Jann Horn reported that SO_PEERCRED and SO_PEERGROUPS implementations
are racy, as af_unix can concurrently change sk_peer_pid and sk_peer_cred.
In order to fix this issue, this patch adds a new spinlock that needs
to be used whenever these fields are read or written.
Jann also pointed out that l2cap_sock_get_peer_pid_cb() is currently
reading sk->sk_peer_pid which makes no sense, as this field
is only possibly set by AF_UNIX sockets.
We will have to clean this in a separate patch.
This could be done by reverting b48596d1dc "Bluetooth: L2CAP: Add get_peer_pid callback"
or implementing what was truly expected.
Fixes: 109f6e39fa ("af_unix: Allow SO_PEERCRED to work across namespaces.")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Jann Horn <jannh@google.com>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Cc: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Cc: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
struct dev_addr_list is used for device addresses, unicast addresses
and multicast addresses. The first of those needs special handling
of the main address - netdev->dev_addr points directly the data
of the entry and drivers write to it freely, so we can't maintain
it in the rbtree (for now, at least, to be fixed in net-next).
Current work around sprinkles special handling of the first
address on the list throughout the code but it missed the case
where address is being added. First address will not be visible
during subsequent adds.
Syzbot found a warning where unicast addresses are modified
without holding the rtnl lock, tl;dr is that team generates
the same modification multiple times, not necessarily when
right locks are held.
In the repro we have:
macvlan -> team -> veth
macvlan adds a unicast address to the team. Team then pushes
that address down to its memebers (veths). Next something unrelated
makes team sync member addrs again, and because of the bug
the addr entries get duplicated in the veths. macvlan gets
removed, removes its addr from team which removes only one
of the duplicated addresses from veths. This removal is done
under rtnl. Next syzbot uses iptables to add a multicast addr
to team (which does not hold rtnl lock). Team syncs veth addrs,
but because veths' unicast list still has the duplicate it will
also get sync, even though this update is intended for mc addresses.
Again, uc address updates need rtnl lock, boom.
Reported-by: syzbot+7a2ab2cdc14d134de553@syzkaller.appspotmail.com
Fixes: 406f42fa0d ("net-next: When a bond have a massive amount of VLANs with IPv6 addresses, performance of changing link state, attaching a VRF, changing an IPv6 address, etc. go down dramtically.")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The process will cause napi.state to contain NAPI_STATE_SCHED and
not in the poll_list, which will cause napi_disable() to get stuck.
The prefix "NAPI_STATE_" is removed in the figure below, and
NAPI_STATE_HASHED is ignored in napi.state.
CPU0 | CPU1 | napi.state
===============================================================================
napi_disable() | | SCHED | NPSVC
napi_enable() | |
{ | |
smp_mb__before_atomic(); | |
clear_bit(SCHED, &n->state); | | NPSVC
| napi_schedule_prep() | SCHED | NPSVC
| napi_poll() |
| napi_complete_done() |
| { |
| if (n->state & (NPSVC | | (1)
| _BUSY_POLL))) |
| return false; |
| ................ |
| } | SCHED | NPSVC
| |
clear_bit(NPSVC, &n->state); | | SCHED
} | |
| |
napi_schedule_prep() | | SCHED | MISSED (2)
(1) Here return direct. Because of NAPI_STATE_NPSVC exists.
(2) NAPI_STATE_SCHED exists. So not add napi.poll_list to sd->poll_list
Since NAPI_STATE_SCHED already exists and napi is not in the
sd->poll_list queue, NAPI_STATE_SCHED cannot be cleared and will always
exist.
1. This will cause this queue to no longer receive packets.
2. If you encounter napi_disable under the protection of rtnl_lock, it
will cause the entire rtnl_lock to be locked, affecting the overall
system.
This patch uses cmpxchg to implement napi_enable(), which ensures that
there will be no race due to the separation of clear two bits.
Fixes: 2d8bff1269 ("netpoll: Close race condition between poll_one_napi and napi_disable")
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Reviewed-by: Dust Li <dust.li@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
lock_sock_fast() and lock_sock_nested() contain lockdep annotations for the
sock::sk_lock.owned 'mutex'. sock::sk_lock.owned is not a regular mutex. It
is just lockdep wise equivalent. In fact it's an open coded trivial mutex
implementation with some interesting features.
sock::sk_lock.slock is a regular spinlock protecting the 'mutex'
representation sock::sk_lock.owned which is a plain boolean. If 'owned' is
true, then some other task holds the 'mutex', otherwise it is uncontended.
As this locking construct is obviously endangered by lock ordering issues as
any other locking primitive it got lockdep annotated via a dedicated
dependency map sock::sk_lock.dep_map which has to be updated at the lock
and unlock sites.
lock_sock_nested() is a straight forward 'mutex' lock operation:
might_sleep();
spin_lock_bh(sock::sk_lock.slock)
while (!try_lock(sock::sk_lock.owned)) {
spin_unlock_bh(sock::sk_lock.slock);
wait_for_release();
spin_lock_bh(sock::sk_lock.slock);
}
The lockdep annotation for sock::sk_lock.owned is for unknown reasons
_after_ the lock has been acquired, i.e. after the code block above and
after releasing sock::sk_lock.slock, but inside the bottom halves disabled
region:
spin_unlock(sock::sk_lock.slock);
mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
local_bh_enable();
The placement after the unlock is obvious because otherwise the
mutex_acquire() would nest into the spin lock held region.
