IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
If a message sent to a PF_KEY socket ended with one of the extensions
that takes a 'struct sadb_address' but there were not enough bytes
remaining in the message for the ->sa_family member of the 'struct
sockaddr' which is supposed to follow, then verify_address_len() read
past the end of the message, into uninitialized memory. Fix it by
returning -EINVAL in this case.
This bug was found using syzkaller with KMSAN.
Reproducer:
#include <linux/pfkeyv2.h>
#include <sys/socket.h>
#include <unistd.h>
int main()
{
int sock = socket(PF_KEY, SOCK_RAW, PF_KEY_V2);
char buf[24] = { 0 };
struct sadb_msg *msg = (void *)buf;
struct sadb_address *addr = (void *)(msg + 1);
msg->sadb_msg_version = PF_KEY_V2;
msg->sadb_msg_type = SADB_DELETE;
msg->sadb_msg_len = 3;
addr->sadb_address_len = 1;
addr->sadb_address_exttype = SADB_EXT_ADDRESS_SRC;
write(sock, buf, 24);
}
Reported-by: Alexander Potapenko <glider@google.com>
Cc: stable@vger.kernel.org
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
syzkaller triggered following KASAN splat:
BUG: KASAN: slab-out-of-bounds in xfrm_hash_rebuild+0xdbe/0xf00 net/xfrm/xfrm_policy.c:618
read of size 2 at addr ffff8801c8e92fe4 by task kworker/1:1/23 [..]
Workqueue: events xfrm_hash_rebuild [..]
__asan_report_load2_noabort+0x14/0x20 mm/kasan/report.c:428
xfrm_hash_rebuild+0xdbe/0xf00 net/xfrm/xfrm_policy.c:618
process_one_work+0xbbf/0x1b10 kernel/workqueue.c:2112
worker_thread+0x223/0x1990 kernel/workqueue.c:2246 [..]
The reproducer triggers:
1016 if (error) {
1017 list_move_tail(&walk->walk.all, &x->all);
1018 goto out;
1019 }
in xfrm_policy_walk() via pfkey (it sets tiny rcv space, dump
callback returns -ENOBUFS).
In this case, *walk is located the pfkey socket struct, so this socket
becomes visible in the global policy list.
It looks like this is intentional -- phony walker has walk.dead set to 1
and all other places skip such "policies".
Ccing original authors of the two commits that seem to expose this
issue (first patch missed ->dead check, second patch adds pfkey
sockets to policies dumper list).
Fixes: 880a6fab8f ("xfrm: configure policy hash table thresholds by netlink")
Fixes: 12a169e7d8 ("ipsec: Put dumpers on the dump list")
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Timo Teras <timo.teras@iki.fi>
Cc: Christophe Gouault <christophe.gouault@6wind.com>
Reported-by: syzbot <bot+c028095236fcb6f4348811565b75084c754dc729@syzkaller.appspotmail.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Currently we allow state updates to competely replace the contents
of x->encap. This is bad because on the user side ESP only sets up
header lengths depending on encap_type once when the state is first
created. This could result in the header lengths getting out of
sync with the actual state configuration.
In practice key managers will never do a state update to change the
encapsulation type. Only the port numbers need to be changed as the
peer NAT entry is updated.
Therefore this patch adds a check in xfrm_state_update to forbid
any changes to the encap_type.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
strparser wants to check socket ownership without producing any
warnings. As indicated by the comment in the code, it is permissible
for owned_by_user to return true.
Fixes: 43a0c6751a ("strparser: Stream parser for messages")
Reported-by: syzbot <syzkaller@googlegroups.com>
Reported-and-tested-by: <syzbot+c91c53af67f9ebe599a337d2e70950366153b295@syzkaller.appspotmail.com>
Signed-off-by: Tom Herbert <tom@quantonium.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
skb_copy_ubufs must unclone before it is safe to modify its
skb_shared_info with skb_zcopy_clear.
Commit b90ddd5687 ("skbuff: skb_copy_ubufs must release uarg even
without user frags") ensures that all skbs release their zerocopy
state, even those without frags.
But I forgot an edge case where such an skb arrives that is cloned.
The stack does not build such packets. Vhost/tun skbs have their
frags orphaned before cloning. TCP skbs only attach zerocopy state
when a frag is added.
But if TCP packets can be trimmed or linearized, this might occur.
Tracing the code I found no instance so far (e.g., skb_linearize
ends up calling skb_zcopy_clear if !skb->data_len).
Still, it is non-obvious that no path exists. And it is fragile to
rely on this.
Fixes: b90ddd5687 ("skbuff: skb_copy_ubufs must release uarg even without user frags")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In commit 42b531de17 ("tipc: Fix missing connection request
handling"), we replaced unconditional wakeup() with condtional
wakeup for clients with flags POLLIN | POLLRDNORM | POLLRDBAND.
This breaks the applications which do a connect followed by poll
with POLLOUT flag. These applications are not woken when the
connection is ESTABLISHED and hence sleep forever.
In this commit, we fix it by including the POLLOUT event for
sockets in TIPC_CONNECTING state.
