netfilter: ipv6: Don't preserve original oif for loopback address

Commit 508b09046c0f ("netfilter: ipv6: Preserve link scope traffic
original oif") made ip6_route_me_harder() keep the original oif for
link-local and multicast packets. However, it also affected packets
for the loopback address because it used rt6_need_strict().

REDIRECT rules in the OUTPUT chain rewrite the destination to loopback
address; thus its oif should not be preserved. This commit fixes the bug
that redirected local packets are being dropped. Actually the packet was
not exactly dropped; Instead it was sent out to the original oif rather
than lo. When a packet with daddr ::1 is sent to the router, it is
effectively dropped.

Fixes: 508b09046c0f ("netfilter: ipv6: Preserve link scope traffic original oif")
Signed-off-by: Eli Cooper <elicooper@gmx.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

netfilter: nf_tables: unbind set in rule from commit path

Anonymous sets that are bound to rules from the same transaction trigger
a kernel splat from the abort path due to double set list removal and
double free.

This patch updates the logic to search for the transaction that is
responsible for creating the set and disable the set list removal and
release, given the rule is now responsible for this. Lookup is reverse
since the transaction that adds the set is likely to be at the tail of
the list.

Moreover, this patch adds the unbind step to deliver the event from the
commit path.  This should not be done from the worker thread, since we
have no guarantees of in-order delivery to the listener.

This patch removes the assumption that both activate and deactivate
callbacks need to be provided.

Fixes: cd5125d8f518 ("netfilter: nf_tables: split set destruction in deactivate and destroy phase")
Reported-by: Mikhail Morfikov <mmorfikov@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

netfilter: nf_nat: skip nat clash resolution for same-origin entries

It is possible that two concurrent packets originating from the same
socket of a connection-less protocol (e.g. UDP) can end up having
different IP_CT_DIR_REPLY tuples which results in one of the packets
being dropped.

To illustrate this, consider the following simplified scenario:

1. Packet A and B are sent at the same time from two different threads
   by same UDP socket.  No matching conntrack entry exists yet.
   Both packets cause allocation of a new conntrack entry.
2. get_unique_tuple gets called for A.  No clashing entry found.
   conntrack entry for A is added to main conntrack table.
3. get_unique_tuple is called for B and will find that the reply
   tuple of B is already taken by A.
   It will allocate a new UDP source port for B to resolve the clash.
4. conntrack entry for B cannot be added to main conntrack table
   because its ORIGINAL direction is clashing with A and the REPLY
   directions of A and B are not the same anymore due to UDP source
   port reallocation done in step 3.

This patch modifies nf_conntrack_tuple_taken so it doesn't consider
colliding reply tuples if the IP_CT_DIR_ORIGINAL tuples are equal.

[ Florian: simplify patch to not use .allow_clash setting
  and always ignore identical flows ]

Signed-off-by: Martynas Pumputis <martynas@weave.works>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

selftests: netfilter: add simple masq/redirect test cases

Check basic nat/redirect/masquerade for ipv4 and ipv6.

Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

selftests: netfilter: fix config fragment CONFIG_NF_TABLES_INET

In selftests the config fragment for netfilter was added as
NF_TABLES_INET=y and this patch correct it as CONFIG_NF_TABLES_INET=y

Signed-off-by: Naresh Kamboju <naresh.kamboju@linaro.org>
Acked-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

sctp: check and update stream->out_curr when allocating stream_out

Now when using stream reconfig to add out streams, stream->out
will get re-allocated, and all old streams' information will
be copied to the new ones and the old ones will be freed.

So without stream->out_curr updated, next time when trying to
send from stream->out_curr stream, a panic would be caused.

This patch is to check and update stream->out_curr when
allocating stream_out.

v1->v2:
  - define fa_index() to get elem index from stream->out_curr.
v2->v3:
  - repost with no change.

Fixes: 5bbbbe32a431 ("sctp: introduce stream scheduler foundations")
Reported-by: Ying Xu <yinxu@redhat.com>
Reported-by: syzbot+e33a3a138267ca119c7d@syzkaller.appspotmail.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>

net: systemport: Fix WoL with password after deep sleep

Broadcom STB chips support a deep sleep mode where all register
contents are lost. Because we were stashing the MagicPacket password
into some of these registers a suspend into that deep sleep then a
resumption would not lead to being able to wake-up from MagicPacket with
password again.

Fix this by keeping a software copy of the password and program it
during suspend.

Fixes: 83e82f4c706b ("net: systemport: add Wake-on-LAN support")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'vsock-virtio-hot-unplug'

Stefano Garzarella says:

====================
vsock/virtio: fix issues on device hot-unplug

These patches try to handle the hot-unplug of vsock virtio transport device in
a proper way.

Maybe move the vsock_core_init()/vsock_core_exit() functions in the module_init
and module_exit of vsock_virtio_transport module can't be the best way, but the
architecture of vsock_core forces us to this approach for now.

The vsock_core proto_ops expect a valid pointer to the transport device, so we
can't call vsock_core_exit() until there are open sockets.

v2 -> v3:
 - Rebased on master

v1 -> v2:
 - Fixed commit message of patch 1.
 - Added Reviewed-by, Acked-by tags by Stefan
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

vsock/virtio: reset connected sockets on device removal

When the virtio transport device disappear, we should reset all
connected sockets in order to inform the users.

Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

vsock/virtio: fix kernel panic after device hot-unplug

virtio_vsock_remove() invokes the vsock_core_exit() also if there
are opened sockets for the AF_VSOCK protocol family. In this way
the vsock "transport" pointer is set to NULL, triggering the
kernel panic at the first socket activity.

This patch move the vsock_core_init()/vsock_core_exit() in the
virtio_vsock respectively in module_init and module_exit functions,
that cannot be invoked until there are open sockets.

Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1609699
Reported-by: Yan Fu <yafu@redhat.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Revert "net: phy: marvell: avoid pause mode on SGMII-to-Copper for 88e151x"

This reverts commit 6623c0fba10ef45b64ca213ad5dec926f37fa9a0.

The original diagnosis was incorrect: it appears that the NIC had
PHY polling mode enabled, which meant that it overwrote the PHYs
advertisement register during negotiation.

Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Tested-by: Yonglong Liu <liuyonglong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge git://git./pub/scm/linux/kernel/git/bpf/bpf

Alexei Starovoitov says:

====================
pull-request: bpf 2019-01-31

The following pull-request contains BPF updates for your *net* tree.

The main changes are:

1) disable preemption in sender side of socket filters, from Alexei.

2) fix two potential deadlocks in syscall bpf lookup and prog_register,
   from Martin and Alexei.

3) fix BTF to allow typedef on func_proto, from Yonghong.

4) two bpftool fixes, from Jiri and Paolo.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

dccp: fool proof ccid_hc_[rt]x_parse_options()

Similarly to commit 276bdb82dedb ("dccp: check ccid before dereferencing")
it is wise to test for a NULL ccid.

kasan: CONFIG_KASAN_INLINE enabled
kasan: GPF could be caused by NULL-ptr deref or user memory access
general protection fault: 0000 [#1] PREEMPT SMP KASAN
CPU: 1 PID: 16 Comm: ksoftirqd/1 Not tainted 5.0.0-rc3+ #37
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
RIP: 0010:ccid_hc_tx_parse_options net/dccp/ccid.h:205 [inline]
RIP: 0010:dccp_parse_options+0x8d9/0x12b0 net/dccp/options.c:233
Code: c5 0f b6 75 b3 80 38 00 0f 85 d6 08 00 00 48 b9 00 00 00 00 00 fc ff df 48 8b 45 b8 4c 8b b8 f8 07 00 00 4c 89 f8 48 c1 e8 03 <80> 3c 08 00 0f 85 95 08 00 00 48 b8 00 00 00 00 00 fc ff df 4d 8b
kobject: 'loop5' (0000000080f78fc1): kobject_uevent_env
RSP: 0018:ffff8880a94df0b8 EFLAGS: 00010246
RAX: 0000000000000000 RBX: ffff8880858ac723 RCX: dffffc0000000000
RDX: 0000000000000100 RSI: 0000000000000007 RDI: 0000000000000001
RBP: ffff8880a94df140 R08: 0000000000000001 R09: ffff888061b83a80
R10: ffffed100c370752 R11: ffff888061b83a97 R12: 0000000000000026
R13: 0000000000000001 R14: 0000000000000000 R15: 0000000000000000
FS:  0000000000000000(0000) GS:ffff8880ae700000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007f0defa33518 CR3: 000000008db5e000 CR4: 00000000001406e0
kobject: 'loop5' (0000000080f78fc1): fill_kobj_path: path = '/devices/virtual/block/loop5'
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
Call Trace:
 dccp_rcv_state_process+0x2b6/0x1af6 net/dccp/input.c:654
 dccp_v4_do_rcv+0x100/0x190 net/dccp/ipv4.c:688
 sk_backlog_rcv include/net/sock.h:936 [inline]
 __sk_receive_skb+0x3a9/0xea0 net/core/sock.c:473
 dccp_v4_rcv+0x10cb/0x1f80 net/dccp/ipv4.c:880
 ip_protocol_deliver_rcu+0xb6/0xa20 net/ipv4/ip_input.c:208
 ip_local_deliver_finish+0x23b/0x390 net/ipv4/ip_input.c:234
 NF_HOOK include/linux/netfilter.h:289 [inline]
 NF_HOOK include/linux/netfilter.h:283 [inline]
 ip_local_deliver+0x1f0/0x740 net/ipv4/ip_input.c:255
 dst_input include/net/dst.h:450 [inline]
 ip_rcv_finish+0x1f4/0x2f0 net/ipv4/ip_input.c:414
 NF_HOOK include/linux/netfilter.h:289 [inline]
 NF_HOOK include/linux/netfilter.h:283 [inline]
 ip_rcv+0xed/0x620 net/ipv4/ip_input.c:524
 __netif_receive_skb_one_core+0x160/0x210 net/core/dev.c:4973
 __netif_receive_skb+0x2c/0x1c0 net/core/dev.c:5083
 process_backlog+0x206/0x750 net/core/dev.c:5923
 napi_poll net/core/dev.c:6346 [inline]
 net_rx_action+0x76d/0x1930 net/core/dev.c:6412
 __do_softirq+0x30b/0xb11 kernel/softirq.c:292
 run_ksoftirqd kernel/softirq.c:654 [inline]
 run_ksoftirqd+0x8e/0x110 kernel/softirq.c:646
 smpboot_thread_fn+0x6ab/0xa10 kernel/smpboot.c:164
 kthread+0x357/0x430 kernel/kthread.c:246
 ret_from_fork+0x3a/0x50 arch/x86/entry/entry_64.S:352
Modules linked in:
---[ end trace 58a0ba03bea2c376 ]---
RIP: 0010:ccid_hc_tx_parse_options net/dccp/ccid.h:205 [inline]
RIP: 0010:dccp_parse_options+0x8d9/0x12b0 net/dccp/options.c:233
Code: c5 0f b6 75 b3 80 38 00 0f 85 d6 08 00 00 48 b9 00 00 00 00 00 fc ff df 48 8b 45 b8 4c 8b b8 f8 07 00 00 4c 89 f8 48 c1 e8 03 <80> 3c 08 00 0f 85 95 08 00 00 48 b8 00 00 00 00 00 fc ff df 4d 8b
RSP: 0018:ffff8880a94df0b8 EFLAGS: 00010246
RAX: 0000000000000000 RBX: ffff8880858ac723 RCX: dffffc0000000000
RDX: 0000000000000100 RSI: 0000000000000007 RDI: 0000000000000001
RBP: ffff8880a94df140 R08: 0000000000000001 R09: ffff888061b83a80
R10: ffffed100c370752 R11: ffff888061b83a97 R12: 0000000000000026
R13: 0000000000000001 R14: 0000000000000000 R15: 0000000000000000
FS:  0000000000000000(0000) GS:ffff8880ae700000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007f0defa33518 CR3: 0000000009871000 CR4: 00000000001406e0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400

Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Cc: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'smc-fixes'

Ursula Braun says:

====================
net/smc: fixes 2019-01-30

here are some fixes in different areas of the smc code for the net
tree.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

net/smc: fix use of variable in cleared area

Do not use pend->idx as index for the arrays because its value is
located in the cleared area. Use the existing local variable instead.
Without this fix the wrong area might be cleared.

Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/smc: use device link provided in qp_context

The device field of the IB event structure does not always point to the
SMC IB device. Load the pointer from the qp_context which is always
provided to smc_ib_qp_event_handler() in the priv field. And for qp
events the affected port is given in the qp structure of the ibevent,
derive it from there.

Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/smc: call smc_cdc_msg_send() under send_lock

Call smc_cdc_msg_send() under the connection send_lock to make sure all
send operations for one connection are serialized.

Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/smc: do not wait under send_lock

smc_cdc_get_free_slot() might wait for free transfer buffers when using
SMC-R. This wait should not be done under the send_lock, which is a
spin_lock. This fixes a cpu loop in parallel threads waiting for the
send_lock.

Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/smc: recvmsg and splice_read should return 0 after shutdown

When a socket was connected and is now shut down for read, return 0 to
indicate end of data in recvmsg and splice_read (like TCP) and do not
return ENOTCONN. This behavior is required by the socket api.

Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/smc: don't wait for send buffer space when data was already sent

When there is no more send buffer space and at least 1 byte was already
sent then return to user space. The wait is only done when no data was
sent by the sendmsg() call.
This fixes smc_tx_sendmsg() which tried to always send all user data and
started to wait for free send buffer space when needed. During this wait
the user space program was blocked in the sendmsg() call and hence not
able to receive incoming data. When both sides were in such a situation
then the connection stalled forever.

Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/smc: prevent races between smc_lgr_terminate() and smc_conn_free()

To prevent races between smc_lgr_terminate() and smc_conn_free() add an
extra check of the lgr field before accessing it, and cancel a delayed
free_work when a new smc connection is created.
This fixes the problem that free_work cleared the lgr variable but
smc_lgr_terminate() or smc_conn_free() still access it in parallel.

Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/smc: allow 16 byte pnetids in netlink policy

Currently, users can only send pnetids with a maximum length of 15 bytes
over the SMC netlink interface although the maximum pnetid length is 16
bytes. This patch changes the SMC netlink policy to accept 16 byte
pnetids.

Signed-off-by: Hans Wippel <hwippel@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/smc: fix another sizeof to int comparison

Comparing an int to a size, which is unsigned, causes the int to become
unsigned, giving the wrong result. kernel_sendmsg can return a negative
error code.

Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

enic: fix checksum validation for IPv6

In case of IPv6 pkts, ipv4_csum_ok is 0. Because of this, driver does
not set skb->ip_summed. So IPv6 rx checksum is not offloaded.

Signed-off-by: Govindarajulu Varadarajan <gvaradar@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

sctp: walk the list of asoc safely

In sctp_sendmesg(), when walking the list of endpoint associations, the
association can be dropped from the list, making the list corrupt.
Properly handle this by using list_for_each_entry_safe()

Fixes: 4910280503f3 ("sctp: add support for snd flag SCTP_SENDALL process in sendmsg")
Reported-by: Secunia Research <vuln@secunia.com>
Tested-by: Secunia Research <vuln@secunia.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge tag 'batadv-net-for-davem-20190201' of git://git.open-mesh.org/linux-merge

Simon Wunderlich says:

====================
Here are some batman-adv bugfixes:

 - Avoid WARN to report incorrect configuration, by Sven Eckelmann

 - Fix mac header position setting, by Sven Eckelmann

 - Fix releasing station statistics, by Felix Fietkau
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

Merge tag 'mac80211-for-davem-2019-02-01' of git://git./linux/kernel/git/jberg/mac80211

Johannes Berg says:

====================
Two more fixes:
 * sometimes, not enough tailroom was allocated for
   software-encrypted management frames in mac80211
 * cfg80211 regulatory restore got an additional condition,
   needs to rerun the checks after that condition changes
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

skge: potential memory corruption in skge_get_regs()

The "p" buffer is 0x4000 bytes long.  B3_RI_WTO_R1 is 0x190.  The value
of "regs->len" is in the 1-0x4000 range.  The bug here is that
"regs->len - B3_RI_WTO_R1" can be a negative value which would lead to
memory corruption and an abrupt crash.

Fixes: c3f8be961808 ("[PATCH] skge: expand ethtool debug register dump")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cfg80211: call disconnect_wk when AP stops

Since we now prevent regulatory restore during STA disconnect
if concurrent AP interfaces are active, we need to reschedule
this check when the AP state changes. This fixes never doing
a restore when an AP is the last interface to stop. Or to put
it another way: we need to re-check after anything we check
here changes.

Cc: stable@vger.kernel.org
Fixes: 113f3aaa81bd ("cfg80211: Prevent regulatory restore during STA disconnect in concurrent interfaces")
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mac80211: ensure that mgmt tx skbs have tailroom for encryption

Some drivers use IEEE80211_KEY_FLAG_SW_MGMT_TX to indicate that management
frames need to be software encrypted. Since normal data packets are still
encrypted by the hardware, crypto_tx_tailroom_needed_cnt gets decremented
after key upload to hw. This can lead to passing skbs to ccmp_encrypt_skb,
which don't have the necessary tailroom for software encryption.

Change the code to add tailroom for encrypted management packets, even if
crypto_tx_tailroom_needed_cnt is 0.

Cc: stable@vger.kernel.org
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

Merge branch 'bpf-lockdep-fixes'

Alexei Starovoitov says:

====================
v1->v2:
- reworded 2nd patch. It's a real dead lock. Not a false positive
- dropped the lockdep fix for up_read_non_owner in bpf_get_stackid

In addition to preempt_disable patch for socket filters
https://patchwork.ozlabs.org/patch/1032437/
First patch fixes lockdep false positive in percpu_freelist
Second patch fixes potential deadlock in bpf_prog_register
Third patch fixes another potential deadlock in stackmap access
from tracing bpf prog and from syscall.
====================

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

bpf: Fix syscall's stackmap lookup potential deadlock

The map_lookup_elem used to not acquiring spinlock
in order to optimize the reader.

It was true until commit 557c0c6e7df8 ("bpf: convert stackmap to pre-allocation")
The syscall's map_lookup_elem(stackmap) calls bpf_stackmap_copy().
bpf_stackmap_copy() may find the elem no longer needed after the copy is done.
If that is the case, pcpu_freelist_push() saves this elem for reuse later.
This push requires a spinlock.

If a tracing bpf_prog got run in the middle of the syscall's
map_lookup_elem(stackmap) and this tracing bpf_prog is calling
bpf_get_stackid(stackmap) which also requires the same pcpu_freelist's
spinlock, it may end up with a dead lock situation as reported by
Eric Dumazet in https://patchwork.ozlabs.org/patch/1030266/

The situation is the same as the syscall's map_update_elem() which
needs to acquire the pcpu_freelist's spinlock and could race
with tracing bpf_prog.  Hence, this patch fixes it by protecting
bpf_stackmap_copy() with this_cpu_inc(bpf_prog_active)
to prevent tracing bpf_prog from running.

A later syscall's map_lookup_elem commit f1a2e44a3aec ("bpf: add queue and stack maps")
also acquires a spinlock and races with tracing bpf_prog similarly.
Hence, this patch is forward looking and protects the majority
of the map lookups.  bpf_map_offload_lookup_elem() is the exception
since it is for network bpf_prog only (i.e. never called by tracing
bpf_prog).

Fixes: 557c0c6e7df8 ("bpf: convert stackmap to pre-allocation")
Reported-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