But that's from the lockdep perspective still the wrong place:
1) The mutex_acquire() is issued _after_ the successful acquisition which
is pointless because in a dead lock scenario this point is never
reached which means that if the deadlock is the first instance of
exposing the wrong lock order lockdep does not have a chance to detect
it.
2) It only works because lockdep is rather lax on the context from which
the mutex_acquire() is issued. Acquiring a mutex inside a bottom halves
and therefore non-preemptible region is obviously invalid, except for a
trylock which is clearly not the case here.
This 'works' stops working on RT enabled kernels where the bottom halves
serialization is done via a local lock, which exposes this misplacement
because the 'mutex' and the local lock nest the wrong way around and
lockdep complains rightfully about a lock inversion.
The placement is wrong since the initial commit a5b5bb9a05 ("[PATCH]
lockdep: annotate sk_locks") which introduced this.
Fix it by moving the mutex_acquire() in front of the actual lock
acquisition, which is what the regular mutex_lock() operation does as well.
lock_sock_fast() is not that straight forward. It looks at the first glance
like a convoluted trylock operation:
spin_lock_bh(sock::sk_lock.slock)
if (!sock::sk_lock.owned)
return false;
while (!try_lock(sock::sk_lock.owned)) {
spin_unlock_bh(sock::sk_lock.slock);
wait_for_release();
spin_lock_bh(sock::sk_lock.slock);
}
spin_unlock(sock::sk_lock.slock);
mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
local_bh_enable();
return true;
But that's not the case: lock_sock_fast() is an interesting optimization
for short critical sections which can run with bottom halves disabled and
sock::sk_lock.slock held. This allows to shortcut the 'mutex' operation in
the non contended case by preventing other lockers to acquire
sock::sk_lock.owned because they are blocked on sock::sk_lock.slock, which
in turn avoids the overhead of doing the heavy processing in release_sock()
including waking up wait queue waiters.
In the contended case, i.e. when sock::sk_lock.owned == true the behavior
is the same as lock_sock_nested().
Semantically this shortcut means, that the task acquired the 'mutex' even
if it does not touch the sock::sk_lock.owned field in the non-contended
case. Not telling lockdep about this shortcut acquisition is hiding
potential lock ordering violations in the fast path.
As a consequence the same reasoning as for the above lock_sock_nested()
case vs. the placement of the lockdep annotation applies.
The current placement of the lockdep annotation was just copied from
the original lock_sock(), now renamed to lock_sock_nested(),
implementation.
Fix this by moving the mutex_acquire() in front of the actual lock
acquisition and adding the corresponding mutex_release() into
unlock_sock_fast(). Also document the fast path return case with a comment.
Reported-by: Sebastian Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: netdev@vger.kernel.org
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Daniel Borkmann says:
====================
pull-request: bpf 2021-09-14
The following pull-request contains BPF updates for your *net* tree.
We've added 7 non-merge commits during the last 13 day(s) which contain
a total of 18 files changed, 334 insertions(+), 193 deletions(-).
The main changes are:
1) Fix mmap_lock lockdep splat in BPF stack map's build_id lookup, from Yonghong Song.
2) Fix BPF cgroup v2 program bypass upon net_cls/prio activation, from Daniel Borkmann.
3) Fix kvcalloc() BTF line info splat on oversized allocation attempts, from Bixuan Cui.
4) Fix BPF selftest build of task_pt_regs test for arm64/s390, from Jean-Philippe Brucker.
5) Fix BPF's disasm.{c,h} to dual-license so that it is aligned with bpftool given the former
is a build dependency for the latter, from Daniel Borkmann with ACKs from contributors.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix cgroup v1 interference when non-root cgroup v2 BPF programs are used.
Back in the days, commit bd1060a1d6 ("sock, cgroup: add sock->sk_cgroup")
embedded per-socket cgroup information into sock->sk_cgrp_data and in order
to save 8 bytes in struct sock made both mutually exclusive, that is, when
cgroup v1 socket tagging (e.g. net_cls/net_prio) is used, then cgroup v2
falls back to the root cgroup in sock_cgroup_ptr() (&cgrp_dfl_root.cgrp).
The assumption made was "there is no reason to mix the two and this is in line
with how legacy and v2 compatibility is handled" as stated in bd1060a1d6.
However, with Kubernetes more widely supporting cgroups v2 as well nowadays,
this assumption no longer holds, and the possibility of the v1/v2 mixed mode
with the v2 root fallback being hit becomes a real security issue.
Many of the cgroup v2 BPF programs are also used for policy enforcement, just
to pick _one_ example, that is, to programmatically deny socket related system
calls like connect(2) or bind(2). A v2 root fallback would implicitly cause
a policy bypass for the affected Pods.
In production environments, we have recently seen this case due to various
circumstances: i) a different 3rd party agent and/or ii) a container runtime
such as [0] in the user's environment configuring legacy cgroup v1 net_cls
tags, which triggered implicitly mentioned root fallback. Another case is
Kubernetes projects like kind [1] which create Kubernetes nodes in a container
and also add cgroup namespaces to the mix, meaning programs which are attached
to the cgroup v2 root of the cgroup namespace get attached to a non-root
cgroup v2 path from init namespace point of view. And the latter's root is
out of reach for agents on a kind Kubernetes node to configure. Meaning, any
entity on the node setting cgroup v1 net_cls tag will trigger the bypass
despite cgroup v2 BPF programs attached to the namespace root.