Fixes: 42b531de17 ("tipc: Fix missing connection request handling")
Acked-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Parthasarathy Bhuvaragan <parthasarathy.bhuvaragan@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The patch(180d8cd942) replaces all uses of struct sock fields'
memory_pressure, memory_allocated, sockets_allocated, and sysctl_mem
to accessor macros. But the sockets_allocated field of sctp sock is
not replaced at all. Then replace it now for unifying the code.
Fixes: 180d8cd942 ("foundations of per-cgroup memory pressure controlling.")
Cc: Glauber Costa <glommer@parallels.com>
Signed-off-by: Tonghao Zhang <zhangtonghao@didichuxing.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Steffen Klassert says:
====================
pull request (net): ipsec 2017-12-22
1) Check for valid id proto in validate_tmpl(), otherwise
we may trigger a warning in xfrm_state_fini().
From Cong Wang.
2) Fix a typo on XFRMA_OUTPUT_MARK policy attribute.
From Michal Kubecek.
3) Verify the state is valid when encap_type < 0,
otherwise we may crash on IPsec GRO .
From Aviv Heller.
4) Fix stack-out-of-bounds read on socket policy lookup.
We access the flowi of the wrong address family in the
IPv4 mapped IPv6 case, fix this by catching address
family missmatches before we do the lookup.
5) fix xfrm_do_migrate() with AEAD to copy the geniv
field too. Otherwise the state is not fully initialized
and migration fails. From Antony Antony.
6) Fix stack-out-of-bounds with misconfigured transport
mode policies. Our policy template validation is not
strict enough. It is possible to configure policies
with transport mode template where the address family
of the template does not match the selectors address
family. Fix this by refusing such a configuration,
address family can not change on transport mode.
7) Fix a policy reference leak when reusing pcpu xdst
entry. From Florian Westphal.
8) Reinject transport-mode packets through tasklet,
otherwise it is possible to reate a recursion
loop. From Herbert Xu.
Please pull or let me know if there are problems.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix memory leak in tipc_enable_bearer() if enable_media() fails, and
cleanup with bearer_disable() if tipc_mon_create() fails.
Acked-by: Ying Xue <ying.xue@windriver.com>
Acked-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Tommi Rantala <tommi.t.rantala@nokia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
RDS currently doesn't check if the length of the control message is
large enough to hold the required data, before dereferencing the control
message data. This results in following crash:
BUG: KASAN: stack-out-of-bounds in rds_rdma_bytes net/rds/send.c:1013
[inline]
BUG: KASAN: stack-out-of-bounds in rds_sendmsg+0x1f02/0x1f90
net/rds/send.c:1066
Read of size 8 at addr ffff8801c928fb70 by task syzkaller455006/3157
CPU: 0 PID: 3157 Comm: syzkaller455006 Not tainted 4.15.0-rc3+ #161
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS
Google 01/01/2011
Call Trace:
__dump_stack lib/dump_stack.c:17 [inline]
dump_stack+0x194/0x257 lib/dump_stack.c:53
print_address_description+0x73/0x250 mm/kasan/report.c:252
kasan_report_error mm/kasan/report.c:351 [inline]
kasan_report+0x25b/0x340 mm/kasan/report.c:409
__asan_report_load8_noabort+0x14/0x20 mm/kasan/report.c:430
rds_rdma_bytes net/rds/send.c:1013 [inline]
rds_sendmsg+0x1f02/0x1f90 net/rds/send.c:1066
sock_sendmsg_nosec net/socket.c:628 [inline]
sock_sendmsg+0xca/0x110 net/socket.c:638
___sys_sendmsg+0x320/0x8b0 net/socket.c:2018
__sys_sendmmsg+0x1ee/0x620 net/socket.c:2108
SYSC_sendmmsg net/socket.c:2139 [inline]
SyS_sendmmsg+0x35/0x60 net/socket.c:2134
entry_SYSCALL_64_fastpath+0x1f/0x96
RIP: 0033:0x43fe49
RSP: 002b:00007fffbe244ad8 EFLAGS: 00000217 ORIG_RAX: 0000000000000133
RAX: ffffffffffffffda RBX: 00000000004002c8 RCX: 000000000043fe49
RDX: 0000000000000001 RSI: 000000002020c000 RDI: 0000000000000003
RBP: 00000000006ca018 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000217 R12: 00000000004017b0
R13: 0000000000401840 R14: 0000000000000000 R15: 0000000000000000
To fix this, we verify that the cmsg_len is large enough to hold the
data to be read, before proceeding further.
Reported-by: syzbot <syzkaller-bugs@googlegroups.com>
Signed-off-by: Avinash Repaka <avinash.repaka@oracle.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When a group member receives a member WITHDRAW event, this might have
two reasons: either the peer member is leaving the group, or the link
to the member's node has been lost.
In the latter case we need to issue a DOWN event to the user right away,
and let function tipc_group_filter_msg() perform delete of the member
item. However, in this case we miss to change the state of the member
item to MBR_LEAVING, so the member item is not deleted, and we have a
memory leak.
We now separate better between the four sub-cases of a WITHRAW event
and make sure that each case is handled correctly.
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We need to check block for being null in both tcf_block_put and
tcf_block_put_ext.
Fixes: 343723dd51 ("net: sched: fix clsact init error path")
Reported-by: Prashant Bhole <bhole_prashant_q7@lab.ntt.co.jp>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In commit 2f487712b8 ("tipc: guarantee that group broadcast doesn't
bypass group unicast") we introduced a mechanism that requires the first
(replicated) broadcast sent after a unicast to be acknowledged by all
receivers before permitting sending of the next (true) broadcast.
