linux-2.6-microblaze.git
6 years agosctp: not allow transport timeout value less than HZ/5 for hb_timer
Xin Long [Tue, 5 Jun 2018 04:16:58 +0000 (12:16 +0800)]
sctp: not allow transport timeout value less than HZ/5 for hb_timer

syzbot reported a rcu_sched self-detected stall on CPU which is caused
by too small value set on rto_min with SCTP_RTOINFO sockopt. With this
value, hb_timer will get stuck there, as in its timer handler it starts
this timer again with this value, then goes to the timer handler again.

This problem is there since very beginning, and thanks to Eric for the
reproducer shared from a syzbot mail.

This patch fixes it by not allowing sctp_transport_timeout to return a
smaller value than HZ/5 for hb_timer, which is based on TCP's min rto.

Note that it doesn't fix this issue by limiting rto_min, as some users
are still using small rto and no proper value was found for it yet.

Reported-by: syzbot+3dcd59a1f907245f891f@syzkaller.appspotmail.com
Suggested-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agobpfilter: switch to CC from HOSTCC
Alexei Starovoitov [Tue, 5 Jun 2018 02:53:41 +0000 (19:53 -0700)]
bpfilter: switch to CC from HOSTCC

check that CC can build executables and use that compiler instead of HOSTCC

Suggested-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet/mlx5e: fix error return code in mlx5e_alloc_rq()
Wei Yongjun [Tue, 5 Jun 2018 02:42:56 +0000 (02:42 +0000)]
net/mlx5e: fix error return code in mlx5e_alloc_rq()

Fix to return error code -ENOMEM from the kvzalloc_node() error handling
case instead of 0, as done elsewhere in this function.

Fixes: 069d11465a80 ("net/mlx5e: RX, Enhance legacy Receive Queue memory scheme")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet/mlx5e: Make function mlx5e_change_rep_mtu() static
Wei Yongjun [Tue, 5 Jun 2018 02:42:45 +0000 (02:42 +0000)]
net/mlx5e: Make function mlx5e_change_rep_mtu() static

Fixes the following sparse warning:

drivers/net/ethernet/mellanox/mlx5/core/en_rep.c:903:5: warning:
 symbol 'mlx5e_change_rep_mtu' was not declared. Should it be static?

Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: qualcomm: rmnet: Fix use after free while sending command ack
Subash Abhinov Kasiviswanathan [Tue, 5 Jun 2018 01:43:38 +0000 (19:43 -0600)]
net: qualcomm: rmnet: Fix use after free while sending command ack

When sending an ack to a command packet, the skb is still referenced
after it is sent to the real device. Since the real device could
free the skb, the device pointer would be invalid.
Also, remove an unnecessary variable.

Fixes: ceed73a2cf4a ("drivers: net: ethernet: qualcomm: rmnet: Initial implementation")
Signed-off-by: Subash Abhinov Kasiviswanathan <subashab@codeaurora.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: ipv6: Generate random IID for addresses on RAWIP devices
Subash Abhinov Kasiviswanathan [Tue, 5 Jun 2018 01:26:07 +0000 (19:26 -0600)]
net: ipv6: Generate random IID for addresses on RAWIP devices

RAWIP devices such as rmnet do not have a hardware address and
instead require the kernel to generate a random IID for the
IPv6 addresses.

Signed-off-by: Sean Tranchetti <stranche@codeaurora.org>
Signed-off-by: Subash Abhinov Kasiviswanathan <subashab@codeaurora.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotcp: refactor tcp_ecn_check_ce to remove sk type cast
Yousuk Seung [Mon, 4 Jun 2018 22:29:51 +0000 (15:29 -0700)]
tcp: refactor tcp_ecn_check_ce to remove sk type cast

Refactor tcp_ecn_check_ce and __tcp_ecn_check_ce to accept struct sock*
instead of tcp_sock* to clean up type casts. This is a pure refactor
patch.

Signed-off-by: Yousuk Seung <ysseung@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet/ipv6: prevent use after free in ip6_route_mpath_notify
David Ahern [Mon, 4 Jun 2018 20:41:42 +0000 (13:41 -0700)]
net/ipv6: prevent use after free in ip6_route_mpath_notify

syzbot reported a use-after-free:

BUG: KASAN: use-after-free in ip6_route_mpath_notify+0xe9/0x100 net/ipv6/route.c:4180
Read of size 4 at addr ffff8801bf789cf0 by task syz-executor756/4555

CPU: 1 PID: 4555 Comm: syz-executor756 Not tainted 4.17.0-rc7+ #78
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+0x1b9/0x294 lib/dump_stack.c:113
 print_address_description+0x6c/0x20b mm/kasan/report.c:256
 kasan_report_error mm/kasan/report.c:354 [inline]
 kasan_report.cold.7+0x242/0x2fe mm/kasan/report.c:412
 __asan_report_load4_noabort+0x14/0x20 mm/kasan/report.c:432
 ip6_route_mpath_notify+0xe9/0x100 net/ipv6/route.c:4180
 ip6_route_multipath_add+0x615/0x1910 net/ipv6/route.c:4303
 inet6_rtm_newroute+0xe3/0x160 net/ipv6/route.c:4391
 ...

Allocated by task 4555:
 save_stack+0x43/0xd0 mm/kasan/kasan.c:448
 set_track mm/kasan/kasan.c:460 [inline]
 kasan_kmalloc+0xc4/0xe0 mm/kasan/kasan.c:553
 kasan_slab_alloc+0x12/0x20 mm/kasan/kasan.c:490
 kmem_cache_alloc+0x12e/0x760 mm/slab.c:3554
 dst_alloc+0xbb/0x1d0 net/core/dst.c:104
 __ip6_dst_alloc+0x35/0xa0 net/ipv6/route.c:361
 ip6_dst_alloc+0x29/0xb0 net/ipv6/route.c:376
 ip6_route_info_create+0x4d4/0x3a30 net/ipv6/route.c:2834
 ip6_route_multipath_add+0xc7e/0x1910 net/ipv6/route.c:4240
 inet6_rtm_newroute+0xe3/0x160 net/ipv6/route.c:4391
 ...

Freed by task 4555:
 save_stack+0x43/0xd0 mm/kasan/kasan.c:448
 set_track mm/kasan/kasan.c:460 [inline]
 __kasan_slab_free+0x11a/0x170 mm/kasan/kasan.c:521
 kasan_slab_free+0xe/0x10 mm/kasan/kasan.c:528
 __cache_free mm/slab.c:3498 [inline]
 kmem_cache_free+0x86/0x2d0 mm/slab.c:3756
 dst_destroy+0x267/0x3c0 net/core/dst.c:140
 dst_release_immediate+0x71/0x9e net/core/dst.c:205
 fib6_add+0xa40/0x1650 net/ipv6/ip6_fib.c:1305
 __ip6_ins_rt+0x6c/0x90 net/ipv6/route.c:1011
 ip6_route_multipath_add+0x513/0x1910 net/ipv6/route.c:4267
 inet6_rtm_newroute+0xe3/0x160 net/ipv6/route.c:4391
 ...

The problem is that rt_last can point to a deleted route if the insert
fails.

One reproducer is to insert a route and then add a multipath route that
has a duplicate nexthop.e.g,:
    $ ip -6 ro add vrf red 2001:db8:101::/64 nexthop via 2001:db8:1::2
    $ ip -6 ro append vrf red 2001:db8:101::/64 nexthop via 2001:db8:1::4 nexthop via 2001:db8:1::2

Fix by not setting rt_last until the it is verified the insert succeeded.

Fixes: 3b1137fe7482 ("net: ipv6: Change notifications for multipath add to RTA_MULTIPATH")
Cc: Eric Dumazet <edumazet@google.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: phy: broadcom: Enable 125 MHz clock on LED4 pin for BCM54612E by default.
Kun Yi [Mon, 4 Jun 2018 20:17:04 +0000 (13:17 -0700)]
net: phy: broadcom: Enable 125 MHz clock on LED4 pin for BCM54612E by default.

BCM54612E have 4 multi-functional LED pins that can be configured
through register setting; the LED4 pin can be configured to a 125MHz
reference clock output by setting the spare register. Since the dedicated
CLK125 reference clock pin is not brought out on the 48-Pin MLP, the LED4
pin is the only pin to provide such function in this package, and therefore
it is beneficial to just enable the reference clock by default.

Signed-off-by: Kun Yi <kunyi@google.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agol2tp: fix refcount leakage on PPPoL2TP sockets
Guillaume Nault [Mon, 4 Jun 2018 16:52:19 +0000 (18:52 +0200)]
l2tp: fix refcount leakage on PPPoL2TP sockets

Commit d02ba2a6110c ("l2tp: fix race in pppol2tp_release with session
object destroy") tried to fix a race condition where a PPPoL2TP socket
would disappear while the L2TP session was still using it. However, it
missed the root issue which is that an L2TP session may accept to be
reconnected if its associated socket has entered the release process.

The tentative fix makes the session hold the socket it is connected to.
That saves the kernel from crashing, but introduces refcount leakage,
preventing the socket from completing the release process. Once stalled,
everything the socket depends on can't be released anymore, including
the L2TP session and the l2tp_ppp module.