Generally, this mutual exclusiveness does not hold anymore in today's user
environments and makes cgroup v2 usage from BPF side fragile and unreliable.
This fix adds proper struct cgroup pointer for the cgroup v2 case to struct
sock_cgroup_data in order to address these issues; this implicitly also fixes
the tradeoffs being made back then with regards to races and refcount leaks
as stated in bd1060a1d6, and removes the fallback, so that cgroup v2 BPF
programs always operate as expected.
[0] https://github.com/nestybox/sysbox/
[1] https://kind.sigs.k8s.io/
Fixes: bd1060a1d6 ("sock, cgroup: add sock->sk_cgroup")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Stanislav Fomichev <sdf@google.com>
Acked-by: Tejun Heo <tj@kernel.org>
Link: https://lore.kernel.org/bpf/20210913230759.2313-1-daniel@iogearbox.net
There is a statement that is indented one character too deeply,
clean this up.
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
pktgen_thread_worker() no longer needs wait variable, delete it.
Fixes: ef87979c27 ("pktgen: better scheduler friendliness")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Daniel Borkmann says:
====================
bpf-next 2021-08-31
We've added 116 non-merge commits during the last 17 day(s) which contain
a total of 126 files changed, 6813 insertions(+), 4027 deletions(-).
The main changes are:
1) Add opaque bpf_cookie to perf link which the program can read out again,
to be used in libbpf-based USDT library, from Andrii Nakryiko.
2) Add bpf_task_pt_regs() helper to access userspace pt_regs, from Daniel Xu.
3) Add support for UNIX stream type sockets for BPF sockmap, from Jiang Wang.
4) Allow BPF TCP congestion control progs to call bpf_setsockopt() e.g. to switch
to another congestion control algorithm during init, from Martin KaFai Lau.
5) Extend BPF iterator support for UNIX domain sockets, from Kuniyuki Iwashima.
6) Allow bpf_{set,get}sockopt() calls from setsockopt progs, from Prankur Gupta.
7) Add bpf_get_netns_cookie() helper for BPF_PROG_TYPE_{SOCK_OPS,CGROUP_SOCKOPT}
progs, from Xu Liu and Stanislav Fomichev.
8) Support for __weak typed ksyms in libbpf, from Hao Luo.
9) Shrink struct cgroup_bpf by 504 bytes through refactoring, from Dave Marchevsky.
10) Fix a smatch complaint in verifier's narrow load handling, from Andrey Ignatov.
11) Fix BPF interpreter's tail call count limit, from Daniel Borkmann.
12) Big batch of improvements to BPF selftests, from Magnus Karlsson, Li Zhijian,
Yucong Sun, Yonghong Song, Ilya Leoshkevich, Jussi Maki, Ilya Leoshkevich, others.
13) Another big batch to revamp XDP samples in order to give them consistent look
and feel, from Kumar Kartikeya Dwivedi.
* https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next: (116 commits)
MAINTAINERS: Remove self from powerpc BPF JIT
selftests/bpf: Fix potential unreleased lock
samples: bpf: Fix uninitialized variable in xdp_redirect_cpu
selftests/bpf: Reduce more flakyness in sockmap_listen
bpf: Fix bpf-next builds without CONFIG_BPF_EVENTS
bpf: selftests: Add dctcp fallback test
bpf: selftests: Add connect_to_fd_opts to network_helpers
bpf: selftests: Add sk_state to bpf_tcp_helpers.h
bpf: tcp: Allow bpf-tcp-cc to call bpf_(get|set)sockopt
selftests: xsk: Preface options with opt
selftests: xsk: Make enums lower case
selftests: xsk: Generate packets from specification
selftests: xsk: Generate packet directly in umem
selftests: xsk: Simplify cleanup of ifobjects
selftests: xsk: Decrease sending speed
selftests: xsk: Validate tx stats on tx thread
selftests: xsk: Simplify packet validation in xsk tests
selftests: xsk: Rename worker_* functions that are not thread entry points
selftests: xsk: Disassociate umem size with packets sent
selftests: xsk: Remove end-of-test packet
...
====================
Link: https://lore.kernel.org/r/20210830225618.11634-1-daniel@iogearbox.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Pablo Neira Ayuso says:
====================
Netfilter updates for net-next
The following patchset contains Netfilter updates for net-next:
1) Clean up and consolidate ct ecache infrastructure by merging ct and
expect notifiers, from Florian Westphal.
2) Missing counters and timestamp in nfnetlink_queue and _log conntrack
information.
3) Missing error check for xt_register_template() in iptables mangle,
as a incremental fix for the previous pull request, also from
Florian Westphal.
4) Add netfilter hooks for the SRv6 lightweigh tunnel driver, from
Ryoga Sato. The hooks are enabled via nf_hooks_lwtunnel sysctl
to make sure existing netfilter rulesets do not break. There is
a static key to disable the hooks by default.
The pktgen_bench_xmit_mode_netif_receive.sh shows no noticeable
impact in the seg6_input path for non-netfilter users: similar
numbers with and without this patch.