The counter for keeping track of the number of acknowledges to expect
is based on the tipc_group::member_cnt variable. But this misses that
some of the known members may not be ready for reception, and will never
acknowledge the message, either because they haven't fully joined the
group or because they are leaving the group. Such members are identified
by not fulfilling the condition tested for in the function
tipc_group_is_enabled().
We now set the counter for the actual number of acks to receive at the
moment the message is sent, by just counting the number of recipients
satisfying the tipc_group_is_enabled() test.
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The rcu_barrier_bh() in mini_qdisc_pair_swap() is to wait for
flying RCU callback installed by a previous mini_qdisc_pair_swap(),
however we miss it on the tp_head==NULL path, which leads to that
the RCU callback still uses miniq_old->rcu after it is freed together
with qdisc in qdisc_graft(). So just add it on that path too.
Fixes: 46209401f8 ("net: core: introduce mini_Qdisc and eliminate usage of tp->q for clsact fastpath ")
Reported-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Tested-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Cc: Jiri Pirko <jiri@mellanox.com>
Cc: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When ip6gre is created using ioctl, its features, such as
scatter-gather, GSO and tx-checksumming will be turned off:
# ip -f inet6 tunnel add gre6 mode ip6gre remote fd00::1
# ethtool -k gre6 (truncated output)
tx-checksumming: off
scatter-gather: off
tcp-segmentation-offload: off
generic-segmentation-offload: off [requested on]
But when netlink is used, they will be enabled:
# ip link add gre6 type ip6gre remote fd00::1
# ethtool -k gre6 (truncated output)
tx-checksumming: on
scatter-gather: on
tcp-segmentation-offload: on
generic-segmentation-offload: on
This results in a loss of performance when gre6 is created via ioctl.
The issue was found with LTP/gre tests.
Fix it by moving the setup of device features to a separate function
and invoke it with ndo_init callback because both netlink and ioctl
will eventually call it via register_netdevice():
register_netdevice()
- ndo_init() callback -> ip6gre_tunnel_init() or ip6gre_tap_init()
- ip6gre_tunnel_init_common()
- ip6gre_tnl_init_features()
The moved code also contains two minor style fixes:
* removed needless tab from GRE6_FEATURES on NETIF_F_HIGHDMA line.
* fixed the issue reported by checkpatch: "Unnecessary parentheses around
'nt->encap.type == TUNNEL_ENCAP_NONE'"
Fixes: ac4eb009e4 ("ip6gre: Add support for basic offloads offloads excluding GSO")
Signed-off-by: Alexey Kodanev <alexey.kodanev@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
skb_copy_ubufs creates a private copy of frags[] to release its hold
on user frags, then calls uarg->callback to notify the owner.
Call uarg->callback even when no frags exist. This edge case can
happen when zerocopy_sg_from_iter finds enough room in skb_headlen
to copy all the data.
Fixes: 3ece782693 ("sock: skb_copy_ubufs support for compound pages")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Call skb_zerocopy_clone after skb_orphan_frags, to avoid duplicate
calls to skb_uarg(skb)->callback for the same data.
skb_zerocopy_clone associates skb_shinfo(skb)->uarg from frag_skb
with each segment. This is only safe for uargs that do refcounting,
which is those that pass skb_orphan_frags without dropping their
shared frags. For others, skb_orphan_frags drops the user frags and
sets the uarg to NULL, after which sock_zerocopy_clone has no effect.
Qemu hangs were reported due to duplicate vhost_net_zerocopy_callback
calls for the same data causing the vhost_net_ubuf_ref_>refcount to
drop below zero.
Link: http://lkml.kernel.org/r/<CAF=yD-LWyCD4Y0aJ9O0e_CHLR+3JOeKicRRTEVCPxgw4XOcqGQ@mail.gmail.com>
Fixes: 1f8b977ab3 ("sock: enable MSG_ZEROCOPY")
Reported-by: Andreas Hartmann <andihartmann@01019freenet.de>
Reported-by: David Hill <dhill@redhat.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
sysctl.ip6.auto_flowlabels is default 1. In our hosts, we set it to 2.
If sockopt doesn't set autoflowlabel, outcome packets from the hosts are
supposed to not include flowlabel. This is true for normal packet, but
not for reset packet.
The reason is ipv6_pinfo.autoflowlabel is set in sock creation. Later if
we change sysctl.ip6.auto_flowlabels, the ipv6_pinfo.autoflowlabel isn't
changed, so the sock will keep the old behavior in terms of auto
flowlabel. Reset packet is suffering from this problem, because reset
packet is sent from a special control socket, which is created at boot
time. Since sysctl.ipv6.auto_flowlabels is 1 by default, the control
socket will always have its ipv6_pinfo.autoflowlabel set, even after
user set sysctl.ipv6.auto_flowlabels to 1, so reset packset will always
have flowlabel. Normal sock created before sysctl setting suffers from
the same issue. We can't even turn off autoflowlabel unless we kill all
socks in the hosts.