The root issue is that, when releasing a connected PPPoL2TP socket, the
session's ->sk pointer (RCU-protected) is reset to NULL and we have to
wait for a grace period before destroying the socket. The socket drops
the session in its ->sk_destruct callback function, so the session
will exist until the last reference on the socket is dropped.
Therefore, there is a time frame where pppol2tp_connect() may accept
reconnecting a session, as it only checks ->sk to figure out if the
session is connected. This time frame is shortened by the fact that
pppol2tp_release() calls l2tp_session_delete(), making the session
unreachable before resetting ->sk. However, pppol2tp_connect() may
grab the session before it gets unhashed by l2tp_session_delete(), but
it may test ->sk after the later got reset. The race is not so hard to
trigger and syzbot found a pretty reliable reproducer:
https://syzkaller.appspot.com/bug?id=418578d2a4389074524e04d641eacb091961b2cf

Before d02ba2a6110c, another race could let pppol2tp_release()
overwrite the ->__sk pointer of an L2TP session, thus tricking
pppol2tp_put_sk() into calling sock_put() on a socket that is different
than the one for which pppol2tp_release() was originally called. To get
there, we had to trigger the race described above, therefore having one
PPPoL2TP socket being released, while the session it is connected to is
reconnecting to a different PPPoL2TP socket. When releasing this new
socket fast enough, pppol2tp_release() overwrites the session's
->__sk pointer with the address of the new socket, before the first
pppol2tp_put_sk() call gets scheduled. Then the pppol2tp_put_sk() call
invoked by the original socket will sock_put() the new socket,
potentially dropping its last reference. When the second
pppol2tp_put_sk() finally runs, its socket has already been freed.

With d02ba2a6110c, the session takes a reference on both sockets.
Furthermore, the session's ->sk pointer is reset in the
pppol2tp_session_close() callback function rather than in
pppol2tp_release(). Therefore, ->__sk can't be overwritten and
pppol2tp_put_sk() is called only once (l2tp_session_delete() will only
run pppol2tp_session_close() once, to protect the session against
concurrent deletion requests). Now pppol2tp_put_sk() will properly
sock_put() the original socket, but the new socket will remain, as
l2tp_session_delete() prevented the release process from completing.
Here, we don't depend on the ->__sk race to trigger the bug. Getting
into the pppol2tp_connect() race is enough to leak the reference, no
matter when new socket is released.

So it all boils down to pppol2tp_connect() failing to realise that the
session has already been connected. This patch drops the unneeded extra
reference counting (mostly reverting d02ba2a6110c) and checks that
neither ->sk nor ->__sk is set before allowing a session to be
connected.

Fixes: d02ba2a6110c ("l2tp: fix race in pppol2tp_release with session object destroy")
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'net-phy-improve-PM-handling-of-PHY-MDIO'
David S. Miller [Tue, 5 Jun 2018 12:50:18 +0000 (08:50 -0400)]
Merge branch 'net-phy-improve-PM-handling-of-PHY-MDIO'

Heiner Kallweit says:

====================
net: phy: improve PM handling of PHY/MDIO

Current implementation of MDIO bus PM ops doesn't actually implement
bus-specific PM ops but just calls PM ops defined on a device level
what doesn't seem to be fully in line with the core PM model.

When looking e.g. at __device_suspend() the PM core looks for PM ops
of a device in a specific order:
1. device PM domain
2. device type
3. device class
4. device bus

I think it has good reason that there's no PM ops on device level.
The situation can be improved by modeling PHY's as device type of
a MDIO device. If for some other type of MDIO device PM ops are
needed, it could be modeled as struct device_type as well.
====================

Tested-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: phy: remove PM ops from MDIO bus
Heiner Kallweit [Sat, 2 Jun 2018 20:37:24 +0000 (22:37 +0200)]
net: phy: remove PM ops from MDIO bus

Current implementation of MDIO bus PM ops doesn't actually implement
bus-specific PM ops but just calls PM ops defined on a device level
what doesn't seem to be fully in line with the core PM model.

When looking e.g. at __device_suspend() the PM core looks for PM ops
of a device in a specific order:
1. device PM domain
2. device type
3. device class
4. device bus

I think it has good reason that there's no PM ops on device level.

Now that a device type representation of PHY's as special type of MDIO
devices was added (only user of MDIO bus PM ops), the MDIO bus
PM ops can be removed including member pm of struct mdio_device.

If for some other type of MDIO device PM ops are needed, it should be
modeled as struct device_type as well.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: phy: add struct device_type representation of a PHY
Heiner Kallweit [Sat, 2 Jun 2018 20:36:06 +0000 (22:36 +0200)]
net: phy: add struct device_type representation of a PHY

A PHY is a type of MDIO device, so let's model it as struct device_type
and place PM ops, attribute groups and release callback on device type
level. For this the attribute definitions have to be moved.
This change allows us to get rid of the PM ops on a bus level in a second
step.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
David S. Miller [Tue, 5 Jun 2018 02:23:35 +0000 (22:23 -0400)]
Merge git://git./linux/kernel/git/davem/net

6 years agoMerge branch '10GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next...
David S. Miller [Mon, 4 Jun 2018 21:35:35 +0000 (17:35 -0400)]
Merge branch '10GbE' of git://git./linux/kernel/git/jkirsher/next-queue

Jeff Kirsher says:

====================
Intel Wired LAN Driver Updates 2018-06-04

This series contains a smorgasbord of updates to documentation, e1000e,
igb, ixgbe, ixgbevf and i40e.

Benjamin Poirier fixes a potential kernel crash due to NULL pointer
dereference in e1000e.

Jeff updates the kernel documentation for e100 and e1000 to correct
default values and URLs which were incorrect in the documentation.  Also
took the time to update these to the new reStructured text format for
kernel documentation.

Joanna Yurdal fixes a missing PTP transmit timestamp by ensuring that
TSICR gets cleared when ICR is cleared.

Sergey updates igb to reset all the transmit queues at one time so that
we only have to wait once for all the queues to be reset.

Alex fixes ixgbevf so that malicious driver detection (MDD) can co-exist
with XDP.

Emil and Tony extend the RTNL lock to ensure we get the most up-to-date
values for the bits and avoid a possible race condition when going down.

YueHaibing from Huawei introduces a helper function in ixgbe for
operation reads to simplify the code a bit more.

Daniel Borkmann adds support for XDP meta data when using build SKB
for i40e.

Shannon Nelson provides twp fixes for the IPSec code in ixgbe, first is
to make sure we do not try to offload the decryption of any incoming
packet that is destined for the management engine.  The other fix is to
resolve a cast problem introduced by a sparse cleanup patch.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: hns: Fix the process of adding broadcast addresses to tcam
Xi Wang [Mon, 4 Jun 2018 16:50:09 +0000 (17:50 +0100)]
net: hns: Fix the process of adding broadcast addresses to tcam

If the multicast mask value in device tree is configured not all
0xff, the broadcast mac will be lost from tcam table after the
execution of command 'ifconfig up'. The address is appended by
hns_ae_start, but will be clear later by hns_nic_set_rx_mode
called in dev_open process.

This patch fixed it by not use the multicast mask when add a
broadcast address.

Fixes: b5996f11ea54 ("net: add Hisilicon Network Subsystem basic ethernet support")
Signed-off-by: Xi Wang <wangxi11@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: sched: return error code when tcf proto is not found
Vlad Buslov [Mon, 4 Jun 2018 15:32:23 +0000 (18:32 +0300)]
net: sched: return error code when tcf proto is not found

If requested tcf proto is not found, get and del filter netlink protocol
handlers output error message to extack, but do not return actual error
code. Add check to return ENOENT when result of tp find function is NULL
pointer.

Fixes: c431f89b18a2 ("net: sched: split tc_ctl_tfilter into three handlers")
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoteam: use netdev_features_t instead of u32
Dan Carpenter [Mon, 4 Jun 2018 14:46:01 +0000 (17:46 +0300)]
team: use netdev_features_t instead of u32

This code was introduced in 2011 around the same time that we made
netdev_features_t a u64 type.  These days a u32 is not big enough to
hold all the potential features.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet_failover: Use netdev_features_t instead of u32
Dan Carpenter [Mon, 4 Jun 2018 14:43:21 +0000 (17:43 +0300)]
net_failover: Use netdev_features_t instead of u32

The features mask needs to be a netdev_features_t (u64) because a u32
is not big enough.

Fixes: cfc80d9a1163 ("net: Introduce net_failover driver")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoqed: use dma_zalloc_coherent instead of allocator/memset
YueHaibing [Mon, 4 Jun 2018 13:10:31 +0000 (21:10 +0800)]
qed: use dma_zalloc_coherent instead of allocator/memset

Use dma_zalloc_coherent instead of dma_alloc_coherent
followed by memset 0.

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Acked-by: Tomer Tayar <Tomer.Tayar@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agowan/fsl_ucc_hdlc: use dma_zalloc_coherent instead of allocator/memset
YueHaibing [Mon, 4 Jun 2018 13:07:59 +0000 (21:07 +0800)]
wan/fsl_ucc_hdlc: use dma_zalloc_coherent instead of allocator/memset

Use dma_zalloc_coherent instead of dma_alloc_coherent
followed by memset 0.

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'for-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/bluetoot...
David S. Miller [Mon, 4 Jun 2018 21:22:17 +0000 (17:22 -0400)]
Merge branch 'for-upstream' of git://git./linux/kernel/git/bluetooth/bluetooth-next

Johan Hedberg says:

====================
pull request: bluetooth-next 2018-06-04

Here's one last bluetooth-next pull request for the 4.18 kernel:

 - New USB device IDs for Realtek 8822BE and 8723DE
 - reset/resume fix for Dell Inspiron 5565
 - Fix HCI_UART_INIT_PENDING flag behavior
 - Fix patching behavior for some ATH3012 models
 - A few other minor cleanups & fixes

Please let me know if there are any issues pulling. Thanks.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agodocs: networking: fix minor typos in various documentation files
Olivier Gayot [Mon, 4 Jun 2018 10:07:37 +0000 (12:07 +0200)]
docs: networking: fix minor typos in various documentation files

This patch fixes some typos/misspelling errors in the
Documentation/networking files.

Signed-off-by: Olivier Gayot <olivier.gayot@sigexec.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: do not allow changing SO_REUSEADDR/SO_REUSEPORT on bound sockets
Maciej Żenczykowski [Sun, 3 Jun 2018 17:47:05 +0000 (10:47 -0700)]
net: do not allow changing SO_REUSEADDR/SO_REUSEPORT on bound sockets

It is not safe to do so because such sockets are already in the
hash tables and changing these options can result in invalidating
the tb->fastreuse(port) caching.