This is a sample of the perf report output:
11.67% kpktgend_0 [ipv6] [k] ipv6_get_saddr_eval
7.89% kpktgend_0 [ipv6] [k] __ipv6_addr_label
7.52% kpktgend_0 [ipv6] [k] __ipv6_dev_get_saddr
6.63% kpktgend_0 [kernel.vmlinux] [k] asm_exc_nmi
4.74% kpktgend_0 [ipv6] [k] fib6_node_lookup_1
3.48% kpktgend_0 [kernel.vmlinux] [k] pskb_expand_head
3.33% kpktgend_0 [ipv6] [k] ip6_rcv_core.isra.29
3.33% kpktgend_0 [ipv6] [k] seg6_do_srh_encap
2.53% kpktgend_0 [ipv6] [k] ipv6_dev_get_saddr
2.45% kpktgend_0 [ipv6] [k] fib6_table_lookup
2.24% kpktgend_0 [kernel.vmlinux] [k] ___cache_free
2.16% kpktgend_0 [ipv6] [k] ip6_pol_route
2.11% kpktgend_0 [kernel.vmlinux] [k] __ipv6_addr_type
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch introduces netfilter hooks for solving the problem that
conntrack couldn't record both inner flows and outer flows.
This patch also introduces a new sysctl toggle for enabling lightweight
tunnel netfilter hooks.
Signed-off-by: Ryoga Saito <contact@proelbtn.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Currently when device is moved between network namespaces using
RTM_NEWLINK message type and one of netns attributes (FLA_NET_NS_PID,
IFLA_NET_NS_FD, IFLA_TARGET_NETNSID) but w/o specifying IFLA_IFNAME, and
target namespace already has device with same name, userspace will get
EINVAL what is confusing and makes debugging harder.
Fix it so that userspace gets more appropriate EEXIST instead what makes
debugging much easier.
Before:
# ./ifname.sh
+ ip netns add ns0
+ ip netns exec ns0 ip link add l0 type dummy
+ ip netns exec ns0 ip link show l0
8: l0: <BROADCAST,NOARP> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
link/ether 66:90:b5:d5:78:69 brd ff:ff:ff:ff:ff:ff
+ ip link add l0 type dummy
+ ip link show l0
10: l0: <BROADCAST,NOARP> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
link/ether 6e:c6:1f:15:20:8d brd ff:ff:ff:ff:ff:ff
+ ip link set l0 netns ns0
RTNETLINK answers: Invalid argument
After:
# ./ifname.sh
+ ip netns add ns0
+ ip netns exec ns0 ip link add l0 type dummy
+ ip netns exec ns0 ip link show l0
8: l0: <BROADCAST,NOARP> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
link/ether 1e:4a:72:e3:e3:8f brd ff:ff:ff:ff:ff:ff
+ ip link add l0 type dummy
+ ip link show l0
10: l0: <BROADCAST,NOARP> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
link/ether f2:fc:fe:2b:7d:a6 brd ff:ff:ff:ff:ff:ff
+ ip link set l0 netns ns0
RTNETLINK answers: File exists
The problem is that do_setlink() passes its `char *ifname` argument,
that it gets from a caller, to __dev_change_net_namespace() as is (as
`const char *pat`), but semantics of ifname and pat can be different.
For example, __rtnl_newlink() does this:
net/core/rtnetlink.c
3270 char ifname[IFNAMSIZ];
...
3286 if (tb[IFLA_IFNAME])
3287 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3288 else
3289 ifname[0] = '\0';
...
3364 if (dev) {
...
3394 return do_setlink(skb, dev, ifm, extack, tb, ifname, status);
3395 }
, i.e. do_setlink() gets ifname pointer that is always valid no matter
if user specified IFLA_IFNAME or not and then do_setlink() passes this
ifname pointer as is to __dev_change_net_namespace() as pat argument.
But the pat (pattern) in __dev_change_net_namespace() is used as:
net/core/dev.c
11198 err = -EEXIST;
11199 if (__dev_get_by_name(net, dev->name)) {
11200 /* We get here if we can't use the current device name */
11201 if (!pat)
11202 goto out;
11203 err = dev_get_valid_name(net, dev, pat);
11204 if (err < 0)
11205 goto out;
11206 }
As the result the `goto out` path on line 11202 is neven taken and
instead of returning EEXIST defined on line 11198,
__dev_change_net_namespace() returns an error from dev_get_valid_name()
and this, in turn, will be EINVAL for ifname[0] = '\0' set earlier.
Fixes: d8a5ec6727 ("[NET]: netlink support for moving devices between network namespaces.")
Signed-off-by: Andrey Ignatov <rdna@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Both SKB_FRAG_PAGE_ORDER are defined to the same value in
net/core/sock.c and drivers/vhost/net.c.
Move the SKB_FRAG_PAGE_ORDER definition to net/core/sock.h,
as both net/core/sock.c and drivers/vhost/net.c include it,
and it seems a reasonable file to put the macro.
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch allows the bpf-tcp-cc to call bpf_setsockopt. One use
case is to allow a bpf-tcp-cc switching to another cc during init().
For example, when the tcp flow is not ecn ready, the bpf_dctcp
can switch to another cc by calling setsockopt(TCP_CONGESTION).
During setsockopt(TCP_CONGESTION), the new tcp-cc's init() will be
called and this could cause a recursion but it is stopped by the
current trampoline's logic (in the prog->active counter).
While retiring a bpf-tcp-cc (e.g. in tcp_v[46]_destroy_sock()),
the tcp stack calls bpf-tcp-cc's release(). To avoid the retiring
bpf-tcp-cc making further changes to the sk, bpf_setsockopt is not
available to the bpf-tcp-cc's release(). This will avoid release()
making setsockopt() call that will potentially allocate new resources.