bpf: fix potential deadlock in bpf_prog_register

Lockdep found a potential deadlock between cpu_hotplug_lock, bpf_event_mutex, and cpuctx_mutex:
[   13.007000] WARNING: possible circular locking dependency detected
[   13.007587] 5.0.0-rc3-00018-g2fa53f892422-dirty #477 Not tainted
[   13.008124] ------------------------------------------------------
[   13.008624] test_progs/246 is trying to acquire lock:
[   13.009030] 0000000094160d1d (tracepoints_mutex){+.+.}, at: tracepoint_probe_register_prio+0x2d/0x300
[   13.009770]
[   13.009770] but task is already holding lock:
[   13.010239] 00000000d663ef86 (bpf_event_mutex){+.+.}, at: bpf_probe_register+0x1d/0x60
[   13.010877]
[   13.010877] which lock already depends on the new lock.
[   13.010877]
[   13.011532]
[   13.011532] the existing dependency chain (in reverse order) is:
[   13.012129]
[   13.012129] -> #4 (bpf_event_mutex){+.+.}:
[   13.012582]        perf_event_query_prog_array+0x9b/0x130
[   13.013016]        _perf_ioctl+0x3aa/0x830
[   13.013354]        perf_ioctl+0x2e/0x50
[   13.013668]        do_vfs_ioctl+0x8f/0x6a0
[   13.014003]        ksys_ioctl+0x70/0x80
[   13.014320]        __x64_sys_ioctl+0x16/0x20
[   13.014668]        do_syscall_64+0x4a/0x180
[   13.015007]        entry_SYSCALL_64_after_hwframe+0x49/0xbe
[   13.015469]
[   13.015469] -> #3 (&cpuctx_mutex){+.+.}:
[   13.015910]        perf_event_init_cpu+0x5a/0x90
[   13.016291]        perf_event_init+0x1b2/0x1de
[   13.016654]        start_kernel+0x2b8/0x42a
[   13.016995]        secondary_startup_64+0xa4/0xb0
[   13.017382]
[   13.017382] -> #2 (pmus_lock){+.+.}:
[   13.017794]        perf_event_init_cpu+0x21/0x90
[   13.018172]        cpuhp_invoke_callback+0xb3/0x960
[   13.018573]        _cpu_up+0xa7/0x140
[   13.018871]        do_cpu_up+0xa4/0xc0
[   13.019178]        smp_init+0xcd/0xd2
[   13.019483]        kernel_init_freeable+0x123/0x24f
[   13.019878]        kernel_init+0xa/0x110
[   13.020201]        ret_from_fork+0x24/0x30
[   13.020541]
[   13.020541] -> #1 (cpu_hotplug_lock.rw_sem){++++}:
[   13.021051]        static_key_slow_inc+0xe/0x20
[   13.021424]        tracepoint_probe_register_prio+0x28c/0x300
[   13.021891]        perf_trace_event_init+0x11f/0x250
[   13.022297]        perf_trace_init+0x6b/0xa0
[   13.022644]        perf_tp_event_init+0x25/0x40
[   13.023011]        perf_try_init_event+0x6b/0x90
[   13.023386]        perf_event_alloc+0x9a8/0xc40
[   13.023754]        __do_sys_perf_event_open+0x1dd/0xd30
[   13.024173]        do_syscall_64+0x4a/0x180
[   13.024519]        entry_SYSCALL_64_after_hwframe+0x49/0xbe
[   13.024968]
[   13.024968] -> #0 (tracepoints_mutex){+.+.}:
[   13.025434]        __mutex_lock+0x86/0x970
[   13.025764]        tracepoint_probe_register_prio+0x2d/0x300
[   13.026215]        bpf_probe_register+0x40/0x60
[   13.026584]        bpf_raw_tracepoint_open.isra.34+0xa4/0x130
[   13.027042]        __do_sys_bpf+0x94f/0x1a90
[   13.027389]        do_syscall_64+0x4a/0x180
[   13.027727]        entry_SYSCALL_64_after_hwframe+0x49/0xbe
[   13.028171]
[   13.028171] other info that might help us debug this:
[   13.028171]
[   13.028807] Chain exists of:
[   13.028807]   tracepoints_mutex --> &cpuctx_mutex --> bpf_event_mutex
[   13.028807]
[   13.029666]  Possible unsafe locking scenario:
[   13.029666]
[   13.030140]        CPU0                    CPU1
[   13.030510]        ----                    ----
[   13.030875]   lock(bpf_event_mutex);
[   13.031166]                                lock(&cpuctx_mutex);
[   13.031645]                                lock(bpf_event_mutex);
[   13.032135]   lock(tracepoints_mutex);
[   13.032441]
[   13.032441]  *** DEADLOCK ***
[   13.032441]
[   13.032911] 1 lock held by test_progs/246:
[   13.033239]  #0: 00000000d663ef86 (bpf_event_mutex){+.+.}, at: bpf_probe_register+0x1d/0x60
[   13.033909]
[   13.033909] stack backtrace:
[   13.034258] CPU: 1 PID: 246 Comm: test_progs Not tainted 5.0.0-rc3-00018-g2fa53f892422-dirty #477
[   13.034964] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.11.0-2.el7 04/01/2014
[   13.035657] Call Trace:
[   13.035859]  dump_stack+0x5f/0x8b
[   13.036130]  print_circular_bug.isra.37+0x1ce/0x1db
[   13.036526]  __lock_acquire+0x1158/0x1350
[   13.036852]  ? lock_acquire+0x98/0x190
[   13.037154]  lock_acquire+0x98/0x190
[   13.037447]  ? tracepoint_probe_register_prio+0x2d/0x300
[   13.037876]  __mutex_lock+0x86/0x970
[   13.038167]  ? tracepoint_probe_register_prio+0x2d/0x300
[   13.038600]  ? tracepoint_probe_register_prio+0x2d/0x300
[   13.039028]  ? __mutex_lock+0x86/0x970
[   13.039337]  ? __mutex_lock+0x24a/0x970
[   13.039649]  ? bpf_probe_register+0x1d/0x60
[   13.039992]  ? __bpf_trace_sched_wake_idle_without_ipi+0x10/0x10
[   13.040478]  ? tracepoint_probe_register_prio+0x2d/0x300
[   13.040906]  tracepoint_probe_register_prio+0x2d/0x300
[   13.041325]  bpf_probe_register+0x40/0x60
[   13.041649]  bpf_raw_tracepoint_open.isra.34+0xa4/0x130
[   13.042068]  ? __might_fault+0x3e/0x90
[   13.042374]  __do_sys_bpf+0x94f/0x1a90
[   13.042678]  do_syscall_64+0x4a/0x180
[   13.042975]  entry_SYSCALL_64_after_hwframe+0x49/0xbe
[   13.043382] RIP: 0033:0x7f23b10a07f9
[   13.045155] RSP: 002b:00007ffdef42fdd8 EFLAGS: 00000202 ORIG_RAX: 0000000000000141
[   13.045759] RAX: ffffffffffffffda RBX: 00007ffdef42ff70 RCX: 00007f23b10a07f9
[   13.046326] RDX: 0000000000000070 RSI: 00007ffdef42fe10 RDI: 0000000000000011
[   13.046893] RBP: 00007ffdef42fdf0 R08: 0000000000000038 R09: 00007ffdef42fe10
[   13.047462] R10: 0000000000000000 R11: 0000000000000202 R12: 0000000000000000
[   13.048029] R13: 0000000000000016 R14: 00007f23b1db4690 R15: 0000000000000000

Since tracepoints_mutex will be taken in tracepoint_probe_register/unregister()
there is no need to take bpf_event_mutex too.
bpf_event_mutex is protecting modifications to prog array used in kprobe/perf bpf progs.
bpf_raw_tracepoints don't need to take this mutex.

Fixes: c4f6699dfcb8 ("bpf: introduce BPF_RAW_TRACEPOINT")
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

bpf: fix lockdep false positive in percpu_freelist

Lockdep warns about false positive:
[   12.492084] 00000000e6b28347 (&head->lock){+...}, at: pcpu_freelist_push+0x2a/0x40
[   12.492696] but this lock was taken by another, HARDIRQ-safe lock in the past:
[   12.493275]  (&rq->lock){-.-.}
[   12.493276]
[   12.493276]
[   12.493276] and interrupts could create inverse lock ordering between them.
[   12.493276]
[   12.494435]
[   12.494435] other info that might help us debug this:
[   12.494979]  Possible interrupt unsafe locking scenario:
[   12.494979]
[   12.495518]        CPU0                    CPU1
[   12.495879]        ----                    ----
[   12.496243]   lock(&head->lock);
[   12.496502]                                local_irq_disable();
[   12.496969]                                lock(&rq->lock);
[   12.497431]                                lock(&head->lock);
[   12.497890]   <Interrupt>
[   12.498104]     lock(&rq->lock);
[   12.498368]
[   12.498368]  *** DEADLOCK ***
[   12.498368]
[   12.498837] 1 lock held by dd/276:
[   12.499110]  #0: 00000000c58cb2ee (rcu_read_lock){....}, at: trace_call_bpf+0x5e/0x240
[   12.499747]
[   12.499747] the shortest dependencies between 2nd lock and 1st lock:
[   12.500389]  -> (&rq->lock){-.-.} {
[   12.500669]     IN-HARDIRQ-W at:
[   12.500934]                       _raw_spin_lock+0x2f/0x40
[   12.501373]                       scheduler_tick+0x4c/0xf0
[   12.501812]                       update_process_times+0x40/0x50
[   12.502294]                       tick_periodic+0x27/0xb0
[   12.502723]                       tick_handle_periodic+0x1f/0x60
[   12.503203]                       timer_interrupt+0x11/0x20
[   12.503651]                       __handle_irq_event_percpu+0x43/0x2c0
[   12.504167]                       handle_irq_event_percpu+0x20/0x50
[   12.504674]                       handle_irq_event+0x37/0x60
[   12.505139]                       handle_level_irq+0xa7/0x120
[   12.505601]                       handle_irq+0xa1/0x150
[   12.506018]                       do_IRQ+0x77/0x140
[   12.506411]                       ret_from_intr+0x0/0x1d
[   12.506834]                       _raw_spin_unlock_irqrestore+0x53/0x60
[   12.507362]                       __setup_irq+0x481/0x730
[   12.507789]                       setup_irq+0x49/0x80
[   12.508195]                       hpet_time_init+0x21/0x32
[   12.508644]                       x86_late_time_init+0xb/0x16
[   12.509106]                       start_kernel+0x390/0x42a
[   12.509554]                       secondary_startup_64+0xa4/0xb0
[   12.510034]     IN-SOFTIRQ-W at:
[   12.510305]                       _raw_spin_lock+0x2f/0x40
[   12.510772]                       try_to_wake_up+0x1c7/0x4e0
[   12.511220]                       swake_up_locked+0x20/0x40
[   12.511657]                       swake_up_one+0x1a/0x30
[   12.512070]                       rcu_process_callbacks+0xc5/0x650
[   12.512553]                       __do_softirq+0xe6/0x47b
[   12.512978]                       irq_exit+0xc3/0xd0
[   12.513372]                       smp_apic_timer_interrupt+0xa9/0x250
[   12.513876]                       apic_timer_interrupt+0xf/0x20
[   12.514343]                       default_idle+0x1c/0x170
[   12.514765]                       do_idle+0x199/0x240
[   12.515159]                       cpu_startup_entry+0x19/0x20
[   12.515614]                       start_kernel+0x422/0x42a
[   12.516045]                       secondary_startup_64+0xa4/0xb0
[   12.516521]     INITIAL USE at:
[   12.516774]                      _raw_spin_lock_irqsave+0x38/0x50
[   12.517258]                      rq_attach_root+0x16/0xd0
[   12.517685]                      sched_init+0x2f2/0x3eb
[   12.518096]                      start_kernel+0x1fb/0x42a
[   12.518525]                      secondary_startup_64+0xa4/0xb0
[   12.518986]   }
[   12.519132]   ... key      at: [<ffffffff82b7bc28>] __key.71384+0x0/0x8
[   12.519649]   ... acquired at:
[   12.519892]    pcpu_freelist_pop+0x7b/0xd0
[   12.520221]    bpf_get_stackid+0x1d2/0x4d0
[   12.520563]    ___bpf_prog_run+0x8b4/0x11a0
[   12.520887]
[   12.521008] -> (&head->lock){+...} {
[   12.521292]    HARDIRQ-ON-W at:
[   12.521539]                     _raw_spin_lock+0x2f/0x40
[   12.521950]                     pcpu_freelist_push+0x2a/0x40
[   12.522396]                     bpf_get_stackid+0x494/0x4d0
[   12.522828]                     ___bpf_prog_run+0x8b4/0x11a0
[   12.523296]    INITIAL USE at:
[   12.523537]                    _raw_spin_lock+0x2f/0x40
[   12.523944]                    pcpu_freelist_populate+0xc0/0x120
[   12.524417]                    htab_map_alloc+0x405/0x500
[   12.524835]                    __do_sys_bpf+0x1a3/0x1a90
[   12.525253]                    do_syscall_64+0x4a/0x180
[   12.525659]                    entry_SYSCALL_64_after_hwframe+0x49/0xbe
[   12.526167]  }
[   12.526311]  ... key      at: [<ffffffff838f7668>] __key.13130+0x0/0x8
[   12.526812]  ... acquired at:
[   12.527047]    __lock_acquire+0x521/0x1350
[   12.527371]    lock_acquire+0x98/0x190
[   12.527680]    _raw_spin_lock+0x2f/0x40
[   12.527994]    pcpu_freelist_push+0x2a/0x40
[   12.528325]    bpf_get_stackid+0x494/0x4d0
[   12.528645]    ___bpf_prog_run+0x8b4/0x11a0
[   12.528970]
[   12.529092]
[   12.529092] stack backtrace:
[   12.529444] CPU: 0 PID: 276 Comm: dd Not tainted 5.0.0-rc3-00018-g2fa53f892422 #475
[   12.530043] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.11.0-2.el7 04/01/2014
[   12.530750] Call Trace:
[   12.530948]  dump_stack+0x5f/0x8b
[   12.531248]  check_usage_backwards+0x10c/0x120
[   12.531598]  ? ___bpf_prog_run+0x8b4/0x11a0
[   12.531935]  ? mark_lock+0x382/0x560
[   12.532229]  mark_lock+0x382/0x560
[   12.532496]  ? print_shortest_lock_dependencies+0x180/0x180
[   12.532928]  __lock_acquire+0x521/0x1350
[   12.533271]  ? find_get_entry+0x17f/0x2e0
[   12.533586]  ? find_get_entry+0x19c/0x2e0
[   12.533902]  ? lock_acquire+0x98/0x190
[   12.534196]  lock_acquire+0x98/0x190
[   12.534482]  ? pcpu_freelist_push+0x2a/0x40
[   12.534810]  _raw_spin_lock+0x2f/0x40
[   12.535099]  ? pcpu_freelist_push+0x2a/0x40
[   12.535432]  pcpu_freelist_push+0x2a/0x40
[   12.535750]  bpf_get_stackid+0x494/0x4d0
[   12.536062]  ___bpf_prog_run+0x8b4/0x11a0

It has been explained that is a false positive here:
https://lkml.org/lkml/2018/7/25/756
Recap:
- stackmap uses pcpu_freelist
- The lock in pcpu_freelist is a percpu lock
- stackmap is only used by tracing bpf_prog
- A tracing bpf_prog cannot be run if another bpf_prog
  has already been running (ensured by the percpu bpf_prog_active counter).