This can have later far reaching consequences wrt. bind conflict checks
which rely on these caches (for optimization purposes).

Not to mention that you can currently end up with two identical
non-reuseport listening sockets bound to the same local ip:port
by clearing reuseport on them after they've already both been bound.

There is unfortunately no EISBOUND error or anything similar,
and EISCONN seems to be misleading for a bound-but-not-connected
socket, so use EUCLEAN 'Structure needs cleaning' which AFAICT
is the closest you can get to meaning 'socket in bad state'.
(although perhaps EINVAL wouldn't be a bad choice either?)

This does unfortunately run the risk of breaking buggy
userspace programs...

Signed-off-by: Maciej Żenczykowski <maze@google.com>
Cc: Eric Dumazet <edumazet@google.com>
Change-Id: I77c2b3429b2fdf42671eee0fa7a8ba721c94963b
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet-tcp: extend tcp_tw_reuse sysctl to enable loopback only optimization
Maciej Żenczykowski [Sun, 3 Jun 2018 17:41:17 +0000 (10:41 -0700)]
net-tcp: extend tcp_tw_reuse sysctl to enable loopback only optimization

This changes the /proc/sys/net/ipv4/tcp_tw_reuse from a boolean
to an integer.

It now takes the values 0, 1 and 2, where 0 and 1 behave as before,
while 2 enables timewait socket reuse only for sockets that we can
prove are loopback connections:
  ie. bound to 'lo' interface or where one of source or destination
  IPs is 127.0.0.0/8, ::ffff:127.0.0.0/104 or ::1.

This enables quicker reuse of ephemeral ports for loopback connections
- where tcp_tw_reuse is 100% safe from a protocol perspective
(this assumes no artificially induced packet loss on 'lo').

This also makes estblishing many loopback connections *much* faster
(allocating ports out of the first half of the ephemeral port range
is significantly faster, then allocating from the second half)

Without this change in a 32K ephemeral port space my sample program
(it just establishes and closes [::1]:ephemeral -> [::1]:server_port
connections in a tight loop) fails after 32765 connections in 24 seconds.
With it enabled 50000 connections only take 4.7 seconds.

This is particularly problematic for IPv6 where we only have one local
address and cannot play tricks with varying source IP from 127.0.0.0/8
pool.

Signed-off-by: Maciej Żenczykowski <maze@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Wei Wang <weiwan@google.com>
Change-Id: I0377961749979d0301b7b62871a32a4b34b654e1
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoqed: Add srq core support for RoCE and iWARP
Yuval Bason [Sun, 3 Jun 2018 16:13:07 +0000 (19:13 +0300)]
qed: Add srq core support for RoCE and iWARP

This patch adds support for configuring SRQ and provides the necessary
APIs for rdma upper layer driver (qedr) to enable the SRQ feature.

Signed-off-by: Michal Kalderon <michal.kalderon@cavium.com>
Signed-off-by: Ariel Elior <ariel.elior@cavium.com>
Signed-off-by: Yuval Bason <yuval.bason@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'bnx2-warnings'
David S. Miller [Mon, 4 Jun 2018 21:07:28 +0000 (17:07 -0400)]
Merge branch 'bnx2-warnings'

Varsha Rao says:

====================
net: bnx2: Fix checkpatch and clang warnings

This patchset fixes NULL comparison and extra parentheses, checkpatch
and clang warnings.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: ethernet: bnx2: Replace NULL comparison
Varsha Rao [Sun, 3 Jun 2018 11:49:52 +0000 (17:19 +0530)]
net: ethernet: bnx2: Replace NULL comparison

This patch fixes the checkpatch issue of NULL comparison. Replace x == NULL
with !x, by using the following coccinelle script:

@disable is_null@
expression e;
@@
-e==NULL
+!e

Signed-off-by: Varsha Rao <rvarsha016@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: ethernet: bnx2: Remove extra parentheses
Varsha Rao [Sun, 3 Jun 2018 11:49:04 +0000 (17:19 +0530)]
net: ethernet: bnx2: Remove extra parentheses

The following coccinelle script removes extra parentheses to fix the
clang warning of extraneous parentheses.

@disable paren@
identifier i;
expression e;
statement s;
@@
if (
-(i == e)
+i == e
 )
s

Suggested-by: Lukas Bulwahn <lukas.bulwahn@gmail.com>
Signed-off-by: Varsha Rao <rvarsha016@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: gemini: fix spelling mistake: "it" -> "is"
YueHaibing [Sun, 3 Jun 2018 08:10:01 +0000 (16:10 +0800)]
net: gemini: fix spelling mistake: "it" -> "is"

Trivial fix to spelling mistake in gemini dev_warn message

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agocls_flower: Fix comparing of old filter mask with new filter
Paul Blakey [Sun, 3 Jun 2018 07:06:14 +0000 (10:06 +0300)]
cls_flower: Fix comparing of old filter mask with new filter

We incorrectly compare the mask and the result is that we can't modify
an already existing rule.

Fix that by comparing correctly.

Fixes: 05cd271fd61a ("cls_flower: Support multiple masks per priority")
Reported-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Paul Blakey <paulb@mellanox.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agocls_flower: Fix missing free of rhashtable
Paul Blakey [Sun, 3 Jun 2018 07:06:13 +0000 (10:06 +0300)]
cls_flower: Fix missing free of rhashtable

When destroying the instance, destroy the head rhashtable.

Fixes: 05cd271fd61a ("cls_flower: Support multiple masks per priority")
Reported-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Paul Blakey <paulb@mellanox.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: skbuff.h: drop unneeded <linux/slab.h>
Randy Dunlap [Sun, 3 Jun 2018 04:40:19 +0000 (21:40 -0700)]
net: skbuff.h: drop unneeded <linux/slab.h>

<linux/skbuff.h> does not use nor need <linux/slab.h>, so drop this
header file from skbuff.h.

<linux/skbuff.h> is currently #included in around 1200 C source and
header files, making it the 31st most-used header file.

Build tested [allmodconfig] on 20 arch-es.

Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: chelsio: Use zeroing memory allocator instead of allocator/memset
YueHaibing [Sun, 3 Jun 2018 02:40:15 +0000 (10:40 +0800)]
net: chelsio: Use zeroing memory allocator instead of allocator/memset

Use dma_zalloc_coherent for allocating zeroed
memory and remove unnecessary memset function.

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agorxrpc: Fix handling of call quietly cancelled out on server
David Howells [Sun, 3 Jun 2018 01:17:39 +0000 (02:17 +0100)]
rxrpc: Fix handling of call quietly cancelled out on server

Sometimes an in-progress call will stop responding on the fileserver when
the fileserver quietly cancels the call with an internally marked abort
(RX_CALL_DEAD), without sending an ABORT to the client.

This causes the client's call to eventually expire from lack of incoming
packets directed its way, which currently leads to it being cancelled
locally with ETIME.  Note that it's not currently clear as to why this
happens as it's really hard to reproduce.

The rotation policy implement by kAFS, however, doesn't differentiate
between ETIME meaning we didn't get any response from the server and ETIME
meaning the call got cancelled mid-flow.  The latter leads to an oops when
fetching data as the rotation partially resets the afs_read descriptor,
which can result in a cleared page pointer being dereferenced because that
page has already been filled.

Handle this by the following means:

 (1) Set a flag on a call when we receive a packet for it.

 (2) Store the highest packet serial number so far received for a call
     (bearing in mind this may wrap).

 (3) If, when the "not received anything recently" timeout expires on a
     call, we've received at least one packet for a call and the connection
     as a whole has received packets more recently than that call, then
     cancel the call locally with ECONNRESET rather than ETIME.

     This indicates that the call was definitely in progress on the server.

 (4) In kAFS, if the rotation algorithm sees ECONNRESET rather than ETIME,
     don't try the next server, but rather abort the call.

     This avoids the oops as we don't try to reuse the afs_read struct.
     Rather, as-yet ungotten pages will be reread at a later data.

Also:

 (5) Add an rxrpc tracepoint to log detection of the call being reset.

Without this, I occasionally see an oops like the following:

    general protection fault: 0000 [#1] SMP PTI
    ...
    RIP: 0010:_copy_to_iter+0x204/0x310
    RSP: 0018:ffff8800cae0f828 EFLAGS: 00010206
    RAX: 0000000000000560 RBX: 0000000000000560 RCX: 0000000000000560
    RDX: ffff8800cae0f968 RSI: ffff8800d58b3312 RDI: 0005080000000000
    RBP: ffff8800cae0f968 R08: 0000000000000560 R09: ffff8800ca00f400
    R10: ffff8800c36f28d4 R11: 00000000000008c4 R12: ffff8800cae0f958
    R13: 0000000000000560 R14: ffff8800d58b3312 R15: 0000000000000560
    FS:  00007fdaef108080(0000) GS:ffff8800ca680000(0000) knlGS:0000000000000000
    CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
    CR2: 00007fb28a8fa000 CR3: 00000000d2a76002 CR4: 00000000001606e0
    Call Trace:
     skb_copy_datagram_iter+0x14e/0x289
     rxrpc_recvmsg_data.isra.0+0x6f3/0xf68
     ? trace_buffer_unlock_commit_regs+0x4f/0x89
     rxrpc_kernel_recv_data+0x149/0x421
     afs_extract_data+0x1e0/0x798
     ? afs_wait_for_call_to_complete+0xc9/0x52e
     afs_deliver_fs_fetch_data+0x33a/0x5ab
     afs_deliver_to_call+0x1ee/0x5e0
     ? afs_wait_for_call_to_complete+0xc9/0x52e
     afs_wait_for_call_to_complete+0x12b/0x52e
     ? wake_up_q+0x54/0x54
     afs_make_call+0x287/0x462
     ? afs_fs_fetch_data+0x3e6/0x3ed
     ? rcu_read_lock_sched_held+0x5d/0x63
     afs_fs_fetch_data+0x3e6/0x3ed
     afs_fetch_data+0xbb/0x14a
     afs_readpages+0x317/0x40d
     __do_page_cache_readahead+0x203/0x2ba
     ? ondemand_readahead+0x3a7/0x3c1
     ondemand_readahead+0x3a7/0x3c1
     generic_file_buffered_read+0x18b/0x62f
     __vfs_read+0xdb/0xfe
     vfs_read+0xb2/0x137
     ksys_read+0x50/0x8c
     do_syscall_64+0x7d/0x1a0
     entry_SYSCALL_64_after_hwframe+0x49/0xbe

Note the weird value in RDI which is a result of trying to kmap() a NULL
page pointer.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoAllow ethtool to change tun link settings
Chas Williams [Sat, 2 Jun 2018 21:49:53 +0000 (17:49 -0400)]
Allow ethtool to change tun link settings

Let user space set whatever it would like to advertise for the
tun interface.  Preserve the existing defaults.