Although the bpf-tcp-cc already has a more powerful way to read tcp_sock
from the PTR_TO_BTF_ID, it is usually expected that bpf_getsockopt and
bpf_setsockopt are available together. Thus, bpf_getsockopt() is also
added to all tcp_congestion_ops except release().
When the old bpf-tcp-cc is calling setsockopt(TCP_CONGESTION)
to switch to a new cc, the old bpf-tcp-cc will be released by
bpf_struct_ops_put(). Thus, this patch also puts the bpf_struct_ops_map
after a rcu grace period because the trampoline's image cannot be freed
while the old bpf-tcp-cc is still running.
bpf-tcp-cc can only access icsk_ca_priv as SCALAR. All kernel's
tcp-cc is also accessing the icsk_ca_priv as SCALAR. The size
of icsk_ca_priv has already been raised a few times to avoid
extra kmalloc and memory referencing. The only exception is the
kernel's tcp_cdg.c that stores a kmalloc()-ed pointer in icsk_ca_priv.
To avoid the old bpf-tcp-cc accidentally overriding this tcp_cdg's pointer
value stored in icsk_ca_priv after switching and without over-complicating
the bpf's verifier for this one exception in tcp_cdg, this patch does not
allow switching to tcp_cdg. If there is a need, bpf_tcp_cdg can be
implemented and then use the bpf_sk_storage as the extended storage.
bpf_sk_setsockopt proto has only been recently added and used
in bpf-sockopt and bpf-iter-tcp, so impose the tcp_cdg limitation in the
same proto instead of adding a new proto specifically for bpf-tcp-cc.
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20210824173007.3976921-1-kafai@fb.com
The source of most of the slow down is the `dev_addr_lists.c` module,
which mainatins a linked list of HW addresses.
When using IPv6, this list grows for each IPv6 address added on a
VLAN, since each IPv6 address has a multicast HW address associated with
it.
When performing any modification to the involved links, this list is
traversed many times, often for nothing, all while holding the RTNL
lock.
Instead, this patch adds an auxilliary rbtree which cuts down
traversal time significantly.
Performance can be seen with the following script:
#!/bin/bash
ip netns del test || true 2>/dev/null
ip netns add test
echo 1 | ip netns exec test tee /proc/sys/net/ipv6/conf/all/keep_addr_on_down > /dev/null
set -e
ip -n test link add foo type veth peer name bar
ip -n test link add b1 type bond
ip -n test link add florp type vrf table 10
ip -n test link set bar master b1
ip -n test link set foo up
ip -n test link set bar up
ip -n test link set b1 up
ip -n test link set florp up
VLAN_COUNT=1500
BASE_DEV=b1
echo Creating vlans
ip netns exec test time -p bash -c "for i in \$(seq 1 $VLAN_COUNT);
do ip -n test link add link $BASE_DEV name foo.\$i type vlan id \$i; done"
echo Bringing them up
ip netns exec test time -p bash -c "for i in \$(seq 1 $VLAN_COUNT);
do ip -n test link set foo.\$i up; done"
echo Assiging IPv6 Addresses
ip netns exec test time -p bash -c "for i in \$(seq 1 $VLAN_COUNT);
do ip -n test address add dev foo.\$i 2000::\$i/64; done"
echo Attaching to VRF
ip netns exec test time -p bash -c "for i in \$(seq 1 $VLAN_COUNT);
do ip -n test link set foo.\$i master florp; done"
On an Intel(R) Xeon(R) CPU E5-2650 v3 @ 2.30GHz machine, the performance
before the patch is (truncated):
Creating vlans
real 108.35
Bringing them up
real 4.96
Assiging IPv6 Addresses
real 19.22
Attaching to VRF
real 458.84
After the patch:
Creating vlans
real 5.59
Bringing them up
real 5.07
Assiging IPv6 Addresses
real 5.64
Attaching to VRF
real 25.37
Cc: David S. Miller <davem@davemloft.net>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Lu Wei <luwei32@huawei.com>
Cc: Xiongfeng Wang <wangxiongfeng2@huawei.com>
Cc: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: Gilad Naaman <gnaaman@drivenets.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We'd like to be able to identify netns from sk_msg hooks
to accelerate local process communication form different netns.
Signed-off-by: Xu Liu <liuxu623@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20210820071712.52852-2-liuxu623@gmail.com
There is no need to synchronize the account updating, so
use the relaxed atomic to avoid some memory barrier in the
data path.
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Acked-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, when creating an ingress qdisc on an indirect device before
the driver registered for callbacks, the driver will not have a chance
to register its filter configuration callbacks.
To fix that, modify the code such that it keeps track of all the ingress
qdiscs that call flow_indr_dev_setup_offload(). When a driver calls
flow_indr_dev_register(), go through the list of tracked ingress qdiscs
and call the driver callback entry point so as to give it a chance to
register its callback.
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Eli Cohen <elic@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
rcu field is not used. Remove it.
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Eli Cohen <elic@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We'd like to be able to identify netns from sockops hooks to
accelerate local process communication form different netns.
Signed-off-by: Xu Liu <liuxu623@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Song Liu <songliubraving@fb.com>
Link: https://lore.kernel.org/bpf/20210818105820.91894-2-liuxu623@gmail.com
Currently, the declaration of fill_imix_distribution() is dependent
on CONFIG_XFRM. This is incorrect.