Eric pointed out that this lockdep splats stops other
legit lockdep splats in selftests/bpf/test_progs.c.

Fix this by calling local_irq_save/restore for stackmap.

Another false positive had also been worked around by calling
local_irq_save in commit 89ad2fa3f043 ("bpf: fix lockdep splat").
That commit added unnecessary irq_save/restore to fast path of
bpf hash map. irqs are already disabled at that point, since htab
is holding per bucket spin_lock with irqsave.

Let's reduce overhead for htab by introducing __pcpu_freelist_push/pop
function w/o irqsave and convert pcpu_freelist_push/pop to irqsave
to be used elsewhere (right now only in stackmap).
It stops lockdep false positive in stackmap with a bit of acceptable overhead.

Fixes: 557c0c6e7df8 ("bpf: convert stackmap to pre-allocation")
Reported-by: Naresh Kamboju <naresh.kamboju@linaro.org>
Reported-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

bpf: run bpf programs with preemption disabled

Disabled preemption is necessary for proper access to per-cpu maps
from BPF programs.

But the sender side of socket filters didn't have preemption disabled:
unix_dgram_sendmsg->sk_filter->sk_filter_trim_cap->bpf_prog_run_save_cb->BPF_PROG_RUN

and a combination of af_packet with tun device didn't disable either:
tpacket_snd->packet_direct_xmit->packet_pick_tx_queue->ndo_select_queue->
  tun_select_queue->tun_ebpf_select_queue->bpf_prog_run_clear_cb->BPF_PROG_RUN

Disable preemption before executing BPF programs (both classic and extended).

Reported-by: Jann Horn <jannh@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

bpf, selftests: fix handling of sparse CPU allocations

Previously, bpf_num_possible_cpus() had a bug when calculating a
number of possible CPUs in the case of sparse CPU allocations, as
it was considering only the first range or element of
/sys/devices/system/cpu/possible.

E.g. in the case of "0,2-3" (CPU 1 is not available), the function
returned 1 instead of 3.

This patch fixes the function by making it parse all CPU ranges and
elements.

Signed-off-by: Martynas Pumputis <m@lambda.lt>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

bnxt_en: Disable interrupts when allocating CP rings or NQs.

When calling firmware to allocate a CP ring or NQ, an interrupt associated
with that ring may be generated immediately before the doorbell is even
setup after the firmware call returns.  When servicing the interrupt, the
driver may crash when trying to access the doorbell.

Fix it by disabling interrupt on that vector until the doorbell is
set up.

Fixes: 697197e5a173 ("bnxt_en: Re-structure doorbells.")
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'ieee802154-for-davem-2019-01-31' of git://git./linux/kernel/git/sschmidt/wpan

Stefan Schmidt says:

====================
pull-request: ieee802154 for net 2019-01-31

An update from ieee802154 for your *net* tree.

I waited a while to see if anything else comes up, but it seems this time
we only have one fixup patch for the -rc rounds.
Colin fixed some indentation in the mcr20a drivers. That's about it.

If there are any problems with taking these two before the final 5.0 let
me know.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

rds: fix refcount bug in rds_sock_addref

syzbot was able to catch a bug in rds [1]

The issue here is that the socket might be found in a hash table
but that its refcount has already be set to 0 by another cpu.

We need to use refcount_inc_not_zero() to be safe here.

[1]

refcount_t: increment on 0; use-after-free.
WARNING: CPU: 1 PID: 23129 at lib/refcount.c:153 refcount_inc_checked lib/refcount.c:153 [inline]
WARNING: CPU: 1 PID: 23129 at lib/refcount.c:153 refcount_inc_checked+0x61/0x70 lib/refcount.c:151
Kernel panic - not syncing: panic_on_warn set ...
CPU: 1 PID: 23129 Comm: syz-executor3 Not tainted 5.0.0-rc4+ #53
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Call Trace:
 __dump_stack lib/dump_stack.c:77 [inline]
 dump_stack+0x1db/0x2d0 lib/dump_stack.c:113
 panic+0x2cb/0x65c kernel/panic.c:214
 __warn.cold+0x20/0x48 kernel/panic.c:571
 report_bug+0x263/0x2b0 lib/bug.c:186
 fixup_bug arch/x86/kernel/traps.c:178 [inline]
 fixup_bug arch/x86/kernel/traps.c:173 [inline]
 do_error_trap+0x11b/0x200 arch/x86/kernel/traps.c:271
 do_invalid_op+0x37/0x50 arch/x86/kernel/traps.c:290
 invalid_op+0x14/0x20 arch/x86/entry/entry_64.S:973
RIP: 0010:refcount_inc_checked lib/refcount.c:153 [inline]
RIP: 0010:refcount_inc_checked+0x61/0x70 lib/refcount.c:151
Code: 1d 51 63 c8 06 31 ff 89 de e8 eb 1b f2 fd 84 db 75 dd e8 a2 1a f2 fd 48 c7 c7 60 9f 81 88 c6 05 31 63 c8 06 01 e8 af 65 bb fd <0f> 0b eb c1 90 66 2e 0f 1f 84 00 00 00 00 00 55 48 89 e5 41 54 49
RSP: 0018:ffff8880a0cbf1e8 EFLAGS: 00010282
RAX: 0000000000000000 RBX: 0000000000000000 RCX: ffffc90006113000
RDX: 000000000001047d RSI: ffffffff81685776 RDI: 0000000000000005
RBP: ffff8880a0cbf1f8 R08: ffff888097c9e100 R09: ffffed1015ce5021
R10: ffffed1015ce5020 R11: ffff8880ae728107 R12: ffff8880723c20c0
R13: ffff8880723c24b0 R14: dffffc0000000000 R15: ffffed1014197e64
 sock_hold include/net/sock.h:647 [inline]
 rds_sock_addref+0x19/0x20 net/rds/af_rds.c:675
 rds_find_bound+0x97c/0x1080 net/rds/bind.c:82
 rds_recv_incoming+0x3be/0x1430 net/rds/recv.c:362
 rds_loop_xmit+0xf3/0x2a0 net/rds/loop.c:96
 rds_send_xmit+0x1355/0x2a10 net/rds/send.c:355
 rds_sendmsg+0x323c/0x44e0 net/rds/send.c:1368
 sock_sendmsg_nosec net/socket.c:621 [inline]
 sock_sendmsg+0xdd/0x130 net/socket.c:631
 __sys_sendto+0x387/0x5f0 net/socket.c:1788
 __do_sys_sendto net/socket.c:1800 [inline]
 __se_sys_sendto net/socket.c:1796 [inline]
 __x64_sys_sendto+0xe1/0x1a0 net/socket.c:1796
 do_syscall_64+0x1a3/0x800 arch/x86/entry/common.c:290
 entry_SYSCALL_64_after_hwframe+0x49/0xbe
RIP: 0033:0x458089
Code: 6d b7 fb ff c3 66 2e 0f 1f 84 00 00 00 00 00 66 90 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 0f 83 3b b7 fb ff c3 66 2e 0f 1f 84 00 00 00 00
RSP: 002b:00007fc266df8c78 EFLAGS: 00000246 ORIG_RAX: 000000000000002c
RAX: ffffffffffffffda RBX: 0000000000000006 RCX: 0000000000458089
RDX: 0000000000000000 RSI: 00000000204b3fff RDI: 0000000000000005
RBP: 000000000073bf00 R08: 00000000202b4000 R09: 0000000000000010
R10: 0000000000000000 R11: 0000000000000246 R12: 00007fc266df96d4
R13: 00000000004c56e4 R14: 00000000004d94a8 R15: 00000000ffffffff

Fixes: cc4dfb7f70a3 ("rds: fix two RCU related problems")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Cc: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Cc: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Cc: rds-devel@oss.oracle.com
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

lib/test_rhashtable: Make test_insert_dup() allocate its hash table dynamically

The test_insert_dup() function from lib/test_rhashtable.c passes a
pointer to a stack object to rhltable_init(). Allocate the hash table
dynamically to avoid that the following is reported with object
debugging enabled:

ODEBUG: object (ptrval) is on stack (ptrval), but NOT annotated.
WARNING: CPU: 0 PID: 1 at lib/debugobjects.c:368 __debug_object_init+0x312/0x480
Modules linked in:
EIP: __debug_object_init+0x312/0x480
Call Trace:
 ? debug_object_init+0x1a/0x20
 ? __init_work+0x16/0x30
 ? rhashtable_init+0x1e1/0x460
 ? sched_clock_cpu+0x57/0xe0
 ? rhltable_init+0xb/0x20
 ? test_insert_dup+0x32/0x20f
 ? trace_hardirqs_on+0x38/0xf0
 ? ida_dump+0x10/0x10
 ? jhash+0x130/0x130
 ? my_hashfn+0x30/0x30
 ? test_rht_init+0x6aa/0xab4
 ? ida_dump+0x10/0x10
 ? test_rhltable+0xc5c/0xc5c
 ? do_one_initcall+0x67/0x28e
 ? trace_hardirqs_off+0x22/0xe0
 ? restore_all_kernel+0xf/0x70
 ? trace_hardirqs_on_thunk+0xc/0x10
 ? restore_all_kernel+0xf/0x70
 ? kernel_init_freeable+0x142/0x213
 ? rest_init+0x230/0x230
 ? kernel_init+0x10/0x110
 ? schedule_tail_wrapper+0x9/0xc
 ? ret_from_fork+0x19/0x24

Cc: Thomas Graf <tgraf@suug.ch>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: netdev@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

l2tp: copy 4 more bytes to linear part if necessary

The size of L2TPv2 header with all optional fields is 14 bytes.
l2tp_udp_recv_core only moves 10 bytes to the linear part of a
skb. This may lead to l2tp_recv_common read data outside of a skb.

This patch make sure that there is at least 14 bytes in the linear
part of a skb to meet the maximum need of l2tp_udp_recv_core and
l2tp_recv_common. The minimum size of both PPP HDLC-like frame and
Ethernet frame is larger than 14 bytes, so we are safe to do so.

Also remove L2TP_HDR_SIZE_NOSEQ, it is unused now.

Fixes: fd558d186df2 ("l2tp: Split pppol2tp patch into separate l2tp and ppp parts")
Suggested-by: Guillaume Nault <gnault@redhat.com>
Signed-off-by: Jacob Wen <jian.w.wen@oracle.com>
Acked-by: Guillaume Nault <gnault@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ieee802154: mcr20a: fix indentation, remove tabs

The are a couple of statments that are one level too deep, fix this by
removing tabs.

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Stefan Schmidt <stefan@datenfreihafen.org>

Merge branch 'stmmac-fixes'

Jose Abreu says:

====================
net: stmmac: Misc fixes

Some misc fixes for stmmac targeting -net.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

net: stmmac: Disable EEE mode earlier in XMIT callback

In stmmac xmit callback we use a different flow for TSO packets but TSO
xmit callback is not disabling the EEE mode.

Fix this by disabling earlier the EEE mode, i.e. before calling the TSO
xmit callback.

Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: stmmac: Send TSO packets always from Queue 0

The number of TSO enabled channels in HW can be different than the
number of total channels. There is no way to determined, at runtime, the
number of TSO capable channels and its safe to assume that if TSO is
enabled then at least channel 0 will be TSO capable.