Signed-off-by: Chas Williams <3chas3@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'sh_eth-fix-and-clean-up-sh_eth_soft_swap'
David S. Miller [Mon, 4 Jun 2018 19:23:26 +0000 (15:23 -0400)]
Merge branch 'sh_eth-fix-and-clean-up-sh_eth_soft_swap'

Sergei Shtylyov says:

====================
sh_eth: fix & clean up sh_eth_soft_swap()

Here's a set of 3 patches against DaveM's 'net-next.git' repo. First one fixes an
old buffer endiannes issue (luckily, the ARM SoCs are smart enough to not actually
care) plus couple clean ups around sh_eth_soft_swap()...
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agosh_eth: use DIV_ROUND_UP() in sh_eth_soft_swap()
Sergei Shtylyov [Sat, 2 Jun 2018 19:40:16 +0000 (22:40 +0300)]
sh_eth: use DIV_ROUND_UP() in sh_eth_soft_swap()

When initializing 'maxp' in sh_eth_soft_swap(), the buffer length needs
to be rounded  up -- that's just asking for DIV_ROUND_UP()!

Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agosh_eth: uninline sh_eth_soft_swap()
Sergei Shtylyov [Sat, 2 Jun 2018 19:38:56 +0000 (22:38 +0300)]
sh_eth: uninline sh_eth_soft_swap()

sh_eth_tsu_soft_swap() is called twice by the driver, remove *inline* and
move  that function  from the header to the driver itself to let gcc decide
whether to expand it inline or not...

Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agosh_eth: make sh_eth_soft_swap() work on ARM
Sergei Shtylyov [Sat, 2 Jun 2018 19:37:42 +0000 (22:37 +0300)]
sh_eth: make sh_eth_soft_swap() work on ARM

Browsing  thru the driver disassembly, I noticed that ARM gcc generated
no  code  whatsoever for sh_eth_soft_swap() while building a little-endian
kernel -- apparently __LITTLE_ENDIAN__ was not being #define'd, however
it got implicitly #define'd when building with the SH gcc (I could only
find the explicit #define __LITTLE_ENDIAN that was #include'd when building
a little-endian kernel).  Luckily, the Ether controller  only doing big-
endian DMA is encountered on the early SH771x SoCs only and all ARM SoCs
implement EDMR.DE and thus set 'sh_eth_cpu_data::hw_swap'. But anyway, we
need to fix the #ifdef inside sh_eth_soft_swap() to something that would
work on all architectures...

Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoixgbe: fix broken ipsec Rx with proper cast on spi
Shannon Nelson [Thu, 31 May 2018 21:12:18 +0000 (14:12 -0700)]
ixgbe: fix broken ipsec Rx with proper cast on spi

Fix up a cast problem introduced by a sparse cleanup patch.  This fixes
a problem where the encrypted packets were not recognized on Rx and
subsequently dropped.

Fixes: 9cfbfa701b55 ("ixgbe: cleanup sparse warnings")
Signed-off-by: Shannon Nelson <shannon.nelson@oracle.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoixgbe: check ipsec ip addr against mgmt filters
Shannon Nelson [Wed, 30 May 2018 18:20:04 +0000 (11:20 -0700)]
ixgbe: check ipsec ip addr against mgmt filters

Make sure we don't try to offload the decryption of an incoming
packet that should get delivered to the management engine.  This
is a corner case that will likely be very seldom seen, but could
really confuse someone if they were to hit it.

Suggested-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Shannon Nelson <shannon.nelson@oracle.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoMerge branch 'mlxsw-Fixes-in-offloading-of-mirror-to-gretap'
David S. Miller [Mon, 4 Jun 2018 17:27:58 +0000 (13:27 -0400)]
Merge branch 'mlxsw-Fixes-in-offloading-of-mirror-to-gretap'

Ido Schimmel says:

====================
mlxsw: Fixes in offloading of mirror-to-gretap

Petr says:

These two patches fix issues in offloading of mirror-to-gretap when
bridge is present in the underlay.

In patch #1, reconsideration of SPAN configuration is not done right at
the point that SWITCHDEV_OBJ_ID_PORT_VLAN deletion notification is
distributed, but is postponed, because the notifications are actually
distributed before the relevant change is implemented in the bridge.

In patch #2, a problem in configuring VLAN tagging in situations when a
VLAN device is on top of an 802.1Q bridge whose egress port is marked as
"egress untagged". In that case, mlxsw would neglect to suppress the
tagging implicitly assumed after the VLAN device was seen.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agomlxsw: spectrum_span: Suppress VLAN on BRIDGE_VLAN_INFO_UNTAGGED
Petr Machata [Sat, 2 Jun 2018 18:09:35 +0000 (21:09 +0300)]
mlxsw: spectrum_span: Suppress VLAN on BRIDGE_VLAN_INFO_UNTAGGED

When offloading mirroring to gretap or ip6gretap netdevices, an 802.1q
bridge is one of the soft devices permissible in the underlay when
resolving the packet path. After the packet path is resolved to a
particular bridge egress device, flags on packet VLAN determine whether
the egressed packet should be tagged.

The current logic however only ever sets the VLAN tag, never suppresses
it. Thus if there's a VLAN netdevice above the bridge that determines
the packet VLAN, that VLAN is never unset, and mirroring is configured
with VLAN tagging.

Fix by setting the packet VLAN on both branches: set to zero (for unset)
when BRIDGE_VLAN_INFO_UNTAGGED, copy the resolved VLAN (e.g. from bridge
PVID) otherwise.

Fixes: 946a11e7408e ("mlxsw: spectrum_span: Allow bridge for gretap mirror")
Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agomlxsw: spectrum_switchdev: Postpone respin on object deletion
Petr Machata [Sat, 2 Jun 2018 18:09:34 +0000 (21:09 +0300)]
mlxsw: spectrum_switchdev: Postpone respin on object deletion

VLAN deletion notifications are emitted before the relevant change is
projected to bridge configuration. Thus, like with VLAN addition,
schedule SPAN respin for later.

Fixes: c520bc698647 ("mlxsw: Respin SPAN on switchdev events")
Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoixgbe: fix possible race in reset subtask
Tony Nguyen [Wed, 30 May 2018 16:05:12 +0000 (09:05 -0700)]
ixgbe: fix possible race in reset subtask

Similar to ixgbevf, the same possibility for race exists. Extend the RTNL
lock in ixgbe_reset_subtask() to protect the state bits; this is to make
sure that we get the most up-to-date values for the bits and avoid a
possible race when going down.

Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agobpf, i40e: add meta data support
Daniel Borkmann [Mon, 28 May 2018 09:07:20 +0000 (11:07 +0200)]
bpf, i40e: add meta data support

Add support for XDP meta data when using build skb variant of
the i40e driver. Implementation is analogous to the existing
ixgbe and ixgbevf support for meta data from 366a88fe2f40 ("bpf,
ixgbe: add meta data support") and be8333322eff ("ixgbevf: Add
support for meta data"). With the build skb variant we get
192 bytes of extra headroom which can be used for encaps or
meta data.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Tested-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoipv6: omit traffic class when calculating flow hash
Michal Kubecek [Mon, 4 Jun 2018 09:36:05 +0000 (11:36 +0200)]
ipv6: omit traffic class when calculating flow hash

Some of the code paths calculating flow hash for IPv6 use flowlabel member
of struct flowi6 which, despite its name, encodes both flow label and
traffic class. If traffic class changes within a TCP connection (as e.g.
ssh does), ECMP route can switch between path. It's also inconsistent with
other code paths where ip6_flowlabel() (returning only flow label) is used
to feed the key.

Use only flow label everywhere, including one place where hash key is set
using ip6_flowinfo().

Fixes: 51ebd3181572 ("ipv6: add support of equal cost multipath (ECMP)")
Fixes: f70ea018da06 ("net: Add functions to get skb->hash based on flow structures")
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoixgbe: introduce a helper to simplify code
YueHaibing [Wed, 23 May 2018 12:08:13 +0000 (20:08 +0800)]
ixgbe: introduce a helper to simplify code

ixgbe_dbg_reg_ops_read and ixgbe_dbg_netdev_ops_read copy-pasting
the same code except for ixgbe_dbg_netdev_ops_buf/ixgbe_dbg_reg_ops_buf,
so introduce a helper ixgbe_dbg_common_ops_read to remove redundant code.

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoRevert "ipv6: omit traffic class when calculating flow hash"
David S. Miller [Mon, 4 Jun 2018 17:20:38 +0000 (13:20 -0400)]
Revert "ipv6: omit traffic class when calculating flow hash"

This reverts commit 87ae68c8b4944d142447b88875c9c412c714434f.

Applied the wrong version of this fix, correct version
coming up.