Move fill_imix_distribution() declaration out of #ifndef CONFIG_XFRM
block.
Signed-off-by: Nick Richardson <richardsonnick@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add gfp_t mask as an input parameter to mem_cgroup_charge_skmem(),
to give more control to the networking stack and enable it to change
memcg charging behavior. In the future, the networking stack may decide
to avoid oom-kills when fallbacks are more appropriate.
One behavior change in mem_cgroup_charge_skmem() by this patch is to
avoid force charging by default and let the caller decide when and if
force charging is needed through the presence or absence of
__GFP_NOFAIL.
Signed-off-by: Wei Wang <weiwan@google.com>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is only one caller for ops_free(), so inline it.
Separate net_drop_ns() and net_free(), so the net_free()
can be called directly.
Add free_exit_list() helper function for free net_exit_list.
====================
v2:
- v1 does not apply, rebase it.
====================
Signed-off-by: Yajun Deng <yajun.deng@linux.dev>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add seq_puts() statement for dev_mcast, make it more readable.
As also, keep vertical alignment for {dev, ptype, dev_mcast} that
under /proc/net.
Signed-off-by: Yajun Deng <yajun.deng@linux.dev>
Signed-off-by: David S. Miller <davem@davemloft.net>
Previously, sockmap for AF_UNIX protocol only supports
dgram type. This patch add unix stream type support, which
is similar to unix_dgram_proto. To support sockmap, dgram
and stream cannot share the same unix_proto anymore, because
they have different implementations, such as unhash for stream
type (which will remove closed or disconnected sockets from the map),
so rename unix_proto to unix_dgram_proto and add a new
unix_stream_proto.
Also implement stream related sockmap functions.
And add dgram key words to those dgram specific functions.
Signed-off-by: Jiang Wang <jiang.wang@bytedance.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Reviewed-by: Cong Wang <cong.wang@bytedance.com>
Acked-by: Jakub Sitnicki <jakub@cloudflare.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20210816190327.2739291-3-jiang.wang@bytedance.com
Turn BPF_PROG_RUN into a proper always inlined function. No functional and
performance changes are intended, but it makes it much easier to understand
what's going on with how BPF programs are actually get executed. It's more
obvious what types and callbacks are expected. Also extra () around input
parameters can be dropped, as well as `__` variable prefixes intended to avoid
naming collisions, which makes the code simpler to read and write.
This refactoring also highlighted one extra issue. BPF_PROG_RUN is both
a macro and an enum value (BPF_PROG_RUN == BPF_PROG_TEST_RUN). Turning
BPF_PROG_RUN into a function causes naming conflict compilation error. So
rename BPF_PROG_RUN into lower-case bpf_prog_run(), similar to
bpf_prog_run_xdp(), bpf_prog_run_pin_on_cpu(), etc. All existing callers of
BPF_PROG_RUN, the macro, are switched to bpf_prog_run() explicitly.
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Yonghong Song <yhs@fb.com>
Link: https://lore.kernel.org/bpf/20210815070609.987780-2-andrii@kernel.org
The { 0 } doesn't clear all fields in the struct, but tells to the
compiler to set all fields to zero and doesn't touch any sub-fields
if they exists.
The {} is an empty initialiser that instructs to fully initialize whole
struct including sub-fields, which is error-prone for future
devlink_flash_notify extensions.
Fixes: 6700acc5f1 ("devlink: collect flash notify params into a struct")
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We can use xarray instead of linearly organized linked lists for the
devlink instances. This will let us revise the locking scheme in favour
of internal xarray locking that protects database.
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The struct devlink itself is protected by internal lock and doesn't
need global lock during operation. That global lock is used to protect
addition/removal new devlink instances from the global list in use by
all devlink consumers in the system.
The future conversion of linked list to be xarray will allow us to
actually delete that lock, but first we need to count all struct devlink
users.
The reference counting provides us a way to ensure that no new user
space commands success to grab devlink instance which is going to be
destroyed makes it is safe to access it without lock.
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Devlink objects are accessible only after they were registered and
have valid devlink_*->devlink pointers.
Remove that check and simplify respective fill functions as an outcome
of such change.
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The devlink_pernet_pre_exit() will be called if net namespace exits.
That routine is relevant for devlink instances that were assigned to
that namespaces first. This assignment is possible only with the following
command: "devlink reload DEV netns ...", which already checks reload support.
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The bps for imix mode is calculated by:
sum(imix_entry.size) / time_elapsed
The actual counts of each imix_entry are displayed under the
"Current:" section of the interface output in the following format:
imix_size_counts: size_1,count_1 size_2,count_2 ... size_n,count_n
Example (count = 200000):
imix_weights: 256,1 859,3 205,2
imix_size_counts: 256,32082 859,99796 205,68122
Result: OK: 17992362(c17964678+d27684) usec, 200000 (859byte,0frags)
11115pps 47Mb/sec (47977140bps) errors: 0
Summary of changes:
Calculate bps based on imix counters when in IMIX mode.
Add output for IMIX counters.
Signed-off-by: Nick Richardson <richardsonnick@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In order to represent the distribution of imix packet sizes, a
pre-computed data structure is used. It features 100 (IMIX_PRECISION)
"bins". Contiguous ranges of these bins represent the respective
packet size of each imix entry. This is done to avoid the overhead of
selecting the correct imix packet size based on the corresponding weights.