To fix this, if IPV6_AUTOFLOWLABEL sockopt is used, we use the
autoflowlabel setting from user, otherwise we always call
ip6_default_np_autolabel() which has the new settings of sysctl.
Note, this changes behavior a little bit. Before commit 42240901f7
(ipv6: Implement different admin modes for automatic flow labels), the
autoflowlabel behavior of a sock isn't sticky, eg, if sysctl changes,
existing connection will change autoflowlabel behavior. After that
commit, autoflowlabel behavior is sticky in the whole life of the sock.
With this patch, the behavior isn't sticky again.
Cc: Martin KaFai Lau <kafai@fb.com>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Tom Herbert <tom@quantonium.net>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
skb_vlan_pop() expects skb->protocol to be a valid TPID for double
tagged frames. So set skb->protocol to the TPID and let skb_vlan_pop()
shift the true ethertype into position for us.
Fixes: 5108bbaddc ("openvswitch: add processing of L3 packets")
Signed-off-by: Eric Garver <e@erig.me>
Reviewed-by: Jiri Benc <jbenc@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, parameters such as oif and source address are not taken into
account during fibmatch lookup. Example (IPv4 for reference) before
patch:
$ ip -4 route show
192.0.2.0/24 dev dummy0 proto kernel scope link src 192.0.2.1
198.51.100.0/24 dev dummy1 proto kernel scope link src 198.51.100.1
$ ip -6 route show
2001:db8:1::/64 dev dummy0 proto kernel metric 256 pref medium
2001:db8:2::/64 dev dummy1 proto kernel metric 256 pref medium
fe80::/64 dev dummy0 proto kernel metric 256 pref medium
fe80::/64 dev dummy1 proto kernel metric 256 pref medium
$ ip -4 route get fibmatch 192.0.2.2 oif dummy0
192.0.2.0/24 dev dummy0 proto kernel scope link src 192.0.2.1
$ ip -4 route get fibmatch 192.0.2.2 oif dummy1
RTNETLINK answers: No route to host
$ ip -6 route get fibmatch 2001:db8:1::2 oif dummy0
2001:db8:1::/64 dev dummy0 proto kernel metric 256 pref medium
$ ip -6 route get fibmatch 2001:db8:1::2 oif dummy1
2001:db8:1::/64 dev dummy0 proto kernel metric 256 pref medium
After:
$ ip -6 route get fibmatch 2001:db8:1::2 oif dummy0
2001:db8:1::/64 dev dummy0 proto kernel metric 256 pref medium
$ ip -6 route get fibmatch 2001:db8:1::2 oif dummy1
RTNETLINK answers: Network is unreachable
The problem stems from the fact that the necessary route lookup flags
are not set based on these parameters.
Instead of duplicating the same logic for fibmatch, we can simply
resolve the original route from its copy and dump it instead.
Fixes: 18c3a61c42 ("net: ipv6: RTM_GETROUTE: return matched fib result when requested")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since commit 0ddcf43d5d ("ipv4: FIB Local/MAIN table collapse") the
local table uses the same trie allocated for the main table when custom
rules are not in use.
When a net namespace is dismantled, the main table is flushed and freed
(via an RCU callback) before the local table. In case the callback is
invoked before the local table is iterated, a use-after-free can occur.
Fix this by iterating over the FIB tables in reverse order, so that the
main table is always freed after the local table.
v3: Reworded comment according to Alex's suggestion.
v2: Add a comment to make the fix more explicit per Dave's and Alex's
feedback.
Fixes: 0ddcf43d5d ("ipv4: FIB Local/MAIN table collapse")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Acked-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When we receive a JOIN message from a peer member, the message may
contain an advertised window value ADV_IDLE that permits removing the
member in question from the tipc_group::congested list. However, since
the removal has been made conditional on that the advertised window is
*not* ADV_IDLE, we miss this case. This has the effect that a sender
sometimes may enter a state of permanent, false, broadcast congestion.
We fix this by unconditinally removing the member from the congested
list before calling tipc_member_update(), which might potentially sort
it into the list again.
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
cls_bpf used to take care of tracking what offload state a filter
is in, i.e. it would track if offload request succeeded or not.
This information would then be used to issue correct requests to
the driver, e.g. requests for statistics only on offloaded filters,
removing only filters which were offloaded, using add instead of
replace if previous filter was not added etc.
This tracking of offload state no longer functions with the new
callback infrastructure. There could be multiple entities trying
to offload the same filter.
Throw out all the tracking and corresponding commands and simply
pass to the drivers both old and new bpf program. Drivers will
have to deal with offload state tracking by themselves.
Fixes: 3f7889c4c7 ("net: sched: cls_bpf: call block callbacks for offload")
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(I can trivially verify that that idr_remove in cleanup_net happens
after the network namespace count has dropped to zero --EWB)
Function get_net_ns_by_id() does not check for net::count
after it has found a peer in netns_ids idr.
It may dereference a peer, after its count has already been
finaly decremented. This leads to double free and memory
corruption:
put_net(peer) rtnl_lock()
atomic_dec_and_test(&peer->count) [count=0] ...