Lets always send TSO packets from Queue 0.

Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: stmmac: Fallback to Platform Data clock in Watchdog conversion

If we don't have DT then stmmac_clk will not be available. Let's add a
new Platform Data field so that we can specify the refclk by this mean.

This way we can still use the coalesce command in PCI based setups.

Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipvlan, l3mdev: fix broken l3s mode wrt local routes

While implementing ipvlan l3 and l3s mode for kubernetes CNI plugin,
I ran into the issue that while l3 mode is working fine, l3s mode
does not have any connectivity to kube-apiserver and hence all pods
end up in Error state as well. The ipvlan master device sits on
top of a bond device and hostns traffic to kube-apiserver (also running
in hostns) is DNATed from 10.152.183.1:443 to 139.178.29.207:37573
where the latter is the address of the bond0. While in l3 mode, a
curl to https://10.152.183.1:443 or to https://139.178.29.207:37573
works fine from hostns, neither of them do in case of l3s. In the
latter only a curl to https://127.0.0.1:37573 appeared to work where
for local addresses of bond0 I saw kernel suddenly starting to emit
ARP requests to query HW address of bond0 which remained unanswered
and neighbor entries in INCOMPLETE state. These ARP requests only
happen while in l3s.

Debugging this further, I found the issue is that l3s mode is piggy-
backing on l3 master device, and in this case local routes are using
l3mdev_master_dev_rcu(dev) instead of net->loopback_dev as per commit
f5a0aab84b74 ("net: ipv4: dst for local input routes should use l3mdev
if relevant") and 5f02ce24c269 ("net: l3mdev: Allow the l3mdev to be
a loopback"). I found that reverting them back into using the
net->loopback_dev fixed ipvlan l3s connectivity and got everything
working for the CNI.

Now judging from 4fbae7d83c98 ("ipvlan: Introduce l3s mode") and the
l3mdev paper in [0] the only sole reason why ipvlan l3s is relying
on l3 master device is to get the l3mdev_ip_rcv() receive hook for
setting the dst entry of the input route without adding its own
ipvlan specific hacks into the receive path, however, any l3 domain
semantics beyond just that are breaking l3s operation. Note that
ipvlan also has the ability to dynamically switch its internal
operation from l3 to l3s for all ports via ipvlan_set_port_mode()
at runtime. In any case, l3 vs l3s soley distinguishes itself by
'de-confusing' netfilter through switching skb->dev to ipvlan slave
device late in NF_INET_LOCAL_IN before handing the skb to L4.

Minimal fix taken here is to add a IFF_L3MDEV_RX_HANDLER flag which,
if set from ipvlan setup, gets us only the wanted l3mdev_l3_rcv() hook
without any additional l3mdev semantics on top. This should also have
minimal impact since dev->priv_flags is already hot in cache. With
this set, l3s mode is working fine and I also get things like
masquerading pod traffic on the ipvlan master properly working.

  [0] https://netdevconf.org/1.2/papers/ahern-what-is-l3mdev-paper.pdf

Fixes: f5a0aab84b74 ("net: ipv4: dst for local input routes should use l3mdev if relevant")
Fixes: 5f02ce24c269 ("net: l3mdev: Allow the l3mdev to be a loopback")
Fixes: 4fbae7d83c98 ("ipvlan: Introduce l3s mode")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Cc: Mahesh Bandewar <maheshb@google.com>
Cc: David Ahern <dsa@cumulusnetworks.com>
Cc: Florian Westphal <fw@strlen.de>
Cc: Martynas Pumputis <m@lambda.lt>
Acked-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

l2tp: fix reading optional fields of L2TPv3

Use pskb_may_pull() to make sure the optional fields are in skb linear
parts, so we can safely read them later.

It's easy to reproduce the issue with a net driver that supports paged
skb data. Just create a L2TPv3 over IP tunnel and then generates some
network traffic.
Once reproduced, rx err in /sys/kernel/debug/l2tp/tunnels will increase.

Changes in v4:
1. s/l2tp_v3_pull_opt/l2tp_v3_ensure_opt_in_linear/
2. s/tunnel->version != L2TP_HDR_VER_2/tunnel->version == L2TP_HDR_VER_3/
3. Add 'Fixes' in commit messages.

Changes in v3:
1. To keep consistency, move the code out of l2tp_recv_common.
2. Use "net" instead of "net-next", since this is a bug fix.

Changes in v2:
1. Only fix L2TPv3 to make code simple.
   To fix both L2TPv3 and L2TPv2, we'd better refactor l2tp_recv_common.
   It's complicated to do so.
2. Reloading pointers after pskb_may_pull

Fixes: f7faffa3ff8e ("l2tp: Add L2TPv3 protocol support")
Fixes: 0d76751fad77 ("l2tp: Add L2TPv3 IP encapsulation (no UDP) support")
Fixes: a32e0eec7042 ("l2tp: introduce L2TPv3 IP encapsulation support for IPv6")
Signed-off-by: Jacob Wen <jian.w.wen@oracle.com>
Acked-by: Guillaume Nault <gnault@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tun: move the call to tun_set_real_num_queues

Call tun_set_real_num_queues() after the increment of tun->numqueues
since the former depends on it. Otherwise, the number of queues is not
correctly accounted for, which results to warnings similar to:
"vnet0 selects TX queue 11, but real number of TX queues is 11".

Fixes: 0b7959b62573 ("tun: publish tfile after it's fully initialized")
Reported-and-tested-by: George Amanakis <gamanakis@gmail.com>
Signed-off-by: George Amanakis <gamanakis@gmail.com>
Signed-off-by: Stanislav Fomichev <sdf@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv6: sr: clear IP6CB(skb) on SRH ip4ip6 encapsulation

skb->cb may contain data from previous layers (in an observed case
IPv4 with L3 Master Device). In the observed scenario, the data in
IPCB(skb)->frags was misinterpreted as IP6CB(skb)->frag_max_size,
eventually caused an unexpected IPv6 fragmentation in ip6_fragment()
through ip6_finish_output().

This patch clears IP6CB(skb), which potentially contains garbage data,
on the SRH ip4ip6 encapsulation.

Fixes: 32d99d0b6702 ("ipv6: sr: add support for ip4ip6 encapsulation")
Signed-off-by: Yohei Kanemaru <yohei.kanemaru@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'virtio_net-Fix-problems-around-XDP-tx-and-napi_tx'

Toshiaki Makita says:

====================
virtio_net: Fix problems around XDP tx and napi_tx

While I'm looking into how to account standard tx counters on XDP tx
processing, I found several bugs around XDP tx and napi_tx.

Patch1: Fix oops on error path. Patch2 depends on this.
Patch2: Fix memory corruption on freeing xdp_frames with napi_tx enabled.
Patch3: Minor fix patch5 depends on.
Patch4: Fix memory corruption on processing xdp_frames when XDP is disabled.
  Also patch5 depends on this.
Patch5: Fix memory corruption on processing xdp_frames while XDP is being
  disabled.
Patch6: Minor fix patch7 depends on.
Patch7: Fix memory corruption on freeing sk_buff or xdp_frames when a normal
  queue is reused for XDP and vise versa.

v2:
- patch5: Make rcu_assign_pointer/synchronize_net conditional instead of
          _virtnet_set_queues.
- patch7: Use napi_consume_skb() instead of dev_consume_skb_any()
====================

Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>

virtio_net: Differentiate sk_buff and xdp_frame on freeing

We do not reset or free up unused buffers when enabling/disabling XDP,
so it can happen that xdp_frames are freed after disabling XDP or
sk_buffs are freed after enabling XDP on xdp tx queues.
Thus we need to handle both forms (xdp_frames and sk_buffs) regardless
of XDP setting.
One way to trigger this problem is to disable XDP when napi_tx is
enabled. In that case, virtnet_xdp_set() calls virtnet_napi_enable()
which kicks NAPI. The NAPI handler will call virtnet_poll_cleantx()
which invokes free_old_xmit_skbs() for queues which have been used by
XDP.

Note that even with this change we need to keep skipping
free_old_xmit_skbs() from NAPI handlers when XDP is enabled, because XDP
tx queues do not aquire queue locks.

- v2: Use napi_consume_skb() instead of dev_consume_skb_any()

Fixes: 4941d472bf95 ("virtio-net: do not reset during XDP set")
Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Acked-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

virtio_net: Use xdp_return_frame to free xdp_frames on destroying vqs

put_page() can work as a fallback for freeing xdp_frames, but the
appropriate way is to use xdp_return_frame().

Fixes: cac320c850ef ("virtio_net: convert to use generic xdp_frame and xdp_return_frame API")
Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Acked-by: Jason Wang <jasowang@redhat.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

virtio_net: Don't process redirected XDP frames when XDP is disabled

Commit 8dcc5b0ab0ec ("virtio_net: fix ndo_xdp_xmit crash towards dev not
ready for XDP") tried to avoid access to unexpected sq while XDP is
disabled, but was not complete.

There was a small window which causes out of bounds sq access in
virtnet_xdp_xmit() while disabling XDP.

An example case of
 - curr_queue_pairs = 6 (2 for SKB and 4 for XDP)
 - online_cpu_num = xdp_queue_paris = 4
when XDP is enabled:

CPU 0                         CPU 1
(Disabling XDP)               (Processing redirected XDP frames)

                              virtnet_xdp_xmit()
virtnet_xdp_set()
 _virtnet_set_queues()
  set curr_queue_pairs (2)
                               check if rq->xdp_prog is not NULL
                               virtnet_xdp_sq(vi)
                                qp = curr_queue_pairs -
                                     xdp_queue_pairs +
                                     smp_processor_id()
                                   = 2 - 4 + 1 = -1
                                sq = &vi->sq[qp] // out of bounds access
  set xdp_queue_pairs (0)
  rq->xdp_prog = NULL

Basically we should not change curr_queue_pairs and xdp_queue_pairs
while someone can read the values. Thus, when disabling XDP, assign NULL
to rq->xdp_prog first, and wait for RCU grace period, then change
xxx_queue_pairs.
Note that we need to keep the current order when enabling XDP though.

- v2: Make rcu_assign_pointer/synchronize_net conditional instead of
      _virtnet_set_queues.

Fixes: 186b3c998c50 ("virtio-net: support XDP_REDIRECT")
Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Acked-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

virtio_net: Fix out of bounds access of sq

When XDP is disabled, curr_queue_pairs + smp_processor_id() can be
larger than max_queue_pairs.
There is no guarantee that we have enough XDP send queues dedicated for
each cpu when XDP is disabled, so do not count drops on sq in that case.

Fixes: 5b8f3c8d30a6 ("virtio_net: Add XDP related stats")
Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Acked-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

virtio_net: Fix not restoring real_num_rx_queues

When _virtnet_set_queues() failed we did not restore real_num_rx_queues.
Fix this by placing the change of real_num_rx_queues after
_virtnet_set_queues().
This order is also in line with virtnet_set_channels().

Fixes: 4941d472bf95 ("virtio-net: do not reset during XDP set")
Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Acked-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

virtio_net: Don't call free_old_xmit_skbs for xdp_frames

When napi_tx is enabled, virtnet_poll_cleantx() called
free_old_xmit_skbs() even for xdp send queue.
This is bogus since the queue has xdp_frames, not sk_buffs, thus mangled
device tx bytes counters because skb->len is meaningless value, and even
triggered oops due to general protection fault on freeing them.

Since xdp send queues do not aquire locks, old xdp_frames should be
freed only in virtnet_xdp_xmit(), so just skip free_old_xmit_skbs() for
xdp send queues.