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoixgbevf: fix possible race in the reset subtask
Emil Tantilov [Tue, 22 May 2018 16:18:46 +0000 (09:18 -0700)]
ixgbevf: fix possible race in the reset subtask

Extend the RTNL lock in ixgbevf_reset_subtask() to protect the state bits
check in addition to the call to ixgbevf_reinit_locked().

This is to make sure that we get the most up-to-date values for the bits
and avoid a possible race when going down.

Suggested-by: Zhiping du <zhipingdu@tencent.com>
Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoipv6: omit traffic class when calculating flow hash
Michal Kubecek [Sat, 2 Jun 2018 07:40:34 +0000 (09:40 +0200)]
ipv6: omit traffic class when calculating flow hash

Some of the code paths calculating flow hash for IPv6 use flowlabel member
of struct flowi6 which, despite its name, encodes both flow label and
traffic class. If traffic class changes within a TCP connection (as e.g.
ssh does), ECMP route can switch between path. It's also incosistent with
other code paths where ip6_flowlabel() (returning only flow label) is used
to feed the key.

Use only flow label everywhere, including one place where hash key is set
using ip6_flowinfo().

Fixes: 51ebd3181572 ("ipv6: add support of equal cost multipath (ECMP)")
Fixes: f70ea018da06 ("net: Add functions to get skb->hash based on flow structures")
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Tested-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoixgbevf: Fix coexistence of malicious driver detection with XDP
Alexander Duyck [Tue, 22 May 2018 15:44:29 +0000 (11:44 -0400)]
ixgbevf: Fix coexistence of malicious driver detection with XDP

In the case of the VF driver it is supposed to provide a context descriptor
that allows us to provide information about the header offsets inside of
the frame. However in the case of XDP we don't really have any of that
information since the data is minimally processed. As a result we were
seeing malicious driver detection (MDD) events being triggered when the PF
had that functionality enabled.

To address this I have added a bit of new code that will "prime" the XDP
ring by providing one context descriptor that assumes the minimal setup of
an Ethernet frame which is an L2 header length of 14. With just that we can
provide enough information to make the hardware happy so that we don't
trigger MDD events.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoigb: Wait 10ms just once after TX queues reset
Sergey Nemov [Fri, 18 May 2018 09:58:40 +0000 (11:58 +0200)]
igb: Wait 10ms just once after TX queues reset

Move 10ms sleep out of function resetting TX queue.
Reset all the TX queues in one turn and
wait for all of them just once.

Use usleep_range() instead of mdelay() in order not to
affect transmission on other interfaces.

Signed-off-by: Sergey Nemov <sergey.nemov@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoigb: Clear TSICR interrupts together with ICR
Joanna Yurdal [Wed, 16 May 2018 11:14:11 +0000 (13:14 +0200)]
igb: Clear TSICR interrupts together with ICR

Issuing "ip link set up/down" can block TSICR interrupts, what results in
missing PTP Tx timestamp and no PPS pulse generation.

Problem happens when the link is set up with the TSICR interrupts pending.
ICR is cleared before enabling interrupts, while TSICR is not. When all TSICR
interrupts are pending at this moment, time_sync interrupt will never
be generated. TSICR should be cleared as well.

In order to reproduce the issue:
1. Setup linux with IEEE 1588 grandmaster and PPS output enabled
2. Continue setting link up/down with random intervals between commands
3. Wait until PPS is not generated ( only one pulse is generated and PPS
dies), and ptp4l complains constantly about Tx timeout.

Signed-off-by: Joanna Yurdal <jyu@trackman.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoDocumentation: e1000: Update kernel documentation
Jeff Kirsher [Thu, 10 May 2018 19:55:38 +0000 (12:55 -0700)]
Documentation: e1000: Update kernel documentation

Updated the e1000.txt kernel documentation with the latest information.

Also convert the text file to reStructuredText (RST) format, since the
Linux kernel documentation now uses this format for documentation.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
6 years agoDocumentation: e100: Update the Intel 10/100 driver doc
Jeff Kirsher [Thu, 10 May 2018 19:20:13 +0000 (12:20 -0700)]
Documentation: e100: Update the Intel 10/100 driver doc

Over the years, several of the links have changed or are no longer valid
so update them.  In addition, the default values were incorrect for a
couple of parameters.

Converted the text file to the reStructuredText (RST) format, since the
Linux kernel documentation now uses this format for documentation.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
6 years agoe1000e: Ignore TSYNCRXCTL when getting I219 clock attributes
Benjamin Poirier [Thu, 10 May 2018 07:28:35 +0000 (16:28 +0900)]
e1000e: Ignore TSYNCRXCTL when getting I219 clock attributes

There have been multiple reports of crashes that look like
kernel: RIP: 0010:[<ffffffff8110303f>] timecounter_read+0xf/0x50
[...]
kernel: Call Trace:
kernel:  [<ffffffffa0806b0f>] e1000e_phc_gettime+0x2f/0x60 [e1000e]
kernel:  [<ffffffffa0806c5d>] e1000e_systim_overflow_work+0x1d/0x80 [e1000e]
kernel:  [<ffffffff810992c5>] process_one_work+0x155/0x440
kernel:  [<ffffffff81099e16>] worker_thread+0x116/0x4b0
kernel:  [<ffffffff8109f422>] kthread+0xd2/0xf0
kernel:  [<ffffffff8163184f>] ret_from_fork+0x3f/0x70

These can be traced back to the fact that e1000e_systim_reset() skips the
timecounter_init() call if e1000e_get_base_timinca() returns -EINVAL, which
leads to a null deref in timecounter_read().

Commit 83129b37ef35 ("e1000e: fix systim issues", v4.2-rc1) reworked
e1000e_get_base_timinca() in such a way that it can return -EINVAL for
e1000_pch_spt if the SYSCFI bit is not set in TSYNCRXCTL.

Some experimentation has shown that on I219 (e1000_pch_spt, "MAC: 12")
adapters, the E1000_TSYNCRXCTL_SYSCFI flag is unstable; TSYNCRXCTL reads
sometimes don't have the SYSCFI bit set. Retrying the read shortly after
finds the bit to be set. This was observed at boot (probe) but also link up
and link down.

Moreover, the phc (PTP Hardware Clock) seems to operate normally even after
reads where SYSCFI=0. Therefore, remove this register read and
unconditionally set the clock parameters.

Reported-by: Achim Mildenberger <admin@fph.physik.uni-karlsruhe.de>
Message-Id: <20180425065243.g5mqewg5irkwgwgv@f2>
Bugzilla: https://bugzilla.suse.com/show_bug.cgi?id=1075876
Fixes: 83129b37ef35 ("e1000e: fix systim issues")
Signed-off-by: Benjamin Poirier <bpoirier@suse.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agonet: mvpp2: mvpp2_percpu_read_relaxed() can be static
kbuild test robot [Fri, 1 Jun 2018 19:46:13 +0000 (03:46 +0800)]
net: mvpp2: mvpp2_percpu_read_relaxed() can be static

Fixes: db9d7d36eecc ("net: mvpp2: Split the PPv2 driver to a dedicated directory")
Signed-off-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet/packet: refine check for priv area size
Eric Dumazet [Fri, 1 Jun 2018 16:23:02 +0000 (09:23 -0700)]
net/packet: refine check for priv area size

syzbot was able to trick af_packet again [1]

Various commits tried to address the problem in the past,
but failed to take into account V3 header size.

[1]

tpacket_rcv: packet too big, clamped from 72 to 4294967224. macoff=96
BUG: KASAN: use-after-free in prb_run_all_ft_ops net/packet/af_packet.c:1016 [inline]
BUG: KASAN: use-after-free in prb_fill_curr_block.isra.59+0x4e5/0x5c0 net/packet/af_packet.c:1039
Write of size 2 at addr ffff8801cb62000e by task kworker/1:2/2106

CPU: 1 PID: 2106 Comm: kworker/1:2 Not tainted 4.17.0-rc7+ #77
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Workqueue: ipv6_addrconf addrconf_dad_work
Call Trace:
 __dump_stack lib/dump_stack.c:77 [inline]
 dump_stack+0x1b9/0x294 lib/dump_stack.c:113
 print_address_description+0x6c/0x20b mm/kasan/report.c:256
 kasan_report_error mm/kasan/report.c:354 [inline]
 kasan_report.cold.7+0x242/0x2fe mm/kasan/report.c:412
 __asan_report_store2_noabort+0x17/0x20 mm/kasan/report.c:436
 prb_run_all_ft_ops net/packet/af_packet.c:1016 [inline]
 prb_fill_curr_block.isra.59+0x4e5/0x5c0 net/packet/af_packet.c:1039
 __packet_lookup_frame_in_block net/packet/af_packet.c:1094 [inline]
 packet_current_rx_frame net/packet/af_packet.c:1117 [inline]
 tpacket_rcv+0x1866/0x3340 net/packet/af_packet.c:2282
 dev_queue_xmit_nit+0x891/0xb90 net/core/dev.c:2018
 xmit_one net/core/dev.c:3049 [inline]
 dev_hard_start_xmit+0x16b/0xc10 net/core/dev.c:3069
 __dev_queue_xmit+0x2724/0x34c0 net/core/dev.c:3584
 dev_queue_xmit+0x17/0x20 net/core/dev.c:3617
 neigh_resolve_output+0x679/0xad0 net/core/neighbour.c:1358
 neigh_output include/net/neighbour.h:482 [inline]
 ip6_finish_output2+0xc9c/0x2810 net/ipv6/ip6_output.c:120
 ip6_finish_output+0x5fe/0xbc0 net/ipv6/ip6_output.c:154
 NF_HOOK_COND include/linux/netfilter.h:277 [inline]
 ip6_output+0x227/0x9b0 net/ipv6/ip6_output.c:171
 dst_output include/net/dst.h:444 [inline]
 NF_HOOK include/linux/netfilter.h:288 [inline]
 ndisc_send_skb+0x100d/0x1570 net/ipv6/ndisc.c:491
 ndisc_send_ns+0x3c1/0x8d0 net/ipv6/ndisc.c:633
 addrconf_dad_work+0xbef/0x1340 net/ipv6/addrconf.c:4033
 process_one_work+0xc1e/0x1b50 kernel/workqueue.c:2145
 worker_thread+0x1cc/0x1440 kernel/workqueue.c:2279
 kthread+0x345/0x410 kernel/kthread.c:240
 ret_from_fork+0x3a/0x50 arch/x86/entry/entry_64.S:412