__put_net(peer) get_net_ns_by_id(net, id)
spin_lock(&cleanup_list_lock)
list_add(&net->cleanup_list, &cleanup_list)
spin_unlock(&cleanup_list_lock)
queue_work() peer = idr_find(&net->netns_ids, id)
| get_net(peer) [count=1]
| ...
| (use after final put)
v ...
cleanup_net() ...
spin_lock(&cleanup_list_lock) ...
list_replace_init(&cleanup_list, ..) ...
spin_unlock(&cleanup_list_lock) ...
... ...
... put_net(peer)
... atomic_dec_and_test(&peer->count) [count=0]
... spin_lock(&cleanup_list_lock)
... list_add(&net->cleanup_list, &cleanup_list)
... spin_unlock(&cleanup_list_lock)
... queue_work()
... rtnl_unlock()
rtnl_lock() ...
for_each_net(tmp) { ...
id = __peernet2id(tmp, peer) ...
spin_lock_irq(&tmp->nsid_lock) ...
idr_remove(&tmp->netns_ids, id) ...
... ...
net_drop_ns() ...
net_free(peer) ...
} ...
|
v
cleanup_net()
...
(Second free of peer)
Also, put_net() on the right cpu may reorder with left's cpu
list_replace_init(&cleanup_list, ..), and then cleanup_list
will be corrupted.
Since cleanup_net() is executed in worker thread, while
put_net(peer) can happen everywhere, there should be
enough time for concurrent get_net_ns_by_id() to pick
the peer up, and the race does not seem to be unlikely.
The patch fixes the problem in standard way.
(Also, there is possible problem in peernet2id_alloc(), which requires
check for net::count under nsid_lock and maybe_get_net(peer), but
in current stable kernel it's used under rtnl_lock() and it has to be
safe. Openswitch begun to use peernet2id_alloc(), and possibly it should
be fixed too. While this is not in stable kernel yet, so I'll send
a separate message to netdev@ later).
Cc: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Fixes: 0c7aecd4bd "netns: add rtnl cmd to add and get peer netns ids"
Reviewed-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Reviewed-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Acked-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The recently added fib_metrics_match() causes a regression for routes
with both RTAX_FEATURES and RTAX_CC_ALGO if the latter has
TCP_CONG_NEEDS_ECN flag set:
| # ip link add d0 type dummy
| # ip link set d0 up
| # ip route add 172.29.29.0/24 dev d0 features ecn congctl dctcp
| # ip route del 172.29.29.0/24 dev d0 features ecn congctl dctcp
| RTNETLINK answers: No such process
During route insertion, fib_convert_metrics() detects that the given CC
algo requires ECN and hence sets DST_FEATURE_ECN_CA bit in
RTAX_FEATURES.
During route deletion though, fib_metrics_match() compares stored
RTAX_FEATURES value with that from userspace (which obviously has no
knowledge about DST_FEATURE_ECN_CA) and fails.
Fixes: 5f9ae3d9e7 ("ipv4: do metrics match when looking up and deleting a route")
Signed-off-by: Phil Sutter <phil@nwl.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
When, during a join operation, or during message transmission, a group
member needs to be added to the group's 'congested' list, we sort it
into the list in ascending order, according to its current advertised
window size. However, we miss the case when the member is already on
that list. This will have the result that the member, after the window
size has been decremented, might be at the wrong position in that list.
This again may have the effect that we during broadcast and multicast
transmissions miss the fact that a destination is not yet ready for
reception, and we end up sending anyway. From this point on, the
behavior during the remaining session is unpredictable, e.g., with
underflowing window sizes.
We now correct this bug by unconditionally removing the member from
the list before (re-)sorting it in.
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now it's using IPV6_MIN_MTU as the min mtu in ip6_tnl_xmit, but
IPV6_MIN_MTU actually only works when the inner packet is ipv6.
With IPV6_MIN_MTU for ipv4 packets, the new pmtu for inner dst
couldn't be set less than 1280. It would cause tx_err and the
packet to be dropped when the outer dst pmtu is close to 1280.
Jianlin found it by running ipv4 traffic with the topo:
(client) gre6 <---> eth1 (route) eth2 <---> gre6 (server)
After changing eth2 mtu to 1300, the performance became very
low, or the connection was even broken. The issue also affects
ip4ip6 and ip6ip6 tunnels.
So if the inner packet is ipv4, 576 should be considered as the
min mtu.
Note that for ip4ip6 and ip6ip6 tunnels, the inner packet can
only be ipv4 or ipv6, but for gre6 tunnel, it may also be ARP.
This patch using 576 as the min mtu for non-ipv6 packet works
for all those cases.
Reported-by: Jianlin Shi <jishi@redhat.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The same fix as the patch "ip_gre: remove the incorrect mtu limit for
ipgre tap" is also needed for ip6_gre.
Fixes: 61e84623ac ("net: centralize net_device min/max MTU checking")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
ipgre tap driver calls ether_setup(), after commit 61e84623ac
("net: centralize net_device min/max MTU checking"), the range
of mtu is [min_mtu, max_mtu], which is [68, 1500] by default.
It causes the dev mtu of the ipgre tap device to not be greater
than 1500, this limit value is not correct for ipgre tap device.
Besides, it's .change_mtu already does the right check. So this
patch is just to set max_mtu as 0, and leave the check to it's
.change_mtu.