Example:
imix_weights 40,7 576,4 1500,1
total_weight = 7 + 4 + 1 = 12
pkt_size 40 occurs 7/total_weight = 58% of the time
pkt_size 576 occurs 4/total_weight = 33% of the time
pkt_size 1500 occurs 1/total_weight = 9% of the time
We generate a random number between 0-100 and select the corresponding
packet size based on the specified weights.
Eg. random number = 358723895 % 100 = 65
Selects the packet size corresponding to index:65 in the pre-computed
imix_distribution array.
An example of the pre-computed array is below:
The imix_distribution will look like the following:
0 -> 0 (index of imix_entry.size == 40)
1 -> 0 (index of imix_entry.size == 40)
2 -> 0 (index of imix_entry.size == 40)
[...] -> 0 (index of imix_entry.size == 40)
57 -> 0 (index of imix_entry.size == 40)
58 -> 1 (index of imix_entry.size == 576)
[...] -> 1 (index of imix_entry.size == 576)
90 -> 1 (index of imix_entry.size == 576)
91 -> 2 (index of imix_entry.size == 1500)
[...] -> 2 (index of imix_entry.size == 1500)
99 -> 2 (index of imix_entry.size == 1500)
Create and use "bin" representation of the imix distribution.
Signed-off-by: Nick Richardson <richardsonnick@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Adds "imix_weights" command for specifying internet mix distribution.
The command is in this format:
"imix_weights size_1,weight_1 size_2,weight_2 ... size_n,weight_n"
where the probability that packet size_i is picked is:
weight_i / (weight_1 + weight_2 + .. + weight_n)
The user may provide up to 100 imix entries (size_i,weight_i) in this
command.
The user specified imix entries will be displayed in the "Params"
section of the interface output.
Values for clone_skb > 0 is not supported in IMIX mode.
Summary of changes:
Add flag for enabling internet mix mode.
Add command (imix_weights) for internet mix input.
Return -ENOTSUPP when clone_skb > 0 in IMIX mode.
Display imix_weights in Params.
Create data structures to store imix entries and distribution.
Signed-off-by: Nick Richardson <richardsonnick@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
After migrating my laptop from 4.19-LTS to 5.4-LTS a while ago I noticed
that my Ethernet port to which a bond and a VLAN interface are attached
appeared to remain up after resuming from suspend with the cable unplugged
(and that problem still persists with 5.10-LTS).
It happens that the following happens:
- the network driver (e1000e here) prepares to suspend, calls e1000e_down()
which calls netif_carrier_off() to signal that the link is going down.
- netif_carrier_off() adds a link_watch event to the list of events for
this device
- the device is completely stopped.
- the machine suspends
- the cable is unplugged and the machine brought to another location
- the machine is resumed
- the queued linkwatch events are processed for the device
- the device doesn't yet have the __LINK_STATE_PRESENT bit and its events
are silently dropped
- the device is resumed with its link down
- the upper VLAN and bond interfaces are never notified that the link had
been turned down and remain up
- the only way to provoke a change is to physically connect the machine
to a port and possibly unplug it.
The state after resume looks like this:
$ ip -br li | egrep 'bond|eth'
bond0 UP e8:6a:64:64:64:64 <BROADCAST,MULTICAST,MASTER,UP,LOWER_UP>
eth0 DOWN e8:6a:64:64:64:64 <NO-CARRIER,BROADCAST,MULTICAST,SLAVE,UP>
eth0.2@eth0 UP e8:6a:64:64:64:64 <BROADCAST,MULTICAST,SLAVE,UP,LOWER_UP>
Placing an explicit call to netdev_state_change() either in the suspend
or the resume code in the NIC driver worked around this but the solution
is not satisfying.
The issue in fact really is in link_watch that loses events while it
ought not to. It happens that the test for the device being present was
added by commit 124eee3f69 ("net: linkwatch: add check for netdevice
being present to linkwatch_do_dev") in 4.20 to avoid an access to
devices that are not present.
Instead of dropping events, this patch proceeds slightly differently by
postponing their handling so that they happen after the device is fully
resumed.
Fixes: 124eee3f69 ("net: linkwatch: add check for netdevice being present to linkwatch_do_dev")
Link: https://lists.openwall.net/netdev/2018/03/15/62
Cc: Heiner Kallweit <hkallweit1@gmail.com>
Cc: Geert Uytterhoeven <geert+renesas@glider.be>
Cc: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Willy Tarreau <w@1wt.eu>
Link: https://lore.kernel.org/r/20210809160628.22623-1-w@1wt.eu
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Enable drivers to publish/unpublish individual parameter.
Signed-off-by: Parav Pandit <parav@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently device configuration parameters can be registered as an array.
Due to this a constant array must be registered. A single driver
supporting multiple devices each with different device capabilities end
up registering all parameters even if it doesn't support it.
One possible workaround a driver can do is, it registers multiple single
entry arrays to overcome such limitation.
Better is to provide a API that enables driver to register/unregister a
single parameter. This also further helps in two ways.
(1) to reduce the memory of devlink_param_entry by avoiding in registering
parameters which are not supported by the device.
(2) avoid generating multiple parameter add, delete, publish, unpublish,
init value notifications for such unsupported parameters
Signed-off-by: Parav Pandit <parav@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Create and use a helper function for one parameter registration.
Subsequent patch also will reuse this for driver facing routine to
register a single parameter.
Signed-off-by: Parav Pandit <parav@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add new device generic parameter to enable/disable creation of
VDPA net auxiliary device and associated device functionality
in the devlink instance.