Similarly virtnet_poll_tx() called free_old_xmit_skbs(). This NAPI
handler is called even without calling start_xmit() because cb for tx is
by default enabled. Once the handler is called, it enabled the cb again,
and then the handler would be called again. We don't need this handler
for XDP, so don't enable cb as well as not calling free_old_xmit_skbs().

Also, we need to disable tx NAPI when disabling XDP, so
virtnet_poll_tx() can safely access curr_queue_pairs and
xdp_queue_pairs, which are not atomically updated while disabling XDP.

Fixes: b92f1e6751a6 ("virtio-net: transmit napi")
Fixes: 7b0411ef4aa6 ("virtio-net: clean tx descriptors from rx napi")
Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Acked-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

virtio_net: Don't enable NAPI when interface is down

Commit 4e09ff536284 ("virtio-net: disable NAPI only when enabled during
XDP set") tried to fix inappropriate NAPI enabling/disabling when
!netif_running(), but was not complete.

On error path virtio_net could enable NAPI even when !netif_running().
This can cause enabling NAPI twice on virtnet_open(), which would
trigger BUG_ON() in napi_enable().

Fixes: 4941d472bf95b ("virtio-net: do not reset during XDP set")
Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Acked-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'erspan-always-reports-output-key-to-userspace'

Lorenzo Bianconi says:

====================
erspan: always reports output key to userspace

Erspan protocol relies on output key to set session id header field.
However TUNNEL_KEY bit is cleared in order to not add key field to
the external GRE header and so the configured o_key is not reported
to userspace.
Fix the issue adding TUNNEL_KEY bit to the o_flags parameter dumping
device info
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

net: ip6_gre: always reports o_key to userspace

As Erspan_v4, Erspan_v6 protocol relies on o_key to configure
session id header field. However TUNNEL_KEY bit is cleared in
ip6erspan_tunnel_xmit since ERSPAN protocol does not set the key field
of the external GRE header and so the configured o_key is not reported
to userspace. The issue can be triggered with the following reproducer:

$ip link add ip6erspan1 type ip6erspan local 2000::1 remote 2000::2 \
    key 1 seq erspan_ver 1
$ip link set ip6erspan1 up
ip -d link sh ip6erspan1

ip6erspan1@NONE: <BROADCAST,MULTICAST> mtu 1422 qdisc noop state DOWN mode DEFAULT
    link/ether ba:ff:09:24:c3:0e brd ff:ff:ff:ff:ff:ff promiscuity 0 minmtu 68 maxmtu 1500
    ip6erspan remote 2000::2 local 2000::1 encaplimit 4 flowlabel 0x00000 ikey 0.0.0.1 iseq oseq

Fix the issue adding TUNNEL_KEY bit to the o_flags parameter in
ip6gre_fill_info

Fixes: 5a963eb61b7c ("ip6_gre: Add ERSPAN native tunnel support")
Signed-off-by: Lorenzo Bianconi <lorenzo.bianconi@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: ip_gre: always reports o_key to userspace

Erspan protocol (version 1 and 2) relies on o_key to configure
session id header field. However TUNNEL_KEY bit is cleared in
erspan_xmit since ERSPAN protocol does not set the key field
of the external GRE header and so the configured o_key is not reported
to userspace. The issue can be triggered with the following reproducer:

$ip link add erspan1 type erspan local 192.168.0.1 remote 192.168.0.2 \
    key 1 seq erspan_ver 1
$ip link set erspan1 up
$ip -d link sh erspan1

erspan1@NONE: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc pfifo_fast state UNKNOWN mode DEFAULT
  link/ether 52:aa:99:95:9a:b5 brd ff:ff:ff:ff:ff:ff promiscuity 0 minmtu 68 maxmtu 1500
  erspan remote 192.168.0.2 local 192.168.0.1 ttl inherit ikey 0.0.0.1 iseq oseq erspan_index 0

Fix the issue adding TUNNEL_KEY bit to the o_flags parameter in
ipgre_fill_info

Fixes: 84e54fe0a5ea ("gre: introduce native tunnel support for ERSPAN")
Signed-off-by: Lorenzo Bianconi <lorenzo.bianconi@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ucc_geth: Reset BQL queue when stopping device

After a timeout event caused by for example a broadcast storm, when
the MAC and PHY are reset, the BQL TX queue needs to be reset as
well. Otherwise, the device will exhibit severe performance issues
even after the storm has ended.

Co-authored-by: David Gounaris <david.gounaris@infinera.com>
Signed-off-by: Mathias Thore <mathias.thore@infinera.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'net-various-compat-ioctl-fixes'

Johannes Berg says:

====================
various compat ioctl fixes

Back a long time ago, I already fixed a few of these by passing
the size of the struct ifreq to do_sock_ioctl(). However, Robert
found more cases, and now it won't be as simple because we'd have
to pass that down all the way to e.g. bond_do_ioctl() which isn't
really feasible.

Therefore, restore the old code.

While looking at why SIOCGIFNAME was broken, I realized that Al
had removed that case - which had been handled in an explicit
separate function - as well, and looking through his work at the
time I saw that bond ioctls were also affected by the erroneous
removal.

I've restored SIOCGIFNAME and bond ioctls by going through the
(now renamed) dev_ifsioc() instead of reintroducing their own
helper functions, which I hope is correct but have only tested
with SIOCGIFNAME.
====================

Acked-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: socket: make bond ioctls go through compat_ifreq_ioctl()

Same story as before, these use struct ifreq and thus need
to be read with the shorter version to not cause faults.

Cc: stable@vger.kernel.org
Fixes: f92d4fc95341 ("kill bond_ioctl()")
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: socket: fix SIOCGIFNAME in compat

As reported by Robert O'Callahan in
https://bugzilla.kernel.org/show_bug.cgi?id=202273
reverting the previous changes in this area broke
the SIOCGIFNAME ioctl in compat again (I'd previously
fixed it after his previous report of breakage in
https://bugzilla.kernel.org/show_bug.cgi?id=199469).

This is obviously because I fixed SIOCGIFNAME more or
less by accident.

Fix it explicitly now by making it pass through the
restored compat translation code.

Cc: stable@vger.kernel.org
Fixes: 4cf808e7ac32 ("kill dev_ifname32()")
Reported-by: Robert O'Callahan <robert@ocallahan.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Revert "kill dev_ifsioc()"

This reverts commit bf4405737f9f ("kill dev_ifsioc()").

This wasn't really unused as implied by the original commit,
it still handles the copy to/from user differently, and the
commit thus caused issues such as
  https://bugzilla.kernel.org/show_bug.cgi?id=199469
and
  https://bugzilla.kernel.org/show_bug.cgi?id=202273

However, deviating from a strict revert, rename dev_ifsioc()
to compat_ifreq_ioctl() to be clearer as to its purpose and
add a comment.

Cc: stable@vger.kernel.org
Fixes: bf4405737f9f ("kill dev_ifsioc()")
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Revert "socket: fix struct ifreq size in compat ioctl"

This reverts commit 1cebf8f143c2 ("socket: fix struct ifreq
size in compat ioctl"), it's a bugfix for another commit that
I'll revert next.

This is not a 'perfect' revert, I'm keeping some coding style
intact rather than revert to the state with indentation errors.

Cc: stable@vger.kernel.org
Fixes: 1cebf8f143c2 ("socket: fix struct ifreq size in compat ioctl")
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'typedef-func_proto'

Yonghong Song says:

====================
The current btf implementation disallows the typedef of
a func_proto type. This actually is allowed per C standard.
This patch fixed btf verification to permit such types.
Patch #1 fixed the kernel side and Patch #2 fixed
the tools test_btf test.
====================

Signed-off-by: Alexei Starovoitov <ast@kernel.org>

tools/bpf: fix test_btf for typedef func_proto case

Fixed one test_btf raw test such that typedef func_proto
is permitted now.

Fixes: 78a2540e8945 ("tools/bpf: Add tests for BTF_KIND_FUNC_PROTO and BTF_KIND_FUNC")
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

bpf: btf: allow typedef func_proto

Current implementation does not allow typedef func_proto.
But it is actually allowed.
  -bash-4.4$ cat t.c
  typedef int (f) (int);
  f *g;
  -bash-4.4$ clang -O2 -g -c -target bpf t.c -Xclang -target-feature -Xclang +dwarfris
  -bash-4.4$ pahole -JV t.o
  File t.o:
  [1] PTR (anon) type_id=2
  [2] TYPEDEF f type_id=3
  [3] FUNC_PROTO (anon) return=4 args=(4 (anon))
  [4] INT int size=4 bit_offset=0 nr_bits=32 encoding=SIGNED
  -bash-4.4$

This patch related btf verifier to allow such (typedef func_proto)
patterns.

Fixes: 2667a2626f4d ("bpf: btf: Add BTF_KIND_FUNC and BTF_KIND_FUNC_PROTO")
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) Need to save away the IV across tls async operations, from Dave
    Watson.

 2) Upon successful packet processing, we should liberate the SKB with
    dev_consume_skb{_irq}(). From Yang Wei.

 3) Only apply RX hang workaround on effected macb chips, from Harini
    Katakam.

 4) Dummy netdev need a proper namespace assigned to them, from Josh
    Elsasser.

 5) Some paths of nft_compat run lockless now, and thus we need to use a
    proper refcnt_t. From Florian Westphal.

 6) Avoid deadlock in mlx5 by doing IRQ locking, from Moni Shoua.

 7) netrom does not refcount sockets properly wrt. timers, fix that by
    using the sock timer API. From Cong Wang.

 8) Fix locking of inexact inserts of xfrm policies, from Florian
    Westphal.

 9) Missing xfrm hash generation bump, also from Florian.

10) Missing of_node_put() in hns driver, from Yonglong Liu.

11) Fix DN_IFREQ_SIZE, from Johannes Berg.

12) ip6mr notifier is invoked during traversal of wrong table, from Nir
    Dotan.

13) TX promisc settings not performed correctly in qed, from Manish
    Chopra.

14) Fix OOB access in vhost, from Jason Wang.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (52 commits)
  MAINTAINERS: Add entry for XDP (eXpress Data Path)
  net: set default network namespace in init_dummy_netdev()
  net: b44: replace dev_kfree_skb_xxx by dev_consume_skb_xxx for drop profiles
  net: caif: call dev_consume_skb_any when skb xmit done
  net: 8139cp: replace dev_kfree_skb_irq by dev_consume_skb_irq for drop profiles
  net: macb: Apply RXUBR workaround only to versions with errata
  net: ti: replace dev_kfree_skb_irq by dev_consume_skb_irq for drop profiles
  net: apple: replace dev_kfree_skb_irq by dev_consume_skb_irq for drop profiles
  net: amd8111e: replace dev_kfree_skb_irq by dev_consume_skb_irq
  net: alteon: replace dev_kfree_skb_irq by dev_consume_skb_irq
  net: tls: Fix deadlock in free_resources tx
  net: tls: Save iv in tls_rec for async crypto requests
  vhost: fix OOB in get_rx_bufs()
  qed: Fix stack out of bounds bug
  qed: Fix system crash in ll2 xmit
  qed: Fix VF probe failure while FLR
  qed: Fix LACP pdu drops for VFs
  qed: Fix bug in tx promiscuous mode settings
  net: i825xx: replace dev_kfree_skb_irq by dev_consume_skb_irq for drop profiles
  netfilter: ipt_CLUSTERIP: fix warning unused variable cn
  ...

MAINTAINERS: Add entry for XDP (eXpress Data Path)

Add multiple people as maintainers for XDP, sorted alphabetically.

XDP is also tied to driver level support and code, but we cannot add all
drivers to the list. Instead K: and N: match on 'xdp' in hope to catch some
of those changes in drivers.

Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: David S. Miller <davem@davemloft.net>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: set default network namespace in init_dummy_netdev()

Assign a default net namespace to netdevs created by init_dummy_netdev().
Fixes a NULL pointer dereference caused by busy-polling a socket bound to
an iwlwifi wireless device, which bumps the per-net BUSYPOLLRXPACKETS stat
if napi_poll() received packets:

  BUG: unable to handle kernel NULL pointer dereference at 0000000000000190
  IP: napi_busy_loop+0xd6/0x200
  Call Trace:
    sock_poll+0x5e/0x80
    do_sys_poll+0x324/0x5a0
    SyS_poll+0x6c/0xf0
    do_syscall_64+0x6b/0x1f0
    entry_SYSCALL_64_after_hwframe+0x3d/0xa2

Fixes: 7db6b048da3b ("net: Commonize busy polling code to focus on napi_id instead of socket")
Signed-off-by: Josh Elsasser <jelsasser@appneta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: b44: replace dev_kfree_skb_xxx by dev_consume_skb_xxx for drop profiles

The skb should be freed by dev_consume_skb_any() in b44_start_xmit()
when bounce_skb is used. The skb is be replaced by bounce_skb, so the
original skb should be consumed(not drop).

dev_consume_skb_irq() should be called in b44_tx() when skb xmit
done. It makes drop profiles(dropwatch, perf) more friendly.

Signed-off-by: Yang Wei <yang.wei9@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: caif: call dev_consume_skb_any when skb xmit done

The skb shouled be consumed when xmit done, it makes drop profiles
(dropwatch, perf) more friendly.
dev_kfree_skb_irq()/kfree_skb() shouled be replaced by
dev_consume_skb_any(), it makes code cleaner.

Signed-off-by: Yang Wei <yang.wei9@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: 8139cp: replace dev_kfree_skb_irq by dev_consume_skb_irq for drop profiles

dev_consume_skb_irq() should be called in cp_tx() when skb xmit
done. It makes drop profiles(dropwatch, perf) more friendly.

Signed-off-by: Yang Wei <yang.wei9@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: macb: Apply RXUBR workaround only to versions with errata

The interrupt handler contains a workaround for RX hang applicable
to Zynq and AT91RM9200 only. Subsequent versions do not need this
workaround. This workaround unnecessarily resets RX whenever RX used
bit read is observed, which can be often under heavy traffic. There
is no other action performed on RX UBR interrupt. Hence introduce a
CAPS mask; enable this interrupt and workaround only on affected
versions.

Signed-off-by: Harini Katakam <harini.katakam@xilinx.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: ti: replace dev_kfree_skb_irq by dev_consume_skb_irq for drop profiles

dev_consume_skb_irq() should be called in cpmac_end_xmit() when
xmit done. It makes drop profiles more friendly.

Signed-off-by: Yang Wei <yang.wei9@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: apple: replace dev_kfree_skb_irq by dev_consume_skb_irq for drop profiles

dev_consume_skb_irq() should be called in bmac_txdma_intr() when
xmit done. It makes drop profiles more friendly.

Signed-off-by: Yang Wei <yang.wei9@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: amd8111e: replace dev_kfree_skb_irq by dev_consume_skb_irq

dev_consume_skb_irq() should be called in amd8111e_tx() when xmit
done. It makes drop profiles more friendly.

Signed-off-by: Yang Wei <yang.wei9@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: alteon: replace dev_kfree_skb_irq by dev_consume_skb_irq

dev_consume_skb_irq() should be called in ace_tx_int() when xmit
done. It makes drop profiles more friendly.

Signed-off-by: Yang Wei <yang.wei9@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: tls: Fix deadlock in free_resources tx

If there are outstanding async tx requests (when crypto returns EINPROGRESS),
there is a potential deadlock: the tx work acquires the lock, while we
cancel_delayed_work_sync() while holding the lock.  Drop the lock while waiting
for the work to complete.

Fixes: a42055e8d2c30 ("Add support for async encryption of records...")
Signed-off-by: Dave Watson <davejwatson@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: tls: Save iv in tls_rec for async crypto requests

aead_request_set_crypt takes an iv pointer, and we change the iv
soon after setting it.  Some async crypto algorithms don't save the iv,
so we need to save it in the tls_rec for async requests.

Found by hardcoding x64 aesni to use async crypto manager (to test the async
codepath), however I don't think this combination can happen in the wild.
Presumably other hardware offloads will need this fix, but there have been
no user reports.

Fixes: a42055e8d2c30 ("Add support for async encryption of records...")
Signed-off-by: Dave Watson <davejwatson@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

vhost: fix OOB in get_rx_bufs()

After batched used ring updating was introduced in commit e2b3b35eb989
("vhost_net: batch used ring update in rx"). We tend to batch heads in
vq->heads for more than one packet. But the quota passed to
get_rx_bufs() was not correctly limited, which can result a OOB write
in vq->heads.

        headcount = get_rx_bufs(vq, vq->heads + nvq->done_idx,
                    vhost_len, &in, vq_log, &log,
                    likely(mergeable) ? UIO_MAXIOV : 1);

UIO_MAXIOV was still used which is wrong since we could have batched
used in vq->heads, this will cause OOB if the next buffer needs more
than 960 (1024 (UIO_MAXIOV) - 64 (VHOST_NET_BATCH)) heads after we've
batched 64 (VHOST_NET_BATCH) heads:
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
=============================================================================
BUG kmalloc-8k (Tainted: G    B            ): Redzone overwritten
-----------------------------------------------------------------------------

INFO: 0x00000000fd93b7a2-0x00000000f0713384. First byte 0xa9 instead of 0xcc
INFO: Allocated in alloc_pd+0x22/0x60 age=3933677 cpu=2 pid=2674
    kmem_cache_alloc_trace+0xbb/0x140
    alloc_pd+0x22/0x60
    gen8_ppgtt_create+0x11d/0x5f0
    i915_ppgtt_create+0x16/0x80
    i915_gem_create_context+0x248/0x390
    i915_gem_context_create_ioctl+0x4b/0xe0
    drm_ioctl_kernel+0xa5/0xf0
    drm_ioctl+0x2ed/0x3a0
    do_vfs_ioctl+0x9f/0x620
    ksys_ioctl+0x6b/0x80
    __x64_sys_ioctl+0x11/0x20
    do_syscall_64+0x43/0xf0
    entry_SYSCALL_64_after_hwframe+0x44/0xa9
INFO: Slab 0x00000000d13e87af objects=3 used=3 fp=0x          (null) flags=0x200000000010201
INFO: Object 0x0000000003278802 @offset=17064 fp=0x00000000e2e6652b

Fixing this by allocating UIO_MAXIOV + VHOST_NET_BATCH iovs for
vhost-net. This is done through set the limitation through
vhost_dev_init(), then set_owner can allocate the number of iov in a
per device manner.

This fixes CVE-2018-16880.

Fixes: e2b3b35eb989 ("vhost_net: batch used ring update in rx")
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bpf, doc: add reviewers to maintainers entry

In order to better scale BPF development on netdev, we've adopted a
reviewer rotation for all BPF patches among the five of us for some
time now. Lets give credit where credit is due, and add Martin, Song
and Yonghong as official BPF reviewers to MAINTAINERS file. Also
while at it, add regex matching for BPF such that we get properly
Cc'ed for files not listed here.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Acked-by: Yonghong Song <yhs@fb.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

sk_msg: Always cancel strp work before freeing the psock

Despite having stopped the parser, we still need to deinitialize it
by calling strp_done so that it cancels its work. Otherwise the worker
thread can run after we have freed the parser, and attempt to access
its workqueue resulting in a use-after-free:

==================================================================
BUG: KASAN: use-after-free in pwq_activate_delayed_work+0x1b/0x1d0
Read of size 8 at addr ffff888069975240 by task kworker/u2:2/93

CPU: 0 PID: 93 Comm: kworker/u2:2 Not tainted 5.0.0-rc2-00335-g28f9d1a3d4fe-dirty #14
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-2.fc27 04/01/2014
Workqueue:            (null) (kstrp)
Call Trace:
 print_address_description+0x6e/0x2b0
 ? pwq_activate_delayed_work+0x1b/0x1d0
 kasan_report+0xfd/0x177
 ? pwq_activate_delayed_work+0x1b/0x1d0
 ? pwq_activate_delayed_work+0x1b/0x1d0
 pwq_activate_delayed_work+0x1b/0x1d0
 ? process_one_work+0x4aa/0x660
 pwq_dec_nr_in_flight+0x9b/0x100
 worker_thread+0x82/0x680
 ? process_one_work+0x660/0x660
 kthread+0x1b9/0x1e0
 ? __kthread_create_on_node+0x250/0x250
 ret_from_fork+0x1f/0x30

Allocated by task 111:
 sk_psock_init+0x3c/0x1b0
 sock_map_link.isra.2+0x103/0x4b0
 sock_map_update_common+0x94/0x270
 sock_map_update_elem+0x145/0x160
 __se_sys_bpf+0x152e/0x1e10
 do_syscall_64+0xb2/0x3e0
 entry_SYSCALL_64_after_hwframe+0x44/0xa9

Freed by task 112:
 kfree+0x7f/0x140
 process_one_work+0x40b/0x660
 worker_thread+0x82/0x680
 kthread+0x1b9/0x1e0
 ret_from_fork+0x1f/0x30

The buggy address belongs to the object at ffff888069975180
 which belongs to the cache kmalloc-512 of size 512
The buggy address is located 192 bytes inside of
 512-byte region [ffff888069975180, ffff888069975380)
The buggy address belongs to the page:
page:ffffea0001a65d00 count:1 mapcount:0 mapping:ffff88806d401280 index:0x0 compound_mapcount: 0
flags: 0x4000000000010200(slab|head)
raw: 4000000000010200 dead000000000100 dead000000000200 ffff88806d401280
raw: 0000000000000000 00000000800c000c 00000001ffffffff 0000000000000000
page dumped because: kasan: bad access detected

Memory state around the buggy address:
 ffff888069975100: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
 ffff888069975180: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
>ffff888069975200: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
                                           ^
 ffff888069975280: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
 ffff888069975300: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
==================================================================

Reported-by: Marek Majkowski <marek@cloudflare.com>
Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com>
Link: https://lore.kernel.org/netdev/CAJPywTLwgXNEZ2dZVoa=udiZmtrWJ0q5SuBW64aYs0Y1khXX3A@mail.gmail.com
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

tools: bpftool: fix crash with un-owned prog arrays

Prog arrays don't have 'owner_prog_type' and 'owner_jited'
fields in their fdinfo when they are created.  Those fields
are set and reported when first program is checked for
compatibility by bpf_prog_array_compatible().

This means that bpftool cannot expect the fields to always
be there.  Currently trying to show maps on a system with
an un-owned prog array leads to a crash:

$ bpftool map show
389: prog_array  name tail_call_map  flags 0x0
Error: key 'owner_prog_type' not found in fdinfo
Error: key 'owner_jited' not found in fdinfo
       key 4B  value 4B  max_entries 4  memlock 4096B
       Segmentation fault (core dumped)

We pass a NULL pointer to atoi().

Remove the assumption that fdinfo keys are always present.
Add missing validations and remove the p_err() calls which
may lead to broken JSON output as caller will not propagate
the failure.

Fixes: 99a44bef5870 ("tools: bpftool: add owner_prog_type and owner_jited to bpftool output")
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Merge branch 'qed-Bug-fixes'

Manish Chopra says:

====================
qed: Bug fixes

This series have SR-IOV and some general fixes.
Please consider applying it to "net"
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

qed: Fix stack out of bounds bug

KASAN reported following bug in qed_init_qm_get_idx_from_flags
due to inappropriate casting of "pq_flags". Fix the type of "pq_flags".