Fixes: 61e84623ac ("net: centralize net_device min/max MTU checking")
Reported-by: Jianlin Shi <jishi@redhat.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
* hwsim:
- set To-DS bit in some frames missing it
- fix sleeping in atomic
* nl80211:
- doc cleanup
- fix locking in an error path
* build:
- don't append to created certs C files
- ship certificate pre-hexdumped
-----BEGIN PGP SIGNATURE-----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=OWs9
-----END PGP SIGNATURE-----
Merge tag 'mac80211-for-davem-2017-12-19' of git://git.kernel.org/pub/scm/linux/kernel/git/jberg/mac80211
Johannes Berg says:
====================
A few more fixes:
* hwsim:
- set To-DS bit in some frames missing it
- fix sleeping in atomic
* nl80211:
- doc cleanup
- fix locking in an error path
* build:
- don't append to created certs C files
- ship certificate pre-hexdumped
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Not only does this remove the need for the hexdump code in most
normal kernel builds (still there for the extra directory), but
it also removes the need to ship binary files, which apparently
is somewhat problematic, as Randy reported.
While at it, also add the generated files to clean-files.
Reported-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Currently the certs C code generation appends to the generated files,
which is most likely a leftover from commit 715a123347 ("wireless:
don't write C files on failures"). This causes duplicate code in the
generated files if the certificates have their timestamps modified
between builds and thereby trigger the generation rules.
Fixes: 715a123347 ("wireless: don't write C files on failures")
Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
This is an old bugbear of mine:
https://www.mail-archive.com/netdev@vger.kernel.org/msg03894.html
By crafting special packets, it is possible to cause recursion
in our kernel when processing transport-mode packets at levels
that are only limited by packet size.
The easiest one is with DNAT, but an even worse one is where
UDP encapsulation is used in which case you just have to insert
an UDP encapsulation header in between each level of recursion.
This patch avoids this problem by reinjecting tranport-mode packets
through a tasklet.
Fixes: b05e106698 ("[IPV4/6]: Netfilter IPsec input hooks")
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
The early call to br_stp_change_bridge_id in bridge's newlink can cause
a memory leak if an error occurs during the newlink because the fdb
entries are not cleaned up if a different lladdr was specified, also
another minor issue is that it generates fdb notifications with
ifindex = 0. Another unrelated memory leak is the bridge sysfs entries
which get added on NETDEV_REGISTER event, but are not cleaned up in the
newlink error path. To remove this special case the call to
br_stp_change_bridge_id is done after netdev register and we cleanup the
bridge on changelink error via br_dev_delete to plug all leaks.
This patch makes netlink bridge destruction on newlink error the same as
dellink and ioctl del which is necessary since at that point we have a
fully initialized bridge device.
To reproduce the issue:
$ ip l add br0 address 00:11:22:33:44:55 type bridge group_fwd_mask 1
RTNETLINK answers: Invalid argument
$ rmmod bridge
[ 1822.142525] =============================================================================
[ 1822.143640] BUG bridge_fdb_cache (Tainted: G O ): Objects remaining in bridge_fdb_cache on __kmem_cache_shutdown()
[ 1822.144821] -----------------------------------------------------------------------------
[ 1822.145990] Disabling lock debugging due to kernel taint
[ 1822.146732] INFO: Slab 0x0000000092a844b2 objects=32 used=2 fp=0x00000000fef011b0 flags=0x1ffff8000000100
[ 1822.147700] CPU: 2 PID: 13584 Comm: rmmod Tainted: G B O 4.15.0-rc2+ #87
[ 1822.148578] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.7.5-20140531_083030-gandalf 04/01/2014
[ 1822.150008] Call Trace:
[ 1822.150510] dump_stack+0x78/0xa9
[ 1822.151156] slab_err+0xb1/0xd3
[ 1822.151834] ? __kmalloc+0x1bb/0x1ce
[ 1822.152546] __kmem_cache_shutdown+0x151/0x28b
[ 1822.153395] shutdown_cache+0x13/0x144
[ 1822.154126] kmem_cache_destroy+0x1c0/0x1fb
[ 1822.154669] SyS_delete_module+0x194/0x244
[ 1822.155199] ? trace_hardirqs_on_thunk+0x1a/0x1c
[ 1822.155773] entry_SYSCALL_64_fastpath+0x23/0x9a
[ 1822.156343] RIP: 0033:0x7f929bd38b17
[ 1822.156859] RSP: 002b:00007ffd160e9a98 EFLAGS: 00000202 ORIG_RAX: 00000000000000b0
[ 1822.157728] RAX: ffffffffffffffda RBX: 00005578316ba090 RCX: 00007f929bd38b17
[ 1822.158422] RDX: 00007f929bd9ec60 RSI: 0000000000000800 RDI: 00005578316ba0f0
[ 1822.159114] RBP: 0000000000000003 R08: 00007f929bff5f20 R09: 00007ffd160e8a11
[ 1822.159808] R10: 00007ffd160e9860 R11: 0000000000000202 R12: 00007ffd160e8a80
[ 1822.160513] R13: 0000000000000000 R14: 0000000000000000 R15: 00005578316ba090
[ 1822.161278] INFO: Object 0x000000007645de29 @offset=0
[ 1822.161666] INFO: Object 0x00000000d5df2ab5 @offset=128
Fixes: 30313a3d57 ("bridge: Handle IFLA_ADDRESS correctly when creating bridge device")
Fixes: 5b8d5429da ("bridge: netlink: register netdevice before executing changelink")
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Whenever a new type of chunk is added, the corresp conversion in
sctp_cname should be added. Otherwise, in some places, pr_debug
will print it as "unknown chunk".