The buggy address belongs to the page:
page:ffffea00072d8800 count:0 mapcount:-127 mapping:0000000000000000 index:0xffff8801cb620e80
flags: 0x2fffc0000000000()
raw: 02fffc0000000000 0000000000000000 ffff8801cb620e80 00000000ffffff80
raw: ffffea00072e3820 ffffea0007132d20 0000000000000002 0000000000000000
page dumped because: kasan: bad access detected

Memory state around the buggy address:
 ffff8801cb61ff00: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 ffff8801cb61ff80: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>ffff8801cb620000: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
                      ^
 ffff8801cb620080: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
 ffff8801cb620100: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff

Fixes: 2b6867c2ce76 ("net/packet: fix overflow in check for priv area size")
Fixes: dc808110bb62 ("packet: handle too big packets for PACKET_V3")
Fixes: f6fb8f100b80 ("af-packet: TPACKET_V3 flexible buffer implementation.")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: aquantia: make function aq_fw2x_get_mac_permanent static
Colin Ian King [Fri, 1 Jun 2018 15:28:34 +0000 (16:28 +0100)]
net: aquantia: make function aq_fw2x_get_mac_permanent static

The function aq_fw2x_get_mac_permanent is local to the source and does
not need to be in global scope, so make it static.

Cleans up sparse warning:
warning: symbol 'aq_fw2x_get_mac_permanent' was not declared. Should it
be static?

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMAINTAINERS: TCP gets its first maintainer
Eric Dumazet [Mon, 4 Jun 2018 13:50:29 +0000 (06:50 -0700)]
MAINTAINERS: TCP gets its first maintainer

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agovrf: check the original netdevice for generating redirect
Stephen Suryaputra [Fri, 1 Jun 2018 04:05:21 +0000 (00:05 -0400)]
vrf: check the original netdevice for generating redirect

Use the right device to determine if redirect should be sent especially
when using vrf. Same as well as when sending the redirect.

Signed-off-by: Stephen Suryaputra <ssuryaextr@gmail.com>
Acked-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: ethernet: mlx4: Remove unnecessary parentheses
Varsha Rao [Fri, 1 Jun 2018 02:00:49 +0000 (07:30 +0530)]
net: ethernet: mlx4: Remove unnecessary parentheses

This patch fixes the clang warning of extraneous parentheses, with the
following coccinelle script.

@@
identifier i;
expression e;
statement s;
@@
if (
-(i == e)
+i == e
 )
s

Suggested-by: Lukas Bulwahn <lukas.bulwahn@gmail.com>
Signed-off-by: Varsha Rao <rvarsha016@gmail.com>
Acked-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: stmmac: Add Flexible PPS support
Jose Abreu [Thu, 31 May 2018 17:01:27 +0000 (18:01 +0100)]
net: stmmac: Add Flexible PPS support

This adds support for Flexible PPS output (which is equivalent
to per_out output of PTP subsystem).

Tested using an oscilloscope and the following commands:

1) Start PTP4L:
# ptp4l -A -4 -H -m -i eth0 &
2) Set Flexible PPS frequency:
# echo <idx> <ts> <tns> <ps> <pns> > /sys/class/ptp/ptpX/period

Where, ts/tns is start time and ps/pns is period time, and ptpX is ptp
of eth0.

Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Joao Pinto <jpinto@synopsys.com>
Cc: Vitor Soares <soares@synopsys.com>
Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Cc: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'qed-next'
David S. Miller [Mon, 4 Jun 2018 14:10:47 +0000 (10:10 -0400)]
Merge branch 'qed-next'

Sudarsana Reddy Kalluru says:

====================
qed: Fix issues in UFP feature commit 'cac6f691'.

This patch series fixes couple of issues in the UFP feature commit,
   cac6f691: Add support for Unified Fabric Port.

Changes from previous version:
------------------------------
v2: Added "Fixes:" tag.

Please consider applying it to "net-next".
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoqed: Fix use of incorrect shmem address.
Sudarsana Reddy Kalluru [Fri, 1 Jun 2018 01:47:37 +0000 (18:47 -0700)]
qed: Fix use of incorrect shmem address.

Incorrect shared memory address is used while deriving the values
for tc and pri_type. Use shmem address corresponding to 'oem_cfg_func'
where the management firmare saves tc/pri_type values.

Fixes: cac6f691 ("qed: Add support for Unified Fabric Port")
Signed-off-by: Sudarsana Reddy Kalluru <Sudarsana.Kalluru@cavium.com>
Signed-off-by: Ariel Elior <ariel.elior@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoqed: Fix shared memory inconsistency between driver and the MFW.
Sudarsana Reddy Kalluru [Fri, 1 Jun 2018 01:47:36 +0000 (18:47 -0700)]
qed: Fix shared memory inconsistency between driver and the MFW.

The structure shared between driver and management firmware (MFW)
differ in sizes. The additional field defined by the MFW is not
relevant to the current driver. Add a dummy field to the structure.

Fixes: cac6f691 ("qed: Add support for Unified Fabric Port")
Signed-off-by: Sudarsana Reddy Kalluru <Sudarsana.Kalluru@cavium.com>
Signed-off-by: Ariel Elior <ariel.elior@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'selftests-mirror_vlan-fixes'
David S. Miller [Mon, 4 Jun 2018 14:08:46 +0000 (10:08 -0400)]
Merge branch 'selftests-mirror_vlan-fixes'

Petr Machata says:

====================
selftests: forwarding: mirror_vlan: Fixlets

This patchset includes two small fixes for the tests that were
introduced in commit 1bb58d2d3cbe ("Merge branch
'Mirroring-tests-involving-VLAN'").

In patch #1, a "tc action trap" is uninstalled after the suite runs,
instead of being installed again.

In patch #2, a test in suite is renamed to differentiate it from another
test of the same name.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoselftests: forwarding: mirror_vlan: Change test description
Petr Machata [Thu, 31 May 2018 22:37:47 +0000 (00:37 +0200)]
selftests: forwarding: mirror_vlan: Change test description

The test description is displayed with the PASS/FAIL resolution after
the test is ran. There however already is one other test described
exactly like this, which makes it unclear which of the tests passed or
failed. Make the description unique.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoselftests: forwarding: mirror_vlan: Uninstall trap
Petr Machata [Thu, 31 May 2018 22:37:44 +0000 (00:37 +0200)]
selftests: forwarding: mirror_vlan: Uninstall trap

Instead of installing a trap before tests run and uninstalling it after
they run, mirror_vlan.sh installs it twice due to a typo. Fix the typo.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'selftests-net-various'
David S. Miller [Mon, 4 Jun 2018 13:50:12 +0000 (09:50 -0400)]
Merge branch 'selftests-net-various'

Willem de Bruijn says:

====================
selftests/net: various

A few odds and ends to network tests:

- msg_zerocopy: run as part of kselftest
- udp gso:      add missing bounds test for minimal sizes
- psocket_snd:  initial basic conformance test
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoselftests/net: add packet socket packet_snd test
Willem de Bruijn [Thu, 31 May 2018 16:14:40 +0000 (12:14 -0400)]
selftests/net: add packet socket packet_snd test

Add regression tests for PF_PACKET transmission using packet_snd.

The TPACKET ring interface has tests for transmission and reception.
This is an initial stab at the same for the send call based interface.

Packets are sent over loopback, then read twice. The entire packet is
read from another packet socket and compared. The packet is also
verified to arrive at a UDP socket for protocol conformance.

The test sends a packet over loopback, testing the following options
(not the full cross-product):

- SOCK_DGRAM
- SOCK_RAW
- vlan tag
- qdisc bypass
- bind() and sendto()
- virtio_net_hdr
- csum offload (NOT actual csum feature, ignored on loopback)
- gso

Besides these basic functionality tests, the test runs from a set
of bounds checks, positive and negative. Running over loopback, which
has dev->min_header_len, it cannot generate variable length hhlen.

Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoselftests/net: udpgso: test small gso_size boundary conditions
Willem de Bruijn [Thu, 31 May 2018 16:14:39 +0000 (12:14 -0400)]
selftests/net: udpgso: test small gso_size boundary conditions

Verify that udpgso can generate segments smaller than device mtu, down
to the extreme case of 1B gso_size.

Verify that irrespective of gso_size, udpgso restricts the number of
segments it will generate per call (UDP_MAX_SEGMENTS).

Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoselftests/net: enable msg_zerocopy test
Willem de Bruijn [Thu, 31 May 2018 16:14:38 +0000 (12:14 -0400)]
selftests/net: enable msg_zerocopy test

The existing msg_zerocopy test takes additional protocol arguments.
Add a variant that takes no arguments and runs all supported variants.
Call this from kselftest.

Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: virtio: simplify the virtnet_find_vqs
Tonghao Zhang [Thu, 31 May 2018 14:16:32 +0000 (07:16 -0700)]
net: virtio: simplify the virtnet_find_vqs

Use the common free functions while return successfully.

Signed-off-by: Tonghao Zhang <xiangxia.m.yue@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge tag 'wireless-drivers-next-for-davem-2018-05-31' of git://git.kernel.org/pub...
David S. Miller [Sun, 3 Jun 2018 15:03:10 +0000 (11:03 -0400)]
Merge tag 'wireless-drivers-next-for-davem-2018-05-31' of git://git./linux/kernel/git/kvalo/wireless-drivers-next

Kalle Valo says:

====================
wireless-drivers-next patches for 4.18

Hopefully the last pull request to 4.18 before the merge window.
Nothing major here, we have smaller new features and of course a lots
of fixes.