[  196.624707] BUG: KASAN: stack-out-of-bounds in qed_init_qm_get_idx_from_flags+0x1a4/0x1b8 [qed]
[  196.624712] Read of size 8 at addr ffff809b00bc7360 by task kworker/0:9/1712
[  196.624714]
[  196.624720] CPU: 0 PID: 1712 Comm: kworker/0:9 Not tainted 4.18.0-60.el8.aarch64+debug #1
[  196.624723] Hardware name: To be filled by O.E.M. Saber/Saber, BIOS 0ACKL024 09/26/2018
[  196.624733] Workqueue: events work_for_cpu_fn
[  196.624738] Call trace:
[  196.624742]  dump_backtrace+0x0/0x2f8
[  196.624745]  show_stack+0x24/0x30
[  196.624749]  dump_stack+0xe0/0x11c
[  196.624755]  print_address_description+0x68/0x260
[  196.624759]  kasan_report+0x178/0x340
[  196.624762]  __asan_report_load_n_noabort+0x38/0x48
[  196.624786]  qed_init_qm_get_idx_from_flags+0x1a4/0x1b8 [qed]
[  196.624808]  qed_init_qm_info+0xec0/0x2200 [qed]
[  196.624830]  qed_resc_alloc+0x284/0x7e8 [qed]
[  196.624853]  qed_slowpath_start+0x6cc/0x1ae8 [qed]
[  196.624864]  __qede_probe.isra.10+0x1cc/0x12c0 [qede]
[  196.624874]  qede_probe+0x78/0xf0 [qede]
[  196.624879]  local_pci_probe+0xc4/0x180
[  196.624882]  work_for_cpu_fn+0x54/0x98
[  196.624885]  process_one_work+0x758/0x1900
[  196.624888]  worker_thread+0x4e0/0xd18
[  196.624892]  kthread+0x2c8/0x350
[  196.624897]  ret_from_fork+0x10/0x18
[  196.624899]
[  196.624902] Allocated by task 2:
[  196.624906]  kasan_kmalloc.part.1+0x40/0x108
[  196.624909]  kasan_kmalloc+0xb4/0xc8
[  196.624913]  kasan_slab_alloc+0x14/0x20
[  196.624916]  kmem_cache_alloc_node+0x1dc/0x480
[  196.624921]  copy_process.isra.1.part.2+0x1d8/0x4a98
[  196.624924]  _do_fork+0x150/0xfa0
[  196.624926]  kernel_thread+0x48/0x58
[  196.624930]  kthreadd+0x3a4/0x5a0
[  196.624932]  ret_from_fork+0x10/0x18
[  196.624934]
[  196.624937] Freed by task 0:
[  196.624938] (stack is not available)
[  196.624940]
[  196.624943] The buggy address belongs to the object at ffff809b00bc0000
[  196.624943]  which belongs to the cache thread_stack of size 32768
[  196.624946] The buggy address is located 29536 bytes inside of
[  196.624946]  32768-byte region [ffff809b00bc0000, ffff809b00bc8000)
[  196.624948] The buggy address belongs to the page:
[  196.624952] page:ffff7fe026c02e00 count:1 mapcount:0 mapping:ffff809b4001c000 index:0x0 compound_mapcount: 0
[  196.624960] flags: 0xfffff8000008100(slab|head)
[  196.624967] raw: 0fffff8000008100 dead000000000100 dead000000000200 ffff809b4001c000
[  196.624970] raw: 0000000000000000 0000000000080008 00000001ffffffff 0000000000000000
[  196.624973] page dumped because: kasan: bad access detected
[  196.624974]
[  196.624976] Memory state around the buggy address:
[  196.624980]  ffff809b00bc7200: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
[  196.624983]  ffff809b00bc7280: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
[  196.624985] >ffff809b00bc7300: 00 00 00 00 00 00 00 00 f1 f1 f1 f1 04 f2 f2 f2
[  196.624988]                                                        ^
[  196.624990]  ffff809b00bc7380: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
[  196.624993]  ffff809b00bc7400: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
[  196.624995] ==================================================================

Signed-off-by: Manish Chopra <manishc@marvell.com>
Signed-off-by: Ariel Elior <aelior@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qed: Fix system crash in ll2 xmit

Cache number of fragments in the skb locally as in case
of linear skb (with zero fragments), tx completion
(or freeing of skb) may happen before driver tries
to get number of frgaments from the skb which could
lead to stale access to an already freed skb.

Signed-off-by: Manish Chopra <manishc@marvell.com>
Signed-off-by: Ariel Elior <aelior@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qed: Fix VF probe failure while FLR

VFs may hit VF-PF channel timeout while probing, as in some
cases it was observed that VF FLR and VF "acquire" message
transaction (i.e first message from VF to PF in VF's probe flow)
could occur simultaneously which could lead VF to fail sending
"acquire" message to PF as VF is marked disabled from HW perspective
due to FLR, which will result into channel timeout and VF probe failure.

In such cases, try retrying VF "acquire" message so that in later
attempts it could be successful to pass message to PF after the VF
FLR is completed and can be probed successfully.

Signed-off-by: Manish Chopra <manishc@marvell.com>
Signed-off-by: Ariel Elior <aelior@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qed: Fix LACP pdu drops for VFs

VF is always configured to drop control frames
(with reserved mac addresses) but to work LACP
on the VFs, it would require LACP control frames
to be forwarded or transmitted successfully.

This patch fixes this in such a way that trusted VFs
(marked through ndo_set_vf_trust) would be allowed to
pass the control frames such as LACP pdus.

Signed-off-by: Manish Chopra <manishc@marvell.com>
Signed-off-by: Ariel Elior <aelior@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qed: Fix bug in tx promiscuous mode settings

When running tx switched traffic between VNICs
created via a bridge(to which VFs are added),
adapter drops the unicast packets in tx flow due to
VNIC's ucast mac being unknown to it. But VF interfaces
being in promiscuous mode should have caused adapter
to accept all the unknown ucast packets. Later, it
was found that driver doesn't really configure tx
promiscuous mode settings to accept all unknown unicast macs.

This patch fixes tx promiscuous mode settings to accept all
unknown/unmatched unicast macs and works out the scenario.

Signed-off-by: Manish Chopra <manishc@marvell.com>
Signed-off-by: Ariel Elior <aelior@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: i825xx: replace dev_kfree_skb_irq by dev_consume_skb_irq for drop profiles

dev_consume_skb_irq() should be called in i596_interrupt() when skb
xmit done. It makes drop profiles(dropwatch, perf) more friendly.

Signed-off-by: Yang Wei <albin_yang@163.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge git://git./pub/scm/linux/kernel/git/pablo/nf

Pablo Neira Ayuso says:

====================
Netfilter/IPVS fixes for net

The following patchset contains Netfilter/IPVS fixes for your net tree:

1) The nftnl mutex is now per-netns, therefore use reference counter
   for matches and targets to deal with concurrent updates from netns.
   Moreover, place extensions in a pernet list. Patches from Florian Westphal.

2) Bail out with EINVAL in case of negative timeouts via setsockopt()
   through ip_vs_set_timeout(), from ZhangXiaoxu.

3) Spurious EINVAL on ebtables 32bit binary with 64bit kernel, also
   from Florian.

4) Reset TCP option header parser in case of fingerprint mismatch,
   otherwise follow up overlapping fingerprint definitions including
   TCP options do not work, from Fernando Fernandez Mancera.

5) Compilation warning in ipt_CLUSTER with CONFIG_PROC_FS unset.
   From Anders Roxell.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

Revert "mm, memory_hotplug: initialize struct pages for the full memory section"

This reverts commit 2830bf6f05fb3e05bc4743274b806c821807a684.

The underlying assumption that one sparse section belongs into a single
numa node doesn't hold really. Robert Shteynfeld has reported a boot
failure. The boot log was not captured but his memory layout is as
follows:

  Early memory node ranges
    node   1: [mem 0x0000000000001000-0x0000000000090fff]
    node   1: [mem 0x0000000000100000-0x00000000dbdf8fff]
    node   1: [mem 0x0000000100000000-0x0000001423ffffff]
    node   0: [mem 0x0000001424000000-0x0000002023ffffff]

This means that node0 starts in the middle of a memory section which is
also in node1.  memmap_init_zone tries to initialize padding of a
section even when it is outside of the given pfn range because there are
code paths (e.g.  memory hotplug) which assume that the full worth of
memory section is always initialized.

In this particular case, though, such a range is already intialized and
most likely already managed by the page allocator.  Scribbling over
those pages corrupts the internal state and likely blows up when any of
those pages gets used.

Reported-by: Robert Shteynfeld <robert.shteynfeld@gmail.com>
Fixes: 2830bf6f05fb ("mm, memory_hotplug: initialize struct pages for the full memory section")
Cc: stable@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

netfilter: ipt_CLUSTERIP: fix warning unused variable cn

When CONFIG_PROC_FS isn't set the variable cn isn't used.

net/ipv4/netfilter/ipt_CLUSTERIP.c: In function ‘clusterip_net_exit’:
net/ipv4/netfilter/ipt_CLUSTERIP.c:849:24: warning: unused variable ‘cn’ [-Wunused-variable]
  struct clusterip_net *cn = clusterip_pernet(net);
                        ^~

Rework so the variable 'cn' is declared inside "#ifdef CONFIG_PROC_FS".

Fixes: b12f7bad5ad3 ("netfilter: ipt_CLUSTERIP: remove wrong WARN_ON_ONCE in netns exit routine")
Signed-off-by: Anders Roxell <anders.roxell@linaro.org>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

netfilter: nfnetlink_osf: add missing fmatch check

When we check the tcp options of a packet and it doesn't match the current
fingerprint, the tcp packet option pointer must be restored to its initial
value in order to do the proper tcp options check for the next fingerprint.

Here we can see an example.
Assumming the following fingerprint base with two lines:

S10:64:1:60:M*,S,T,N,W6:      Linux:3.0::Linux 3.0
S20:64:1:60:M*,S,T,N,W7:      Linux:4.19:arch:Linux 4.1

Where TCP options are the last field in the OS signature, all of them overlap
except by the last one, ie. 'W6' versus 'W7'.

In case a packet for Linux 4.19 kicks in, the osf finds no matching because the
TCP options pointer is updated after checking for the TCP options in the first
line.

Therefore, reset pointer back to where it should be.

Fixes: 11eeef41d5f6 ("netfilter: passive OS fingerprint xtables match")
Signed-off-by: Fernando Fernandez Mancera <ffmancera@riseup.net>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

netfilter: ebtables: compat: un-break 32bit setsockopt when no rules are present

Unlike ip(6)tables ebtables only counts user-defined chains.

The effect is that a 32bit ebtables binary on a 64bit kernel can do
'ebtables -N FOO' only after adding at least one rule, else the request
fails with -EINVAL.

This is a similar fix as done in
3f1e53abff84 ("netfilter: ebtables: don't attempt to allocate 0-sized compat array").

Fixes: 7d7d7e02111e9 ("netfilter: compat: reject huge allocation requests")
Reported-by: Francesco Ruggeri <fruggeri@arista.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

net: dsa: mv88e6xxx: Fix serdes irq setup going recursive

Duec to a typo, mv88e6390_serdes_irq_setup() calls itself, rather than
mv88e6390x_serdes_irq_setup(). It then blows the stack, and shortly
after the machine blows up.

Fixes: 2defda1f4b91 ("net: dsa: mv88e6xxx: Add support for SERDES on ports 2-8 for 6390X")
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>