Fixes: cc16f00f65 ("sctp: add support for generating stream reconf ssn reset request chunk")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo R. Leitner <marcelo.leitner@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now when reneging events in sctp_ulpq_renege(), the variable freed
could be increased by a __u16 value twice while freed is of __u16
type. It means freed may overflow at the second addition.
This patch is to fix it by using __u32 type for 'freed', while at
it, also to remove 'if (chunk)' check, as all renege commands are
generated in sctp_eat_data and it can't be NULL.
Reported-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A group member going into state LEAVING should never go back to any
other state before it is finally deleted. However, this might happen
if the socket needs to send out a RECLAIM message during this interval.
Since we forget to remove the leaving member from the group's 'active'
or 'pending' list, the member might be selected for reclaiming, change
state to RECLAIMING, and get stuck in this state instead of being
deleted. This might lead to suppression of the expected 'member down'
event to the receiver.
We fix this by removing the member from all lists, except the RB tree,
at the moment it goes into state LEAVING.
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Group messages are not supposed to be returned to sender when the
destination socket disappears. This is done correctly for regular
traffic messages, by setting the 'dest_droppable' bit in the header.
But we forget to do that in group protocol messages. This has the effect
that such messages may sometimes bounce back to the sender, be perceived
as a legitimate peer message, and wreak general havoc for the rest of
the session. In particular, we have seen that a member in state LEAVING
may go back to state RECLAIMED or REMITTED, hence causing suppression
of an otherwise expected 'member down' event to the user.
We fix this by setting the 'dest_droppable' bit even in group protocol
messages.
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Daniel Borkmann says:
====================
pull-request: bpf 2017-12-17
The following pull-request contains BPF updates for your *net* tree.
The main changes are:
1) Fix a corner case in generic XDP where we have non-linear skbs
but enough tailroom in the skb to not miss to linearizing there,
from Song.
2) Fix BPF JIT bugs in s390x and ppc64 to not recache skb data when
BPF context is not skb, from Daniel.
3) Fix a BPF JIT bug in sparc64 where recaching skb data after helper
call would use the wrong register for the skb, from Daniel.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
One example of when an ICMPv6 packet is required to be looped back is
when a host acts as both a Multicast Listener and a Multicast Router.
A Multicast Router will listen on address ff02::16 for MLDv2 messages.
Currently, MLDv2 messages originating from a Multicast Listener running
on the same host as the Multicast Router are not being delivered to the
Multicast Router. This is due to dst.input being assigned the default
value of dst_discard.
This results in the packet being looped back but discarded before being
delivered to the Multicast Router.
This patch sets dst.input to ip6_input to ensure a looped back packet
is delivered to the Multicast Router.
Signed-off-by: Brendan McGrath <redmcg@redmandi.dyndns.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Stable bugfixes:
- NFS: Avoid a BUG_ON() in nfs_commit_inode() by not waiting for a
commit in the case that there were no commit requests.
- SUNRPC: Fix a race in the receive code path
Other fixes:
- NFS: Fix a deadlock in nfs client initialization
- xprtrdma: Fix a performance regression for small IOs
-----BEGIN PGP SIGNATURE-----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=XhEC
-----END PGP SIGNATURE-----
Merge tag 'nfs-for-4.15-3' of git://git.linux-nfs.org/projects/anna/linux-nfs
Pull NFS client fixes from Anna Schumaker:
"This has two stable bugfixes, one to fix a BUG_ON() when
nfs_commit_inode() is called with no outstanding commit requests and
another to fix a race in the SUNRPC receive codepath.
Additionally, there are also fixes for an NFS client deadlock and an
xprtrdma performance regression.
Summary:
Stable bugfixes:
- NFS: Avoid a BUG_ON() in nfs_commit_inode() by not waiting for a
commit in the case that there were no commit requests.
- SUNRPC: Fix a race in the receive code path
Other fixes:
- NFS: Fix a deadlock in nfs client initialization
- xprtrdma: Fix a performance regression for small IOs"
* tag 'nfs-for-4.15-3' of git://git.linux-nfs.org/projects/anna/linux-nfs:
SUNRPC: Fix a race in the receive code path
nfs: don't wait on commit in nfs_commit_inode() if there were no commit requests
xprtrdma: Spread reply processing over more CPUs
nfs: fix a deadlock in nfs client initialization
Pull networking fixes from David Miller:
1) Clamp timeouts to INT_MAX in conntrack, from Jay Elliot.
2) Fix broken UAPI for BPF_PROG_TYPE_PERF_EVENT, from Hendrik
Brueckner.
3) Fix locking in ieee80211_sta_tear_down_BA_sessions, from Johannes
Berg.
4) Add missing barriers to ptr_ring, from Michael S. Tsirkin.
5) Don't advertise gigabit in sh_eth when not available, from Thomas
Petazzoni.
6) Check network namespace when delivering to netlink taps, from Kevin
Cernekee.
7) Kill a race in raw_sendmsg(), from Mohamed Ghannam.
8) Use correct address in TCP md5 lookups when replying to an incoming
segment, from Christoph Paasch.