Major changes:

ath10k

* add memory dump support for QCA9888 and QCA99X0

* add support to configure channel dwell time

* support new DFS host confirmation feature in the firmware

ath

* update various regulatory mappings

wcn36xx

* various fixes to improve reliability

* add Factory Test Mode support

brmfmac

* add debugfs file for reading firmware capabilities

mwifiex

* support sysfs initiated device coredump
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agomlx4_core: restore optimal ICM memory allocation
Eric Dumazet [Thu, 31 May 2018 12:52:24 +0000 (05:52 -0700)]
mlx4_core: restore optimal ICM memory allocation

Commit 1383cb8103bb ("mlx4_core: allocate ICM memory in page size chunks")
brought two regressions caught in our regression suite.

The big one is an additional cost of 256 bytes of overhead per 4096 bytes,
or 6.25 % which is unacceptable since ICM can be pretty large.

This comes from having to allocate one struct mlx4_icm_chunk (256 bytes)
per MLX4_TABLE_CHUNK, which the buggy commit shrank to 4KB
(instead of prior 256KB)

Note that mlx4_alloc_icm() is already able to try high order allocations
and fallback to low-order allocations under high memory pressure.

Most of these allocations happen right after boot time, when we get
plenty of non fragmented memory, there is really no point being so
pessimistic and break huge pages into order-0 ones just for fun.

We only have to tweak gfp_mask a bit, to help falling back faster,
without risking OOM killings.

Second regression is an KASAN fault, that will need further investigations.

Fixes: 1383cb8103bb ("mlx4_core: allocate ICM memory in page size chunks")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Tariq Toukan <tariqt@mellanox.com>
Cc: John Sperbeck <jsperbeck@google.com>
Cc: Tarick Bedeir <tarick@google.com>
Cc: Qing Huang <qing.huang@oracle.com>
Cc: Daniel Jurgens <danielj@mellanox.com>
Cc: Zhu Yanjun <yanjun.zhu@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agovlan: use non-archaic spelling of failes
Thadeu Lima de Souza Cascardo [Thu, 31 May 2018 12:20:20 +0000 (09:20 -0300)]
vlan: use non-archaic spelling of failes

Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: axienet: remove stale comment of axienet_open
YueHaibing [Thu, 31 May 2018 11:51:15 +0000 (19:51 +0800)]
net: axienet: remove stale comment of axienet_open

axienet_open no longer return -ENODEV when PHY cannot be connected to
since commit d7cc3163e026 ("net: axienet: Support phy-less mode of operation")

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet/ncsi: Avoid GFP_KERNEL in response handler
Samuel Mendoza-Jonas [Thu, 31 May 2018 07:02:54 +0000 (17:02 +1000)]
net/ncsi: Avoid GFP_KERNEL in response handler

ncsi_rsp_handler_gc() allocates the filter arrays using GFP_KERNEL in
softirq context, causing the below backtrace. This allocation is only a
few dozen bytes during probing so allocate with GFP_ATOMIC instead.

[   42.813372] BUG: sleeping function called from invalid context at mm/slab.h:416
[   42.820900] in_atomic(): 1, irqs_disabled(): 0, pid: 213, name: kworker/0:1
[   42.827893] INFO: lockdep is turned off.
[   42.832023] CPU: 0 PID: 213 Comm: kworker/0:1 Tainted: G        W       4.13.16-01441-gad99b38 #65
[   42.841007] Hardware name: Generic DT based system
[   42.845966] Workqueue: events ncsi_dev_work
[   42.850251] [<8010a494>] (unwind_backtrace) from [<80107510>] (show_stack+0x20/0x24)
[   42.858046] [<80107510>] (show_stack) from [<80612770>] (dump_stack+0x20/0x28)
[   42.865309] [<80612770>] (dump_stack) from [<80148248>] (___might_sleep+0x230/0x2b0)
[   42.873241] [<80148248>] (___might_sleep) from [<80148334>] (__might_sleep+0x6c/0xac)
[   42.881129] [<80148334>] (__might_sleep) from [<80240d6c>] (__kmalloc+0x210/0x2fc)
[   42.888737] [<80240d6c>] (__kmalloc) from [<8060ad54>] (ncsi_rsp_handler_gc+0xd0/0x170)
[   42.896770] [<8060ad54>] (ncsi_rsp_handler_gc) from [<8060b454>] (ncsi_rcv_rsp+0x16c/0x1d4)
[   42.905314] [<8060b454>] (ncsi_rcv_rsp) from [<804d86c8>] (__netif_receive_skb_core+0x3c8/0xb50)
[   42.914158] [<804d86c8>] (__netif_receive_skb_core) from [<804d96cc>] (__netif_receive_skb+0x20/0x7c)
[   42.923420] [<804d96cc>] (__netif_receive_skb) from [<804de4b0>] (netif_receive_skb_internal+0x78/0x6a4)
[   42.932931] [<804de4b0>] (netif_receive_skb_internal) from [<804df980>] (netif_receive_skb+0x78/0x158)
[   42.942292] [<804df980>] (netif_receive_skb) from [<8042f204>] (ftgmac100_poll+0x43c/0x4e8)
[   42.950855] [<8042f204>] (ftgmac100_poll) from [<804e094c>] (net_rx_action+0x278/0x4c4)
[   42.958918] [<804e094c>] (net_rx_action) from [<801016a8>] (__do_softirq+0xe0/0x4c4)
[   42.966716] [<801016a8>] (__do_softirq) from [<8011cd9c>] (do_softirq.part.4+0x50/0x78)
[   42.974756] [<8011cd9c>] (do_softirq.part.4) from [<8011cebc>] (__local_bh_enable_ip+0xf8/0x11c)
[   42.983579] [<8011cebc>] (__local_bh_enable_ip) from [<804dde08>] (__dev_queue_xmit+0x260/0x890)
[   42.992392] [<804dde08>] (__dev_queue_xmit) from [<804df1f0>] (dev_queue_xmit+0x1c/0x20)
[   43.000689] [<804df1f0>] (dev_queue_xmit) from [<806099c0>] (ncsi_xmit_cmd+0x1c0/0x244)
[   43.008763] [<806099c0>] (ncsi_xmit_cmd) from [<8060dc14>] (ncsi_dev_work+0x2e0/0x4c8)
[   43.016725] [<8060dc14>] (ncsi_dev_work) from [<80133dfc>] (process_one_work+0x214/0x6f8)
[   43.024940] [<80133dfc>] (process_one_work) from [<80134328>] (worker_thread+0x48/0x558)
[   43.033070] [<80134328>] (worker_thread) from [<8013ba80>] (kthread+0x130/0x174)
[   43.040506] [<8013ba80>] (kthread) from [<80102950>] (ret_from_fork+0x14/0x24)

Fixes: 062b3e1b6d4f ("net/ncsi: Refactor MAC, VLAN filters")
Signed-off-by: Samuel Mendoza-Jonas <sam@mendozajonas.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: netcp: ethss: remove unnecessary pointer set to NULL
YueHaibing [Thu, 31 May 2018 03:48:48 +0000 (11:48 +0800)]
net: netcp: ethss: remove unnecessary pointer set to NULL

If statement has make sure the 'slave->phy' is NULL

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet/smc: fix error return code in smc_setsockopt()
Wei Yongjun [Thu, 31 May 2018 02:31:22 +0000 (02:31 +0000)]
net/smc: fix error return code in smc_setsockopt()

Fix to return error code -EINVAL instead of 0 if optlen is invalid.

Fixes: 01d2f7e2cdd3 ("net/smc: sockopts TCP_NODELAY and TCP_CORK")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet/mlx5: Make function mlx5_fpga_tls_send_teardown_cmd() static
Wei Yongjun [Thu, 31 May 2018 02:31:12 +0000 (02:31 +0000)]
net/mlx5: Make function mlx5_fpga_tls_send_teardown_cmd() static

Fixes the following sparse warning:

drivers/net/ethernet/mellanox/mlx5/core/fpga/tls.c:199:6: warning:
 symbol 'mlx5_fpga_tls_send_teardown_cmd' was not declared. Should it be static?

Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agohv_netvsc: fix error return code in netvsc_probe()
Wei Yongjun [Thu, 31 May 2018 02:04:43 +0000 (02:04 +0000)]
hv_netvsc: fix error return code in netvsc_probe()

Fix to return a negative error code from the failover register fail
error handling case instead of 0, as done elsewhere in this function.

Fixes: 1ff78076d8dd ("netvsc: refactor notifier/event handling code to use the failover framework")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
David S. Miller [Sun, 3 Jun 2018 13:31:58 +0000 (09:31 -0400)]
Merge git://git./linux/kernel/git/davem/net

Filling in the padding slot in the bpf structure as a bug fix in 'ne'
overlapped with actually using that padding area for something in
'net-next'.

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: phy: consider PHY_IGNORE_INTERRUPT in state machine PHY_NOLINK handling
Heiner Kallweit [Wed, 30 May 2018 20:13:20 +0000 (22:13 +0200)]
net: phy: consider PHY_IGNORE_INTERRUPT in state machine PHY_NOLINK handling

We can bail out immediately also in case of PHY_IGNORE_INTERRUPT because
phy_mac_interupt() informs us once the link is up.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf-next
David S. Miller [Sun, 3 Jun 2018 12:24:27 +0000 (08:24 -0400)]
Merge git://git./linux/kernel/git/pablo/nf-next

Pablo Neira Ayuso says:

====================
Netfilter updates for net-next

The following patchset contains Netfilter updates for your net-next tree:

1) Get rid of nf_sk_is_transparent(), use inet_sk_transparent() instead.
   From Máté Eckl.

2) Move shared tproxy infrastructure to nf_tproxy_ipv4 and nf_tproxy_ipv6.
   Also from Máté.