User who prefers to disable such functionality can disable it using below
example.
$ devlink dev param set pci/0000:06:00.0 \
name enable_vnet value false cmode driverinit
$ devlink dev reload pci/0000:06:00.0
At this point devlink instance do not create auxiliary device for the
VDPA net functionality.
Signed-off-by: Parav Pandit <parav@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add new device generic parameter to enable/disable creation of
RDMA auxiliary device and associated device functionality
in the devlink instance.
User who prefers to disable such functionality can disable it using below
example.
$ devlink dev param set pci/0000:06:00.0 \
name enable_rdma value false cmode driverinit
$ devlink dev reload pci/0000:06:00.0
At this point devlink instance do not create auxiliary device for the
RDMA functionality.
Signed-off-by: Parav Pandit <parav@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add new device generic parameter to enable/disable creation of
Ethernet auxiliary device and associated device functionality
in the devlink instance.
User who prefers to disable such functionality can disable it using below
example.
$ devlink dev param set pci/0000:06:00.0 \
name enable_eth value false cmode driverinit
$ devlink dev reload pci/0000:06:00.0
At this point devlink instance do not create auxiliary device for the
Ethernet functionality.
Signed-off-by: Parav Pandit <parav@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently there's support for filtering neighbours/links for interfaces
which have a specific master device (using the IFLA_MASTER/NDA_MASTER
attributes).
This patch adds support for filtering interfaces/neighbours dump for
interfaces that *don't* have a master.
Signed-off-by: Lahav Schlesinger <lschlesinger@drivenets.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20210810090658.2778960-1-lschlesinger@drivenets.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Daniel Borkmann says:
====================
bpf-next 2021-08-10
We've added 31 non-merge commits during the last 8 day(s) which contain
a total of 28 files changed, 3644 insertions(+), 519 deletions(-).
1) Native XDP support for bonding driver & related BPF selftests, from Jussi Maki.
2) Large batch of new BPF JIT tests for test_bpf.ko that came out as a result from
32-bit MIPS JIT development, from Johan Almbladh.
3) Rewrite of netcnt BPF selftest and merge into test_progs, from Stanislav Fomichev.
4) Fix XDP bpf_prog_test_run infra after net to net-next merge, from Andrii Nakryiko.
5) Follow-up fix in unix_bpf_update_proto() to enforce socket type, from Cong Wang.
6) Fix bpf-iter-tcp4 selftest to print the correct dest IP, from Jose Blanquicet.
7) Various misc BPF XDP sample improvements, from Niklas Söderlund, Matthew Cover,
and Muhammad Falak R Wani.
* https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next: (31 commits)
bpf, tests: Add tail call test suite
bpf, tests: Add tests for BPF_CMPXCHG
bpf, tests: Add tests for atomic operations
bpf, tests: Add test for 32-bit context pointer argument passing
bpf, tests: Add branch conversion JIT test
bpf, tests: Add word-order tests for load/store of double words
bpf, tests: Add tests for ALU operations implemented with function calls
bpf, tests: Add more ALU64 BPF_MUL tests
bpf, tests: Add more BPF_LSH/RSH/ARSH tests for ALU64
bpf, tests: Add more ALU32 tests for BPF_LSH/RSH/ARSH
bpf, tests: Add more tests of ALU32 and ALU64 bitwise operations
bpf, tests: Fix typos in test case descriptions
bpf, tests: Add BPF_MOV tests for zero and sign extension
bpf, tests: Add BPF_JMP32 test cases
samples, bpf: Add an explict comment to handle nested vlan tagging.
selftests/bpf: Add tests for XDP bonding
selftests/bpf: Fix xdp_tx.c prog section name
net, core: Allow netdev_lower_get_next_private_rcu in bh context
bpf, devmap: Exclude XDP broadcast to master device
net, bonding: Add XDP support to the bonding driver
...
====================
Link: https://lore.kernel.org/r/20210810130038.16927-1-daniel@iogearbox.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Currently page pool only support page recycling when there
is only one user of the page, and the split page reusing
implemented in the most driver can not use the page pool as
bing-pong way of reusing requires the multi user support in
page pool.
Those reusing or recycling has below limitations:
1. page from page pool can only be used be one user in order
for the page recycling to happen.
2. Bing-pong way of reusing in most driver does not support
multi desc using different part of the same page in order
to save memory.
So add multi-users support and frag page recycling in page
pool to overcome the above limitation.
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
For 32 bit systems with 64 bit dma, dma_addr[1] is used to
store the upper 32 bit dma addr, those system should be rare
those days.
For normal system, the dma_addr[1] in 'struct page' is not
used, so we can reuse dma_addr[1] for storing frag count,
which means how many frags this page might be splited to.
In order to simplify the page frag support in the page pool,
the PAGE_POOL_DMA_USE_PP_FRAG_COUNT macro is added to indicate
the 32 bit systems with 64 bit dma, and the page frag support
in page pool is disabled for such system.
The newly added page_pool_set_frag_count() is called to reserve
the maximum frag count before any page frag is passed to the
user. The page_pool_atomic_sub_frag_count_return() is called
when user is done with the page frag.
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Currently, page->pp is cleared and set everytime the page
is recycled, which is unnecessary.
So only set the page->pp when the page is added to the page
pool and only clear it when the page is released from the
page pool.
This is also a preparation to support allocating frag page
in page pool.
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