9) Add schedule points to BPF map alloc/free, from Eric Dumazet.
10) Don't allow silly mtu values to be used in ipv4/ipv6 multicast, also
from Eric Dumazet.
11) Fix SKB leak in tipc, from Jon Maloy.
12) Disable MAC learning on OVS ports of mlxsw, from Yuval Mintz.
13) SKB leak fix in skB_complete_tx_timestamp(), from Willem de Bruijn.
14) Add some new qmi_wwan device IDs, from Daniele Palmas.
15) Fix static key imbalance in ingress qdisc, from Jiri Pirko.
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (76 commits)
net: qcom/emac: Reduce timeout for mdio read/write
net: sched: fix static key imbalance in case of ingress/clsact_init error
net: sched: fix clsact init error path
ip_gre: fix wrong return value of erspan_rcv
net: usb: qmi_wwan: add Telit ME910 PID 0x1101 support
pkt_sched: Remove TC_RED_OFFLOADED from uapi
net: sched: Move to new offload indication in RED
net: sched: Add TCA_HW_OFFLOAD
net: aquantia: Increment driver version
net: aquantia: Fix typo in ethtool statistics names
net: aquantia: Update hw counters on hw init
net: aquantia: Improve link state and statistics check interval callback
net: aquantia: Fill in multicast counter in ndev stats from hardware
net: aquantia: Fill ndev stat couters from hardware
net: aquantia: Extend stat counters to 64bit values
net: aquantia: Fix hardware DMA stream overload on large MRRS
net: aquantia: Fix actual speed capabilities reporting
sock: free skb in skb_complete_tx_timestamp on error
s390/qeth: update takeover IPs after configuration change
s390/qeth: lock IP table while applying takeover changes
...
Move static key increments to the beginning of the init function
so they pair 1:1 with decrements in ingress/clsact_destroy,
which is called in case ingress/clsact_init fails.
Fixes: 6529eaba33 ("net: sched: introduce tcf block infractructure")
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Acked-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since in qdisc_create, the destroy op is called when init fails, we
don't do cleanup in init and leave it up to destroy.
This fixes use-after-free when trying to put already freed block.
Fixes: 6e40cf2d4d ("net: sched: use extended variants of block_get/put in ingress and clsact qdiscs")
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Acked-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We must ensure that the call to rpc_sleep_on() in xprt_transmit() cannot
race with the call to xprt_complete_rqst().
Reported-by: Chuck Lever <chuck.lever@oracle.com>
Link: https://bugzilla.linux-nfs.org/show_bug.cgi?id=317
Fixes: ce7c252a8c ("SUNRPC: Add a separate spinlock to protect..")
Cc: stable@vger.kernel.org # 4.14+
Reviewed-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Commit d8f532d20e ("xprtrdma: Invoke rpcrdma_reply_handler
directly from RECV completion") introduced a performance regression
for NFS I/O small enough to not need memory registration. In multi-
threaded benchmarks that generate primarily small I/O requests,
IOPS throughput is reduced by nearly a third. This patch restores
the previous level of throughput.
Because workqueues are typically BOUND (in particular ib_comp_wq,
nfsiod_workqueue, and rpciod_workqueue), NFS/RDMA workloads tend
to aggregate on the CPU that is handling Receive completions.
The usual approach to addressing this problem is to create a QP
and CQ for each CPU, and then schedule transactions on the QP
for the CPU where you want the transaction to complete. The
transaction then does not require an extra context switch during
completion to end up on the same CPU where the transaction was
started.
This approach doesn't work for the Linux NFS/RDMA client because
currently the Linux NFS client does not support multiple connections
per client-server pair, and the RDMA core API does not make it
straightforward for ULPs to determine which CPU is responsible for
handling Receive completions for a CQ.
So for the moment, record the CPU number in the rpcrdma_req before
the transport sends each RPC Call. Then during Receive completion,
queue the RPC completion on that same CPU.
Additionally, move all RPC completion processing to the deferred
handler so that even RPCs with simple small replies complete on
the CPU that sent the corresponding RPC Call.
Fixes: d8f532d20e ("xprtrdma: Invoke rpcrdma_reply_handler ...")
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
If pskb_may_pull return failed, return PACKET_REJECT instead of -ENOMEM.
Fixes: 84e54fe0a5 ("gre: introduce native tunnel support for ERSPAN")
Cc: William Tu <u9012063@gmail.com>
Signed-off-by: Haishuang Yan <yanhaishuang@cmss.chinamobile.com>
Acked-by: William Tu <u9012063@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Let RED utilize the new internal flag, TCQ_F_OFFLOADED,
to mark a given qdisc as offloaded instead of using a dedicated
indication.
Also, change internal logic into looking at said flag when possible.
Fixes: 602f3baf22 ("net_sch: red: Add offload ability to RED qdisc")
Signed-off-by: Yuval Mintz <yuvalm@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Qdiscs can be offloaded to HW, but current implementation isn't uniform.
Instead, qdiscs either pass information about offload status via their
TCA_OPTIONS or omit it altogether.
Introduce a new attribute - TCA_HW_OFFLOAD that would form a uniform
uAPI for the offloading status of qdiscs.
Signed-off-by: Yuval Mintz <yuvalm@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
