags() for use with
TCA_FLOWER_KEY_ENC_FLAGS{,_MASK}.
This patch adds an encap argument, similar to fl_set_key_ip/
fl_dump_key_ip(), and determine the flower keys based on the
encap argument, and use them in the rest of the two functions.
Since these functions are so far, only called with encap set false,
then there is no functional change.
Signed-off-by: Asbjørn Sloth Tønnesen <ast@fiberby.net>
Tested-by: Davide Caratti <dcaratti@redhat.com>
Reviewed-by: Davide Caratti <dcaratti@redhat.com>
Link: https://patch.msgid.link/20240713021911.1631517-5-ast@fiberby.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In prevision to add new UAPI for hwtstamp we will be limited to the struct
ethtool_ts_info that is currently passed in fixed binary format through the
ETHTOOL_GET_TS_INFO ethtool ioctl. It would be good if new kernel code
already started operating on an extensible kernel variant of that
structure, similar in concept to struct kernel_hwtstamp_config vs struct
hwtstamp_config.
Since struct ethtool_ts_info is in include/uapi/linux/ethtool.h, here
we introduce the kernel-only structure in include/linux/ethtool.h.
The manual copy is then made in the function called by ETHTOOL_GET_TS_INFO.
Acked-by: Shannon Nelson <shannon.nelson@amd.com>
Acked-by: Alexandra Winter <wintera@linux.ibm.com>
Signed-off-by: Kory Maincent <kory.maincent@bootlin.com>
Link: https://patch.msgid.link/20240709-feature_ptp_netnext-v17-6-b5317f50df2a@bootlin.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Replace a comma between expression statements by a semicolon.
Signed-off-by: Chen Ni <nichen@iscas.ac.cn>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20240709072838.1152880-1-nichen@iscas.ac.cn
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Cross-merge networking fixes after downstream PR.
Conflicts:
net/sched/act_ct.c
26488172b0 ("net/sched: Fix UAF when resolving a clash")
3abbd7ed8b ("act_ct: prepare for stolen verdict coming from conntrack and nat engine")
No adjacent changes.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Pedro Pinto and later independently also Hyunwoo Kim and Wongi Lee reported
an issue that the tcx_entry can be released too early leading to a use
after free (UAF) when an active old-style ingress or clsact qdisc with a
shared tc block is later replaced by another ingress or clsact instance.
Essentially, the sequence to trigger the UAF (one example) can be as follows:
1. A network namespace is created
2. An ingress qdisc is created. This allocates a tcx_entry, and
&tcx_entry->miniq is stored in the qdisc's miniqp->p_miniq. At the
same time, a tcf block with index 1 is created.
3. chain0 is attached to the tcf block. chain0 must be connected to
the block linked to the ingress qdisc to later reach the function
tcf_chain0_head_change_cb_del() which triggers the UAF.
4. Create and graft a clsact qdisc. This causes the ingress qdisc
created in step 1 to be removed, thus freeing the previously linked
tcx_entry:
rtnetlink_rcv_msg()
=> tc_modify_qdisc()
=> qdisc_create()
=> clsact_init() [a]
=> qdisc_graft()
=> qdisc_destroy()
=> __qdisc_destroy()
=> ingress_destroy() [b]
=> tcx_entry_free()
=> kfree_rcu() // tcx_entry freed
5. Finally, the network namespace is closed. This registers the
cleanup_net worker, and during the process of releasing the
remaining clsact qdisc, it accesses the tcx_entry that was
already freed in step 4, causing the UAF to occur:
cleanup_net()
=> ops_exit_list()
=> default_device_exit_batch()
=> unregister_netdevice_many()
=> unregister_netdevice_many_notify()
=> dev_shutdown()
=> qdisc_put()
=> clsact_destroy() [c]
=> tcf_block_put_ext()
=> tcf_chain0_head_change_cb_del()
=> tcf_chain_head_change_item()
=> clsact_chain_head_change()
=> mini_qdisc_pair_swap() // UAF
There are also other variants, the gist is to add an ingress (or clsact)
qdisc with a specific shared block, then to replace that qdisc, waiting
for the tcx_entry kfree_rcu() to be executed and subsequently accessing
the current active qdisc's miniq one way or another.
The correct fix is to turn the miniq_active boolean into a counter. What
can be observed, at step 2 above, the counter transitions from 0->1, at
step [a] from 1->2 (in order for the miniq object to remain active during
the replacement), then in [b] from 2->1 and finally [c] 1->0 with the
eventual release. The reference counter in general ranges from [0,2] and
it does not need to be atomic since all access to the counter is protected
by the rtnl mutex. With this in place, there is no longer a UAF happening
and the tcx_entry is freed at the correct time.
Fixes: e420bed025 ("bpf: Add fd-based tcx multi-prog infra with link support")
Reported-by: Pedro Pinto <xten@osec.io>
Co-developed-by: Pedro Pinto <xten@osec.io>
Signed-off-by: Pedro Pinto <xten@osec.io>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Cc: Hyunwoo Kim <v4bel@theori.io>
Cc: Wongi Lee <qwerty@theori.io>
Cc: Martin KaFai Lau <martin.lau@kernel.org>
Link: https://lore.kernel.org/r/20240708133130.11609-1-daniel@iogearbox.net
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
At this time, conntrack either returns NF_ACCEPT or NF_DROP.
To improve debuging it would be nice to be able to replace NF_DROP verdict
with NF_DROP_REASON() helper,
This helper releases the skb instantly (so drop_monitor can pinpoint
exact location) and returns NF_STOLEN.
Prepare call sites to deal with this before introducing such changes
in conntrack and nat core.
Signed-off-by: Florian Westphal <fw@strlen.de>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
If the action has a user_cookie, pass it along to the sample so it can
be easily identified.
Reviewed-by: Michal Kubiak <michal.kubiak@intel.com>
Reviewed-by: Aaron Conole <aconole@redhat.com>
Acked-by: Eelco Chaudron <echaudro@redhat.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Adrian Moreno <amorenoz@redhat.com>
Link: https://patch.msgid.link/20240704085710.353845-3-amorenoz@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