3) Add hashtable to speed up chain lookups by name, from Florian Westphal.

4) Patch series to add connlimit support reusing part of the
   nf_conncount infrastructure. This includes preparation changes such
   passing context to the object and expression destroy interface;
   garbage collection for expressions embedded into set elements, and
   the introduction of the clone_destroy interface for expressions.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Sun, 3 Jun 2018 00:35:53 +0000 (17:35 -0700)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) Infinite loop in _decode_session6(), from Eric Dumazet.

 2) Pass correct argument to nla_strlcpy() in netfilter, also from Eric
    Dumazet.

 3) Out of bounds memory access in ipv6 srh code, from Mathieu Xhonneux.

 4) NULL deref in XDP_REDIRECT handling of tun driver, from Toshiaki
    Makita.

 5) Incorrect idr release in cls_flower, from Paul Blakey.

 6) Probe error handling fix in davinci_emac, from Dan Carpenter.

 7) Memory leak in XPS configuration, from Alexander Duyck.

 8) Use after free with cloned sockets in kcm, from Kirill Tkhai.

 9) MTU handling fixes fo ip_tunnel and ip6_tunnel, from Nicolas
    Dichtel.

10) Fix UAPI hole in bpf data structure for 32-bit compat applications,
    from Daniel Borkmann.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (33 commits)
  bpf: fix uapi hole for 32 bit compat applications
  net: usb: cdc_mbim: add flag FLAG_SEND_ZLP
  ip6_tunnel: remove magic mtu value 0xFFF8
  ip_tunnel: restore binding to ifaces with a large mtu
  net: dsa: b53: Add BCM5389 support
  kcm: Fix use-after-free caused by clonned sockets
  net-sysfs: Fix memory leak in XPS configuration
  ixgbe: fix parsing of TC actions for HW offload
  net: ethernet: davinci_emac: fix error handling in probe()
  net/ncsi: Fix array size in dumpit handler
  cls_flower: Fix incorrect idr release when failing to modify rule
  net/sonic: Use dma_mapping_error()
  xfrm Fix potential error pointer dereference in xfrm_bundle_create.
  vhost_net: flush batched heads before trying to busy polling
  tun: Fix NULL pointer dereference in XDP redirect
  be2net: Fix error detection logic for BE3
  net: qmi_wwan: Add Netgear Aircard 779S
  mlxsw: spectrum: Forbid creation of VLAN 1 over port/LAG
  atm: zatm: fix memcmp casting
  iwlwifi: pcie: compare with number of IRQs requested for, not number of CPUs
  ...

6 years agonetfilter: nf_tables: handle chain name lookups via rhltable
Florian Westphal [Sat, 2 Jun 2018 21:41:06 +0000 (23:41 +0200)]
netfilter: nf_tables: handle chain name lookups via rhltable

If there is a significant amount of chains list search is too slow, so
add an rhlist table for this.

This speeds up ruleset loading: for every new rule we have to check if
the name already exists in current generation.

We need to be able to cope with duplicate chain names in case a transaction
drops the nfnl mutex (for request_module) and the abort of this old
transaction is still pending.

The list is kept -- we need a way to iterate chains even if hash resize is
in progress without missing an entry.

Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
6 years agonetfilter: nf_tables: add connlimit support
Pablo Neira Ayuso [Sat, 2 Jun 2018 19:38:51 +0000 (21:38 +0200)]
netfilter: nf_tables: add connlimit support

This features which allows you to limit the maximum number of
connections per arbitrary key. The connlimit expression is stateful,
therefore it can be used from meters to dynamically populate a set, this
provides a mapping to the iptables' connlimit match. This patch also
comes that allows you define static connlimit policies.

This extension depends on the nf_conncount infrastructure.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
6 years agoMerge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Linus Torvalds [Sat, 2 Jun 2018 22:54:49 +0000 (15:54 -0700)]
Merge tag 'scsi-fixes' of git://git./linux/kernel/git/jejb/scsi

Pull SCSI fix from James Bottomley:
 "Eve of merge window fix: The original code was so bogus as to be
  casting the wrong generic device to an rport and proceeding to take
  actions based on the bogus values it found.

  Fortunately it seems the location that is dereferenced always exists,
  so the code hasn't oopsed yet, but it certainly annoys the memory
  checkers"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: scsi_transport_srp: Fix shost to rport translation

6 years agoMerge tag 'drm-fixes-for-v4.17-rc8' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Sat, 2 Jun 2018 22:24:45 +0000 (15:24 -0700)]
Merge tag 'drm-fixes-for-v4.17-rc8' of git://people.freedesktop.org/~airlied/linux

Pull drm fixes from Dave Airlie:
 "A few final fixes:

  i915:
   - fix for potential Spectre vector in the new query uAPI
   - fix NULL pointer deref (FDO #106559)
   - DMI fix to hide LVDS for Radiant P845 (FDO #105468)

  amdgpu:
   - suspend/resume DC regression fix
   - underscan flicker fix on fiji
   - gamma setting fix after dpms

  omap:
   - fix oops regression

  core:
   - fix PSR timing

  dw-hdmi:
   - fix oops regression"

* tag 'drm-fixes-for-v4.17-rc8' of git://people.freedesktop.org/~airlied/linux:
  drm/amd/display: Update color props when modeset is required
  drm/amd/display: Make atomic-check validate underscan changes
  drm/bridge/synopsys: dw-hdmi: fix dw_hdmi_setup_rx_sense
  drm/amd/display: Fix BUG_ON during CRTC atomic check update
  drm/i915/query: nospec expects no more than an unsigned long
  drm/i915/query: Protect tainted function pointer lookup
  drm/i915/lvds: Move acpi lid notification registration to registration phase
  drm/i915: Disable LVDS on Radiant P845
  drm/omap: fix NULL deref crash with SDI displays
  drm/psr: Fix missed entry in PSR setup time table.

6 years agonetfilter: nf_tables: add destroy_clone expression
Pablo Neira Ayuso [Sat, 2 Jun 2018 21:38:50 +0000 (23:38 +0200)]
netfilter: nf_tables: add destroy_clone expression

Before this patch, cloned expressions are released via ->destroy. This
is a problem for the new connlimit expression since the ->destroy path
drop a reference on the conntrack modules and it unregisters hooks. The
new ->destroy_clone provides context that this expression is being
released from the packet path, so it is mirroring ->clone(), where
neither module reference is dropped nor hooks need to be unregistered -
because this done from the control plane path from the ->init() path.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
6 years agonetfilter: nf_tables: garbage collection for stateful expressions
Pablo Neira Ayuso [Sat, 2 Jun 2018 21:38:49 +0000 (23:38 +0200)]
netfilter: nf_tables: garbage collection for stateful expressions

Use garbage collector to schedule removal of elements based of feedback
from expression that this element comes with. Therefore, the garbage
collector is not guided by timeout expirations in this new mode.

The new connlimit expression sets on the NFT_EXPR_GC flag to enable this
behaviour, the dynset expression needs to explicitly enable the garbage
collector via set->ops->gc_init call.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
6 years agonetfilter: nf_tables: pass ctx to nf_tables_expr_destroy()
Pablo Neira Ayuso [Sat, 2 Jun 2018 21:38:48 +0000 (23:38 +0200)]
netfilter: nf_tables: pass ctx to nf_tables_expr_destroy()

nft_set_elem_destroy() can be called from call_rcu context. Annotate
netns and table in set object so we can populate the context object.
Moreover, pass context object to nf_tables_set_elem_destroy() from the
commit phase, since it is already available from there.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
6 years agonetfilter: nf_conncount: expose connection list interface
Pablo Neira Ayuso [Sat, 2 Jun 2018 21:38:47 +0000 (23:38 +0200)]
netfilter: nf_conncount: expose connection list interface

This patch provides an interface to maintain the list of connections and
the lookup function to obtain the number of connections in the list.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
6 years agonetfilter: nf_tables: pass context to object destroy indirection
Pablo Neira Ayuso [Sat, 2 Jun 2018 21:38:46 +0000 (23:38 +0200)]
netfilter: nf_tables: pass context to object destroy indirection

The new connlimit object needs this to properly deal with conntrack
dependencies.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
6 years agonetfilter: Libify xt_TPROXY
Máté Eckl [Fri, 1 Jun 2018 18:44:56 +0000 (20:44 +0200)]
netfilter: Libify xt_TPROXY

The extracted functions will likely be usefull to implement tproxy
support in nf_tables.

Extrancted functions:
- nf_tproxy_sk_is_transparent
- nf_tproxy_laddr4
- nf_tproxy_handle_time_wait4
- nf_tproxy_get_sock_v4
- nf_tproxy_laddr6
- nf_tproxy_handle_time_wait6
- nf_tproxy_get_sock_v6

(nf_)tproxy_handle_time_wait6 also needed some refactor as its current
implementation was xtables-specific.

Signed-off-by: Máté Eckl <ecklm94@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
6 years agonetfilter: Decrease code duplication regarding transparent socket option
Máté Eckl [Fri, 1 Jun 2018 12:54:07 +0000 (14:54 +0200)]
netfilter: Decrease code duplication regarding transparent socket option

There is a function in include/net/netfilter/nf_socket.h to decide if a
socket has IP(V6)_TRANSPARENT socket option set or not. However this
does the same as inet_sk_transparent() in include/net/tcp.h

include/net/tcp.h:1733
/* This helper checks if socket has IP_TRANSPARENT set */
static inline bool inet_sk_transparent(const struct sock *sk)
{
switch (sk->sk_state) {
case TCP_TIME_WAIT:
return inet_twsk(sk)->tw_transparent;
case TCP_NEW_SYN_RECV:
return inet_rsk(inet_reqsk(sk))->no_srccheck;
}
return inet_sk(sk)->transparent;
}

tproxy_sk_is_transparent has also been refactored to use this function
instead of reimplementing it.

Signed-off-by: Máté Eckl <ecklm94@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>